Production and Electrolysis of Light Metals. Proceedings of the International Symposium on Production and Electrolysis of Light Metals, Halifax, August 20â€“24, 1989

Titles of Related Interest-Other CIM Proceedings Published by Pergamon Bickert REDUCTION AND CASTING OF ALUMINUM Chalkley TAILING AND EFFLUENT MANAGEMENT Dobby PROCESSING OF COMPLEX ORES Jaeck PRODUCTION AND ELECTROLYSIS OF LIGHT METALS Jonas DIRECT ROLLING AND HOT CHARGING OF

STRAND CAST BILLETS Kachaniwsky IMPACT OF OXYGEN ON THE PRODUCTIVITY OF NON-FERROUS METALLURGICAL PROCESSES

Macmillan QUALITY AND PROCESS CONTROL IN REDUCTION AND CASTING OF ALUMINUM AND OTHER LIGHT METALS

Mostaghaci PROCESSING OF CERAMIC AND METAL MATRIX COMPOSITES

Plumpton PRODUCTION AND PROCESSING OF FINE PARTICLES Rigaud ADVANCES IN REFRACTORIES FOR THE

METALLURGICAL INDUSTRIES Ruddle ACCELERATED COOLING OF ROLLED STEEL Salter GOLD METALLURGY Thompson COMPUTER SOFTWARE IN CHEMICAL AND

EXTRACTIVE METALLURGY Twigge-Molecey PROCESS GAS HANDLING AND CLEANING Tyson FRACTURE MECHANICS Wilkinson ADVANCED STRUCTURAL MATERIALS

Related Journals (Free sample copies available upon request)

ACTA METALLURGICA CANADIAN METALLURGICAL QUARTERLY MINERALS ENGINEERING SCRIPTA METALLURGICA

PROCEEDINGS OF THE INTERNATIONAL SYMPOSIUM ON PRODUCTION AND ELECTROLYSIS OF LIGHT METALS HALIFAX, AUGUST 20-24, 1989

Production and Electrolysis of Light Metals Editor Bernard Closset Timminco Metals, Toronto, Ontario

Symposium organized by the Light Metals Section of The Metallurgical Society of CIM

28th ANNUAL CONFERENCE OF METALLURGISTS OF CIM 28e CONFERENCE ANNUELLE DES METALLURGISTES DE L'ICM

Pergamon Press New York Oxford Beijing Frankfurt Säo Paulo Sydney Tokyo Toronto

Pergamon Press Offices:

U.S.A.

U.K.

Pergamon Press, Inc., Maxwell House, Fairview Park, Elmsford, New York 10523, U.S.A.

Pergamon Press pic, Headington Hill Hall, Oxford 0X3 OBW, England

PEOPLE'S REPUBLIC Pergamon Press, Room 4037, Qianmen Hotel, Beijing, OF CHINA People's Republic of China

FEDERAL REPUBLIC Pergamon Press GmbH, Hammerweg 6, OF GERMANY

BRAZIL

AUSTRALIA

JAPAN

CANADA

D-6242 Kronberg, Federal Republic of Germany

Pergamon Editora Ltda, Rua Ega de Queiros, 346, CEP 04011, Säo Paulo, Brazil

Pergamon Press Australia Pty Ltd., P.O. Box 544, Potts Point, NSW 2011, Australia

Pergamon Press, 8th Floor, Matsuoka Central Building, 1-7-1 Nishishinjuku, Shinjuku-ku, Tokyo 160, Japan

Pergamon Press Canada Ltd., Suite 271, 253 College Street, Toronto, Ontario M5T 1R5, Canada

Copyright © 1989 by The Canadian Institute of Mining and Metallurgy

All rights reserved. No part of this publication may be reproduced, stored in a retrieval system or transmitted in any form or by any means: electronic, electrostatic, magnetic tape, mechanical, photocopying, recording or otherwise, without permission in writing from the publishers.

First edition 1989

Library of Congress Cataloging in Publication Data

ISBN 0-08-037295-3

In order to make this volume available as economically and as rapidly as possible, the authors' typescripts have been reproduced in their original forms. This method unfortunately has its typographical limitations but it is hoped that they in no way distract the reader.

Printed in the United States of America

The paper used in this publication meets the minimum requirements of American National Standard for Information Sciences - Permanence of Paper for Printed Library Materials, ANSI Z39.48-1984

Session Chairmen

Smelter Operations C. Bickert

Pechiney Corporation, Greenwich, Connecticut, U.S.A.

D. MacMillan Alcan Rolled Products, Cleveland, Ohio, U.S.A.

Aluminum Casting E. Ozberk

Sherritt Gordon Ltd. Fort Saskatchewan, Saskatchewan, Canada

P. Aylen Alumax Company Ferndale, Wisconsin, U.S.A.

Aluminum Melt Treatment and Control M. Sahoo

CANMET Ottawa, Ontario, Canada

L. Larouche Canadian Reynolds Metals Co. Ltd. Baie-Comeau, Quebec, Canada

Electrolysis of Light Metals R. Guthrie

McGill University Montreal, Quebec, Canada

P. Pinfold Norsk Hydro Becancour, Quebec, Canada

Reduction and Production of Light Metals P. Tremblay

Alcan International Ltd. Jonquiere, Quebec, Canada

M. Bouchard Universite du Quebec ä Chicoutimi Chicoutimi, Quebec, Canada

Performance prediction of the aluminum casting furnace R. Bui, A. Charette, G. Simard, A. Larouche, Y. Kocaefe Universita du Quabec ä Chicoutimi, Chicoutimi, Quebec, Canada, G7H 2B1

E. Dernedde, W. Stevens Alcan International limitae, Jonquiere, Quabec, Canada, G7S 4K8

ABSTRACT

A comprehensive three-dimensional overall transient model has been built for the aluminum casting furnace. It results from the coupling of two models, repre-senting the combustion chamber and the metal respectively. The chamber model takes account of gas flow, combustion and heat transfer, mainly radiative. The metal model includes the melting of a solid charge. The overall model can readily accommodate the various operational procedures and can be used to study furnace performance as well as solve design problems.

KEYWORDS

Casting furnace, computer model, performance analysis, furnace design.

INTRODUCTION

The aluminum casting furnace plays a central role in the fabrication of primary aluminum. It receives liquid aluminum coming from the electrolytic cells, brings it up to a desired temperature and keeps it there while preparations are made (fluxing, alloying, skimming) before casting takes place. Also, a solid charge is introduced and melted. This is why the furnace is sometimes referred to as a melter-holder.

Figure 1 presents the cutaway views of a typical casting furnace. It is composed essentially of two main parts: the chamber of combustion and the metal. The process going on in the casting furnace is quite complex. On the one hand, in the chamber there is the gas flow, combustion of the fuel (natural gas), heat transfer by convection and more importantly radiation, and conductive heat transfer through the roof refractories. On the other hand in the metal, there is the heat transfer by conduction and convection, both natural and forced. Forced convection is due to stirring and also to the various operational proce-dures such as the introduction of liquid metal siphoned in from the crucibles. The solid metal also undergoes a solid-to-liquid phase change.

Each of these individual physical phenomena deserves on its own to be treated as a substantial research subject. The aim here, however, it to model all of these phenomena together and obtain a mathematical tool for the purpose of analysis and design of the furnace. What is more, the model must be reasonably thrifty in computer time in order to make sense in an industrial environment.

The modelling work at hand is no simple task. However the alternative would be to conduct plant experiments, which are long, costly and sometimes risky, without giving all the answers that a mathematical model can give.

When dealing with a complex process, it is often helpful to split it into two or more parts and model each of them separately. Intuition confirmed by experience shows that each such partial model is more manageable. Also, each model can be conceived as a self-containing simulator that can be used separately for partial design problems. Finally, in the context of today's fast moving

3

4 PRODUCTION AND ELECTROLYSIS OF LIGHT METALS

6,7

1 ,2 :

®

Figure 1- Cutaway views of the casting furnace (A) longitudinal (B) cross section

roof and its refractories 9: and insulation 10: exit duct (to stack) 11: burner 12: doors 13: floor and its refractories and insulation

melt level solid metal syphon pouring spout thermocouples

computer technology, parallel computing allows these models to be run simulta-neously and their outputs coupled interactively, thus considerably shortening the simulation time. For these reasons, the casting furnace has been modelled in two parts: the chamber and the metal.

In a previous publication (Bui, Charette, Dernedde 1988), these two models were presented together with the analysis and validation of their respective outputs. The next step was to couple the two models by an appropriate technique to obtain the overall model. This paper shows how this overall model is built and how the physical and simulation parameters are handled. The overall model is next validated on a real furnace, then model capabilities in terms of improving the operation or the design of the furnace are discussed.

THE OVERALL MODEL

The model of the chamber and that of the metal are built separately but based on a common computational tool, namely the general-purpose fluid-flow code

3: 4: 5:

,8 :

PRODUCTION AND ELECTROLYSIS OF LIGHT METALS 5

PHOENICS, an acronym for Parabolic, Hyperbolic or Elliptic Numerical Integration Code Series (Rosten and Spalding 1986).

In the chamber model, the differential equations that describe the transport of momentum, energy and concentration in 3D are solved by PHOENICS. The radiative heat transfer is calculated by the Imaginary Planes (IP) method (Charette, Erchiqui, Kocaefe 1987), a simplified version of the classical zone method that yields a computer time reduced by a factor of 20 compared to the classical method of zones. A model of combustion based on the combustion kinetics, incorporated in PHOENICS, is used for the simulation of the flame.

In the metal model, the solid-to-liquid phase change is treated by the effective thermal properties (ETP) method (Simard, Bui, Potocnik 1987), while the effect of natural and forced convection on the heat transfer is accounted for by the use of an augmented conductivity in the liquid metal (Bui et al 1989).

The overall model of the furnace is obtained by a coupling of the model of the chamber and that of the metal. The primary aim of coupling is to ensure the continuous heat transfer between chamber and metal, but it also has to take account of the fact that the geometry of the metal changes with time due to the addition of the liquid metal from the crucibles and the gradual melting of the solid blocks. Strictly speaking, that part of the metal space still unoccupied by the liquid metal should be seen as part of the chamber, and as a consequence, chamber geometry should change with time. But it would be unrealistic to recalculate the gas flow and heat transfer for a new chamber geometry at each time step.

The solution is to choose a fixed parallelepiped geometry for the chamber, let the metal geometry change with time, and perform the coupling through an "equiv-alent plane" separating chamber and metal and corresponding to the bottom plane of the chamber parallelepiped. This equivalent plane represents all parts of the furnace underneath itself, namely the metal and the empty space filled with the gas before being gradually filled with liquid metal. In the coupling process, the information exchange at the equivalent plane level takes place once every coupling time step known as the period of interaction between the two models. A 5-minute interaction period was used with success. The chamber model is run for 5 minutes and its heat flow outputs are transferred to the metal model which then runs for 5 minutes before sending its emissivity and tempera-ture outputs back to the chamber model, and the recurrent process continues.

The effective emissivities are calculated by successive steps to represent two or more surfaces by one equivalent surface until the equivalent plane is finally reached. For each step, the effective emissivities are determined from the radiative heat balances making use of the concept of interchange areas (Hottel, Sarofim 1967). Figure 2 illustrates the method. It represents a two-dimensional case obtained with a cross section of the metal made at a location where a solid block exists. In element II we have:

*\ * < S 1 2 S 7 + S 3 S 7 > I kl < 2 )

where

S .S . = total exchange area between areas i and j Cm ] A = area of equivalent surface 7 Cm 1 ε' = effective emissivity

Similarly for element III:


= <SA + W ' \ (2)

Note that in this case both surfaces 4 and 5 are refractories and if ε - ε , relation (2) simplifies further. Finally for element I:

ε' l <δΛ + W / Al (3)

and ε' is the final effective emissivity to be sent to the chamber model.

EQUIVALENT PLANE-

S5.«5

Figure 2- Principle for determining the effective emissivities

The temperature distribution on the equivalent plane is done by projection, as illustrated by Figure 3 where T , T , T are the average emerged solid surface temperature, average liquid surface temperature and average refractory surface temperature respectively.

The chamber model thus receives the effective emissivities and temperatures coming from the metal model. It must return a heat flow distribution to the metal model. Figure 3 illustrates the simple two-dimensional case where heat flow distribution is done by projection. Q , Qn»CL are the heat flows calcu-lated by the chamber model for the sections of tne equivalent plane facing the solid, the liquid and the refractories respectively. They are applied directly to each of these three surfaces.

SECTIONS: SOLID LIQUID i REFRACTORIES

Figure 3- Convention used for assigning temperatures and heat flows


For those furnace cross sections falling in between two solid blocks and thus containing no solid metal, the situation is more complex due to the radiative exchanges between the two blocks. The recipe for determining the effective emissivities and the heat flow distribution is then slightly more complicated arithmetically but still based on the same principle.

The usefulness of such coupling process cannot be overemphasized. Due to it, the overall model not only can account for all the physical phenomena taking place in the chamber (gas flow, combustion, heat transfer) and in the metal (melting, heat transfer) but also can accommodate the various operating conditions such as different heel levels and temperatures, different solid loading techniques, preheating of the metal or the refractories, different time schedules of the arrivals of crucibles, different crucible liquid metal tempera-tures. The casting furnace operational procedures are so varied and flexible that a model that cannot accommodate them sees its usefulness seriously reduced.

PHYSICAL AND SIMULATION PARAMETERS

Figure 4 shows the handling of the physical parameters and the simulation parameters. Figure 4a explains the preparation of the data files. For the chamber, three different cartesian grids are used, one for the heat transfer and fluid flow calculations by PHOENICS, another, coarser, for radiative heat transfer calculations. The latter consume more CPU time due to the need to determine the shape factors and therefore must be done on a coarser grid. Still a third grid, this one one-dimensional, is used to calculate the transient heat conduction through the refractories. For the metal, body-fitted coordinates are used to fit the non-cartesian geometry. However the depthwise discretization is kept horizontal to accommodate the successive arrivals of crucibles and the ensuing changes in liquid metal level. The mesh generation is done with PATRAN, a finite element pre- and post-processing software interfaced with PHOENICS through a software called PAPH, or its improved version NEWPAPH. The various details of the operational procedure such as preheating, crucibles arrivals, heating... are spelled out in the code PIL, an acronym for PHOENICS Input Language.

Figure 4b is the continuation of Figure 4a. It shows the chamber model in part A, the metal model in part B and the coupling in between. In the coupling, the metal model sends the effective values of emissivities and temperatures to the chamber model and in return, receives from it the heat flow distribution.

VALIDATING SIMULATION

In this section, the overall model is run to simulate a complete batch using the data from a 1987 plant test conducted on an operating castshop furnace of Alcan Smelters and Chemicals, Jonquiere, Quebec, Canada. The purpose of the simu-lation is to validate the model against the plant test data, and also to show the model capabilities in terms of operation and design improvements.

Physical parameters of the batch

The batch starts with a 20-ton liquid heel at 750°C. The refractories have been preheated to a 800°C initial inner surface temperature, and 100°C for outer surface. Three 4-ton blocks of solid metal are introduced at equidistance along the furnace and leaning against the walls on the door side (Figure 8). Their initial temperature is 100°C. The liquid metal coming from the crucibles is pure aluminum with 660°C melting point. The solid blocks are an alloy of average melting point 642°C.

PA

RA

ME

TER

S O

F T

HE

BA

TCH

-load

ing

of s

crap

met

al:

-num

ber

of b

lock

s -p

ositi

on o

f bl

ocks

-k

ind

of a

lloy

-leve

l an

d m

ass

of t

he h

eel

-dis

trib

utio

n of

th

e ar

rival

s of

cr

ucib

les

-cha

ract

eris

tics

of

the

burn

er

(fuel

,mas

s fl

ow

) -c

hara

cter

istic

s of

st

irrin

g (c

ontin

uous

or

in

term

itte

nt,

inte

nsi

ty)

inpu

t of

dat

a an

d pr

epar

atio

n of

the

mes

hes

: ra

diat

ion

and

refr

acto

ries

calc

ulat

ion

of t

he s

hape

fa

ctor

s (M

onte

-Car

lo)

|PH

OE

NIC

S|

gene

ratio

n of

the

mes

h w

ith

PA

TRA

N e

t NE

WP

AP

H

fluid

flo

w

with

he

at

tran

sfer

(c

ombu

stio

n,ra

diat

ion,

cond

uctio

n,

conv

ectio

n)in

st

eady

st

ate

Θ

|PIL

c

od

es

|

1 pr

ehea

ting

1 ar

rival

of

cr

ucib

le

1 ar

rival

of

cr

ucib

le #1

#2

arriv

al

of

cruc

ible

#n

heat

ing

1

T Θ

Figu

re 4

a-

Prep

arat

ion

of

the

data

fi

les

IPH

OE

NJC

SI

heat

tr

ansf

er

(com

bust

ion,

radi

atio

n,

cond

uctio

n,co

nvec

tion)

in

un

stea

dy

sta

te

conv

ectio

n :

velo

city

fie

ld

pre-

calc

u la

ted

1 ca

lcul

ate

the

effe

c-tiv

e em

issi

vitie

s

i ef

fect

ive

tem

pera

-|^

—|

ture

s at

the

equ

iva-

i le

nt

plan

e ,

' he

at f

lux

at t

he

» ,

cont

rol

volu

me

sur-

,

1 fa

ce o

f m

etal

'

i geo

met

rica

l, [ p

aram

eter

s '

JPH

OE

NIC

SJ

heat

tra

nsfe

r by

con

duct

ion

with

ph

ase

chan

ge a

nd

augm

ente

d-k

to r

e-pr

esen

t co

nvec

tion

verif

y of

the

set

ting

of t

he b

lock

s

sim

plify

m

odel

to

add

liq

uid

met

al

from

cr

ucib

les

Figure 4b- The recurrent coupling process

PRODUCTION AND ELECTROLYSIS OF LIGHT METALS


The temperature and mass of the crucibles as well as their arrival times are random, as seen from Figure 5. As the grid is fixed, in some cases, two crucibles must be regrouped in order to fit the grid, in which case the crucible temperature is taken as the weighted average of those of the individual crucibles.

TEMPERATURE OF CRUCIBLES MASS OF CRUCIBLES 1000 5000-

0 4 17 30 34 39 45 49 57

TIME [min]

17 30 34 39 45 49 57 88

TIME [mlnj

Figure 5- Temperature and mass of crucibles

Outputs from the chamber model;

A selected number of outputs from the chamber model is presented. Figure 6 shows the flame shape based on a 1% residual fuel contour. The overall flame shape and length agree with plant observations, although the real flame observed seems to point slightly upwards toward the roof. Introducing a flotation term into the model of combustion may help solve this discrepancy.

STACK SYPHON

SPOUT

BURNER

Figure 6- Flame shape

Figure 7 gives the measured and calculated roof temperatures at furnace midlength, at the inner roof location identified as thermocouple no. 3. Calculated values are lower than measured ones by 40°C, due likely to the low calculated flame position already mentioned.


■a— MEASURED * — CALCULATED

i i I I I I I 50 100 150 200 250 300 350

TIME[min]

Figure 7- Roof temperature at midlength

Outputs from the metal model:

Figure 8 shows the phase change boundary, i.e. the shape of the solid blocks, after 100 minutes of heating. Some solidified metal is observed between the blocks due to their low initial temperature. Figure 9 shows the liquid metal surface temperature, at the level of the spout. The maximum difference between calculated and measured values is 7%. Figure 10 gives the liquid metal bottom temperature, at the level of the stack. The maximum difference between calculated and measured values is 6%, Figure 11 shows the variation of the mass of solid metal with time. The time to total melting of 200 minutes agrees with plant observations.

Figure 8- Phase-change boundary after 100 minutes of heating

Computing time:

On the VAX/11-785, the computing time required for the simulation of 320 minutes of real time is 20 hours. On top of this, the precalculation of the flowfield in the chamber takes one hour. The chamber model takes five times more CPU time than the metal model. In fact in the chamber not only the energy and combustion equations but also the radiative heat transfer must be solved. The model as it is, is discretized into 3000 cells in the metal part while in the chamber, there are 120 cells for radiative transfer calculation and 720 cells for flow and

BURNER

O

HI QC

\-< DC W Q.

SYPHON STACK

1100

1000

900 H

800 H

700 H

600 H

500·]

400

1


SYPHON STACK 0 50

F i g u r e 9 - L iqu id m e t a l s u r f a c e t e m p e r a t u r e

BUFNm

UJ t£ ID

< LU

o. 5 LU

100 150 200 250 300 350 TIME[min]

100 150 200 250 300 350 TIME[min]

Figure 10- Liquid metal bottom temperature

15000Γ

1200<

en 9000l·

SI 6000h

3000h

60 120 180 240 300

TIME (min) Figure 11- Variation of the mas? of solid metal with time


energy calculation. The computing time of 20 hours is long yet may be even longer if a finer grid is used to enable the model to address such questions as the effect of a small tilt imposed to the burner. In this case, it will be realistic to consider the use of either a faster computer or parallel computing. In the latter, each model (of the chamber and of the metal) is run on a separate CPU simultaneously and coupled by a periodic exchange of information.

CAPABILITIES OF THE MODEL

It has been shown that the model is capable of simulating the operation of the casting furnace with good representativity. Due to its discretization scheme, its modular structure and its coupling process, the model can accommodate the various operational procedures involved. Thus, any arrival schedule of the crucibles can be accommodated: the resulting liquid metal geometry changes with time, but this difficulty is avoided by the use of an equivalent plane for the coupling of the metal with the chamber. The use of a coarse grid for radiative heat transfer calculations and a finer grid for flow calculations enables the model to account for the effect of inleakage of fresh air into the chamber in the case of a negative chamber pressure. For this purpose, the door gaps and other openings will be represented by an appropriate number of surface discretizations. The use of body-fitted coordinates in the metal makes it possible to accommodate the various solid charge loading techniques and also, the different solid charge geometries.

As for questions related to furnace design, the model can be used to study the effect of a change in the geometry of the chamber or in the geometry of the metal part of the furnace. Questions such as the effect of a change in the location of the stack can also be answered by the model.

CONCLUSION

The proposed model of the aluminum casting furnace is shown to be a useful tool for analysis and design. Its modular structure comprising a model for the chamber and a model for the metal causes an additional burden, that of coupling; but the advantages outweigh the cost. Each model is less cumbersome, less error-prone, more manageable, more transparent, more easily calibrated and validated separately before being coupled. Each model can be used as a stand-alone simulator for partial design purpose, or used in connection with the other for global design problems. A feature inherent to complex industrial processes is the difficulty in validating the model due to difficulties in obtaining the experimental measurements and also in interpreting them in terms of the corre-sponding model outputs. It can be seen that the proposed casting furnace model withstands the validation test and shows a good degree of representativity.

ACKNOWLEDGEMENTS

The content of this article is drawn from a joint project between the Universite du Quebec ä Chicoutimi (UQAC) and Alcan International Limited, Jonquiere, Quebec. The many contributions from Alcan Smelters and Chemicals of Jonquiere, Quebec, are sincerely appreciated. The code PHOENICS comes from CHAM Limited. Financial support from the NSERC of Canada is acknowledged. The authors thank Alcan for authorizing this publication.

REFERENCES

Bui, R.T., A. Charette and E. Dernedde (1988). Computer assisted process analysis of the casting furnace. Proceedings of the Metallurgical Society of CIM, Vol 8, 151-165.


Bui, R.T., G. Simard, M. Lacroix, A. Charette, E. Dernedde (1989). Modelling the melting of aluminum. Light Metals, AIME, 531-536.

Charette A., F. Erchiqui and Y. Kocaefe (1987). ''The imaginary planes method for the radiative heat transfer analysis in furnaces1 \ Proceedings of 37th Canadian Congress of Chemical Engineering, Montreal, 391-393.

Hottel, H.C. and A.F. Sarofim (1967). Radiative Transfer, McGraw-Hill, New York.

Rosten, H.I. and D.B. Spalding (1986). PHOENICS beginner's guide and user manual, CHAM, London, U.K.

Simard, G., R.T. Bui and V. Potocnik (1987). Solving moving boundary problems using PHOENICS with effective thermal properties. 2nd International PHOENICS Users Conference, London, U.K.

15

Computer start-up of a cold Soderberg potline

Nelson Dub6, Luden J. Larouche and Yves iêmee Sociätä canadienne des mataux Reynolds, limitae, Case Postale 1530, Baie-Comeau, Quabec, Canada, G4Z 2H7

ABSTRACT

The first phase of a Soderberg cell start-up consists in raising the amperage up to full line load and melting down the metal pad until the anode is free. The molten metal pad is then heated up over a long period from a temperature of approximately 660 °C to around 950 'C. This is done by increasing the resistance of the cell, normally by using manual controls. This operation requires constant attention, to ensure a gradual and continuous resistance increase, and to prevent high voltage swings which result in poor anode current distribution and consequently, anode cracking.

This paper describes a method that uses the potline computer to perform this critical operation. The computer runs a specially adapted resistance control program. This automatic control permits a very consistent increase in heat input to the cells throughout the heating period and brings them to a self-fluxing state.

KEYWORDS

Vertical stub Soderberg cells, computer control, start-up, heat input, resistance control, anode current distribution, temperature.

INTRODUCTION

Societe Canadienne de Metaux Reynolds' South Plant in Baie-Comeau has 542 vertical stub Soderberg cells divided into three (3) potlines. For its part, the North Plant has one (1) potline consisting of 240 Pechiney prebaked anode cells. The plant's annual production capacity stands at 279,000 metric tons of aluminum but a major expansion of the facilities begun during the spring of 1989 will bring that capacity to 400,000 metric tons per year by 1991 with the addition of another complete potline of 240 Pechiney prebaked anode cells.

Table 1 gives the main operating characteristics of the three (3) Soderberg potlines.


TABLE 1 Soderberg Potiines

Annual production capacity (metric tons) 160,000

Potiines 3

Cells 542

Amperage (A) 115,000

dc kV/h/kg Al 15.75

Volts/Cell 4.70

Current density (A/cm2) 0.68

Gross carbon consumption (kg/kg Al) 0.52

In order to improve the operating performance and emission control of the Soderberg cells, the three (3) potiines were completely modernized starting in 1980 according to a technology package purchased from the Sumitomo Aluminum Smelting Company. Consequently, the Soderberg cells now operate with a dry paste anode under low current density with a highly insulated cathode. The cells are controlled by microcomputers linked to a host computer. This computer controls the cell's voltage. It also controls the cell's feeding (regular crust breaking, anode effects and anode effect predictions) by means of a bar breaker on each side of the cell with independently operated pneumatic cylinders at each end of the bar. The computer also acts upon anode effect, operating an automatic termination system by air blowing which includes one (1) air pipe near each end of the anode and solenoid valves which can be activated either by the computer or manually. Figure 1 gives the layout of a potline.

NATURE OF THE PROBLEM

Shutting down one or more potiines is generally done because of economic or labour relations considerations. It is a process that must be systematic to prevent damages to the cells. The normal steps for the shutdown are: adjusting the height of the bus bars; making sure no stubs are too low in the anode; lowering the level of the metal (approximately 3000 kg are tapped from each cell); slow positioning of the anodes in the remaining molten aluminum (under 2.0 volts); shutting down the current to the potline; and, raising the anodes thirteen (13) hours after the shutdown.

The start-up process for the potline is equally complex and requires much attention. The main steps are: perfect cleaning of the anode underface and of the hardened aluminum pad; levelling and positioning of the anodes in the cathode cavity; preheating the cells; cell preparation (cryolite and crushed bath); transfer of molten metal into the cell (approximately 4000 kg); energizing the potline; gradual increase of power to heat the cell; and, cut-in with liquid bath.

The gradual increase of power represents a critical phase in the start-up of a potline. It can be divided into two (2) stages: the period leading to maximum amperage; and, the subsequent period. The second stage requires maximum control of the power dissipated by the potline.


Fig. 1 The Baie-Comeau Soderberg potline.

At about the 100-kA level, slightly short of the target amperage, the metal pads are sufficiently melted down to free the anodes. We then begin to observe a decline in the potline resistance and a subsequent decline in energy input which requires the first manual action on the adjustment of the cell voltage. As the anodes are free to move in the molten metal, the cell voltage is adjusted to approximately 2.8-3.0 volts (below this level, resistance drops suddenly) and maintained there until the line amperage reaches the target level of 115 kA. Figure 2 shows the general evolution of voltage, amperage, resistance and power during the start-up.

Fig. 2 General evolution of voltage, amperage, resistance and power during a start-up.


It was at this stage that the unusual application of the computer was made using a specially conceived voltage control program. This program allowed the computer to take over the cell voltage control operation from the point the first manual movement of the anode was made and the first voltage target was set on a cell, to the point where the cell was ready for cutting in with liquid bath. This is a critical operation where the computer performed exceptionally well and gave results considerably superior to the conventional manual operation.

SOLUTION

Conventional Manual Operation

Let us have a look first at the conventional manual operation required during the metal heating phase.

The planned start-up schedule shows that the cell cut-in operation is done at a rate of 20-22 cells per 8-hour shift and takes a total of seven (7) to eight (8) shifts. The limitations for the cut-in rate are: the maximum line voltage, the near maximum line amperage, the manpower and the equipment available. For these reasons, the potline is divided into eight (8) sections of 20-22 cells, and each section must be 8 hours apart in terms of cell voltage and metal temperature in order to assure an orderly cell cut-in operation in the final phase of the start-up. Table 2 shows the schedule used during our last start-up.

TABLE 2 Typical Schedule for the Voltage Increase Phase

GROUP OF CELLS CELLS DAY SHIFT CUT-IN OF CELLS DAY SHIFT

First group 501-519 6 1 601-619

Second group 502-520 6 2 602-620

Third group 521-543 6 3 621-643

Fourth group 522-544 7 1 622-644

Fifth group 545-567 7 2 645-667

Sixth group 546-568 7 3 602-620

Seventh group 569-589 8 1 601-619

Eighth group 570-590 8 2 670-690

At this stage, a supervisor is assigned to each section of cells, his responsibility being to patrol continuously the section and gradually raise the voltage of the cells. In fact, his work consists in maintaining each cell at a medium level of instability until the warmer metal pad permits a slightly higher voltage without exceeding the acceptable level of instability.

7

7

8

8

8

9

9

9

2

3

1

2

3

1

2

3


During the metal heating phase, the temperature of the metal will go from slightly less than 700 SC to approximately 950 6C, while the cell voltage will go from 3.0 volts to around 5.5 volts.

The degree of success in this operation is mainly related to the ability of the supervisor to maintain a steady increase of the energy input into the cells. It is obvious that patrolling the section continuously and adjusting cell voltage is of extreme importance. High levels of instability or high voltage swings will be experienced if a steady warm-up trend is not obtained. If not properly done, the anodes, which are then even more fragile than usual, will be affected considerably by the high instability, and cracking will likely occur. These situations will make subsequent operations much more difficult and they must be avoided.

Experience shows that the main difficulty frequently encountered in a manually controlled operation during the metal heating phase is that individual cells usually behave very differently one from another. The cell voltage adjustment will not always hold until the next round, thus requiring some catching up and inevitably resulting in very high instability and possibly damaging the anodes and cathodes.

Unfortunately, even with the best effort from the participating employees, the damage suffered by the anodes during the heating phase of the metal will make the cell cut-in operation and the following cell operation more difficult for a period of time.

Automatic Voltage Control

The computer control system for the Soderberg cells consisted of a host computer and distributed microcomputer control units. A dual and completely separate communication architecture was designed into the system. One of the links was used for communicating with another microcomputer which served as an entry and reporting unit in the potroom office, while the second was connected to a minicomputer which served as a host data collector for long term report generation. Figure 3 shows a simplified schematic of the control system used in Baie-Comeau.

The program was inserted into the microcomputers of the distributed system and it was ready for use on a cell basis at a signal from the computer terminal located in the line office. The two (2) principal functions of the computer program were: first, to maintain a cell voltage at a set target and second, to increase the voltage at a set pace respecting programmed limitations. The voltage target used by the computer for a given cell was the voltage at which that cell was adjusted manually prior to positioning the selector switch on the automatic mode. This meant that the voltage target of a cell could easily be changed either at the cell control box or at the computer terminal.

The cell voltage control was set on automatic mode after the first manual movement of the anode and at that point the voltage was adjusted to 3.0 volts. The role of the computer was then to maintain that voltage until all the cells were on automatic mode and the potline had reached the target load.

Since the amperage target had been reached and we needed to keep the total energy input on an upward trend, it was time to start increasing the cell voltage. As described previously in the manual operation, the line was divided into eight (8) equal sections. On the first shift, the first section was set on "voltage increase" mode; on the second shift, the second section was set on "voltage increase" mode and so on... to finally obtain the necessary distribution for the


final cut-in schedule. From then on, the computer brought the cells to around 5.5 volts in approximately 48 hours with minimum human surveillance. The temperature of the metal was then around 950 °C and the cells were ready to receive liquid bath. At this temperature, the voltage required for cutting in a cell should not exceed 12-15 volts, which is a relatively mild shock for the anodes and cathodes.

Fig. 3 General arrangement of the control system.

The schematic of the algorithm used for computer control of the voltage during the metal heating is shown in Fig. 4.


FIX Tfflf TARGET OF TH|B CELL

FOB ALL GfcLLS OF THE MICRO

tS REAOWG* RESISTANGE! DETERMINE, MSTABIUTY ,i

I CBIL ON AUTOMATIC 7

ir CELL UNDER CONTROL ?

Ι-Γ 3T

HfGK WSTA&tlTY *

3Γ

LAST CELL 7

UPDATE TRANSMiSS40N BUFFER

TARGET - TARGET - 0V RESTRICTIONS: Φ VOLTAGE * SET POWT ?

3E S5T VOLT AGB

TIME INCREASE TARGET ? !

-MAXIMUM INSTABILITY ■ 1 VOLT -SO mV / 30 minute· -MAXMUM TARGET VOLTAGE ■ 5.5 VOLTS NO DOWN MOVE UP MOVE ■ 0.6 SECOND

® Fig. 4 Algorithm used for computer control of the voltage.

The start-up algorithm services its complement of cells once every two (2) minutes.

The voltage of the first cell is read once every second for fifteen (15) seconds. The lowest value is subtracted from the highest and the result is called instability.

Depending on its value, instability is classified as normal or high.

The computer verifies whether or not this particular cell is on automatic control. If such is the case, the cell voltage is compared to the target and the anode is raised slightly (0.5 mm) if the voltage is found to be a little low.

The computer will now raise the voltage target if it is time to do so but only if the instability is found to be normal.

All new readings pertaining to the cell under consideration are now copied to the BASIC variables before tackling the next cell.

When all cells have been serviced, the transmission buffer is sent to the host computer in order to update the cells1 data files.


The computer functions and the safety of the operation were assured by the following restrictions:

Rate of voltage increase. A rate of voltage increase of 50 mV per 30 minutes was found to be excellent. Ideally, the overall 2.5-volt increase would occur over a 25-hour period but due to periods of greater instability, the real period was around 48 hours.

Maximum instability. This parameter was set at 1.0 volt which was relatively low when compared to a manual operation and it contributed greatly to reduce anode stresses caused by high voltage at that stage.

Upward movements of the anodes. All upward movements of the anodes were set to a 0.6-second action of the anode jack motors. The anode movements were always in the upward direction so there was no need to compensate for backlash on the jack mechanisms. The duration of the movement was equivalent to the finger touch action in the manual operation. The 0.6-second movement corresponded to a displacement of the anode of approximately 0.5 mm.

Downward movements of the anode. No downward movement of the anode was allowed in the automatic mode. This restriction assured a continuous gain in energy input. If necessary, a new higher or lower target voltage was given to the computer.

Maximum voltage. The 5.5-volt level was normally reached after approximately 48 hours and the level of instability at this point limited further voltage increases. The temperature of the metal was close to 950 "C and the cell was considered ready for liquid bath addition.

COMPARISON BETWEEN THE TWO METHODS

Comparing the data between the manual operation and the automatic control clearly shows the superiority of the latter method. Figure 5 shows the evolution of the power for manual and automatic actions.

POWER (Megawatts)

0 28 58 84 112 140 168 186 224 252

14 42 70 88 126 154 182 210 238 266

HOURS SINCE THE BEGINNING OF THE START-UP

Fig. 5 Evolution of the power dissipated by the line under manual and automatic actions.


Among other things, we can see the constant increase in power in the automatic mode. Contrary to what happened in the manual anode, there was no decrease in the power dissipated by the potline. This allowed for a continuous increase in power which in turn ensured a progressive increase in temperature. This is specially true starting from the 120th hour when we began using the automatic voltage control.

The same trend was noticed in the evolution of the line voltage as demonstrated by Fig. 6.

VOLTAGE (VOLTS!

1000 ,

o i 1 0 28 58 84 112 140 168 196 224 252

14 42 70 98 126 154 182 210 238 268

HOURS SINCE THE 0EGINNING OF THE START-UP

Fig. 6 Evolution of the l i n e voltage under manual and automatic ac t ions .

Figure 7 i l l u s t r a t e s the evolution of the temperature of the metal during the voltage increase phase.

Also, with automatic con t ro l , we were able t o c u t - i n a l l the c e l l s of the same group within 42 hours a f t e r the beginning of the systematic voltage increase phase. This was 30 hours e a r l i e r than with the manual cont ro l .


TEMPERATURE (°C)

500 12 16 20 24 2Θ 32 36 40 44 48 52 56 60 62 66 70

HOURS SINCE BEGINNING OF VOLTAGE INCREASE

Fig. 7 Evolution of the temperature of the metal under manual and automatic actions.

Another parameter clearly demonstrates the superiority of computer control. Each cell's particular voltage curve shows that the resistance control is remarkably consistent with what we had set for each cell.

Figure 8 shows the evolution of voltage in automatic mode with the voltage graph used at the time. This graph is still being used today for our regular operation. We can see the systematic increase of the set point of the cell for every 30-minute period and the corresponding voltage value. There are no high voltage swings or high instability periods. The left part of the graph indicates the current amperage, set point and voltage values.

Manual interventions for modifying voltage targets were very seldom required.

There were a few exceptional cells that behaved a little differently from the others. It is estimated that 6-7 % of the cells experienced high instability during the whole operation and prevented a normal increase of the voltage. It is believed that too high a metal pad had been left in these cells. The only inconvenience was a longer heating period. Figure 9 shows the evolution of the resistance of these cells.

The 8-hour interval between sections of 20-22 cells was well followed until the end. Except for the last 7-8 cells that were held up due to high line voltage, the operation was on schedule and the cut-in was done in a very orderly manner.

We were all surprised the operation.

to see that only minimal surveillance was required during

Did Murphy's law apply to this new technique? It most certainly did. Even if computer control proved to be very advantageous, we almost had to reconsider its use because of an incident early on.


Fig. 8 Voltage and set point evolution in automatic mode for normal cells.


Fig. 9 Voltage and set point evolution in automatic mode for unstable cells.


At the time of the first trial in the field, four (4) cells were set on automatic mode. Everything went very well for approximately three (3) hours until two (2) anodes started moving downward, two (2) others went upward, the bar breakers were activated and the air lances started blowing air. It. was nearly the end of our novel development but fortunately, the problem was quickly traced to a faulty socket on a printed circuit board. We thus kept the program running.

CONCLUSION

Optimal control of the power dissipated by the potline is necessary to ensure a problem-free start-up. That is exactly what the use of a computer easily allowed us to get.

By reducing manpower requirements for the cell-to-cell voltage control, we were able to follow and even get ahead of the scheduled work because people could perform other tasks.

Broken pieces of carbon during the cut-in operation were practically inexistent. The cells were warm and ready for liquid bath additions and the voltage necessary for this operation did not exceed 12-15 volts.

Finally, the majority of people who had expressed reservations on the practicability of the operation were very surprised and pleased with the results.

29

Eight-and-a-half-hour power failure and subsequent restart of the 180 kA prebaked aluminum potline in Baie-Comeau

Luke Tremblay and Germain Leblanc Sociata canadienne des matawc Reynolds, limitie, Case Postale 1530 Baie-Comeau, Quabec, Canada, G4Z 2H7

ABSTRACT

To have a potline accidentally shut down for 8j hours and then re-energized and brought back into normal operation within a couple of days would be considered an unlikely event by most experts.

In August 1985, a fire caused by a faulty potential transformer in the switchgear room of a main auxiliary power substation interrupted power supply to the entire 180 kA prebaked potline in Baie-Comeau, Quebec, for 8 hours and 28 minutes. The line current was then re-established but the auxiliary electrical circuits remained out of service for another 9 hours. The potline was therefore energized, but hibernated. Most of the anodes were in contact with liquid metal but, without any auxiliary power supply, there were no lights, pot tending machines, microcomputers, anode jack motors or gas collection for routine operation.

The restart of the line, which required 60 hours, was done by creating a controlled opening of the circuit by slowly tapping metal from the pots and adding bath as necessary. In this process, nine (9) pots were cut out.

This paper describes the actions taken during the shutdown and outlines the technique used to restart the pots. Finally, possible effects of the shutdown, such as cathode drops, cathode block deteriorations and pot failures, over the last three (3) years, will be discussed.

KEYWORDS

Potline, prebaked, anode, cathode, cut-in, shutdown, restart, power, current, metal pad, bath.

INTRODUCTION

The Baie-Comeau North Plant consists mainly of a 180 kA prebaked Pechiney type potline comprising 240 pots. Construction of the plant was completed in November 1984. As seen in Fig. 1, it was built just 90 meters north of the existing South Plant which comprises 542 vertical pin Soderberg pots.


Fig. 1 The Baie-Comeau plant.

Start-up of the new potline began in November 1984 and was virtually complete when, on August 28, 1985, a fire caused by a faulty potential transformer in the switchgear room of the main 13.8 kV auxiliary power substation interrupted power supply to the entire North Plant. The outage began at 8:52 p.m., August 28 and ended at 5:20 a.m., August 29, when the rectiformers were re-energized.

Even though re-establishment of the line current began at 5:20 a.m., the auxiliary electrical circuits remained out of service until about 2:00 p.m., preventing operation of the compressor room, potline auxiliary equipment, fume treatment system, new cast house, rodding room and maintenance building.

The potline therefore was energized, but hibernated. Most of the anodes were in contact with liquid metal but without any auxiliary power supply which means that there were no lights, pot tending machines, microcomputers, anode jack motors or gas collection for routine operation.

When the shutdown occurred, 238 pots were in operation while the remaining two (2) were on resistor bake. The average age of the pots was only 171 days.

This paper describes the causes of the failure, the actions taken during the shutdown and outlines the technique used to restart the pots. The short and long term effects are also discussed.


DESCRIPTION OF THE SHUTDOWN

The shutdown originated from a fire in a main substation which supplies 13.8 kV power to the auxiliary transformers of the potline rectiformers as well as two (2) other substations which feed power to the auxiliary equipment of the potline, the fume treatment system, the compressor room, the rodding room, the new cast house, the maintenance shop, and the personnel facilities.

Due to its importance, this main substation consists of two (2) 161 kV/13.8 kV transformers, each of them capable of supplying the entire North Plant's requirements. The switchgear for these units is located in a building underneath. The normal mode of operation was to have an open tie circuit breaker between the two (2) transformers so that each carried about half the load.

The fire was evidently started by a faulty potential transformer that ignited the entire tie circuit breaker located just below the transformer. It is unknown how long the fire smoldered before the smoke and soot contamination caused the breakers on both sides to trip.

When the fire was extinguished, the electrical department decided to cut the bus which joined the two (2) units behind the circuit breakers, and clean out the best they could the switchgear of the least contaminated transformer. An attempt was then made to close the breakers of the transformer and thus re-energize the line. Unfortunately, the breakers did not hold as they were evidently too contaminated.

At this point, the situation was becoming critical since the potline had been shut down for more than 4J hours. Realizing the magnitude of the necessary clean-up, the quick-thinking electrical department decided to run temporary cables to provide a temporary power supply to the auxiliaries of the rectiformers. Two (2) 1200-meter cables were used to connect a 400 A/600 V output in a substation located close to the rectiformers of the South Plant with a 600-volt distribution panel in the main North Plant substation. The cables were left over from the construction of the new plant; without them, the plant would probably have remained shut down.

Completed around 5:00 a.m., August 29, the connection permitted the operation of the rectiformers and re-energizing of the potline at 5:20 a.m.. At this point, the line had been shut down for 8 hours and 28 minutes.

ACTIVITY DURING THE POWER OUTAGE

Shortly after the power failure, it became obvious that there was considerable damage to the main substation and that the potline would be shut down for at least a few hours. The following actions were taken during the outage:

a) Immersion of the anodes in the bath was periodically verified through the tap hole. Around 1:00 a.m., the liquid bath had diminished to such an extent that it was doubtful the anodes would be in contact with the bath once power was re-established. It was therefore decided that the anodes should be lowered but the means of doing so were very limited since the motorization of the anode jacks was not functional and no lighting was available. The anodes of only eighty-five (85) pots were lowered approximately one centimeter with the use of a pneumatic impact wrench.


b) The alumina feed holes were periodically verified to ensure proper feeding upon re-establishment of power. This was not a problem since the holes remained open during the entire outage.

c) As of midnight, pot temperatures were monitored on two (2) pots, one of which was equipped with graphitized cathode blocks (pot 7080), the other with standard anthracite blocks (pot 7081). As indicated below, the temperatures dropped rapidly and, by 2:00 a.m. and 4:00 a.m. respectively, we had metal/bath inversion in both pots. At this point, the temperatures climbed slightly before dropping to 870 eC, which was the lowest reading registered during the shutdown.

TEMPERATURE (°C)

TIME

12:15 a.m.

12:45 a.m.

1:15 a.m.

1:45 a.m.

2:15 a.m.

2:45 a.m.

3:15 a.m.

3:45 a.m.

4:15 a.m.

4:45 a.m.

5:15 a.m.

6:00 a.m.

6:30 a.mf

7:00 a.m.

7:30 a.m.

8:00 a.m.

* Metal/Bath

POT 7080 (GRAPHITIZED CATHODE BLOCKS)

894

892

888

884

902*

896

894

888

892

872

-

871

874

874

874

878

inversion

POT (ANTHRACITE

7081 CATHODE BLOCKS)

923

909

887

880

870

868

849

844

882*

870

-

882

892

882

890

877

Pot 7080 was later cut out.

d) Preparations were under way in the South Plant adding cryolite and solid bath to the pots in order to have the maximum amount of liquid bath available for transfer.


RESTORATION OF POWER AND RESTART

The line voltage just prior to power restoration was 115 volts while the line current was evidently zero. At this point, the general feeling was that we had a very remote chance of re-energizing the line. As mentioned earlier, only one third of the anode bridges had been lowered (just 1 centimeter) and we feared that the circuit was completely ruptured and the resistance too high to overcome.

When power was re-established, the line resistance was effectively very high, creating several impressive arcs to ground caused mainly by the poor insulation between the concrete slabs around the pots and the potroom floors. However, the line current successfully found a path, so the anodes were presumably in contact with the metal pad. This situation created a few moments of anguish and hesitation but the decision to keep the direct current on the line was maintained. For several minutes, the line current varied between 140 and 180 kA while the line potential varied from 400 to 500 volts providing on the average 2.15 volts per pot.

Restart of the pots began around 7:00 a.m., August 29.

The pots restarted during the day shift came onto anode effect immediately following re-establishment of the line current and they succeeded in melting down their bath. The means of restarting were very limited since none of the auxiliary equipment was operational. However, anode bridges were moved with pneumatic impact wrenches, portable voltmeters were used for voltage readings and alumina was fed to the pots by manually activating the twin-hopper solenoid valves. By the end of the day shift, fourty-four (44) pots had been restarted that way.

For the restart of the remaining pots, the auxiliary electrical power was re-established and all auxiliary equipment was available. Solid bath around the anodes prevented any upward movement of the bridge even though the anode jacks were operational. The only available method to restart was to create a controlled opening of the circuit by slowly tapping metal from the pots and letting them heat up until hot bath could be added. We were fortunate that the metal was still warm enough to be tapped with regular tapping equipment.

With this plan in mind, the following guidelines were therefore given to the start-up crews:

- Slowly tap metal from the pot until the voltage reaches approximately 10 volts.

- Let the voltage oscillate until the metal heats up and the pot starts making its own bath .

- The solid bath will melt and, within approximately 30 minutes, free the anodes.

- The anode bridge can then be adjusted so that the provoked anode effect can be controlled and the voltage maintained below 20 volts.

- Within approximately 30 minutes, the bath temperature should reach 940 °C.

- Add one ton of liquid bath to the pot (the liquid bath was available from the South Plant).

- At this point, the pot can be considered restarted.


This "cut-in" procedure, although long (see Fig. 2), was used throughout the restart of 187 pots.

AUG. 29 AUG. 30 AUG. 31 Nl9ht

Fig. 2 Cumulative number of restarted pots vs shifts.

To cut in the pots, 426,5 tons of metal were tapped, most of which were poured into regular transport crucibles. Slightly more than 150 tons of bath, provided by the South Plant, were added to the pots (not all pots required the addition of bath).

As expected with this method of restarting, many anode stems were stripped at the metal transition piece. By the end of the start-up, 314 stems had failed. This stem damage caused the cut-out of nine (9) pots. In most of these cases, ten (10) or more metal transition pieces were broken leaving the pot with only six (6) anodes to carry 180 kA. Too much work and risk would have been involved to try to restart a pot in this condition so it was decided that they should be cut out, cleaned and restarted at a later date.

Following its restart, each pot's resistance was adjusted as follows:

Hours since restart Target Resistance (μΩ) *

0 to 24 Rk + 0.5 μΩ

24 to 48 Rk + 0.3 μΩ

48+ Rk + 0.2 μΩ

* Target resistance prior to the shutdown was Rk

Bath levels, which were slightly out of control for a few shifts were progressively brought down to 20 cm.

The restart was completed by the end of the day shift on August 31, some 60 hours following the re-establishment of line current.


EFFECTS OF THE SHUTDOWN

Except for a short transition period observed at the beginning of September, the potline regained normal behavior within a few weeks following the shutdown. The technical results, although less impressive than those of the first half of 1985, were nonetheless respectable.

However, we did notice a change in the cathode drop evolution during the Fall of 1985 (see Fig. 3), at which point there was a sudden increase in the cathode drops, particularly in the case of pots equipped with amorphous cathode blocks.

mV

450 425 j 400 J 375 J 350 J 325 J 300 J 275 J 250 4

... | Amorphous >cathode I blocks

T — i — i — i — r I I Sep

Graphitized cathode blocks

Nov T—r

Mar 85 May Jui Sep Nov Jan 86

Fig. 3 Cathode drop evolution.

Mar

Pots equipped with graphitized blocks did not show and have remained at the same level to this date.

any increase in cathode drops

The deteriorating condition of some of the cathodes became obvious when, on March 30, pot 8069 failed at only 362 days. A complete dry delining of the cathode, performed by experts from Aluminium Pechiney and Reynolds Metals Company, showed extensive lamination-type cracking and impressive heaving. Molten aluminum had penetrated the blocks through the cracks and had attacked the collector bars (see Fig. 4).

Vertical cracks

Horizontal cracks with metal infiltration

frsf-rtT* '»re Λ ' . Ί ' . Ί Ί Ί Ί τ Alumina mixed

with the first 'firebrick layer

Fig. 4 Laminated cathode block.


Researchers from Pechiney and Reynolds agreed that the cathode lamination was caused by an unusual expansion of the blocks, where the top part of the blocks expanded differently than the bottom.

The reasons for this abnormal expansion are not clear, although it has been suggested by some experts that it is a result of the power failure occurring when some of the pots were in a critical phase of high sodium impregnation.

Since March 1986, several pieces of laminated cathode blocks, typically illustrated in the Fig. 5 below, have been removed from the pots.

Fig. 5 Piece of a laminated cathode block.

There have also been many failures particularly in 1986. As of April 1988, 19.6 % of the original pots had failed.

Baie-Comeau has registered, by far, the highest rate of pot failures compared to other smelters using the Pechiney 180 kA technology. However, this high rate of pot failures has occured only on two (2) types of cathode blocks while the other two (2) types, comprising 40 % of the potline, have not shown any failures to date. Thus, the hypothesis that the early failure is solely the result of the power outage remains debatable.


CONCLUSION

Few people expected the Baie-Coraeau 180 kA prebaked anode pot line to be operational only a few days after such a long shutdown. The prebaked operation, considered by many to be extremely vulnerable to long shutdowns demonstrated an unexpected performance. Not being able to properly lower the anodes into the metal pad during the outage led us to believe that the potline would have to be restarted from scratch after all the pots had been cleaned out. However, even after the bath/metal inversion, the anodes of every pot remained in contact with the metal pad. This phenomenon kept the overall potline resistance at a level that permitted re-energizing of the line.

There is no doubt that the Pechiney 180 kA pot demonstrated extraordinary capacity in facing such a difficult situation. The only immediate damage was isolated cases of stripped anodes and limited failures of anode stem metal transition pieces.

Whether or not the early pot failures are directly related to the shutdown remains hypothetical.

Personnel of Societe Canadienne de Metaux Reynolds displayed enormous effort throughout the restart. The expertise of Mr. Gilbert Roma, of Aluminium Pechiney, was of great importance and without him, the restart would have been much more of a burden.

39

Measurement of anode and cathode resistance to thermal shock

J. Bigot, G. Georgalas Aluminum Pechiney, Laboratoire de recherches des fabrications, Saint-Jean-de-Maurienne, France D. Dumas, P. Lacroix Carbone savoie, Laboratoire de recherches et essais, Venissieux, France

ABSTRACT

During the first few hours of the anode life in pot operation and during the rodding of cathodes, cracking may occur if the carbon resistance to thermal stress is not sufficient. An apparatus to measure the resistance to thermal shock in anodes and cathodes is of great interest in aluminum production. Such a device would permit to evaluate the thermal stress resistance of various anodes and cathodes. It would also provide a useful means to determine anode and cathode manufacturing conditions, avoiding carbon thermal stress damage. The information obtained could also help in the specification of raw materials and in the production of anodes and cathodes with sufficient resistance to thermal stress.

KEYWORDS

Primary aluminium, prebaked potlines, anode manufacture, cathode manufacture thermal shock, cracking process, cathode rodding.

INTRODUCTION

To produce aluminum by the Hall-Heroult process, prebaked potlines require anodes and cathodes. The anodes made of petroleum coke are consumed in the electrolytic cell and have to be regularly replaced. Changing anodes consists in replacing a consumed anode and setting up a new one under ambient temperature conditions. The low part of the new anode comes into contact with the hot liquid bath, which is at a temperature above 950°C. During the first few hours of the anode life in the pot, cracking may occur if the anode resistance to thermal shock is not sufficient. If an anode cracks, it has to be replaced by a new one. This leads to an increased workload for potline operators and a higher anode consumption. The cathode mainly composed of calcined anthracite is set in the electrolytic cell for 5 years or more. Before setting up the cathode in the cell, steel rods are secured to the cathode with cast iron. Liquid cast iron at a temperature above 1350°C is poured between the rod and the carbon material. During this rodding operation, if the cathode resistance to thermal shock is not sufficient, cracking may occur. Because of the costly consequences, it is necessary to produce anodes and cathodes, which are resistant to thermal shock. Due to the importance of these last aspects, this cracking phenomenon has already been studied. The usual approach consists in measuring characteristics in relation to thermal shock on anodes and cathodes : elasticity, coefficient of thermal expansion, resistance to rupture,... Then, according to Kingery or Hasselman theory, materials can be characterized with the formula presented by Brown (1975) and Schmitt-Hatting (1988). Because of the lack of homogeneity of the material, as usually just one sample is taken, the values obtained for some of the characteristics are not very precise and this leads to difficulties in the interpretation of the formula results. These characteristics have also been introduced in a mathematical model developed to describe the stresses distribution in materials as shown by Wong (1984), Michard (1986) and Larsen (1989).

Cracks in corners

Vertical- and horizontal cracks


With a method allowing to describe all the aspects of the thermal shock sensitivity, it is possible to take a sample of both cathode and anode on line and then to characterize it. And this is of great interest. The measurement of the thermal sensitivity with an apparatus will permit to determine the anode and cathode manufacturing conditions. The suitable raw materials can be selected to avoid carbon thermal stress damage. Thanks to low cost of this measurement it is possible to characterize bench scale anodes in significant quantity.

TYPES OF CRACKS

During the first few hours of the anode life in the pot, cracking may occur. Several types of cracks can be observed as shown in Figure 1 (cracks in corners, horizontal cracks, vertical cracks or any of the above combination).

Fig. 1. Types of cracks in anodes due to thermal shock


During the rodding of a cathode, cracking may occur as shown by Dumas (1973). Several types of cracks can be observed as shown in Figure 2 (wing cracks, corner cracks and transversal cracks).

Fig. 2. Types of cracks in cathodes due to thermal shock

CRACKING PROCESS

When a new and cold anode is set in a pot, the anode receives a thermal shock coming from the hot liquid bath. When a steel bar is secured in a cathode, the cathode receives a thermal shock coming from the hot liquid cast-iron. The temperature increases inside the carbon blocks with a distribution depending on the position of the heat source. This distribution changes in function of duration. Thermal expansion takes place and develops mechanical stresses in the anode and the cathode. If the stresses are beyond the carbon material resistance, a crack appears.


CHARACTERIZATION OF THE CARBON STRENGTH

A way to characterize the carbon strength, is to apply a thermal shock of ajustable power to a sample of carbon. A simple mean to create a thermal shock is to establish an electrical arc at a center of a carbon disc as shown by Sato (1975). The mathematical analysis of the stresses developed within a disc heated at its center has been made by Riney (1961). The center of the disc is submitted to compression forces while the exterior of the disc is submitted to tensile forces. The crack is radially propagated from the circumference. The test duration has to remain rather low to avoid the sample Oxydation and the erosion of the graphite electrode in arc. This is obtained with a suitable choice of the disc diameter and thickness and of the injected power.

DESCRIPTION AND USE OF THE APPARATUS

The apparatus permits to arc a carbon disc at a given power and during a given time. To execute the arcing in such conditions, the apparatus needs to include a Iransformer, a regulation system for the power and a test cubicle where arcing is done (Figure 3). This apparatus is patented.

Fig. 3. Apparatus to measure the thermal shock resistance


In the use of the apparatus, arcing duration and power level are adjusted. The carbon disc is set on a graphite block. The arcing is started manually. Once the arc is established, the power is adjusted automatically. When the preset time is passed the system stops and the operator observes the disc which can be either cracked or not cracked (Figure A).

Area red by heat

Crack

Fig. 4. Cracked carbon disc

DETERMINATION OF THE CRACKING POWER

The cracking power value characterizes the carbon material resistance to thermal shock. It is. the value of the lowest power, at which all the tested discs crack. The first step in this determination is to find an approximate cracking power. This is done by Lhe arcing of a few discs at different power levels. This preliminary search covers a large range of powers. From this approximative cracking power, two or more power levels are tested with a relatively high number of discs coming from the base sample. These tests lead to the determination of the cracking power, by following an algorithm shown in Figure 5.


At each step corresponding to a precise power, we have two possibilities :

YES : all the disks are cracked NO : one or more disc is not cracked

The algorithm describes the range of power levels, step by step, and shows how one ot the power levels of this range is reached.

AP - S

AP

YES

'YES CP < AP - IS

NO CP = AP

AP + S NO

YES CP = AP + S

NO CP £ AP + 2S

AP = Approximate power

CP = Cracking power

S = Chosen step

Fig. 5. Algorithm to determine cracking power


ANODES STUDIES WITH THE APPARATUS

The thermal stress resistance of anodes consumed in various smelter pots can be evaluated with the apparatus. These anodes are produced with cokes of different origins and in different manufacturing conditions. The table 2 presents different types of anodes used in various smelters. With the power level obtained by the test, the cracking phenomenon can be forecasted in Potlines. For example, the level 950W on the table 2 indicates a high probability of cracking in the pots and this is effectively observed.

TABLE 2. Evaluation of the Manufactured Anodes Thermal Stress Resistance

Smelter

Power level

A

1500 W no cracking

in pot

B

1370 W no cracking

in pot

C

950 W cracking in pot

The apparatus allows to study with bench scale anodes the influence of raw materials and anodes manufacturing parameters on the thermal shock. Figure 6 shows the significant influence of cokes of different origins and the influence of the percentage of pitch in the mix.

2100 X

1900 X

1700 -l·

1500 i

1300 1

1100 1

900 1

Power (W)

10 12 1A 16 18 - I > 20 % Pitch

Figure. 6. Influence of coke origin and pitch percentage on thermal shock resistance


CATHODES STUDIES WITH THE APPARATUS

Thanks to this apparatus it is possible to study with bench scale cathodes the influence of the components on the thermal shock. As an example, we have selected different samples to show the influence of graphite content on the material behaviour.

TABLE 3. Influence of Graphite Content on Thermal Shock Resistance

Grades

Thermal conductivity (W/mK)

at 20°C

Power

(w)

A

4

1500

B

7

1700

C

9

2100

D

15

3000

E 1

26

-

In this table, the cracking power value is the lowest one at which 50 % of the discs are cracked. Sample E presents no disc cracked at 3000 W.

CONCLUSION

The use of this apparatus is of great interest to understand how the raw materials and the manufacturing processes have an impact on anode or cathode cracking due to the thermal shock. This works efficiently in the prevention of cracking occurences for anodes in contact with the hot bath and for cathodes in contact with the hot cast iron. Not only is anode cracking prevented, but also the quality of carbon material to resist to thermal shock is improved. The main interest of the device resides in a reduction of the costs generated by this cracking phenomenon which leads to anode and cathode rejections.


REFERENCES

Leroy M. (1987). Some causes of cathode failures and ways to present them - third aluminium electrolysis workshop. Carnegie Mellon University.

Dumas D. Vallon J, (1973). Improvement of the casting with cast-iron of collector bars in large length cathodic carbon blocks. Light Metals, 793-304.

Dumas D. (1987). Cathode failure modes - third aluminium electrolysis workshop. Carnegie Mellon University.

Brown J.A. Elhedey P.J. (1975). Characterization of prebaked anodes carbon by mechanical and thermal properties. Light Metals, 253-269.

Koohman A.A. Schmidt-Hatting W. Van Den Bogerd P. (1988). Sensitivity of anodes for electrolysis aluminium production to thermal shock. Light Metals, 253-257.

Hsu M.B. Weng T.L. (1984). Temperature and stresses in carbon anodes in an aluminium reduction cell. Light Metals, 977-987.

Michard L. (1986). Modelling of the sealing of cathodes bottom block. Light Metals, 699-704.

Larsen B. Sorlie M. (1989). Stress Analysis of cathode bottom block. Light Metals, 641-646.

Sato S. (1975). Determination of the thermal shock resistance of graphite by arc discharge heating. Carbon, Vol 13, 309-316.

Riney T.D. (1961). Journal of Applied Mechanics, E 28,631.

ALUMINIUM PECHINEY - CARBONE SAVOIE, N° 88/10710 - French Patent Pending.

49

The Venalum 230 kA pot line H. Medina and H. Lacourt Gerencia de Coordinacion Tecnologica, CVG Industria Venezolana del Aluminio C.A., CVG-Venalum, Apdo. 312, Puerto Ordas, Edo. Bolivar, Venezuela

ABSTRACT

In 1986, VENALUM iniciated a smelter modernization and expansion program scheduled to be completed in 1988. This program comprises the construction of a 230 KA pot line with 180 pots resulting in a capacity increase of 110.000 MT. of aluminium per year. The VENALUM 230 KA pot line incorporates leading edge technologies regarding working conditions, enviromental protection, automated operations higher productivity and lower energy consumption. This paper deals with the experience in the execution of the program and with the characteristic features of the new systems installed.

KEY WORDS

High amperage pots, point feeders, dense phase, dry scrubbing, ventilation.

INTRODUCTION

Venezolana de Aluminio C.A. (VENALUM), located at Ciudad Guayana, Venezuela, was commissioned in 1978 with an annual production capacity of 280,000 metric tons of primary aluminium. In 1986 the Venalum smelter decided to increase its production capacity up to 456,000 MT a year, with the modernization of the 720 existing pots and the start-up of the new potline of 180 electrolytic pots of 230 KA, which has incorporated the latest developments in terms of reduction technology such as: high amperage, point feeders, alumina and additives feeding by dense phase and the use of microcomputers to control the process. The expansion project of the Venalum 230KA potline has been established on the basis of:

1. Low-Cost hydroelectric power ava i l ab i l i t y . 2. Projection of a world's de f ic i t of primary

aluminium and with a increasing world market demand. 3. Alumina production in Venezuela (Interalumina Plant) with a capacity of

1,000,000 MT/year. 4. Local experience in the management of aluminium production smelters.


The technology evaluation for the new potline (Fifth Line) was primarily based on real- estate inventory of existing capital of the plant infrastructure with advantages for the expansion such as: carbon plant, baking furnaces, rodding-and crucible-services. The characteristics of the design of the 230 KA pot permits the maximum production capacity in the available physical space for the expansion and the advantages are taken to make use of the same type of anodes (1400 mm) foreseen for the modernization of the existing pots (Ref. 1). It also offers the advantages in terms of direct investment cost, permiting the reduction of the construction volume of the potrooms with the following savings in earth-movement, civil construction and material supplies. The principal specifications of the 230KA potline are shown in the following table:

REDUCTION POT

Production capacity 1720 Kg/Al/Day Nominal current (DC) 230 KA Current efficiency 93% Energy consumption 13.6 Kwh/TMAL Alumina consumption 1930 Kg/TMAL Carbon consumption gross 580 Kg/TMAL Dimensions anode carbon 1400 x 790 x 560 mm Number of anodes 26 Voltage per pot 4.18 Volts Current density 0.80 A/cm2

Cathode lining insulating/refractory bricks Cathode type Semigraphyte Number point feeders 5

PQTR00M

Number of pots 180 Pot layout Side by Side Potroom dimensions 47 x 650 m Center distance between pots 6.3 m Passage along the pots 5.6 m

PROJECT EXECUTION

In order to optimize the project execution and comply with a narrow project schedule, it was decided to import the know-how, design and basic engineering, local manufacturing and fabrication, erection of all equipment and all the associated required services. The first steps in the construction of the Venal urn 230 KA pot line were taken in July 1986 with the earthmoving. The implentation of this particular project for a shorter period of time became a challenge in order to overcome difficulties such as:

-. Shortage of construction raw-materials, concrete etc. -. Contracting skillfull workers for construction like carperters, plumbers,

bricklayers. -. Hoisting of the alumina conveyor. On January 27th of 1988, the first

technical pot start-up was carried out after 18 months of the project execution which means a time record for this kind of project.


CELL DESIGN

BUSBARS

One of the basic requirements to seccesfully design a high amperage cell is to have the ability to find a design hydrodynamics equilibrium in the liquid zone of the busbar configuration. For the 230 KA cell, busbar locations were examined using a computer model developed by Hydro aluminium. The magnetic fields calculated are shown in Fig. 1. It can be seen that system almost compensate the magnetic influence of the next row with an average value Bz = 4.58 Gauss.

CATHODE

The cathode design was developed to reduce heatloss. The steel shell is lined with refractory bricks, insulation bricks and semi-graphite carbon blocks with two collector bars per block.

PROCESS CONTROL

The process control system was developed by Hydro Aluminium. It is based on a distributing concept, which means that all the programs for controlling the 180 pots of fifth (V) potline are handled by microcomputers located at each pot. (Fig. 2) The central computer only receives status data/messages, which coordinates the data to be used by the foremen. In addition, the computer system has an automatic control of alumina feeding, anode adjustment and anode effect quenching. It also includes an adaptive process control software (Ref. 2) installed in the microcomputer that can be able to calculate the bath resistance and the concentrations of AI2O3 to perform a control of the feed rate resulting in a significant reduction of anode effects, energy consumption and an increase in the current efficiency.

GAS CLEANING

Fume control and venti lat ion systems play a major part in the efforts of Venal urn to reduce the health-hazards to which workers in the new potline are exposed. (Fig.3).

DRY GAS SCRUBBING

The fume treatment and fluoride recorvery systems have been dimensioned and design by Flakt. This includes all recent optimizing and modernization-technologies such as snaprings filterbags, integrated reactors for high efficiency, pneumatic and fluidized transportation systems. The process pot gas is collected from the hooded cells with a collection efficiency in excess of 98%.


Fig. 1 Venalum 230 KA. Calculated Magnetic Z Field

VAX C O M P U T E R

CURRENT C O N V E R T E R

MICRO-COMPUTER

P O T I

COMUNICATION C O N T R O L L E R S

D I S T R I B U T I O N BOXES

J~T 3_Π MICRO

|COMPUTEF|

MICROr

ICOMPUTERl

A N O D E E F F E C T ALARM SYSTEM

MICRO ICOMPUTERl

MICRO

(COMPUTER!

P O T 2 POT 3 POT 4 POT 5

Fig. 2 230 KA. Pot Line Process Control System Configurati on


VENTILATION AIR

DRY SCRUBBER

VENTILATION AIR

FANS

Fig. 3 Fume Control and Venti lation System

53


ALUMINA PLANT

AIRLIFT 4 0 0 T/h

AIR FROM BLOWER

AIR FROM BLOWER

AIRLFT 400 T/h

DRY SCRUBBER J ADDITIVES

4 0 T

PRESSURE VESSEL

\PRESSURE VESSEL

SECUNDARY ALUMINA

POINT FEEDER

L - .

POTS HW\ ADDITIVES

mmss^m PRIMARY ALUMINA

SECUNDARY ALUMINA

· _ « · · ADDITIVES

Fig. 4 Alumina and Additives Transportation System


VENTILATION

The ventilation system, developed by Hydro Aluminium, is based on segregated air layers. In the case of Yenalum 230 KA pot line, the introduction of the ventilation air through the potroom floor using fans, provides a "vertical flow pattern" which takes the anode fumes and pot gases upwards with crossflow of approximately 3 m'above the potroom floor. This accomplishes operators area with fresh air and minimum amounts of pollution.

ALUMINA TRANSPORTATION

The fundamental characteristics for this system is that the primary alumina is transferred from the alumina plant to the smelter storage silos via 2.6 Km long conveyor belt. The charging of secondary alumina and additives to pot hoppers is done by a pneumatic dense phase conveying system (Fig. 4) designed by Alesa Alusuisse for handling dry bulk materials of grain size up to 1 mm with compressed air at a low velocity in the conveying pipe. Each pot is connected to the main pipe on the wall by a valve and a branch pipe. The conveying system is controlled with a level indicator for each pot·

CONCLUSIONS

I t can be concluded that the Venal urn 230 KA potl ine project eminently sat isf ies the test of technical and economic f e a s i b i l i t y based on:

I t u t i l i zed leading edge technologies to obtain optimum perfomance with respect to large scale smelter operations.

F ina l ly , the expansion of the Venal urn plant is not only a positive contribution to primary metal production in our country, but also proves we are considered to be one of the mayor plants producer of the world primary aluminium industry with 456,000 metric tons a year, and one of the most prof i table companies of the world.

REFERENCES

1 . - H. MEDINA N. ELARBA "Modernization of Venal urn Pots" Light Metals 1988, 623-626

2 . - T. Μ0ΕΝ, J . AALBU, P. BORG "Adaptative Control of Alumina Reduction Cells with Point Feeders" Light Metals 1985, 459-469

57

Vertical flue ring furnaces: energetic implications of the correlations between operating characteristics

F. Gheorghiu, I. Oprescu and M. Georgescu Polytechnical Institute of Bucharest, Bucharest, Romania

ABSTRACT

The authors present the correlations between the operating charac-teristics of the vertical flue ring furnaces, based on kiln experi-mental identification, the result being the possibility of fast de-termination of combustible flow and pressure difference values corresponding to an imposed productivity or a step in time. Also, the paper includes the energetic implications of these correlations under the incidence of baking expenses and the obtained results at a furnace in operation·

KEYWORDS

E l e c t r o d e bak ing ; baking f u r n a c e s ; fu rnace o p e r a t i o n ; p roces s e f f i -c i e n c y ; energy consumpt ion .

I1MR0DUCTI0N

I n o r d e r to bake ca rbon e l e c t r o d e s i n v e r t i c a l f l u e r i n g f u r n a c e s , under c o n d i t i o n s of i n c r e a s i n g p roces s e f f i c i e n c y , the fo l l owing f a c t o r s must be s i m u l t a n e o u s l y ensured : i ) the e s t a b l i s h i n g of a the rma l regime - h e a t i n g , e q u a l i z i n g , c o o l i n g - which w i l l a l low the baking p roces s to be c a r r i e d out f o r the e n t i r e furnace charge (Oprescu, 1987) ; i i ) the i n c l u s i o n of the h e a t i n g v e l o c i t i e s d u r i n g d i s t i l l i n g pe r iod 0 ο ο · . · 7 ο ο C gas t e m p e r a t u r e ) i n the range 1 . 5 · · · 5 C/h; i i i ) the combust ion of the h i g h e s t p o s s i b l e q u a n t i t y of d i s -t i l l e d v o l a t i l e s ; i v ) r a t i o n a l energy consumpt ion.

I n what f o l l ows we w i l l p r e s e n t some p o s s i b i l i t i e s of the i n c r e a s e of the baking p r o c e s s e f f i c i e n c y , based on the c o r r e l a t i o n s between the o p e r a t i n g c h a r a c t e r i s t i c s of the f u r n a c e .

CORRELATIONS BETWEEN OPERATING CHARACTERISTICS

G e n e r a l l y , the above f a c t o r s depend on the f o l l o w i n g s * e s t a b l i s h e d thermal c o n d i t i o n s of the furnace (Gheorghiu , 1986) ; c o n t r o l l e d

58 PRODUCTION AND ELECTROLYSIS OF LIGHT METALS combustion of v o l a t i l e s a t the l e v e l of the d i s t i l l i n g chamber (Ghe-o r g h i u , 1987a) ; c o n s t a n t t empera ture of gases a t the e q u a l i z i n g chamber i n l e t (Gheorghiu, 1987b) ; c o n t r o l l e d c o o l i n g of the c h a r g e .

I n o rde r to e s t i m a t e the thermal and gas-dynamic c o n d i t i o n s of f u r -nace o p e r a t i o n , the baking curve chosen must be c o r r e l a t e d wi th the s p e c i f i c hea t consumption (q s 4-35o -f 445o k<J/kg; Gheorghiu, 1987c) and wi th the s t e p i n t ime , £ ; the value of the p r e s su re d i f f e r e n c e , Δ ρ , we determine on the bas"is of the c o r r e l a t i o n wi th the f u e l flow r a t e B, under c o n d i t i o n of c o n t r o l l e d combustion of d i s t i l l e d v o l a -t i l e s (Gheorghiu, 1987a) · I n t h i s r e s p e c t , the i d e n t i f i c a t i o n of the furnace c o n s i s t s of ma in t a in ing the f u e l flow r a t e a t a c o n s t a n t l e v e l BQ ,

B o ~ q · m c / ( H i * t ) j m 5 / h ( 1 )

and i n de te rmin ing the oxygen amount of the gases a t the i n l e t of chamber i n i n t e n s e d i s t i l l a t i o n ; t h i s amount cor responds to the ma-ximum va lue of the p r e s s u r e d i f f e r e n c e Δ ρ n e c e s s a r y f o r the c o n t r o l -led combustion of d i s t i l l e d v o l a t i l e s (Gheorghiu, 1987a) · The d e -pendence of the p r e s s u r e d i f f e r e n c e on the a c c i d e n t a l or neces sa ry v a r i a t i o n s of the f u e l flow r a t e i s expressed by

Δ ρ ^ [ ( Β + Κ")/Κ·] 2 , N/m2 (2 ) where

and Κ· = ( Β 0 · ν Β + ν κ + 1 6 χ ) / ( ν Β · Δ ρ ° · 5 )

*" - (VK + L e X > ^ B · The term I i s the excess a i r which ensures the c o n t r o l l e d combus-t i o n of t h l v o l a t i l e s a t the l e v e l of the chamber i n i n t e n s e d i s t i l -l a t i o n and i t can be determined by

Iea-l.ol6(B0. vB + V O ° 2 ] / {0°2] a" [*°2]} * ^ h · «) I n e q u a t i o n (2) the f u e l flow r a t e JB may vary around the value B , depending on the s p e c i f i c behaviour of every chamber which has reached the cons idered l e v e l · As a r e s u l t of the furnace o p e r a t i o n on the b a s i s of t h i s c o r r e l a t i o n , the q u a n t i t y of burned v o l a t i l e s r i s e s and thus the thermal e f f i c i e n c y of the p rocess i n c r e a s e s t o o · I n the imposed case of vary ing of h o u r l y p r o d u c t i o n φ* ma in ta in ing the same chamber number of furnace s e c t i o n ( h e a t i n g , e q u a l i z i n g , c o o l i n g ) , i t i s n e c e s s a r y to vary the s t e p i n time i , i t s range of va lues being determined s i m u l t a n e o u s l y by p r oduc t i on requ i rements and by h e a t i n g r a t e s i n the range 1 · 5 · · · 5 C/h 0 ο ο . · · 7 ο ο C, gas tem-p e r a t u r e ) · Thus , under c o n d i t i o n s of the h e a t i n g of the chamber charge wi th the same tempera ture g r a d i e n t ΔΤ, us ing two s t a t i o n a r y t imes tn and t , on the b a s i s of the thermal balance e q u a t i o n , we can express the" r e l a t i o n s h i p

V B n = V * l <*> where the f u e l flow r a t e B-, being known and cor respond ing t o a c o r -r e c t baking reg ime . I n the two ca se s of furnace o p e r a t i o n -t-, and t - between the flow r a t e s Ii , , Ii „ , V,f Ί and V „ the f o l i o -xi ΘΧ , j . ex , η ν , χ ν , η wing r e l a t i o n s h i p must be observed :

L , / L Ä V , /V ^ t / t - , ( 5 ) e x , l / J J e x , n ν , Γ ν , η η ' 1* KJ)

PRODUCTION AND ELECTROLYSIS OF LIGHT METALS 59 Knowing t h a t i n the two cases of furnace o p e r a t i o n , the gas flow r a t e s a t the chamber i n l e t i n i n t e n s e d i s t i l l a t i o n a re expressed by (Gheor-g h i u , 1987a) ι

▼l = V V B + Les,l + \ , 1 > *3'h >

and that

then V n A l - t x / t n . (7)

On the basis of the gas-dynamic conception of the furnace which en-sures the relationship

the value of the pressure difference corresponding to the stationary time t can be determined with the equation

Δ Ρ Π - Δ Ρ Ι - C W ? (8) Taking into account that in the cooling area and in the first heating chambers, the surface heat exchange is mostly connective, determined by the convection heat transfer coefficient «C^v , the equations (7) and (8) should be corrected, thus t

(νη/ν°·8-ν*ηί (9)

Äi>n^ ΔΡ Χ · ( V V 2 · 5 . <1Ο> As t ^ tn , the new value Δρ determines a supplimentary air flow rate, aspirated through the cooling area, flow which can be calculated with one of the following equations t

Xn s La,l · C W ' [< W 0 · 2 5 - 1 ! · ^ <U> ^ a . n ^ a . l *(1/Δρ5· 5 ) · (ΔΪ° · 5 -Δρ°·5) , m5/h (12)

where La,lÄBl' V W + mK " V * l ' m3/h · ( 1 3 )

This supplimentary air flow must be heated with the temperature gra-dient ΔΤ_ (from atmospheric temperature to the temperature at the in-

a let-of the equalizing area) which entails an additional fuel flow rate

SB Ä sLa>n · oa· AÂR^y , m5/h . (14)

Thus, i t results that the necessary fuel flow rate for the fyrnace operation with stationary time t and pressure difference Δρ is ex-pressed by:

Bn = Bn ♦ SB , m5/h . (15)

In these new conditions of furnace operation ( Δρη and t n ) , in order to ensure the controlled combustion of volatiles i t is necessary to

60 PRODUCTION AND ELECTROLYSIS OF LIGHT METALS v e r i f y the observance of the c o r r e l a t i o n S* = f ( Δ ρ η ) us ing one of the fo l lowing equa t ions (Gheorghiu, 1987a):

S n ~ [<M + AV * ( Δ Ρ , , / Δ Ρ ! ) 0 · 5 - Aj· t j / t n ] · 1 Λ Β , m3/h (16)

where Α ί " V * B » Α ϊ = V K , 1 + W *'

KJ = (AJ + A j ) / ( A P ? ' 5 - v B ) , K» . A»/v B .

By de te rmin ing the f u e l flow r a t e 5 * wi th e q u a t i o n (16) o r (17) > t h e r e r e s u l t the fo l l owing a l t e r n a t i v e s of fu rnace o p e r a t i o n _i_i) B^ ^ 5*· case when t h e o p e r a t i n g c h a r a c t e r i s t i c s a r e IL and Δ ρ ^ ; xi a* φ. 4. -w n ^*·η i i ) S n < B n , the o p e r a t i n g c h a r a c t e r i s t i c s a re B and Δ ρ ·

The a p p l i c a t i o n of the t e c h n o l o g i c a l a s p e c t s p resen ted above, under c o n d i t i o n s of a reduced s t a t i o n a r y time £ , has the fo l lowing e n e r g e -t i c and p r o d u c t i o n i m p l i c a t i o n s x i ) the""increase of the h o u r l y p r o -d u c t i o n ; i i ) the i n c r e a s e of the s p e c i f i c f u e l consumption; i i i ) the i n c r e a s e of the e l e c t r i c i t y consumption neces sa ry f o r the p a s s i n g of gases through the f u r n a c e .

The s t a t i o n a r y t ime £ was thus chosen a s a r e s u l t of the s imul taneous a n a l y s i s of the economical and t e c h n o l o g i c a l f a c t o r s .

I t should be mentioned t h a t the i n c r e a s e of the hou r ly p r o d u c t i o n based on the r e d u c t i o n of the s t a t i o n a r y time i s l i m i t e d by the con-d i t i o n s of d i s t i l l i n g p rocess of the v o l a t i l e s , which a r e determined by the e l e c t r o d e d i a m e t e r (Gheorghiu, 1987c) , t h e i r arrangement i n the box, the p e r m e a b i l i t y of the packing coke (Dernedde, 1987; Opres-c u , 1988) , e t c ·

EXPSRIMBNTAL RESULTS

The t he rmo techno log i ca l a s p e c t s p resen ted above were a p p l i e d on a v e r t i c a l f l u e r i n g f u r n a c e , whose o p e r a t i n g d a t a a r e inc luded i n Table 1 .

TABLE) 1 Opera t ing Data of Furnace ( t ^ s 48 h o u r s )

No of chambers p e r furnace - 28 No of f i r e s ( s e c t i o n s ) pe r furnace - 2 No of burne r b r idges pe r f i r e - 2 S t a t i o n a r y time ( t ) - 48 hours No of chambers p e r f i r e ( s e c t i o n )

- p r e h e a t i n g - 6 (P 1 . . . P 6) - e q u a l i z i n g - 1.25 (B) - c o o l i n g (covered) ~ 2 # 75 - c o o l i n g (uncovered) - 1

P r e h e a t i n g time - 288 hours E q u a l i z i n g time - 6o hours

PRODUCTION AND ELECTROLYSIS OF LIGHT METALS 61 Cool ing time - 28o hours E q u a l i z i n g t e m p e r a t u r e , T~» - l o jo C Combustible flow r a t e pe r f i r e , J3 ~ 94 n r / h (average v a l u e ) P r e s s u r e d i f f e r e n c e , Δ ρ (measured i n P6 chamber) - 15 N/m By-pass d i ame te r va lue - 9°o mm

On the b a s i s of the h e a t i n g curve ob ta ined under the o p e r a t i n g con-d i t i o n s p r e s e n t e d i n Table 1 , i t i s cons idered t h a t the a p p l i c a t i o n of c o r r e l a t i o n B = f ( Δ ρ ) f o r chamber P3 ( sense : from B to P6) e n -s u r e s the c o n d i t i o n s r e q u i r e d by the c o n t r o l l e d combust ion of the d i s t i l l e d v o l a t i l e s . Thus, the flow r a t e of excess a i r a t the l e v e l of chamber P3 was determined wi th the e q u a t i o n ( 3 ) , r e s u l t i n g !u~ i = 2558 n r / h · This c a l c u l a t i o n was r e a l i z e d on the b a s i s of the

βΧ , 1 following data t the oxygen percent of the gases at the chamber P3 i n l e t [% 02]= 1ο·3%; the oxygen percent of the atmospheric a i r

Γ% 02] = 18.5%; the humidity of the atmospheric a i r xQ c=r o .ol fcg/fcg . I t should be mentioned that the subscript n=l represents the referen-ce conditions (t-, = 48 hours) · In order to estimate the p o s s i b i l i t i e s of the production increase , based on a reduction of the s t a t ionary time, the following operat ional c h a r a c t e r i s t i c s were determined with equations (4) and *8) : i ) heating ve loc i ty , w (for gases) , in °C/h; i i ) pressure di f ference, Δρ„ , in N/m ; i i i ) fuel flow r a t e , Bn , in n r /h ·

Technological and production considerat ions have imposed a s ta t ionary time t c as 4o hours , which resul ted in a production increase of 2o% as compared to the operat ional regime with t^ = 48 hours· Under these new condi t ions , for an acceptable value of the heating veloci ty (Wj- ss 2*5°C/h) and based on the thermal balance, there resul ted Δρ,- SS 2l»6 N/m and B^ = 112#8 n r / h · In order to ensure the surface heat t r ans fe r , equations ( l o ) , (12) , (14) and (15) were used to r e -ca lcula te the values of the pressure difference and fuel flow r a t e , r e su l t i ng thus Δ Ρ 5 = 23·66 N/m and B^ ss 119·3 m5/h. Figure 1 presents comparatively the dependence of the cha r ac t e r i s t i c s w, Ap> Δρ, B and B on the s t a t iona ry time t · Knowing the values , under reference condi t ions , of the constants K£ = 84·729 and K£ as 234·153» the co r re l a t ion B*" ss f ( A p ) , given by equation (17) , can be rewr i t t en as s

Έ*~ 84.729 · Δ ρ £ · 5 - 234.153 - t x / t n , m3/h (18) equation which expresses the va r i a t ion of the fuel flow ra te B* necessary to the control led combustion of v o l a t i l e s depending <>n the s t a t i ona ry time t (n ss 1 . . .6) and the pressure difference ΔΡη ·

Figure 2 presents the dependence of the cha rac t e r i s t i c s B* and Δρ

62 PRODUCTION AND ELECTROLYSIS OF LIGHT METALS on the s t a t i o n a r y time t .

I n the new conditions_impose.d ( t c = 4o h ) , f o r a c a l c u l a t e d p r e s -su re d i f f e r e n c e of Δ ρ 5 = 23#66 N/m2, t h e r e r e s u l t s B? =3.31.15 m3 /h .

130

CO

£120 E

ICO110

Έ m100

90

30

* 2 5

20

Γ\^~

Γ ^ ^ ^ | \ B 5 d ! 2 J

B5= 119.3 m3/h

) m 3 / h ^ ^ "

B

^ / ^ B

> v 1 /

^ ^ >

: \Δρ=21.6 | \ - P 5 —

66 N/m2

N/m2

Δρ

ι .

40 42 U 46 STATIONARY TIME "to , hrs.

48

F i g · 1· The dependence of the cha rac t e r i s t i c s νν,Δρ, Δ ρ , Ja and B on the s ta t ionary time t n .

The experiments were carr ied out for a durat ion of z = 5 · t r = 2ooh, introducing in the host computer of t-he furnace the values Δ Ρ 5 S 24 N/m2 , B^ s 131 m5/h and the co r re l a t ion B = f (Z\p) corresponding to chamber P3 (tj- = 4o hours ) .

The r e su l t s obtained are in accordance to the heating curve imposed, under conditions Δ ρ ^ 2 4 N/m5, wR = 2.5°C/h and T{? = lloo°C, thus


r e s u l t i n g f o r the f u e l flow r a t e B^ the f requency r e p a r t i t i o n p r e s e n t e d i n F i g . 3 · p

63

ft?

150

140

130

120

110

100

90 30

«Ί 25

13 20

15

L _ ^ i

| \l3W5

|\23^=g j p y

^Λ-38 48 40 42 44 46

STATIONARY TIME " t ^ h r s .

F i g . 2 . The dependence of the c h a r a c t e r i s t i c s §*~andAp n on the s t a t i o n a r y time t ·

n

The average va lue of the f u e l flow r a t e d u r i n g the experiment per iod was (Bc)M = 13o.662 a r / h , ve ry c l o s e to the one c a l c u l a t e d (Be = 131.15 n r / h ) . That the d i s t r i b u t i o n of the f r e q u e n c i e s p r e -sented i n F i g . 3 does not correspond to a normal Gauss d i s t r i b u t i o n i s exp la ined by the v a r i a t i o n , s t a r t i n g from B*= f ( A p ) , of the f u e l flow r a t e d u r i n g s t a t i o n a r y time t^ = 4o hours (Gheorghiu, 1987a) , as fo l lows : a ) the f i r s t 18 hours of t ? > (B^ ) M ^ l 2 9 m 5 / h ; b) the l a s t 22 hours of t r , ( B * )M—133 m 5 / h .

ENERGETIC IMPLICATIONS

From the e n e r g e t i c po in t of view, the i n c r e a s e of the hour ly p r o -d u c t i o n due to the dec rea se of the s t a t i o n a r y time jt , b r ings about an i n c r e a s e of the f u e l and e l e c t r i c i t y consumption nece s sa ry to the p a s s i n g of gases through the f u r n a c e . The v a r i a t i o n s of the a b s o l u t e va lues of the h o u r l y p r o d u c t i o n Q , the f u e l flow r a t e


uo

30 T

o w20lt

10H

33,5 MV-Mean value

,19.5

MV: 126.32

3

9 MV:128. 5 MV:13070 MV:132.15

13.5

MV:133.73

125 127 129 131

B? m3/h

133 135

F i g · 3· The frequency r e p a r t i t i o n of the fuel flow r a t e ·

S n and the pressure difference Δ ρ η , the last two being presented

in Fig· 2, are not illustrative as regards the energetic implica-tions of the problem discussed. For this reason, the relative va-riations (depending on the t ) of the following characteristics

were analysed : i) hourly production Q f in tons/h; ii) specific

fuel consumption bn, in nr/t; iii) specific electricity consumption

for passing of gases through furnace, e , in kWh/t. The relative

variations £q, £ β and E^ of the above characteristics are calcu-lated with the equations s

are presented in Fig. 4.

The comparative analysis of the relative variations 6nf £ Ώ and £ ,, . ^ b Ii»

points out the increase of the specif ic energy consumption in the conditions of r i s ing hourly production Q , which is determined by the reduction of the s ta t ionary time. By studying the absolute va-lues of the specif ic fuel and e l e c t r i c i t y consumptions for t r=4o h, the following can be s ta ted : i ) r e l a t ive increase of specif ic fuel consumption £ B c—16.3% i s acceptable for a production i n -


100

50

0 30

£ 20

CD

ω 10

0

30

\ § 5 ,

\ H 3

0 >

20

10 »5

38 40 LI U Z.6 STATIONARY TIME"t*' HRS.

F i g . 4 . The dependence of £ Q» H B a n d

the s ta t ionary time t .

48

£ s on

crease of 6^^ s 2o'£; ii) the relatively high increase of specific electricity consumption £ E - = 65% does not significantly influence the cost of the electrodes as, in absolute value, the specific elec-tricity consumption necessary for the movement of gases in the fur-nace comprises a small part of the energetical expenses.

On the data presented above, the technological conditions of baking can be quickly determined so as to satisfy the requirements of pro-duction at a particular time, while the energetic expenses foi? the solution chosen can be calculated at the same time.

CONCLUSIONS

The theoretical aspects presented as well as the experimental re-sults regarding the production and energetic implications determined by the correlations between the operational characteristics of ver-tical flue ring furnaces impose the determination of the characte-ristics ( Δρ and B) -under conditions of varying stationary times t - by simultaneously taking into consideration the followings : trrermal balance; surface heat transfer; controlled combustion in highest possible quantity of the volatiles. In a new imposed hea-ting curve (a new tn), the minimum fuel flow rate B* , corresponding

to a controlled combustion of volatiles (j T calculated and correc-ted by experiment, being at a minimum), can be determined, based on

66 PRODUCTION AND ELECTROLYSIS OF LIGHT METALS the reference heating curve, by application of the theoretical and experimental considerations presented· The analysis of the energetic implications of the reduction of the stationary time t, must be made on the basis of the specific fuel consumption J) and tKe specific e-lectricity consumption e, knowing that, in absolute value, the elec-tricity necessary for tHe movement of gases in the furnace is not an important part of the energetic expenses·

NOTATIONS

: mass of e lectrodes of a furnace chamber, kg. z : i n fe r io r ca lo r i f i c value of fue l , J/nr or J /kg.

: gas specif ic volume r e su l t i ng from fuel burning, nr/nr z

or nr /kg · : volume flow ra te of gases r e su l t ing from packing coke

z combustion, n r /h · : volume percent of ©2 in burning gases, a t d i s t i l l i n g chamber i n l e t ·

: volume percent of O2 in atmospheric a i r · t volume flow ra t e of gases resu l t ing from d i s t i l l e d vola-

z tile combustion, nr/h·

: burning gas velocity (through heating channels), m/s. : volume flow rate of aspirated air (through cooling zone),

m / k · z z z * a i r specif ic volume for fuel combustion, nr/nr or nr /kg · * mass of burnt packing coke, kg· -, : a i r speci f ic volume for packing coke combustion, nr /kg ·

Z Q x specific heat of hot air, J/nrdeg. C· x thermal efficiency of the air heating process·

H i VB

[%o2]

V La

■K

5 a 7

REFERENCES

Dernedde, E . , T· Bourgeois, R.T· Bui, and A. Charette (1987)· Light Metals. 591-595·

Gheorghiu, F · , and I · Oprescu (1986)· Metalurgia (Romania), 8, 381-585· ~

Gheorghiu, F · , and I . Oprescu (1987)· Light Metals, 615-618· Gheorghiu, F · , and I · Oprescu (1987)· Quality and Process Control i n

The Reduction and Casting of Aluminium and Other Light P e t a l s , Edited by D· W· MacMillan. Pergamon Press , New York·

Gheorghiu, F · (1987). Unpublished works. Oprescu, I · , F· Gheorghiu, and M. Georgescu (1987)· Temperature f i e ld s

in the furnace rooms, a basic fac tor i n bake conditions e s t a b l i -shing. Proceedings 116th T» M» S· of A· I · Μ· Ε· Annual Meeting, Denver, Colorado·

Oprescu, I · , and F · Gheorghiu (1988)· Metalurgia (Romania)· Accepted for publ ica t ion .

[%o

[%o

[%o

[%o[%o

[%o[%o

69

Near-net shape casting of Al-Si alloy E. Essadiqi, S. Caron, F.G. Hamel and J. Masounave Industrial Materials Research Institute, National Research Council Canada, 75 de Mortagne, Boucherville, Quabec, Canada, J4B 6Y4

ABSTRACT

IMRI has been developing thin strip casting with a horizontal twin-roll machine. The A356 aluminum alloy was cast. The strip was 3 to 5 mm in thickness and 100 mm in width. The microstructure of the as-cast material was characterized. The secondary dendrite arm spacing (SDAS) was found to vary from 3 to 10 μπι close to the wheel surface and at the central zone, respectively. The mechanical properties were determined. The ultimate tensile strength (UTS) and ductility increase with decreasing SDAS while the yield strength is almost constant. The UTS varies from 196 to 233 MPa, the elongation from 7.5 to 16.8% and the value of the yield strength is around 95 MPa. The mean value of the SDAS varies from 6 to 4.32 μιη when the speed goes from 7 to 3.4 RPM. Thermal analysis of the process was conducted by reproducing cyclic thermal variation of the roll surface. The heat transfer coefficient was found to be around 30kW/m2-K.

KEYWORDS

Strip casting, twin-roll caster, Al alloys, solidification, microstructure, heat transfer, mechanical properties.

INTRODUCTION

The objectives of the near net shape casting processes are elimination of at least a part of hot rolling and improvement of mechanical properties of the strip. There is strong correlation between secondary dendrite arm spacing (SDAS) and mechanical properties. The refinement of the structure achieved by increasing cooling rate produces an increase in strength and ductility (Spear and Gardner, 1963; Oswalt and Misra, 1980; Armstrong and Jones, 1979). The present study uses a horizontal twin-roll caster for strip production of A356 aluminum alloy. Particular attention is given to the microstructure of the as-cast alloy, the mechanical properties and thermal simulation of the process.

EXPERIMENTAL PROCEDURE

The configuration of the twin-roll caster shown in Fig. 1 is used to cast a horizontal strip. The twin-rolls are made of 4340 steel with no cooling system. The radius and width of the rolls are 30 and 10 cm respectively. The process is controlled by the speed of the rolls which is decreased from 15 RPM until a solid strip could be produced. The chemical composition of Al-Si cast and the experimental conditions are given in Tables 1 and 2, respectively. The physical properties of the cast alloy and the roll material are summarized in Table 3. The cast strip was 3 to 5 mm in thickness and 100 mm in width. 150 kg of aluminum alloy was melted in an induction furnace. The melt was poured into a tundish and injected into the twin-roll caster through a refractory nozzle. The casting speed was 2 to 9 RPM depending on the


Fig. 1. Schematic configuration of the horizontal twin-roll caster

Table 1 Chemical Composition (wt%) of A356 Al Alloy

Al

Rem.

Si

7.39

Fe

0.33

M£

0.48

Mn

0.14

Zn

0.24

Cu

0.19

strip thickness and the contact angle which was set in the range of 10 to 20 degrees. In order to determine the yield strength, the ultimate tensile strength and the elongation of the as-cast A356, tensile test were performed on three longitudinal and three transverse samples for each condition using standard procedures (ASTM Standards). The temperature of the two rolls was measured by means of two chromel-alumel thermocouples that were inserted 2 mm below the roll surface.

Table 2 Physical Properties of A356 Al Alloy and Roll Materials 4340

Material

A356A1 AUoy

4340 steel

Properties Density p (kg/m3) 2600 Specific heat Cp (J/kg-K) 1100 Solidus Ts (°C) 555 LiquidusTL(°C) 615 Thermal conductivity k (W/m-K) 221 Latent heat of fusion Hf (kJ/kg) 389 Density p (kg/m3) 7870 Specific heat Cp (J/kg-K) 585 Thermal conductivity k (W/m-K) 36 Convection coefficient (W/m2-K) 20 Emissivity ε 0.6

References 1 Stefanescu and co-workers (1988) * Metals Handbook (1979) Metals Handbook (1979) estimated from Touloukian (1970) Metals Handbook (1979) Metals Handbook (1979) Touloukian (1970) estimated from Smithells (1967) Kreith (1980) Kreith (1980)

* Average value of the liquid and solid specific heat (Stefanescu and co-workers, 1988; Touloukian, 1976).


Table 3 Experimental Conditions for the Strip Casting of A356

Contact Angle (degrees)

10

20

Liquid Temperature (°Q

630 630 630 630 625 625 635 625 635 625

Speed ] (RPM) 1

7.20 6.00 4.70 3.37 1 8.60 6.00 5.46 3.95 3.23 2.40 |

EXPERIMENTAL RESULTS AND DISCUSSION

Microstructure

A typical microstructure of a longitudinal section of the as-cast Al-Si alloy is presented in Fig. 2. Dendrites close to the chill surface are finer and oriented almost vertically parallel to the heat flow. In the central zone the structure is equiaxial (Fig. 3). Some porosity is present in the interdendritic area, particularly in the central zone which solidifies last (Fig. 4). The amount depends strongly on the experimental conditions such as the contact angle, the speed and the roll force. Under rolling conditions the dendrites are tilted as shown in Fig. 2. This is probably due to the liquid flow between rolls and the deformation which starts when the central zone is still in the mushy state. The structure consists of aluminum solid solution (white area) with Al-Si eutectic in the interdendritic spaces (dark area). The SDAS varies from 3 to 10 μιη close to the wheel surface and at the central zone, respectively. The nature of the variation of the SDAS with respect to the distance from one wheel surface is shown in Fig. 5. The secondary dendrite arm spacing increases as the distance from the substrate surface increases until the

Fig. 2. As-cast structure of A356 alloy in the longitudinal section close to the wheel surface (5.1 RPM, 4.5 mm in thickness).

72 PRODUCTION AND ELECTROLYSIS OF LIGHT METALS central zone is reached, and then starts decreasing again. The influence of the roll speed on the mean values of SDAS, in the case of 10° as constant angle, is shown in Fig 6. This diagram illustrates the refinement of the structure with decreasing speed. The SDAS varies from 6 to 4.32 μπι when the speed goes from 7 to 3.4 RPM. It is of interest that Fig 6 also indicates the saturation of the refinement of the dendrite spacing at a value of 4.5 μπι when the speed reaches 5 RPM. Decreasing the speed below this value will result in no further refinement of the structure. This refinement of the structure correlates very well with the roll force which increases with decreasing speed as can be seen in Fig. 7 in the case of 10° as contact angle. As the speed is decreased, the residence time of the strip between the rolls increases; therefore shell thickness between rolls increases with decreasing exit temperature and increasing roll force. The latter makes the heat transfer more efficient and enhances the cooling rate with refinement of the solidification structure when the central zone is still in the mushy state. At lower speed, when the

Fig. 3. As-cast structure of A356 alloy at the central zone (5.1 RPM, 4.5 mm in thickness).

Fig. 4. As-cast structure of A356 alloy at the center of the strip showing some porosity (A and B).


alloy solidifies completely before the exit, the distance between the secondary dendrites, already formed will be reduced by deformation.

The relationship between the secondary dendrite arm spacing and the solidification time of the A356 alloy was determined by Spear and Gardner (1963). It can be given by:

λ = 10.48 t|?·337

where λ is SDAS in μιη and tf the solidification time in s. 10

8

9

(1)

E 6

< 5 Q

l·

r V* !■ w*

7

/ · <·

-

• •

» ·

· · · •

AS CAST A356

• • ·

—^ ·

# w

• \

•

DISTANCE mm

Fig. 5. Dendrite arm spacing versus distance from one wheel surface of the as-cast A356 alloy (7.2 RPM, 2.7 mm in thickness).

ROLL SPEED RPM

Fig. 6. Dependence of the mean SDAS on roll speed for as-cast A356 in the case of 10° contact angle.

200

z

LU Ü

p 100

o

AS CAST A356

CONTACT ANGLE: 10°

4 5 6 7 ROLL SPEED RPM

Fig. 7. Effect of roll speed on roll force for A356 cast alloy under 10° contact angle.


The solidification times estimated from the SDAS by using this relation were found to vary between 0.03 and 0.87 s close to the wheel surface and the central zone, respectively. The corresponding cooling rates vary approximately from 2000 to 100°C/s.

Heat Flow Model

In order to explain the experimental results a thermal numerical model of the strip casting process was developed. The main objective was to accurately model the macro variables of the process such as the exit temperature of the strip, the thermal gradient in the rolls and the solidification rate. The mathematical evaluation presented here is obtained from a one dimensional non-steady-state heat flow model. It takes into account the variable geometry of the metal pool over the contact angle and permits the use of non linear variation of all the thermodynamical properties over the temperature range. For simplicity, constant parameters were used in the calculations, averaged between the solid and liquid phases of the alloy (Table 2). The general equation for unidirectional transient conduction heat transfer may be written as (2):

pCp — = _ k — U < J (2) dt dx \ dxj

where Q is a heat source term, T, the temperature evaluated at a single point, p? the density, Cp, the specific heat and k, the thermal conductivity of the material. The term Q is used to express the latent heat release for solidification. The "enthalpy method" has been retained here(Voller and Cross, 1981; Gutierrez and Szekely, 1986). The general equation was solved by the finite difference method using the explicit scheme. The heat transfer coefficient between the rolls and the strip was calculated by reproducing the cyclic thermal variation close below the roll surface. The recording of roll temperature reading obtained during the casting of 4.5 mm thick strip with 20° contact angle is given in Fig. 8 (continuous line). The rapidly ascending slopes represent the moment when the thermocouples region is getting in contact with the liquid pool and the decreasing part of the curve occurs when the surface of the roll cools down away from the strip. The simulated curve (dashed line) is in good agreement with the experimental one, using a heat transfer coefficient value of 30 kW/m2-K except for the first thermal peak where 17 kW/m2-K was used. This is mainly due to the fact that at the start-up of the process the metal pool is not yet filled. The above heat transfer coefficient value of 30 kW/m2K is used in this study to simulate the strip casting of A356 alloy.

300

200 LU QC

1-< tu Q.

HI H 100

0 0 1 0 2 0 3 0 4 0

TIME SEC.

1 ■

-

l· N i

N N

l\

\

■

CALCULATED

EXPERIMENTAL

-

-

Fig. 8. Roll temperature variation during the casting of 4.5 mm thick strip of A356 with 20° contact angle.


Mechanical Properties

The results from the tensile tests performed on the as-cast A356 alloy are collected in Table 4. The UTS and the yield strength are plotted versus roll speed at selected contact angle in Figs. 9a and b, respectively. It is very obvious from Fig. 9a that the yield strength is not affected by speed while the UTS increases as speed decreases. At speeds lower than 5 RPM the ultimate tensile strength curve displays asymptotic behavior due to the saturation of the secondary dendrite arm spacing (Fig. 6). In a similar manner, as indicated in Fig. 9b, the ductility increases as the roll speed is decreased, reaching an asymptotic value at a speed of about 5 RPM. It can also be seen that in the case of a contact angle of 20° the ductility is higher than in the case of 10°. The UTS varies from 196 to 233 MPa, the elongation from 7.5 to 16.8% and the value of the yield strength is around 95 MPa. Available data on the mechanical

Table 4 Mechanical Properties of Strip-Cast A356 Alloy

Contact Angle (degrees)

10

20

Speed (RPM)

7.20 6.00 4.70 3.37 8.60 6.00 5.46 5.10 3.95 3.23 2.40

Thickness (mm)

2.70 3.10 3.25 3.62 4.50 4.65 4.50 4.50 4.65 5.00 5.30

Yield Strength 0.2% (MPa) 103.7 101.5 100.7 103.6 116.0 98.0 92.0 93.3 87.2 84.0 82.0

Ultimate Tensile Strength (MPa) 196.0 205.0 226.2 232.5 163.0 197.0 228.0 228.0 233.0 227.1 230.0

Elongation %

7.5 11.0 11.2 12.0

1.3 9.7

15.1 14.5 16.8 16.7 16.6

300

200

CC

100

AS CAST A356

CONTACT ANGLE

■ YIELD STRENGTH • TENSILE STRENGTH D YIELD STRENGTH A TENSILE STRENGTH

□ D

4 6 8 ROLL SPEED RPM

20

< 2 10

1 0

\ » · ·

1

b ■ 1 1 l

AS CAST A356 ■ 10° contact angle • 20° contact angle

1 \

4 6 8 ROLL SPEED RPM

1 0

Fig. 9. Dependence of the mechanical properties of the as-cast A356 on roll speed, a) yield strength and ultimate tensile strength; b) ductility.


properties of the as-cast A356 alloy obtained by different casting processes are listed in Table 5 (Rooy, 1988; Chadwick, Yue and Granger, 1988). Our results are comparable to those of the squeeze-cast alloy, particularly at low speeds. This agreement is related to the effect of pressure on heat transfer between the liquid and the substrate which results in high cooling rates and thus refines the solidification structure. The effect of SDAS on the UTS, yield strength and elongation in the case of a contact angle of 10° is shown in Fig. 10a and b. The UTS and ductility increase as the SDAS decreases while the yield

Table 5 Mechanical Properties of the As-Cast A356 Produced Using Various Processes

Casting Process

Sand cast Drill cast

Squeeze-cast

Ultimate Tensile Strength MPa

159 180 195

Yield Strength MPa 83 90

124

Elongation % 6 5

15

References 1

E.L. Rooy (1988) Chadwick and Yue (1989) Granger (1988)

2

o z o

29U

240

190

140

"*"^

-

■

AS CAST A356

■ YIELD STRENGTH

• ULTIMATE TENSILE STRENGTH

• ""~*~ - ^

_ ■

-

a

5 6

SECONDARY DENDRITE ARM SPACING m

20

18

16

14

12

10

8

6

4

2h

AS CAST A356

Contact angle: 10

1 5 6 7

SECONDARY DENTRITE ARM SPACING urn

Fig. 10. Effect of SDAS on the mechanical properties of the as-cast A356. a) yield strength and UTS; b) ductility.


strength is almost constant. It is well known that the mechanical properties of Al alloys are improved by structure refinement (Spear and Gardner, 1963; Oswalt and Misra, 1980). Particularly, the refinement of SDAS increases the tensile strength and ductility keeping the yield strength constant. Also, as the dendrite arm spacing decreases, the porosity and the intermetallic constituents such as iron aluminide crystals become smaller and hence do not exhibit their normal embrittling effects (Chadwick and Yue, 1989). The mechanical properties of the strip-cast A356 are improved by increasing roll force. This is due to the elimination of microscopic pores in the interdendritic zone by hot working of the strip (Singh and Flemings, 1969).

Table 6 summarizes the roll force, the calculated volume fraction of the solid at the exit and the measured and calculated exit temperature of the strips cast under different conditions using 30 kW/m2-K as heat transfer coefficient. At high speed (> 4 RPM) and volume fraction of solid lower than 85% there is agreement between the calculated and the experimental exit temperatures. At low speed (< 4 RPM), the experimental temperatures are lower than the calculated ones. Under the condition of roll speeds of 3.23 and 2.43 RPM, the strip solidifies completely before the exit and deformation of the plate follows, which increases the roll force. During rolling of the strip the heat transfer coefficient was higher than the value of 30 kW/m2-K used during solidification of the alloy, which can explain the difference between the calculated and experimental exit temperatures.

Table 6 The Roll Force, the Calculated Volume Fraction of the Solid and the Measured and Calculated Exit Temperatures of the As-Cast A356 Alloy

Roll Force (kN)

17 89 73 82

182 1 171

Speed (RPM)

6.00 5.46 5.10 3.95 3.23 2.40

Volume Fraction of Solid % (calculated)

78 83 73 98

100 100

Exit Temperature (°C)

experimental Calculated 557 553 569 520 487 431

560 1 558 565 549 541 527 |

CONCLUSION

The present investigation involved a study of the microstructure and the mechanical properties of a strip-cast A356 alloy produced by a horizontal twin-roll caster. The following conclusions were reached:

- The secondary dendrite arm spacing varies from 3 to 10 μπι close to the wheel surface and at the central zone, respectively. The corresponding cooling rates vary approximately from 2000 to 100°C/s.

- The mean value of the SDAS decreases with decreasing roll speed. An asymptotic value of 4.5 was reached at approximately 5 RPM.

- The heat transfer coefficient involved in the strip casting of A356 alloy was found to be 30 kW/m2-K.

- The UTS and elongation were improved by refinement of the structure. They vary from 169 to 233 MPa and from 7.5 to 16.8%, respectively. The yield strength is almost constant around 95 MPa.

ACKNOWLEDGMENT

The authors would like to thank R. Lavallee, J.-P. Nadeau, D. Simard G. St-Amand and J. Tremblay for their technical help.


REFERENCES

Armstrong, G.R. and H. Jones (1979). Solidification and Casting of Metals. Metal Society, London, pp. 454-458.

ASM Metals Handbook. 9th ed., Vol. 2, ASM International, Metals Park, Ohio, 44073, p. 165. ASTM Standards. (1986), E8M, Vol. 01.02. Chadwick, G.A. and T.M. Yue (1989). Metals and Materials. 5, pp. 6-12. Croft D.R. and D.G. Lilley (1977). Heat Transfer Calculations Using Finite Difference Equations.

Applied Science Publishers, London, p. 107. Elwin, P. and L. Rooy (1988). Metals Handbook. 9th ed., Vol. 15, "Casting", ASM International,

Metals Park, Ohio 44073, pp. 743-770. Granger, D.A. (1988). Metals Handbook. 9th Ed., Vol. 15 "Casting", ASM International, Metals Park,

Ohio 44073, pp. 159-168. Gutierrez, E.M. and J. Szekely (1986). Metall. Trans.. 17b. pp. 695-703. Kreith, F. and W.Z. Black (1980) Basic Heat Transfer. Harper and Row, New York, pp. 258 and 289. Oswalt, K.J. and M.S. Misra (1980). AFS Transactions. 89, pp. 845-862. Singh, S.N. and M.C. Flemings (1969). Trans. Met. Soc. AIME, 245, pp. 1803-1809. Spear, R.E. and G.R. Gardner (1963). AFS Transactions. 71, pp. 209-215. Smithells, C.J. (1967). Metals Reference Book. 4th ed., Vol. 3, London Buttherworths, pp. 8.1,

14.16. Stefanescu, D.M., D. Bandyopadhyay and G. Upadhya (1988). Casting of Near Net Shape Products,

Y. Sakai et al. (Ed.), The Metallurgical Society, pp. 153-165. Touloukian, Y.S. and E.H. Buyco (1970). Thermophvsical Properties of Matter. Vol. 4, New York,

E1F1 Plenum, pp. 917, 1152. Voller, V. and M. Cross (1981). Int. J. Heat Mass Transfer. 24, pp. 545-556.

79

Lost foam casting of aluminum alloys: metallurgical aspects L. Wang, S. Shivkumar and D. Apelian Aluminum Casting Research Laboratory, Department of Materials Engineering, Drexel University, 32nd and Chestnut Streets, Philadelphia, Pennsylvania 19104, U.S.A.

ABSTRACT

The metallurgical quality of simple lost foam castings has been evaluated. Strips of constant thickness were produced under carefully controlled conditions with alloys 319, 356 and 390. The castings were examined by optical and scanning electron microscopy. The mechanical properties were estimated from machined tensile specimens. The results indicate that because of endothermic degradation of the polymer pattern at the metal front, steep thermal gradients are present in the casting. Consequently, dendrite arm spacing and grain size decrease with increasing distances along the length of flow. Furthermore, the distribution and the morphology of Si, Cu and Fe-rich phases depend on location in the casting. The porosity levels in the casting have been estimated by two different techniques to be of the order of 1.5 to 2%. Porosity levels of up to 7% have been detected near the surface of the casting. The ultimate tensile strength and %elongation increase slowly with distance from the downsprue. Defects such as folds and black inclusions are often observed on fracture surfaces.

INTRODUCTION

Foamed polymer patterns which are coated with a refractory slurry and buried in loose sand are used for the production of the casting in the lost foam process. On contact with the molten metal, the polymer pattern undergoes degradation through a series of complex transitions and is gradually replaced by the molten metal to yield the desired component after solidification. This innovative process offers several advantages and has generated considerable attention among foundrymen. The technological features and the operating advantages of this unique casting technique have been summarized in numerous reviews in the published literature t1'6!.

Despite the commercial use of the process for over a decade, there is a paucity of reliable data and information regarding the

fundamental aspects that govern the metallurgical integrity of the casting. The degradation of the polymer in the mold introduces certain unique features during solidification of the casting that affect the quality of the component. Because degradation of the foamed polymer is highly endothermic, requiring energies in excess of 1000 J/g, steep thermal gradients are established in the casting after the mold is filled. An example of the thermal gradients in a strip casting is shown in Fig. 1 t7l. The thermal gradients promote directional solidification in the casting. Under these conditions, microstructural characteristics in lost foam molds can be expected to be significantly different from that in conventional empty-cavity molds [8]. In addition, the interaction of degradation products formed in the mold with the


solidifying metal may lead to the formation of unique defects in the casting. Consequently, an analysis has been conducted to evaluate the metallurgical quality of simple castings produced by the lost-foam process. This work is part of a larger program on the lost foam process and on cast aluminum alloys being conducted under the auspices of the Aluminum Casting Research Laboratory (ACRL) at Drexel University.


The patterns and their gating systems (Fig. 2) were prepared by means of a hot-wire cutter from solid expanded polystyrene blocks and plates with a nominal density of 0.02 g/cm3

and a bead size of about 2 mm (T beads) PI. Strips 1.3 x 2.5 cm in cross section were attached along their periphery to a hollow downsprue with a low-ash hot-melt wax. The assembly of the downsprue and the strips was glued to a pouring cup made of the same material. The patterns were coated by dipping in a commercial refractory slurry (Styrokote 145.3 PM) of fixed density (43 Baume')· The coated pattern was dried overnight at a temperature of 45C in a circulating hot air stream. The prepared pattern was placed in a steel box and filled with loose unbonded sand with a AFS grain size number of 55. A ceramic pouring basin was then positioned on the pattern and secured tightly to avoid displacement during pouring. The metal consisting of 100% pig alloy was melted in a clay graphite crucible using a high-frequency (10 kHz) induction furnace. The metal was degassed thoroughly with high purity (99.999%) nitrogen. A pouring temperature of 750±5C was used in all experiments.

The cast strip was subjected to extensive metallographic analysis. The strip was sectioned at several locations and each section was examined separately by optical and electron microscopy to determine the variation in microstructure as a function of the distance from the downsprue. The LeMont image analysis system was used to characterize the distribution and morphology of the Si particles. A minimum of 30 fields were analyzed for a single specimen. Density measurements were carried out to determine

the extent of porosity in the casting. Image analysis techniques were also used to quantify the amount of porosity. Tensile specimens (ASTM E8) were machined from the cast strip to determine yield strength (0.2% proof strength), ultimate tensile strength and %elongation. At least two tests were conducted under identical conditions. Various defects in the casting were identified and analyzed by microscopic techniques. Other pertinent experimental details are presented elsewhere PI.

RESULTS AND DISCUSSION

The local solidification time varies within the casting and is inversely proportional to distance from the downsprue. Typical solidification times in strip castings poured at a temperature of 800C have been measured to be 250 s and 70 s at a distance of 1 cm and 52 cm from the downsprue respectively $] . Microstructures in the strip at two different locations are shown in Fig. 3 for hypoeutectic (A356.2) alloys. The as cast structure consists of aluminum rich dendrites and other eutectic phases in the interdendritic regions. At locations close to the downsprue, a coarser dendritic structure is observed than at distances far away from the downsprue. The secondary dendrite arm spacing (DAS) is of the order of 20 to 50 μπι and varies inversely with the distance from the downsprue W . In Fig. 4, the microstructures observed in lost foam strip castings are compared with those obtained in sand and permanent mold castings produced under identical conditions. At the entrance of the strip, microstructures in lost foam castings are comparable to those in sand castings. At locations far away from the downsprue, the microstructure in lost foam castings is finer than in sand castings, but is not comparable to those in permanent mold castings. A similar behavior is detected in hypereutectic alloys. In this case, the size of the primary silicon particle decreases gradually along the length of flow.

The grain size in the casting is determined by the extent of undercooling observed during solidification. At low undercoolings, very few nuclei become stable during solidification and consequently, a large grain size is


obtained. At large undercoolings, frequent interior nucleation may occur, resulting in a fine equiaxed structure. The endothermic losses at the metal front lead to significant undercooling of the liquid metal. The undercooling is a function of the location in the casting and is directly proportional to the distance from the downsprue. Since the undercooling increases along the length of the strip, the grain size decreases (Fig. 5). The ASTM grain size number near the downsprue is measured to be about 5.0 and 12.0 at distances of 0 cm and 45 cm respectively W. Hence, in most sections of lost foam castings, the microstructure is finer than in sand castings. The grain size in areas close to the metal front are comparable to the grain size in well grain refined castings produced by conventional bonded sand techniques. The AluDelta instrument was also used to assess the differences in the grain structure between lost foam and sand castings. This device utilizes cooling curves to evaluate grain size and the extent eutectic modification in the casting t10!. A piece of of expanded polystyrene which was cut so that it fit snugly into the AluDelta cup was used to study the solidification characteristics in lost foam castings. When an empty cup was used, the grain size number was of the order of 4 to 5. When the cup was filled with the polystyrene pattern, grain sizes were about 11 to 12.

The local solidification time has a strong influence on the size and morphology of various phases present in the casting. The gradual reduction in local solidification time along the length of the strip leads to a progressive refinement of interdendritic phases in hypoeutectic alloys. Typical microstructures in alloy 319 are shown in Fig. 6 for unmodified and Sr modified alloys. In unmodified castings, Si particles are present as coarse needles. A fine and fibrous eutectic structure is obtained in modified alloys. The copper-rich phase exhibits a typical eutectic structure. The Fe-rich phase primarily adopts a Chinese script morphology. The iron containing compound was identified to be (Fe,Mn)3Si2Ali5. Some (Fe,Mn)3Si2Ali5 particles were present within the aluminum dendrites indicating that there was primary precipitation of this phase

along with aluminum dendrites. Small amounts of FeSiAls needles were also detected in the microstructure. In unmodified alloys (both hypo and hypereutectic), the size and morphology of silicon particles depend on location in the casting. The variation of aspect ratio (AR) and average equivalent diameter (D) of unmodified Si particles in alloy 319 with distance from the downsprue is plotted in Fig. 7. The aspect ratio is defined as the ratio between the maximum and minimum dimensions of the particle. The average equivalent diameter was calculated from the average area of all the silicon particles. The aspect ratio and the average equivalent diameter do not vary significantly up to about 30 cm along the length of the strip. Subsequently, however, there is rapid decrease in both AR and D. The aspect ratio and average particle diameter in lost foam castings are lower than in green sand castings produced under identical conditions. In modified alloys, the Si particle characteristics do not vary significantly with location the casting. In both unmodified and modified alloys, (Fe,Mn)3Si2Ali5 particles are progressively refined along the direction of flow. The concentration of Fe-rich phases decreases gradually with distance from the downsprue. At short distances along the length of the strip (< 5 cm), the (Fe,Mn)3Si2Ali5 phase exhibited a strong tendency to form clusters. With increasing distances, however, the clusters gradually disintegrate and a uniform distribution of (Fe,Mn)3Si2Ali5 particles is observed after a distance of 35 cm.

A strong segregation of Si, Cu and Fe-rich phases was observed in a region near the metal front (Fig. 8). The length of this region was around 1.5 cm. In alloys 356 and 319, silicon concentrations of up to 13% have been measured in this region. By comparison, the concentration of Si particles at other locations in the casting is of the order of 5%. In this region, often, many primary silicon particles were also detected (in hypoeutectic alloys) as can be seen from Fig. 8. Several Si needles with aspect ratios greater than 20 and large amounts of FeSiAls needles with aspect ratios of about 50 to 100 are also observed in this region. The proportion of the (Fe,Mn)3Si2Ali5 phase is


much smaller than that at other locations. The segregation of alloying elements in a region near the metal front may result from the flow of interdendritic liquid metal. It has been established that when an alloy with a wide freezing range (such as alloy 319) is poured against a chill, a solute rich region is obtained in the vicinity of the chill because of interdendritic fluid flow t11].

An examination of the micros true ture indicates significant amounts of porosity. Density measurements and image analysis data yield porosity values of about 1.5 to 1.8% in lost foam castings and about 1.4 to 1.7% green sand castings . In lost foam castings, density is affected by large amounts of gas bubbles that are present in the casting. In green sand cast components, however, extensive interdendritic porosity influences density. The porosity in lost foam samples does not show any significant dependence on the distance from the downsprue. The average pore diameter in the bulk of the casting is of the order of 57 μπι. In this respect, it should be noted that the glue used to join the polymer may have a significant influence on the porosity levels in lost foam castings. Excessive amounts of glue at the joints may lead to increased amounts of porosity in the casting.

The evolution of large amounts of gaseous products in the mold promotes increased porosity levels in the casting. The amount of gas produced in the mold corresponds to about 205 cm3 (STP) per gram of solid expanded polystyrene I 1 2 ] . In addition, liquid degradation products formed in the mold may be trapped in the molten metal. The liquid products are eventually vaporized to gaseous products and may add to the porosity in the casting. Traces of liquid styrene have been observed (characteristic smell and color) in some large gas bubbles. The presence of liquid degradation products in the mold may lead to extensive surface porosity in the casting. It has been established that liquid degradation products may accumulate preferentially at the metal/mold interface and can remain at this location for times of the order of several seconds t 1 2] . The liquid products are gradually vaporized and the gases formed may be trapped in the

solidifying casting. Extensive surface porosity was detected in all the castings. Porosity levels of up to 7% were measured in a region near the surface of the casting. The thickness of this region was of the order of 5 mm. The size of the pore in this region was much smaller than the rest of the cross section. The average equivalent pore diameter in this region was estimated to be about 35 to 40 μπι. The problem of porosity is also exacerbated in the vicinity of the metal front. These sections solidify very rapidly and the probability of a gas bubble generated at the metal front being trapped in the solidifying metal is very high.

The variation of yield strength, ultimate tensile strength and %elongation with the location in the casting is shown in Fig. 9. The yield strength data correspond to 0.2% proof strength values. The yield strength is relatively constant at about 90 MPa along the length of the strip. Both UTS and %elongation increase slowly with distance from the downsprue. The hardness generally increases with distance of flow 19]. The improvement in tensile properties along the length of flow may be attributed to a reduction in the dendrite arm spacing with increasing distances from the downsprue. As DAS decreases, porosity and second phase constituents are dispersed more finely and evenly resulting in an enhancement of mechanical properties.

Two types of defects are often detected on fracture surfaces: folds and black inclusions. Folds are discontinuities in the casting extending inward from the casting surface and severely affect pressure tightness. In some respects, folds are analogous to cold shuts and form because of improper fusion of two streams of liquid metal. A typical photograph of a fold is presented in Fig. 11 (a). It can be seen that the area of the fold contains a relatively smooth and striated surface. The size of a fold is typically of the order of a few mm. In extreme cases, folds whose largest dimension is greater than a few cm have also been identified. The mechanism of fold formation is very complex and has not been clearly understood. The collapse of a gas bubble perhaps offers the best explanation for the formation of a fold


PI. A typical black inclusion on the fracture surface is shown in Fig. 11 (b). This inclusion is generally observed in sections that were filled towards the end of pouring. The size of this defect varied from about a few mm to a cm. Clusters of black inclusions were also observed in some areas of the casting. Preliminary results indicate that the inclusions are introduced into the casting through the coating material ^13λ

CONCLUSIONS

The degradation of the polymer pattern in the mold introduces several unique features during the solidification of the casting. The degradation process is highly endothermic and requires energies in excess of 1000 J/g. Consequently, steep thermal gradients are established in the casting after the mold is filled. Because of the thermal gradients, dendrite arm spacing and grain size in the casting decrease with distance from the downsprue. The dendrite arm spacing and grain size in most of the sections in the casting are smaller than in conventional green sand molds.

In unmodified alloys, the distribution and morphology of Si particles varies with the location in the casting. The aspect ratio and the average Si particle diameter decrease along the length of flow. The silicon particle characteristics do not vary significantly in modified castings. The copper and iron-rich phases are progressively refined along the length of flow. Both FeSiAls needles and Chinese script (Fe,Mn)3Si2Ali5 particles are detected in the microstructure. At locations close to the downsprue, clusters of large particles of (Fe,Mn)3Si2Ali5 are observed. There is strong segregation of Si, Q1AI2 and FeSiAls in a region near the metal front. The length of this region is around 1.5 cm.

Porosity levels in the casting do not vary significantly along the length of flow and are of the order of 1.5 to 2%. The average pore diameter is measured to be about 55 to 60 μπι. Porosity levels are comparable to those in conventional green sand molds. Large amounts of porosity may be present near the surface of the casting. Porosity values of up to 7% have been observed in a region near

the surface of the casting. The thickness of this region is about 5 mm. The size of the pore in this region is much smaller than in the rest of the cross section. The ultimate tensile strength, %elongation and hardness improve with distance from the downsprue. The yield strength is relatively unaffected. Defects such as folds and black inclusions are often detected on fracture surfaces.

REFERENCES

1. AJ. Clegg, Foundry Trade Journal Int., 9(30), (1986), 51-69

2. HJ . Heine, Foundry M & T, 114(10), (1986), 36-41

3. G. del Gaudio, G. Serramoglia, G. Caironi and G. Tosi, Metall. Sei. and Tech., 3(3), (1985), 76-86

4. AJ. Clegg, Foundry Trade Journal, 145(3143), (1978), 393-403

5. AJ. Clegg, Foundry Trade Journal, 145(3144), (1978), 149-160

6. E J . Sikora, Trans AFS, 86, (1978), 65-68

7. S. Shivkumar, B. Gallois and E. Cesmebasi, submitted for publication in Met Trans B

8. S. Shivkumar , Ph .D. thes i s : "Fundamental characteristics of metal flow in the full mold casting of aluminum alloys", Stevens Institute of Technology, Hoboken, NJ, (1987)

9. S. Shivkumar, L. Wang and B. Steenhoff, Trans AFS, 9 7 , 1989, pending publication

10. D. Apelian, G. Sigworth and K . Whaler, Trans AFS, 92, (1984), 197-207

11. M.C. Flemings: "Solidification Processing", McGraw Hill Inc., New York, (1974)

12. S. Shivkumar and B. Gallois, Trans AFS, 95, (1987), 791-800

13. L. Wang, B. Steenhoff, S. Shivkumar and D. Apelian, Fourth annual conference on "Evaporative foam pattern casting", Rosemont, Illinois, June 6-7, 1989


1110,

1050

ro 990

E

POURING BASIN

9 3 0 k ή

870 20 3 0 4 0

Distance (cm)

STRIP

HOLLOW DOWNSPRUE

Fig. 1 Temperature gradients in a strip casting as a function of distance of flow. Symbols correspond to measured values while lines are calculated from a theoretical model V\

Fig. 2 Schematic of strip pattern and gating system.

Fig. 3 Typical microstructures in the casting (A356.2, 100X). a) 1 cm from downsprue b) 60 cm from downsprue.

Fig. 4 Microstructures in castings produced by different processes (A356.2, 100X). a) Sand casting b) Permanent mold casting.


Fig. 5 Photographs illustrating the grain structure in lost foam castings at two different locations (319, 3X). a) 1 cm from downsprue b) 54 cm from downsprue

Fig. 6 Optical micrographs in unmodified and Sr modified castings (Alloy 319, 500X). a) 1 cm from downsprue (unmodified) b) 46 cm from downsprue (unmodified) c) Modified (20 cm from downsprue)


.2 3.3 "a DC ü S. 2.9 <

2.5

• Diameter o Aspect Ratio

14 28 42 Distance from downsprue (cm)

56

4.0

3.5 ^

0)

2-5 I 5

2.0

Fig. 7 Variation of Si particle aspect ratio and average equivalent diameter with distance from the downsprue (Alloy 319).

2.0

σ> 1.0Η c .o LU

0.0

150

"a" Q. w 120H jr σ> c Φ W 901

60

D % Elongation

# Tensile

o Yield

-() o σ σ o

Fig. 8 Optical micrograph illustrating the segregation of Si in a region near the metal front (A356.2, 200X).

27.0

22.5 _

JC

|-18.0 j? 0)

13.5

9.0 0 10 20 30 40 50

Distance from downsprue (cm)

Fig. 9 Variation of yield strength, ultimate tensile strength and %elongation with distance from downsprue (Alloy 319).

Fig. 10 Photograph of defects on fracture surface (3X). a) Fold 1: Fracture surface 2: Fold area b) Inclusion l:Fracture surface 2: Inclusion area

87

Evaluation of feeding characteristics in cast aluminum alloys J. Zou, K. Tynelius, D.G. Kim, S. Shivkumar and D. Apelian Aluminum Casting Research Laboratory, Department of Materials Engineering, Drexel University, 32nd and Chestnut Streets, Philadelphia, Pennsylvania 19104, U.S.A.

ABSTRACT

The production of sound castings often entails supplying adequate amounts of feed metal during solidification of the casting. If the feed metal is insufficient, several defects such as porosity, shrinkage and surface sinks can occur in the casting. The extensive data and information in this area has been reviewed in order to obtain a fundamental understanding of the complex phenomena occurring during feeding of aluminum castings. The influence of process parameters such as alloy composition, grain refinement, modification and cooling rate on feeding properties has been analyzed. Various modeling techniques that have been adopted to study the feeding behavior have been described.

INTRODUCTION

The production of sound aluminum castings often entails supplying adequate amounts of feed metal during solidification. The mode of solidification in aluminum alloys is such that there is significant volume contraction upon freezing. At low solid fractions, this shrinkage can be accommodated by a mass movement of liquid and solid phases. The extent of this feeding process can be enhanced by incorporating risers in the gating system or by using chills to promote directional solidification. But as the fraction of solid increases, the dendrites grow to form a coherent network and the shrinkage can only be compensated by the transport of liquid through the channels between the dendrites. When these channels become too narrow to permit the flow of liquid, microshrinkage defects are created. In addition, aluminum alloys are highly susceptible to the absorption of hydrogen. The dissolved hydrogen is liberated during solidification and results in the formation of voids in the casting. In the interdendritic regions, the combination of solidification shrinkage and gas generation also leads to microporosity defects in the casting.

Because of the importance of the problem, a vast body of technical information is available in this area. It is our intent to review the

available information in order to obtain a fundamental understanding of the physical phenomena occurring during the feeding of aluminum castings. The existing data will be analyzed to determine the influence of several processing variables. This work is part of a larger program on the solidification characteristics of cast aluminum alloys.

FEEDING MECHANISMS

Various feeding mechanisms that may operate during solidification of aluminum alloys have been described by Campbell1 (Fig. 1). These mechanisms can be classified as:

• Liquid feeding • Mass feeding • Interdendritic feeding • Solid feeding

Liquid feeding involves the transport of liquid metal to compensate for the solidification shrinkage. Both liquid and solid phases are transported during mass feeding. Liquid feeding and mass feeding operate at low solid fractions and consequently, the resistance to flow is relatively small. Interdendritic feeding occurs after the formation of the dendrite network. Solid feeding is the inward movement


of the solidification shell to compensate for volume contraction.

An inherent deficiency with Campbell's representation of feeding is that it does not account for the diffusion of hydrogen liberated during solidification. Furthermore, Campbell emphasizes fluid flow along the longitudinal direction (Fig. 1). During solidification, dendrites grow from the mold walls and fluid flow should also occur in the transverse direction. It should be noted that the transverse flow occurs parallel to the primary dendrite arms and compensates for the solidification shrinkage in interdendritic regions. A schematic illustration of the various types of fluid flow during the solidification of the casting is presented in Fig. 2. In this figure, bulk flow of liquid from the riser into the casting is classified as macrofeeding. This flow occurs normal to primary dendrite arms which are growing from the mold walls and includes both liquid transport and liquid/solid transports. Fluid flow parallel to the primary dendrite arms (interdendritic flow) is termed as microfeeding.

Since diffusion rates for hydrogen are much larger than the solidification rate, the hydrogen content in the interdendritic regions can be considered to be almost uniform. After diffusion of hydrogen from the interdendritic region to the bulk liquid, hydrogen atoms have to travel relatively large distances to escape into the atmosphere through the riser. If the hydrogen passage is blocked because of solidification, the hydrogen can be trapped in the casting and may promote porosity formation. This condition may be satisfied towards the end of solidification. The dendritic network may restrict the diffusion of hydrogen because of the low diffusivity of hydrogen in the solid phase. The entrapment of hydrogen becomes more probable when the distance between the casting and riser (L in Fig. 2) becomes large. This phenomenon is especially important in thin-walled castings. Hence the effects of hydrogen may be neglected during microfeeding but should be included in models simulating the macrofeeding phenomena.

SOLIDIFICATION AND FEEDING

In most commercial alloys, liquid metal has to flow through a mushy zone in order to

compensate for the volume contraction at the solid/liquid interface. In equiaxed structures which are commonly observed in most castings, the mushy zone exists over the entire casting. The length of the mushy zone is determined primarily by the freezing range of the alloy and the thermal gradients in the liquid. Pure metals and eutectics freeze at a constant temperature and hence the mushy zone is essentially negligible. Alloys of intermediate compositions, however, solidify over a range of temperature and the fluid has to pass through the mushy zone during solidification. Recently, Thevoz2 has shown that a dendritic network is established at a solid fraction as low as 0.2 for equiaxed solidification and Iwahori et al's experimental data indicate that limiting fraction solid for porosity formation is about 0.41 for Al-7%Si alloy (Fig. 3)3. This critical value varies inversely with the silicon content. Hence, it is seen that in long freezing range alloys resistance to interdendritic fluid flow is encountered even during the early stages of solidification. These results highlight the detrimental effect of equiaxed solidification on feeding properties.

The solid/liquid interface morphology influ-ences the formation of the dendritic network during solidification. Planar metal fronts ob-served in pure metals and eutectics contain smooth surfaces and thus exhibit good feeding properties. In long freezing range alloys, the interface is irregular and this interface mor-phology inhibits feeding. Engler and Hen-richs4 have studied the feeding behavior of bi-nary Al-Cu alloys in metallic molds. Their re-sults suggest that feeding properties improve when the solidification front changes from an irregular to planar morphology.

It has been established that fluid flow through the mushy zone is similar to that through a porous media and can be described by DArcy's equation5'6:

where V is the velocity of the fluid, K is the permeability of the media, P is pressure, gr is acceleration due to gravity, μ is viscosity, and fl is volume fraction of liquid. The principal


parameters in Eq. [1], K, pi, μ, ΔΡ and fi have been individually studied by several researchers5"8. The permeability depends on the fraction of liquid and on microstructural parameters such as grain size and dendrite arm spacing (Fig. 4). Al-4.5%Cu alloy freezes in a "pasty" fashion and hence exhibits low permeability. A high value of permeability is observed with Al-4%Si alloy because of the relatively small freezing range. It should be noted that the results reported for Al-4%Si alloy are for columnar structure. For the three alloys compared in Fig. 4, Al-4%Si-0.25%Ti with an equiaxed structure possesses the lowest permeability. This result suggests that permeability in columnar structures may be greater than in equiaxed structures. Furthermore, permeability of the interdendritic liquid varies inversely with the cooling rate and hence decreases with grain size.

Several investigators9"14 have used Blake Kozeny equations15 to develop empirical relationships for permeability. When fluid flow occurs in a direction of parallel to the primary dendrite arms, the permeability equation becomes:

Κ=4.53·10 +4.02-10 (fÔ.l) - ί - j - [2] (l_ r l)

where di is the primary dendrite arm spacing. For fluid flow perpendicular to the primary dendrite arms permeability can be written as:

K=1.73xl0"3.(di/d2)1-09d22fi3(l-fi)-°·749 [3]

U2 is the secondary dendrite arm spacing. When flow is normal to primary dendrite arms, permeability depends on the both di and U2.

The above equations can be used to describe the interdendritic fluid flow in columnar structures. For equiaxed structures, the permeability K under steady-state flow conditions can be expressed as5»7:

K = { [4] d-fi) · Sv

where Sv is the number of intersections (the surface area of solid per unit volume of the solid).

The thermal gradient in the casting can also

affect the length of the mushy zone. A positive thermal gradient from the solid-liquid interface to the riser promotes directional solidification. Conversely, a negative thermal gradient will promote the formation of equiaxed dendrites in the liquid and thus establish barriers for fluid flow. Niiyama et al1 6 have developed an empirical parameter, G/VR (G is the thermal gradient and R is the cooling rate), to characterize the influence of thermal gradient. It was observed that feeding properties decreased significantly when this empirical parameter attained a critical value.

The experimental results of Niiyama et al16 can be explained as follows. The parameter G/VR can be transformed to an interface morphology parameter, G/V, according to the relation:

R = aT/3t = dT/dx · dx/dt = G · V [5a]

where V is the interface growth velocity. Hence:

G2/R = G/V [5b]

When G/V is very high (> (Ti-Ts)/D ~ 104

Ks/mm2), the solid-liquid interface is planar. When G/V decreases to about 1 Ks/mm2, the interface transforms from columnar grain structure to equiaxed grain structure. These theoretical predictions are confirmed from the experimental results of Mahapatra and Weinberg17.

EFFECT OF CASTING PARAMETERS

Alloy Composition

The main alloying elements in cast aluminum alloys such as Si and Cu have a strong influence on the feeding properties. Si is the most common alloying element added to pure aluminum. Silicon increases in volume during solidification and consequently, the susceptibility of the casting to shrinkage defects is reduced. The addition of silicon influences the feeding properties of the liquid metal significantly. As the silicon concentration increases, the morphology of the solidification front is altered. The planar metal front observed with pure metals has a smooth interface. In Al-Si alloys, the formation of


dendrites yields a rough solidification front. The dendritic grains formed in the melt tend to attach to this rough solidification front and thus, a rigid dendritic network will form at an early stage of solidification. Hence, the feeding capacity decreases until the alloying content reaches a critical value, which has been estimated to be about 5% by Hashihura and Komatsu18 (Fig. 5) and by Kotlyarskii et all9.

At Si concentrations greater than about 5%, the feeding properties are determined primarily by the amount of eutectic liquid. The amount of eutectic liquid increases with silicon content20. As the amount of eutectic liquid increases, the roughness of the solid/liquid interface decreases and the feeding properties begin to improve. Feurer21 has calculated relative values of feeding properties for three different Al-Si alloys (4%, 7% and 10% Si) and Al-4.5%Cu. His results suggest that feeding properties in this composition range generally increase with the amount of eutectic liquid. Furthermore, the values for Al-Si alloys are significantly higher than for Al-4.5%Cu. Vazielle and Morice22

have measured the microshrinkage properties in various Al-Si alloys and obtained similar results. The morphology of pores also depends on the Si content. As the silicon concentration increases, the pores become more rounded and their aspect ratio decreases. It has been reported that a minimum of 20% eutectic liquid is necessary for the alloy to be castable23. Best results are obtained when the concentration of eutectic liquid is less than about 80 to 90%. Although alloys with eutectic composition contain almost 100% liquid and solidify essentially at constant temperature, they are susceptible to shrinkage cavities.

Copper reduces the solidification shrinkage slightly, but the effect is not comparable to that of Si. Copper additions generally reduce the corrosion resistance and castability. In addition aluminum-copper alloys are prone to hot tear-ing. When copper is added to Al-Si alloys, the eutectic temperature is lowered and hence, the freezing range increases. Drossel et al24 have shown that in Al-Si alloys, the amount of porosity varies inversely with the eutectic tem-perature. Consequently, several problems are encountered in feeding Al-Si-Cu alloys which are essentially long freezing range alloys.

Iron is present as an impurity in many hypoeutectic alloys. The fraction of insoluble phases increases with the Fe content and hence, feeding characteristics are adversely affected at high iron26 concentrations. A recent publication indicates that in Al-Si-Cu (319) castings containing more than 0.5% Fe, the tendency for shrinkage porosity formation increases with the iron content. This behavior has been attributed to the formation of FeSiAls needles during the early stages of solidification. It is claimed that needle-shaped FeSiAls particles form at temperatures above the eutectic arrest and prevent the compensation of shrinkage during the early stages of solidification.

Initial Hydrogen Content

During solidification of aluminum alloys, hydrogen solubility decreases drastically and pores may form because of increased gas pressure in the melt. Porosity formation depends on gas content and is described by the relation, ΔΡ=2σ/Γ where σ is the surface tension and ΔΡ is the critical pressure that must be exceeded in the pore if a pore nuclei of radius r can grow. A higher hydrogen content in the melt will enhance ΔΡ and will decrease the critical radius of pore nuclei. In this case, the amount of porosity in the casting increases27»28.

Cooling Rate

The cooling rate determines the local solidification time and hence establishes the microstructure of the casting. The length of the mushy zone varies inversely with the cooling rate. Consequently, the fraction of porosity decreases as the cooling rate is increased28"3* (Fig. 6). This behavior can be explained from surface tension effects which can again be represented by the relation ΔΡ=2σ/Γ. Because of the dendritic refinement at higher cooling rates, the critical radius of the pore nuclei (r) decreases. Consequently, the critical pressure required for pore formation will increase and thereby introduce barriers for the nucleation of pores. The amount of hydrogen that can be tolerated before the onset of porosity in the casting depends on the cooling rate. At high cooling rates, even hydrogen contents of about 0.3 cc/100 g of aluminum do not generate


significant porosity in the casting28.

Grain Refinement

Small amounts of grain refiners are added to aluminum alloys in the form of Al-Ti-B master alloys to control the grain structure of castings. It has been reported by several investigators that grain refining significantly enhances the feeding properties and reduces shrinkage related defects. The improved feeding properties of grain refined alloys are associated with increased mass feeding in the casting. When the alloy is grain refined the dendritic network is broken down into small equiaxed grains. The relative ease with which these small fragments of dendrites are transported within the casting contributes significantly to the mass feeding process. Thus mass circulation of these small equiaxed crystals by convection currents reduces the need for liquid feeding and therefore lower amounts of porosities are observed in grain refined castings. The experimental data of Fang and Granger28 illustrate that in A356 alloys, the pore size decreases with the addition of the grain refiner.

Inclusion Removal

Several types of non-metallic inclusions can be present in the melt. These inclusions include foreign particles that are added to the melt such as grain refiners or refractory materials, dross, furnace oxides, unmelted elements or flux material. The presence of inclusions in the melt reduces the feeding properties appreciably. Ruddle and Cibula3% Blondike and Hess33 and Celik and Bennet34 have all shown that inclusions promote porosity formation arid tend to lower the threshold hydrogen content by providing nucleation sites for the formation of porosity. Inclusions may significantly alter the amount and distribution of shrinkage in the casting, although it must be recognized that the total amount of shrinkage is independent of melt cleanliness35. The amount of pipe and sponge type of shrinkage is increased in the presence of inclusions.

Modification

The morphology of eutectic Si in aluminum alloys is generally controlled by the addition of

small amounts of Na or Sr to the melt. The addition of these elements changes the morphology of silicon from acicular needles to a fibrous structure. The influence of modification on the feeding properties has been an area of controversy. While many investigators have shown that modification increases porosity in the casting27»28»36»37

investigations where modification did not have any effect on porosity have also been repor ted 2 4 . The most often cited result regarding modification with Na and Sr is that surface shrinkage is counteracted and that there is a redistribution of micro and macro porosity.

Argo and Gruzleski36 have used the Tatur test to identify differences in the distribution of porosity and shrinkage between unmodified and Sr-modified A356 alloy. Their data show that modified samples are more prone to microporosity and less susceptible to macroshrinkage than unmodified alloys. The data from radiographic examination of castings and from density measurements indicate that modification causes a widespread dispersion of microshrinkage. In this case, pores are relatively large and isolated while in unmodified samples, pores appear to be concentrated, fine and interconnected (Fig. 7). Closset38 also reported a similar change in the distribution of porosity between unmodified and Sr modified A356.0 alloys. Fang and Granger28 have shown that the pore size and volume fraction increase with Sr modification in A356 alloys.

The variation in the amount and distribution of porosity upon modification may be attributed to several factors including a depression of the eutectic temperature and of the surface tension in modified alloys. Argo and Gruzleski36 have proposed a mechanism to explain the formation of microshrinkage in the casting which involves interdendritic feeding and eutectic solidification. In unmodified alloys, the eutectic exhibits an irregular solid/liquid interface (Fig. 7). Because of the irregular interface, small pockets of liquid become trapped between advancing solidification fronts resulting in fine concentrated microporosity in the last areas to freeze. In modified alloys, the solid/liquid interface is regular or planar and porosity is more widely dispersed and appears to be larger. This effect was also noted by


Flood and Hunt39 and by Fredriksson et al40. Modification has a marked effect on the shape of the solid/liquid interface, so it is reasonable to presume that this factor is responsible for the differences in microporosity noted above.

MODELING OF FEEDING BEHAVIOR

A casting process is generally described by three balances: heat, mass and momentum balance. Each balance is written with a partial differential equation to describe the global phenomena to obtain a "macroscopic model".

Basic equations

Heat balance

If the heat transfer due to interdendritic fluid flow is negligible, heat conduction equation can be written as:

„ 3T J 3 T 3 TI 3f!

9t \3x 2 dy2) 3t where κ is thermal conductivity, Lv is latent heat, fi is volume fraction liquid, p is density Cp is specific heat. The first term on the left hand side accounts for heat accumulation. The first term on the right hand side represents the heat flux by conduction, and the second term is the evolved latent heat. In order to solve Eq. [6] the relation between fraction liquid and temperature is required. To achieve this objective, it can be assumed that solute diffusion in liquid is much faster than in solid and Scheil's equation can be used for macroscopic modeling. Recently, Zou41, Rappaz and coworkers42 '44 have developed techniques that interrelate micro and macro models such that they can be used to simulate the microstructure evolution in the bulk casting.

Mass balance (continuity equation)

In the liquid , a mass balance equation can be written to account for the flow of liquid. This equation should include the effects of porosity formation and the variation of density during solidification. Hence:

( i _ ^ W + ^ + ^ =^v [7] I pjdt dx dy 3t

where ps and pi are density of solid and liquid respectively, fv is fraction of porosity45. The first term on the left is the amount of shrinkage, and the second and third terms represent the amount of liquid input by interdendritic flow. The term on the right is the rate of of formation porosity.

Conservation equation Of gas content

If the amount of hydrogen diffusing into the atmosphere can be neglected, a mass balance equation can be written for the hydrogen45:

[HJ = (1-f^HJ + f, [HJ + α Η *-φ^ [8]

where Ho is initial hydrogen content, Hs and Hi are hydrogen contents in the solid and liquid, OCH is constant, Pg is gas pressure. The last term on the right represents porosity formation. The hydrogen content in the solid and liquid are given by Sie vert's law45: [HJ = k sH- J>1'2 [9a]

[HJ = k1H- P* / 2 [9b] where kSH and kin are equilibrium constants in solid and liquid.

Application in microporosity

Kubo and Pehlke45 (1985) have reported their simulation results (FDM) for the microporosity in Al-4.5%Cu alloy by using Eqs. [1] and [6] to [9]. There appears to be a good agreement between calculated and measured results. According to their calculation, the amount and the radius of porosity in Al-4.5%Cu plate castings increase with initial hydrogen content and decrease with cooling rates.

Application in centerline porosity

Centerline porosity is often found in steel casting46-4^ and sometimes in die castings48. The three basic equations (heat balance, mass balance and D Arcy's equation) are also used to predict centerline porosity in castings. As shown in Fig. 849, the solidification diagram in a plate casting depends on the freezing range of the alloy. Pure metals such as aluminum and copper, near eutectic Al-Si alloys and steel, have a short freezing range. Hence the feeding distances are short and the alloys are fed with


relative ease. But Al-7%Si and Al-6%Si-4%Cu have long freezing range and are difficult to feed. In particular, several practical feeding problems are experienced with Al-Si-Cu (319) alloys. These alloys freeze over a range of 90 °C and are prone to severe microporosy.

Davies and Moe 4 8 have calculated the solidification pattern of some practical aluminum alloy die castings (e.g. Al-5%Si-3%Cu and Al-7%Si-Mg). Moosbrugger and Berry49 have calculated the feeding range of A357 alloy using FEM method (FDM method is used by Davies and coworkers47»48). Their data are in good agreement with the results of Davies47. Niiyama et al16»50 and Minakawa et al4 6 have used Davies's model to predict the centerline porosity in steel castings. Three parameters were used to assess porosity formation for several casting geometries: tf (local solidification time), G (thermal gradient), and G and R (cooling rate). It was found that G/VR is the best parameter to define the onset of centerline porosity in steel castings (Fig. 9).

Prediction of microporosity

The formation of different types of microporosity have been individually addressed by several researchers28»45»51. Poirier et a l 5 i

have studied the microporosity between primary dendrite arms. The microporosity between discrete dendrite grains has been studied by Fang and Granger28 , while Murakami et al52»53 have investigated surface porosity. The results of these investigations have shown that porosity increases with distance from the chill to the casting. It should be noted that although several nucleation mechanisms of porosity in aluminum alloys have been proposed by different inves t iga tors* 7 » 3 * 1 » 5 4 ' 5 5 , only simple mechanisms of porosity formation are still used in computer modeling28'45»51.

Prediction pfinter-dendrite-grm micrpporpsjty

Inter-dendrite-arm microporosity often exists in alloys which have a fully-developed dendritic structure at a low solid fraction and sometimes in directionally solidified eutectics. Poirier et al51 assumed that a gas pore is stable (will not shrink) provided that the supersaturation, or the excess pressure, in the gas is sufficiently great

to overcome the surface tension when the gas phase has a radius (concave to liquid) that is small enough to fit in the interdendritic space. This requirement is expressed as:

P g - P = 4 a / ( f p d i ) [10]

The formations of porosity during solidification can then be predicted with the following assumptions:

• D'Arcy's law (K=C-fi2) is valid for interdendritic fluid flow.

• The variation of fraction solid with temper-ature can be described by Scheil equation.

• The hydrogen content is uniform in the liquid and solid phases.

From their analysis, Poirier et al51 have shown that gas bubble pressure increases with volume fraction of liquid. Under conditions that the gas pressure is below the pressure in the liquid, porosity does not form. The calculated volume percent porosity increases with primary dendrite arm spacing. This result indicates that porosity is proportional to primary dendrite arm spacing which is expected to vary inversely with the cooling rate, solidification time and thermal gradient. The predicted values of porosity as function of initial hydrogen concentration compare well with experimental data, although the conditions of the experiment are not stated clearly.

Prediction ofinter-dendrite-grain microporosity

The inter-dendrite-grain microporosity exists between equiaxed dendrite grains for eutectic structures. For alloys with a large eutectic volume ratio (e.g. Al-7%Si, Al-4%Si-3%Cu alloys), the porosity forms in the eutectic liquid rather than between the interdendritic arms28. In this case, the model is based on mass balance between the amount of hydrogen which enters the pore and the amount of hydrogen which is rejected at the solidification front due to the difference in solubilities of hydrogen in the solid and liquid states. Fang and Granger28

have utilized these principles to evaluate pore size and distribution by considering the growth of pores during solidification. The growth of the pore is related to the hydrogen content and to the local solidification characteristics, especially the grain size. It is assumed that: a)


hydrogen content is uniform b) density of the pore is proportional to the density of grain d) spherical gas pore d) hydrogen solubility in eutectic is same as that in dendrite. Under these conditions, pore growth equations for various solidification modes can be written as:

• during equiaxed dendrite solidification

d(4/3-7cr3.n)/dt=-toR2.dR/dt(piCi-psCs)[H]

where r is radius of pore, n is the density of hydrogen gas inside the pore, R is the radius of grains, Q and Cs are hydrogen concentrations in the liquid and solid.

• during eutectic solidification

3 /4 ·πτπ3 = 3/4-7C-n3+(fi- fs)-3/4-7t-R03 ·

(Peu1 * Ceu1 - Peus · Ceus) · K-Teu/pt [12]

where η and rn are the pore radii at the end of stages I and Π, respectively, Ceu1 and v eu are the hydrogen solubilities in eutectic liquid and solid, pei? and peus are the densities of the eutectic liquid and solid, P t is the total pressure inside the pore.

• during end of eutectic solidification

3/4.Krm3=3/4.KTn3+p e uf r .3/4^Ro3 [13]

where ηπ is the final pore radius, ß e u is the volumetric shrinkage coefficient associated with the phase change, and fr is the residual eutectic volume fraction. The calculated poros-ity (diameters and volume fraction) as a func-tion of hydrogen content are compared with the experimental data for A356 alloys. Four inter-esting points are worth noting in the results:

• until hydrogen content > 0.05 or 0.03 cc/100 g, the pore diameter is constant and independent of the hydrogen content the critical hydrogen content depends only on the amount of the residual eutectic liquid assumed, because pores are presumed to form between dendrite grains in the eutectic liquid

• amount of porosity varies inversely with cooling rate;

• amount of porosity increases drastically when hydrogen content exceeds 0.25 cc/100 g.

The variation of pore diameter with grain size for unmodified and modified A356 alloys are plotted in Fig. 10. Sr decreases the surface tension between hydrogen gas and the liquid metal and hence increases porosity. Although values of γ (surface tension) used in the model may not be accurate (since yis very difficult to measure), the calculated values are in a good agreement with experimental data24»30»56.

Conclusions

The feeding behavior of cast aluminum alloys has been studied extensively by experimental and computer modeling techniques and a vast amount of information is available. Experimental methods have been devised to acquire data during and after the solidification of the casting. Cooling curves obtained during solidification of the casting in a variety of mold designs have been analysed to deduce data on equisolidification contours and on thermal gradients in the casting. These results have been utilized innovatively to assess the feeding characteristics in the casting. Techniques based on measuring the extent of interdendritic fluid flow have also been employed to determine the feeding properties since many commercial foundry alloys freeze over a range of temperature and the feeding properties are essentially controlled by the interdendritic fluid flow. Numerous investigators have examined the solidified casting to procure data on the amount, distribution and morphology of feeding related defects.

As a result of these efforts, it is now established that various processing parameters such as alloy composition, cooling rate, initial hydrogen content, grain refinement and eutectic modification have a significant influence on the feeding properties. Si additions greater than about 4 to 5% improve the feeding properties appreciably. In Al-7%Si alloys, the critical solid fraction at which feeding becomes difficult is of the order of 40%. The addition of Cu to Al-Si alloys lowers the eutectic temperature and adversely affects feeding characteristics. Iron which may form insoluble phases during solidification also has a similar effect. The amount of porosity in the casting increases with the initial hydrogen content and is inversely proportional to the cooling rate. The initial hydrogen content should be less than


about 0.08 cc/100 g of aluminum for the production of premium quality castings. While there is general agreement that grain refinement improves feeding properties, the role of eutectic modification is not clear.

Macroscopic models utilizing heat, mass and momentum balance equations have been developed to predict the onset of porosity in the bulk casting. Recently, microscopic models are also being formulated to simulate the evolution of porosity in the casting. The data calculated from the model compare well with experimental results. However, these models are generally valid for columnar structures and need to be extended to simulate the development of equiaxed structures which are present in most commercial castings.

ACKNOWLEDGEMENTS

This research was conducted as a part of an ongoing research project at the Aluminum Casting Research Laboratory (ACRL). The authors would like to gratefully acknowledge the financial support of the consortium of companies supporting the Aluminum Casting Research Laboratory.

REFERENCES

1. J. Campbell, AFS Cast Met. Res. J., Mar. 1969, 1-8

2. Ph. Thevoz, "Modelisation de la solidification dendritique equiaxe", Ph. D. Thesis, No. 765, Ecole Polytechnique Federale de Lausanne, Switzerland, 1988

3. H. Iwahori, K. Yonekura, Y. Sugiyama, Y. Yamamoto and M. Nakamura, Trans. AFS, 1985, 93, 443-452

4. S. Engler and L. Henrichs, AFS Cast Met. Res. J., Sept. 1973, 122-126

5. D. Apelian, M. C. Flemings and R. Mehrabian, Met. Trans., 1974, 5, 2533-2537

6. T. S. Piwonka and M. C. Flemings, TMS-AIME, 1966, (236), 1157-1165

7. T. Takahashi, US-Japan Cooperative Science Program Seminar on Solidification Processing, Dedham, Ma, 1983,45-60

8. K. Murakami, A. Shiraishi and T. Okamoto, Acta Metall, 1984, 32(9), 1423-1428

9. D. R. Poirier, Met Trans., 1987, 18B, 245-255

10. K. Murakami and T. Okamoto, Acta Metall, 1983, 31, 1741-1744

11. R. Nasser-Rafi, R. Deshmukh and D. R. Poirier, Met. Trans., 1985, 16A, 2263-2271

12. K. Murakami, A. Shiraishi and T. Okamoto, Acta Metall, 1983, 31(9), 1417-1424

13. K. Murakami, T. Fujiyama, A. Koike and T. Okamoto, Acta Metall, 1983, 31 (9), 1425-1432

14. B. V. Karlekar and R. M. Desmond, Engineering Heat Transfer, West Publishing Co., St. Paul. Mn, 1977, 321-324

15. G. H. Geiger and D. Poirier, Transport Phenomena in Metallurgy, Addison-Wesley, Reading, Ma, 1973, 94

16. E. Niiyama, T. Uchida, M. Morikawa and S. Saito, AFS Int. Cast Met. J., Sept. 1982, 52-63

17. R. B. Mahapatra and F. Weinberg, Met. Trans., 1987, 18B(6), 425-432

18. K. Hashiura and M. Komatsu, / . Jpn. Inst. Light Met., Nov. 1966, 16(6), 293-297

19. F. M. Kotlyarskii, G. P. Borisov and V. I. Belik, Sov. Cast. Techn., 1986(4), 9-11

20. R. Mai and G. Drossel, Giessereitechn., 1983, 29(2), 46-49

21. U. Feurer, Quality Control of Engineering Alloys and the Role of Metals Science, 1977,131-145

22. P. Grandier Vazielle and J. Morice, Rev. Alum., 1971, (2), 177-187

23. Metals Handbook, Vol. 15, Ninth Edition, American Society for Metals, 1988.

24. G. Drossel, R. Mai and U. Liesenberg, Giessereitechn., 1981,27(6), 167-170

25. K. A. Jackson and J. D. Hunt, TMS-AIME, 1966,236, 1129

26. H. Iwahori, H. Takamiya, K. Yonekura, Y. Yamamoto and M. Nakamura, / . Jpn. Inst. Light Met., 1988, 60(9), 590-596

27. H. Shahani, Scand. J. Met., 1985, 14, 306-312

28. Q. T. Fang and D. A. Granger, to appear in Trans. AFS, 1989

29. P. M. Thomas and J. E. Gruzleski, Met. Trans., 1978, 9B, 139-141


30. J. E. Gruzleski, P. M. Thomas and R. A. Entwistle, Br. Foundryman, Apr. 1978, 71(4), 69-78

31. H. Fredriksson and I. Svensson, Met. Trans., 1976, 7B, 599-606

32. R. W. Ruddle and A. Cibula, / . Inst. of Metals, 1962, 91, 48-57

33. K. J. Brondyke and P. G. Hess, Trans. AIME, 1962,230, 1542-1551

34. M. C. Celik and G. H. J. Bennett, Met. Tech., 1977, 4, 138-144

35. D. E. Groteke, Mod. Cast., 1983, 25-27 36. D. Argo and J. E. Gruzleski, Trans AFS,

1988, 96, 65-74 37. W. Meyer, Aluminium, 1978, 54(11),

700-70 38. B. Closset, to appear in Trans. AFS,

1988 39. S. C. Flood and J. D. Hunt, Met. Sei.,

1981, 16, 287-294 40. H. Fredriksson, M. Hillert and N .

Lange, / . of Inst. of Metals, 1973, 101, 285-299

41. J. Zou, "Simulation de la solidification eutectique equiaxe", Ph. D. Thesis, No. 774, Ecole Polytechnique Federale de Lausanne, Switzerland, 1988

42. M. Rappaz, Ph. Thevoz, J. Zou, J-.P. Gabathuler and H. Lindschied, State of the Art of Computer Simulation of Casting and Solidification Processes, E-MRS, Strasbourg, France, June 1986, Ed. H. Fredriksson, 277-284

43. Ph. Thevoz, J. Zou and M. Rappaz, Solidification Processing 1987, Ranmoor House, Sheffield, U. K, 21-24 Sept. 1987

44. Ph. Thevoz, J. Desbiolles and M. Rappaz, Met. Trans., 1989, 20A, 371-337

45. K. Kubo and R. D. Pehlke, Met. Trans., 1985, 16B, 359-366

46. S. Minakawa, I. V. Samarasekera and F. Weinberg, Met. Trans., 1985, 16B, 823-829

47. V. de L. Davies, AFS Cast Met. Res. J., June 1975, 33-44

48. V. de L. Davies and R. M o e , International Conference on Solidification and Casting, Sheffield, July 1977, 357-362

49. J. C. Moosbrugger and J. T. Berry, Trans. AFS, 1986, 94, 373-380

50. E. Niiyama, T. Uchida, M. Morikawa

and S. Saito, AFS Int. Cast Met. J., June 1981, 16-22

51. D. R. Poirier, K. Yeum and A. L. Maples, Met. Trans., 1987, 18A, 1979-1987

52. C. Y. Liu, K. Murakami and T. Okamoto, Acta Metall. 1986, 34(7), 1173-1178

53. K. Murakami, C. Y. Liu and T. Okamoto, Solidification Processing, Eds. J. Beech and H. Jones, Inst. Metals, London, 1988, 454-457

54. J. Campbell, The Solidification of Metals, ISI Report No. 110, 1967, 18-26

55. H. Shahani and H. Fredriksson, Scand. J. Met., 1985, 14, 316-320

56. G. K. Sigworth, Trans. AFS, 1987, 95, 73-78


1.0

0.8

0.6

o Defects free • Shrinkage porosity defects

4 6 8 1 0 1 2

Si content [%]

Fig. 1 Various feeding mechanisms in the casting1.

Fig. 3 Effect of silicon content on the critical fraction of solid which inhibits feeding in Al-Si alloys3.

Fig. 2 Schematic of macro and micro feeding mechanisms associated w i th interdendritic fluid flow and hydrogen diffusion in the liquid metal.

E

a) Borneol-perofiln alloys b) Street end Weinberg c) Apelfen et el _d) Plwonka end Flemings *y e) Tekehesht

0.05 0.1 1.0

f, Fraction of liquid

Fig. 4 Permeability of dendritic network as a function of fraction liquid for Borneol-paraffin (a), Pb-Sn (b), Al-Si (c,e) and Al-Cu (d) alloys*.


Fig. 5 Effect of silicon content on feeding capacity of Al-Si alloys18.

-Al dendritc

—i Intcrdcndritic| feeding

(a)

Λ1 dendritc

Unmodified

Al dendritc

Intcrdendritic feeding

Fig. 7 Schemat ic i l lustrat ion of microshrinkage formation in A356 alloys36. a) Unmodified castings with short interdendritic feeding distance and small but dispersed porosity. b) Modified castings with l o n g interdendritic feeding distance leading to the formation of large and isolated porostiy.

0.1 0.2 0.3

Hydrogen content [cc/100 gm]

(b)

Fig. 6 Effect of initial hydrogen content on volume fraction of porosity at" several cooling rates for A356 alloy28. Fig. 8 Evolution of the calculated solidification

pattern during the feeding process for short freezing range alloys (a) and for long freezing range alloys (b)49.


SOUND

• > i m m m

SHRINKAGE

V'U1 10 100 1000 10000

tf [mn]

Fig. 9 Interface morphology parameter (G/VR) as a function of local solidification time (tf). Shrinkage defects are observed when (G/VR) attains a critical value16.

Fig. 10 Variation of pore diameter with grain size in unmodified and Sr modified A356 alloys2«.

1.5 •E

Nu

»E 1.0

Q,= 0.31 cc/100g

Ύ =8X10 J/cm

)0 Grain size [μπ\]

101

Production of aluminum alloy castings using evaporative pattern and vacuum techniques J.L. Dion, R.D. Warda, R.K. Buhr, J.R. Emmett and M. Sahoo CANMET, Metals Technology Laboratories, 555 Booth Street, Ottawa, Ontario, Canada, K1A 0G1

ABSTRACT

New casting processes were developed to produce a variety of aluminum castings using evaporative patterns and unbonded sand A vacuum was applied to hold the sand, fill the mould cavity and extract the gases. Various casting parameters studied were melt temperature, degree of vacuum, foam density and mould filling time. Gravity casting by the lost foam process was briefly investigated with and without vacuum to establish the temperature profile near the foam/sand interface. The vacuum pouring was extended to conventional sand moulds. The quality of the castings produced by these methods is described.

INTRODUCTION

Considering the rapidly growing interest in the evaporative pattern casting (EPC) process, new approaches were investigated at CANMET. Initially, aluminum-based automotive castings were studied (1-3). The EPC process has numerous advantages over conventional foundry techniques: namely, great design freedom due to the elimination of core and parting line, reusable sand, no binders or sand additives are required, dimensional accuracy can be achieved, inexpensive single piece flasks can be used, automation might be possible, yield is increased because castings can be produced in clusters from a single sprue and casting cleaning is minimised. As a result, capital equipment and operating costs are reduced (4-6).

METALS TECHNOLOGY LABORATORIES

REPORT: 89-37(OP-J)


The first part of this work consisted of the development of a new low-pressure disposable mould-casting process which produced commercially acceptable castings. This process combines low-pressure die casting and volatile pattern casting techniques with vacuum assistance. Experiments were conducted with different patterns, casting media and gating configurations. The casting parameters studied were: melt temperature, degree of vacuum, foam densities and mould filling time. Results of the above studies led to the design of a modified ladle system to simulate the characteristics of the low-pressure unit for the production of aluminum and copper alloys and grey iron castings. The initial results also led to the vacuum pulling of full mould castings with disposable tubes. Finally, a larger container where green sand, pepset or a CO2 mould could be inserted was designed and fabricated for vacuum pouring through disposable tubes. Attempts have been made to study the cooling characteristics of different sands with or without vacuum assistance using foam patterns.

EQUIPMENT AND PROCESSES

LOW PRESSURE DISPOSABLE MOULD CASTING PROCESS In the low-pressure die casting process, a permanent mould is positioned typically

above a crucible of molten metal contained in a hermetically-sealed holding furnace (7). A feed tube extends from well below the molten metal surface through the furnace cover to the bottom of the die cavity. Compressed air or inert gas is fed into the sealed container to fill the die. This forces the metal up the tube into the die. In the present process, as illustrated in Fig. 1, the die is replaced by a steel container, which is pressed on the head of the feed tube during the casting operation, by the moving platen.

In die-casting operations, this platen holds the top portion of the die. The container is spring mounted on a trolley which is rolled out for loading and unloading. The spring lifts the container off the head of the feed tube when releasing the pressure from the moving platen. The container can also be rotated for ease of unloading. This is better illustrated in Figs. 2 and 3 which show the two types of containers mounted on the trolley with the air-operated vibrator bolted on. The round container has an inside diameter of 30.5 cm with a depth of 47 cm. The rectangular one is 30.5 cm wide by 50 cm long and 38 cm deep. Figure 4 shows one container with a foam pattern ready to be filled with loose sand. The sprue is connected to a core sand donut which is inserted in the flange at the bottom of the flask. These details are shown more clearly in Fig. 5 which shows a pattern ready to be used. Vacuum was used to hold the sand together as the few inches of sand covering the top portion of the pattern would have been insufficient to overcome the pressure from underneath and prevent wall movement. A secondary aspect was the extraction of gases. The vacuum was applied through drilled holes in the inner wall of the vacuum manifold (Fig. 4). A fine mesh screen prevents the sand from entering the vacuum system. A plastic sheet was used to vacuum seal the exposed sand on the top of the container.


LADLE SYSTEM The ladle system, which was used to duplicate the low pressure operation, is shown in

Fig. 6. Basically it consists of the same container being positioned to the side of a ladle. First, with the container in an upright position, the pattern is inserted, the box is filled with sand, vibrated and the vacuum applied. The container is then wedged against the side of the ladle. The hole seen in Fig. 7 duplicates the head of the feed tube in the low-pressure operation. Casting was done by tilting the whole assembly after molten metal had been poured into the ladle. A metal shield had to be inserted in the hole to prevent hot gases from the melt, prior to pouring, from being pulled inside the mould by vacuum. These gases burn the foam leading to rapid erosion of the mould cavity. This metal shield did also permit the metal at the surface of the melt to flow past the hole before melting, thus ensuring that the metal was drawn from under the surface of the melt.

VACUUM POURING OF FULL MOLD Figure 8 shows the rectangular container with a disposable tube for vacuum pouring.

The tube was attached to the donut as shown in Fig. 9. A piece of Al-foil was wrapped at the end of the tube in order to maintain the desired level of vacuum and to allow the tube to go through the surface of the melt without pulling any metal in. The preparation of the container was the same as that described previously. The tube was lowered into the melt with the container under vacuum.

VACUUM POURING OF SAND MOULD Vacuum pouring of a regular mould was accomplished in the assembly shown in

Fig. 10. The container was 92 cm x 92 cm and 76 cm deep. Because of its size, the regular vacuum pump was unusable, and the container was attached to another large vacuum furnace by the 9 cm O.D. hose seen on the right side. The vacuum was set in the container by the large valve on the container. The vacuum in the unit was controlled by the degree of vacuum in the much larger furnace. Steel and sand snorkles were used. In the present work, the snorkle was inserted in the flange at the bottom of the container, then the mould was wedged on top of it. The cover resting on a rubber gasket was loose. It was held in place during casting by the vacuum. Vacuum was applied with the tube inside the melt

EXPERIMENTAL PROCEDURES

The main part of the study was carried out on the low-pressure unit with three patterns provided by the private sector. The castings produced for the automotive industry using these patterns are shown in Figs. 11-13; they weighed 3.2 kg, 2.7 kg and 5 kg, respectively. Patterns such as plates cut from foam sheet or foam cup were also used especially in the initial period. Later, other patterns from the automotive industry were attempted. The biggest one weighed 11.3 kg.


The parameters investigated were: melt temperature, degree of vacuum, foam density, filling time, casting media and gating configuration. The casting media include steel grit, silica, zircon, olivine and McConnellsville sand of different AFS numbers.

The patterns were coated by dipping in a refractory slurry, and then air dried Castings were also produced without coating.

Vacuum filling with a disposable mould was done either on the container fitted with a tube or on the low pressure unit With the low pressure unit, the pressurization system was shut off, and a leak opened in the sealed chamber. The ladle system was used mainly to study reduction in sprue which leads to vacuum filling of a regular moulds

Vacuum filling of regular moulds was done in CO2 or green sand mould. Throughout these experiment, four alloys were tried (Table 1). The first two alloys,

6290 and 135, were used in the low-pressure die-casting process and later, alloys 319 and 331 were suggested by the private sector.

Although some of the castings were X-rayed, the casting soundness was verified mainly by macro-examination of cut sections.

The possibility of studying the cooling characteristics of casting media was investigated with two patterns. Figure 14 shows a casting where two thermocouples were attached with wires at a distance of 1 cm from thin and thick sections. The second pattern, a step block is shown in Fig. 15. The overall length and width are 23 cm and 7.7 cm, respectively. The thermocouples were attached to the middle of every step. Section thicknesses were 6 mm, 19 mm and 38 mm.

RESULTS

SOUNDNESS The three castings shown in Figs. 11-13 were successfully cast at temperatures as low

as 655,655 and 675°C, respectively. Sound castings were obtained with the different alloys and with different casting media. These results were duplicated with the vacuum-pulled system either on the low-pressure unit or the container with a tube. Only pattern #2 was tried in the ladle system. Other types of castings which were vacuum pulled are shown in Figs. 16 and 17.

Three types of casting defects were observed. One type was large pore (1-3 mm diam) probably caused by trapped gases during burning of the foam and their associated turbulence. These pores did not have the shiny appearance of typical gas porosity, although sometimes is was associated with much smaller gas porosity. Examples are shown in Fig. 18 (sectioned casting) and Fig. 19 which came from the box shown in Fig. 17. These large pores were observed either at low melt temperatures in association with misrun or high melt temperatures. Occasionally, they would show up in otherwise sound castings. The second type of defects were surface defects at high melt temperatures. A high degree of vacuum increases the surface defects. This indicates that the problem is probably related to turbulence. The third type of defect was caused by sand penetration. It was related to the casting media, the melt temperature and the degree of vacuum.


There was also some type of edge effect, or sand penetration, at sharp edges or glued parting lines. This is shown in Fig. 20. Reducing the melt temperature or degree of vacuum had some positive effect This could be related to poor adherence of the coating at sharp edges.

CASTING MEDIA It became obvious, in the initial period, that sand penetration was a problem closely

associated with the casting media, and a series of plates was cast to evaluate the influence of the different casting media. Plates, 15 cm high, 10 cm wide and 2.5 cm thick were cast vertically from a central sprue (3.8 cm diam) at the bottom. A constant melt temperature with alloy 6290 was maintained. The results, summarized in Table 2 show that finer grain size of the casting media would be beneficial to eliminate or minimize the sand penetration. An AFS grain fineness of 25-50 is recommended for gravity pouring.

The main part of the experiment was then carried out with the silica 45 or an AFS 126 McConnellsville sand. Even with these finer sands, a combination of high melt temperature and high degree of vacuum could lead to sand penetration. Some data to corroborate this are given in Table 3. However, with the degree of vacuum generally used (510-560 mm Hg) there was no problem with an AFS grain fineness above 50.

The chilling effect of casting media such as zircon sand or steel grit compared with silica 45 was also demonstrated. For similar pattern and gating system, the minimum pouring temperature to avoid misrun was higher with the above two casting media.

FOAM DENSITY Patterns #1, 2 and 3 in Figs. 11-13 had a density of 0.026 g/cm3 (1.6 lb/ft3). Pattern

#2 was also obtained with a density of 0.032 g/cm3 (2 lb/ft3). The plates had a density of 0.029 g/cm3 (1.8 lb/ft3). Other patterns with densities of 0.048 - 0.064 g/cm3 (3 or 4 lb/ft3), such as the cup seen in Fig. 21, were also cast.

Gas problems appeared to increase with an increase in foam density during gravity pouring. However, such problems did not arise during vacuum pouring irrespective of the foam density.

CAST TIME In the low-pressure system, cast time refers to interval of time between the molten metal

reaching the sprue and the end of pressurization. In the vacuum-pulled system, it refers to the interval of time, the vacuum is applied to the molten metal. For most castings, it varied between 45-65 s. It was much higher with the plates. However, the sprue with a diameter of 3.8 cm was also acting as the riser. The cast time did not appear critical provided the main part of the sprue had solidified. If too short, it would permit the vacuum to suck air inside the casting through the partially molten sprue.


MOULD FILLING The effect of vacuum on mold filling capability was briefly investigated by pouring the

cup shown in Fig. 21 under different casting conditions. The cup was poured vertically with a single sprue (1.9 x 1.9 cm) attached to the base of the cup with alloy 331. The main wall thickness was 2.15 mm. The results are shown in Table 3 where it is evident that the presence of a vacuum facilitated mould filling in direct proportion to the degree of the vacuum. This effect has also been reported in the literature (8).

FOAM COATING Dipping appeared to be the best approach to achieve some consistency in coating

thickness or adherence. Although the manufacturer's recommendations, including viscosity measurements were followed, consistency was not completely obtained.

The coating is considered a key element in the process in gravity pouring. It must have sufficient permeability to permit the gases to escape and enough strength to prevent sand penetration of the coarse sand used. It also combines with the gas pressure and the liquid metal in supporting the loose sand while the pattern is burning. The situation is, however, different with regard to vacuum pouring where a vacuum as low as 632 mm still helped in holding the sand. In fact, vacuum assistance could change the basic requirements of a good coating. Its ability to extract gases could permit the use of less permeable coating with increase in strength and/or a reduction in thickness.

It may be noted that successful castings were also produced without applying any coating to the three patterns studied and other shapes such as the box shown in Fig. 17. This was accomplished by keeping the melt temperature to a minimum with the vacuum at 585 mm. An AFS McConnellsville and an AFS 180 olivine sand were used. Despite the difference in grain fineness, there were no advantages in using the olivine sand. This could be due to the screen distribution. The olivine sand had a four-screen distribution while the other had a three-screen distribution.

Although there was no sand penetration, light brushing or sand blasting was required to remove some sand adhering to the castings.

GATING Different gating systems were tried, mainly with patterns #1 and #2, the emphasis

being on configuration rather than optimum size. The aim in the design of the gating systems was to promote directional solidification towards the sprue.

To cast pattern #1 in a vertical position, the best approach was as shown in Fig. 9. The size of the gate was about 2.8 cm2. It was also cast from both ends in that position with one or more ingates. The minimum melt temperature required was 675°C. In another method, vertical gating was also used, except from a different location (point A in Fig. 8). The sprue or gate was 2.5 cm in diam. Double gating from points A and B was also tried. Here the minimum melt


temperature was 655°C. For comparison, this pattern was also cast with its original gating

(Fig. 22). Multiple gating was advantageous in reducing hot spots which promote sand

penetration.

Pattern #2 was cast with a single sprue (2.5 cm diam) which also acted as the gate as

shown in Fig. 5. A better design consisted of gating through the side of the heavy flange. The

casting was also gated successfully through point A in the inverted position.

Pattern #3 was cast in a vertical position using part of the original design. In other

attempts, each gate was reduced up to about one third of the original size without affecting the

quality of the casting. ■

Introducing a constriction in the sprue was useful in reducing the cast time. However,

this had an adverse effect on casting quality. It is believed that the reduction is promoting

turbulence.

YIELD

All the castings were produced with yield in excess of 85% which is much higher than

the yield calculated from the gating design attached to the patterns in the as-received condition.

However, this comparison could be misleading, because in industry, especially with automated

process, extra ribs and gates are used for support, as low density pattern also mean low strength.

FILLING TIME

Filling time was measured with a probe located at the base of the sprue and the top of

the casting. In the low-pressure and vacuum-pulled systems, it was about 8 s while in the ladle

system, 5 s was the average value. These measurements in fact lead to the development of the

vacuum-pulled systems. In the low-pressure unit, the rate of pressure built up is controlled by a

flow valve which is adjusted according to the maximum pressure desired at the end of the cast

time. In the present operation, the average maximum pressure was 0.04-0.05 MPa, and it would

take about 8 s for the molten metal to reach the container or the sprue, by which time the vacuum

had already forced the mould to fill. As a result, the low-pressure unit acts as a holding furnace or

back-up system. However, the extra pressure could have a positive effect on the soundness of the

casting.

VACUUM PULLING OF REGULAR MOULD

Figure 24 shows the casting used to study this process. As it was a symmetrical

elbow, two types of riser were used at the same time, a blind riser on the runner and a trapezoidal

riser on top of the flange for the other side. One of the advantages of the blind riser is that it

reduces the amount of sand needed in comparison with the top riser. This system is already in use

for chemically bonded sand. As a riser cannot act under vacuum, a mechanical device was used to

pinch the feed tube after filling the mould, then the vacuum was released

Based on the data obtained with the other systems, the mechanical device was replaced

by a freezing spot on the runner. This is shown in Fig. 25. The thickness of the runner in one


spot on both sides of the sprue was reduced to 0.32 cm, and in order to maintain the same cross-section to avoid increased turbulence, its width was increased. This worked well with a cast time of about 10 s. No difficulties were experienced in producing castings in the CO2 mould. However, with green sand mould, there were two major problems. First, the mould was collapsing very fast. Vacuum tests in a smaller vacuum chamber showed that it was caused by two factors. The mould was drying rapidly under vacuum, and with a leak in the system such as the one provided by the feed tube, very rapid erosion of the mould occurred. Under static conditions, collapsing of the mould was not so fast. This problem was solved by introducing the feed tube inside the melt before setting up the vacuum. The second problem with green sand mould was the pulling of the tube inside the mould under vacuum. This was due to the lower compressive strength of the green sand. A steel plate was then interposed between the mould and the tube. With the CO2 mould using the silica 45 sand, a wash was needed to avoid sand penetration.

In green sand, sound castings were produced with the AFS 126 McConnellsville sand without wash. However, the pouring temperature was lowered to 665°C, and the vacuum adjusted to 510 mm to avoid sand penetration. The casting with the gating systems weighed 6.4 kg in alloy 319. With this system, the trapezoidal riser on the flange was preferred as occasionally some gases get trapped at the top of the other flange with the side riser. The mould filling time was about 6 s.

COOLING CHARACTERISTICS Cooling curves for castings produced in lost foam showed a differences in cooling

characteristics for a silica 45 and a silica 24 sand. However, the difference due to section thickness was not consistent from one casting to the others. As this could interfere with the accuracy of the overall picture, it was decided to use the step block shape because it was much easier to locate the thermocouple at the same reference point. Figure 26 is an example of the typical cooling curves obtained with a silica 45 sand with and without vacuum assistance using aluminum alloy 319. It is evident that, for any section thickness, the sand reaches the peak temperature faster and the cooling rate of sand is also higher when cast under vacuum.

DISCUSSION

Satisfactory castings were produced with all different processes tried. For disposable pattern, the low-pressure process appears promising, because is could easily be automated as a carousel-type operation, and it does permit close control over the casting parameters. It could be used with or without pressurization. However, the beneficial effect of pressurization along with vacuum pulling would need investigation. The use of a permanent feed tube is also an advantage over direct vacuum pulling where disposable tubes are used. Use of protective atmosphere and a stirrer in the molten bath for metal-matrix composites is also easy.


The scope of the study with disposable patterns was limited to castings with relatively thin sections where risers were not needed. However, with regard to vacuum filling of regular moulds, it was observed that the introduction of freezing spots in the gating system could cut down the cast time, i.e., reduce the time of vacuum application, thus, permit the use of risers.

Besides filling the mould, vacuum assistance exhibited some very interesting beneficial effects. Even in low amounts, it helped to hold the sand together.

This effect could be a very important asset in the casting operation. In evaporative foam pattern casting, to prevent mould collapse and avoid misruns or gas entrappment, a proper balance must be maintained between the mould filling rate (gating design), the gas pressure inside the mould (permeability of coating and casting medium) and coating strength. The above requirement appears much less stringent with vacuum assistance. In fact, good castings were produced without coating, although close control of the casting parameters was necessary. Removal of the coating would result in considerable saving in cost.

The ability to extract gases would permit the use of higher density patterns, i.e., stronger patterns. Stronger patterns need less support such as extra ribs or gate, especially in automated operations, and hence, the casting yield will be increased. It could also change the basic requirement of coating such as a decrease in permeability with a resulting increase in strength. Vacuum assistance also had a positive influence on mould filability which probably accounts for the much lower melt temperature used.

Another interesting feature of vacuum assistance lies in the fact that the vacuum pump could be used to collect the combustion gases. It is highly likely that, in the near future, more critical environmental control might be applied to EPC processes as some of the combustion product are carcinogenic (9). The above observations on the effect of vacuum should also apply to gravity pouring.

The study of the cooling curves obtained with the casting and the step blocks showed that, this could be a method of studying the cooling characteristics of different casting media with or without vacuum assistance and hopefully the insulating properties of different types of coatings.

In vacuum filling of regular moulds, the elimination of a pinching mechanism by freezing spots in the gating system should prove to be an asset

CONCLUSION

Sound castings were produced with all new processes investigated. Vacuum assistance proved beneficial to all EPC processes. It made it possible to cast foam patterns without coating.

Recording of cooling curves could be used to study the cooling characteristics of different casting media with or without vacuum assistance.


REFERENCES

1. Bailey, R. "Understanding the evaporative pattern casting process (EPC)"; Modern Casting 72:56-61; April 1982.

2. Clegg, AJ. "Expanded-polystyrene moulding - A status report"; Foundry Trade Journal 159:177-180,183-184,187,196; Sept. 1985.

3. Heine, H.S. "Evaporative pattern casting developments"; Foundry M & T pp 36-41; Oct. 1986.

4. Sikora, EJ. "Evaporative casting using expandable polystyrene pattern and unbonded-sand casting techniques"; Trans AFS 86:65-68; 1978.

5. Köhler, P.G. "Lost foam process offers jobbing opportunity"; Foundry M & T pp 60-64; Oct. 1984.

6. Arzt, A.M. and Bralower, P.M. "Questions about EPC vaporize with proper practice"; Modern Casting 77:21-24; Jan. 1987.

7. Dion, J.L., Emmett, J.R. and Sahoo, M. "Low pressure die casting of zinc-aluminum foundry alloy"; Trans AFS 95:813-818; 1987.

8. Capadona, J. A. and Albright, D.L. "Review of fluidity testing as applied to lost-polystyrene investment casting"; Trans AFS 86:43-54; 1978.

9. Gressel, M.G., O'Brien, D.M. and Tenaglia, R.D. "Emission characteristics of the evaporative pattern casting process"; Trans AFS 95:503-514; 1987.


Table 1. Chemical compositions of alloys

Alloy

6290

135

319

331

Element (wt %)

Si

12

6.7-7.5

5.5-6.5

6.5-8

Cu

3-4

3-4

Fe

0.15-0.6

1

0.8

Table 2. Effect of casting media on metal penetration in plate castings.

Casting

1 media

Steel grit 120

AFS 63.5

Steel grit 80

AFS 40.6

Steel grit 50

Silica 45

AFS 52

Süica 24

AFS 28

Zircon

AFS 112

Olivine 50

1 AFS 43.4

Casting parameters

Melt

temp

(°Q

700

700

700

700

705

695-705

700

700

705

Casting

time

(s)

130

110

110

130

130

130

130

130

130

Degree of

vacuum

mmHg

255

380

380

510

635

380

510

380

380

Metal

penetration

Nil

Nil

Some - light

Heavy

Heavy

NÜ

Heavy

Nu

Moderate in

some areas |


Table 3. Effect of casting parameters on mould fiUing

Casting media

SUica45 AFS 52

McConneUsviUe AFS 126

Casting parameters Melt temp (°Q 735 734 720 736 700 681 735 734

750

L_7_37

Casting time (s) 60 55 65 55 50 50 50 50

55

50

Degree of vacuum (mm Hg)

255 585 585 280 255 255 406 458

535

[ 535

Metal penetration

Some Nu NU

Some NU NU NU NU

NU

1 NU

Comments FuU 1

l/4füled 1/3 fiUed

FuU 1 FuU

l/2fiUed FuU

Some misrun at top

Heavy misrun at top

1/3 fiUed

HYDRAULIC CYLINDER

SCHEMATIC OF LOW PRESSURE DISPOSABLE MOULD CASTING PROCESS

Fig. 1. Schematic of low-pressure disposable mould casting process, (neg 19230)


Fig. 2. Round container, (neg 8741)

Fig. 3. Rectangular container. The bulge area on bottom section is the vacuum manifold, (neg 11533)


Fig. 4. Round container with pattern ready for sand filling. (neg8745)

Fig. 5. Details of pattern, sprue and core donut assembly for low-pressure disposable mould casting, (neg 19325)


Fig. 6. Ladle system, (neg 13703)

Fig. 7. Details of hole on the side of the ladle. (neg!3705)


Fig, 8. Rectangular container for vacuum pouring of full mould, (neg 19249-6)


Fig. 9. Details of pattern assembly for vacuum pouring, (neg 19326)

Fig. 10. Container use for vacuum pouring of regular mould, (neg 19249-40)


Fig. 11. Casting from pattern #1. Overall length and width are 42 cm and 21.5 cm respectively, (neg 19232)

Fig. 12. Casting from pattern #2. Overall height and width are 21 cm and 15.5 cm, respectively, (neg 19328)


Fig. 13. Top and bottom views of casting from pattern #3. Overall length and width are 49 and 21.5 cm respectively, [(a) neg 19237; (b) 19233]


Fig. 14. Casting used to record cooling curves, (neg 18711)

Fig. 15. Step block, (neg 19214)


Fig. 17. Section of box cast without coating.

Wall thickness is 6 mm. (neg 19327)

Fig. 16. Complex shape vacuum-pulled. Overall height 16 cm. (neg 19231)

Fig. 18. Examples of pores, (neg 19311)


Fig. 19. Examples of pores, (neg 19312)

Fig. 20. Photograph of casting showing edges effect, (neg 19313)

Fig. 21. Photograph of cup used in mould-filling experiments. Height 8.3 cm. (neg 14407)


Fig. 22. Casting #1 with original gating system, (neg 19236)

Fig. 23. Casting from pattern #3 showing gating design used, (neg 19235)


Fig. 24. Elbow casting used in vacuum pulling of regular mould, (neg 19271)

Fig. 25. Sprue and runner system showing flattened area, (neg 19270)


STEP BLOCK - AFS 28 SILICA SAND

0 200 400 600 800 1000 1200 1400 1600 1800

Fig. 26. Example of temperature profiles in the step block, (neg 19346)

125

127

Production of rheocast structures by partial melting A. Damasco and M.H. Robert UNICAMP, Universidade Estadual de Campinas, Caixa Postal 1170,13100 Campinas SP, Brazil

ABSTRACT The possibility of producing Al-7 Cu and Cu-30 Zn rheocast slurries by simply heating cold deformed dendritic structures is analyzed. Heating at temperatures above the solidus leads to the transition of the structure from deformed dendritic to globular, with the solid phase immersed in liquid. The influence of deformation, treatment temperature and time in the produced rheo-cast structure is examined. The degree of globularization and globulae dimensions are depen-dent on the process parameters as follows: the higher the initial deformation, the faster the structure evolution kinetics, and the smaller the globulae diameter. Increasing time and temperature treat-ment leads to undesirable coarsening of globulae.

131

Aluminum-silicon eutectic modification—sodium or strontium?

J.E. Gruzleski Department of Mining and Metallurgical Engineering, McGill University, 3450 University Street, Montraal, Quabec, Canada, HS A 2A7

ABSTRACT

A detailed comparison of the effects of sodium and strontium on the structure and mechanical properties of AT-Si casting alloys is given. Methods of addition and the subsequent recovery which can be expected in the melt are discussed, as are the effects of each modifier on foundry properties and the generation of porosity in the final cast product. A recommendation of which modifier is better is made, based on current knowledge of the modification process.

KEYWORDS

Al-Si alloys, modification, modifiers, sodium, strontium, mechanical properties, porosity, modifier recovery.

INTRODUCTION

It was in 1921 that Aladar Pacz (Pacz, 1921) was granted a United States patent on the discovery that Al-Si alloys containing between 5% and 15% Si could be treated with alkali fluoride fluxes to yield alloys of improved ductility and machinability. The preferred flux was sodium fluoride. Treatment with this flux gave rise to the so-called sodium modification of aluminum foundry alloys. Although many of the benefits of modification were established before the Second World War, this process was not used extensively until the 1960's and 1970's. At that time, a rapid growth in the aerospace industry necessitated the production of an increasing number of complex structural castings manufactured from aluminum. In the auto-motive industry, the energy crisis of the early 1970's set off a determined effort to replace heavy steel and cast iron with lighter-weight aluminum. Many of these aluminum parts were castings, and modification was seen as a simple method to extract the best mechanical properties from these, and so to make them as light as possible. The trend continues today. New casting techniques particularly suited to aluminum, such as the lost-foam process, encourage the manufacture of complex parts by casting rather than by other methods. To the producer of these castings, modification remains a process which must be seriously considered in order to obtain the best possible mechanical properties from the alloy.

At the beginning of this increase in activity within the aluminum casting business, certain problems with the use of sodium as a modifier were recognized. Researchers sought an alternative, and strontium quickly came to the fore. The use of strontium, as strontium fluoride, is actually mentioned in Paca's original patent, and so separate patents dealing with strontium have not been issued. Strontium


was more or less ignored by the foundry industry until Alcoa undertook a research project which culminated in the now classic paper by Hess and Blackmun (1975) in which they described how strontium could be used and how it influenced the alloy microstructure. Since that time, and particularly during the 1980's, there has been strong growth in the use of strontium modifiers. Some strontium has replaced sodium, but much has been used in new applications for modification which have arisen out of the recent growth within the aluminum foundry industry.

As new applications for aluminum castings develop, the question of choice of modifier, sodium or strontium, inevitably arises. In this paper, the pros and cons of each are explored along with their similarities. At the end of the paper a reasoned recommendation for which modifier to use is given.

EFFECTS ON MICROSTRUCTURE AND PROPERTIES

Modification changes the morphology of the silicon phase in the Al-Si eutectic. In an unmodified alloy, the silicon forms as coarse, acicular plates which act as internal stress raisers. Ummodified alloys are, therefore, characterized by re-latively poor mechanical properties, specifically low ductility and impact strength. When a modifier is added to the alloy in the correct amount, the silicon solidifies with a fine, interconnected fibrous morphology. The fineness and overall roundness of this structure reduce the stress raising capacity of the silicon resulting in an alloy having significant increases in both ductility and impact resistance.

The transition from an unmodified to a modified structure is not a sharp one, and a lamellar structure which is intermediate between acicular plates and fibres is formed if too little modifier is used, or if other elements, such as antimony1 are present.

There are hundreds of research papers dealing with why modifiers exert this dramatic effect on the cast microstructure. The most successful explanation is due to Lu and Hellawell (1987) who have shown that atoms of the modifier are incorporated in-to the silicon crystal structure at the solid-liquid interface and cause a multi-tude of growth twins to form in the crystal. These twins allow the silicon to bend and to twist, forming a fibrous instead of a plate-like structure.

Of the two modifiers, sodium is the more powerful in that it produces the most uniform fibrous structure at the lowest concentration (typically 0.005% to 0.01%). Strontium modification requires somewhat higher modifier levels (0.01% to 0.02%), and strontium modified alloys often exhibit a rather non-uniform structure which consists of well modified fibrous regions interspersed with undermodified lamellar areas. This behaviour is predicted by the Lu and Hellawell model. The size of the sodium atom is more correct for producing growth twins in silicon than is the size of the strontium atom.

The important question for our purposes is whether or not this difference is of any practical significance. On the basis of our current knowledge it appears that it is not. In Table 1 are summarized typical tensile property data for some

Antimony is used in several European alloys to produce a lamellar structure. It is not considered here as an alternative to either sodium or strontium because of a strong resistance to its use in North America. This resistance is based on its toxicity and on the negative effects it has on both sodium and strontium modifica-tion.


aluminum foundry alloys having different microstructures. The quantity, Q, re-ferred to in this Table is known as the quality index. It is a commonly used method of combining the tensile strength and elongation of Al-Si-Mg foundry alloys into one parameter defined as:

Q = UTS + 150 log elongation

The improvement rendered by changing the microstructure from acicular (unmodified)

TABLE 1 Some Properties of As-Cast A l l o y s ^

Alloy Chemistry

Silicon Structure

Acicular

Lamellar

Fibrous

UTS (MPa)

180

200

200

AT -7%Si-0.3%Mg

E (%)

7

12-16 ( i i i )

16

Q (MPa)

307

362-381

381

ΑΊ-

UTS (MPa)

150

170

170

-ll%Si( Mi)

E (%)

6

14-18 ( i i i )

18

(i) cast in permanent molds. (ii) the data to determine Q for Al-ll%Si alloys is not available. (iii) according to the fineness of the lamellar eutectic.

to either lamellar or fibrous is clear. It is important to note that fine lamellar structures yield properties equivalent to those measured with the fibrous (well modified) structure. No significant difference is expected between a sodium and a strontium modified as-cast structure, even though the sodium modification may be more uniformly fibrous while the strontium one may contain some areas of fine lamellar silicon. This hypothesis is supported by the tensile test data in Table 2 for strontium and sodium treated A356 alloy (antimony treated is included for comparison). In the as-cast state, sodium and strontium yield essentially the same value of the quality index. Heat treatment tends to equalize all of the microstructures and to narrow the differences between unmodified and modified. Accordingly, some of the advantages of modification are diminished, and effects of small differences in the cast structure become even less important.

TABLE 2 Mechanical Properties of A356 Alloy Treated with Different Modifiers (Bercovici, 1979).

Modifier Structure As Cast Heat Treated* UTS E Q UTS E Q (MPa) (%) (MPa) (MPa) (%) (MPa)

None Sodium Strontium Antimony

acicular fibrous fibrous lamellar

180 195 196 201

6.8 16.4 15.9 11.9

305 377 376 362

304 292 301 293

11.8 15.1 14.4 16.5

465 469 475 476

solution treated at 540 C for 10 hours, quenched, aged 6 hours at 160°C.


We can therefore quite safely conclude that sodium and strontium are equally effective in mechanical property improvement, even though sodium is the more powerful modifier. The basis for this equality is the very large change in the microstructure brought about by almost any degree of modification. The change from acicular to fibrous, or to fibrous mixed with lamellar, is of such a magnitude that small differences in degree of modification become unimportant.

EFFECTS ON FLUIDITY

It is often claimed that modifiers decrease the fluidity of foundry alloys. A careful and critical review of the literature reveals that there is no convincing evidence to support this hypothesis. Fluidity can be measured by a variety of techniques all good to about +10% to +20%. Reported decreases in fluidity when modifiers are used all fall into this'range. We can infer that any effect of either sodium or strontium on fluidity is small, and if it does exist cannot be accurately measured at this time. Recent results obtained at McGill University on sand cast fluidity of a variety of alloys reveal no effect of either modifier. Some results for an A356 alloy in the range 700°C to 750°C are presented in Fig. 1.

90

80

E TO

σ» 60

"5 4 50 ω

40

30 700 710 720 730 740 750

Temperature (eC)

Fig. 1 Sand cast fluidity of unmodified, sodium modified and strontium modified A356 alloy.

COMPATIBILITY

Both strontium and sodium have been proven to be incompatible with antimony. This is an important emerging problem since antimony, which is used extensively in Europe and Japan, is gradually finding its way into the scrap cycle. Handiak, Gruzleski and Argo (1987) and Wang and Gruzleski (1989) have demonstrated that the

■ Net modified ▲ Sr modified • Unmodified

J I I I I I L


interaction takes place through a chemical reaction in the melt fo form solid intermetallic phases. These dense compounds which contain both antimony and the modifier sink to the bottom of the melt. Identification is difficult, but so far MgoSb2Sr and NaMgSb2_5 have been found as products of the interaction. When the modifier becomes combined as a compound it is unable to change the silicon growth form and hence is completely ineffective.

Modification of antimony containing melts requires the addition of very large quantities of modifier since it is first necessary to chemically combine all of the antimony. The exact amount needed depends on the freezing rate. Some experi-mental results are given in Figs. 2 and 3. Here, the modification rating is a quantitative assessment of the microstructure. A rating of 2 indicates a lamellar structure while 4 to 5 is modified to an acceptable level. If 0.01% Sb is present, strontium concentrations which are 4 to 6 times the normal are needed for modifi-cation. The same is true if a sodium modifier is used. Neither appears to present any particular advantage in combating the antimony problem.

8.0

σ> 6.0

o Q:

| 4.0 σ #o <^-O 2 2.0

0.0 0.00 0.02 0.04 0.06 0.08 0.10

Initial W t % Sr

Fig. 2 Modification rating of strontium modified A356 con-taining antimony. These samples were cast one hour after the strontium treatment and cooled at 0.5°C sec-' through the solidification range.

■ 0 . 0 0 % Sb

A 0 . 0 1 % Sb

• 0 . 0 4 % Sb

J i I i I i L


c σ

c o o

O 2

8.0

6.0

4.0

2.0

0.0

? 1 1

-—-— *-"

^ 1

_ ι 1 i 1 ■

■ 0.00% Sb 1 A 0 . 0 1 % Sb • 0.04% Sb

._. A ., ■ ·

1 i. , 0.00 0.02 0.04 0.06

Initial Wt% Na

0.08 0.10

Fig. 3 The modification rating of A356 melts containing various levels of antimony cast 10 minutes after sodium treatment.

Happily, both sodium and strontium are compatible with each other. Scrap contain-ing both modifiers may be mixed and remelted without difficulty. There is even some indication of an advantage to the intentional use of both together (Wang and Gruzleski, 1989). Sodium provides quick short term modification while strontium requires a few minutes in the melt before it becomes fully effective. The use of both together will give a consistent long term modification.

ADDITION AND RECOVERY

Sodium is added to foundry melts either through flux treatment or in metallic form. Metallic sodium can be purchased prepackaged in small aluminum cans. Since sodium has only a very low solubility in aluminum, the manufacture of Al-Na master alloys for melt treatment is impractical. Sodium melts at 98°C, and consequently, readily enters melts which are normally treated in the temperature range 740°C to 800°C. Dissolution of sodium is practically instantaneous at these temperatures, but it has a very high vapour pressure (0.2 atm. at 730°C), and large quantities boil off almost immediately. Sodium recoveries are therefore poor and erratic (20% to 30% of the addition) despite the excellent dissolution characteristics. A lowering of the melt temperature is- of little use as dissolution slows significantly below 700°C. In summary, sodium as a modifier is characterized by eacy dissolution above


700°C accompanied by a poor and unpredictable recovery. Furthermore, the boiling action generates large quantities of fume which pollute the workplace, and the accompanying agitation of the melt facilitates hydrogen pick-up from the atmos-phere.

Alloys of aluminum and strontium are fairly easy to fabricate, and most strontium additions are made by means of master alloys. Two distinct types can be purchased, each having its own dissolution characteristics. High strontium alloys such as 90%Sr - 10%A1 dissolve by reactive dissolution involving an exothermic reaction in the melt between aluminum, silicon and strontium (Pekguleryuz and Gruzleski, 1988). This reaction is favoured at lower rather than higher temperatures, and while some dissolution does occur in the absence of the exothermic reaction, the best re-coveries and most rapid dissolution are found when the reaction is most intense. Addition temperatures below 7009.C are recommended for A356 alloys. Under these conditions, reproducible recoveries of 90% to 95% are readily achieved.

Low strontium containing master alloys such as 90%A1 - 10%Sr or 95%A1 - 5%Sr are also available on the market. Unlike their high strontium counterparts,' these undergo classical dissolution (Pekguleryuz and Gruzleski, 1989) in which the disso-lution rate increases with melt temperature. In an A356 alloy, melt temperatures of 7750C yield repeatable recoveries of 90% or better.

The use of strontium master alloys allows high recoveries and controlled dissolu-tion. The trick in working with these alloys is to match the master alloy to the addition temperature, using low temperatures for high strontium alloys, and high temperatures for low strontium alloys. Dissolution does require somewhat longer times than if sodium is used (typically 15-20 minutes), but this is a small price to pay for the process control which is possible with these additives. Strontium addition is fumeless and can be done without excessive agitation of the bath. The working environment is preserved, and the opportunities for hydrogen dissolution in the liquid are minimized.

MODIFIER FADING

The concentration and effectiveness of both sodium and strontium decrease with time after their initial addition. This fading is due either to vaporization of the species or to selective oxidation. In the case of sodium, vapourization is the main fading mechanism. The vapour pressure of strontium is 200 times lower than that of sodium at melt treatment temperatures. While vapourization is not a problem, oxidation of strontium is, since its oxide is slightly more stable than that of either silicon or aluminum.

The exact rate of fading depends \/ery much on the melt conditions - melt size, surface area, degree of stirring etc. Nevertheless, it is well established that the fading rate of sodium is many times that of strontium. Sodium concentrations in the melt are halved in times of the order of 20-30 minutes while the half life of strontium is several hours. This rapid rate of sodium fade necessitates the pouring of castings immediately after the modifier is added. Control of modifier concentration is yery difficult, and castings poured late in a run may be less well modified than those poured soon after the sodium addition. Furthermore, as the phosphorus content of the alloy increases, fading becomes of even greater concern. High phosphorus alloys are more difficult to modify and require higher modifier concentrations. As a result, the allowable delay time after sodium treatment is shortened and the time window for casting is narrowed.

Strontium, on the other hand, has a semi-permanent effect in the melt. Its rate of fading is so low that ample time is available for other melt treatment opera-tions.


REFRACTORY WEAR

Sodium exerts a corrosive effect on crucibles and refractories. Precise data is difficult to find in the literature, but anecdotal information indicates that strontium is far less corrosive. Improvements in crucible life of 100% have been reported when sodium is replaced by strontium.

HYDROGEN PICKUP AND DEGASSING

When any solid addition is made to an aluminum alloy melt, there is a risk of hydrogen pick-up either from the additive itself, or from turbulence generated during the addition process. It has been amply demonstrated (Gruzleski and co-workers, 1986; Dimayuga and co-workers, 1988) that hydrogen is not transferred in-to the liquid from 90%Sr - 10%A1 master alloy additions. On the other hand, the addition of metallic sodium has been shown to add up to 0.1 ml Ho/100 g Al into the melt (Mulazimoglu, Handiak, Gruzleski, 1989). This occurs if a violent addi-tion reaction takes place, and is believed due to the presence of NaOH on the sur-face of the additive.

Of course, aluminum alloy melts can always be degassed to remove dissolved hydro-gen. Inert gases such as nitrogen or argon are most commonly used since these do not react with the modifying element. Unfortunately, sodium treated melts cannot readily be degassed after the sodium addition. Degassing requires 15-20 minutes, and sodium fade is important in this time period. In addition, the high vapour pressure of sodium in the melt causes sodium removal by the degassing agent. Hydrogen is often added during sodium treatment, and this hydrogen cannot sub-sequently be removed. Sodium modified castings may therefore contain significant amounts of hydrogen induced porosity.

When strontium is used, there is no tendency to add hydrogen to the melt, and standard degassing methods can be used after the addition and immediately before casting. This degassing ability is due to the low fading rate and vapour pressure of strontium. Control of hydrogen and porosity is evidently easier with strontium modification.

Finally, it is worth mentioning that neither sodium nor strontium influence the rate at which a melt absorbs hydrogen from the atmosphere above it (Mulazimoglu, Handiak, Gruzleski, 1989).

REDISTRIBUTION OF POROSITY AND SHRINKAGE

Modified castings have a reputation in the industry for being more porous than their non modified counterparts. Until recently, it was thought that modified alloys simply contained more hydrogen. Certainly, hydrogen can be added by sodium treatment, but this is not always the case, and hydrogen addition is not associa-ted with strontium treatment. There is now a growing body of evidence (Argo and Gruzleski, 1988; Closset and Gruzleski, 1989; Fang and Granger, 1989) to indicate that at equal levels of hydrogen, a modified casting does indeed contain more microporosity than the same casting unmodified. This increased microporosity comes about because, with modification, some of the shrinkage which normally occurs in the primary pipe is redistributed as dispersed micropososity. At pre-sent, the reasons for this phenomenon are not well understood. It appears to be a characteristic of the modification process, and may be due to a decrease in sur-face energy brought on by the modifier.

A quantitative demonstration of the effect is possible by use of the Tatur Test. This specialized foundry test measures the various types of shrinkage found in


foundry alloys. Some results taken from the work of Argo and Gruzleski (1988) are summarized in Table 3. Most important is the increase in the total amount of microshrinkage which is accompanied by a decrease in the pipe volume. While each modifier causes an increase in microporosity, sodium is worse in this regard. In these tests, use of sodium causes a 100% increase in the amount of microporosity expressed as a percent of total shrinkage. When strontium is used the increase is less at 33%.

TABLE 3 Tatur Test Results for Sodium and Strontium Treatment of A356 Alloy

Parameter Unmodified Sr-Modified Na-Modified

Π76.2

442.9

2.655

4.0

6.0

31.2

41.2

9.7

1180.2

446.3

2.645

5.8

5.3

28.4

39.4

14.6 rin

SUMMARY AND RECOMMENDATION

The effects of both sodium and strontium on the various aspects of the foundry process and on the metallurgy of Al-Si foundry alloys are summarized in Table 4. In their metallurgical effects they are equivalent, but it is in the control of the modification process that sodium fails. Its poor recovery coupled with rapid fading and disagreeable influence on the working environment make it a poor choice compared to strontium. Coupled with this is the demonstrated tendency of metallic sodium treatment to result in higher hydrogen concentrations.

Emphasis has been placed on technical factors in making this comparison, and so far cost has not been mentioned. On the surface, sodium appears to be consider-ably cheaper than strontium; however, its cost advantage disappears when poor recovery, poor process control and increased crucible and refractory wear are taken into account. In most applications, the true costs of using either modifier are comparable, and given its technical advantages, strontium emerges as clearly superior.

Casting Weight (g)

Casting Volume (cm )

Density (g.cm )

Microshrinkage (cm )

Pipe (cm )

Slumping and 3 Contraction (cm )

Total Shrinkage (cm )

Microshrinkage as Percent of Total

1175.2

442.9

2.660

2.9 10.6

26.6

40.0

7.3


TABLE 4 Summary and Comparison

Property Strontium Sodium

Effect on microstructure less powerful than sodium

Effect on tensile pro-perties

Effect on fluidity

Compatibility with antimony

Compatibility with each other

Ease of addition

same as sodium

slight, if at all

negative

positive

requires reasonably good temperature control for the master alloy used

Pollution characteristics quiet with no pollution

Recovery

Hydrogen addition on treatment

Fading

Ease of degassing after treatment

reproducible and up to 95%

none

êry slow

excellent

Crucible and refractory little effect wear

Effect on regassing rate none

Tendency to create micro- yes porosity

produces very fine fibrous structure

same as strontium

slight, if at all

negative

positive

easy if temperature ex-ceeds 700°C

violent boiling and con-siderable fume

erratic and poor, in range 20% to 20%

usually

rapid

not usually possible

significant

none yes, but worse than strontium

REFERENCES

Argo, D., and J.E. Gruzleski (1988). Trans. Amer. Foundrymen's Society, 96, in press.

Closset, B., and J.E. Gruzleski (1989). Proc. Int. Foundry Congress, Düsseldorf, May 1989, in press.

Bercovi, S. (1979). Revue de 1'Aluminium, February 1979, 85-99. Dimayuga, F., N. Handiak, and J.E. Gruzleski (1988). Trans. Amer. Foundrymen's

Society, 96, in press. Fang, Q.T., and D. Granger (1989). Trans. Amer. Foundrymen's Society, 97, in press. Gruzleski, J., N* Handiak, H. Campbell, and B. Closset (1986). Trans. "Amer.

Foundrymen's Society, 94, 147-154.


Handiak, N., J.E. Gruzleski, and D. Argo (1987). Trans. Amer. Foundrymen's Society, 95, 31-38.

Hess, P.D., and E.V. Blackmun (1975). Trans. Amer. Foundrymen's Society, 84, 87-90.

Lu, Shu-Zu, and A. Hellawell (1987). Met. Trans. 18A, 1721-1733. Mulazimoglu, H., N. Handiak, and J.E. Gruzleski (1989). Trans. Amer. Foundrymen's

Society, 97, in press. Pacz, A., United States Patent, 1,387,900, Aug. 16, 1921. Pekguleryuz, M., and J.E. Gruzleski (1988). Trans. Amer. Foundrymen's Society, 96,

in press. Pekguleryuz, M., and J.E. Gruzleski (1989). Can. Met. Quart., 28, 55-65. Wang, W., and J.E. Gruzleski (1989). Mat. Sei, and Technology 5, in press.

143

Effect of strontium modifiers composition on dissolution rate in Al-Si casting alloys Bernard Closset Timminco Metals, 130 Adelaide Street West, Toronto, Ontario, Canada, M5W 1G5

ABSTRACT

The dissolution rates in liquid A356.0 and 413.0 alloys of several strontium based modifiers has been studied. It has been shown that the composition of the modifier has a direct influence on the dissolution mechanism of the modifiers in Al-Si alloys. The effect of the melt temperature and composition has also been investigated. The dissolution parameters are dramatically affected by the melt temperature. A change in the silicon content of the melt can alter significantly the dissolution rate of each strontium modifier. Depending on the alloy type and the modifier composition, an optimal melt temperature range will be given for strontium addition.

KEYWORDS

Foundry alloys ; strontium alloys dissolution ; thermal analysis ; microstructures ; eutectic modification.

INTRODUCnON

The usage of aluminum cast parts is increasing rapidly in the automotive industry. Progress in the molding technology and molten metal treatment has resulted in higher quality castings.

Recent work has shown the positive effect of strontium modification on the properties of Al-Si cast metal. The addition of a small amount of strontium to a level ranging from 0.01% to 0.03% can result in a significant improvement in elongation (B. Closset and J.E. Gruzleski, 1982), tensile and impact strength (B. Closset, 1988), and cycles to failure (G.A. Hoskin, J.W. Provan and J.E. Gruzleski, 1988).

The preferred method to add strontium to an Al-Si melt is by using low strontium containing master alloys (Al-3.5%Sr; Al-10%Sr) or high strontium containing alloys (Al-90%Sr). In the present work the dissolution in liquid A356.0 and 413.0 alloys of two binary master alloys (Al-3.5%Sr ; Al-10%Sr) and a binary alloy (Al-90%Sr) are examined. Dissolution tests were conducted at three different melt temperatures (650°C, 700°C, and 750°C).


The dissolution experiment of the three modifier types were conducted on A356.0 and 413.0 alloys having the composition given in Table 1.

Table 1. Chemical Composition of A356.0 and 413.0 alloys.

Alloy Elements (wt. %) _ Sr Fe Cu Mn Mg Zn Ti Sr Al

A356.0 7.5 0.15 0.007 0.006 0.35 0.015 0.1 -- Bal. 413.0 11.5 0.7 0.30 0.2 0.04 0.2 0.03 - Bal.


The composition of the three modifiers is shown in Table 2.

Table 2. Composition of Strontium Modifiers. Modifier Type Elements (wt. %)

Sr Master Alloy A 3.2 to 3.8 Master Alloy B 9.0 to 11.0 Alloy C 88 to 92

Si Fe Ba Ca 0.1 0.2 —

0.15 0.30 0.20

0.05 0.1 0.7

0.01 0.03 0.2

0.005 -0.01 -0.0005 0.2

Mg N

0.5

Cylindrical samples of each strontium modifier were cast in steel molds maintained between 300°C and 400°C. Because of the different strontium levels in each modifier type it was decided to cast three samples size in order to add to the melt an amount of strontium between 0.02% and 0.03%. A thermocouple protection sheet was cast in the center and at mid-height of each sample.

For each dissolution experiment approximately 10 kg. melts were prepared in SiC crucibles using a gas-fired furnace. Before strontium addition, the melt was degassed with a mixture N2-5% Freon for 30 minutes using a graphite tube with a perforated head. The strontium sample was placed in a sample holder and immersed vertically 10 cm below the molten metal surface. The experimental set-up was similar to the one developed by Pekguleryuz, et al. (1984) and is shown in Figure 1. A K type thermocouple protected by an alumina sheet, was located in the melt. All the experiments were conducted isothermally at 650 + 5C, 700 + 5C and 750 + 5C. After a 20 minute immersion period under static condition, the sample was withdrawn from the melt.

Ba t h T h t r m o c o u p l ·

Fig. 1. Schematic of experimental set-up and data acquisition system.

A data acquisition system (S. Argyropoulos, R.S. Valivetti, B. Closset, 1983) was adapted to measure both the bath temperature (No. 1) and the sample temperature (No. 2) During the samples immersion spectrochemical samples were taken at regular time intervals for strontium analysis. After removing the sample holder from the melt, a thermal analysis was conducted on the remaining melt.

P


The hydrogen level in the melt was measured by Telegas after melting, and before and after strontium addition. RESULTS

Immersion experiments were conducted with the three modifiers at different bath temperatures in liquid A356.0 and 413.0 alloys.

Dissolution in A356.0 Alloys

Master Alloy A : Al-3.5%Sr Figure 2 represents the strontium uptake by melts held at 700°C and 750°C. At both melt

temperatures the immersed sample dissolves rapidly and reaches a strontium level in the order of 0.02% which is close to 0.024% which corresponds to a 100% strontium recovery. The dissolution is completed after a 3 to 5 minute immersion period and the strontium level remains constant until the end of the experiment.

Fig. 2. Dissolution of Al-3.5%Sr Master Alloy. Strontium uptake in liquid A356.0 alloy at 700°C and 750°C.

Master Alloy B : Al-10%Sr

Strontium uptake by the melt depends on temperature. Figure 3. At 700°C, there is no noticeable dissolution during the first 5 minute immersion period. This is followed by a slow dissolution during which time the strontium level increases only from 0 to 0.012%. However, considerable dissolution occurs at 750°C. After a 10 minute immersion time the strontium level increases from 0% to a maximum level of 0.018%.


STRONTIUM (%) 0.03

0.02 h

0.01 k

0.0 δ 10 15

TIME (Minutes) 20

Fig. 3. Dissolution of Al-10% Master Alloy. Strontium uptake in liquid A356.0 alloy at 700°C and 750°C.

Allov C: Al-90%Sr

At 700°C, a complete dissolution of the binary Al-90%Sr sample is attained in approximately 5 minutes and the strontium content increases from 0% to 0.022% which corresponds to a 100% strontium recovery. Figure 4. A complete different result is obtained at 750°C. There is very little strontium uptake.

STRONTIUM (%) 0.03

0.02

0.01

O^-""""""" O S^ °

' n -0.0 3 ' w '—" ■ ■ ■

0 5 1 0 TIME (Minutes)

Ü 750 °C 1

O 700 °C 1

0

-, π i 16 20

Fig. 4. Dissolution of Al-90%Sr. Strontium uptake in liquid A356.0 alloy at 700°C and 750°C.


Dissolution in 413.0 Alloys

Master Allov A : Al-3.5%Sr

At 700°C and 750°C the strontium uptake reaches a maximum after a 3 to 4 minute immersion period. At both temperatures the strontium level is in the order of 0.02% at the end of the immersion period. Figure 5.

Fig. 5. Dissolution of Al-3.5%Sr Master Alloy. Strontium uptake in liquid 413.0 alloy at 700°C and 750°C.

Master Allov B : Al-10%Sr

The melt temperature is again an important factor in the dissolution of an Al-10%Sr master alloy. There is no strontium uptake after a 15 minute immersion period in a 413.0 melt at 700°C. At the end of the addition period there is only 0.006% strontium present in the melt. Figure 6. An increase in the melt temperature to 750°C results in an increase in the dissolution rate. After a 5 minute immersion period the maximum strontium level of 0.022% is attained.


STRONTIUM (%) 0.03

0.02 h

0.01

0.0 5 10 15

TIME (Minutes)

Fig. 6. Dissolution of Al-10%Sr Master Alloy. Strontium uptake in liquid 413.0 alloy at 700°C and 750°C.

Alloy C: A1-90%A1

Only a strontium uptake is obtained at 650°C. Figure 7. A maximum strontium level of 0.018% is reached 10 minutes after the sample immersion. A melt temperature increase to 700°C or 750°C does not result in a significant strontium uptake.

STRONTIUM (%) 0.03

0.02

0.01

Ü 700 C

O 6 5 0 °C

M t Q - r 5 10 15

TIME (Minutes) 20

Fig. 7. Dissolution of Al-90%Sr Alloy. Strontium uptake in liquid 413.0 alloy at 650°C and 700°C.


Thermal Analysis

In previous work it has been shown that thermal analysis is a good technique to evaluate the effect a strontium modifier on the microstructure (S. Argyropoulos, B. Closset, J.E. Gruzleski and H. Oger, 1983). Simple thermal analysis curves were obtained before and after each modifier addition. The strontium addition changes only the portion of the curves corresponding to the eutectic transformation located in the temperature range 580-550°C. In order to compare these curves we have considered certain characteristic temperatures; namely, the temperature for eutectic nucleation (Tc) and the eutectic growth temperature (Tg). The quantity Δθ= ( Τ Ε - Tc) and the difference

ΔΤ between the eutectic growth temperature before and after strontium addition are given in Table 3 along with the values of Tg and T Q

Table 3. Eutectic Transformation Characteristics of Simple Thermal Analysis Curves.

Alloy Melt Temperature

(°C)

A356.0 700

750

700

750

700

750

413.0 700

750

700

750

650

700

Modifier Type

A A A A B B B B C C C C

A A A A B B B B C C C C

Strontium (%)

0 0.021 0 0.019 0 0.012 0 0.017 0 0.021 0 0

0 0.021 0 0.016 0 0.006 0 0.022 0 0.016 0 0

Eutectic Nucleation

Temperature TC(°C)

576.8 567.4 576.0 566.8 576.5 569.4 N/A 565.9 576.8 569.1 575.7 N/A

578.8 574.5 577.1 574.0 579.1 574.8 578.0 574.0 577.4 573.4 578.2 578.5

Eutectic Growth

TE(°C)

577.1 569.7 576.0 569.4 577.4 571.1 N/A 568.2 577.4 571.7 576.0 N/A

579.4 576.0 577.7 575.1 579.7 576.8 578.5 574.8 578.2 575.4 578.8 578.8

Δ θ = TE-TC

(°C)

0.3 2.3 0. 2.6 0.9 1.7 N/A 2.3 0.6 2.6 0.3 N/A

0.6 1.5 0.6 1.1 0.6 2.0 0.5 0.8 0.8 2.0 0.6 0.3

ΔΤ

(°C)

7.4 -6.6 -6.3 -N/A -5.7 -N/A

3.4 -2.6 -2.9 -3.7 -2.8 -0

When the strontium level is increased by the addition of different modifier types to A356.0 alloys, the eutectic temperature decreases in the order of 6 to 7°C and the value of Δ θ = Tc - TJJ increase from a range of 0°C to 0.9°C to levels comprised between 1.7°C and 2.6°C.


In the case of 413.0 alloys the increase of the strontium level results only in a decrease of eutectic temperature in the order of 3°C while the values of Δθ= T c - Tg increase from a range of 0.5°C to 0.8°C to levels comprised between 1.1°C and 2.0°C.

Hydrogen Content

Aluminum alloys are susceptible to pick-up hydrogen when held in the molten state. Melts can be readily degassed by using inert gases (N2 or Ar) or gas mixtures (N2~Freon or N2-CI2). In the present work the hydrogen level was measured by a Telegas instrument before degassing, after degassing and after the strontium modifiers addition. Table 4 shows the variation of the hydrogen content upon strontium modifiers addition.

Table 4. Variation of Hydrogen Level In Al-Si Alloys With Strontium Modifiers Addition.

Alloy

A356.0

413.0

Melt Temp. (°C)

700 750 700 750 700 750

700 750 700 750 650 700

Modifier Type

A A B B C C

A A B B C C

: Strontium (%)

Before After

0 0 0 0 0 0

0 0 0 0 0 0

0.021 0.019 0.012 0.017 0.021 0

0.021 0.016 0.006 0.022 0.016 0

Hydrogen Before

Level (mlH2/100gAl) After

Degassing Degassing

0.28 0.29 0.25 0.32 0.26 0.30

N/A 0.21 N/A 0.23 0.16 N/A

0.17 0.21 0.18 0.20 0.16 0.21

0.17 0.15 0.14 0.16 0.05 0.11

After Sr

Addition

0.20 0.26 0.20 0.22 0.21 0.22

0.17 0.20 0.14 0.19 0.08 0.13

Hydrogen Increase upon Sr Addition

0.03 0.03 0.02 0.02 0.05 0.01

0 0.05 0 0.03 0.03 0.02

At 700°C and 750°C, the hydrogen content of non degassed A356.0 and 413.0 melts lies generally between 0.20 and 0.30 ml H2/100g Al with the exception of a 413.0 melt at 650°C which has a hydrogen level of only 0.16 ml H2/100g. Al. After degassing with a mixture of N2~Freon the hydrogen content is reduced to levels comprised between 0.10 and 0.20 ml H2/100g. Al. A hydrogen level of only 0.05 ml H2/100g Al is obtained for a 413.0 alloy held at 650°C.

The addition of strontium modifiers increase only slightly the hydrogen of the melt by values ranging from 0 to 0.05 ml H2/100g Al. The type of modifier does not seem to have a particular influence on the hydrogen increase.

DISCUSSION

Master Alloy A

The dissolution of master alloy A in both A356.0 and 413.0 alloys is achieved 5 minutes after the immersion in melts held at 700°C and 750°C. Figures 2 and 5. The type of alloy and the melt temperature does not have a significant effect on the dissolution rate of master alloy A. Master alloy A presents a liquidus temperature Lj = 670°C (B. Closset, H. Dugas, M. Peguleryuz and J.E. Gruzleski, 1986) which is well below the melt holding temperatures of 700°C and 750°C. The shell


formed around the sample during immersion and the sample itself start to melt immediately after the immersion. Therefore the mass transfer between the sample and the molten bath is governed by melting which results in a fast dissolution rate.

Thermal analysis confirms the good degree of modification obtained after the dissolution is completed in A356.0 and 413.0 alloys. The increase of the strontium level in both melts from 0% to approximately 0.02% results in typical eutectic temperature decrease in the order of 7.0°C and 3.0°C corresponding respectively to A356.0 and 413.0 melts. The addition of master alloy A to both A356.0 and 413.0 alloys results in a small hydrogen increase in the melts. The increase ranging from 0 to 0.05 ml H2/100g. Al corresponds to the natural regassing of the melt which occurs even in the absence of a strontium modifier addition (F. Dimayuga, N. Handiak and J.E. Gruzleski, 1988).

Master Alloy B

The dissolution rate of master alloy B increases with the increase of the melt temperature (Figures 3 and 6). At 700°C, there is a significant strontium uptake only after an immersion period ranging from 5 minutes for A356.0 alloys to 15 minutes for 413.0 alloys. Only when the melt temperature is increased to 750°C, a complete dissolution occurs after a 20 minute immersion period. Nevertheless, the dissolution rate of master alloy B at 750°C in both A356.0 and 413.0 alloys is significantly lower than the dissolution rate of master alloy A under similar conditions. Dissolution of master alloy A is achieved in less than 5 minutes while the addition of master alloy B needs an immersion time ranging from 5 to 10 minutes before the maximum strontium level of approximately 0.02% is obtained.

Master alloy B shows a high liquidus temperature L2 = 790°C which is well above the melt temperature of 700°C and 750°C. The heat exchange between the immersed sample and the liquid metal is certainly retarded by the formation of a solid shell around the sample immediately after immersion. The dissolution of master alloy B can start only after the shell has molten back. At 750°C, 40°C below the liquidus temperature, the mass transfer between the sample and the liquid is accelerated, and results in a strontium level increase from 0% to 0.02%.

Thermal analysis results are similar to those obtained with master alloy A. Even at 700°C with strontium levels of 0.012% in A356.0 alloys and 0.006% in 413.0 alloys a good modification is obtained.

Hydrogen levels are only slightly increased upon master alloy B addition. The increase ranging from 0 to 0.03 ml H2/100g Al is close to values obtained for non-modified melts.

Allov C

The strontium modifier, alloy C is of an eutectic composition and presents a transformation E3 = Sr + AlSr, at 575°C. The presence of elemental strontium results in an exothermic reaction at the sample solid-liquid melt interface (B. Closset and S. Kitaoka, 1987). At 700°C a complete dissolution is achieved in less than 5 minutes after the immersion of the sample in A356.0 alloys. Figures 4 and 7. At the same temperature no dissolution occurs in 413.0 alloys. It is necessary to decrease further the melt temperature to 650°C to obtain rapid dissolution in 413.0 alloys. By increasing the melt temperature to 750°C for A356.0 alloys or 700°C for 413.0 alloys, the exothermic reaction does not occur and this results in insignificant strontium dissolution. The optimal melt temperature range in which exothermic dissolution occurs depends also on the silicon content of the molten metal. Lower melt temperatures are needed to dissolve readily the modifier alloy C in 413.0 alloys. In the case of both A356.0 and 413.0 alloys there is definitely a change in the dissolution mechanism which depends on the melt temperature but which is also influenced by the silicon content of the melt.


Thermal analysis data shows good modification when the strontium is dissolved in the melt. The decrease of the eutectic temperature are respectively 5.7°C and 2.8°C for A356.0 and 413.0 alloys. These values of ΔΤ are similar to the results obtained after the addition of master alloys A and B.

The addition of strontium modifier C does not add significant amounts of hydrogen to the melt. Measurements show that the hydrogen level in the melt increases by values ranging from 0.01 to 0.05ml H2/100g. Al. Again these increases are similar to results obtained for unmodified melts.

CONCLUSIONS

1. The addition to a liquid A356.0 or 413.0 alloy of both low strontium containing Al-3.5%Sr and Al-10%Sr master alloys is characterized by an increase of the dissolution rate with an increase of the bath temperature from 700°C to 750°C.

2. The high strontium containing binary Al-90%Sr alloy dissolution is characterized by an exothermic reaction during addition to liquid A356.0 and 413.0 alloys maintained respectively at 700°C and 650°C. The increase of the silicon content of the melt from 7.5% to 11.5% necessitates a decrease in the melt temperature by approximately 50°C to obtain good dissolution rates.

3. Thermal analysis is a good technique to characterize the degree of modification of the melt after the dissolution period. The decrease in the eutectic temperature of A356.0 alloys is comprised between 6°C and 7°C while the eutectic temperature of 413.0 alloys is only decreased in the order of 3°C.

4. The addition of the three strontium modifiers does not contribute significantly to the hydrogen increase in the melt. The hydrogen level increase of unmodified melts is comparable to the results obtained after the three modifiers addition.

REFERENCES

- B. Closset and J.E. Gruzleski (1982). Met. Transactions. 13A, 945-951.

- B. Closset (1988). In C. Bickert (ed.). Reduction and Casting of Aluminum. CIM 8, 243-254.

- G.A. Hoskin, J.W. Provan and J.E. Gruzleski (1988). Theoretical and Applied Fracture Mechanics. 10, 27-41.

- M. Pekguleryuz, B. Closset, J.E. Gruzleski (1984). AFS Transactions. 92,109-118.

- S. Argyropoulos, R.S. Valivetti, B. Closset (1983). Journal of Metals. 35, 30-35.

- S. Argyropoulos, B. Closset, J.E. Gruzleski, H. Oger (1983). AFS Transactions. 91, 351-358.

- B. Closset, H. Dugas, M. Pekguleryuz and J.E. Gruzleski (1986). Met Transactions. 17A, 1250-1253.

- F. Dimayuga, N. Handiak and J.E. Gruzleski (1988). AFS Transactions. 95, 83-88.

- B. Closset and S. Kitaoka (1987). AFS Transactions. 95, 233-240.

153

Grain refining response surfaces for three commercial aluminum alloys W.C. Setzer, G. W. Boone and B. H. Wilson KB Alloys Corporate Technology, Henderson County Plant, McDonald Road, Robards, Kentucky 42452, U.S.A.

ABSTRACT INTRODUCTION

This paper presents grain refining data using the recently standardized Aluminum Associa-tion test to show the grain refining response of three alloys; commercial purity alumi-num (P1020A), and alloys AA6063 and AA3004. The grain refining master alloys evaluated were of various ratios (titanium to boron).

Grain refiner, response surface, prediction of grain refiner addition, grain refining mechan-ism, P1020A aluminum, AA3004, AA6063, titanium to boron ratio, aluminum master alloy.

Boone, et al. (1989) utilized the recently instituted Aluminum Association (AA) grain size test (TP-1, 1989) and the Golf Tee test to examine the effect of various process parameters on grain size (Kirby, et al. 1986).

In this paper the AA test pro-cedure was exclusively used to describe the grain refining effectiveness of a TITALR (Al-Ti) and various TIBORR (Al-Ti-B) grain refiners in 99.7% pure Aluminum (P1020A), AA3004 and AA6063. Response surfaces were developed in order to show similarities and differ-ences between these three alloy systems. Also, the number of grains generated per cubic millimeter was calculated in order to assist in developing a

Grain refining response surfaces were developed for each of the three alloys. There appears to be a linear relationship between grain refining addition and number of grains per cubic mm generated. In addition, each of the three alloys developed a unique response surface with different curvature, position-ing, and slope.

In all three alloys, the change in grain refining response with respect to a change in master alloy ratio was gradual and continuous with the exception of the boron free titanium addition wherein substantially more titanium was required than was predicted by the model.

KEY WORDS

This is the third in a series of papers utilizing the concept of response surfaces as a means for predicting the effectiveness of various ratio (titanium to boron) master alloy grain re-finers. Setzer, et al. (1989) presented the concept whereby the grain refining capability of various ratio grain refiners could be reviewed on a single response surface plot similar to a topographical map. The data was derived from an earlier paper by Mollard, et al. (1987) and utilized KB Alloys' (KBA's) calibrated ring test to evaluate grain refiner performance.

154 PRODUCTION AND ELECTROLYSIS OF LIGHT METALS further understanding of the mechanism of grain refining.


Table 1 gives the chemical composition of the materials used in this work.

data for various combinations of titanium and boron were the basis for the quadratic response surface models developed utilizing the software program Design-ExpertR.


The AA test procedure was used to produce cast samples for all the grain refining results re-ported. Testing variables such as ladle temperature, water temperature, flow rate, heat weights, etc., were closely controlled. In addition, each test run was qualified by testing a blank and control sample.

Specimens were cast after an in-cubation time of 5 minutes. The as-cast specimens were section-ed, mechanically polished and anodized in a 2.5% fluoroboric acid solution. Grain size was determined on the anodized sur-face under polarized light at 100X by measuring the average intercept distance (AID) in microns using the linear inter-cept procedure described in ASTM E-112. The master alloy grain refiners were tested over a range of addition levels in each alloy system. The resultant (AID) data for each of the TITAL and TIBOR grain refiners were transformed to grains per cubic millimeter (grains/mm3) and fitted by least squares regres-sion. The fitted grains/mm3

By using titanium and boron additions as the abscissa and ordinate, one can depict all of the TIBOR and TITAL alloys on one graph thus allowing commer-cial grain refining alloys to be easily compared in terms of uniqueness or unusual features.

Commercial master alloy grain refiners for wrought products have a titanium to boron ratio greater than 2.2:1, the stoi-chiometric ratio for ΊίΒζ. Stoichiometric ratio alloys are not effective grain refiners for wrought aluminum alloys (Lu, et_ a_i, 1981; Sigworth and Guzowski, 1985). Also, sensitivity to processing conditions make them susceptible to unacceptably coarse T1B2 particles. Conse-quently, the titanium to boron ratios in commercial TIBOR alloys typically range from 3:1 to 25:1, 25:1 being represented commercially by both the alloys 5ΤΪ/0.2Β and 10Ti/0.4B. It is generally noted that the grain refining characteristics of higher concentration alloys such as 10Ti/0.4B are less effective. These alloys contain coarser (Campbell and Sutker, 1986) and

Alloy

TABLE 1. Chemical Composition

Si Fe Cu Mn MS. Ti

P1020A AA3004 AA6063

3/1 TIBOR 5/1 TIBOR

5/0.6 TIBOR 6.0.4 TIBOR 3/0.2 TIBOR 5/0.2 TIBOR

6% TITAL

0.06 0.22 0.47 0.08 0.09 0.09 0.10 0.08 0.10 0.09

0. 19 0.50 0.25 0.10 0.12 0.14 0.12 0. 11 0.10 0. 11

__ 0. 15 0.001 --------------

„

1.12 0.002 --------------

„

1.10 0.61 ------__ ------

0.005 0.005 0.005 3.1 5.2 5. 1 6. 1 3.0 5.2 6.3

----

0.98 1.12 0.67 0.45 0.22 0. 24 0.00

B

155

800

120

150

2 0 0

(D 0) a L <D >

CE

0.000 0.005 0.010 TI t a n i urn Rdd i t i on

0.015 CZ1

Fig. 1. Relationship between grain refiner titanium addition and number of grains (grains/mm3) for P1020A commercial purity aluminum.

PRODUCTION AND ELECTROLYSIS OF LIGHT METALS Figure 1 shows the number of grains per cubic millimeter generated in P1020A aluminum for various addition levels of a 6% TITAL alloy and several different TIBOR alloys. There is a linear relationship between the number of grains per cubic millimeter or, the number of active nuclei, and the grain refiner addition level over the range investigated. Each incre-mental grain refining addition is as effective as the previous one in adding additional active nuclei. TITAL was tested to a 0.05% titanium level, almost twice the normal addtion rate, and still exhibited this linear relationship. Similar results were obtained for AA3004 and AA6063 although there was slightly more scatter in the data.

These results suggest that it is possible to achieve a fine grain

consequently, fewer TiAl3 particles or effective nuclei.

At KB Alloys, a substantial effort has been made to improve the effectiveness and consis-tency of grain refiners by adaptation of new process con-cepts and implementation of sta-tistical process and chemistry control. As a result, the response surfaces developed in this study are substantially different from those which were first presented based on data developed in 1986 (Mollard, et al., 1987). Also, Kiusalaas and Backerud (1987) and Snyder (1989) have shown that produc-tion parameters utilized to manufacture Al-Ti-B master alloys markedly influence grain refining performance so that caution should be taken in applying the response surfaces described below to other manu-facturers products.


800

0.001 0.002 0.003 Boron A d d i t i o n C2!)

0.004

-MOO +»

<D D )

a L (D >

CE

0.005

Fig. 2. Relationship between grain refiner boron addition and the number of grains (grains/mm3) for P1020 commercial purity aluminum.

size by adding a sufficient quantity of a TITAL grain refiner. In practice however, only the use of a TIBOR product will allow the attainment of a very small grain size.

50% as might be expected.

Since the number of grains generated is proportional to the inverse cube of the Average Intercept Distance (AID), a doubling of the amount of grain refiner added will produce twice as many grains but only reduce the AID grain size by 20%, not

The data for a straight TITAL addition was not used in the model as the extrapolated curves for ratios higher than 25:1 suggest a strong dependence for grain size on titanium content. This strong dependence is not observed for boron free alloys. For example, on the basis of the P1020A model a 170 micron AID grain size would be predicted for a 100 ppm titanium addition

On the basis of titanium con-tent, the smallest grain sizes are attained by lower ratio TIBOR grain refiner alloys. Conversely, on the basis of allowable boron level, higher ratio grain refiners are best (Fig. 2). Also, if the current addition rate and grain size are known grain size can be pre-dicted for alternate ratio grain refiner or addition levels.

The fitted data developed from the TIBOR grain refiner alloys were modeled to develop the relationship between titanium addition, boron addition and grain size (grains/mm3). Figure 3 is the quadratic grain refining response surface of titanium versus boron on which ratio lines are drawn and whereupon grain size is shown as AID (micron) contours for P1020A aluminum.


10

E a a

[

0 /

;

»· /

180

1

<b'

170

400 r-i

' β ο

1

^ Ν . ;

^ S ö ,-

Rluminum

L_

"°

— &

flssoc.

325 — □

\<fc ,··'""

Test Method

275 1 1

30 60 90 TI t a n I urn RddI 11 on C ppm}

120 150

Fig. 3. Grain refining response surface for P1020A aluminum.

of 6% TITAL. In fact, a grain size of over 300 micron was measured. 0rf stated another way, less than 20% of the number of grains predicted by the TIBOR model were nucleated by the straight titanium addition. It is on the basis of these results that we examined a KBA grain refiner alloy with a 65:1 ratio, 6.5/0.1 TIBOR. Limited data for this alloy supports the TIBOR model given in Fig. 3.

From Fig. 3, cast house person-nel can predict which grain refiner and addition level should be used to provide the required grain size in P1020A aluminum or in similar commer-cial purity alloys. Cast house personnel are now in a position to choose the most metallur-gically desirable or least expensive refiner alloy based on existing chemistry constraints or pricing conditions.

Figures 4 and 5 show the grain sizes achieved in AA3004 and AA6063 as a function of grain

refiner addition. The left ordinate scale, grains/mm3, is linear whereas the right ordinate scale, AID grain size, expands rapidly. Consequently, finer grain sizes accentuate any error in the grain size values measured and depicted in a plot of this type. Whereas a 5 or 10 micron measurement error will have little influence on the position of the individual curves for large grain sizes (smaller additions), it will influence the shape and position of the response surface some distance away from the origin since with larger additions the expanding AID scale exaggerates even a small deviation. With this in mind an examination of the raw data for all three alloys suggests that force ranking by ratio is justified with 3Ti/lB being the superior alloy {more nuclei per unit titanium addition) for the master alloys tested. On the basis of effective boron addi-tion, force ranking by ratio also appears to be justified


800

120

150

a CD Ü

(D

9 L (D > (E

0.000 0.005 0.010 0.015 T i t a n i u m P d d i t i o n QZ1

0.020

Fig. 4. Relationship between grain refiner titanium addition and number of grains (grains/mm3) for alloy AA3004.

800

700

Θ00

500

CD

a CD

c D L

CD

400

300

200

100

0.000 0.005 0.010 0.015 T i t a n i u m RddI11 an CZ1

TOO v

120

150

2 0 0

0.020

a CD Ü

Fig. 5. Relationship between grain refiner titanium addition and number of grains (grains/mm3) for alloy AA6063.

PRODUCTION AND ELECTROLYSIS OF LIGHT METALS 159 with 5Ti/0.2B being the superior alloy. The exception to the force ranking may be 3/1 in that an extensive effort has been made to improve its effective-ness, which had not been applied to 5/1 at the time this data was developed.

A comparison of Figures 6 and 7 with Figure 3 for P1020A aluminum shows that the grain refining response in going from one TIBOR ratio alloy to another is continuous and gradual. However, curvature, position and slope vary for each of the alloys.

Both P1020A and AA6063 show a gradual change in grain size with addition rate and ratio, with grain size being dependent on both titanium and boron content. However, AA6063 is much more dependent on titanium content as shown by the steeper

slope. Also, as is well known, a smaller grain size is achieved for the same addition level than in pure aluminum. AA3004 is interesting in that the high ratio TIBOR alloys are more titanium dependent while for the low ratio alloys 3Ti/lB and 5Ti/lB, grain size is dependent on boron level. Table 2 shows the predicted relative volumes of grain refiner required in order to achieve a 150ju grain size in AA3004. Because of the boron equivalency for 3:1 and 5:1, the volume of grain refiner required is about the same. As noted in Table 2 there is a substantial cost reduction in going from a TITAL to a TIBOR alloy with the cost advantage being further enhanced (in reduced volume of aluminum and titanium) in going to a low ratio alloy in either P1020A, AA3004 or AA6063.

150

Fig. 6. Grain refining response surface for alloy AA3004


60 90 TI tan i um RddI11 on C ppm )

150

Grain refining response surface for alloy AA6063.

TABLE 2 Grain Refiner Required for 150 u AID Grain Size

Alloy P1020A AA3004 AA6063

Addition # Reg. Rel. /1000JL Cost

# Reg. Rel. /1000# Cost

# Reg. Rel. /1000# Cost

3/1 5/1

5/0.6 6/0.4 3/0.2 5/0.2

6%

TIBOR TIBOR TIBOR TIBOR TIBOR TIBOR TITAL

0 0 1 1 2 2

.87

.77

.10

.40

.80

.46 >5

1.06 1.00 1.39 1.79 3.08 2.77 >5

1 . 3 0 1 . 0 0 1 . 2 1 1 . 4 3 2 . 6 1 2 . 2 0

>5

CONCLUSIONS

There appears to be a linear relationship between the number of as-cast grains per cubic millimeter (number of active nuclei) and the grain refiner addition level in P1020A aluminum and AA3004 and AA6063.

2. A desired grain size can be predicted if an addition rate and grain size are known when other variables are held constant.

3. On the basis of the least total titanium added or least cost, the smallest grain sizes are achieved with lower ratio TIBOR grain

0 . 7 0 0 . 7 4 1 . 2 6 1 . 4 3 2 . 8 6 1 . 9 2

>5

1 . 0 0 1 . 12 1 . 8 6 2 . 14 2 . 5 5 2 . 6 9

>5

0 . 9 3 0 . 6 8 0 . 8 4 0 . 9 8 1 . 9 7 1 . 6 2

>5

F i g . 7 .

1 .


The authors would like to thank Mrs. T. Shofner for her prep-aration of the manuscript, Mike Patton for his accurate labora-tory work and the management of KB Alloys for giving permission to publish this work.

REFERENCES

Aluminum Association: Standard Test Procedure for Aluminum Alloy Grain Refiners, TP-1, 1987.

Boone, G. W., Carver, R. F. and Setzer, W. C , Factors Affecting As-Cast Grain Size in AA3004 Presented at the Aluminum Association "Ingot and Continuous Casting Process Technology Seminar for Flat Rolled Products", New Orleans, LA, May 10-12, 1989.

Campbell, G. T. and Sutker, S.

Mollard, F. R., Lidman, W. G. and Bailey, J. C , Systematic Selection of the Optimum Grain Refiner in the Aluminum Cast Shop, Light Metals 1987, 116th Annual Meeting AIME Proc., February 24-26, 1987, pp. 749-755.

Setzer, W. C , Boone, G. W. , Carver, R. F. and Wilson, B. H., Grain Refining Response Surfaces in Aluminum Alloys, Light Metals 1989, 118th Annual Meeting TMS Proc, February 27-March 3, 1989, pp. 745-748.

Sigworth, G. K. and Guzowski, M. M., Grain Refining of Hypoeutectic Al-Si Alloys, Transactions of the American Foundrymens Society, Vol. 93, 1985, pp. 907-912.

Snyder, W.J., Private Communication.

refiner alloys. Conversely, if total boron content must be controlled, high ratio alloys are best.

4. Extremely high ratio alloys are markedly improved over boron free titanium grain refiners.

5. The grain refining response surfaces for P1020Af AA3004 and AA6063 are similar but differ in that iso-contour lines are displaced and have different curvature and slopes. This minimizes the possibility of a single grain refiner alloy being ideal for all aluminum alloy systems and operating con-ditions.

A., The effect of Titanium and Boron Concentration on Aluminum Grain Refiners, Light Metals, 1986, 115th Annual Meeting TMS Proc, March 2-6, 1986, pp. 741-749.

Kirby, J. L., McCarthy R. W., and Levy, S. A.: Grain Size Test Methods Comparisons, Light Metals 1986, AIME, pp. 749-757.

Kiusalaas, R. and Bäckerud, L., Influence of Production Parameters on Performance of Al-Ti-B Master Alloys, Solidification Processing 1987, The Institute of Metals, London, 1988, pp. 137-140.

Lu, H. T., Wang, L. C. and Kung, S. K., Grain Refining in A356 Alloys, Journal of Chinese Foundrymen's Association, Vol. 29, June 1981, pp. 10-18. ACKNOWLEDGEMENTS

163

Telegas II™ for on-line measurement of hydrogen in aluminum alloy melts D.A. Anderson, D.A. Granger and J.G. Stevens Alcoa Laboratories, Alcoa Center, Pennsylvania 15069, U.S.A.

HYDROGEN EFFECTS IN ALUMINUM ALLOYS

Even at low hydrogen concentrations there can be serious degradation in properties of an aluminum alloy because the solubility of the gas is so low in the solid compared with the liquid (Eichanauer, Hattenbach, and Pebler, 1961; Ransley and Neufeld, 1948; Talbot and Anyalebechi, 1988). This marked difference in solubility of hydrogen at the melting point results in the rejection of gas from the solid during freezing, and leads to its concentration in the liquid where gas porosity forms once the solubility level is exceeded. The relationship between gas content and porosity formation is principally a function of alloy composition and local solidification rate. Both qualitative (Ransley and Talbot, 1955) and quantitative (Fang and others, 1988; Thomas and Gruzleski, 1978) relationships have been developed for the "threshold hydrogen content," which is described as the hydrogen content below which macroscopic porosity is avoided. When quantitative microscopy is used in place of density measurements, porosity may be detected at lower hydrogen levels (Fang and others, 1988; Fang, Anyalebechi and Granger, 1988). Hydrogen porosity is reported to have a detrimental effect on: (i) mechanical properties in plate and forgings (Hess and Turnbull, 1974; Talbot, 1975; Turner and Bryant, 1967), (ii) tensile properties, particularly the ultimate tensile strength, of castings (Jay and Cibula, 1956), (iii) resistance to fatigue crack propagation (Renon and Calvet, 1961), and (iv) surface finishing characteristics (Hess and Turnbull, 1974; Talbot, 1975).

METHODS OF HYDROGEN REDUCTION

Considerable time and effort have been devoted to reducing hydrogen content of molten aluminum and its alloys. In the majority of commercial processes existing in the aluminum industry today, a gas, typically a mixture of argon (or nitrogen) and a small amount of reactive gas such as chlorine, is introduced into the melt to reduce the hydrogen content. Driven by the concentration difference of hydrogen in the melt and its absence, or minute concentration in the fluxing gas bubble, hydrogen diffuses into the bubble, and is carried out of the melt. This results in the desired decrease of melt hydrogen content, provided that the rate of hydrogen removal is greater than the pick-up induced when the escaping bubble breaks the metal surface. The fluxing gas may be introduced into the molten aluminum in a variety of ways. In some foundry applications gas is generated in the melt by immersion of tablets containing gas evolving salts such as hexachloroethane, aluminum fluoride, and aluminum chloride. The disadvantages of this method of hydrogen reduction include the fact that it is expensive, excessive skim is generated, there may be noxious fumes, and there is rarely a uniform distribution of the fluxing gas throughout the melt. In most fluxing processes, gas is injected under pressure through a pipe, lance, diffuser plug, nozzles, or some other type of disperser. Inert gases may be used to afford the hydrogen reduction; however, in many cases a small amount of reactive gas is added to the flux. The most common reactive gases used include chlorine, Freon® 12, and sulfur hexafluoride. The concentration of reactive gas used is typically less than 10%. In certain cases, however, fluxing has been carried out with 100% reactive gas; this practice has shown no well-documented advantage. The primary disadvantage is, of course, the fumes generated. The classic hydrogen reduction method is simple batch-type furnace fluxing using straight flux tubes, or pipes. This is a very inefficient procedure because of the large flux bubbles generated. These bubbles are rarely uniformly distributed, and the disturbances they cause in the bath result in an increase of particles suspended in the melt. A "settling time" must therefore be added to the production cycle (Martin and others, 1989). Hydrogen reduction is greatly facilitated by the generation and uniform dispersion of small flux bubbles.


Many in-line commercial processes are available to address this need. They include: MINT (Dore and Milligan, 1986) (Melt In-line Treatment System) which utilizes high pressure gas injection nozzles, SNIF (Dokken and Griffin, 1985) (Spinning Nozzle Inert Flotation), and the Alcoa 622 Process (Stevens and Yu, 1986; Stevens and Yu, 1988), which will be discussed in detail in a later section of this paper. Units of this type are currently the most popular and efficient means to accomplish adequate hydrogen reduction. Less conventional methods of hydrogen reduction such as vacuum degassing and ultrasonic vibrations have also been examined to a limited extent. The use of ultrasonic vibrations (Cleg, 1986) has not been pursued commercially to date, due to the amount of time necessary to degas aluminum melts in this fashion. Vacuum degassing (Van Wijk and Ackermann, 1978), while it is not popular in North America, has been practiced in Europe. While this method has the advantage of minimal dross formation, its disadvantages include extremely high capital and operating costs. These disadvantages limit the use of this method considerably.

HYDROGEN MEASUREMENT

There are a number of techniques for assessing hydrogen content in aluminum alloys ranging from the strictly qualitative to quantitative methods. The qualitative tests such as the Straube-Pfeiffer (Brondyke and Hess, 1964) and initial bubble test (Dardel, 1948) are not direct measurements of the hydrogen content and can be misleading when used for this purpose. As discussed by Brondyke and Hess,20

these techniques are useful as measurements of "metal quality," but should not be used for quantifying the hydrogen content of the melt. This is because tests, such as the initial bubble method, depend on the presence of nuclei in the melt to assist in the creation of a hydrogen bubble. Attempts to avoid this problem by inducing nucleation have had limited success and can only be described as semi-quantitative (Hess, 1973). The two most successful quantitative methods for measuring hydrogen in aluminum are: (i) the hot-extraction method (Ransley and Talbot, 1956), and (ii) the Telegas instrument (Ransley, Talbot and Barlow, 1958). The former requires a sample taken from the melt, casting or wrought product, which must be carefully prepared by surface machining or etching, after which the hydrogen is extracted by heating it to a sub-solidus temperature in a good vacuum and measuring the gas evolved. It is a time-consuming and expensive laboratory method not suitable for everyday use in the ingot plant or foundry. In contrast, the Telegas method provides a means of obtaining a direct reading of the hydrogen content of the melt. The major drawbacks limiting its use on the shop floor, particularly for process control purposes, have been its fragility and difficulty of use (Granger, 1986). As described later, extensive changes in its construction, and introduction of microprocessor controls have provided the casting shop operator with a more reliable, easier-to-use and portable instrument.

REDUCING HYDROGEN CONCENTRATION WITH THE ALCOA 622 PROCESS

The process used for hydrogen reduction in most Alcoa plants is the Alcoa 622, shown in Figure 1. The refractory-lined reaction vessel consists of inlet and outlet compartments, separated by a silicon carbide baffle. Molten aluminum flows down through the inlet and up through the outlet, while the temperature of the unit is maintained with gas fired immersion heaters. Argon or argon with a small amount of chlorine is introduced into the metal treatment unit through the disperser shaft. The gas exits the underside of the disperser and is broken into small bubbles, one to two millimeters in diameter, as it passes up through the rotating vanes. The Alcoa 622 Process is typically run as a one, two or three stage unit, depending on the molten metal flow rate and the hydrogen reduction requirements of the system.


Fig. 1 One Stage Alcoa 622 Unit

As explained by Stevens and Yu (1988), hydrogen reduction in the Alcoa 622 Process is a combination of various modes of mass transfer. The mass balance on the hydrogen not only includes a term to account for hydrogen stripped from the melt by fluxing, but also a term to describe the melt's interaction with the atmosphere. Coupling this with the hydrogen in the bulk flow terms leads to the following equation for each stage.

Amks(k3(H2O))(0-5) + F(H)0 - (Amks+F+VKa)(H)i = 0 (1)

where:

Am = surface area of melt ks = mass transfer constant k3 = reaction rate constant (H2O) = water concentration of atmosphere F = metal flow rate (H)o = hydrogen concentration of inlet stream (H)i = hydrogen concentration of outlet stream V = mass of aluminum in stage and

( ° · 1 0 4 5 0 3) (1,3601)

*-im «a' (2)

with:

z = a constant, dependent on alloy N = rotor speed D = rotor diameter G = gas flow rate


Equation (2) describes the collective influence of rotor speed, rotor diameter and gas flux on the product of the reduction rate constant (due to fluxing), and the fluxing gas interfacial area. This relationship is a characteristic of the Alcoa 622 Process, and its development is detailed in previous work (Stevens and Yu, 1986; Stevens and Yu, 1988). Equations (1) and (2) are adequate to describe hydrogen reduction in the Alcoa 622 Process. The process model therefore consists of N equations of the form of equation (1), where N is the number of stages in the system. This process model can be used with an in-line hydrogen detection unit, such as the Telegas Π instrument, to provide control of the Alcoa 622 Process. System upsets, such as an increase in incoming hydrogen level, can be detected by the Telegas Π unit. The model can then be used to give information necessary to control the argon flow rate so that the outlet hydrogen level stays below that dictated by the product specifications.

DEVELOPMENT OF TELEGAS II

The concentration of hydrogen in solution in molten aluminum alloys is proportional to the square root of the partial pressure of hydrogen at a free surface within the metal (Sieverts' law):

where:

So is the solubility at 760 mm Hg at a given temperature Pi is the equilibrium internal pressure of hydrogen

The Telegas instrument was developed by Ransley, Talbot, and Barlow (1958) to measure the partial pressure of hydrogen in molten aluminum. The basic principle, shown in Figure 2, was to create a circulating gas volume within the molten metal into which hydrogen could diffuse. A suitable inert carrier gas such as nitrogen is bubbled through the melt and collected by a probe with an inverted bell shape. Hydrogen diffuses into the carrier gas bubbles until it has reached its equilibrium value. The partial pressure of hydrogen in the carrier gas is measured by a differential thermal conductivity technique. The thermal conductivity of hydrogen is more than an order of magnitude greater than that of most inert gases, such as nitrogen and argon, which makes this type of measurement feasible. The other key factor is the relationship of the hydrogen partial pressure to the liquid phase gas content which is a strong function of temperature. For pure aluminum this relationship is defined by the equation:

logioSo = x + B

where:

A and B are constants and T is the temperature in degrees Kelvin

For alloys of aluminum, a factor is empirically determined specific to that alloy which relates the solubility of the alloy to that of pure aluminum. Direct measurement of the hydrogen content of aluminum alloys therefore involves a Telegas measurement, a temperature measurement, and a computational means of relating these measurements to liquid phase gas content.

PRODUCTION AND ELECTROLYSIS OF LIGHT METALS Fig. 2

Schematic Drawing of the Alcoa Telegas Instrument

167

Wheatstone Bridge

Voltage Regulator 4V Output

Diaphram , Check / Pump

Molten Metal Level N

Telegas Instrument

Pneumatic Operator

Purge Valve

—$l·-

££> Nitrogen Cylinder

Alcoa licensed the right to manufacture and sell instruments based on the Telegas principle and has done so for two decades. This instrument is basically the Ransley instrument with a pneumatically-driven pump rather than the original hand-cranked pump. A separate temperature measurement is required to relate the Telegas reading to liquid phase gas content via a set of charts in the instrument manual. Telegas has proven reliable and effective in industrial environments for many years. However, it has several drawbacks: it is quite bulky because of the large compressed gas cylinder used to power the pump; the conversion of the Telegas meter reading into a gas content is error prone; the probe is delicate and expensive; and the instrument uses many components which require factory maintenance. A new instrument, Telegas Π, has recently been developed which addresses these problems.

TELEGAS II - NEW FEATURES

The Telegas Π instrument is a completely redesigned instrument sharing only the basic circulating gas and differential thermal conductivity principles with the original instrument. The objective was to design an instrument which could be carried in one hand to any location in a plant. Additionally, the instrument was designed to have computational ability, relieving the operator of the tedious duty of converting Telegas readings to liquid phase gas contents. It was designed so that the instrument could be interfaced directly to a mainframe computer, if desired. This feature simplifies statistical compilation of plant data or feedback control of fluxing units. The new instrument is battery powered, compact, and includes an onboard printer for hard copy output. The original Telegas instrument has two platinum filaments arranged in a Wheatstone Bridge circuit. One filament is exposed to air or pure carrier gas while the other is exposed to the gas circulating through the molten metal. Because of the substantially higher thermal conductivity of hydrogen, the sensing filament cools when hydrogen is introduced. Since the temperature of the filament is directly proportional to its resistance, the resistance change is reflected in a signal on the Telegas meter. A single commercial hot film sensor is used in Telegas Π in a constant temperature circuit, shown in Figure 3. This has two advantages: 1) the change in sensor voltage is greater for a given change in gas concentration; 2) the other half of the Wheatstone Bridge is eliminated along with its substantial current drain. However, it requires a measurement of the ambient temperature in the sensor cavity. Because the sensor reading is compensated for the changes in ambient temperature, there is no need to wait for the


system to warm up and come to equilibrium. This problem with the Telegas instrument was previously addressed by Terai and others (1984). The present configuration has all of the advantages of quick warm-up, but without the bulky vacuum pump of the Terai and others instrument.

Fig. 3 Constant Temperature Filament Excitation Circuit

+ Battery

Constant Current

To Measuring

Circuit

Ultra Stable Voltage Reference

Temperature Controller

v V Hot Film Sensor

Constant temperature operation of hot film sensors is the most common configuration for velocity sensors. However, the relative change in sensor output for a velocity sensor is orders of magnitude higher than the small change in natural convective heat transfer caused by changes in gas thermal conductivity. Therefore, commercial anemometer circuits could not be used in Telegas Π. High gain and high thermal stability in the feedback loop are not enough to protect the instrument from short term or thermal drift. An ultra stable voltage reference circuit is fed with a constant current and used as the reference for a voltage regulator at the input of the constant temperature portion of the circuit. A U.S. patent (Anderson, Warchol, and Wojnar, 1989) is pending for this circuit and foreign patents have been applied for.

TELEGAS II - MECHANICAL DESIGN

The physical design of Telegas Π is completely different from the Telegas instrument. The instrument is packaged in a 25.4 cm by 26.7 cm by 17.8 cm box with a hinged lid and weighs only 9.1 kg. The main power source is a rechargeable battery (an AC powered version is currently being designed) as opposed to a large nitrogen cylinder. Gas circulation is accomplished by an electric motor driven metal bellows pump. The carrier gas is stored in a 150 cc high pressure sample gas cylinder and is admitted by a miniature solenoid valve. The sensor is contained in a standard in-line filter assembly. This provides a compact mechanical enclosure for the delicate sensor and provides a very effective way of isolating the sensor from the flow. Using concentric tubes, circulating flow is forced into one end of the filter assembly and back out the same end. The sensor is isolated from the flow effects by the fine filter element, but hydrogen quickly diffuses through to the sensor side. The instrument, shown in Figure 4 both in its case and with the internal assembly removed, is built on a base to which all internal components and the operator console are attached. The base is attached to the bottom of the instrument enclosure with one machine screw. Removing the instrument for service is thus quite easy. The sensor assembly can be removed by plant personnel for service, rather than sending the whole instrument back for repair. Recalibration is quite easy and requires only the carrier gas and one gas standard. The calibration procedure is menu driven and easy to follow. The pump check valves are an assembly and can be replaced in the field. These are only a few examples of the design principles for ease of maintenance and field serviceability.


Fig. 4 Telegas II

TELEGAS II - MICROPROCESSOR CONTROL

All instrument functions on the Telegas II instrument are controlled by a microprocessor based single card computer. The operator communicates to the computer via a low power micro-terminal which has a 4 line by 16 characters/line display and 20 function keypad. A custom designed interface board has the sensor excitation, sensor block temperature circuit, melt thermocouple amplifiers, and three solid state switches. The switches actuate the carrier solenoid valve, the printer and the circulating pump motor, and are automatically controlled by the microprocessor. The processor card has a real time clock which is active whenever the battery is engaged. When not in use, the instrument can be placed in a low power mode in which all analog systems are off, the microprocessor is in a low current mode, but the real time clock is active. In this mode the instrument can retain a charge on its battery for a week or more. The operating program for Telegas Π is burned onto an Erasable reprogrammable Read Only Memory (EPROM) circuit contained in a socket on the microprocessor board. Updates in the software can therefore be incorporated in the field by simply changing the EPROM. For example, new software will soon be available which allows the user to create (and change) a custom library of up to 15 alloys and correction factors (in addition to the present on-board library). The instrument also has an Erasable and Electrically reprogrammable Read Only Memory (EEPROM) in which calibration constants are stored. The calibration procedure automatically rewrites the EEPROM with new constants and they are held even when power is removed. This is done from the keypad and does not require any disassembly. The instrument has 64K of Random Access Memory (RAM) in which current readings can be stored. This memory is volatile and is lost when the battery is removed. As an option, the instrument has a switch and connector which allows it to talk to a mainframe computer via RS232. In this mode, the onboard terminal is disconnected. This allows the user to access RAM where current readings are stored. Additionally, the user can reprogram the instrument to suit the application or operate it from a remote terminal. The instrument is programmed in a modified version of BASIC, which is quite easy to learn.

TELEGAS II - INSTRUMENT OPERATION

The instrument was designed for ease of use and does not require constant attention by the operator. All software is menu driven, prompting the operator through the measurement procedure. The first menu, which appears automatically when the instrument is powered up, is shown below:


1 - MEASURE 2 - SETUP 3 - SETTIME 4 - QUIT _

The operator simply enters the number of the option he wishes to select. When the Measure option is selected, the instrument automatically starts purging with carrier gas and prompts the operator to put the probe in the melt. When the instrument senses that the probe has been inserted into the melt (from the temperature), it automatically starts the pump and continues taking readings until the operator presses the reset switch. The analyses are automatically printed on the onboard 20 column printer. The Setup option allows the operator to select the alloy being tested (from an internal library of over 40 alloys), the carrier gas being used (argon or nitrogen), and the type of probe. The Quit option puts the instrument in a low power mode in which all analog systems and the terminal are off. The original Telegas probe is delicate and expensive, but works well and with proper care has good life. Several new probes have been reported (Martin, Tremblay, and Dube, 1989; Pelton, 1986) and Telegas II is compatible with these new designs. Alcoa is working on several new probe designs which will reduce the cost and will be more rugged than the present ceramic probe. Some of these new probes will allow another feature of the instrument to be exploited, namely the use of argon carrier gas. The original Telegas instrument is restricted to nitrogen carrier gas because it is referenced to air on one leg of the bridge circuit. Telegas Π does not have this restriction and can therefore be used with any inert carrier which has substantially different thermal conductivity compared to hydrogen. Argon has the advantage of being available in many casting pits where it is used as a fluxing gas, and therefore it is a simple matter to connect the instrument to plant argon. This relieves the operator of the task of refilling carrier gas cylinders.

SUMMARY

The Telegas principle is a proven and effective technique for direct measurement of hydrogen concentration in molten aluminum alloys. The development of the Telegas II instrument represents a major improvement in instrument operation. Combining accurate in-line measurement of hydrogen content with an in-line removal system makes feedback control possible. To date this has been accomplished by operator action; however, with intelligent instruments like Telegas II, the loop can be closed automatically. The ability to communicate analyses directly to a computer via RS 232 also makes possible automation of quality control charts. The benefits of these procedures are well-established and have the potential of significantly improving ingot plant performance.

ACKNOWLEDGEMENTS

The authors wish to thank M. F. A. Warchol and R. C. Wojnar for their assistance in designing the Telegas Π instrument. A special thanks to M. L. Hritz, who builds the Telegas instrument, for his assistance in developing the Telegas II instrument.

REFERENCES

Anderson, D. A., Warchol, M. F. A., and Wojnar, R. C. (1989). Filament drive circuit for measuring gas content in molten metal, U.S. Patent No. 4.829.810.

Brondyke, K. J. and Hess, P. D. (1964). Interpretation of vacuum gas test results for aluminum alloys. Trans. Met. Soc. AIME. 1542.

Clegg, A. J. (1986). Aluminum degassing practice. Conf. Proc. International Molten Aluminum Processing. 375, 17-18.

Dardel, Y. (1948). Hydrogen in aluminum. Trans. AIME. Metals Technology. T.P. No. 2484.

Dokken, R. N. and Griffin, J. V. (1985). In-line refining with SNIF. Proceedings of Aluminum Conference '85. 1-7.


Dore, J. E. and Milligan, B. R. (1986). MINT: an in-line treatment system for removing impurities from aluminium alloy melts. Aluminum Technology '86 Conf. Proa. 101-110.

Eichanauer, W., Hattenbach, K., and Pebler, A. (1961). Solubility of hydrogen in solid and liquid. Z. Metallkunde. 52, 682.

Fang, Q. T. and others (1988). Effect of solidification conditions on hydrogen porosity formation in unidirectionally solidified aluminum alloys. Solidification Processing. 33-36, The Institute of Metals.

Fang, Q. T., Anyalebechi, P. N., and Granger, D. A. (1988). Measurement of hydrogen porosity in unidirectionally solidified aluminum alloys. Light Metals 1988. Proceedings of the 117th TMS-AIME Annual Meeting. 477-486.

Granger, D. A. (1986). Telegas for determining hydrogen in the foundry industry. Proceedings of the AFS/CMI Conference on Molten Metal Processing.

Hess, P. D. and Turnbull, G. K. (1974). Effects of hydrogen on properties of aluminum alloys. Int'l Conf. on Hydrogen in Metals. ASM, 277-287.

Hess, P. D. (1973). Measuring hydrogen in aluminum alloys. J. of Metals. 10, 2-6.

Jay, R. and Cibula, A. (1956). Influence of magnesium content on layer porosity and tensile properties of sand-cast aluminium/magnesium alloys B.S. 1490-L.M.10 and B.S. (aircraft) L.53. Foundry Trade J.. 101, 131-142.

Martin, J.-P., Dube, G., Frayce, D., and Guthrie, R. (1989). Settling phenomena in casting furnaces: a fundamental and experimental investigation. Light Metals 1989. Proceedings of the 118th TMS-AIME Annual Meeting. 445-455.

Martin, J.-P., Tremblay, F., and Dube, G. (1989). A1SCAN: a new and simple technique for in-line analysis of hydrogen in aluminium alloys. Light Metals 1989. Proceedings of the 118th TMS-AIME Annual Meeting. 903-912.

Pelton, J. F. (1986). Hydrogen probe. U.S. Patent No. 4.624.128. Assigned to Union Carbide Corp.

Ransley, C. E. and Neufeld, H. (1948). The solubility of hydrogen in liquid and solid aluminium. J. Inst. Metals. 74, 682.

Ransley, C. E. and Talbot, D. E. J. (1955). Hydrogen porosity in metals with particular reference to aluminium and aluminium alloys. Z. Metallkunde. 46, 328.

Ransley, C. E. and Talbot, D. E. J. (1956). The routine determination of the hydrogen content of aluminium and aluminium alloys by the hot-extraction method. J. Inst. Metals. 84,445-452.

Ransley, C. E., Talbot D. E. J., and Barlow, H. C. (1958). An instrument for measuring the gas content of aluminium alloys during melting and casting. J. Inst. Metals. 86,212-219.

Renon, C. and Calvet, J. (1961). On cavities observed in aluminum and its alloys. Mem. Sei. Rev. Metallurg.. 58, 835-851.

Stevens, J. G. and Yu, H. (1986). A computer model of a stirred tank reactor in trace alkaline elements removal from aluminum melt - the Alcoa 622 process. Light Metals 1986. Proceedings of the 117th TMS-AIME Annual Meeting. 837-845.

Stevens, J. G. and Yu, H. (1988). Mechanisms of sodium, calcium, and hydrogen removal from an aluminum melt in a stirred tank reactor - the Alcoa 622 process. Light Metals 1988. Proceedings of the 117th TMS-AIME Annual Meeting. 437-443.

172 PRODUCTION AND ELECTROLYSIS OF LIGHT METALS Talbot, D. E. J. (1975). Effects of hydrogen in aluminium, magnesium, copper, and their alloys, Int'l Metallurgical Rev.. 20, 166-184.

Talbot, D. E. J. and Anyalebechi, P. N. (1988). Solubility of hydrogen in liquid aluminium. Materials Science and Technology. 4, 1-6.

Terai, S, Sato, S., Kato, S., Imai, M., Inumaru, S., and Yoshida, M. (1984). Apparatus for measuring the content of hydrogen dissolved in a molten metal. U.S. Patent No. 4.454.748. Assigned to Sumitomo Light Metal Industries, Ltd.

Thomas, P. M. and Gruzleski, J. E. (1978). Threshold hydrogen for pore formation during the solidification of aluminum alloys. Met. Trans.. 9B. 139.

Turner, A. N and Bryant, A. J. (1967). The effect of ingot quality on the short transverse mechanical properties of high strength aluminium alloy thick plate. J. Inst. Metals. 95, 353.

Van Wijk, G. W. M. and Ackermann, D. M. (1978). Light Metals 1978. Proceedings of the 107th TMS-AIME Annual Meeting. 235.

173

Influence of solution treatment on tensile properties of Sr-modified Al-Si-Mg alloys S. Shivkumar, S. Ricci, Jr. and D. Apelian Aluminum Casting Research Laboratory, Department of Materials Engineering, Drexel University, 32nd and Chestnut Street, Philadelphia, Pennsylvania 19104, U.S.A.

ABSTRACT

An investigation has been conducted to determine the influence of solution heat treatment parameters on Ύ6 properties of unmodified and Sr-modified A356.2 alloys. The effects of solution temperature and time were studied with ASTM B-108 test bars cast in sand molds and in permanent molds. The microstructural changes occurring during heat treatment have been investigated by metallographic and image analysis techniques. The results indicate that the effect of Sr modification is to increase the spherodization rate and lower the coarsening rate of Si particles. Silicon particle characteristics are strongly influenced by the solution temperature. Extremely high coarsening rates can be obtained at temperatures greater than 560C. Solution treatment at these temperatures, however, has a detrimental effect on mechanical properties because of grain boundary melting. It has been established that solution times can be reduced significantly in modified samples. Increasing the solution temperature from 540C to 550C enhances the strength properties.

INTRODUCTION

Cast aluminum components are generally heat treated to the T6 condition in order to enhance mechanical properties. Sand castings are solution treated at 540C for 10 to 18 hr, quenched in water and aged at 154C for 4 to 8 hr. Permanent mold castings are solution treated at 540C for 4 to 8 hr, quenched in water and aged at 171C for 4 to 8 hr M. The primary purpose of the long solution treatment is to thermally alter silicon particle characteristics. The solution treatment changes the morphology of silicon from a polyhedral to globular structure and enhances mechanical properties appreciably. In addition, when the castings are solutionized, magnesium and some silicon dissolve to produce a homogeneous solid solution. The Mg and Si which are in solid solution precipitate as Mg2Si during the aging treatment. The influence of various parameters on tensile properties of the heat treated product have recently been reviewed by Apelian et al 121.

Molten metal processing treatments which have been adopted by foundrymen such as

grain refinement and modification may have a significant influence on the heat treatment procedure. It has been reported that modification of the melt with Sr or Na alters the kinetics of spherodization and coarsening processes during solution treatment ß-5]. As a result, it may be possible to reduce solution times considerably in fully modified castings [6-8] j n addition, preliminary investigations indicate that T1AI3 particles, which are introduced into the melt through grain refiner additions, may delay the precipitation kinetics of Mg2Si phase during aging t9 l . The practical implications of these observations are that heat treatment standards which are presently being used in the cast shop were developed several decades ago and need to be revised to suit current foundry practice. In the present contribution, the results of a detailed investigation conducted to optimize solution heat treatment parameters in modified and unmodified castings have been reported. This study is part of a larger program on the heat treatment and feeding characteristics of cast aluminum alloys being carried out at the


Aluminum Casting Research Laboratory (ACRL), Drexel University.


The effects of solution treatment time and temperature were evaluated for two different casting techniques: sand casting and permanent mold casting. In both cases, unmodified and Sr-modified ASTM B-108 A356.2 test bar samples were used to examine the influence of selected variables on tensile properties. Sand cast test bars were produced at Littlestown Hardware & Foundry Co., Inc. in Littlestown, Pennsylvania. Permanent mold test bars were cast at Stahl Specialty Co. in Kingsville, Missouri NO].

The castings were heat treated at Drexel University in a resistance-heated air circulating box-type muffle (Lucifer) furnace. The T6 heat treatment cycle is summarized below:

• Solution treatment: Initially, the test bars were solution treated at 540±2C for times ranging from 25 min to 1600 min. In subsequent experiments, the effect of both solution temperature and time was studied. The solution temperature was varied from 520C to 570C.

• Quench in water at 60C • Natural age at room temperature for 24

hr • Age sand cast bars at 154±2C for 4 hr • Age permanent mold test bars at

171±2Cfor4hr

The as-cast and heat treated samples were analysed to assess microstructural changes occurring during heat treatment. The LeMont image analysis system was used to measure structural characteristics of the Si-rich particles. The OASYS linescan algorithm was utilized for this purpose. The fields of observation were selected randomly. At least 10 fields were analysed from a single specimen. The electrical conductivity of solution heat treated samples was measured in order to monitor structural changes occurring during solution treatment. The Magnaflux FM-140 digital conductivity meter was used for measuring conductivity. The yield strength (YS), ultimate tensile strength (UTS), %elongation and Rockwell (F) hardness were

estimated from heat treated samples. A minimum of 10 bars were tested under each condition. Other pertinent details regarding the experimental procedure are summarized elsewhere NO],


Microstructures illustrating the influence of solution treatment time on Si particle morphology in sand cast specimens are presented in Fig. 1. It can be seen that as the solution time increases, Si particles undergo necking and are broken down into smaller fragments. The fragmented particles are gradually spherodized. Prolonged solution treatment leads to extensive coarsening of the particles. Modification has a strong influence on spherodization and coarsening of Si particles. A high degree of spherodization is observed in modified alloys after only 50 min of solution treatment. In unmodified specimens, however, even after 800 min of solution treatment several long needles of silicon are visible. A similar behavior is observed in permanent mold test bars. In this case, the time required for spherodization are much smaller than in sand cast specimens. Interfacial instabilities cannot readily occur in plate-like (unmodified) eutectics and the structure is resistant to spherodization tH]# Fibrous eutectics (modified), however, are susceptible to shape perturbations and the particles are easily spherodized. The spherodization and coarsening processes occur because of the instability of interfaces between two different phases and are driven by a reduction in the total interfacial energy.

Image analysis data for both sand cast and permanent mold test bars are shown in Fig. 2. In this figure, Si particle characteristics such as average particle diameter, aspect ratio and number of particles are plotted as a function of solution time. The average diameter initially registers a decrease because of fragmentation of the particles. Subsequently, particle coarsening leads to a significant increase in the average diameter. Because of the higher cooling rate, average equivalent diameter is much larger in permanent mold than in sand cast specimens. The average silicon particle diameter is much smaller in modified than in unmodified samples. The large diversity in particle size and shape in unmodified alloys provides a greater driving force for coarsening


of Si particles than in modified samples. Consequently, higher growth rates are observed in unmodified than in modified alloys. Typical growth rates have been estimated to be of the order of 0.04 to 0.1 μητ/ητ and 0.02 to 0.05 μιη/hr in unmodified and modified alloys respectively. In modified castings, particle growth obeys equations developed for diffusion-controlled growth and particle volume is proportional to v-ß t12l. In unmodified sand castings,the dependence of particle size on t1/3 is observed only after large solution times (> 200 min). In this case, the assumption of spherical morphology for Si particles is not valid at short solution times. Spherical particles begin to dominate only after the casting has been solutionized for extended periods. Meyers 141 has conducted a detailed study of the coarsening process in unmodified A357 samples and has shown that the t1/3 dependence of particle radius is valid after a solution time of 5 hr. The number of particles varies inversely with the particle diameter since the total volume of Si remains constant. Up to about 100 min of solution treatment, there is a rapid reduction in the aspect ratio. Subsequently, there is relatively small decrease in the aspect ratio. As can be expected, unmodified alloys exhibit a higher aspect ratio than modified alloys. Also, aspect ratio is greater in sand cast specimens than in permanent mold samples.

The segregation of Si and Mg in Al-Si-Mg alloys is not severe and, consequently, homogenization and dissolution of Mg2Si occur in a relatively short period PI. Electron microprobe analysis was conducted on as-cast and heat treated samples in order to investigate the kinetics of the homogenization process. The Si and Mg concentrations were determined across the dendrites. The results are shown in Fig. 3. In the as-cast samples, the highest concentration of Si is found at locations close to the center of the dendrite. Relatively large fluctuations are observed in the Mg concentration. In both sand cast and permanent mold castings, Si and Mg concentration becomes almost uniform after 50 min of solution treatment indicating that homogenization is essentially complete within this short period. No significant differences can be detected between modified and unmodified alloys. Similar results have been reported by Closset et al Π3]# At solution temperatures commonly employed in

commercial castings (540C), about 0.6% Mg can be placed in solid solution. This value is lower than the magnesium concentrations in most castings and hence, all of the magnesium is in solid solution after the solution heat treatment.

The electrical conductivity of the test bars is plotted as a function of solution time in Fig. 4. In the as-cast condition, sand cast samples exhibit a lower conductivity than permanent mold specimens. Note also that higher conductivities are obtained in modified than in unmodified specimens because of the differences in eutectic Si morphology i14l· As the solution time increases, conductivity initially decreases, attains a minimum value and then begins to increase. The initial decrease is due primarily to the dissolution of solutes in the matrix and the subsequent increase results from changes in the Si particle morphology. A minimum in conductivity is observed after 50 min of solution treatment in sand cast specimens and after 25 min in permanent mold specimens. These results also confirm that homogenization of the casting and dissolution of Mg2Si are essentially complete within 50 min of solution treatment.

The solution treatment temperature also plays a major role in determining the Si particle characteristics (Fig. 5). The spherodization and coarsening rates are directly proportional to the solution temperature. At temperatures greater than 550C, extremely large silicon particles are observed in the microstructure. The activation energy for coarsening has been measured to be of the order of 80 kcal/mole indicating that the coarsening process is very sensitive to temperature fluctuations t4l. At a temperature of 570C, particles exhibit faceting even at short times. The temperature may not be increased indefinitely because of the formation of complex eutectics which melt at temperatures below the equilibrium eutectic temperature (Fig. 5). If the temperature of solution treatment exceeds 560C, the melting of the ternary eutectic phases occurs at the grain boundaries. These ternary phases are predominantly iron-rich particles.

The average yield strength, ultimate tensile strength and %elongation are plotted in Fig. 6 as a function of time at 540C. Both YS and UTS generally improve with solution time and


attain a maximum after about 200 to 400 min in sand cast alloys and about 100 min in permanent mold samples. Subsequently, strength properties register a slight decrease and then remain essentially constant. Modified alloys exhibit higher YS and UTS than unmodified alloys t1^]. The hardness of the test bars does not show any significant dependence on solution treatment time t10L The property that is most affected by modification and by solution treatment is %elongation. As can be expected, permanent mold test bars exhibit a higher elongation than sand cast bars. In addition, modified bars possess a higher elongation than unmodified samples. In sand cast bars, %elongation gradually increases with solution time. After about 400 min of solution treatment, modified alloys possess an elongation which is equivalent to that observed after 1600 min in unmodified alloys. Modification of sand cast alloys improves the tensile properties despite the introduction of porosity in the casting t10L Density measurements in the casting indicated that porosity levels in modified samples were higher (~ 1.5%) than in unmodified test bars (~ 0.8 to 1%). It should be noted, however, that these porosity values are typically observed in most commercial sand castings.

The variation of YS, UTS and elongation with solution temperature is shown in Fig. 7. Both YS and UTS improve with temperature and maximum properties are attained at 550C. At this temperature, maximum properties are obtained at times of the order of 50 to 100 min in sand cast alloys and about 25 to 50 min in permanent mold castings. In unmodified sand cast alloys, substantial improvements in elongation can be obtained by increasing temperature from 540C to 550C. In modified sand cast alloys, however, elongation registers a small decrease when solution temperature is increased from 540 to 550C. In permanent mold specimens, elongation is essentially unaffected as temperature is increased from 540 to 550C. Because of grain boundary melting, YS and UTS and elongation are lowered appreciably at 560C and 570C.

This work demonstrates that solution times can be reduced significantly in modified castings. Optimum solution times at 540C are of the order of 50 to 100 min in permanent mold castings and about 200 to 400 min in

sand castings. The solution times can be reduced further by increasing the solution temperature. For example, at 550C, optimum solution times are about 25 to 50 min in permanent mold castings and about 50 to 100 in sand castings. When the solution temperature is increased above 540C, it becomes imperative that temperature in the parts being heat treated be controlled precisely in order to avoid grain boundary melting. Differential thermal analysis experiments indicate that the first liquid begins to form in the temperature range 560 to 563C. But this temperature depends on chemical composition of the alloy, primarily copper and iron concentrations. Hence, the temperature of eutectic melting should be determined accurately before elevating the solution temperature. In this respect, several other investigators have reported that with proper control of melt quality and heat treatment conditions, solution treatment times can be reduced appreciably in modified castings [6,8,15]# Researchers in Europe t16l have proposed the total elimination of solution treatment in selected castings which are modified. Instead the casting is removed from the mold as soon as possible (when the metal temperature is still close to the eutectic temperature), and quenched. A subsequent aging treatment gives a good combination of strength and ductility. Heat treatment costs are typically of the order of 5 to 10 c/lb. The potential economic savings resulting from a reduced heat treatment cycle are substantial.

CONCLUSIONS

Recent developments in molten metal processing technology have a significant impact on the heat treatment response of cast Al-Si-Mg (A356.2) alloys. The influence of Sr modification on solution heat treatment has been illustrated in this contribution. Modified alloys exhibit a large spherodization rate and a high degree of spherodization is obtained after only about 1 hr of solution treatment at 540C. By comparison, long needles of silicon are observed even after 25 hr of solution treatment. The spherodization and coarsening rate increase with solution temperature. The solution temperature cannot be increased above 560C since differential thermal analysis indicates that the first liquid begins to form at the grain boundaries in the temperature range 560 to 563C. Solution treatment times can be


reduced significantly in modified alloys. At 540C, optimum solution times are of the order of 3 to 6 hr in sand cast alloys and 1 to 2 hr in permanent mold samples. These times can be reduced further by increasing the solution temperature. At 550C, optimum solution times are around 1.5 to 3 hr in sand cast alloys and 0.5 to 1 hr in permanent mold castings. The strength properties in the casting can be enhanced by increasing the temperature from 540C to 550C.

ACKNOWLEDGEMENTS

This research was conducted as a part of an ongoing research project at the Aluminum Casting Research Laboratory (ACRL). The authors would like to gratefully acknowledge the financial support of the consortium of companies supporting the Aluminum Casting Research Laboratory.

REFERENCES

1. Metals Handbook, Volume 15, Ninth edition, American Society for Metals, 1988

2. D. Apelian, S. Shivkumar and G. Sigworth, Trans AFS, 1989, 97, pending publication

3. B.A. Parker, D.S. Saunders and J.R. Griffiths, Metals Forum, 1982, 5(1), 48-53

4. F.N. Rhines and M. Aballe, Met Trans A, 1986,17A, 2139-2152

5. M.M. Tuttle and D.L. Mclellan, Trans AFS, 1982, 90, 13-23

6. HJ. Li, S. Shivkumar, XJ. Luo and D. Apelian, Cast Metals, 1989, 1(4), 227-234

7. M. Adachi, Alutopia(Jpn.), 1984, 14(12), 16-22

8. E. Mauveaux and M. Lafargi, Trait. Therm., 1982, 169, 31-35

9. M.S. Misra and K.J. Oswalt, Trans AFS, 1982, 90, 1-10

10. S. Shivkumar, S. Ricci, Jr., B. Steenhoff, D. Apelian and G. Sigworth, Trans AFS, 1989, pending publication

11. J.W. Martin and R.D. Doherty: "Stability of Microstructures in Metallic Systems", Cambridge University Press, London, 1980

12. I.M. Liftshitz and V.V. Sloyozov, J. Phys. Chem. Solids, 1961,19, 35-47

13. B. Closset, R.A.L. Drew and J.E. Gruzleski, Trans AFS, 1986, 94, 9-16

14. M.H. Mulazimoglu, R.A.L. Drew and J.E. Gruzleski, Met Trans, 1987, 18A, 941-947

15. M. Tsukuda, S. Koike and M. Harada, J. of Japan Inst. of Light Metals, 1978, 28(1), 8-14

16. M. Kaczorowski and R. Szostak, Aluminium, 1983, 59, 304-306


Fig. 1. Typical microstructures of sand cast test bars as a function of solution time (Solution temperature = 540C) (500X).


Sand Cast Permanent Mold

0 500 1000 1500 0 200 400 600 800 Solution time (min) Solution time (min)

Fig. 2. Variation of silicon particle average diameter, aspect ratio and number of silicon particles with solution time (Solution temperature = 540C).


2.4-

0.8-

1.2

0.8

As cast Heat treated

0.5

0.4

E k0.3 .3

(/> <D

0.2 |

0.1

o Unmodified • Modified

~τ 1 1 1— 10 20 30 40

Distance ( mm)

o.o

0.4

0.3

l· 0.2

0.1

E 3

Φ c

o.o 50

Si and Mg concentrations across a dendrite in as-cast and solution treated (50 min at 540O test bars.


35 35

34 ü <

"> 32 '■=

h31 o ϋ

30 500 1000 1500 0

Solution time (min)

100 200 300 400 500

Solution time (min)

Variation of electrical conductivity immediately after quenching (T4) as a function of solution treatment time.

180

Fig. 3.

Fig. 4.


Fig. 5. Typical microstructures of sand cast samples as a function of solution temperature (Solution time = 200 min) (500X).



400 800 1200 1600 0 200 400 600 800 1000 Solution time (min) Solution time (min)

Fig. 6. Variation of tensile properties with solution time (Solution temperature = 540CV


540 550 560 Solution temperature (°C)

540 550 560 Solution temperature (°C)

Fig. 7. Variation of tensile properties with solution temperature (Solution time =100 min).

182

185

Economic considerations in the production of electrolytic magnesium J .C. Agarwal, F.E. Katrak and M.J . Loreth Charles River Associates Incorporated, John Hancock Tower, 200 Clarendon Street, Boston, Massachusetts 02116, U.S.A.

All the current producers of magnesium use either the electrolysis of anhydrous magnesium chloride or the silicothermic reduction of dolime (and its variations), the two dominant processes of producing primary magnesium metal. All processes are energy-intensive and require low-cost electricity to be competitive. The major raw materials are dolime, seawater, chlorine, brines, and magnesite. The recovery of chlorine is extremely important in the economics of magnesium production.

All of the currently-used electrolytic processes require an anhydnms magnesium chloride feed to the electrolytic cells. The purity of the magnesium chloride feed has become extremely important and will become more so, because the heavy metal content of the final primary metal (i.e., the amount of nickel, copper, or iron, in parts per million) determines the corrosiveness and chemical behavior of magnesium and its alloys. The production of high-purity magnesium is the major requirement for the growth of the magnesium consumption is in its structural use. The importance of producing high-chemical-purity metal without flux contamination cannot be overemphasized.

The silicothermic process and its major modification, the magnetherm process, are batch processes that generate large slag volumes, require a metal purification step to remove inclusions, and are generally more labor-intensive than the electrolytic processes, these processes become uneconomical for capacities of more than 30,000 to 40,000 tonnes per year.

Magnesium production from 1984 through 1987 (see Table 1) had been relatively constant in the Western world, at about 235,000 tonnes per year. In 1988 and during the first quarter of 1989, there has been steady increase in consumption and production. It is currently running at the rate of about 255,000 tonnes per year. About two-thirds of the total production is located in the United States. Because of the past market discipline in the industry, the recent 5-year operating rate has been about 85 to 90 percent throughout the industry with a stable price history (see Table 2) and a market value of $824 million. This paper presents the cash costs of current producers.

TABLE 1 NCW Primary Magnesium Market: Overall Structure

Primary Production Primary Consumption Secondary Production Net Imports from the Communist

World

Figures in thousands of tonnes. SOURCE: I.MA. and Charles River Associates.

1975

174 166 38

1980

239 219 64

1985

234 225 70

1988

241 251 70

1989

255 256 70


TABLE 2 Noncommunist World Primary Magnesium Market Size

1975

Consumption (000 Tonnes) Producer Price (US$/lb.) NCW Market Size (US$ Million)

SOURCE: Charles River Associates.

1980

166 219 0.82 1.16 299 559

1985 1988

225 240 1.48 1.56 733 824

ELECTROLYTIC MAGNESIUM PRODUCTION PLANTS

Three plants in the noncommunist world, Dow, AMAX, and Norsk Hydro, currently produce magnesium by the electrolytic reduction of magnesium chloride. In addition, there are two potential new entrants, Norsk Hydro's Becancour plant in Canada and MagCan's plant near High River in Alberta, Canada. Norsk Hydro's plant will come onstream in 1989 and will have a capacity to produce 60,000 tonnes per year of magnesium by 1992. MagCan's plant will have an initial capacity of 12,500 tonnes of magnesium and may come on stream by 1990. MagCan's intentions are to increase capacity to 37,500 tonnes in the second stage and to 62,500 tonnes in the third stage, which could be as soon as 1996. With the addition of these two new plants, electrolytic magnesium capacity will increase from about 165,000 tonnes in 1987 to about 294,000 tonnes by 1996. In addition, Noranda is considering production of electrolytic magnesium in Quebec. When the Noranda plant comes on stream, Canada will replace the United States as the largest producer of magnesium metal.

The five electrolytic magnesium plants are quite different from one another in many respects: they use somewhat different sources of magnesium; plant designs differ significantly; and the costs of process materials, utilities, and labor also vary.

In all the electrolytic magnesium plants, the feed to the electrolytic cells is high-purity magnesium chloride. In the AMAX and in Norsk Hydro's Norway plants, concentrated brine is evaporated to prepare the feed. In Dow's plant, magnesium hydroxide is first produced and then chlorinated. MagCan does not use an acqueous process at all, but chlorinates the magnesite ore directly to produce anhydrous magnesium chloride. Labor costs are highly dependent on the production capacity and marketing support provided.

ECONOMICS

The direct operating costs included in this paper include all cash operating costs at the facility. They do not include depreciation, corporate changes, interest expenses, or costs associated with working capital fluctuations.

Direct operating costs for the five electrolytic magnesium plants are summarized in Table 3. The magnesium capacities of the electrolytic plants are quite different. Dow Chemical has the largest magnesium operation at 86,000 tonnes per year (a reduction from a previous capacity of 118,000 tonnes). Currently AMAX's magnesium plant has the smallest production of the three operating magnesium plants at 35,500 tonnes per year. However, the initial capacity of MagCan's magnesium plant is expected to be only 12,500 tonnes per year.


TABLE 3 Costs for Electrolytic Magnesium Plants

Direct Operating Cost (U.S. Cents per Pound M&

Materials Utilities Labor Maintenance & Repair Technical & Administrative

Expenses Property Taxes, Insurance

Other Subtotal

Credits

Total*

U.S. Dow Chemical 86,000 Tonnes

24 35

9 18

6

4 96

0

96

U.S. AMAX

35,000 Tonnes

20 49 12 20

4

4 1.09

(4)

1.05

Norway Norsk Hydro

50,000 Tonnes

24 31

8 19

3

5_ 90

(21)

69

Canada Norsk Hydro

60,000 Tonnes

40 16 7

15

2

8 82

0

82

Canada MagCan

125,000 Tonnes

32 21 18 23

5

2 1.08

0

1.08

Excludes corporate overhead, depreciation, and interest expense.

SOURCE: Charles River Associates, May 1989.

Magnesium-bearing raw material costs are high at Norsk Hydro's Norway and Canadian plants, and at MagCan's plant compared to the materials costs at Dow and AMAX. Dow uses sea water and AMAX uses solar evaporated salt lake brine, both relatively cheap sources of magnesium. The major cost item is energy. Energy costs at Dow and AMAX are the largest at 35 and 49 cents per pound, respectively. High energy costs for electricity and natural gas contribute substantially to these high costs. The electricity cost at Dow is based on rates charged to large industrial users in Texas. Dow obtains electricity from cogenerated power from nonmagnesium-related facilities in its Freeport complex. The magnesium facility is charged the industrial rate, since Dow could contract to sell this electricity at this rate to other users. The two potential magnesium plants in Canada have low energy costs due to both low-cost electricity contracts and energy-efficient operations.

The lowest-cost electrolytic magnesium producer is the Norsk Hydro plant in Norway. Its direct operating cost is 69 cents per pound of magnesium after byproduct credits. The substantial byproduct credit of 21 cents per pound results from selling chlorine to another operating unit within the complex. Norsk Hydro's plant in Canada is potentially the second-lowest-cost producer. However, the cost of electricity will increase from an estimated 21 cents per kilowatt hour in the first year of operation to 2 cents in the sixth year. This will increase direct operating costs by about 20 cents per pound of magnesium.

Direct operating costs at AMAX have decreased because new solar ponds are used again as a magnesium source instead of purchased brine. As MagCan increases capacity from 12,500 tonnes to 37,500 and 62,500 tonnes per year of magnesium, direct operating costs will be reduced. Total direct operating costs will decline to about 80 cents per pound at 62,500 tonnes.

TECHNICAL AND ECONOMIC ISSUES

Seawater, used by Dow, has the appearance of a low-cost raw material for magnesium ions. However, any seawater-based process must first calcine dolomite to precipitate magnesium hydroxide, which then has to be chlorinated.


Solar evaporation of the Great Salt Lake brine or any other brine also has the potential to be a low-cost source of magnesium. However, all these processes require extensive purification and evaporation steps to produce anhydrous magnesium chloride. Therefore, the thermal energy requirements are high. In addition, many brines, especially seawater, contain boron, an impurity which must be removed to produce acceptable-quality magnesium.

When magnesium is produced from a naturally occurring brine, nearly three pounds of chlorine are produced per pound of magnesium. A significant amount of this chlorine can be recovered and marketed if conditions so warrant. Therefore, Norsk Hydro's plant in Norway has an inherent advantage. The AMAX plant does not recover as much chlorine for sale as it could.

In all other electrolytic processes, makeup chlorine is required, and must be purchased or produced internally at a substantial cost.

Magnesite-based processes use a naturally occurring ore, which is either dissolved in hydrochloric acid (Norsk Hydro, Canada) or carbochlorinated, as by MagCan. All naturally occurring magnesite ores have impurities such as heavy metal ions, which must be removed before electrolysis to produce acceptable-quality magnesium metal.

All electrolytic processes require a solid magnesium chloride feed with minimum quantity of oxygen either as oxychlorides or as water of hydration. The electrolytic cells do not operate economically if this condition is not met. The electricity consumption in the electrolytic cells is generally in the range of 5 to 7 kWh per pound of magnesium. Thus, the electricity unit cost has a dominant impact on the economics of magnesium production. With the new discoveries of magnesite ores in Australia, perhaps Australia could challenge Canada as a major producer of magnesium because it too can supply relatively inexpensive electric power.

Finally, in many respects, magnesium production is similar to the aluminum production. It is capital- and energy-intensive.

189

Interfacial phenomena in aluminum electrolysis T. Utigard*, S. Rolsetht, J. Thonstadtt and J.M. Toguri* ^Department of Metallurgy and Materials Science, University of Toronto, Toronto, Ontario, Canada, M5S 1A4 1SINTEF, Division of Metallurgy, N-7034 Trondheim, Norway ^Laboratories of Industrial Electrochemistry, Norwegian Institute of Technology, N-7034 Trondheim, Norway

ABSTRACT

An X-ray radiographic technique has been employed to study the behaviour of the three phases: anode gas, bath, and aluminium, in a laboratory aluminium electrolysis cell. Experiments in which small particles (1-2 mm) of Ti32 were placed at the bath - metal interface show that the particles can move along the interface when current is applied. The direction of the particle movement is reversed on current reversal. This shows that the Marangoni effect can occur during electrolysis, i.e. the movement is caused by interfacial tension gra-dients due to a non-uniform current density at the aluminium surface.

KEYWORDS

Aluminium electrolysis; X-ray radiography; Marangoni effect; interfacial tension; mass transfer; current efficiency.

INTRODUCTION

It is now generally accepted that the loss in current efficiency is mainly controlled by the rate of dissolution of aluminium into the bath phase. Thus all factors which affect the bath-metal boundary will have an influence on the current efficiency. In recent years many improvements in current efficiency have been achieved through appropriate efforts to reduce disturbances at the bath - metal interface. Most works on improving the current efficiency have concentrated on the various factors that cause convection in the bulk of the bath and metal phases. However, among other possible effects are interfacial tension gradients at the bath-metal interface which will establish a flow directly at the interface. This surface phenomenon, the so called Marangoni effect, is well documented (Ludviksen, 1971) and it plays an important role in many metallur-gical processes where interfacial reactions occur (Richardson, 1982; Berg 1982; Belton 1982; Elliot 1985, Mukai, Toguri and Yoshitomi 1986). The motion set up at the interface when interfacial tension gradients occur is always in the direction from regions of low to high surface tension.

Richardson (1982) and Berg (1982) concluded that interfacial phenomena induced by surface tension differences may exert significant effects on the rate of reactions occurring across an interface. A non-reacting surface active species may tend to keep reactants which are less surface active out of the interface and so retard the reaction. Surface tension forces may resist the motion imposed by other forces, as is the case in the growth of gas bubbles in liquids and the damping of turbulent fluctuations at an interface (Szekely 1979). On the other hand, surface active solutes which enter the reaction, may speed up surface renewal and accelerate reactions causing interfacial turbulence.


In the present study, efforts were directed at observing this type of motion at the bath-metal interface in a laboratory aluminium cell by use of an X-ray radiographic technique.

EXPERIMENTAL

Experiments were carried out in graphite crucibles with inner diameters of 20 or 38 mm. The lower part of the crucibles was lined inside with sleeves of either sintercorundum (Al203) or hot pressed boron nitride (BN) to prevent passage of current to the side of the crucible. In addition to cryolite, the bath contained 5 mass-% CaF2, 8 mass-% A1F3 and 3 mass-% AI2O3. The bath was assumed to be saturated with alumina when sintercorundum linings were used.The required amount of premelted bath and aluminium was placed in the crucible before the crucible was placed inside the furnace. In some cases, small particles (1-3 mm) of TiB2 were added on top of the premelted crushed bath prior to placement in the furnace, but more often the particles were added after melting had occurred. The purpose of the addition of TiB2 was to obtain "markers" at the bath metal interface which would serve as movement indicators. The TiB2 was well suited for this purpose because it is one of the few materials inert towards aluminium and cryolite and because it gives good X-ray contrast in relation to the surroundings which are less dense. When electrolysis was conducted, one lead from the DC power supply was connected to the graphite crucible and the other to a centrally located graphite electrode. When the smaller crucible was used the central electrode consisted of a 3 mm rod, while in the case of the larger crucible a 9 mm rod with pointed tip was used.

Images of the crucible and its content were obtained on a TV monitor. The assembly is shown in Fig. 1. The cell was placed inside a gas tight vertical tube of quartz or mullite. The furnace was heated by a split tube graphite element which is easily penetrated by the X-rays. A horizontal beam of X-rays was directed through a window in the central part of the furnace. The beam is collected on an image intensifier after passage through the cell. The intensifier transforms the X-rays to a visible image which can be filmed by a TV-camera. The camera is connected to the TV-monitor and a video recorder.

X-ray source

Fig. 1. Furnace and X-ray apparatus.


ALUMINIUM-CRYOLITE SYSTEM In the aluminium-cryolite system, most workers have found that the interfacial tension increases with increasing A1F3 content in the bath (Zhemchuzhina and Belyaev, 1960; Dewing and Desclaux, 1977; Utigard and Toguri, 1985). The effect of the cryolite ratio (CR = mole ratio NaF/AlF3) is shown in Fig. 2.

Previous experiments in the same apparatus have shown that the interfacial tension is affected when the interface is polarized (Utigard and Toguri, 1986). Direct observation of the aluminium drop during electrolysis by the use of the TV monitor showed that on applying a positive current (aluminium drop cathode) the drop apex rose, indicating an increase in the interfacial tension. The opposite effect was observed when a negative current was applied. This appeared to be an instantaneous effect which indicates it is associated with charge effects at the interface.

Mole ratio NaF/AIF

Fig. 2. Interfacial tension of aluminium in cryolite melts at 1273 K. 1- Utigard and Toguri (1985), 2- Zhemchuzhina and Belyaev (1960), 3- Gerasimov and Belyaev (1958), 4- Dewing and Desclaux (1977)

In addition, there appeared to be a second, but not so rapid, effect which to some degree counteracted the charging effect. It was evident that the interfacial tension decreased with time from its initial high value immediately after a positive current was applied. Due to the scatter in the data obtained 10 -30 seconds after the current was applied , it was not possible to study this effect in more detail. However, less disturbances were obtained when the electrolysis was interrupted and the results indicated that a period of 200-300 seconds was needed to establish a new stable condition after interruption (Utigard and Toguri, 1986). This second effect can be ascribed to NaF enrichment at the cathode (Thonstad and Rolseth, 1978). When current is applied, a certain time period is needed to establish a stable concentration gradient in the boundary layer at the bath side of the bath-metal interface. The gradient will depend upon the convection in the bath and current density at the aluminium cathode surface. The CR at the interface will increase with increasing current density. Thus if the current density varies over the liquid metal surface an interfacial tension gradient will be set up giving rise to the Marangoni effect. As mentioned previously, the motion of the flow is from low to high interfacial tension. Figure 2 indicates that the direction of the flow will be from a region of high current density to a region of low current density.


RESULTS Direct observation of the motion of the TiB2 succeeded only in the small graphite crucible. In the larger crucible the particles became more difficult to observe because of loss in contrast. The X-rays had to pass through a larger mass of the surrounding materials in addition to thicker walls of the mullite reaction tube compared to a more thin-walled quartz tube used with the smaller crucibles. This was a drawback because the narrow dimensions of the smaller crucible restrict the possibility of obtaining an uneven current distribution along the aluminium surface. Another disadvantage was the curvature of the bath-metal interface. In the smaller crucible, it was very difficult to position a TiB2 particle at the interface without it sliding down the inclined metal surface. The density of the TiB2 particles used was 4.1 g/cm

2.

Figure 3 shows the movement obtained with a 2-3 mm TiB2 particle placed at the bath metal interface. Before current was applied the particle rested in position A. Within seconds after a current of 3.1 A was applied (graphite rod anode), the particle began moving towards the wall of the crucible. It travelled the distance from A to B in approximately 1 minute. However, the first half of the distance was covered within 20 s, which represents an average speed of «7 mm/minute. The arrest of the particle in position B resulted because it touched the wall of the crucible. It moved in a plane that was not normal to the X-rays. The electrolysis was interrupted after approximately two minutes and the cell was disconnected for another five minutes before electrolysis was resumed with reversed polarity (aluminium positive). Movement of the particle could first be detected after 12 seconds. It started to move towards position C. This position was reached 28 seconds after the current was applied, which leads to an average speed of «7.5 mm/min. The particle stayed in position C for approximately another 10 seconds before it suddenly slid back along the interface to B, where it remained attached to the crucible wall. The movement from position C to B was rapid (« 1 s). The reason for the sudden drop to position B is not clear, but it might have been caused by some minor vibrations in the furnace. This sensitivity to vibrations made it impossible to turn the crucible in the furnace to obtain a 90° cross

Fig. 3. Movement of a TiB2 particle at the bath-metal interface when a current of 3.1 A was applied. Particle moved from A to B when normal polarity was applied (aluminium negative), and from B to C when the current was reversed.

PRODUCTION AND ELECTROLYSIS OF LIGHT METALS 193 reference in order to determine in which plane the movement had occurred. Observations were also made when particles were dropped into the bath. In some cases they would float or move around on top of the bath surface for a few seconds before they dropped to the interface. Here they would either slide down the inclined interface or "bounce" back to the bath surface and then immediately sink and slide down the interface. In some cases when larger particles (~ 3mm) were involved, they remained approximately half a minute at the interface before they rose rather slowly (1-2 seconds) to the bath surface. Immediately after the particle reached the surface it dropped rapidly (<0.5 s) back to the bath-metal interface. The reason for this "bouncing" behaviour of particles is not clear. The particles were not heated or dried prior to addition, and thus the expansion of gas or vapour captured in the pores of the TiB2 particle may explain the rise of the particle immediately after addition to the melt, but this is not likely the case for the slow rise of particles which remained more than 30 s at the interface.

As mentioned above, it was difficult to trace the movement of the small particles when the larger crucible was used because the contrast became diffuse. During long electrolysis experiments, it was evident that some particles disappeared and other appeared in new positions. It was not possible to trace the particle route, however. It was also difficult to determine whether the particles rested on the interface or if they adhered to the wall of the crucible (the BN lining). The experiments performed in the larger crucible was therefore not very helpful in determining interfacial tension driven flow.

DISCUSSION

Laboratory experiments

The experiments in the smaller crucible indicate that interfacial tension driven flow can occur during aluminium electrolysis. The observed direction of the movement of the TiB2 can be predicted from Fig. 2. The movement is from a region of high to low current density (cd) when "normal" electrolysis is carried out. The boundary layer in the high current density region becomes more enriched with NaF than the region with low current density due to the high transference number of sodium ions in cryolitic melts. It might be argued that the convection created by the gas evolved at the anode can generate some disturbances at the bath metal interface, causing the TiB2 to move away from the region close to the anode, but the fact remains that the movement reverses when the current is reversed. Under this condition the high current density region close to the graphite electrode will be more depleted in sodium ions and more enriched with aluminium ions than the low current density region closer to the wall of the crucible. The cryolite ratio at the interface increases as one moves from the center towards the wall. This will create an interfacial tension gradient which will drive the interface from the wall to the centre and causing the TiB2 particle to move up the curved interface.

The observed movement was rather slow, but this can be ascribed to several factors. In a small crucible it is difficult to maintain an uneven current distribution because the size of the anode gas bubbles sets a limit to how close the anode can be positioned to the bath-metal interface. Due to the small dimension, the thickness of the boundary layer will probably be very thin compared to the dimension of the particle. The overvoltage effect and the interfacial driven flow by itself will try to even out the variation in the cryolite ratio over the interfacial area. The fact that the particle itself is an electronic conductor and probably acts as an active part of the aluminium electrode is also a concern. It therefore seems fruitless to try a more quantitative approach from these laboratory findings other than to establish that interfacial driven flow may occur during electrolysis in aluminium


laboratory cells. However, in industrial cell where the dimensions are more than two orders of magnitude larger, the interfacial effect and flow will be magnified, and some estimates can be made on how the Marangoni effect might affect the performance of the cell. In the following, a simple approach is attempted in order to determine if the Marangoni effect can significantly affect the current efficiency.

Effect on current efficiency in industrial cells

Figure 4 shows the current distribution obtained when the metal pad extends beyond the projection of the anode. It is generally accepted that this is detrimental to the performance of the cell for several reasons (Tabereaux and Gagnon, 1988), but the situation is not uncommon in practice because of the difficulties involved in maintaining a stable side ledge. The current distribu-tion in the bath was obtained by mapping the field on a conducting paper. This method is based on the fact that the current and potential lines are determined by the geometrical form only, and the fields in a cross section of the bath can be simulated on the paper when the geometrical proportions are similar. The actual current distribution will be a little less biased because the anodic and cathodic overvoltage tend to even out the distribution, so this can be regarded as a worst case "scenario" for the current distribution. The figure shows large differences in the current density in the bath under the anode and in the outer channel.

Fig. 4. Current distribution in the outer channel of an aluminium cell determined by using conducting paper. Overvoltage effects are neglected. Anode-cathode distance = 5cm and the metal pad extends approximately 25 cm out of the projected area of the anode.

From Fig. 4 the current density at the aluminium surface can be determined. The result is shown in Fig. 5 where the current density (cd) is plotted as a function of the distance from the projection of the anode edge. It is assumed that the current density beneath the central part of the anode is 1 A/cm2.

Figure 5 shows that the cd varies from «0.15 to 1 A/cm2 over a distance of about 30 cm. According to data reported by Haupin and Frank (1981), and Haupin (1987), for industrial cells, the overvoltage corresponding to cds of 0.15 and 1 A/cm2

will be 0.01 V and 0.1 V. Because the overvoltage is caused by concentration gradients in the boundary layer at the bath metal interface (Thonstad and Rolseth, 1978), the composition of the bath at the interface can be estimated using data for the emf of the aluminium electrode in cryolitic melts.


0 10 20

Distance (cm)

30

Fig. 5. Calculated current density at the bath-metal interface plotted as function of distance from the projection from the anode edge. The current density is determined from Fig.4 assuming 1 A/cm2 under the central part of the anode.

From literature data (Yoshida and Dewing, 1972; Thonstad and Rolseth, 1978; Sterten and coworkers, 1982), it can be estimated that for an industrial bath where the bulk composition is CR=2.5, the composition at the interface would be CR=2.55 and CR=3.2 for the corresponding overvoltages. Figure 2 can then be used to estimate the interfacial tension for the regions where the cd is 0.15 A/cm2 and 1 A/cm2. By using curve 1, a difference in the interfacial tension of about 80 mN/m is found. A reservation may be made about using the data in Fig. 2 because they were obtained without the passage of any current through the interface. With no other data available, it is assumed that the charging effect is constant and independent of the bath composition.

The velocities created by the surface stress at the interface can be estimated by an approximate solution of the Navier Stokes equation (Utigard, 1987). The boundary layer thickness, b, becomes

b = 0.78 (i/L/u)*i (1)

where v is the kinematic viscosity of the bath, L the distance over which the interfacial tension gradient occurs and u the horizontal velocity component. The maximum horizontal velocity, v^^, is given by

âx = (7l-72)/[L-(Âl/bAl + / V ^ B ) ] (2)

where (7i~72) interfacial tension difference between the two locations and μ is the viscosity of the liquids. The subscript Al and B represents aluminium and bath.

From eqs. (1) and (2) the interfacial velocity becomes:

u ^ - [0.78( 7 I-72)/((MAIPAID% + (/«B/>BL)

%)]2/3 (3)

In the calculation of v^^ the following physical constants were used: μΒ=2.5·10"

3 kg/ms, μΑ1=1.18Ί0"3 kg/ms, pB=2050 kg/m

3, />A1=2280 kg/m3, (7l-72)=80

mN/m and L=0.3 m, which give 10,^=0.095 m/s. Inserting this velocity into eq. (1) the thickness of the bath's boundary layer becomes b=0.0015 m at L=0.3 m.


The movement of the interface will be directed out from the projection of the anode towards the side of the cell. This leads to a continuous transport of dissolved metal out from the central part of the cell because the boundary layer will be more enriched with metal than the bulk of the bath. In order to estimate this transport of dissolved metal it is assumed that the mean velocity of the boundary layer is H u ^ , and that the boundary layer acts as 'a moving film with this velocity, Ηιι χ. In the region where the movement starts, a renewal of the boundary layer will occur bringing electrolyte with bulk concentration of dissolved metal in contact with the interface. By using the moving film concept the dissolution of dissolved metal can be treated as a non-stationary diffusion into the film assuming a time of contact between the film and liquid metal of L/HUjnax. This is a simplification because the horizontal velocity gradient in the boundary layer also affects the vertical transport of dissolved metal. However, the present calculations are intended to determine only the relative importance of the Marangoni effect.

The Orstrand - Dewey solution was used for the calculation of the concentration gradient in the moving film. The concentration, c, of dissolved metal at a distance x and time t of contact between the film and the metal surface is given by

c = cb+(c*-cb)-erfc[x/(27(Dt))] (4)

In eq.(4) cb is the bulk concentration and c* is the saturation concentration of dissolved metal in the bath, D is the diffusion coefficient of the dissolved metallic species and erfc is the error function complement. The calculations were performed using a spreadsheet (Quattro) whereby the erfc was fitted to a fifth degree polynomial using the tabulated data. Some results from the spreadsheet calculation are shown in Fig. 6 where the concentrations of dissolved metal are plotted as a function of the distance from the interface for various times of contact between the moving film and interface. The following data were used in the calculations: Saturation concentration of dissolved metal, c*=38 mol/m3 (0degard, 1988); concentration in bulk of the bath 10% of saturation (Rolseth, Thonstad, 1981), cb=3.8 mol/m

3; diffusion coefficient of dissolved metal (Dewing and Yoshida, 1976), D=310~8 m2/s.

0l 1 1 1 1 1 0 0.001 0.002 0.003 0.004 0.005

Distance from bath-metal interface (m)

Fig. 6. Calculated concentration of dissolved metal in a moving film of bath as function of distance from the bath metal interface at various times of contact between the moving film and interface.


The amount of metal which is dissolved during the time of contact is plotted as a function of time in Fig.7. The results were obtained by numerical integration of the data shown in Fig. 6.

Based upon the previously made assumption that the time of contact between the moving film and the metal will be L/ih'v^^) , a time of t = 6.3 s is obtained. Figure 6 shows that for this time of contact the concentration profile will be within the boundary layer thickness of 1.5 mm as calculated from eq. (1). From Fig. 7 it is found that the excess amount of metal transported out by this film will be 0.016 mol dissolved AI per m of anode periphery. Assuming an anode

0. 06

0.04

0.02

0. 00

20 40 60

Time/sec

80

Fig. 7. Amount of dissolved aluminium diffused into a moving film of bath as a function of the time of contact between the film and the bath-metal interface.

periphery of 2-(7.5+3.5) m and an average flow rate of 0.047 m/s the metal loss due to this transport becomes:

M 0.0160.047-22 = 0.0165 mol Al/s = 0.447 g Al/s

Compared with the theoretical production rate from a 150 kA cell of 14 g Al/s, this accounts for a loss in current efficiency of 3.2 %. The value of the diffusion coefficient of dissolved aluminium used in these calculations may be

questioned, and if it is reduced by one order of magnitude to D=3*10~9 m 2/s, the

calculated CE loss is 1.6 %. This is still a significant part of the overall loss in a Hall-Heroult cell.

It may appear that the calculated loss is rather high, and it must be stressed that the calculations are based on the worst case of cathode current distribution. The anodic and cathodic overvoltages will tend to even out the current distribution. In addition, the mass flow created by the Marangoni effect will partially even out the difference in cryolite ratio of regions with different current density thus reducing the driving force for the interfacial flow. However, these calculations show that the Marangoni effect cannot be ruled out as one of the sources of the loss in current efficiency in industrial cells. Further they indicate that the Marangoni effect is an additional reason why a configuration where the bath-metal interface extends beyond the projection of

198 PRODUCTION AND ELECTROLYSIS OF LIGHT METALS the anode should be avoided. Such a configuration is detrimental to the CE also because of the excessive aluminium surface area established in a region where both magnetically and gas driven flow velocities are high (Solheim et al., 1989). One should also be aware of the uneven current distribution which results when prebaked anodes are replaced. After installation a new anode will be more or less covered by layer of frozen bath which will reduce the current through the anode. It might be argued that the interfacial flow generated here will be less detrimental to the CE because the flow will be directed to a region under the anode where the rate of gas evolution is reduced, giving rise to less turbulence and reoxidation between gas and dissolved metal than in the outer channel. However, one should keep in mind that all metal that dissolves is lost, no matter what happens after the dissolution. More work is needed, especially if one wants to incorporate the Marangoni effect into mathematical models fo^ the loss in CE during aluminium electrolysis.

ACKNOWLEDGEMENTS.

This work was performed during one of the authors (S. Rolseth) stay at the University of Toronto. He gratefully acknowledges financial support from the Royal Norwegian Council for Scientific and Industrial Research. The authors are also grateful to NSERC for financial support.

REFERENCES

Belton, G.R. (1982). The interplay between strong adsorption of solutes and interfacial kinetics at the liquid metal surface. Can. Met. Quart., 21. 137-143.

Berg, J. (1982). Interfacial hydrodynamics: An overview. Can. Met. Quart., 21. 121-136.

Elliot, J.F. (1985). The role of interfaces in pyrometallurgical processes. Trans. Inst. Min. Metall, Sect. C., 94, 171-178.

Dewing, E.W.,and K. Yoshida (1976). Electronic conductivity in cryolite -alumina melts? Can. Met. Quart., 15^ 299.

Dewing, E.W.,and P. Desclaux (1977). The interfacial tension between aluminium and cryolite melts saturated with alumina. Met. Trans. B., 8B, 555- 561.

Gerasimov, A.D.,and A.I. Belyaev (1958). Investigation of the interfacial tension at the boundary of metal and electrolyte in the electrolytic production and refining of Al. Izv. Vyssh. Uchebn. Zaved, Tsvet. Met., 1, No.5, p.58.

Haupin, W.E., and W.B. Frank (1981). In J.OM. Bockris, B.E. Conway, E. Yeager and White (Ed), Comprehensive Treatise of Electrochemistry, Vol. 2, Plenum Publishing Corporation, New York. pp. 301-325.

Haupin, W.E. (1987). Electrode reactions in Hall-Heroult cells. In A.R. Burkin (Ed.), Production of Aluminium and Alumina. Critical Reports on Applied Chemistry, Vol. 20, John Wiley & Sons, New York, pp.120-133.

Ludviksson, V., and E. N. Lightfoot (1979). The dynamics of thin liquid film in the presence of surface tension gradients. AIChE, Γ7, 1166-1173.


Mukai, K., J.M. Toguri, and J. Yoshitomi (1986). Corrosion of alumina-graphite refractories at the slag-metal interface. Can. Met. Quart., Z5, 265-275.

0degard, R. (1988). On the solubility of aluminium in cryolitic melts. Met. Trans., 19B, 449-457.

Richardson, F. D. (1982). Interfacial phenomena and metallurgical processes. Can. Met. Quart., 21, 111-119.

Rolseth, S., and J. Thonstad (1981). On the mechanism of the reoxidation reaction in aluminium electrolysis. Proceedings of the 110th AIME Annual Meeting, Chicago, Feb. 22-26. Light Metals 1981, pp. 289-301.

Sterten, Ä., A .Maland, and K. Hamberg (1982). Activities and phasediagram data of NaF-AlF3-Al203 mixtures derived from electromotive force measurements. Acta Chem. Scand, A 36. 329-344.

Solheim, A.S, S.T. Johansen, S. Rolseth, and J. Thonstad (1989). Gas driven flow in Hall-Heroult cells. Proceedings of the 118 TMS Annual Meeting, Las Vegas, Nevada, Feb. 27-March 3, Light Metals 1989, pp. 245-252.

Szekely, J. (1979). Surface tension effects in dynamic systems. In J. Szekely, Fluid flow phenomena in metals processing. Academic Press, New York. pp.238-253.

Tabereaux, A.T., and H. Gagnon (1988). Control of the side ledge freeze in vs Soderberg cells. Proceedings of the International Symposium on Reduction and Casting of Aluminum, 27th Annual Conference of Metallurgists, Montreal, August 28-31, pp.219-227.

Thonstad, J., and S. Rolseth (1978). On the cathodic overvoltage in cryolite -alumina melts - I. Electrochim Acta, 23, 223-231.

Thonstad, J., and S. Rolseth (1978). On the cathodic overvoltage in NaF-AlF3-A1203 melts - II. Electrochim Acta, 23, 233-241.

Utigard, T., and J.M. Toguri (1985). Interfacial tension of aluminum in cryolite melts. Met. Trans., 16B, 333-338.

Utigard, T., and J.M. Toguri (1986). Electrocapillarity in the aluminum reduction cell. Met. Trans., 17B, 547-552.

Utigard, T. (1987). Mass transfer in Hall-Heroult electrolysis induced by interfacial tension gradients. Aluminium, 63, 608-613.

Yoshida, K., and E.W. Dewing (1972). Activities in NaF-AlF3 melts saturated with A1203. Met. Trans., 3, 683.

Zhemchuzhina, E.A., and A.I. Belyaev (1960). Interfacial tension at the boundary between liquid Al and molten salts. Fiz. Kim. Rasplav. Solei i Shlakov, Akad Nauk SSSR, Uralsk Filial, Inst. Elextrochim., Tr. Vses. Soveshch., Sverdlovsk, 1960, (Publ. 1962), pp. 207-214.

201

Industrial plant for the production of electrolytic titanium— Ginatta technology M.V. Ginnata, G. Orsello, R. Berruti, M. Gaido and E. Serra Ginatta Torno Titanium, C.P. 543 - 10100 Torino-Centro, Italy

ABSTRACT

Due to its excellent many properties titanium could be more widely utilized, but traditional thermochemical processes for its production are obsolete technologies which in turn generate high production costs. The innovative Ginatta electrolytic process is a valid solution for resolving this problem. This paper describes the plant.

KEYWORDS

Titanium, electrowinning, molten salts, titanium tetrachloride, automation.

SUMMARY

Electrowon titanium has reached industrial commercialization. In this paper we review the development stages which led to the construction of the electrowinning plant. Our first electrolytic industrial pilot plant ("Modex I") was built in 1980. It was followed by a second plant (Modex II) in 1983, and in 1986 we constructed the plant now in operation (Modex III), which has a nominal capacity of 70 tons of titanium per year. This year (1989) Modex IV has been started up in RMI plant, Ashtabula (Ohio). The core of the plant is its extraction module ("Modex") which comprises a chamber and a pre-chamber with controlled atmospheres and ancillary equipment. The interior of the chamber is horizontally divided by removable covers into two parts. The lower part contains the electrolytic cells operating at temperatures up to 950°C and with current intensities reaching 50,000 A. In the upper part, operating at temperatures lower than 100°C, a hydraulic manipulator handles the electrodes and allows continuous production. The overall operation of the plant has a simplicity comparable to that of aqueous solution tankhouses. In comparison with other processes and plants for the production of titanium, the metal produced by a Ginatta plant has the advantages of lower costs and higher quality. Costs are lower because of: a) lower overall energy consumption; b) lower labour requirements due to the continuous character of the process and its high degree of automation; c) high rate of throughput; d) lower capital costs.

INTRODUCTION

The literature on electrolytic cells for the production of titanium from molten salts is quite extensive.


Although some of the associated developments reached the pilot plant stage, the lack of specifically designed hardware did not allow their full exploitation on a commercial scale. Important examples of this situation are provided by the activities of the U.S. Bureau of Mines (1-2), and of Companies such as New Jersey Zinc (3), Timet (4), Cezus (5) and D-H Titanium (6). The Ginatta electrolytic plant was specifically designed and constructed for titanium. Our development work started from experimental studies (7-9) in prototype cells. The results confirmed that it is difficult to produce reactive metals electrolytically in small closed cells with production cycles long enough to be of industrial interest. Clearly, too many conflicting tasks and functions were demanded of the few general-purpose components. The main recurring problems were:

mechanical strength of the equipment at working temperature; corrosion of materials; handling of the cathodes for the continuous operation of the process; accurate data logging for all process parameters.

In this paper we describe and illustrate the hardware which allows an easy operation of the process.

DESCRIPTION OF THE PROCESS

The raw material fed into the electrolytic plant is titanium tetrachloride. It dissolves in the electrolyte in the Dissolution Cell (10) according to the reaction :

TiCl, -> TiCln + Cl„ 4 2 2

The electrolytic titanium is deposited on cathodes in the Extraction Cell (11) according to the reaction:

TiCl. -> Ti + Cl2

The Dissolution Cell is separated from the Extraction Cell. Their common electrolyte (Sodium-Titanium-Chloride) circulates in closed circuit. The cells have Heterogeneous Bipolar Electrodes, generating a high titanium tetrachloride dissolution rate in the electrolyte and a low average valence of the titanium species dissolved. When a steady state is reached, they maintain very low Titanium chloride concentrations in the anode compartment (anolyte). The operating temperature (830°C) results in:

Low drag out. High current density. High titanium concentration in the electrolyte.

The electrolyte is inexpensive (NaCl technical grade) and easy to handle.

DESCRIPTION OF THE PLANT

The present design of our electrolytic plant (12) has enabled us to achieve the ease of operation of an aqueous solution tankhouse. The plant required had to have high versatility so electrochemical measurements could be made and samples obtained under reproducible and steady state conditions of real industrial operation. Our Modex III plant has the flexibility required for long production runs and for the rapid change of many key parameters: cell configuration, type of electrode, electrolyte composition operating temperature, pressure and composition of the gas atmosphere, current density and voltage. The plant is the result of a design integrating many components, each one specializing in a specific function.


The main tasks of the Modex plant are: providing an inert atmosphere in the electrolytic cell; melting the electrolyte and keeping the electrolytic cell at the operating temperature; allowing energy and mass transfer between the electrolytic process and the external environment; controlling the process.

The Modex plant comprises: the external shell, formed by a chamber and a pre-chamber; the electrolytic cell, inside the chamber; the removable covers of the cell; a structure for supporting the electrodes and feeding electric current to them; the electrodes; the hydraulic manipulator, for moving the electrodes as well as maintenance and ancillary operations.

The Shell

The shell provides a protected environment in which the titanium electrolytic process can be operated in open cells. The pre-chamber has the purpose of permitting transfer of material from the Modex to the exterior under a controlled atmosphere. Windows allow observation inside the chamber and into the cells. Consequently, the electrodes can be photographed during operation, and reference and standard electrodes can be exactly positioned for accurate measurements.

The Cell

Unlike traditional designs, the Cell here has only one function, which is to contain the molten electrolyte since gas tightness is assured by the shell. This results in two very important operating advantages; the process can be run:

at higher temperatures, and under negative pressure.

The Cell structure has been entirely built of carbon steel, the latter being compatible with the electrolytes of titanium production. The structural weakness of steel at the operating temperature has been overcome by refractories supporting the outside of the Cell. The Cell is rectangular, a geometry typical of aqueous electrolytic plants (such as Pb, Zn, Cu...) and placed inside an electric furnace. To avoid corrosion and impurities, heaters and refractories are not in contact with the gases generated by the electrolysis. On start-up of the plant, the furnace melts the electrolyte. The current for electrolysis keeps the Cell at the operating temperature through the joule effect, but the furnace allows tests with other working temperatures to be carried out if wished.

The Electrodes

The assembly of the electrodes is such that each one has an independent electric control and can be easily replaced. Repeated collection of the cathode deposits allows the production to be continuous. In the Modex III Extraction Cell there are six cathodes, each one having a total immersed surface of two square meters. Good electric contact is provided by the weight of the electrode fitting onto the power supply bar; the shape of the contact ensures its cleanliness and a negligibly

204 PRODUCTION AND ELECTROLYSIS OF LIGHT METALS small junction voltage drop. The bearing bars are fed by high intensity-low voltage electric feedthroughs, across the shell. Busbars connect the feedthroughs to rectifiers.

Ancillary Equipment

All power-mechanisms of our Modex III use proportional hydraulics, that have proved to be very reliable. The main movements are associated with the two pre-chamber ports, for introduction and removal of electrodes, the removable covers of the cell which thermally insulate the upper zone of the chamber, the manipulator which handles the electrodes and performs various maintenance tasks inside the module. The inert atmosphere in the Modex is created by producing a vacuum (by means of pumps) during the plant start-up stage, and then by filling it with argon. The anodic gas is continuously recovered with a chlorine pump. The rectifiers can be current or voltage controlled; reference electrodes can be used to control the power supplied. TiCl, feed is introduced either by argon gas pressure, by metering pump or by negative pressure intake. Various alloy thermocouples measure the temperature at several strategic points of electrolysis, while linear piezo-resistive transducers monitor the pressure. Logging and control equipment (PC and PLC) are located in a Control Room.

Materials

The Modex plant has been designed with the goal of cost effectiveness; consequently, low cost materials have been used. Low carbon steel has been selected for the equipment in contact with the electrolyte or with cell atmosphere: since iron reacts with the electrolyte and forms a highly stable and protective intermetallic compound. This reaction is accelerated by means of a pre-electrolysis period , at low current density in which the steel of the cell as a cathode operates. The steel is protected from anodic gas corrosion, at operating temperatures, because of the formation of a compact, high-melting compound (of the type Fe-Ti-O-Cl), which adheres to the metal and is generated by the reaction of the steel with the atmosphere of the cell at the start-up of operation. Low-cost refractories have been used, since they are not in contact with either the electrolyte or the cell atmosphere. The electrical insulators of the electrodes feedthroughs are the only high quality materials.

OPERATION OF THE PLANT

Continuous steady state production is obtained by supplying TiCl, to the Dissolution Cell housed in the shell. The electrodes of this Cell are supplied with direct current from a specific section of the rectifier. The electrolyte is composed of a mixture of sodium and titanium chlorides at an operating temperature of about 830°C. This temperature is maintained by the Joule effect of electrolytic current. The titanium in solution is then deposited on the cathodes of the Electrowinning Cell, while chlorine gas is simultaneously released at the graphite anodes. The electrodes of the Cell are supplied with direct current from a section of the rectifier which is independent of that for the Dissolution. The electrolyte, containing a high concentration of titanium in solution progresses from the Dissolution Cell to the Extraction Compartment through convection


movements in the electrolyte. Samples of the electrolyte are periodically taken on a scheduled program and sent to the analytical laboratory in order to determine the concentration of titanium and its average valence state. The chlorine produced is pumped to a plant for its recovery. When the titanium metal deposited on a cathode has reached a predetermined mass, this cathode is removed from the bath by the manipulator. The "mature" cathodes are individually taken to the stripping machine, in order to harvest the product, and then immediately repositioned in the Extraction Cell to continue the electrowinning process. To remove the harvested titanium, the pre-chamber is connected with the chamber by opening the separating door. Before starting the operation an inert atmosphere in the pre-chamber is provided at the same pressure as that of the chamber. The product is loaded in the crusher, and then treated in the leaching plant. The titanium crystals are dried at low temperature and packed under argon.

CONCLUSION

The operating experience we have gained through the Modex III plant allows us to conclude that:

the positioning and handling of electrodes is very efficient; the equipment is reliable. Present hydraulic components ensure a very low probability of failure; furthermore maintenance does not interfere with production; the molten-salt electrolytic cell can be operated with the same ease as that of an aqueous solution electrolytic tankhouse. It is possible to raise the electrodes, visually examine the deposit, take truly representative samples, without affecting the electrolytic system, and immerse them again; the inert gas volume above the cell has seal surfaces which are at room temperature; energy losses associated with the electrolytic process (e.g. ohmic potential drops and heat losses) or with ancillary equipment (e.g. manipulator) are very low. Consequently, the overall energy consumption is also very low;

- i the design of the plant permits the operation to be carried out at high temperature, thus allowing:

the use of pure and inexpensive NaCl as electrolyte; high density currents with reduced voltages; low metal levels in drag-out salt;

the design of the plant also allows a high level of automation. The process and equipment yields an excellent quality of titanium metal. Typically, the impurities are in the following range: 0, 200 to AOOppm; N, 30 to 50ppm; H, 200ppm; C, 50 ppm; Cl, 200 to 400ppm; Fe, 50ppm. In comparison with the titanium produced by thermochemical process (Kroll or Hunter) plants, only the core of the cake attains such a high quality. Our work has demonstrated that this new design of electrolytic plant is cost effective on an industrial scale (13), because of significantly lower capital and operating costs. We have installed at RMI Co. in U.S.A. a larger plant (Modex IV) with a rated capacity of 150 tons of titanium sponge per year (13).

REFERENCES

(1) Haver F.P., and Baker Jr. D.H. (1961). U.S. Bureau of Mines R.I., 5805. (2) Leone 0.Q.,and Couch D.E. (1972). U.S. Bureau of Mines'R.I.y 7648. (3) Myhren A.J. (1968), J. of Metals. (4) Priscu J.C. (1968). TMS-AIME Extractive Metallurgy Symposium, Cleveland, Ohio


(5) Champin B., Graff, and Molinier R. (1980). Memoires et Etudes Scientifiques, Revue Metallurgie, Mai , p. 681.

(6) CobeJ G., Fisher J., and Snyder L.E., Titanium *80 Science and Technology, p. 1969.

(7) Ginatta M.V., (1970). Master Thesis T 1342, Colorado School of Mines, Golden, Colorado.

(8) Ginatta M.V. (1970). TMS-AIME Annual Meeting, J. of Metals, December, p. 22 A. (9) Ginatta M.V. (1972). Electrochemical Society Spring Meeting, Houston, Texas,

May , Abstract 192. (10) Ginatta M.V. (1983). U.S. Patent 4.400.247, Aug. 23. (11) Ginatta M.V., Orsello~G., "and Berruti R., (1988). Italian Patent Application

67.364-A/88, April 19. (12) Ginatta"ΜΪν'. , and Orsello G., (1987). U.S...J at nt_4_j57Q_.i21, June 2. (13) Ginatta U., (1987). Metal Bulletin Conference, Los Angeles, March 25-27. (14) Metal Bulletin, 25 February"T988"

Fig. 1. Electrolytic titanium plant (Modex III) at Santena (Torino, Italy) facility, G.T.T. S.A.


Fig. 2. Electrolytic titanium cathode pulled up trom moJten sait bath, two square meters suriace

Fig. 3. The Control Room


Fig. 4. Electrolytic titanium sponge

Fig. 5. The hydraulic manipulator which handles the electrodes inside the Modex

209

Lithium: the lightweight champion Richard W. Barnes Lithco, Lithium Corporation of America, 449 North Cox Road, Gastonia, North Carolina 28054, U.S.A.

INTRODUCTION

Since the late 1960's, the metallurgists' primary exposure to the element Lithium has been it's addition to aluminum reduction pots and, more recently, the promotion of aluminum-lithium alloys for aerospace applications. The object of this presentation is to broaden that perception.

Much has been written about the sourcing and production of lithium ores and chemicals. This paper, while reviewing that body of information will address the unique properties of lithium and how they relate to the diverse uses of the metal and it's compounds.

THE LITHIUM BUSINESS

Lithium is in round numbers, $200 million business, worldwide—relatively small as compared with others in the chemical industry. The element is marketed as one of its ores, the metal itself, or one of the many lithium derivatives. Because it is offered in such varied forms, the producers like to talk about volume in terms of pounds of lithium carbonate equivalent (CE). In 1988 the world market was between 72 and 7^ million pounds of CE for ores and chemicals.

Ores The lithium ores sold on the merchant market are all alumino-silicates mined from pegmatites. Spodumene (LiAlSipO^) and petalite (LiAlSiK0in) satisfy essentially

all of the world ore demand. This represents approximately 1/6 of the total lithium demand in CE. A large majority of lithium used in the manufacture of glass is consumed as lithium ore.

Chemicals & Metal

The remaining 5/6 of the lithium markets are in the various chemical forms. The initial product is lithium carbonate, which is sold as produced or upgraded further into the hydroxide, chloride, etc. In 1988 the world lithium chemical business totaled less than $200 million in revenues.

Less Than

LITHIUM CARBONATE and BY-PRODUCTS $70 million INORGANICS 70 million ORGANICS l\0 million BATTERY PRODUCTS 20 million

210 PRODUCTION AND ELECTROLYSIS OF LIGHT METALS Supply/Demand

Manufacturing lithium for the world marketplace is shared by several producers. Those which make a significant impact are shown in the exhibits below. It is worth noting that two of the major suppliers are not basic in a resource, but are upgraders of purchased lithium carbonate.

TABLE 1 Producers, Ores/Chemicals

ORES LOCATION

Tanco Canada Lithium Australia Ltd. Australia Bikita Minerals Zimbabwe

CHEMICALS

Lithco (Subsy. of FMC Corp.) USA, UK Cyprus/Foote-SCL USA, Chile *Chemetall (Subsy. of Metalgesellshaft) W. Germany *Alco (Lithco/Honjo Joint Venture) Japan Brazilero de Litio Brazil USSR government Germany PRC government China

* Not basic in resource

Figure 1 shows the supply distribution of lithium ores and chemicals by major producer and the free world demand. Lithco and Cyprus/Foote have an approximately equal supply share. Over 3/^ of these products are used in North America, Europe, and Asia.

Figure 2 illustrates the distribution of demand for lithium products and their revenues. It is interesting to note from this figure the value added products constitute a majority of total lithium revenue.

GEOGRAPHIC SOURCES

Throughout the world there are many sources of lithium. The element is found in pegmatites, clays, underground brines, salars or, even sea-water. What limits its availability and adds to the cost of production are the relatively few deposits which combine the necessary ingredients for exploitation: size of resource; grade; ease of extraction; accessibility and/or proximity to markets. Those resources which currently impact the lithium industry can be seen in figure 3·

PRODUCTION

Ore Benefication

Production methods for ores and chemicals vary widely with the type of resource. Some ore is produced and marketed "as mined." Most applications require further upgrading both for enhancing lithium concentration and for removal of impurities. Figure 4 outlines a typical ore beneficiation circuit for spodumene.


Figure 1

Figure 2


• Current Ore Operation

O Current Brine Operation

Figure 3

Mining Operation

[örg

Crushing Circui t

·* Grinding Circui t

Spodumene Flotat ion

[Mixed Ore|

Ore Sorter

By Product Flotation

Spodumene By Products R u n - o f -Mine Ore

Figure 4

212

PRODUCTION AND ELECTROLYSIS OF LIGHT METALS 213 Sulfuric Acid Extraction of Ore

There are two economic methods known for the production of lithium carbonate from whence all subsequent compounds are derived. Figure 5 presents a flow diagram detail of a sulfuric acid extraction of ore referred by some as the Ellestad process. This method involves the decrepitation (that is the changing of crystal structure of the ore by heat) of the ore concentrate at 1100 C followed by milling, then reaction with sulfuric acid for 1-2 hours at 250 C. The resulting lithium sulfate is neutralized, leached and separated from the residue by a rotary drum vacuum filter. The weak liquor is purified and concentrated before precipitation as lithium carbonate at high temperature with soda ash.

Schematic (low sheet Tor Li,CO) production from ore

Figure 5 ("Lithium and Lithium Compounds" Encyclopedia of Chemical Processing and Design 1988 pg. 332)

Extraction from Brines

The technology of separating lithium from mixed chloride brines has been known for many years. Cyprus/Foote Mineral has been producing lithium carbonate at their Silver Peak, Nevada site since 1966. Separation of lithium from the Great Salt Lake brines was developed by Lithco, but never exploited. Today, large, high quality sources of lithium are found in the desiccated lake beds (salars) located in the Andes of South America. Cyprus/Foote is producing at their site at the Salar de Atacoma near Antafagasta, Chile since 1984. All brine extraction processes, existing or proposed, involve the use of solar evaporation as an energy source for most of the evaporation. For this reason production of lithium carbonate from brines is the lowest cost process today.

Figures 6 and 7 show conceptual flow sheets for the extraction of lithium from a brine resource, purification of concentrated brine and production of lithium carbonate.


EVAPORATION EVAPORATION

7 f \BI5HOriTt7 \ STORAGE 7 — * ■- BRINE

TOCAROONATE PLANT

SOLIDS I SEPARATION |

I ^ N.CI 1 * KCI-MoCI·»· KCI · MgCI2 · 6HjO

MgCI, · 6 H ,0 H J B OJ

Conceptual process k>w sheet for solar evaporation of a lithium-rich brine.

Figure 6 (Ibid pg. 337)


Other Lithium Chemicals

A schematic product flow sheet for an integrated lithium plant is shown on figure 8. The carbonate is the starting material for other lithium chemicals now in use. The carbonate is reacted with lime to form the hydroxide, hydrochloric acid to form the chloride, hydrobromic acid to form the bromide and so forth. From the chloride, metal can be produced by electrolysis of a molten LiCl/KCl eutectic. The metal is precursor to a range of products including butyllithium, lithium hydride, lithium amide, and a limitless variety of specialty organic compounds.

PROPERTIES

At this point it should be apparent that a significant property of lithium is that it will never be inexpensive to produce. It is because of its other unique properties that the benefits of lithium outweigh its cost.

What makes lithium unique is that its properties are a study in extremes and contrasts. It is the purpose of this paper to present a discussion of those properties and how they relate to the myriad of needs this versatile element satisfies.


LITMti/4 I fcl Lt2C03 I J[ Γ J — H L-*- Η»Πϋ ^κ 1

'

'

1

Z] r

'

f

'

ILi CASTINGS 1

^| Li BATTERY |

. 1 LITHIUM 1 * | SANO/SMOT 1

1 BUTYL I ILITMIUMIS). 1

K.

Schematic (low shccl of an integrated lithium plant.·


Reactivity

Lithium's place of honor in the periodic chart almost assures it of uniqueness. Of all the solids, it has the lowest atomic number and atomic weight. But, it is also it's position on the electromotive series that is responsible for many of it industrial applications.

Single Electrode Potentials of Selected Metals

(Langes* Handbook 1973)

L i + Rb+ K+ Ca++ Ba++ Na+ Mg++ Mn++ Zn++ Fe++

Metal

> L i + e > Rb+ e

—>> K+ e > Ca+2e > B a + 2 e > Na+ e > Mg+2e > Mn+2e > Zn+2e > F e + 2 e

E ( V o l t s )

-3.04 -2.92 -2.92 -2.87 -2.9O -2 .7I -2.37 -1.18 -O.76 -OAH

Despite the willingness of lithium to give up an electron, it behaves in a similar way to aluminum, forming a protective layer of oxide at is surface which partially inhibits its reaction to atmosphere. Unlike sodium or potassium, an ingot of lithium can be held in the hand. Still, the metal is always shipped under mineral oil or dry argon since the inhibiting layer does not protect it from a specific tendency to react with nitrogen in the air-forming lithium nitride (Li^N).

Organolithium compounds are generally less reactive than the other alkali metal analogs, but more reactive than the group II A organometals.

216 PRODUCTION AND ELECTROLYSIS OF LIGHT METALS Atomic Radius

The small size of the lithium ion relative to other metals has a significant impact on the commercial attractiveness of its compounds. Reactions that would not normally be possible due to the size of the ion (steric hindrance) are completed with lithium compounds. The permeability of membranes by the lithium ion is another important characteristic.

Alkali Metals Alkaline Earth Metals

Ionic Radii of

Selected Metals

Figure 9

Density

The specific gravities of most lithium compounds do not differ much from other common chemicals. The density of the metal, however, is dramatically different from any other element. Lithium metal has a specific gravity of 1/2 water (.53). and will float on the lightest of liquids. Alloys of lithium and heavier metals are reduced in density. Often a combination of lithium's low density and other properties will assure a special application. Some examples will be described here.

Solubility

One thing seems to be certain about the solubility of lithium compounds: there are no rules. To be sure there seems to be little correlation between lithium salts and their common Group I analogs:

Lithium Sodium Potassium

Chloride

v.sol v.sol v.sol

Fluoride

insol sol sol

Sulfate

v.sol v.sol v.sol

Carbonate

si.sol v.sol v.sol

Hydroxide

si.sol v.sol v.sol

PRODUCTION AND ELECTROLYSIS OF LIGHT METALS 217 A curious phenomenon of lithium carbonate is its inverse solubility in relation to temperature:

ioo°c 0.7% o°c 1.5%

This property is exploited in the production of the carbonate. As it is being precipitated from lithium sulfate and soda ash (NapCO~) the process is kept at a

maximum temperature. Lithium bromide, on the other hand, is so soluble that concentrations in excess of 60$ (wt) are possible. The product is commonly sold as the 5 W solution.

Eutectics

Lithium chemicals tend to reduce the melting temperature of systems in which they are a part as they form eutectics of commercial advantage. There is probably no other single property that has had a greater impact on the commercialization of lithium. As will be seen in the discussion of applications, a large part of the world demand for the element is due to the fact that lithium chemicals are strong fluxes.

Conductivity

Lithium dissociates in water and many other solvents, as well as in the molten salt form. This characteristic, and probably the small atomic radius of the ion, make it's compounds excellent electrolytes. The metal and several salts are excellent conductors.

Hydroscopicity

Two of lithium's compounds, the chloride and the bromide, have a strong ability to absorb moisture from the atmosphere. The bromide is so efficient in this regard that a 50% solution of the compound, left in a partially full open beaker, will eventually overflow the container as it continues to absorb moisture from the air.

Coefficient of Expansion

Possibly directly related to ionic radius, lithium compounds, particularly the oxide, impart an inhibiting effect on expansion coefficients for other oxide systems such as glasses and ceramics.

Ease of Identification

While the quantitative analyses of lithium, particularly in the company of other group I and II A metals, can be difficult, the qualitative identification of the element is straightforward via fast and inexpensive methods. The "red flame" denotes the presence of the lithium ion in almost any system.

218 PRODUCTION AND ELECTROLYSIS OF LIGHT METALS APPLICATIONS

It is the objective of this presentation to inform the reader of the enormous number and variety of applications for lithium and its chemicals. To detail each use currently being employed by industry would be tedious reading, at best. The following are some are those uses which are currently significant to the lithium market we know today, and how they relate to the properties just discussed.

Ceramics and Glass

The major products used by these industries are lithium ore (spodumene or petalite) and lithium carbonate. In most cases the lithium materials are used as a fluxing additive which lowers the melting temperature of the raw batch. Additions to glass batches on the order of 0.15$ Li-0 have been shown to reduce

melt temperatures by 20-30 C. Other benefits are obtained, as well. Lithium is similarly added to porcelain enamel frit in order to improve firing characteristics and adhesion. One well known lithium use is in pyroceram developed by Corning Glass Works and known to the general public as CORNINGWARE.

Aluminum Smelting

The Heroult-Hall process for the electrolytic reduction of alumina in a molten cryolite bath can be enhanced by the addition of lithium in the form of lithium carbonate. The carbonate is converted to the fluoride in the bath. Here lithium fluoride, in concentrations of 2-3$ by weight in the cell, calls on its eutectic property to lower the electrolyte temperature by about 10-15 C. This lower operating temperature results in reduced fluorine emissions. By making adjustments to other operating parameters, current efficiencies can be increased by 2-3$· Alternatively, the high conductivity of the molten LiF can be exploited to increase amperage (and thus output of the cell) by as much as 15$.

Grease

One product of WW II was the development of the "all-purpose" lithium grease. Grease, which is a base oil thickened with a soap, has the undesirable tendency to exhibit a widely varying viscosity (hence lubricity) with temperature. Those made with lithium soaps (as opposed to sodium or calcium) do not.

Just as important is the relative insolubility of the lithium soap in water. This gives the lithium grease considerably more resistance to weather deterioration than its counterparts. Lithium grease comprises 60$ of all grease manufactured in North America and close to ^0% of the world's grease production.

Batteries

For many years, battery manufacturers have exploited the superior conductivity of the lithium ion in the electrolyte of the alkaline battery. In recent years a number of new lithium batteries have been developed. These not only employ lithium in the electrolyte, but as the anode itself. In these cells, the property of conductivity is joined by light-weight (density) and high voltage drop to provide a battery with increased energy per unit weight (energy density). This is particularly important for military applications.


Air Treatment

Lithium chemicals are employed in three distinct air treatment applications:

— Cooling. Large, institutional facilities, such as shopping malls and hospitals often employ absorption chillers as their source of air conditioning. These units utilize the property of lithium bromide to absorb water from its immediate atmosphere. By this device the evaporation of water from the surface of a heat exchanger is accelerated thereby cooling the refrigerant. An advantage of this system is the near absence of moving parts.

— Dehumidification. Large areas which must operate at very low humidity often employ systems which take advantage of lithium chloride's thirst for moisture. Here, the moist air is passed through a series of baffles, screens or felts impregnated with lithium chloride which absorb the moisture. Typical installations are gelatin manufacturers, pharmaceutical companies, breweries and hospitals.

— Air Regeneration. Space vehicles, submarines and miners safety devices all have systems to remove carbon dioxide from human exhaust. The vigorous, non-reversible reaction of anhydrous lithium hydroxide with CO« combined with low atomic weight of the Li ion results in the most efficient removal of the gas for the weight of absorbent employed.

Pharmaceuticals

Lithium therapy for the psychiatric treatment of mania and depression (unipolar and bipolar mood swings) has become quite well known and has become the drug of choice. While the mechanism of how this works is still disputed, most agree that it involves the migration of the ion through a human cell owing to lithium's small ionic radius.

Alkyllithium chemicals are being used as synthetic agents in the manufacture of organic intermediates for drugs and agricultural chemicals. The small lithium ion will often react at sites inaccessible to other metal analogs. Often the metal halide is a by-product of the reaction, where the lithium salt is preferred for reasons of efficient separation of reaction products.

Effluent Tracing

The increasing public awareness of the need to protect our environment has heightened the need to trace sources of ground and surface water pollution. The relative low "background" of lithium in most water streams plus the ease of analytical identification of Li make it an ideal way to "tag" a waste stream and follow its progress and rate of dilution. Lithium chloride or sulfate are commonly employed for this use.

Sanitation

While lithium displays no intrinsic sanitizing properties, it combines with the hypochlorite anion (0C1 ) to form a stable solid. This is used primarily in the disinfection of spa and swimming pool water. Other hypochlorites are available at lower price, however, the lithium salt has a solubility and hard-water

220 PRODUCTION AND ELECTROLYSIS OF LIGHT METALS compatibility advantage over the calcium salt and availability in solid form advantage over the sodium salt (NaOCl is stable only in solution—usually sold in 5% concentration).

Metal Joining

Two major metal joining methods employ lithium, both exploiting it's fluxing (eutectic) properties:

— Dip brazing. The fabrication of complex aluminum shapes such as wave guides and heat exchangers employs dip brazing techniques. Molten baths of lithium chloride or fluoride salts with other salts are formulated to operate at temperatures specific to the melting temperature of the brazing alloy. Lithium is the key to controlling this temperature.

— Electric welding. Lithium halides and carbonate are used as fluxing coatings for welding rod. Again, the eutectic properties of lithium are employed to advantage.

Polymerization Initiation

The manufacture of styrene butadiene copolymer (SBR) is accomplished by the use of a butyllithium initiation in the solution polymerization process. These and other block co-polymers (thermoplastic elastomers) rely on butyllithium to impart certain stereo-specific properties to the polymer chain. The lithium atom radius and the solubility of alkyllithiums in organic solvents are key influences to this phenomenon.

There are many other uses for the 70-odd lithium chemicals commercially available. They include dyes, photoprocessing, cosmetics, textiles, bleaches, and coatings, among others.

THE FUTURE

Alloys

The desirable attributes of lithium can be transferred with the metal in alloys of other metals. Often the best qualities of both metals are retained. In the next decade, rechargeable lithium batteries may utilize an anode of Al-Li alloy, combining the energy and light weight of lithium with the relative strength of aluminum.

The new low weight Al-Li alloys for aerospace applications will have tensile strengths in excess of either metal while exhibiting weight savings of about 8-10$ versus comparable aluminum alloy parts. When a part is re-designed to take advantage of its increased properties, weight savings of 15-20$ are achieved.

Fusion Energy

The global thirst for low-cost, safe energy drives continued research on fusion. Working reactors by the early to mid 21st century are predicted if some formidable technical problems can be overcome. In fusion, a deuterium-tritium reaction


produces neutrons captured by a lithium metal "blanket". These neutrons also react with lithium to produce more tritium fuel. Large quantities of lithium metal would be needed for this use (1).

Fuel Cells

Another larger scale energy source being developed is the molten carbonate fuel cell. Here, hydrogen and carbon monoxide are fed to the anode of a molten lithium carbonate-potassium carbonate electrolyte system. Hydrogen reacts with carbonate ion to produce water, carbon dioxide and energy. Carbon monoxide reacts with water to regenerate more hydrogen fuel and carbon dioxide. Again, the eutectic properties of lithium are utilized (2).

In summary, one thing is certain, these and unforeseen new uses will continue to develop and assure lithium its continuing reign as the Lightweight Champion.

(1) D. A. Dingee, Fusion Power, Chemical Engineering News, April 2, 1979» P. 32-45.

(2) Krumpelt, G. M. Cook, A.D. Pierce, VP Ackerman, "Molten Carbonate Fuel Cells for Coal & Natural Gas Fuels" Energy Sources Technology Conference, February 1984, New Orleans, LA.

223

Preparation of lithium metal by molten salt electrolysis Y.C. Höh, T.M. Chiu and Z.J. Chung Institute of Nuclear Energy Research CAEC, Lung-Tan, Taiwan, Republic of China

ABSTRACT

Molten salt electrolysis is the most practical method for the preparation of light metal· Due to the density of lithium metal lower than molten salt that float on the surface of molten salt. So it is necessary collect-ed by means of special way. This study would design and manufacture a fused salt system , and inspect the possibility of molt salt electrolysis for lithium preparation . In add-ition , operating characteristics was discussed for current density/ Cathode-Anode gap and some effect factors. The electrolytic system consisted of three parts, a collector chamber , an electrolysis cell and an absorption system for the off-gas . The collector chamber was located inside the electrolytic cell. Both of the chamber and the cell were heated by a furnace. Lithium metal was formed in the steel cathode after electrolytic reaction. As the liquid metallic lithium formed, it passed through a steel riser pipe and discharged into a collector chamb-er. In the other end, the chlorine gas from the graphite anode was neutra-lized with 15 % sodium hydroxide solution by an absorption system. The results of experiment showed that lithium could be prepared from a melt composition of 60 % lithium chloride mixed with 40 % potassium chloride

eutectic mixture at the temperature range of 420 - 550 °C. The current efficiency could be reached to 71 % .

KEYWORDS

Lithium metal; Molten salt electrolysis; Reduction; Absorption system.

INTRODUCTION

The specially properties of lithium metal play a very important position in light metal alloy(1), battery industry(2) and energy supply(3) etc.. Generally there are four methods to prepare lithium metal, i.e.,

electrolysis(4), reduction(5), in which Al N Ca * and Mg metals or Si N C elements were used as the reductants, thermal dissociation(6), in which low melting point and corrosion resistant material such as carbonate complex (M2C03) was selected as the reactant and plasma reaction(7). Both of the 3rd and 4th methods were not commonly used for the sake of lower production yield and reaction rate .


The electrolysis and reduction methods have advantages and disadvantages (8). Electrolysis method might require more energy, however, the yield is relatively high. According to the difference of intermediate product, the electrolysis method included aqueous and molten salt systems (9). When the reaction takes place in an aqueous system, the metal product must be removed immediately in order to prevent from reacting with water. This problem can be overcome by using mercury as the cathode (10) and the metal product can be dissolved in mercury to form an amalgam. The metal can then be separated from mercury by distillation. However, this method was difficult to control and operate, it was not suitable for high activity metal. Molten salt electrolysis was the common technique to prepare light metal. Especially, alkali metal could be rapidly prepared by electrolysis. Molten salt electrolytes have many advantages over aqueous system. The absence of water eliminates anode effect and the reaction of the product with water. The present study relates to a method for the production of lithium through an electrolysis of molten salt mixtures. This process was developed for both batchwise and continuous operations etc.. The two methods of operation involved different cell designs and experiment conditions. The collection chamber was located underneath the electrolysis cell in order to prevent the riser pipe from obstructing the continuous operation. However, it has been verified that the construction of the cell was satisfied for lithium preparation. The details of the process was discussed in the text.

EXPERIMENTAL

Design and Manufacture of Molten Salt Electrolysis System

A Molten salt electrolysis system was designed and manufactured to study the electrolysis of lithium. Due to the character of the light metal, the product collection system should be designed properly. All of the construction and electrode materials and the control valves should be corrosion resistance at an elevated temperature. In this study, batchwise and continuous operation systems were used. Both of them consisted of three parts, an electrolysis cell, a collection chamber and an absorption system. The details of the system were discussed as follows: Electrolysis cell. The whole electrolytic cell was made of 304 stainless

steel, including anode * cathode and diaphram. The diaphram was used to separate lithium metal and chlorine gas. High density graphite was used as the anode material in order to avoid the corrosion effect. The cathodes were a cylinder stainless steel tube surround the anode and a stainless steel rod for continuous and batch operations, respectively. The anode and cathode were separated by a stainless steel diaphram in order to prevent lithium and chlorine from recombining. The diagram of the electrolytic cell was shown in Figs. 1 and 2.


Fig. 1. The diagram of molten salt electrolysis cell for batch operation.

Fig. 2. The diagram of molten salt electrolysis cell for continuous operation. 1. Gas Collector Tube. 2. Anode. 3. Cathode. 4. Metal Collector. 5. Collection Tube. 6. Diaphragm. 7. Steel Shell.


Collection chamber for lithium metal. In batch system, a steel plug was used as a control device. When reaction completed, the plug was opened and lithium metal was flowed into a collection chamber which was located outside the cell. During continuous operation, the produced metallic lithium was in the cathode and floated on the surface of molten salt. Then, passing through a steel riser tube and discharged continuously into a holding tank. Argon gas was purged into the tank in order to prevent the produced lithium from oxidation. The collector was made of carbon steel.

Because the melting point of molten lithium was 180 °C, the riser pipe was

kept at a temperature of 200 °C . Both of the collection chamber and the riser tube were located inside the electrolytic cell and were heated by a furnace. Chlorine absorption. When the electrolytic reaction started, chlorine was produced continuously at the anode. Due to chlorine reactes easily with lithium and forms lithium chloride, and therefore, the yield decreases. In order to avoid this problem , a molten salt electrolytic cell with a diaphram to separate cathode and anode was designed. Chlorine gas was then removed from anode chamber by slightly suction. The evolved chlorine gas was scrubbed by caustic solution . The absorption system was consisted of

three packed columns with 1 meter length inside diameter of 18 cm and a 50-liter plastic drum containning 30 liters of 15 % NaOH solution. To forbid corrosion effeet, a nickel pipe was used as duct. The chlorine absorption system was shown in Fig. 3.

Suction

Cl2 Cl2

Fig. 3. The Chlorine absorption system for lithium metal preparation.


Lithium Metal Preparation

The preparation of lithium metal is quite similar with the sodium preparation method(11). Lithium chloride was used as a raw material. The decomposition potential of lithium chloride decreases with electrolysis temperature increasing indicates that the reduction reaction happens easier at a higher temperature. However, corrosion rate of the anode became significantly with temperature rising. To overcome this problem, LiCl-KCl eutectic mixture was used as an electrolyte , the mixture melted at 360 °C

and electrolytic reaction took place at 450 °C.The results of molten salt electrolysis showed that lithium and chlorine were produced at the cathode and anode, respectively, the reaction equation can be expressed as follows

Anode reaction 2 Cl" > Cl; 2 e"

Cathode reaction Li + e --» Li

The chlorine removal and the lithium metal collection were the two important aspects in the system studied. The specific gravity of lithium metal was lighter than the molten salt, lithium metal floated on the surface of the molten salt. A special collection method was required. According to the literature(12), 5 kg of chlorine gas was produced per 1 kg of lithium metal was prepared. Therefore, an absorption system was required . Figure. 4 shows the flow diagram of the molten salt electrolysis process for lithium.

Molten Salt Preparation Eutectic Mixture

sl· Pre-electrolysis

I Electrolysis -> Chlorine Absorption

I Lithium Product

Storage

Fig. 4 . The flow diagram of molten salt electrolysis process

It consisted of four steps, molten salt preparation, pre-electrolysis, molten salt electrolysis and storage. These steps were discussed as follows Molten salt preparation. Due to the melting point of lithium chloride is 606 °C, a two component electrolyte ( LiCl-KCl ) was used in this investigation in order to reduce the melting point (13). The molten salt


composition contained 60% LiCl and 40% KCl. However, lithium chloride could be added 10 % more with respect to the eutectic composition of the mixture.

The melting point of the eutectic mixture was reduced to 360 °C. Pre-electrolysis. A pre-electrolysis step was needed in order to remove the impurities such as water and other metal oxides in the electrolyte. These impurities caused electrolytic efficiency decrease.-* Molten salt electrolysis.Two kilograms of LiCl-KCl eutectic mixture were fed into the electrolytic cell. Operation temperature, current and voltage

were kept at 420°C-SB0°C, 80A, 6^10 V, respectively. Argon gas was purged into the surface of the molten salt in order to prevent corrosion and oxidation reactions from taking place. The operation conditions for lithium metal preparation were listed in table 1.

TABLE 1 The Main Operation Conditions of Lithium Metal Preparation

Operation Data Lithium

Parameters

Raw Material Electrolyte

LiCl (%)

KCl (%)

Temp. (°C) Current (A) Current density

(A/cm2 )

Voltage (V) Anode Cathode

Electrolysis Cells

Conditions

LiCl

40^60

60—40

420—ΒΒ0 80 0.40

6—10 graphite Steel

During the electrolytic reaction, LiCl was dissociated to Li and Cl Ions , where Li ion moved toward cathode and reduced to lithium metal. After reduction, lithium metal rised up to the surface of the molten salt. In the other end, Cl ion moved toward anode and oxidized to chlorine gas. Chlorine gas was ducted through a scrubbing system before discharge to the atmosphere. Storage. Metallic lithium product was storaged in liquid petroleum in order to prevent from oxidation with air or water.


Due to the accotipanied side effects together with the electrolytic reaction, the current efficiency of molten salt electrolysis was never

reached to 100 % . Usually, It was only 70—80 % (14). Current efficiency could be calculated by using the following equation(IB):


G SP

229

C. I. t S a

n : current efficiency G : weight of metal (g)

S P : cathode total area (cm2 )

S a : total conductive area of cathode (cm2 ) t : time (sec) I : current (A) C : 0.000072 (g/A.sec)

In this research the current efficiency was reached to 71%. Silver-white and high reactive lithium metal was collected in the collector and then stored in liquid petroleum. The emission and absorption spectra of the product were showed in Tables 2 and 3, respectively. Table 2 presented the data of qualitative analysis by emission spectra.

TABLE 2 The Results of Emission Spectra Qualitative Analysis for

Metallic Lithium Sample

Item

Al

Ca

Cr

Cu

Fe

K

Li

I

O

o

I

o

o

o

o o

o

o

• o

Item

Mg

Mn

Na

Ni

Si

Sn

I I

O

o

o

o

o

o

Key : O : present * *k: very strong

I : sample I (11L-1), I : sample I (11L-2)

230 PRODUCTION AND ELECTROLYSIS OF LIGHT METALS TABLE 3: The Impurity Contents in Lithium Metal Product

Item Mn Mg Fe

Si Al Ni Cr Cu Ca

I ( zxg/ml)

< 0.40 0.35

3.80 19.00

< 4.00 3.90 1.40 1.20 2.40

I (ug/ml)

2.00 4.00 60.00

6.00 < 4.00 5.00 1.50 40.00 17.00

* I : sample I (11L-1), I : sample I (11L-2)

Sample I (11L-1) was prepared from lithium metal dissolving in H2O and sample I (11L-2) was the filtrated residue dissolved in dilute HCl solution. The results of spectra analysis showed that the impurities contained Mn *Mg ^ Fe Si Al Ni Cr Cu and Ca etc. and most of the contents were down to 10 ppm except Fe Cu and Ca. Among the impurity concentrations. Fe was much higher than any others, this was attributed to the corrosion effect of S.S 304 made electrolytic cell. This could be improved by using 310 stainless steel as the material of the electrolytic cell.

There were many factors such as current density ^ anode-cathode gap temperature and anode effect(16) etc. affecting the molten salt electroly-sis for lithium metal preparation. Figure 5 showed the variances of lithium chloride percentage with respect to reaction time and cathode current density.

50.00

40.00

Time I'min)

Fig. 5. The variance of lithium chloride percentage with reaction time in the molten salt electrolysis process.

( Current Density : 0.2, 0.3, 0.4, 0.5 A/ci2 )


In this figure, the results showed that electrolytic reaction is slow in the beginning, and the reaction rate became faster as time increasing. Figure 6 presented the relationship between the residual LiCl % in molten salt and current density after 4 hours reaction.

4 9 . 0 0 I " ' ' ' ■ ' ' ' I ' ' ' ' i i i i i i i i I I i i i i i i i i i 1 i i i

48.00

47.00

O 46.00

45.00

rxn time r 240

4 . 0 0 | 11 i f i r i r i 11 i i i 11 i i i | 11 i i i i i i i 11 i i i 11 i i i 11 i i i i i i i i

0.10 0.20 0.30 0.40 6.50 0.60

Current density ( A / c m * )

Fig. 6. The relationship of residue LiCl % and current density after 4 hours reaction.

From this figure, it is concluded that the higher the current density, the faster the reaction rate is reached. However the anode graphite was destroyed more seriously at high current density. Therefore, high density graphite was generally selected for the anodeused. The best current density

was set between 0.3^0.4 A/cm2 . The distance between cathode and anode also affected the electrolytic efficiency. The larger the anode-cathode gap, the slower the reaction rate was resulted. If the gap is too small, lithium and chlorine will recombine. The results indicated that the proper gap between anode and cathode was

3^5 cm. Both of the electrodes were separated by a steel diaphram to prevent lithium and chlorine from reaction. Anode effect plays an important role in the molten salt electrolysis. This effect was attributed to the moisture existing in the electrolyte and the high resistance gas film presented between the electrodes and electrolytes. These problems could be solved by drying the raw materials and by a pre-electrolysis step. Also slight rise the reaction temperature may avoid high

potential from gas film. However the temperature should not be over 550 °C in order to reduce the corrosion problem. During operation, the applied voltage was larger than the decomposition voltage. The decomposition potential(17) was defined as the smallest voltage which the metallic ion deposits on the cathode. Base on Helmholtz Thomson rule(18), the energy required to decompose a mole of salt might be


expressed as the product of this decomposition voltage "E" and the number of coulombs , equivalent to the heat of formation.

E x 96494 = energy = heat of formation

The relationship between operation voltage and decomposition voltage is presented in Fig.7.

g QQ -111 I I I > -J. ' I I I i t 1.1 1, 1 I

CD

o - 4 -'

o >

7.00

6.00 H

5.00 H

4.00

! 1 1.1 M I I I I 1 I I I 1 1,1 1.1 i I 1 1 I 1 I I I

cell potential

decomposition potential

3 . 0 0 "T~T—i—r—i—i' i )—i—i—|—i—ι—i—ι—i—i—>—!—ι—j—ι—ι—i—!—i—i—i—i—r-400.00 450.00 500.00 550.00

Temperature ( °C)

Fig. 7. The relationship between operation temperature and voltage.

The difference between the above mentioned two kinds of voltages was called the polarization voltage . It was caused by the electrolyte concentration gradient nearby the electrode and bubble on the electrode surface. It was helpful to solve these phenomena by suitable stirring the molten salt electrolyte. Figure 7 also indicated that the decomposition voltage of lithium chloride decreased with increasing of temperature. Potassium chloride was not reduced during the electrolysis period due to its high potential. It was the reason why the potassium chloride was chosen as the fused salt. From the results of absorption spectra analysis, it was found that the potassium content in the lithium metal product is less than 1 % . The relationship between the operating current and voltage was shown in Fig.8. It can be seen that the voltage only changes a little as the current increased rapidly. It may be implied that the system resistance is very low.


>

CD

Ό ■+->

O >

8.00 :

-7.00 : ■

\

6. CO -

>

5 X 0 \

10.

1 I l.t

00

. l . l l l l H L

20.00

I J U I J I , I I , I , J I I . I I I I l i l i i n

S X

X s

/ / /

30.00 40.00 50.00

M I L»

: :

" I

':

V

-

\

6 0 0 0

Current (A)

Fig. 8. The relationship between current and voltage for molten salt electrolysis process.

CONCLUSIONS

The following conclusions can be drawn from this study:

(1) The optimum operation temperature is 420^550 °C.

(2) The best current density was between 0.3^0.4 A/ci2.

(3) The proper gap between anode and cathode was 3^5 cm. (4) The decomposition voltage of lithium chloride decreases with

increasing temperature.

REFERENCES

(1). Unknown( 1985). Muminum Lighter than Aluminum.Foote Prints, 48,19. (2). D.A.J. Swinkel( 1966) .Lithium -chloride Battery J. Electrochem. Soc.

, 113, 6. (3). I.Y. Borg and L.G. OfConnell(1976). Energy Sources., 2, 347. (4). S. John Newman (1973) .Electrochemical System, Prentice Hall, Inc;

Englewood Cliffs, New Jersey, P.29 . (5). P. S. Baker, F. R. Duncan, and H. B. Greene (1953). Lithium Metal

Production,Oak Ridge National Laboratory, Tenn; Rept. CF-53-4-185 (6). K. Luther, J. Troe and H. G. Wagner(1972). Ber. Bbunsenges.Phys.Chem

.76, 53. abstracted in Chemical Abstracts 76, 77081 (1972) . (7). F. K. Mctaggart (1970). Use of Plasma in production of Metals from

their Halides,U.S. patent 3,533,777 (Oct. 13). (8). P. Mahi,A.A.J. Smeets, D.J. Fray and J.A. Charles (1986) .Journal of

Metals., November , P.20 . (9). D. Inman, S. H. White(1978) .Journal of Applied Electrochemistry. ,8,

375.


(10).L.F. Audrieth (1939). Inorgan ic Sys theses , Mcgraw-Hill Book Co, Inc; New York, Vol, p 7.

(11).P. S. Baker, G. F. Wells and W. R. Rathkamp(1954).J. chem. edu . , 515. (12).W. A. Averi l l and D. L. Olson (1978). Energy . , 3, 305. (13) .Richard Bauer (1972) .Chemi-Ingenieur Techn ik . , 44, 147. (14).D.W.F. Hardie (1975) . E l e c t r o l y t i c manufacture of Chemicals from s a l t

,The Chlor ine I n s t i t u t e , I nc ; New York. 2nd ed p.5 . (15).S. Tokumoto, E. Tanaka and K. Ogisu(1975). J . Mining. Met. I n s t . Japan

^ , 24,18. (16).D. G. Lovering (1982) .Molten s a l t Technology, Plenum publishing

corporation , New York. (17).S. N. Flengas (1964).Fused S a l t E lec t romot ive Force S e r i e s , in the

theEncyclopedia of E l ec t rochemis t ry , Reinhold publishing Corporation, New York, p 644-649 .

(18).D.W.F. Hardie (1975) . E l e c t r o l y t i c manufacture of chemicals from s a l t , 2nd edi t ion, The Chlorine I n s t i t u t e , Inc. 342 Madison Avenue, New. York.

235

A carnallite-based magnesium project for Canada Richard B. Stein Bechtel, Inc., Fifty Beale Street, San Francisco, California 94119, U.S.A.

ABSTRACT

The Mining & Metals Operations of Bechtel, Inc. has performed a technical and economic feasibility for a world-scale, grassroots, carnallite-based magnesium project for Western Canada. Starting material is from underground deposits of bedded carnallite occurring in the Province of Saskatchewan. Carnallite is extracted by conventional solution mining and the ensuing process scheme comprises:

o Decomposition of carnallite in methanol (Bechtel Process) with recovery of KC1 co-product for sales,

o Dehydration of MgCl2-6H20 using glycol and ammonia (St. Joe Process), o High-efficiency, high-amperage electrolytic cells to produce magnesium metal and chlorine, o Liquefaction of chlorine co-product for sales.

The conceptual project is based on annual production of:

25,000 short tons of magnesium 71,000 short tons of chlorine 100,000 short tons of potash

Depending on plant location, electric power is either imported or produced by cogeneration with process steam used for solution mining and feed preparation. The economics and financial analysis were performed by J.M. Pryde, Ltd. of Calgary, Alberta and indicated a very attractive rate of return.

239

Economics of the production of gallium J.C. Agarwal, F.E. Katrak, F.C. Brown and MJ. Loreth Charles River Associates Incorporated, John Hancock Tower, 200 Clarendon Street Boston, Massachusetts 02116, U.S.A.

Gallium is a minor element in the earth's crust, constituting a total of 0.0015 percent. Gallium is, however, more abundant than molybdenum, lead, and tungsten. The tendency of gallium to substitute in other mineral structures and the rarity of gallium-containing minerals, however, has led to widespread gallium dispersal.

The major association of gallium is with zinc sulfide (sphalerite) mineralization and with bauxite (aluminum) deposits. Not all sphalerite or bauxite deposits have sufficient gallium concentrations to warrant extraction. For those that are sufficiently gallium-rich, however, gallium is extracted after it has been further concentrated during the processing of these raw materials to final metal.

From bauxite sources, the gallium is extracted mostly from Bayer alumina plant production; a smaller portion is obtained from aluminum metal process residues. The European ores, particularly from Eastern Europe, and the bauxite ores of Africa and Australia provide the raw materials from which gallium is extracted, principally in French, German, and Japanese alumina plants.

A minor source of production is the gallium obtained from zinc refinery residues. This constitutes only a small fraction of total production, and is generally found in association with germanium, which utilizes these zinc ores as the major source of primary germanium.

A new source of gallium that only recently has been exploited is the primary gallium-germanium ores of Musto Exploration's St. George mine in Utah. This constitutes the only nonbyproduct source of primary gallium in the world.

RESERVES

The U.S. Bureau of Mines estimates total world reserves of gallium to be 110,000 tonnes, principally located in Africa and Australia. This statement is somewhat misleading, as the gallium can be extracted only as rapidly as the associated bauxite or zinc ores are exploited. Since a considerable portion of these ores are not currently processed for gallium, they actually constitute only potential reserves. Nevertheless, the determination of whether gallium can be extracted from Bayer alumina liquors or zinc residues is largely a function of price. At a price of over $250 per kilogram, several of the Bayer alumina liquor process streams around the world become economic, and extraction could commence. In addition, current capacity significantly exceeds demand, and additional capacity could become operational with minimum capital investment with gallium prices in the range of $300 per kilogram.

SCRAP

Prompt (or new) scrap is a significant source of supply for gallium, principally because the process required to make wafers for a range of gallium products yields a significant amount


of scrap. Since these yields may average only 30 percent, up to 70 percent of the material used in production becomes metal scrap. A portion of this scrap is lost in the production process; the remainder is lost in the reprocessing of the scrap. The net scrap losses become a significant source of supply when compared to most metals.

PRODUCTION CAPACITY

There has been a considerable increase in world production capacity of gallium in response to the sharply increasing demand of the 1980s. Production capacity actually declined, particularly in the United States, in the early part of the 1980s. The rapid rise in demand, however, has led to a considerable increase in production capacity since that time. A fraction of the stated capacity is utilized simply to upgrade lower-quality gallium metal to higher-quality (6Ns) metal for more rigorous end-use specification requirements.

Sixty-five percent of 1987 primary active world capacity is concentrated in Western Europe and Japan, which depend heavily on African and Australian feedstock for Bayer liquor production of gallium. Eastern Europe and China rely primarily on domestic bauxite ores to provide an additional 20 percent of world gallium production capacity.

Of the approximately 107 mt of primary plus secondary gallium metal capacity that existed in 1988, about 30 to 35 percent has been added from 1982 to 1987. Table 1 lists the major world processors of gallium as of 1987; Table 2 includes likely new capacity additions and deletions. The new gallium developments that have occurred in the last 5 years include

Tonnes

Sumitomo Chemical (Japan) 10 primary Ingal International Gallium (West Germany) 8 (expansion)

primary/secondary Alcan Aluminum (Ontario, Canada) 4-5 scrap Sumitomo Metal Mining (Japan) 3-4 scrap Ote Metal (Japan) 6 primary Tingzhou Aluminum (China) 3+ primary

TOTAL ADDED CAPACITY 34-36


TABLE 1 Major World Processors of Gallium: 1987

Country

Canada China Czechoslovakia France West Germany

Hungary Japan

Switzerland United States

Startup

1987

10/83 7/87

1/87


Company

Alcan Government Impemex Rhone-Poulenc INGAL (Schwandorf) INGAL (Lunen) Preussag Penaroya Hungarian Aluminum Corp. Sumitomo Chemical Sumitomo Metal Mining Rasa Ind Ote Metal Dowa Kogyo Alcan Musto Exploration Eagle Picher Alcoa

TABLE 2 Likelv New Additions and Old Plant Closures. 1988-1990

Country

Australia/ United States

Canada

France

West Germany

Japan

Norway

Startup

1988-1989

1988-1989

1988

Company

Rhone-Poulenc

Alcan

Rhone-Poulenc

INGAL (Lunen)

Dowa Kogyo

Elkem


In addition, the world's only production of gallium mined solely for its gallium/germanium content was initiated in St. George, Utah in 1986. This mine/plant has a capacity to produce 9 tonnes per year of gallium. To date, however, capacity has not been achieved, and the operation has been shut down to improve plant recovery of gallium and germanium.

The overall increase in gallium capacity that has occurred from 1982 to 1987 is substantial. Even with rapidly rising demand, substantial capacity is in place to handle rapid demand growth.


In addition to the significant new developments for gallium production in the 1980s, several other projects are planned to increase capacity or replace obsolete capacity. The most significant will be the Rhone-Poulenc primary gallium extraction plant at Pinjarra, Western Australia. This plant will use Rhone-Poulenc liquid-liquid extraction technology to process Bayer plant liquors from Alcoa of Australia. It will produce a 3N to 4N product, which subsequently will be upgraded to 6N or 7N gallium at Rhone-Poulenc's Freeport, Texas facility. This plant will replace the 15 tonne-per-year Rhone-Poulenc facility at Salindres, France, which is scheduled for closure in 1988. The Australian facility will likely have an initial design output of 25 tonnes per year, which could be modified to produce 50 tonnes per year if demand warranted it.

In addition to the above net capacity increase at Rhone-Poulenc, several other new facilities will be commissioned from 1988 through 1990 (Table 2). The Elkem facility in Norway will be a primary facility processing aluminum smelter waste streams. In addition, a process change will be introduced at Dowa Kogyo to increase primary capacity to 8 tonnes per year of gallium in its Iijiama zinc refinery; germanium will also be extracted.

Other processing capacity additions likely to occur include Alcan in Quebec and Penaroya in France. In addition to new sources, there is progress toward improving yields of gallium in gallium arsenide substrate utilization. In reality, the actual consumption of gallium in products is significantly lower than annual requirements.

PRODUCTION

Production of primary gallium on a worldwide basis was on the order of 37 tonnes in 1986. Primary production has been growing steadily since the early 1980s from the 25-tonne level. Production of primary gallium is below our reported primary active capacity of 83 tonnes. However, a portion of this capacity (about 20 tonnes) was effectively idle in 1986-1987. So in reality the actual active primary production capacity in 1985-1986 was only about 62 tonnes.

A breakdown of production from primary and secondary sources and imports is given in Table 3. Since Japan is a large consumer of gallium in product manufacture, it uses a significant amount of refined scrap and imports to meet demand requirements. Scrap production in 1986 was 9 tonnes, and current scrap capacity in Japan is about 16 tonnes. Gallium scrap production in the United States is less precisely known, but it can be estimated at about 4 tonnes based upon consumption data. The major difference in the U.S. primary requirement is that it is met largely by imports of metal, rather than from domestically produced primary metal, as is the case for a significant portion of Japan's requirements. Little scrap is generated for recycle in Europe, as most primary gallium is exported to Japan and the United States and not consumed internally.

Of the "new scrap" generated, we assume that a fraction is permanently lost and the balance is used during the year following the one in which it was generated. We further assume that scrap satisfies demand first, and primary purchases satisfy the balance. Western world scrap capacity is on the order of 30 to 35 tonnes per year, which is approximately twice the current production level.


TABLE 3 Gallium Production and Imports (Kilograms)

Japan Primary Scrap Recovery Imports (6Ns)

(3-4NS)

Total

United States Primary Scrap Recovery

Imports

Total

Europe» Primary Scrap Recovery Imports

Total

China Primary Secondary Imports

Total

1980

2,500 3,500 3,500 4,500

14,000

w

6,175

6,175

Net

0

1981

3,000 5,000 3,500 5,500

17,000

w

5,536

5,536

Exporter

0

1982

3,000 4,000 2,500 6,580

16,080

1,560

5,199

6,759

0

1983

3,000 5,000 4,820 9,380

22,200

7,924

7,924

16,000 1,000

17,000

7,000

0

7,000

1984

10,000 7,000 7,435 6,300

30,735

9,669

9,669

18,000 2,000

20,000

6,000

0

6,000

1985

10,000 10,000

8,514 8,085

36,599

2,000-4,000(6)

7,964

11,964

20,000 3,000(6)

23,000

6,000

0

6,000

1986

10,000 9,000 7,000 6,300

32,300

750 4,000

17,202

21,952

19,500(6) 3,000(6)

22,500

3,000

0

7,000

»Includes Hungary and Czechoslovakia, (e) estimate.


COST ESTIMATES

If the current $400 price for 3-4N gallium remains the same, it is already economical for more gallium to be extracted from Bayer liquor in aluminum production. Whether this is reflected in expansion of existing capacity, reactivation of idle capacity, or development of new capacity, it is»clear from recent developments that significant economic recovery of gallium is possible at the existing price levels. Our methodology for estimating the future supply of gallium is described below.

Primary gallium of 3-4N purity will be produced using one of the following technologies: solvent extraction (Rhone-Poulenc); solid ion exchange (Sumitomo); electrochemical reduction based on mercury (Alusuisse); or chemical dissolutions (various processes suitable for treatment of specific drosses). Secondary gallium will be produced by a variety of chemical dissolution processes suitable for treatment of specific scraps. We characterize the various processes in terms of the estimated unit capital requirements for each and the unit consumptions of materials, utilities, labor, and capital-related charges. The unit consumptions are summarized in Table 4. Then, unit costs are determined for each of the items consumed, and production costs are estimated from the sum of the unit consumptions times the unit costs. Costs are computed for all plants in operation, or assumed to have come into operation, and the results are ordered to show cumulative capacity at increasing costs. Direct operating cost estimates for current primary and secondary producers are shown graphically in Figure 1. The lower-cost primary gallium producers (e.g., Rhone-Poulenc ~ $166 per kilogram - tend to have larger-scale plants. High-cost facilities, such as Elkem's plant ($450 per kilogram) are relatively small with a high labor cost.


T A B LE 4 unit consumption and other operating cost Assumptions for Production of

cat Iium by Process

CALL IUM EXTRACTION PROCESS

PROCESS MATER 1/

HCI NaOH HN03 H2S04 H2S Hg RESIN-IX RESIN-SX SX REACENTS OTHER ·

UTILITIES

ELECTRICITY FUEL. OTHER

LS

• LABOR (FOR 10-TONNE

UNIT

TONNES/KC TONNES/KC TONNES/KC TONNES/KC TONNES/KC TONNES/KC M3/KC M3/KC TONNES SUS 000

KWH/KC SUS 000

PLANT)

ALUSUISSE

0.0065

0.0015

0.00075

0.00015

100

190 100

ION EXCHANCE

0.0233

0.0255

0.0023

50

77 50

SOLVENT EXTRACTION

0. 102 0.112

0.0001 1

0.00121

50

90 50

CHEMICAL DISSOLUTION

0.261 0

0.286 0

0

0.0001 1

50

90 50

SCRAP

00177 00194

00004

100

30 50

DIRECT OPERATINC EMPLOYEES

DIRECT MAINT. EMPLOYEES

SUPERVISION EMPLOYEES

LOCAL OVERHEAD 50 % OF ABOVE

• BASED ON 10 TONNE-PER-YEAR CALLIUM PLANT.

SOURCE: CHARLES RIVER ASSOCIATES.

DIRECT OPERATING COST ($US PER KG GALLIUM)

500

400

300

200 l·

100 l·

0 10 20 30 40 50 60 70 80

CUMULATIVE CAPACITY (TONNES GALLIUM PER YEAR)


Fig. 1. World primary gallium supply: 1988.

244


These costs are estimated for production at full capacity. Costs estimated at current production rates are shown graphically in Figure 2, and confirm the current supply/demand balance.

DIRECT OPERATING COST ($US PER KG GALLIUM)

1100 -

1000 -

900 -

800 -

700 -

600 -

500 -

400 -

300 -

200 -

100 -

0 -

r-T Γ

H I

' 1 ■ ' ■ ' ' 1 ■

0 10 20 30 40 CUMULATIVE PRODUCTION (TONNES GALLIUM PER YEAR)


Fig. 2. World primary gallium supply: 1988 production.

In addition to the supply curves developed for primary gallium, one must be cognizant of the impact of secondary gallium supplies. Current secondary gallium capacity is approximately 32.5 tonnes, and it can be produced at costs from about $120 to $240 per kilogram above the cost of gallium contained in the scrap. It is possible that more scrap processing capacity will be added in the future as more scrap becomes available, but with capital recovery charges in the vicinity of $110 to $120 per kilogram, plants in the 5 tonne-per-year range will be at a significant disadvantage vis-a-vis 25 tonne-per-year primary plants. A new 25 tonne-per-year scrap plant would have capital charges of only $60 per kilogram and realize significant economies of scale in labor, reducing labor costs to about $50 per kilogram. This would reduce total costs not including scrap costs to the $200-per-kilogram range, making secondary gallium slightly more costly than primary gallium from large new plants.

247

The problem of limited recoveries in the Pidgeon process for magnesium production J.R. Wynnyckyj, E. Tackie and G. Chen Department of Chemical Engineering, University of Waterloo, Waterloo, Ontario, Canada, N2L 3G1

ABSTRACT

A hypothesis is proposed concerning the reasons for the failure of the reaction.

2MgO + 2CaO + Si — 2 Mg(g) + Ca2Si04

to go to completion. Typical recovery of Mg is about 80% in spite of the usual addition of about 15% excess silicon. The hypothesis derives from recent laboratory investigations of the above reaction in a fixed bed with vapour entrainment into an inert gas. Radial temperature gradients within the retort (or, in the case of the Bolzano/Brasmag process, within layers of charge between heating elements) cause transfer of, effectively, magnesium oxide from the hotter to the colder portions of the charge. The local MgO excess, so created, fails to react. The local transfer of magnesium oxide is via reaction of the product vapour (Mg) with calcium silicate, i.e.

2Mg(g) + Ca2Si04 — 2 CaO + 2MgO + S i ^ )

The Ca2Si04 having been formed by the primary reaction, thus:

4CaO + (2X + l)Si — 2CaSix(sm) + Ca2Si04

Both experimental and modelling evidence will be presented to substantiate the above hypothesis.

249

An evaluation of the aluminothermic production of magnesium N.E. Richards, A.F. Saavedra Reynolds Metals Company, Manufacturing Technology Laboratory, Sheffield, Alabama, U.S.A.

R.L. Potter Reynolds Metals Company, Richmond, Virginia, U.S.A.

L.M. Ruch Cabot Corporation, Boyertown, Pennsylvania, U.S.A.

ABSTRACT

A process for producing magnesium metal at atmospheric pressure, using aluminum as a reductant, was proposed by Avery (1974). In the bench scale experiments, dolomite and magnesite in varying proportions were reacted with aluminum in the temperature range 1475° to 1700°C with conversion efficiencies up to 86%.

KEYWORDS

Magnesium production, metallothermic reduction, aluminothermic reduction, electric furnace.

INTRODUCTION

In the recent past, when magnesium-containing alloys were in demand for both beverage containers and components in the trim of automobiles, there was a risk of either shortage or increased pricing of magnesium.

Presently, in addition to that magnesium extracted electrolytically from at least three upstream methods for the feedstock, Northwest Alloys has modified its initial Magnatherm process to use both aluminum and ferrosilicon to reduce dolomite, CaO.MgO, in an electric furnace, under reduced pressure.

We were interested in discovering what the parameters would be for a potentially continuous process for extracting magnesium from dolomite also, but at atmospheric pressure using aluminum as reductant. A route using aluminum silicon alloy (approximately 30% Al) had been patented by Avery (1974); consequently, even for our research and development, his position was acknowledged.

The objective of this presentation is to present the summaries of the laboratory and pilot scale developments we made as essential components for estimating and projecting the feasibility of the aluminothermic production of magnesium.

EXPERIMENTAL

For the small scale experiments involving 6-8 kg of slag and reactants, after a few tests, chemically successful, we abandoned inductively heated, air-tight, graphite crucibles and


developed an internally heated system shown schematically in Fig. 1. While this equipment worked and, like the internally heated system with which we replaced it, could sustain a vacuum of 29 in. of Hg, we could not be sure that in the temperature range 1500° to 1700°C, the carbon of the crucible walls was not contributing to the reaction.

The graphite-lined resistance furnace designed for these experiments comprises two sections joined at electrically insulated, clamped flanges at the top of the 8 in. internal diameter graphite tube, cemented to the graphite base and lower stub for connection to the power supply. Purotab (calcium aluminate) refractory lined the thermally insulated upper assembly which was fitted with a side arm condenser tube, a sight tube for temperature measurement by optical pyrometer (or ray-o-tube), an insulated entry tube for the upper graphite electrode for heating current, and a charging port.

By mounting in a counter weighted pivoting frame, the furnace could be tilted up to 60° for tapping, although the tap hole was located midway up the crucible wall so that when tapping without tilt, enough slag, about 4.4 kg, would remain for the option of continued operation after a tap.

A piece of carbon rod was preferred over graphite for plugging the 1 in. i.d. taphole because of the excessive heat loss and freezing too much slag at the tap hole. This was ready assurance that there was a skull of solid slag around the walls of graphite crucible which, when filled to a depth of 10 in., had a capacity of 22 kg. Opening the taphole without venting the pressure in the furnace would result in a gush of molten slag, followed by a violent, torch-like jet of burning Mg vapor. (During an experiment, a positive pressure of about 1 psi was maintained with a small (2 - 10 scft/hr) flow of Ar. This flow was introduced across the glass observation port to prevent condensation/fogging.

Fig. 1. Electric Furnace for Aluminothermic Magnesium Production


We had difficulty in making temperature measurements. Sight tubes electrically insulated from the top lid and thoroughly prebaked to avoid fume deposition from carbon cement were helpful until the tube wall failed for reasons of arcing out or corroding through at the surface of the slag. There was a steep temperature gradient from the electrode to the frozen sidewalls.

Use of cooling on the condenser was discontinued because the grey product was very powdery. As our knowledge progressed, we added heating elements to the outboard portion of the condenser arm to heat it above the melting point of Mg, leading to recovery of small amounts of coalesced metal.

The starting oxide mixtures (slags) were prepared by premelting the materials, viz.,

Dolomite Magnesite Alumina

MgO 37.9 98.0 0.004

^ 3

0.38 0.26 98.5

CaO 57.4 1.2 0.03

Si02

0.49 0.46 0.03

NOI 0.34 0.11 0.9

in about 10 kg batches in a separate resistance furnace. Half the slags were initially 0.2 to 0.4 wt.% Mg and the remainder, 5 +.1.5 wt.%. These were always crushed, riffled for consistency in composition, and analyzed for MgO, CaO, A1203 and C.

The ratio of Ca0:Al203 of the slag, and that of the dolomite:magnesite of the charge was chosen according to three criteria:

Minimize the volume of slag or slag/Mg ratio. Maintain the fluidity of the slag. Maximize the use of the cheapest source of MgO, viz., dolomite.

Aluminum was charged in the form of weighed, 1 cm cubes sawn from EC grade busbar material. Another known amount of aluminum was added to the system since tubing was used to encapsulate the CaO.MgO charges in units of about 210 or 480 g.

The ratio of dolomite to magnesite was 1.72:1 in Experiments 2 to 4, and 2.3:1 for 5 to 10, giving an initial content of MgO, 62 and 58% in the oxides charges added to the starting slag.

Each test of the aluminothermic reduction was started by shorting the central electrode to the hearth (bottom current stub) on which was placed 10 to 15 kg of the initial premelted slag. The electrodes (run negative in Experiments 1-4, and then positive, 5-10) after preheating in contact at 1000 amp, 8-10 volts, was raised about 0.5 in. when, at 15-20 V, arc began melting the slag. Continuous heating was necessary to avoid freezing the starting pool of liquid which took 2 to 4 hours at 15-20 kW power input to melt. Then, the current was increased to 1200-1500 amp and charging began at a temperature of 1550 to 1600°C. Units of charge (Al + dolomite + magnesite) added at 10 minute intervals, was equivalent to 4 to 8 g Al/minute. There was an immediate reaction: evolution of fume which obscured the melt, accompanied by an increase in pressure, venting of Mg vapor through the overpressure vent, and a 1 to 2 volt increase in the potential drop across the electrodes.

After a few minutes, the fuming subsided, the voltage restored which we interpreted to mean that the exothermic reaction was finished. The current was controlled to an almost constant value during the sequential additions of charges. Power was varied by vertical adjustment of the electrode. In the first group of our experiments with the center electrode negative, the fast rate


of consumption of the electrode necessitated frequent electrode adjustment and checking of the liquid level. With the electrode positive, the delivery of consistent power was much easier and the electrode needed raising only to compensate for the increasing depth of the molten pool.

After tapping the slag, the crucible was cooled with argon flowing to impede the oxidation of the highly reactive Mg condensate. Despite these precautions, the powdery condensate often formed bright yellow Mg3N2, which was prone to hydrolyze to MgO and NH3.

Due to the unavoidable side reactions during cooling, and Mg vapor losses venting during the incremental charging and cumulative reaction process, collection of the exactly stoichiometric Mg product was not possible. In these experiments, magnesium vapor could not be prevented from condensing in and around the upper insulation of the furnace assembly. All forms of magnesium were recovered, their weight and composition determined for the best estimate of material balance.

With the central electrode negative, the slag wetted the crucible well and actually would creep up the crucible. Portions of this were analyzed chemically and by X-ray diffraction.

The major crystalline phase in the slag tapped after carrying out the reduction of MgO was 12 Ca0.7Al203.

RESULTS

Experiments in Externally Heated Crucibles

The experiments done in the induction heated crucibles served to prove initially that at atmospheric pressure, under Ar, aluminum charged directly to the surface of a molten slag (15 mol % MgO; 32 mol % AL03; 53 mol % CaO) produced magnesium metal at temperatures in the range 1520 to 1640°C. Minimum temperature necessary for observable Mg vapor evolution was 1475 ±40°C.

When Al was added, vigorous reaction began. The slag temperature, measured at the bottom of the crucible, dropped 30-50°C within the first two minutes of charging and restabilized within 10 minutes. Droplets of aluminum, 2-12 mm in diameter, decreased in diameter as they migrated slowly over the surface. From photographs of the melt surface for a 6.8 g addition of Al at 1630° C, the droplets were shown to be roughly hemispherical and that the rate of decrease in the radius of the largest droplet was linear with time (depending upon temperature, it took from 29 to 49 minutes for Al to disappear), -.03 cm/min. Adapting Rosenqvist's (1974) analysis for the kinetics of a reactant with diminishing surface area,

d = do - 2R/e t where d is the diameter of the droplet at time, t, R is the rate of reaction per unit area of slag-metal interface, e is the density of Al, 2.1 at 1630°C. Thus the photographic record gave an estimated reaction rate equivalent to the production of 2.5 grams Mg ernähr"1 (or 0.1 moles Mg cm^hr"1).

This low reaction rate confirmed our low efficiencies in these first experiments, 7-70%, due to Al vaporization and A14C3 formation. This was one reason we needed to design around the direct contact between slag-Al-Mg-carbon. Another was that with gas analysis, for CO, we proved that with the approximately 270 cm2 area of interface, slag - crucible wall, there was carbothermic reduction parallel to aluminothermic reduction, although the latter was by far predominant when ratio was about 1 to 18.


Experiments in Internally Heated Crucibles

In this configuration, possibly better simulating an eventual continuously charged process, the aluminum was in contact with the chemical source of MgO rather than in dilute solution in the slag, and the reaction rate was roughly two orders of magnitude (consuming up to 8 g Al/min. versus 6 g Al/hr) over the externally heated experiments.

Although the condenser captured the Mg vapor from the reaction zone, our design and operation were not suitable for coalescing or quantitative collection. Condensates were heterogeneous mixtures of fine, grey, highly-reactive powder, strongly crystalline material, and small metallic globules. During each experiment, our mass balance could never be perfect because some metal vapor would blow out during tending of the sight tube and charging ports. Magnesium condensate was also found in the gaps in the refractory in the upper lid and on the flange connecting the two main sections. Consequently, the aluminum conversion efficiency was determined from the amount of MgO reduced from the charge (as determined by careful recovery and chemical analyses) and the aluminum additions.

The average overall closeness for the Al mass balance given in Table 1 was 96.3% indicating probable losses of Al to the container and distillation to the condenser, quite good procedures for sampling and chemical analysis. For experiments 2-4, the condensate averaged 4.5% Al and 0.2% Al for Tests 5-10. When the center electrode was negative (Runs 2-4) the A14C3 was 5.1%, and only 0.3% when this polarity was reversed. The residual aluminum metallic content of the slags at the end of an experiment was very low, 0.1% Al or less.

Our balances for MgO or reaction efficiencies of Al are summarized in Table 2. The Al reaction efficiencies are inflated in Runs 2-4 due to the parallel carbothermic reduction and as also indicated by the high electrode consumption and presence of carbon in the condensates found for these (up to 5% C). Condensates from Experiments 5-10, when the electrode was positive, showed no carbon and the carbon balances suggest that the small amounts from the electrode were dissolved or incorporated into the slag.

TABLE 1 Aluminum Balances for Internally Heated Furnace Experiments

Al Charged (g) Al Recovered (g) Net Al Run No.

2 3 4

5 6 7

8 9 10

Initial Slag*

1351 2673 2634

2795 1997 2928

2173 1221 1295

Reduction Charge

988 736 714

741 741 1051

876 2891 1752

Total

2349 3414 3347

3536 2738 3979

3049 4112 3047

AsAl Metal

90 24 22

27 7 81

_ 64.2 46

AsAl 4C 3

in Slag

433 146 210

10 19 24

_ 95 31

As A1 2 0 3

in Slag

1919 3240 2292

3357 2594 3793

2939 3525 2890

Total

2242 3386 3224

3394 2620 3898

2939 3684 2967

Consumption

(g)

908 712 692

731 734 970

876 2827 1706

*Plus A1203 impurities in dolomite and magnesite.


TABLE 2 Magnesium Oxide Balances and Al Reaction Efficiencies for Internally Heated Furnace Experiments

Run No.

2 3 4

5 6 7

8 9

10

Starting Slag

20 570 453

41 29 26

19 300 331

MgO Charge (z)

Dolomite

747 673 726

874 874

1254

1045 3444 2081

Magnesite

1115 1003 1085

980 980

1405

1171 3861 2340

Total

1882 2246 2264

1895 1883 2685

2235 7605 4752

MgO Recovered (g) Tapped Dug

Slag Slag Total

12 54 18

569 703 629

1173 839

19 491 262

238 46

204

1221 539

31 545 280

807 749 833

679 2384 1378

Total MgO

Reduced

1851 1701 1984

1088 1134 1852

1556 5221 3374

Aluminum Reaction Efficiency

83.7 100.4 92.7

76.2 68.4 78.7

79.3 80.7 86.0

Electrode Consumption

Carbon usage from the central electrode was drastically affected by the electrode polarity. When we reversed the polarity, after Run 4, to positive, the use of carbon was lowered from 2.6 to 0.15 g/min. In terms of the total aluminum charged, the average consumption of carbon was 1.04 g C/g Al and 0.07 g C/g Al, respectively. The lowest electrode consumption, perhaps partly as a consequence of accumulating experience in managing these reactions, was in No. 10, 0.04 g C/gAl.

With the central electrode cathodic, of the major metal ions, Al3+, Mg2+, and Ca2+, in the oxide electrolyte, Al3+ is the most electropositive. The current density was about 15 amp cm"2, and most of the 20 V drop was right near the electrode, so metal ions had to carry the current. Conduction and discharge of Al would result in the rapid formation of A14C3. This rationalization is consistent with the higher concentrations of both A14C3 and C that were found in the slags with the electrode negative over when positive. Also, the slag tapped from the former set of experiments contained twice as much A14C3 as that slag forming the frozen layer or skull on the sidewalls. Conduction by Mg ions would result in electrolytic reduction to Mg vapor which would enhance and explain the higher values for conversion efficiency found with the electrode negative.

In the case of a positive electrode, the current may be carried by electrons or oxide ions, in any event, with orders of magnitude less attack.

The erosion rates observed were only 4% of those calculated for carbon acting as an anode generating CO at 100% current efficiency. Thus, we can conclude that the current is predominantly electronic. The cathodic current density on the walls of the crucible, where the electrolyte was in the solid, to certainly no higher than "plastic" state, was 1-2 amp cm"2 and the A14C3 (much reduced, of course) in that region was enriched by a factor of 4 over that in the bulk of the liquid slag.

Our carbon balances considered two sources of carbon: the carbon in the premelted slag, the carbon from the electrode, and two forms of carbon in the final slag - total carbon combusted to C 0 2 and A14C3 hydrolyzed to CH4. The balances were not perfect because CO was not measured.


At this scale, specific power is not relevant but the average was about 24 kWh/lb Mg reduced.

DISCUSSION

These experiments confirmed that it is technically possible to reduce MgO, sourced from dolomite and/or magnesite, two minerals easily obtainable at adequate grades, with aluminum at atmospheric pressure, at rates commensurate with temperatures in the range 1550-1660°C. Since an electric arc or resistance is essential, a process metallurgy based on this could be called "electroaluminothermic." At a larger scale, the electrothermal energy could be provided by resistance rather than partial or low voltage arc, and with understanding of power factors, a.c. might be more effective than d.c.

We did not put proportional effort in, nor did we solve the important problem of recovering useful, coalesced, Mg metal. The next scale, and any industrial reactor, will have to be extremely air tight for reasons of recovery and safety. From the little we did learn, the condenser will have to be heated and operate probably above the melting point of Mg, 649°C.

Aluminum is a very reactive metal and there are several parasitic reactions in this system. The main possibilities are reduction of impurities, vaporization reaction with carbon, and any air leaking in. The latter two can be controlled by reactor design and we have shown that by arranging for the slag itself (at an intelligently chosen composition) as the ultimate container, distance and slow diffusion should impede reaction with the carbon walls, an otherwise amenable material of construction.

The principal reducible impurities in dolomite and magnesite, Fe203, Si02, and H20, can be present in significant quantities, and they would result in the consumption of 0.34, 0.6, and 1 gram Al per gram of oxide feed, respectively, if completely reacted with Al. The reduction of silica by Al is thermodynamically possible for activities as low as 3X10"4. For examples in the mass balance for Experiment 10, Si02 and H20 reduction consumed 23 and 22 g Al, respectively, at complete conversion amounting to a 2.6% loss in the MgO conversion efficiency. Fe203 and Si02 are present in all dolomites, magnesitic dolomite, so specifications subject to economic constraints will have to be set realistically.

Reaction of Al to form A14C3 was to be expected since Al and C react rapidly above about 1200°C. However, when we switched polarity, the amount was not only greatly diminished because the carbon was no longer at the hottest part of the furnace, it was concentrated in the solid - mushy state at the container walls, accounting for a maximum of about 1.7% loss in use of Al as a reductant. In large furnaces with a well managed frozen skull, the carbon-Al contact area to Al-charge contact area would be small and the proportionate A14C3 greatly decreased. Alternating current would obviate electrolysis at the sidewalls.

Air leaks accounted for a major part of the inefficiencies in these small furnaces. The estimated loss in Run 10 was 6.2%. In a larger furnace in which any water cooled seals would not impact the heat balance so much, nor be so stressed as these ones were, air leaks could be reduced.

So the challenges in a scaled-up version of these experiments are managing the power input commensurate with the incremental charges of MgO and Al, to avoid the formation of A14C3, to keep the system at a very low oxygen partial pressure, to move the Mg vapor at atmospheric pressure down a thermal gradient so that useful Mg can be condensed. The furnace heat balance has to take into account keeping a frozen skull, the dynamics of a changing composition (and therefore, liquidus) of the slag for the aluminothermic reduction, and the considerable enthalpy, and possible radiant heat transfer to the condenser.


REFERENCES

Avery, J. M., U.S. Patents 3,761,247 (1973); 3,782,922 (1974); 3,994,717 1976.

Rosenqvist, T. (1974). Principles of Extractive Metallurgy, McGraw-Hill, New York. pp. 133-134.

257

Role of magnesium in the aluminothermic reduction of spodumene B. Jena, E. Mast and R. Harris Department of Mining and Metallurgical Engineering, McGill University, 3450 University Street, Montreal, Quebec, Canada, HS A 2A7

ABSTRACT

A process has been developed for the extraction of lithium from spodumene using excess molten aluminum as a reductant. The extraction was found to improve with the addition of magnesium to the excess reductant. The degree of extraction increased linearly up to 8 moles of magnesium addition per mole of spodumene reduced. It was also observed that the rate of reaction increased with increasing magnesium addition. The reaction rate decreased considerably after some time irrespective of the amount of magnesium addition. The reaction mechanisms are discussed and a kinetic model is developed.

KEYWORDS

Lithium, extraction, spodumene, aluminothermic reduction, molten metal reductant, magnesium addition, reaction mechanism.

INTRODUCTION

Lithium metal has found increased application in the aerospace, battery and nuclear power industries in recent years ' ' ' which has resulted in a growing demand for the metal. The present methods of metal extraction are costly and complicated . This warrants for the development of an alternative, cheaper and simpler process for lithium extraction. An alternative process has been developed in our laboratory for the extraction of lithium from spodumene, Li 0.A1 0 .4SiO-, by metallothermic reduction using

the "Melt leach" technique . In this novel technique, excess molten metal is used as a reductant and solvent. The metal recovered from the ore body is dissolved in the molten metal after reduction. The dissolved metal is separated from the excess reductant by vacuum distillation . Excess molten aluminum was chosen as the reductant to extract lithium from spodumene via the following chemical reaction :

Li 0.A1 O .4SiO + xsAl = {<xs - 6>Al + 2Li + 4Si} * (alloy)

+ 4A12°3 . (solid residue)

During the process development, it was observed that the extraction of lithium was improved with the addition of magnesium to the excess reductant. In this paper the role of magnesium in the reaction mechanism of lithium extraction will be discussed.



The reaction materials used in the experiment were :

- High grade spodumene concentrate from the pegmatite deposit of Bernic Lake, Manitoba.

- Commercial grade aluminum ingot, ALCOA 89775 AL, obtained from Alcoa, Pittsburgh.

- Commercial grade magnesium ingot supplied by Timminco Metals, Haley, Ontario.

The size and chemical analysis of the spodumene and the chemical analysis of aluminum ingot are shown in Tables 1 and 2, respectively. The experimental conditions are shown in the Table 3. About 2.2 kilograms of aluminum and approximately 370 grams of spodumene which corresponded to one mole of spodumene, were used for each experiment. The magnesium concentration was varied by adding magnesium ingot such that the magnesium : spodumene molar ratio was 0, 2, 4, 6, 8 and 10 : 1. In order to improve the reactivity of spodumene, α-spodumene was converted to ß-spodumene by heating the spodumene concentrate in a muffle furnace at 1050 C for 3 hours.

TABLE 1 Chemical and Size Analysis of Spodumene

size analysis

+ 95 % - 212 jam

Wt. %

Li20

7 - 7.5

Na20

0.16

κ2ο

0.06 - 0.15

Fe2°3

0.05

P2°5

0.15 - 0.2

TABLE 2 Chemical Analysis of the Commercial Aluminum Ingot

Wt. %

Si

0.033

Fe

0.029

C

0.004

Mn

0.001

Mg

<0.001

Zn

0.005

Ti

0.004

v

0.012


TABLE 3 Summary of the Experimental Conditions

Experimental Conditions

Reduction

Reduction

Protective Atmosphere

Temperature : 900 C

Time : 90 minutes

Gas : Argon

Amounts of reactants used for each experiment :

Spodumene

Aluminum :

Magnesium :

: Approximately 370 grams

(1 mole) after converting

a - spodumene to ß - spodumene

at 1050° C for 3 hours

About 2.2 Kgs

In the magnesium : spodumene

molar ratio of 0, 2, 4, 6, 8

10 : 1

A schematic diagram of the experimental set up for spodumene reduction is shown in the Fig. 1. The aluminum charge was cut into small pieces, placed in a bonded alumina crucible and was heated in a 100 kW induction melting furnace to 900 C. After the temperature of the melt reached 900 C, small pieces of magnesium ingots were added to the melt. When the magnesium had completely dissolved in the melt, the ß-spodujnene powder was charged on the top of the melt and the crucible was covered with a water cooled mild steel cover. To prevent oxidation, argon gas was passed through the reactor. The entire charge was thoroughly mixed with a stainless steel impeller coated with high temperature refractory cement. The mixing was carried out for 90 minutes at 900 C. When the mixing was stopped, the crucible was allowed to cool to approximately 750 C. The entire charge which now consisted of molten metal and residue solid was then poured in to a permanent mold. The residue powder which settled on the top was skimmed off.

Samples were drilled from different areas of the ingot and were analyzed to determine lithium, magnesium and silicon contents in the ingot by wet chemical analysis. An International Instruments atomic absorption spectrophotometer was used for the analysis. The residue powder was also analyzed to estimate lithium, aluminum and silicon content.

Samples were also collected from the melt at certain time intervals during reduction experiments and were analyzed for lithium.


SHAFT SUPPORT

ALUMNA CRUCBLE

W E L L EF DRIVE & SUPPORT VEHCLE

Fig. 1. A schematic diagram of the experimental set up. The residue powder was characterized by using X-ray diffractometry. The spodumene powder (both a and ß spodumene) and the residue samples were mounted in epoxy resin, polished and their microstructures were examined under optical and scanning electron microscopes. The chemical composition of the residue spodumene crystal was determined semi-quantitatively by using SEM - EDS. A Jeol JSM-300 scanning electron microscope with Tracor Northern micro analyzer TN 5400 was used for that purpose.


The percentage of lithium extracted into the molten excess reductant at various magnesium additions is shown in Table 4. The experimental lithium extraction values are compared with thermodynamically predicted values in Fig. 2. The thermodynamically predicted values were obtained by using equilibrium program of the F*A*C*T . The experimental values agree well with the predicted values. It is also seen from the Fig. 2 that the degree of lithium extraction increased from 0.4% to 53% with the increase in magnesium to spodumene ratio from 0 to 10. The beneficial effect of magnesium addition on the lithium extraction is due to thermodynamic factors. From the Ellingham diagram, Fig. 3, it is seen that reduction of lithium oxide is thermodynamically more favorable with magnesium than with aluminum. The addition of magnesium is also believed to increase lithium extraction by decreasing surface energy of the molten aluminum . The latter effect resulted in better powder - melt contact due to improved wetting. The improvement in the wetting was evident from the clusters of metallic aluminum coated with residue spodumene powder at the outer surface which were found in the residue at higher magnesium additions.

Silicon was also found to be dissolved in the ingots. The silicon concentration in the ingots is shown in the Table 5. It can be seen that silicon was extracted from spodumene even with no magnesium addition. Some metallic silicon was also detected in the powder residue by XRD and micro analysis. A summary of the major constituents of the solid powder residue at different magnesium addition levels is shown in Table 6. Since reduction of silica is thermodynamically favorable and metallic silicon was observed in the


TABLE 4 Lithium Extraction at Different Levels of Magnesium Addition

Magnesium Addition ( Moles )

0

2

4

6

8

10

Lithium Extraction (%)

0.4

8.4

25

37

48

53

70

s I X Φ

30

20

10

A predicted

■ experimental

Mg, moles

Fig. 2. Lithium extraction as a function of magnesium addition; comparison of experimental values with thermodynamic predictions.


Fig. 3. Comparison of free energy - temperature curves of different oxides

residue, it appears that the silica present in the spodumene was reduced to silicon but a part of the reduced silicon did not dissolve in the melt. The distribution of silicon between the ingot and residue as a function of magnesium addition is shown in Fig. 4. It is seen that as the magnesium to spodumene molar ratio was increased from 0 to 4, the concentration of silicon increased in the ingot with a decrease in the silicon concentration in the residue. At higher molar ratios of magnesium to spodumene, little variation in the silicon concentration is noticed due to magnesium addition.

From the summary of the constituents of the residue powder in Table 6 it is seen that the reaction product was mainly a spinel (MgAl 0.) and unreacted

spodumene at lower magnesium additions. At higher magnesium additions, the reaction product also contained some amounts of magnesium oxide along with the spinel. This indicates that most of the silica was reduced by aluminum at low magnesium concentrations. As the concentration of magnesium increased in the melt, increased proportions of silica were reduced by magnesium due to its increased activity as opposed to be reduced by the aluminum. This also indicates that a major proportion of the magnesium was used for silicon extraction. Therefore only a small amount of magnesium was available for the extraction of lithium.

From the above observations, the extraction of lithium from the spodumene is envisaged to occur according to the following chemical reaction :

Li 0.A1 0 .4SiO + aMg + xsAl -

{<xs + x - 2y - 2u>Al + xLi + wSi + <a - y - z>Mg} (alloy

) + [yMgO.AL 0 + zMgO + <2-x>LiAlO + <4-w>Si + uAl 0 ]

(solid residue)


Where the values of u, v, w, x, y and z are dependant on the magnesium concentration, a.

Aluminum appears to be playing a minor role in the extraction of lithium as a reductant. With no magnesium addition, a negligible amount of lithium was extracted. Lithium was mainly extracted by magnesium.

TABLE 5 Silicon Concentration in the Ingot at Various Levels of Magnesium Addition

Magnesium addition ( Moles )

0

2

4

6

8

10

Silicon concentration in the ingot (Wt. %)

2.7

1.5

3.8

3.5

3.7

3.0

TABLE 6. Summary of the Constituents Found in the Powder Residue as Detected by XRD

Mg addition (moles)

0

2

4

8

10

Constituents

Spodumene (LiAlSi 0.) , Al 2 6 1

Spodumene, MgAl 0 , Al

Spodumene, MgAl 0 , Al, Si

Spodumene, MgAl 0 , MgO, Si, Al

Spodumene, MgAl 0 , MgO, Si, Al


c

Φ o c o o <7>

A in the residue

■ in the Ingot

8 10

Mg, moles

Fig. 4. Distribution of silicon in the ingot and residue as a function of magnesium addition.

Mineralogical Study

The typical microstructures of a and ß spodumene particles are shown in Figures 5 and 6, respectively. A comparison between the microstructures shows that the crystals of a and ß spodumene are different. While the a-spodumene crystals are dense in nature, ß-spodumene crystals are porous. The ß-spodumene crystals appear to have numerous intra granular cracks which are believed to have formed during the a to ß transformation. The a-spodumene crystals are monoclinic and these crystals undergo polymorphic transformation to tetragonal structure during the a to ß conversion. This transformation is accompanied by anisotropic volume expansion and stress development(8> which results in cracking of the crystals and an increase in porosity. The volume expansion was also observed during α-β conversion experiments. A significant increase in the volume of the original α-spodumene powder was noticed and in some cases the pressure due to the expansion caused cracks in the crucible containing the powder.

The high reactivity of the ß-spodumene is attributed to the high porosity in the crystals. Due to pore formation the surface area is increased and the transport of the reactants inside the crystal is facilitated. This resulted


in increased reaction rate. Some dense crystals were found to be present in the ß-spodumene specimen and were thought to be α-spodumene. This indicates that complete α-β transformation might not have been achieved.

Typical microstructures of the residue powder with 10 moles of magnesium addition are shown in Figures 7 and 8. The residue is seen to contain metallic aluminum randomly distributed among the spodumene particles. A reacted spodumene grain embedded in the metallic aluminum is shown in Figure 8. This grain was apparently trapped in the molten metal during solidification. The pores of the grain are seen to be filled with the aluminum. This indicates that the melt flowed to the interior of the spodumene grain through the pores due to capillary action and the reaction took place deep inside the grain.

Some metallic silicon was seen to be present in the embedded grain. It is likely that since amount of melt flowing through the pores of grain was small in quantity, it became saturated with dissolved silicon upon reaction. Therefore, some metallic silicon might have precipitated.

Semi-quantitative micro analysis of the powder residue grains was carried out using SEM-EDS. The chemical composition of the grain measured at many different locations were variable but, the assays could be grouped into three distinct sets whose average compositions are given in the Table 7. The presence of magnesium throughout the grain indicates that there was some transport of magnesium into the grain from bulk excess reductant. No conclusion was drawn to the manner and mechanism by which the magnesium was finally distributed in the reacted grain. It is likely, however, that the regions corresponding to the high Mg contents were also those giving rise to the spinels with X-ray diffraction.

Kinetic Studies

Samples were collected every 20 minutes from the melt and were analyzed for lithium content. The amount of lithium present in the melt after a certain time is shown in the Table 8. The lithium extraction as a function of time and magnesium concentration in the melt is shown in the Fig. 9. It can be seen from the Figure that the rate of lithium extraction was high during the initial period. The extraction rate slowed down considerably after some time. It is also seen that the initial rate of extraction was increased due to the presence of higher amount of magnesium in the melt. At the magnesium to spodumene ratio of 10 : 1 the rate of extraction during the initial period was the highest. However, the decrease in the extraction rate after a certain period of time was also most noticeable at this level of magnesium addition.


Fig. 5. Typical microstructure of a-spodumene. (a = a-spodumene, E = epoxy)

Fig. 6. Typical microstructure of ß-spodumene. (a = a-spodumene, ß = ß-spodumene, E = epoxy)


Fig. 7. Typical microstructure of a residue powder reacted with 10 moles of magnesium. (Al = aluminum, Sp = spodumene, E = epoxy)

Fig. 8. The microstructure of a residue grain found trapped in the aluminum. (Al = aluminum, Sp - spodumene, Si = silicon, E - epoxy)


TABLE 7 Average Chemical Composition of the Residue Grain (From Micro Analysis)

Mg

(Wt.%)

20

8

3

Al

(Wt. %)

54

63

69

Si

(Wt. %)

26

29

28

TABLE 8 Lithium Concentration in the Melt at Different Time Intervals

Mg Addition (moles)

2

6

8

10

Lithium concentration (Wt. % )

After 20 minutes

0.05

0.14

0.21

0.25

After 40 minutes

0.09

-

0.29

0.26

After 60 minutes

0.05

0.21

0.31

0.3

Based on the kinetic data, observations under optical microscope, micro analysis results, and XRD data following reaction steps are postulated :

1. Transport of reactants (Al and Mg) from the bulk liquid phase to the spodumene grain - liquid interface.

2. Reaction of Mg and Al with silica. 3. Transport of the silicon away from the interface. 4. Dissolution of the silicon in the melt. 5. Formation of the spinel MgO.Al 0-.

6. Transport of Mg and Al into the spodumene grain. 7. Reaction of LiAlO with magnesium.

8. Transportation of the reduced lithium in to the bulk liquid phase. 9. Dissolution of the lithium in aluminum.


i

$

i Ώ

0 20 40 60

Time, minutes

Fig. 9. Lithium extraction as a function of time and magnesium concentration in the melt.

Among the above mentioned steps, chemical reaction of magnesium with the oxides at the interface appears to be the rate limiting step. Due to the presence of intra granular pores in the ß-spodumene grain and thorough mixing of the charge, the transportation of reactant to the reaction interface was very fast. As the reaction occurred at the interface, the concentration of magnesium decreased. Since the magnesium concentration in the melt was finite, the chemical potential of magnesium decreased after some time as the magnesium was consumed. Therefore, the rate of reaction decreased considerably. As magnesium : spodumene ratio was increased, the initial magnesium concentration in the melt increased. This increased the rate of reaction due to a higher chemical potential of magnesium. Thus the initial reaction rate was highest for the magnesium :spodumene ratio of 10 : 1.

CONCLUSIONS

From the present study, following conclusions are made :

1. The extraction of lithium from the spodumene is improved by magnesium addition. The degree of lithium extraction increases with increasing amounts of magnesium addition.

2. Lithium extraction is accompanied by extraction of Silicon. 3. Silica reduction is an important factor in lithium extraction as it is

preferentially reduced by magnesium. 4. At 900°C, Aluminum plays a minor role in lithium extraction as a

reductant.

0.4

0.3

0.2

0.1

Λ

— B - 2 motes Mg

Δ 6 moles Mg

—■— 8 moles Mg

- A - 10 moles Mg ^ ^ ^ ^ ^ ^ ^ ^ ^

- /F\^^^ - I/^^^_ ' ]>^^^^^~^.


ACKNOWLEDGEMENTS

The authors would like to express their gratitude to Tanco Ltd. , Manitoba for supplying spodumene concentrate for the experiments and carrying out a part of the chemical analysis. The financial support provided by the NSERC through Strategic Research Grant to carry out this work is also gratefully acknowledged.

REFERENCES

1. Sanders T. H. and E. S. Balmuth (1978). Aluminum - Lithium Alloys : Low Density and High Stiffness, Metal Progress, 113, March 1978, p 32.

2. Ober J. A. (1986). Lithium, Minerals Year Book, United States Department of Interior, Bureau of Mines, pp 619-625.

3. Grady H. R. (1980). Distribution, Consumption, Pricing and Outlook of the Lithium Industry, 1st International Al-Li Conference, Stone Mountain, Georgia, May 19-21, 1980, pp 1-8.

4. Mahi P. , A. A. J. Smeets, D. J. Fray and J. A. Charles (1986). Lithium - Metal of the Future, Journal of Metals, November, 1986, pp 20-26.

5. Mast E. , R. Harris and J.. Toguri (1989). Lithium Extraction from Spodumene by Metallothermic Reduction, To be presented at the International Symposium on Productivity and Technology in the Metallurgical Industry, Cologne, FRG, September 17-22, 1989.

6. Harris R. , J. Toguri and A. Wraith (1988). Producing Volatile Metals, U.S.Patent Application S. N. 201,446, Applied on June 2, 1988.

7. Richardson F. D. (1974). Physical Chemistry of Melts in Metallurgy, Volume 2_, Academic Press, London, p 430.

8. Deer W. A. , R. A. Howie and J. Zussman (1978). Rock Forming Minerals, Volume 2A, Single Chain Silicates, Longman, New York, pp 526-545.

9. Thompson W. T. , A. D. Pelton and C. W. Bale (1988). F*A*C*T, Facility for Analysis of Chemical Thermodynamics, CRCT, Centre for Research in Computational Thermodynamics, Ecole Polytechnique, Montreal.

271

The preparation of Nd metal from Taiwan black monazite Y.W. Miao, J.S. Horng and Y.C. Hoh Institute of Nuclear Energy Research CAEC, Lung-Tan, Taiwan, Republic of China

ABSTRACT

There are about a million tons of heavy sand deposited along the south-west and north-west coasts of Taiwan. Among the deposits,ten percent of the sand is black monazite. The institute has developed a process to recover the individual rare earths from the local monazite. Nd203 content in the local black monazite is 18.68 %. Neodymium metal is one of the raw materials to prepare Nd-Fe-B supermag-net. A process has been developed to prepare metallic neodymium. Anhydrous neodymium chloride was first prepared by reacting neodymium oxide with ammonium chloride. Results indicated that 99 % of Nd203 is converted to NdCl3 . Neodymium chloride was then dissolved in molten alkali chlorides in a graphite cell. Finally, the metallic neodymium was obtained by molten salt electrolysis. Batch process was conducted and an average yield of 50 % was obtained.

KEYWORDS

Black monazite; Neodymium oxide; Neodymium chloride; Metallic neodymium; Molten salt electrolysis.

INTRODUCTION

Along the south-west and north-west coasts of Taiwan,there are about a millon metric tons of heavy sand deposited were found and about 10 % of the sand is black monazite. The Institute has developed a separation process to recover the individual rare earths successfully. The developed process has been commercialized by a local private company. Neodymium oxide is about 18 percent weight of the total rare-earth oxide contained in the black monazite. Table 1 shows the composition of the Taiwan black monazite.


TABLE 1 Composition of Taiwan Black Monazite

Item Composition (wt. %)

ΊΉΕΟ

P 2 O 5

S1O2

Th0 2

U 3 0 s

4 8 . 6 2

2 0 . 1 4

1 8 . 6 6

0 . 4 1

0 . 2 8

REO/TREO(%)

Y 2 O 3 0 . 9 1

Ce0 2 4 7 . 9 9

La 2 0 3 2 1 . 0 3

N d 2 0 3 1 8 . 6 8

Pre Oil 5 . 4 2

Sm 2 0 3 3 . 2 9

Eu 2 0 3 0 . 5 4

G d 2 0 3 1 . 6 3

T b i O ? 0 . 1 9

D y 2 Ü 3 0 . 3 5

H 0 2 O 3 0 . 0 3

Er 2 O 3 0 . 0 3

Y b 2 0 3 0 . 0 7

Rare-earth metal alloys are becoming increasingly important, particularly samarium and neodymium, because of their magnetic properties. In the mid of 1960s, SmCo5 was the most powerful magnet material with an energy

product ( BHm ) of 143 kj/m3. However, cobalt is quite expensive and only in a limited amount of supply. Fe like Co, has a large magnetic moment but either have high magnetocrystalline anisotropy. It is found that the compounds formed from transition metals and rare earths can improve the anisotropy energies, while boron solves the disadvantage of iron in its in-ability to form many compounds with rare-earths. In the beginning of 1984,

the highest BHm value of 341 kj/m3 for Nd-Fe-B magnet is obtained both on technological and economic aspects.The most likely tendency is the develop-ment of a family of magnets with varying compositions with Nd-Fe-B magnets at one extreme and Sm-Co magnets at the other (1). The production of magnet alloys by the R/M process requires a rare earth metal feedstock that, in addition to being of a size suitable for remelting, contains at least as much rare earth as desired in the alloy being made.The most direct approach is to use "pure" rare earth metal(2). There are two basic routes to prepare metallic rare earth alloys adapted commercially. These two major routes are based on rare earth oxides as the


starting material but they differ in the reduction step. Oxides reduction can be carried out either by calcium as the reductant ( Calciothermic reduction ) or by fused salt electrolysis techniques(3). In molten salt electrolysis, the electrolytes are multicomponent melts of either chlorides or fluorides. Chlorides offer the advantages of a lower operating temperature a greater choice of electrode and container materials, suitability for continuous operation and lower cost of electri-city when compared to metallic reductants (2,3). The only disadvantage is the extremely hygrosopic of chlorides(4). In this study, the anhydrous NdC13 was first prepared from Nd203 through the chlorination of NH4C1 and then fed into the fused LiCl and KCl melts to prepare Nd metal by electrolysis method(5). The results of the preliminary laboratory scale study to prepare neodymium metal by molten salt electrolysis method were presented. Batch operation was carried out and an average conversion rate of 50 % was obtained. The preparation of anhydrous neodymium chloride was also discussed.

EXPERIMENTAL

Preparation of Anhydrous Neodymium Chloride

Neodymium oxide with a purity of greater than 99% was supplied by INER. All other chemicals used were of industrial grades. In this investigation a kilogram order bench scale chlorinator was used to prepare NdCl3 powder. Nd203 was mixed intimately with an excess amount of NH4C1 by a v-type blender. After blending, the mixture was put into a 10-liter pyrex beaker which was then placed in a stainless steel cylindric chlorinator(o.d♦= 51 cm ). Two thermal couples were placed at the different places to control and record the temperatures. The temperature

were raised up by electric power to 250 °C and the chlorinator was also purged with an inert gas. It took about 10 hours to convert neodymium oxide to chloride form.When there was no more water vapor and ammonia odor vent-ed from the ventilation system, it indicated that the chlorinating reaction

has completed and the temperature is up to 350 °C. This temperature was set by a programable temperature controller to perform the purification step. After another 10 hours, the reactor was cooled down to room temperature and the beaker was removed from the reactor to collect anhydrous NdC13 powder. Samples were taken to analyse the purity of the product by dissolv-ing it into water. After filtration, if there was no undissolved solid remained, the solution appeared a very clear and transparent violet color. It indicated that nearly a 100 % conversion of Nd203 to NdCl3 was achieved. The purity of NdCl3 could be calculated by titrating the filtrate in terms of rare earth concentration. Meanwhile a sample was sent to the analytical group to test the purity by ICP as a cross check of the purity of the NdCl3 . Fig. 1. shows the block diagram of anhydrous NdC13 pre-paration .


NH4CI -> Nd 2 0 3 -»

Inert Gas ->

Inert Gas ->

BLENDING

4,

CHLORINATION 250 °C

I

PURIFICATION 350 °C

Vent ΐ

-> Water Vapor -> NH3 Inert Gas

t 1

WATER

SCRUBBER

Anhydrous Neodymium Chloride <- ANALYSIS

Fig. 1. Flow Sheet of Anhydrous NdCl 3 Preparation

Molten Salt Electrolysis

Industrial grade LiCl and KCl with a weight ratio of 1.5 was mixed intimately and put into a graphite cell. When the temperature of the mixed

salt melts raised up to 400 °C by an electrical heater, the cathode(Mo rod) was inserted into the melts . The reaction temperature began to raise up to

900°C, at this time, anhydrous NdCl3 was fed into the cell gradually. After anhydrous NdCl3 dissolved into the melting salt completely, D. C. power supply was turned on to proceed the electrolytic reaction. A yellow green C12 gas was released to the scrubber. This indicated that the electrolytic reaction was proceeding. At the end of the electrolysis, the

temperature of the cell was raised up to 1050 °C and the produced Nd sponge melting together to end up with the Nd ingot. The cell then was cooled down

to 700°C and the cover together with the cathode was removed from the cell. The Nd ingot form was obtained from the solid salt by leaching it with water. The collected Nd ingot was submerged into petroleum in order to pre-vent it from oxidation. The prepared Nd metal was then sampled to analyze the purity by ICP method.


Preparation of Anhydrous Neodymium Chloride

The preparation of anhydrous neodymium chloride is of considerable impor-tance due to its wide use in neodymium metal production both by electrolyt-ic or metallothermic reduction method. In general, anhydrous neodymium chloride can be obtained either by dehydration of the hydrated chloride using a dehydration reagent(6) or by direct chlorination of the oxide using a chlorinating agent (3). However, for dehydration method, time consuming and difficulty in temperature control limited this method for a plant scale use.The direct conversion of oxide to chloride is by passing H d or C12 gas stream over a hot intimate mixture of neodymium oxide. The metal product obtained by this method was contaminated by carbon (7). Among the various


chlorination methods, Hopkins (1935) (8) first developed a rather simple method to prepare anhydrous NdCl3 by heating a mixture of rare earth oxides

and an excess amount of ammonium chloride at a temperature of 200 °C or higher. The overall reaction can be expressed as (9,10,11):

N d 2 0 3 + 6 NH4C1 -» 2 NdCl3 ( s > + 3 H 2 0 ( 9 > + 6 N H 3 ( 9 > (1)

This method has been extensively used by Lunex Corporation (12) to prepare anhydrous rare earth chlorides for metal production.The process parameters, such as,the temperature for chlorination and purification, the weight ratio of Nd203 to NH4C1 , and the time for chlorination and purification etc.were studied intensively. Table 2 shows the experimental results.

TABLE 2 Anhydrous NdCl Preparation Results

Nd 2 O 3 / Chlorination Purification Yield

Run No. NH4CI

(Weight) Temp.,°C Time,hrs Temp.,°C Time,hrs %

1 2 3 4 5 6 7 8 9 10 11

1/2 1/2.5 1/3 1/3 1/3 1/3 1/3 1/3 1/3 1/3 1/3

150 200 200 200 250 300 350 250 250 250 250

4 8 10 10 10 10 10 10 10 11 10

250 300 350 400 350 350 350 350 350 350 350

8 10 10 10 10 10 10 10 10 10 10

78 84 91 80 94 74 70 92 96 99

> 99

The table shows that because of the sublimation nature of NH4C1 , an excess amount of NH4C1 should be used in order to get a high yield. The optimum weight ratio of Nd203 to NH4C1 is 1 to 3.

Electrolysis of the Fused Chloride

The decomposition voltage of neodynium chloride was found to be 8 volts and the operation current varied from 110 to 140 amperes. When the current was below 110 amperes, Nd ingot can not be obtained. However, while the current was over 140 amperes, the molten salt was boiling rigorously and flooding over the cell. Table 3 collected the experimental data of electrolysis. The results indicate that the best yield of the runs is 50 %. A subchloride of neodymium with a formula of NdCl2 might be formed especially in an earlier electrolytic stage(7).Therefore, it is recommended that the electrolytic operation should be kept a longer time in order to reduce this subchloride to metal form. This could be seen from the last eight runs in Table 3. The longer the reaction time the higher the yield was obtained. After the electrolytic operation, the temperature of the cell was raised up to over the melting point of Nd in order to arrive an ingot form. This caused a part of the Nd sponge dissolved into the melts again at this high temperature. Thus a relatively lower yield was resulted(2).


TABLE 3 Experimental Runs of Electrolysis

Run Current Density Time Temperature Yield Remarks

No. Amp/«2 hrs °C %

1 2 3 4 5 6 7 11 12 13 14 15

4.3 8.6 11.8 11.0 9.4 9.4 9.4 9.4 9.4 9.4 9.4 9.4

7.0 2.0 2.0 1.5 2.0 2.0 2.6 3.0 3.0 2.5 2.5 1.5

900 960 890 1000 900 900 900 905 920 910 937 926

6 16 31 21.5 40 47 50 36.5 42 46 47 26

Free of Iron Powder Sponge Sponge Ingot Ingot Ingot Ingot Ingot Ingot Ingot Ingot

Form Form Form Form Form Form Form Form Form Form Form

A graphite cell was used in all experimental runs in order to avoid the iron contamination. However,it is inevitable to get minor neodymium carbide . The final Nd ingot contained 115 mg/ml of chlorine. A further vacuum dis-tillation should perform in order to get pure metal. The purity of the produced ingot was sampled and analyzed by ICP. The results are shown in Table 4. The average purity of Nd is 94.6 %, the iron contamination came from the cover, vent and feeding pipings of the cell. As for the Al and Mg contamination, probably came together with the industrial KCl and LiCl. Thousand ppm order of molybdenum involved in Nd ingot came from the moly-bdnum cathode. While the potassium and lithium are very few contaminated the Nd ingot. The impurity of other rare earth elements is depended on how pure the Nd203 starting material is. As for the Nd-Fe-B magnet purpose 90% purity of Nd203 will meet the need.

TABLE 4 Purity Analytical Data of Nd Ingot

Run Nd K Li Fe Mo Al Mn Mg Cr No. % % % % % % % % %

11

12

13

14

15

16

94.03

97.25

96.44

89.43

98.98

91.53

0.002

0.015

0.015

0.009

0.04

0.11

<0.001

0.002

0.002

0.001

0.003

0.02

0.9

0.8

0.4

0.7

0.9

1.15

0.23 **>*

1.42

0.19

— 0.52

~~ 1.02

— 0.12

0.18 ^

0.23

0.006

0.05

0.2

0.04

0.03

'"VX -"N-/·

0.016 —

0.03 —

0.11 —

0.34 0.74

0.46 —

PRODUCTION AND ELECTROLYSIS OF LIGHT METALS 277 CONCLUSIONS

The preliminary study on the molten salt electrolysis of the neodymium chloride to produce ingot neodymium was reported. The results indicated that in the stage of preparation anhydrous neodymium chloride, the optimum

chlorination and purification temperatures were in the range of 250 °C to

350 °C. Whereas, the optiinum electrolytic temperature was 900 °C and the optimum current and voltage is 110 to 140 amperes and 8 volts,respecti-vely. The best yield is 50%. The average purity of the Nd ingot produced is 94.6 %. For solving the problem of low yield, recently, Li(1987) (15) suggested that the electrowinning-distillation method to produce Nd from Nd-Mg alloy is adapted. This alloy was first obtained by fused chloride salt electrowinning and followed by the Nd-Mg alloy distillation process. Chambers ( 1989 ) (16) used the molten-metal cathode of Mg-Zn,Mg-Cd from a molten-chloride electrolyte to electrowin the Nd alloy first and followed by distillation. The electrowinning of Nd-Fe alloy by using the consumable iron cathode has been used commercially ( Ronson, U.S.A. (3); Sumitomo Light Metal Industries,LTD.,Japan). However, the most direct approach is to use the pure Nd metal to achieve the desired Nd-Fe-B magnet alloy. Because of the abundance of Nd and the lower cost of alkali salt compensate the econo-mic lost of low yield. However, A further study to achieve of a high yield is a state of art and an attractive challenge.

ACKNOWLEDGE1MENTS

The author wish to thank Mr. Yuan, Mr. Hsue, Mr. Chen, Mr. Hwang for their experimental works. Thanks are due to the Analytical Group of INER for the analytical data and also due to Mr. Lu for his kind supply of the raw material for these experiments. The permission to publish this paper by INER is also appreciated.

REFERENCES

(1) .Unknown(1987). Rare-Earth Information Center News,Vol. x x n ,1. (2).Jones, F. G.( 1987).9th International Workshop on Rare-Earth

magnets and their Applications, Bad Soden, FRG. 737-751. (3).Agarwal, J. C.,J. M. Loreth and F. E. Katrak (1989).Rare Earth

Extraction, Preparation and Application,R.G.Bautista and M.M.Wong, A Publication of TMS, 281-289.

(4). Sadoway,D.R. (1989) .Rare Earth Extraction,Preparation and Application R.G.Bautista & M.M.Wong,A publication of TMS, 281-289.

(5). Miao, Y. W.,S. L. Chern and Y. C. Hoh (1988). Proceeding of 4th Annual Symposium on Magentism and Magnetic Materials,85-98.

(6). Kleinheksel, J. H. and H. C. Kremers (1928). J. Am. Chem. Soc.,50, 959.

(7). Kremers, H. C. (1925) .Trans. Am. Electrochem.Soc. ,47,365. (8). Hopkins, B. S.(1935).J. Am. Chem. Soc. ,57,1159. (9). Reed, J. Β.,Β. S. Hopkins and L. F. Andrieth (1939). In Inorganic

Synthesis, Vol. I,J. B. Reed McGraw-Hill.28. (lO).Remeika, J. P. (1956). J. Am. Chem. Soc. ,78,4259. (11).Speeding F. ~H. and J. E. Powell (1952) .J. Am. Chem. Soc,74,856. (12).Moriarty, J. L.(1988).J. of Metals,20,41. (13).Kim Y. S., F. Planinsek, B. J. Beandry & K. A. Gschneidner,Jr. (1980)

The Rare Earths in Modern Science and Technology,Vol.2,G.M.Mc-

278 PRODUCTION AND ELECTROLYSIS OF LIGHT METALS Carthy, J. J. Rhyne and H. B. Silber (Eds.), Plenum Publishing Co.53-58.

(14) .Kennedy(1984) .Nd-Fe Permanent Magents: Their Present and Future Applications,I. V. Mitchell,Elsevier Applied Science Publishers, London and New York.41-48.

(15).Li, Z., G. Yang and Zongan Li(1987).9th International Workshop on Rare-Earth Magnets and Their Applications,Bad Soden,FRG, 325-303.

(16) .Chambers M. F. and J. E. Murphy(1989) .Rare Earths, Extraction, Preparation and Applications,R. G. Bautista & M. M. Wong, Ed. by TMS, 369-376.

(17).Singh, S. and J. Balachandra (1973).J. Electrochem. Soc. India,22 -3,222-225.

Production and Electrolysis of Light Metals. Proceedings of the International Symposium on Production and Electrolysis of Light Metals, Halifax, August 20â€“24, 1989

Documents