CVDLabReport

17 February, 2015

TO: Professor A. Drews, PhD

FROM: Group B4: Brandon Sanchez, Saman Hadavand, Janet Mok, Liliana Busanez

SUBJECT: Chemical Vapor Deposition Reactor, Final Report

Attached is the final report and our recommendations concerning the Low Pressure Chemical Vapor

Deposition Reactor. This report includes a study on the effect that temperature, pressure, wafer spacing,

inlet mole fraction on the uniformity of silicon deposition on wafers throughout the reactor. We hope this

report will satisfy the desired expectations. If you have any questions or concerns, please contact us.

Sincerely, Group B-4 Saman Hadavand, Part I Brandon Sanchez, Part II Janet Mok, Part III Liliana Busanez, Part IV

4

Chemical Vapor Deposition Reactor Saman Hadavand-Part I, Brandon Sanchez-Part II, Janet Mok-Part III, Liliana Busanez-Part IV

The goal of the experiment was to understand the characteristics and design of a chemical vapor deposition (CVD) reactor, as well as to evaluate the chemical kinetics of silicon deposition. A CVD reactor was designed using the Comsol simulation. The effects of the wafer bundle characteristics, transport properties, and operating conditions were evaluated based on their effects on deposition rate, uniformity, and the effectiveness. It was inferred from the data that temperature and pressure had an inverse relationship with uniformity and wafer spacing. While inlet velocity and silane mole fraction had a direct relationship to uniformity within the CVD reactor. It was concluded that temperature had the greatest effect on deposition rate, since a 5% increase in temperature showed a high average deposition rate of 0.086 nm/s.

Advisor: Professor Aaron Drews

17 February, 2015

1

1 Introduction

Chemical vapor deposition (CVD) is a process used to synthesize thin films of various materials

onto surfaces through chemical reactions between gas molecule.1 Low pressure chemical vapor deposition

(LPCVD) is a CVD process that is used to synthesize deposition of thin films on semiconductors that

range from the nanometer and micrometer scales.2 CVD has a history of many different applications, such

as its use in the electric lamp industry in 1880 in which a CVD process was used for the deposition of

various metals in order to produce stronger lamp filaments.3 Another notable historic CVD process is the

deposition of silicon by hydrogen reduction of SiCl4, where from 1909 through 1927 thin silicon films

were used in the electronic industry of silicon-based photo cells and rectifiers.3 In today’s industry CVD

processes are used for applications such as hard coating on metal cutting tools, modern computer chips,

and coatings for window glass.1

CVD is a growing field for researchers. A popular research topic in the field is plasma-enhanced

CVD (PECVD). PECVD is a method used to deposit thin films on wafers by adding plasma in a

deposition chamber with reactive gases to create the necessary surface on a substrate.2 Research in this

field focuses on developing higher ionized PECVD by applying plasmas with higher densities.4

Researchers are improving this process through experimenting with many factors such as deposition

conditions. Deposition conditions can be designed to provide a higher degree of energy for the growing

field.5 The high level of energy is important for specific types of films such as diamond-like carbon.5 In

recent years CVD has played an important role in breakthroughs in two-dimensional (2D) nanomaterials,

more specifically in transition metal dichalcogenides (TMD).6 TMD’s are important because of the

special properties that make it an untapped source of 2D nanomaterials.6 These semiconducting TMD’s

are very important because they have shown to be feasible for future electronics.6 CVD has shown great

promise to generate high-quality TMD layers with a scalable size, controllable thickness, and excellent

2

electronic properties.6 These are some of the advances being made in the field of CVD reactors and

research breakthroughs in the application processes of CVD reactors.6

For this experiment, a LPCVD reactor was designed in order to optimize the deposition of silicon

onto wafers in a three-zone boat reactor. Variables were manipulated such as: temperature, pressure, inlet

velocity, wafer spacing, and inlet mole fraction of silane to study the deposition rates in the reactor.

2 Background & Theory

CVD is a process commonly used to decompose silane gas to solid silicon via reaction on wafer

surfaces within a low pressure vessel, usually around 100 Pa.7 A thin film of silicon is deposited onto the

wafers which are uniformly distributed throughout the reactor in series.7 Hydrogen gas is also produced in

the decomposition reaction as displayed by Eq 1.

𝑆𝑖𝐻! → 𝑆𝑖 𝑠 +  2𝐻!(𝑔) (1)

A schematic of the boat reactor is displayed in Figure 1. The reactor contains three different

regions which may have varying temperatures. In Figure 1, YAA represents the silane gas mole fraction

flowing through the annular region, YA represents the mole fraction of reacting gas between the wafers,

WAr represents the mass flow rate flowing back into the annular region, and RW represents the wafer

radius, which will increase with increasing silicon deposition.7 The inlet gas feed flows into the reactor

and is absorbed onto the wafer bundles. At the wafer surface, the silane reacts to form solid silicon. The

remaining gas then flows out of the exhaust.

3

Figure 1: Boat reactor. Wafer discs are uniformly distributed within the horizontal reactor. Vapor feed gas is fed to

the reactor on the left and reacts at the surface of wafers to form a thin film of solid product. Exhaust leaves at right.

The following assumptions are made in this simplified CVD boat reactor model. Only an

isothermal reactor is considered. The density of silane is assumed constant, mass transport of silane only

by diffusion is considered, the boat reactor medium is directionally dependent, and vapor flow is laminar.8

The silicon deposition rate is of crucial importance in analyzing the system and can be

represented by Eq. 2.8 ∆Si is the deposition rate (nm/min), k is silane rate constant (m/s), c is

concentration of silane (mol/m3), MSi is the molar mass of silicon (kg/mol), and � is density of silicon

(kg/m3).

∆𝑆𝑖 =   !"!!"!

(2)

Langmuir-Hinshelwood (LH) kinetics are used to more accurately model the CVD reactor. The

bimolecular kinetics depend on competitive adsorption, that is that two gases are competing to deposit

onto the same wafer surface site.9 The two gases are SiH4 and H2 in our system, and the LH model

assumes that equilibrium reactions occur. A reaction rate can be given by Eqn. 3, where [SiH4] and [H2]

4

are silane and hydrogen gas concentrations, respectively, Ki are the equilibrium rate constants, and k is

the normal rate constant of the reaction. 9

𝑟𝑎𝑡𝑒 =   !!!"!!!!! !"!! [!!](!!!!"!! !"!! !!! !! )

!         (3)

A simple model only accounts for silicon being deposited onto the wafer surfaces. In practice,

silicon will deposit onto the reactor walls and on the boat support.7 To account for this, an effectiveness

factor, denoted η, is introduced. The effectiveness factor can be defined as the ratio of the actual rate of

reaction to the rate of reaction when the entire wafer is is exposed to the maximum concentration of silane

gas present in the reactor annulus.7

3 Methods

This experiment involved building and designing a Chemical Vapor Deposition reactor using a

COMSOL simulation. An isothermal boat reactor was first built based on a Comsol tutorial,

boat_reactor_tutorial.mph. In this experiment, in order to understand the kinetics of silane deposition,

multiple variables were adjusted including: temperature, wafer spacing, pressure, inlet velocity, and mole

fraction of hydrogen present in the inlet. These parameters were varied in order to determine how it

affected the deposition rate. The deposition rate results were then exported to and analyzed using Matlab.

Many trials were then performed using the improved model, boat_3zone_LH_2011.mph. Various

zone temperatures were chosen in order to measure and obtain the deposition rate profiles. The effects of

the pressure, inlet velocity, and the inlet hydrogen mole fraction were then studied to see which variables

had a substantial effect on the deposition rates, uniformity, and the effectiveness factor.

4 Results and Discussion

The effectiveness factor was one of the ways used to observe and analyze the effects each

variable had on the silicon deposition rate and the uniformity of the wafer. The closer the effectiveness

factor gets to 1, the more uniform the silicon deposition rate gets.

5

Figure 2 shows that the effectiveness factor increases as wafer spacing increases. The more space

between the wafers, the more uniform the silicon deposition on the wafer is. An increase in wafer spacing

also shows an increase in the deposition rate. The uniformity of deposited silicon on the wafer increases

as wafer spacing increases because the more space between the wafers, the more silane gas gets passed

through. This caused more silicon to get deposited in the center of the wafer. If the spacing between the

wafers decreased, more silicon would be deposited on the edges, which lowers uniformity across the

wafers. Figure 2 shows a direct relationship between wafer spacing and the uniformity of the wafer.

Figure 2: Effectiveness Factor versus Wafer Spacing (mm). Increasing effectiveness factor and increasing

uniformity.

Figure A2 and A3 also show a direct relationship to the effectiveness factor. As the inlet velocity

increases, the effectiveness factor also increases. This is because as the inlet velocity of the silane gas

increases, the more uniform the silicon deposition would be. Figure A3 shows that as the silane mole

6

fraction increases the uniformity of the wafer increases as well. A theory on why uniformity increases as

the silane mole fraction increases is that when there is a larger concentration of silane, there is a higher

rate that silane will be deposited across the whole wafer.

Figure 3 shows that as the temperature increases, the effectiveness factor decreases. A higher

temperature means a faster reaction causing more silicon to be deposited on the wafer. Since temperature

causes a faster reaction, more silicon will be deposited on one side of the wafer. This shows poor

uniformity as the temperature increases because more silicon will be deposited on one side of the wafer

leaving little silane to deposit silicon on the other side of the wafer. As the temperature increases, the

deposition rate would decrease.

Figure 3: Effectiveness Factor versus Temperature (degrees Celsius). Decreasing effectiveness factor and decreasing

uniformity

Both temperature and pressure have inverse relationships with the effectiveness factor. This is

shown in Figure 3 and Figure A1. As pressure increases, uniformity would decrease because different

7

pressures would lead to more silicon being deposited on various parts of the wafer, while less is deposited

on other parts.

The effects of each variable on the average deposition rate were analyzed using the slope

analysis. A value of -1 and 1 were assigned to each variable. A -1 meant there was a 5% decrease from

the nominal value, while a +1 meant there was a 5% increase from the nominal value. The nominal values

used were 560 degrees Celsius, 70 bar, 2.5 mm, 0.05 m/s, and 0.1 for reactor temperature, pressure, wafer

spacing, inlet velocity, and silane mole fraction. Table A1 shows the 28 different trials conducted using a

design of experiment method. From Table A1, Figure 4 was constructed. Figure 4 showed the individual

effects that each variable had on the average deposition rate. For example, a 5% increase on the

temperature from the nominal value showed a high average deposition rate of 0.086 nm/s. A 5% decrease

on the nominal temperature showed a very low average deposition rate of 0.026 nm/s. The other 4

variables had average deposition rates fairly close to each other when a 5% variation was applied. Figure

4 shows that the pressure, wafer spacing, inlet velocity, silane mole fraction, and the temperature/pressure

had a very small effect on the average deposition rate.

8

Figure 4: Average Deposition Rate (nm/s) versus Parameter Change. Temperature had a higher +1 value and a lower

-1 value leading to a larger slope.

The Pareto Chart in Figure 5 shows that temperature had the largest effect on the silicon

deposition rate. All the other variables showed minimal effect on the deposition rate. The slope of the

deposition rate was 0.0605 while all the other variables had a decreasing slope of less than 0.0087. This

showed that pressure, wafer spacing, inlet velocity, silane mole fraction, and the temperature/pressure had

small to minimal effect on the deposition rate. After temperature, the variables that affected the deposition

rate from greatest to least effective were silane mole fraction, wafer spacing, pressure, inlet velocity, and

then temperature/pressure.

9

Figure 5: Pareto Chart Effect of Variables on Average Deposition Rate

5 Conclusion

The variables that affected the uniformity of the wafer in the CVD reactor from most effective to

least effective were temperature, silane mole fraction, wafer spacing, pressure, and then inlet velocity.

Increasing wafer spacing, inlet velocity, and the silane mole fraction increased uniformity within the CVD

reactor. The variables that decreased uniformity within the CVD reactor were increasing temperature and

increasing pressure. The amount of uniformity on the wafer was analyzed using the effectiveness factor.

Using slope analysis it was found that pressure, wafer spacing, inlet velocity, silane mole fraction, and

pressure had very minimal effect on the deposition rate, while temperature had the greatest effect.

6 References

[1] “Chemical Vapor Deposition.” Linkoping University Department of Physics, Chemistry and Biology

(IFM). N.p., 07 December 2012. Web. 14 February 2015.

10

[2] Curley, Ronald, Thomas McCormack and Matthew Phipps. “Low-pressure CVD and Plasma-

Enhanced CVD.” Ece.umd.edu. N.p., Web. 14 February 2015.

[3] Jones, Anthony C. and Michael L. Hitchman. Overview of Chemical Vapour Deposition. Glasgow:

Thin Film Innovations LTD. Print.

[4] “Plasma Enhanced Chemical Vapor Deposition (PECVD).” Linkoping University Department of

Physics, Chemistry and Biology (IFM). N.p., 30 June 2014. Web. 14 February 2015.

[5] Pedersen, Henrik, Petter Larsson, Asim Aijaz, Jens Jensen and Daniel Lundin. “A novel high-power

pulse PECVD method.” Surface & Coatings Technology 206.22 (2012): 4564-4566. Print.

[6] Shi, Yumeng, Henan Li and Lain-Jong Li. “Recent Advances in Controlled Synthesis of Two-

dimensional Transition Metal Dichalcogenides via Vapour Deposition Techniques.” Chemical Society

Reviews. N.p., 20 October 2014. Web. 14 February 2015.

[7] H. Scott Fogler, Elements of Chemical Reaction Engineering, 4th edition, Prentice Hall, 2008.

[8] “Boat Reactor for Low Pressure Vapor Chemical Vapor Deposition.” Comsol.com. N.p., Web. 10

February 2015.

[9] “Kinetics IV.” UVM.edu. N.p., Web. 10 February 2015.

11

7 Appendices

Figure A1: Effectiveness Factor versus Pressure (Pa). Decreasing effectiveness factor and decreasing uniformity.

Figure A2: Effectiveness Factor versus Inlet Velocity (m/s). Increasing effectiveness factor and increasing uniformity.

12

Figure A3: Effectiveness Factor versus SiH4 Inlet Mole Fraction. Increasing effectiveness factor and increasing uniformity. Table A1: DOE Slope Analysis Data Temperature (T) Pressure (P)

Wafer Spacing (W) Inlet Velocity (v)

SiH4 Mole Fraction (m)

Avg Deposition (nm/s)

-1 -1 -1 -1 -1 0.025254 1 -1 -1 -1 -1 0.080697

-1 1 -1 -1 -1 0.025678 -1 -1 1 -1 -1 0.025678 -1 -1 -1 1 -1 0.025612 -1 -1 -1 -1 1 0.026318 -1 -1 -1 1 1 0.026319

-1 -1 1 -1 1 0.026376

-1 1 -1 -1 1 0.026503 1 -1 -1 -1 1 0.086571

-1 -1 1 1 -1 0.025801 -1 1 -1 1 -1 0.025801

1 -1 -1 1 -1 0.080731

13

-1 1 1 -1 -1 0.025939 1 -1 1 -1 -1 0.084618 1 1 -1 -1 -1 0.082378 1 1 1 -1 -1 0.086467

1 1 -1 1 -1 0.082418 1 -1 1 1 -1 0.084651

1 -1 -1 1 1 0.086606

1 1 -1 -1 1 0.088369 1 -1 1 -1 1 0.090398 1 1 1 1 -1 0.086505 1 1 1 -1 1 0.092347

1 1 -1 1 1 0.088409 1 -1 1 1 1 0.09043

-1 1 1 1 1 0.026623 1 1 1 1 1 0.092383

Table A2: Constants used in the COMSOL Simulation

R_a 40[mm] Wafer radius d_cc 2.5[mm] Wafer spacing d_w 0.5[mm] Wafer thickness S_a 2*(R_a^2+R_a*d_w)/(R_a^2*d_cc) Specific surface area p_tot 70[Pa] P D1_amb 5e-5[m^2/s] SiH4 diffusivity in air at 1 atm D2_amb 26e-5[m^2/s] H2 diffusivity in air at 1 atm D1 D1_amb*1.013e5[Pa]/p_tot D corrected for P D2 D2_amb*1.013e5[Pa]/p_tot D1_eff D1*(1-d_w/d_cc) SiH4 effective diffusivity in wafer bundle D2_eff D2*(1-d_w/d_cc) H2 effective diffusivity in wafer bundle M_H2 2[g/mol] Molar mass, N2 M_N2 28[g/mol] Molar mass, N2 M_Si 28[g/mol] Molar mass, Si M_SiH4 32[g/mol] Molar mass, SiH4 y0_SiH4 0.2 Inlet mole fraction, SiH4 (y1) y0_H2 0.01 Inlet mole fraction, H2 (y2) y0_N2 1-y0_SiH4-y0_H2 M_mix y0_SiH4*M_SiH4+y0_H2*M_H2+y0_N2*M_N2 Molar mass, mixture R_g 8.314[J/(mol*K)] Ideal gas constant T0 607[degC] Inlet T c10 y0_SiH4*p_tot/(R_g*T0) Inlet molar concentration, SiH4 (c1) c20 y0_H2*p_tot/(R_g*T0) Inlet molar concentration, H2 (c2) rho_no_T p_tot*M_mix/R_g To calc gas density in scalar expression rho0 p_tot*M_mix/(R_g*T0) Initial gas density

14

eta 3.1e-5[Pa*s] Dynamic viscosity v0 0.05[m/s] Inlet velocity Re 0.02*v0*rho0/eta Reynolds number at inlet (R = 0.02 m) rho_Si 2e3[kg/m^3] Silicon density mrho_Si rho_Si/M_Si Silicon molar density Ao 131296[m/(s*K)] Pre-exponential factor EaR 18500[K] E/R k0 Ao*T0*exp(-EaR/T0) Checking k_given 8.06[mm/s] Comsol given value ks0 k0*S_a More numbers for checking Delta_Si_0 k0*c10/mrho_Si RateO0 ks0*c10 Tw_1 607[degC] Zone 1 temperature Tw_2 607[degC] Tw_3 607[degC] emiss 0.85 Quartz emissivity sigma 5.6704e-8[W/(m^2*K^4)] Stefan-Boltzmann constant k_cond 0.0715[W/(m*K)] Gas mixture thermal conductivity Cp 1560[J/(kg*K)] Gas mixture heat capacity k_cond_Si 60[W/(m*K)] Si thermal conductivity Cp_Si 860[J/(kg*K)] Si heat capacity void_frac_wfb (1-d_w/d_cc) Wafer bundle void fraction k_condwfb k_cond_Si*void_frac_wfb Wafer bundle thermal conductivity Cpwfb Cp_Si*void_frac_wfb Wafer bundle heat capacity Kat 5.74[m^3/(mol*K)] L-H absorption, parts for calc, SiH4 Kbt 0.544[(m^3/(mol*K))^(1/2)] Hrxn 45600[J/mol] -ve of Heat of reaction near 800K Table A3: MATLAB code for Effectiveness Factor load xsections.txt %assuming the data file is named "xsections.txt" a=xsections; %and we rename it to a generic "a" clear xsections; x = a(:,1); x1=floor(0-x)+1; izeros=find(x1); clear x, x1; iset=size(izeros,1); iend=size(a,1); izeros=[izeros; iend+1]; %izeros stores what comes next hold on for i=1:iset r = a(izeros(i):izeros(i+1)-1,1); %radial xdata

15

ydata = a(izeros(i):izeros(i+1)-1,2); plot(r,ydata) %calculate simple difference last=size(ydata,1); surfvalue=ydata(last); percent_diff = 100*(surfvalue - ydata(1))/surfvalue; radius=r(last); ydatar = ydata.*r; % for trapz() to 'int ca r dr' ymean = 2*trapz(r,ydatar)/radius^2; eff = ymean/surfvalue; % need to format long or formatted sprintf() for tiny numbers display([i percent_diff eff]) end hold off

16

17 February, 2014 TO: Professor A. Drews, PhD FROM: Group B4: Brandon Sanchez, Saman Hadavand, Janet Mok, Liliana Busanez SUBJECT: Technical memorandum regarding Reverse Osmosis The purpose of the reverse osmosis is to apply the single membrane (ROM1) set-up to include the inlet stream, and the retentate (second) stream and a permeate stream. For the first stream, a large inlet pressure is applied, while for the second and last operations, both pressure and flow rate will be varied such that this will encourage varying concentration flows. Also to be applied to two membranes in series, adding 2 more streams for a total of 5 streams will carry high pressure from one stream to another stream to operate control on flow rate and pressure. As an effect, the 5th stream will have more salt content released from reverse osmosis membrane 2 (ROM2). Expected observations and trends include a constant flow rate for the permeate and retentate, but change in the concentration in the feed with time. The course of action to test optimizing performances includes measuring the water and solvent flux across the membrane and therefore energy by controlling permeate pressure and feed pressure. This includes changing the concentration of the feed to measure a permeate concentration, and later the solute flux. Utilizing the surface area of the membrane from the manufacturer, water flux can be measured using water as the feed only.Ultimately, our objective is to calculate recovery of retenant in the permeable (product) stream. Whereby changing parameters such as pressure and flow rate enables change of concentration for the feed, ROM1 and ROM2 can be tested for recovery technology in optimized design and parameters. Sincerely, Liliana Busanez