SOUTHWEST TEST 2006 1 Multiple resort effects on Cu bump technology Weida Qian Intel Corp. Zhongkai Xu Intel Corp. James Stiehl Intel Corp. Chris Tran Intel Corp. June 2006
SOUTHWEST TEST 2006
1
Multiple resort effects on Cu bump technology
Weida Qian Intel Corp.Zhongkai Xu Intel Corp.James Stiehl Intel Corp.Chris Tran Intel Corp.
June 2006
SOUTHWEST TEST 2006
2
The objectives of this presentation
This presentation isis intended to:Present the Cu bump resort effect;Present our current understanding of the issue and its impact;Seek help from the probing industry to find a solution for this effect;
This presentation is notnot about:Cu bump probing process development;Other Cu bump probing challenges.
SOUTHWEST TEST 2006
3
Outline
Introduction: TV and product Cresmeasurements
The Cu bump multiple probing effect on Cres: TVs and logic products
The hypotheses explaining the observed effect
Impact of the multiple resort effect – system correlation and tool IQ methodology
Summary
SOUTHWEST TEST 2006
4
☞ Introduction
SOUTHWEST TEST 2006
5
Cres: A key measurement for process dev and HVM monitoring
Test vehicle (TV) Cres for process dev:Special Special CresCres measurement test vehicles (TV) are used measurement test vehicles (TV) are used during probing process development;during probing process development;All tester IO channel resources can be used for All tester IO channel resources can be used for CresCresmeasurements (measuring path resistance and deducing measurements (measuring path resistance and deducing CresCres););Both the single channel max (addresses product IO Both the single channel max (addresses product IO concerns) and die mean (addresses device power delivery) concerns) and die mean (addresses device power delivery) CresCres are used as key responses.are used as key responses.
Product Cres for HVM monitoring:Product IO ESD protection diodes used for Product IO ESD protection diodes used for CresCres estimation;estimation;Product Product CresCres monitored through onmonitored through on--line and offline and off--line PCS line PCS systems. systems.
SOUTHWEST TEST 2006
6
HVM Cu bump probing technology requirements
Must meet electrical performance requirements Must meet electrical performance requirements for a broad range of products;for a broad range of products;Must achieve a low and stable probing Must achieve a low and stable probing CresCres in a in a production environment;production environment;Consistent performance over probe card lifetime; Consistent performance over probe card lifetime; Consistent performance across multiple probe Consistent performance across multiple probe cards and modules;cards and modules;Adequate process margin to overcome variations Adequate process margin to overcome variations in Cu bump characteristics;in Cu bump characteristics;Consistent performance under repeated probing Consistent performance under repeated probing (multiple probing).(multiple probing).
SOUTHWEST TEST 2006
7
☞☞ The effect The effect ……
SOUTHWEST TEST 2006
8
Cu bumps show unique behavior under multiple sorts:
Probing Cres deteriorates significantly with relatively few repeated tests on the same bumps (wafers);The same wafer can only be used to yield reliable results for up to ~ 14 resorts;
Cres goal
Cu bump to probe Cres increases with multiple resorts
Die average Cres distribution at eachsort for the same Cu bump TV wafer.
Sort Order0
SOUTHWEST TEST 2006
9
This resort effect with Cu is observed with different probing technologies and processes
Technology B, process 1
Technology B, process 2
2
SOUTHWEST TEST 2006
10
Characteristics of the multiple resort effect
Uniform Cres increase on all pins on all dies was not observed;Probability of pins to have high Cres (SCF rate for pins) increases with the number of resorts;The magnitude of the effect depends on probing processes, individual wafers, probe cards and other random factors ;Was not due to specific probes, bumps;
SCF Probability Per Pin
-0.01
0
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
0.09
1 3 5 7 9 11 13 15 17 19 21Number of Resorts
SCF
prob
abili
ty SCF Probability
Pin Level Cres in Sort Order(2 TV wafers sorted 21 times)
SCF = single channel failure Pin Cres Control Limit
SOUTHWEST TEST 2006
11
Product Cres change vs multiple resorts
Similar trend on product Cres observed on different probing technologies;The die average std Cres increases with multiple resorting.
Technology N
Same wafer15 resortsDie ave Cres in sort order
SOUTHWEST TEST 2006
12
Product Cres change vs multiple resorts: Same technology but different products
SOUTHWEST TEST 2006
13
Product IO channel resistance change vs the number of resorts
• Some channels were more sensitive to the number of resorts;• But the overall trend does exist for all channels.
SOUTHWEST TEST 2006
14
☞ The hypotheses …
SOUTHWEST TEST 2006
15
What causes the multiple probing effect?
Probe Mark Observations:More flat areas on the bump surface after multiple probing with flat probes;Wider and deeper probe marks after multiple probing with wedge/blade probes;
Flat probes Wedge/Blade probes
1st probing multiple probing 1st probing multiple probing
SOUTHWEST TEST 2006
16
Probe mark examples
Probe markProbe markafter 20 resortsafter 20 resorts
Probe markProbe markafter 1 sortafter 1 sort
SOUTHWEST TEST 2006
17
What causes the multiple probing effect (cont’d)?
Hypothesis I:A minimum pressure is needed to overcome the copper oxide layer (a few nm) for good contacts. Multiple probing increases the “contact” area, reduces the contact pressurepressure, resulting in poor contacts;
Hypothesis II:Multiple probing smoothes the bump surface on a micro-scale;The probes sometimes may slideslide on the smooth Cu surface, resulting in poor contacts;
Hypothesis III:The Cu bumps act as a polisher for probes, and the accumulated debris may cause poor contacts;
SOUTHWEST TEST 2006
18
☞ The impacts …
SOUTHWEST TEST 2006
19
Key impacts of the effect
Sort module “golden” correlation methodologyCu correlation wafer lifetime is reduced compared to that of PbSn wafers;Bin switches (Kappa) is sensitive to this effect;
Sort module IQ methodologyCu IQ wafer lifetime is reduced compared to that of PbSnwafers;Parametric reproducibility and kappa are both affected by the increased Cres after multiple resorts; It is difficult to qualify a whole fleet of tools when the wafers you are using as “standards” could not yield reproducible results after small number of testing.
SOUTHWEST TEST 2006
20
Corr wafer lifetime example: Bin switches (Kappa) vs. number of resorts for logic products
Kappa deteriorates after multiple resorts;The rate of deterioration is not consistent;Potential factors impacting deterioration rate: 1. Probing technology; 2. Individual SIU; 3. Probing process; 4. Material
Resort Kappa vs the # of Resorts
0.6
0.65
0.7
0.75
0.8
0.85
0.9
0.95
1
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20Number of resorts
Kap
pa R
elat
ive
to th
e 1s
t Sor
t
Tech B, Wafer 1Tech B, wafer 2Tech A
Goal
SOUTHWEST TEST 2006
21
Module 1 Module 4Module 3
Module 2 Module 5 Module 6
IQ wafer lifetime example: Kappa, Cres vs resorts
2 IQ wafers2 resorts on each module
Kappa Goal
Kappa
SOUTHWEST TEST 2006
22
Potential Solutions
Further probing process optimization to minimize the effect;
Alternative correlation wafer methodology;Revisit system correlation usage cases;Alternatives to “golden” corr methods;
Alternative tool IQ methodology;Optimize IQ wafer usage;Revisit IQ test parameters and their success criteria;
New advances in probing technology for Cu;New advances in probing technology for Cu;Innovative correlation/module matching solutions;Alternative probe tips may increase corr/IQ wafer lifetime;
SOUTHWEST TEST 2006
23
SummaryThe Cu bump multiple resort effect was observed on both TV and products, and on different probing technologies;
The magnitude of the effect depends on probing technologies/processes, individual wafers, probe cards and other random factors;
We believe the effect is due to the micro-scale surface condition changes on the probing area: pressure, slidingand debris are the three main hypotheses.
Both the system correlation and IQ methodologies are being challenged by the effect. Innovative solutions are needed to increase corr/IQ wafer lifetime.