A reliable copper wire bond interface for microelectronic packaging

1-1-2001

A reliable copper wire bond interface for microelectronic

A reliable copper wire bond interface for microelectronic

packaging

packaging

Nicole Leigh DeBlieck Cavanah

Iowa State University

Follow this and additional works at: https://lib.dr.iastate.edu/rtd

Recommended Citation Recommended Citation

Cavanah, Nicole Leigh DeBlieck, "A reliable copper wire bond interface for microelectronic packaging" (2001). Retrospective Theses and Dissertations. 21106.

https://lib.dr.iastate.edu/rtd/21106

This Thesis is brought to you for free and open access by the Iowa State University Capstones, Theses and

Dissertations at Iowa State University Digital Repository. It has been accepted for inclusion in Retrospective Theses and Dissertations by an authorized administrator of Iowa State University Digital Repository. For more information, please contact [email protected].

A reliable copper wire bond interface for microelectronic packaging

by

Nicole Leigh DeBlieck Cavanah

A thesis submitted to the graduate faculty

in partial fulfillment of the requirements for the degree of MASTER OF SCIENCE

Major: Materials Science and Engineering Major Professor: David M. Martin

Iowa State University Ames, Iowa

2001

Graduate College Iowa State University

This is to certify that the Masters thesis of

Nicole Leigh DeBlieck Cavanah

h~s met the thesis requirements of Iowa State University

111

DEDICATION

I have been blessed with many sources of support in my life so I have several dedications with this thesis:

First, I dedicate this to my advisor, Dr. David Martin, who put the goal before me on the day I received my Bachelor of Science in Ceramic Engineering. He said, "you will be back for your graduate degree." I knew I did not want to do it without him. Dr. Martin had expectations of me, which caused me to stretch my goals in many directions. Dr. Martin taught by example. He has shown me the value of lifelong learning, active participation in community, and a well-rounded education about the world in which we live.

I also dedicate this to my parents, Michele and Charles DeBlieck, whose

unconditional support has given me confidence to succeed by encouraging me to take advantage of many opportunities. And to my husband, Andrew Cavanah, who has a great belief in my abilities, I say simply, "thank you."

TA.BLE OF CONTENTS LIST OF FIGURES

LIST OF TABLES

ACKNOWLEDGMENTS ABSTRACT

CHAPTER 1. GENERAL INTRODUCTION Statement of Problem

General Overview Review of Related Work

CHAPTER 2. MATERIALS AND METHODS Material Preparation

Material Analysis

Design of Experiment (DOE)

CHAPTER 3. RESULTS AND DISCUSSION

As-Bonded Optical Photographs and SEM Analysis of Wire Bonds Optical and SEM Analysis of Wire Bonds After HAST

Data Analysis of Electrical Resistance Measurements CHAPTER 4. CONCLUSION

Summary

Recommendations Future Research

APPENDIX LITERATURE SURVEYED REFERENCES CITED V vu vm IX 1 1 1 3 5 5 10 14 16 16 23

27

31 31 31 31 33 35

V

LIST OF FIGURES

Figure 1. Aluminum test die for electrical verification 5

Figure 2. Close-up of aluminum test die 5

Figure 3. Cu ball bond on a Cu bond pad with a Au cap of 1,000A 8

Figure 4. Close-up of Figure 3 8

Figure 5. Copper-Aluminum Phase Diagram, ASM 9

Figure 6. Copper-Gold Phase Diagram, ASM 10

Figure 7. As-bonded Pull-test Results 12

Figure 8. As-bonded Shear-test Results 12

Figure 9. After Aging Pull-test Results 13

Figure 10. After Aging Shear-test Results 13

Figure 11. Cu wire on a Cu bond pad with only a gold cap of 1,000A 16

Figure 12. Cu wire on a 10, 000A Au capped pad 16

Figure 13. Cu wire on a 10,000A Al capped pad 17

Figure 14. Au wire on a 10,000A Au capped pad 17

Figure 15. Au wire on a 10,000A Al capped pad 17

Figure 16. Secondary electron SEM image of Cu wire on a Cu pad with a 18 1,000A Au cap, bond 1

Figure 17. Secondary electron SEM image of Cu wire on a Cu pad with a 18 1,000A Au cap, bond 2

Figure 18. SEM of one side of a Au capped test die with Au wire 19 Figure 19. SEM of Au capped Cu test die with Au wire as-bonded 19 Figure 20. SEM of Al capped Cu test die with Au wire as-bonded 19 Figure 21. SEM of Au capped Cu test die with Cu wire as-bonded 19 Figure 22. SEM of Al capped Cu test die with Cu wire as-bonded 19 Figure 23. Optical image of cross-sectioned Au capped test die with Au wire as- 20

bonded

Figure 24. Optical image of cross-sectioned Al capped test die with Au wire as- 20 bonded

Figure 25. Optical image of cross-sectioned Au capped test die with Cu wire as- 20 bonded

Figure 26. Optical image of cross-sectioned Al capped test die with Cu wire as- 20 bonded

Figure 27. Optical image of cross-sectioned Au capped test die with Au wire as- 21 bonded and microetched

Figure 28. Optical image of cross-sectioned Al capped test die with Au wire as- 21 bonded and microetched

Figure 29. Optical image of cross-sectioned Au capped test die with Cu wire as- 21 bonded and microetched

Figure 30. Optical image of cross-sectioned Al capped test die with Cu wire as- 21 bonded and microetched

Figure 31. Secondary electron SEM image of cross-sectioned Au capped test die 22 with Au wire as-bonded

Figure 32. Secondary electron SEM image of cross-sectioned Al capped test die 22 with Au wire as-bonded

Figure 33. Secondary electron SEM image of cross-sectioned Au capped test die 22 with Cu wire as-bonded

Figure 34. Secondary electron SEM image of cross-sectioned Al capped test die 22 with Cu wire as-bonded

Figure 35. SEM of Au capped test die with Au wire after HAST 23 Figure 36. SEM of Al capped test die with Cu wire after HAST 23 Figure 37. Optical Image of Al capped Cu pad with Cu wire bond after HAST 23 Figure 38. Optical image of cross-sectioned Au capped test die with Au wire 24

after HAST

Figure 39. Optical image of cross-sectioned Cu capped test die with Cu wire 24 after HAST

Figure 40. Optical image of cross-sectioned Au capped test die with Cu wire 24 after HAST

Figure 41. Optical image of cross-sectioned Al capped test die with Cu wire 24 after HAST

Figure 42. Optical image of cross-sectioned Au capped test die with Au wire 25 after HAST and microetched

Figure 43. Optical image of cross-sectioned Cu capped test die with Cu wire 25 after HAST and microetched

Figure 44. Optical image of cross-sectioned Au capped test die with Cu wire 25 after HAST and microetched

Figure 45. Optical image of cross-sectioned Al capped test die with Cu wire 25 after HAST and microetched

Figure 46. SEM of cross-sectioned 1,000A Au capped test die with Cu wire 26 after HAST

Figure 47. SEM of cross-sectioned 1 0,000A Au capped test die with Cu wire 26 after HAST

Figure 48. SEM map of cross-sectioned 1,000A Au capped test die with Cu wire 26 after HAST

Figure 49. SEM map of cross-sectioned 1 0,000A Au capped test die with Cu 26 wire after HAST

Figure 50. Wire Bonding Diagram Detailing 4 Resistance Measurements per 28 Package

Table I. Table 2. Table 3.

Vll

LIST OF TABLES

Design of Experiment for HAST Samples 15

-Summary of Resistance Measurements of all Samples through HAST 30

ACKNOWLEDGMENTS

I am greatly appreciative of my major professor, Dr. David Martin and to the other members ofmy advisory committee, Professor Alan Russell and Professor Gary Tuttle.

I am grateful to the following Rockwell Collins managerial personnel: Russ J.

Wagner and Christopher G. Olson, who recognized the need for this investigation and for their support of my continuing education; to Alan Boone and Dave Hillman for technical support; and John Chihak, Jim Cook, Marsha Hilton, Mike Kanellis, and Dwayne Koch for testing and laboratory support.

I am thankful to Dr. Timothy Ellis and Rudy Wicen of Kulicke & Soffa. Their technical insight and laboratory support were critical in the completion of this project.

lX

ABSTRACT

Silicon wafer fabrication facilities are developing copper metallized wafers. Advantages of copper circuits on wafers are:

• Performance benefits as a result of device speed improvements • Denser circuitry because of reduced trace cross-section

• Cost reduction because of lower process and material costs.

The change to copper metallized wafers requires interconnect changes from the die to the next level of assembly. The majority of wafer fabrication facilities that have switched to copper metallized wafers are still terminating bond pads with aluminum because of technical challenges terminating with copper.

An example of a technical challenge with copper bond pads is dishing during

chemical mechanical planarization (CMP) of the wafer (1Colgan). The copper thins faster than the surrounding dielectric material during CMP. The bond pad midsection thins fastest leaving a non-uniform pad. This is unacceptable for interconnection to the die.

However, benefits such as electrical performance and bond strength due to lack of intermetallic junctions are sufficient to motivate industry to solve the technical challenges of using copper bond pads. Electrical performance will increase if there are no dissimilar metals conducting the current. A total copper interconnect solution is a copper metallized wafer with copper bond pads connected with copper wires to a copper lead frame or directly to a copper circuit board. A total copper solution eliminates brittle intermetallic compound formation of interconnects that are a failure mechanism in microelectronics.

However we always must keep in mind that the intermetallic failure mechanism should not be exchanged for a different failure mechanism with a greater failure rate. It is

essential to the total copper solution to develop a reliable wire bond interconnect with copper wire to copper bond pads.

My work investigated how to bond copper wire to copper bond pads. The reliability of the copper bonds was validated by passing HAST (Highly Accelerated Stress Test).

Copper produces an oxide relatively quickly. This fact presents a double-edged sword for this application. The wire and bond pads will oxidize quickly inhibiting bonding and perhaps contributing to early failures. However, if an intimate bond can be made and an oxide grows over the pad and wire and not through the bond interface, it could act as a passivation layer and protect the bond interface from corrosion. Thus, this oxide could provide a protective layer similar to the self-passivating layer on stainless steel.

CHAPTER 1. GENERAL INTRODUCTION Statement of Work

The majority of silicon wafer fabrication facilities use aluminum bond pads for their wire bonding processes. Copper has several distinct advantages over aluminum but copper wire to copper bond pad reliability has not been proven. The intent of this thesis is to prove a copper wire bond to a copper metallized silicon interface can survive 250 hours of Highly Accelerated Stress Testing (HAST).

General Overview

Silicon wafer fabrication facilities have developed the capability of making copper metallized wafers. Some of the facilities are IBM, Motorola, Texas Instruments, and AMD. Advantages of copper circuits on wafers are:

• Performance benefits because of improvements in device speed • Denser circuitry because of reduced trace cross-section

• Cost reduction because of lower process and material costs

The change to copper metallized wafers requires interconnect changes from the die to the next level of assembly. The majority of wafer fabrication facilities that have switched to copper metallized wafers are still terminating the bond pads with aluminum because there are technical challenges terminating with copper at the bond pads.

An example of a technical challenge with copper bond pads is bond pad dishing. Dishing occurs during wafer processing. The middle of the bond pad thins faster leaving a non-uniform pad. This is not acceptable for interconnection to the die from those wafers.

However, benefits such as enhanced electrical performance and elimination of intermetallic bonds provide sufficient motivation to solving the technical challenges of

copper bond pads. Electrical performance will increase if there are no changes in the metal conducting the current. This can be achieved with a total copper interconnect solution. Total copper interconnect solution is a copper metallized wafer with copper bond pads connected with copper wires to a copper lead frame or directly to a copper circuit board. This total copper solution eliminates brittel intermetallic bonds that are a source of failure in microelectronics.

Therefore the next step is to ensure the intermetallic failure mechanism is not exchanged for a different failure mechanism with a greater failure rate. It is essential to the total copper solution to develop a reliable wire bond interconnect with copper wire to copper bond pads. Therefore this work investigates how to make copper bonds with copper wire to copper bond pads. The reliability of the copper bonds was validated by passing HAST (Highly Accelerated Stress Test).

HAST has become a current testing standard for component level testing. This test has replaced the more traditional 85% humidity/85° C environmental test. The greatest benefit of the change is length of test time. The HAST testing requirements for Rockwell Collins components is 250 hours at 130° C, 85% humidity, and biased. I selected a 5-volt bias because the majority of Rockwell Collins components operate at either 5 or 3 volts.

The fact that copper produces an oxide relatively quickly is a double-edged sword for this application. The wire and bond pads will oxidize quickly, inhibiting bonding and, perhaps, contributing to early failures. However, if an intimate bond can be made and an oxide grows over the pad and wire and not through the bond interface, it could act as a passivation layer and protect the bond interface from corrosion. Thus, this oxide provides a protective layer similar to the benefit stainless steel has from its self-passivating layer.

3

Assembly benefits of copper wire are: • Lower material costs

• Reduced wire diameter because of greater strength • Longer wire bonds because of higher stiffness values • Reduced wire sweep because of higher stiffness values

These assembly benefits align well with the higher device densities allowed by copper wafers and interconnects. The reduction of wire diameters accommodate wire-bonding pitches (center-to-center spacing of adjacent bond pads) below 60 µm. Copper wire strength and stiffness values allow for longer wire lengths and less wire sweep during overmolding or encapsulation.

Review of Related Work

The appendix presents a list of references regarding copper interconnect on wafers and copper wire interconnect between an integrated circuit and a package. The majority of papers reviewed were about the challenges and benefits of replacing aluminum metallization with copper metallization on the wafer in the integrated circuit. However, most were still terminating the circuit at the bond pads with an aluminum alloy instead of a monometallic system of copper. Several of the papers in Appendix A discuss the

process development of copper wire bonding including comparisons to the standard process of gold wire bonding. Few of these papers discuss the reliability of the bond for long term use at elevated or cycling temperatures. None of the papers reports information regarding the reliability of a copper-to-copper bond during and after HAST. Motorola presented an accelerated reliability study of copper-to-copper bonding (2Tan). The presentation reported on a copper-to-copper bond aging studying up to 2000 hours at

1) Cu-Cu development in its infancy 2) Good level of feasiblity

3) Potential areas of concern include lifted bond pad metal and robust methods for preventing or cleaning pad oxides

5

CHAPTER 2. MATERIALS AND METHODS Material Preparation

The test die design was created based on several key parameters for HAST. The die needs to be biased during exposure to temperature and humidity. A bias voltage of 5 volts was selected because it was an average voltage requirement for current Rockwell Collins products. Industry is looking to go to 5 volts or less for improved battery life.

Initial test die were fabricated with aluminum metallization to prove die design and development of electrical testing. Figures 1 and 2 show an overall view of the test die and the daisy chain pattern.

Two additional metallization stack-ups were fabricated for the test die, 1 O,OOOA Cu over 500A Ti and 1 O,OOOA Cu over 500A Ta. Titanium seeding layer is Rockwell Collins

Figure 1. Aluminum test die for electrical verification

microelectronics laboratory typical wafer fabrication process. The literature review revealed the majority of the copper wafer fabricators were using Ta or TaN as a barrier layer between the copper metal layer and the dielectric.

Rockwell Collins microelectronics laboratory had difficulty getting the Cu over Ta to work. There was metal peeling, and the Ta layer did not etch completely. After several months of effort, we choose to use only the Cu over Ti test die rather than allocate additional resources to make Ta successful since that was not the purpose of the thesis.

Six test die with Cu over Ti metallization were sent to Kulicke & Soffa (hereafter K&S) for parameter set-up. After die attach, 1 hour at 150° C, and a week after

fabrication, the copper oxide was thick enough to be visually observed as a color change on the bond pads. The thermocompression bonder could not break through the oxide layer to make a metallurgical bond. At that point I decided a thin layer of aluminum or gold would be needed to protect the copper metallization for the short period of time through die attach and transport.

I determined a capping metal would be needed to prevent or limit copper oxidation during die attach, storage, and wirebonding. The first capping metal attempted was a 1000A layer of gold over the 10,000A layer of copper over 500A seed layer of Ti. Gold was selected because it dissolves into copper readily. This solution of gold into the bulk copper helped to achieve a copper-to-copper bond.

The gold cap significantly improved the integrity of the copper bond pad. However there were still bonding challenges. There were three technical challenges with bonding this set of samples:

7

2) Limiting the time of the die package on the hot plate (of the wire bonder), to prevent oxidizing through the thin gold layer

3) The hardness of the copper thin film

The first issue while bonding the sample set was creating an oxide-free copper ball. The wire bonder had been retrofitted with a glass tube designed to encompass the area surrounding the tip of the wire during electric flame off (EFO) with nitrogen. The challenge with this system was controlling the wire length and tear-off so the end of the wire always was in the nitrogen environment. If the color of the ball on the end of the wire was anything but shiny copper (like a new penny), bonding was not successful.

The second issue, while bonding this sample set, was limiting the time of the die package on the hot plate ( of the wire bonder), to prevent oxidizing through the thin gold layer. In the process of developing wire bonding parameters, half of the packages spent several minutes on the hot plate at 17 5° C. This was enough time for the copper

metallization under the thin gold cap to visually change color, i.e. oxidize. The majority of these samples did not bond. Ifbonding did occur, it was with extreme parameters and inconsistent results. This pad oxidation issue could be addressed in two ways: Showering the entire bonding area with nitrogen to prevent oxidation of die pads or an alternative wire bond pad cap to prevent oxidation during wire bonding.

The third issue, while bonding this sample set, was the hardness of the copper thin film. The hardness comparison of bulk material for copper metallization is 85 (Hardness Vickers) which is greater than gold, of 60 to 90 (Hardness Vickers); or aluminum, of 20 to 50 (Hardness Vickers) (3HARMAN). Studies have shown that as the hardness of the die metallization increases, bondability decreases (4_{HARMAN). Visual inspection and}

metallization did not move or deform during wire bonding. In addition, a few of the samples had evidence of the wire moving back up the capillary during bonding. Figures 3 and 4 show a line indicating extrusion in the wire above the ball. We believe the copper pad not deforming during bonding contributes to the difficulties in bonding to copper. If the path of least resistance for material to flow is back up the capillary tool, a reliable bond cannot be made. Even with these difficulties several copper wires were bonded to thinly gold-coated copper bond pads. The close-up in Figure 4 shows extrusion lines. Extrusion lines are evidence that the wire tried to retreat up the capillary instead of spreading out onto the pad.

Figure 3. Cu ball bond on a Cu bond pad with a Au cap of 1,000

A.

9

To overcome the hardness, the capping layer thickness was increased from 1,000A to I O,OOOA. The literature search found several references to copper wire bonding to Al bond pads. Therefore, I decided to compare an aluminum cap with a gold cap. Aluminum bond pads create a reliability concern because copper and aluminum form intermetallic compounds as shown in Figure 5.

Atomic Percent Copper

0 10 20 30 -40 :I() 60 70 80 90 100 1100 _J!JN67 1000 L 900 c..; 0 600 Ql '-"5 700 '-V 600 :)()() 400 ,, (Cu) JOO 0 10 20 30 40 50 60 70 80 90 JOO

Al Weight Percent Copper Cu Figure 5. Copper-Aluminum Phase Diagram, ASM.

Both capping metals create major problems. The longevity of a Cu and Al bond is important because that bond is available in commercial components which are being integrated in highly reliable products. Gold is typically not allowed in silicon wafer fabrication facilities because the poisoning effect gold has on silicon. Gold was chosen because it is soft, offers an oxide prevention layer, and it was presumed that there were no intermetallics between copper and gold. Review of the copper-gold phase diagram (Figure 6) found several intermetallics. Apparently these intermetallics are not a problem based on centuries of practice by jewelers to gold plate copper jewelery.

1000 900 800 p 700 ;:I 600~ Cl I-Q.. E V f-L (Au,Cu}

Figure 6. Copper-Gold Phase Diagram, ASM. Material Analysis

Bond integrity was characterized by several analytical and mechanical techniques. As-bonded samples were optically inspected, cross-sectioned through the wire bonds for metallography analysis, and the microstructure was analyzed with SEM. As-bonded samples were also pull-tested and ball-sheared to verify bond strength for comparison to industry standards. An industry reference for destructive pull-testing is Mil-Std-883 Method 2011 which requires a pull strength of 3 grams minimum for 1 mil (25 .4 microns) gold wire. An industry reference for ball-shear is ASTM F1269, which has a minimum shear strength of 15 grams for 1 mil gold wire on an aluminum bond pad. The pull-test and ball-shear graphs, Figures 7 and 8, chart both the average and the average minus three times the standard deviation. This second number is generally required to be above the industry reference minimums for a process to be considered capable. The copper wire samples bonded to gold capped copper pads were initially stronger than the gold wire bonded samples.

11

In addition to as-bonded shear and pull testing, it is common to age the wire bonds and repeat pull-testing. These results are shown in Figures 9 and 10. Aging at 125° C for 16 hours can provide some insight into bond pad contamination. If a good gold-to-gold bond is made, the pull strengths will increase with time and temperature. If there is an organic contamination on the pad or poor plating, the pull strengths will significantly decrease with time and temperature.

The Cu/Cu samples were from the first bonding attempt at K&S. Cu/Cu samples did not have optimum ball bond shape. The bond shape affects the pull and shear test results. This should be considered when comparing the Cu/Cu mechanical test results with the other samples.

As mentioned in the Material Preparation section, hardness has been identified as a key characteristic ofbondability. In an effort to quantify the hardness of the wire bond pads, three samples were prepared and sent to the IT Characterization Facility at the University of Minnesota. The three samples were a layer of copper, gold, or aluminum of

1 0,000A over 500A of Ti on silicon. Even if quantifiable results cannot be achieved at least comparable results of the three materials can be. These results provide a qualified baseline for future development samples. For example, in an effort to soften the copper bond pads, increasing the purity of the copper source could be changed from 99.99% to 99.999%.

16.00 14.00 12.00

E

10.00

e

00 i:: 8.00 '"3 0... _6.00 <+-₀ ..c _4.00 bl) i:: (/) 2.00 0.00

-2.00 Cu/Cu Cu/Au Cu/Al Au/Au Au/Al

-4.00

Ball/Pad Metallization

-+- Average A verage-3 Standard Deviations

Figure 7. As-bonded Pull-test Results

80.00 70.00

s

60.00 e 00 .::: 50.00 1-, ro Q) ..c _40.00 C/'J <+-0 ..c _30.00 bl) i:: (/) _20.00 10.00 0.00

Cu/Cu Cu/Au Cu/Al Au/Au Au/Al

Ball/Pad Metallization

-+- Average A verage-3 Standard Deviations

16.00 14.00 12.00

6

10.00

e

oJ) C _8.00 '"3 p..., _6.00 4-, 0 -:5 _{oJ) 4.00} C 2.00 VJ 0.00 -2.00 -4.00 80.00 70.00

8

60.00

e

oJ)

·=

50.00 1-. ro 0 ..c: _40.00 VJ 4-, 0 ..c: _30.00 bl) C VJ 20.00 10.00 0.00 C Cu Cu/Cu 13

Cu/Au Cu/Al Au/Au

Ball/Pad Metallization

J--

Average --- Average-3 Standard Deviations I Figure 9. After Aging Pull-test Results

Cu/Au Cu/Al

Ball/Pad Metallization

Au/Au

-- Average --- Average-3 Standard Deviations

Figure 10. After Aging Shear-test Results

Au/Al

Design of Experiment (DOE)

Design of Experiments, DOE, is an effective method of analysis for process or product development. DOE was conducted for this experiment. The purpose was to highlight significant reliability differences between a standard wire bond process with gold wire and a copper wire bond process on various copper bond pads. The reliability was determined through 250 hours of HAST. Table 1 is the full factorial DOE of samples exposed to the HAST environment. A full factorial DOE includes all variable level combinations.

There were 28 biased sites on the 44J leaded HAST board which determined the number of samples that could be tested at one time. The twelve combinations listed in Table 1. were replicated. The four additional sites were filled with samples from the set with only a 1000

A

capping gold layer. Two of these four were hermetically sealed, and two were unprotected. The variables selected of wire material and environmental protection provided enough differences between samples to characterize a successful bond and an unsuccessful bond.

Table 1. Design of Experiment of HAST samples

Combinations Wire Material Bond Pad Environmental Metallization Protection

1 Gold Au/Cu/Ti Open Cavity

2 Copper Au/Cu/Ti Open Cavity

3 Gold Au/Cu/Ti Encapsulated

4 Copper Au/Cu/Ti Encapsulated

5 Gold Au/Cu/Ti Hermetic

6 Copper Au/Cu/Ti Hermetic

7 Gold Al/Cu/Ti Open Cavity

8 Copper Al/Cu/Ti Open Cavity

9 Gold Al/Cu/Ti Encapsulated

10 Copper Al/Cu/Ti Encapsulated

11 Gold Al/Cu/Ti Hermetic

15

Gold wire is an industry standard in both hermetic and plastic encapsulated parts and was considered a baseline in this experiment. A baseline in an experiment is often

considered insurance. If an unforeseen variable occurs while testing, the baseline will not perform as expected. Then one would know to investigate for additional factors not first considered as significant in the experiment or test set-up.

The three protection levels provided a good cross-reference of bond degradation. 1.) No protection from the environment will allow for the greatest amount of

oxidation to form.

2.) Encapsulation will provide exposure to chlorides, which are thought to be an aggravation of corrosion development.

3.) Hermetic protection is a baseline for most electric packages based on many years of use in harsh environments.

CHAPTER 3. RESULTS AND DISCUSSION

Prior to and after HAST, the samples were visually inspected and documented. After visual inspection, SEM analysis was done on wire bonds before and after

cross-sectioning. Ball bonds were cross-sectioned to investigate the bond interface between the wire and the pad. Analysis provided interesting results. Observations included 1)

extensive metal deformation during bonding and 2) how the bond interface changed due to HAST.

Specific cross-sections were selected to identify failed interfaces. For example, the failed samples of copper wire and aluminum capped bond pads were separated. Others were selected to document successful and reliable bonds.

As-Bonded Optical Photographs and SEM Analysis of Wire Bonds

The as-bonded samples Cu/Cu, Cu/ Au, Cu/ Al, Au/ Au, and Au/ Al are all included in this section. The first sets of photographs, Figures 11-15, are low magnification photos of several wire bonds made to each of the samples of test chips.

\ \ J

Figure 11. Cu wire on a Cu bond pad

Figure 13. Cu wire on a 1 O,OOOA Al capped pad.

17

-.:...

Figure 14. Au wire on a 1

o,oooA

Au capped pad.

Figure 15. Au wire on a 10,000A Al capped pad.

Figures 16 and 17, are of cross-sectional and secondary electron SEM analysis of the as-bonded samples of copper wire on a l ,OOOA capped gold/copper bond pad. The gold spike up into the copper ball was found on 4 out of 4 ball bonds cross-sectioned and was an enormous surprise to us. This demonstrates the movement of the metals during bonding. The small gold waves into the copper look similar to the effect of explosion bonding (Kren). The gold spike was confirmed with energy-dispersive spectrometry (EDS) and dot mapping. Figures 18-22, are SEM photos showing representative wire bonds made for this experiment.

10 Jlffi

P:1,000x 15.0 kV P:10,000x i0000

Figure 16. Secondary electron SEM image of Cu wire on a Cu pad with a 1,000A Au cap, bond 1.

10J.lm

P:1,000X 15.0 kV P:10,000X i0000

Figure 17. Secondary electron SEM image of Cu wire on a Cu pad with a 1,000A Au cap, bond 2.

19

Figure 18. SEM of one side of a Au capped test die with Au wire .

...

D:Sb0• 12 FEB 101

P:200>< 15.0 hV AMRAY #000[)

Figure 19. SEM of Au capped test die with Au wire as-bonded.

Figure 21. SEM of Au capped test die with Cu wire as-bonded.

Figure 20. SEM of Al capped test die with Au wire as-bonded.

D:5GD.< 14 FEB 101

P:200, 15.0 kV AMRAY 10000

Figure 22. SEM of Al capped test die with Cu wire as-bonded.

In Figures (23-26), bonds were cross-sectioned to show the bond interface of the as-bonded samples. In some of the photos there is evidence of excessive wire bond force. The crack under the bond in the silicon material (Figures 24 and 25) can cause latent failures and may be discovered by wire bond pull-tests or ball-shear tests.

Figure 23. Optical image of cross-sectioned Au capped test die with Au wire as-bonded.

Figure 25. Optical image of cross-sectioned Au capped test die with Cu wire as-bonded.

Figure 24. Optical image of cross-sectioned Al capped test die with Au wire as-bonded.

Figure 26. Optical image of cross-sectioned Al capped test die with Cu wire as-bonded.

21

In Figures 27-30, bonds were cross-sectioned and copper microetched to highlight the grain structure and metal bonding in the copper wire. The copper microetch used was 1 part NH4OH, 5 parts water, and 1 part 3% solution ofH2O2•

Secondary electron SEM imaging was used to qualitatively identify metal movement shown in Figures 31-34. ~. •'

..

-·

.... : · t - .... \,

..

.. , • '.,

Figure 27. Optical image of cross-sectioned Au capped test die with Au wire as-bonded and microetched.

Figure 29. Optical image of cross-sectioned Au capped test die with Cu wire as-bonded and microetched.

Figure 28. Optical image of cross-sectioned Al capped test die with Au wire as-bonded and microetched.

Figure 30. Optical image of cross-sectioned Al capped test die with Cu wire as-bonded and microetched.

D:14,000:< 1--;;; 26 FEB 101

P:5,000>< 15.0 kV A:wfRM 101300

Figure 31. Secondary electron SEM image of cross-sectioned Au capped test die with Au wire as-bonded.

Figure 33. Secondary electron SEM image of cross-sectioned Au capped test die with Cu wire as-bonded.

Figure 32. Secondary electron SEM image of cross-sectioned Al capped test die with Au wire as-bonded.

Figure 34. Secondary electron SEM image of cross-sectioned Al capped test die with Cu wire as-bonded.

23

Optical and SEM Analysis of Wire Bonds After HAST

The HAST exposed samples Cu/Cu, Cu/ Au, Cu/ Al, Au/ Au, and Au/ Al are all included in this section. The SEM images were produced using secondary electrons.

The majority of the open cavity packages after HAST had all the ball bonds in place as seen in Figure 35. Figure 36 is a SEM photograph of an aluminum capped bond pad with copper ball bonds after HAST in an open cavity package. The intermetallic growth was so great that several ball bonds completely separated from their pads as seen in Figures 36 and 37.

Figure 35. SEM of Au capped test die with Au wire after HAST.

0: 140" 10a";,m 1 MAR 101 P:S0.0x 15.0 I..V AM.RI« 10000

Figure 36. SEM of Al capped test die with Cu wire after HAST. Note separated bonds at several locations.

Figure 37. Optical Image of Al capped Cu pad with Cu wire bond after HAST Bond wire separated from pad during HAST.

Figures 38-41, were cross-sectioned to show the bond interface of the after HAST samples. Figures 42-45, were cross-sectioned and copper microetched to highlight the grain structure and metal bonding after HAST exposure. Figures 46 and 47, were cross-sectioned and had SEM analysis to identify metal qualitative movement after HAST.

Figure 38. Optical image of cross-sectioned Au capped test die with Au wire after HAST

Figure 40. Optical image of cross-sectioned Au capped test die with Cu wire after HAST

Figure 39. Optical image of cross-sectioned Cu capped test die with Cu wire after HAST

Figure 41. Optical image of cross-sectioned Al capped test die with Cu wire after HAST

Figure 42. Optical image of cross-sectioned Au capped test die with Au wire after HAST and microetched

Figure 44. Optical image of cross-sectioned Au capped test die with Cu wire after HAST and microetched

25

Figure 43. Optical image of cross-sectioned Cu capped test die with Cu wire after HAST and microetched

Figure 45. Optical image of cross-sectioned Al capped test die with Cu wire after HAST and microetched

D:28 , 200• T;;- :? APR 101 P: 10 , 100,. 20.0 kV ANIRAY #0000

Figure 46. SEM of cross-sectioned 1,000A Au capped test die with Cu wire after HAST

C H

Figure 4 7. SEM of cross-sectioned 1 0,000A Au capped test die with Cu wire after HAST

0 I( Cul

;~:~·-.-;:::' •,•·~. ~>'•.a.,· ..

:·;\::?/

.

~-

"•.

. -~uM . iSi AuCu · CuK

Figure 48. SEM dot map of cross-sectioned 1,000A Au capped test die with Cu wire after HAST

C l< 0 I(

AuM Pdl .CuK

Figure 49. SEM dot map of cross-sectioned 1 0,000A Au capped test die with Cu wire after HAST

27

Data Analysis of Electrical Resistance Measurements

Resistance measurements were made to evaluate the integrity of the wire bonds on the die pad. As the wire bond begins to separate from the pad due to intermetallic growth or corrosion, the resistance of the circuit increases. The increase will continue until the circuit reads open. Resistance measurements of the packages were done on each side of the package. This provided 4 measurements per package. Figure 48 details the four measurement locations on the package. The resistance measurements were all taken at room temperature. Measurements were made prior to preconditioning, after

preconditioning, and after 250 hours of HAST.

Preconditioning was done to simulate three reflow passes through a oven. In

production, package components are reflowed at least once onto a printed circuit board with lead-tin eutectic solder paste with a peak solder joint temperature of 193° C for at least 30 seconds in a nitrogen environment. Instead of three passes through a reflow oven, an alternative preconditioning cycle of 100° C for 40 minutes in a nitrogen filled box oven was used. This maintained the integrity of the open cavity samples. The samples were removed from the hot oven to cool in a nitrogen cabinet. Table 2 lists all of the resistance measurements. An open reading means that at least one of the eleven wire bonds in series was not successful. In the case of the copper-to-copper samples,

resistance measurements were taken to resolve the number of successful bonds to within two bonds.

There were 4 of 32 package sites not biased with 5-volts during the 250 hours of HAST. The 4 packages not biased were denoted with a(*) designator.

1) Significant bond failures due to formation of copper/aluminum intermetallics between the copper wire and alumninum capped copper bond pads

2) Hermetic protection provides superior protection

3) More samples needed to increase confidence level of successful combinations The 95% Upper Confidence Limit column describes to a 95% confidence level how many defects in a population. For example, in the case of the Cu/Au encapsulated samples, we are 95% confident there will be 3.4 % defects in a given population.

The Confidence Level for 99% Lasting greater than 250 hours column describes how confident we are that 99% of the population will survive 250 hours of HAST. For

example, in the case of the Cu/ Au encapsulated samples, we are 59% confident that 99% of a population will survive 250 hours of HAST.

The Cu/Cu samples in hermetic packages may have been opened prior to sealing, or the temperatures of the sealing process caused excessive oxidation between the bond and the pad to provide open circuits. It would be important in future research to verify electrical resistance prior to sealing or encapsulation.

t

Resistance of Pins 1-11

r

I

Resistance of Resistance of Pins 12-22

L

.l

Resistance of Pins 23-33

Table 2. Summary of Resistance Measurements of all Samples

INITIAL TEST@ROOM TEMP POST PRECONDITIONING TEST@ _{ROOM TEMP}

Wire/Pad Resistance Resistance Resistance Resistance Resistance Resistance Resistance Resistance

Metal Protection SIN of Pins 1 to of Pins 12 to of Pins 23 to of Pins 34 to of Pins 1 to of Pins 12 of Pins 23 Klf Pins 34 to _{11 22 33 44 11 to 22 to 33 44}

(ohms) (ohms) (ohms) (ohms) (ohms) (ohms) (ohms) (ohms)

Au/Au Hermetic 1 6.0 7.0 5.9 6.4 7.3 5.9 5.8 5.7 Cu/A.I Hermetic 2 6.4 6.4 7.4 6.7 6.5 6.4 6.5 9.5 Au/A.I Hermetic 3 6.7 7.1 8.7 6.9 6.6 8.4 8.7 6.5 Cu/AJJ Hermetic 4 5.7 5.8 6.3 5.5 5.7 5.8 5.8 5.6 Cu/J\I Encap. 5 6.5 7.3 6.6 6.4 5.5 5.7 5.6 5.4 Cu/AJJ Encap. 6 5.5 5.9 6.1 5.4 6.4 6.4 7.5 6.4 Au/J\I Encap. 7 8.1 7.0 8.0 6.7 6.9 7.1 7.1 6.8 Au/AJJ Encap. 8 6.5 5.8 6.0 5.8 5.9 5.8 6.0 5.8 Cu/J\I Encap. 9 6.5 6.4 6.6 6.7 6.5 6.7 6.6 6.5 Cu/AJJ Encap. 10 5.4 5.6 5.6 7.0 5.3 5.7 5.7 5.4 Au/A.I Encap. 11 6.6 7.1 6.9 8.0 6.6 6.9 6.8 6.5 Au/AJJ En cap. 12 6.0 6.2 6.1 5.7 5.9 5.9 6.0 5.7 Au/J\I Hermetic 13 7.2 7.1 7.1 6.9 7.1 7.1 7.1 6.9 Au/AJJ Hermetic 14 6.7 6.0 5.9 6.0 6.0 6.0 5.9 6.4 Au/J\I Hermetic 15 6.8 6.8 6.9 6.6 6.8 6.8 6.9 6.6 Au/AJJ Hermetic 16 5.7 6.5 6.7 5.6 5.8 5.9 6.1 5.6 Cu/AJJ Hermetic 17 5.4 8.0 5.6 5.4 5.4 5.5 5.5 5.5 Cu/A.I Hermetic 18 6.2 6.2 6.7 6.0 6.2 6.2 6.2 6.0 Cu/AJJ Hermetic 19 5.4 5.9 6.6 5.4 5.4 5.6 5.6 5.4 Cu/A.I Hermetic 20 6.6 6.7 6.8 6.8 6.6 6.7 7.3 6.7

Cu/Cu Hermetic 21 * open open open open open open open open

Cu/Cu Hermetic 22* 6.1 open 6.2 open 6.1 open 6.2 open

Cu/Cu Open Cav 23 6.3 5.9 6.0 6.0 5.8 6.3 6.5 5.8

Cu/J\I Open Cav 24 6.6 6.9 8.0 7.4 6.7 6.7 7.7 6.7

Cu/J\I Open Cav 25 6.7 10.5 13.5 6.7 6.7 6.9 6.8 6.7

Au/AJJ Open Cav 26 5.9 6.2 7.2 6.7 5.9 6.0 7.5 5.8

Au/J\I Open Cav 27 6.7 6.7 7.0 6.7 6.7 6.6 7.2 6.8

Au/AJJ Open Cav 28 6.0 6.3 6.3 6.0 6.0 6.0 6.6 5.9

Au/J\I Open Cav 29 6.8 6.8 6.9 6.6 6.8 6.7 9.1 6.6

Cu/AJJ _{Open Cav 30} 5.5 5.6 6.1 5.4 5.9 5.5 5.4 5.4

Cu/AJJ Open Cav 31 * 5.4 5.5 5.7 5.5 5.4 5.5 5.8 5.5

Cu/Cu Open Cav 32* open open 5.8 open open open 5.9 open

POST 250 Hours HAST TEST@ ROOM TEMP

Resistance Resistance Resistance Resistance of Pins 1 to pf Pins 12 to of Pins 23 of Pins 34 to

11 22 to 33 44

(ohms) (ohms) (ohms) (ohms)

6.2 5.9 7.0 5.7 6.5 6.8 6.5 6.8 6.9 6.8 6.8 6.7 5.8 5.9 6.0 5.7 115 270 345 77 5.6 5.7 5.8 5.8 7.4 7.6 7.8 7.4 6.0 6.0 7.2 6.0 3000 500 980 750 5.5 5.8 6.0 5.8 7.1 7.3 7.4 7.0 6.1 6.5 6.2 6.3 7.6 7.6 7.4 7.3 6.2 6.3 6.1 6.0 7.2 7.4 7.3 7.0 6.0 6.0 6.3 5.9 5.7 5.7 5.7 5.7 6.5 6.6 6.7 6.5 5.6 5.8 5.9 5.7 6.9 7.7 7.8 7.0

open open open open

6.4 open 6.5 open

6.3 6.4 6.4 6.3

open 8.1 open open

open open open open

6.8 6.4 6.3 6.3 8.2 10.1 10.0 7.9 6.4 6.3 7.3 6.1 7.5 7.7 7.4 7.4 5.8 5.8 5.8 200 5.8 6.6 5.8 5.9

open open 6.3 open

N

Table 3. Copper Test Die Resistivity Study

Wire/Pad Protection Wirebonds Wire bonds Fraction 95% Upper Confidence Level Metal Passing 250 Initially Good Defective Confidence for 99% Lasting

hrs. ofHAST Limit >250 hrs. of HAST

Au/Al Encap. 88 88 0.0 0.034 0.59

Au/Al Hermetic 132 132 0.0 0.023 0.73

Au/Al Open Cav 88 88 0.0 0.034 0.59

Au/Au Encap. 88 88 0.0 0.034 0.59

Au/Au Hermetic 132 132 0.0 0.023 0.73

Au/Au Open Cav 88 88 0.0 0.034 0.59

Cu/Al Encap. 0 88 1.0 No value 0.0

Cu/Al Hermetic 132 132 0.0 0.023 0.73

Cu/Al Open Cav 11 88 0.875 0.928 No value

Cu/Au Encap. 88 88 0.0 0.034 0.59

Cu/Au Hermetic 132 132 0.0 0.023 0.73

Cu/Au Open Cav 77 88 0.125 0.198 No value

Cu/Cu Hermetic 22 22 0.0 0.127 0.20

31

CHAPTER 4. CONCLUSION Summary

The intent of this thesis was to prove a copper wire bond to a copper metallized silicon interface can survive 250 hours of HAST. Copper wire bonded samples were built and electrical verification of connection was made prior to and after HAST. All initially successful copper wire bonded samples survived HAST.

Recommendations

There is sufficient interest and success to continue investigation into copper wire bonding. This investigation should lead to furthering the life of wire bonding by allowing the diameter of the wire to decrease to be compatible with decreasing bond pad pitch on the die. Furthermore knowledge gained here is transferable to stud bumping and direct chip attach. Direct chip attach is anticipated to become mainstream in future

microelectronic interconnect because of cost and technical performance. Future Research

Future research may include items to increase bondability and reliability of copper wire bonding to copper bond pads. These items could include organic coatings, metal caps, controlled environments, or processes to remove copper oxidation just prior to wire bonding.

Organic coatings such as organic solder protection in circuit boards are a possibility to temporarily protect the copper bonds pads from oxidation until wire bonding.

However, usually any organic contamination on wire bond pads inhibits bondability and reliability.

Metal caps such as palladium do not readily oxidize. This would be a wafer level solution.

Inert atmosphere from wafer fabrication until after wire bonding is a possibility, but would require substantial infrastructure and process control changes. The necessary infrastructure is not currently in place.

Processes such as ROSA, Reduced Oxide Soldering Activation, to remove surface copper oxidation just prior to wire bonding are already available (6Tench). ROSA would require capital investment and an additional processing step. However ROSA would require the least amount of change to the current wire bonding process.

33

APPENDIX LITERATURE SURVEYED

"Copper Wire Bonding", SEMI CON Singapore 2000, May 10, 2000, Timothy W. Ellis, Ph.D., Lee Levine, Rudy Wicen, and Larbi Ainouz.

Atmospheric Degradation and Corrosion Control, Philip A. Schweitzer, P .E., Marcel Dekker, Inc. Copyright 1999.

"Copper bonding could outshine gold", Choen, Charles L., Electronics v. 58 July 29 '85 p. 27-28.

"High-Density Packaging for Copper Chips", Kime J. Blackwell, Irving I. Memis, and Ronald P. Kuracina, HDI August 1998 p. 20-22.

"Development of Copper Wire Bonding Application Technology", Kenji Toyozawa, Kazuya Fujita, Syozo Minamide, and Takamichi Maeda, IEEE Transaction s on Components, Hybrids, and Manufacturing Technology, Vol,. 13, NO. 4, December 1990.

"A Comparison of Copper and Gold Wire Bonding on Integrated Circuit Devices". Salim L. Khoury, David J. Burkhard, David P. Galloway, and Thomas A. Scharr, 40th Proceeding ECTC, May 1990, p. 768-776.

"Optimization of Copper Wire Bonding on Al-Cu Metallization", Luu T. Nguyen, David Mc Donald, Anselm R. Danker, and Peter Ng, IEEE Transactions on Components, Packaging, and Manufacturing Technology -art A, Vol. 18, NO. 2, June 1995, p. 423-429.

"Optimizing the Wire-Bonding Process for Copper Ball Bonding, Using Classic Experimental Designs", Michael Sheaffer, Lee R. Levine, and Brian Schlain, IEEE Transactions on Components, Packaging, and Manufacturing Technology, VOL. CHMT-10, NO. 3, September 1987, p. 321-326.

"Ag and Cu Migration Phenomena on Wire-Bonding", Ker-Chang Hsieh and Theo Martens, Journal of Electronic Materials, Vol. 29, No. 10, 2000, p. 1229-1232. "The long awaited "copper crunch" has now came and although copper based

interconnect will continue its dominance in short distance (up to 10 meters) and medium bandwidth (up to 5Gbps) applications" pg. 148 The Electronics Industry Report 2000, Prismark Partners LLC

APPENDIX LITERATURE SURVEYED

"Introducing IBM's First Copper Wiring Foundry Technology: Design, Development, and Qualification of CMOS 7SF, a 0.18um Dual Oxide Technology for SRAM, ASICs, and Embedded DRAM", Nivo Rovedo, Terence Hook, Michael Armacost, Gary Hueckel, Douglas Kemerer, and Joseph Lukaitis. IBM Microelectronics Fourth Quarter 2000. Pg.34-39.

"Building an ESD Strategy: Bulk CMOS to SOI, Copper, Low-k, and SiGe", Steven H. Voldman IBM Microelectronics Fourth Quarter 2000. Pg. 21-23.

"Fine-Pitch Copper Ball Bonding on BGA substrates", Binner, Ralph and Weal, Tan Kah, HDI September 2000, Pg. 26-32.

"Practical Issues in the Probing of Copper Pads", Timothy Turner, Solid State Technology March 2001, Pg. 97-102.

"Using Copper Materials in IC Semiconductor Packaging", Michael Sheaffer, Kulicke & Soffa News Connections, www.kns.com.

"The Use of Copper Wire as an Alternative Interconnection Material Advaced

Semiconductor Packaging", Larbi Ainouz and Muller Feindraht AG, Kulicke & Soffa Resources, www.kns.com.

35

REFERENCES CITED

1_{Colgan, Micheal; Morath, Chris; Tas, Ouray; Grief, Malcolm; "An Ultrasonic Laser}

Sonar Technique for Copper Damascene CMP Metrology", Solid State Technology, February 2001, Pg. 67.

2_{Tan, K.H.; Lee, K. Y.; Mui, K. W.; Tran, Tu Anh; Yong, C. C.; Tan, L. C.; Harum,}

Fuaida; Faizari, Mohd; "Cu-Cu Bonding- A Feasibility Investigation", presented by Motorola, July 15, 1999.

3_{Harman, George; Wire Bonding in Microelectronics Materials, Processes, Reliability,}

and Yield, McGraw-Hill, Pg. 212.

4_{Harman, George; Wire Bonding in Microelectronics Materials, Processes, Reliability,}

and Yield, McGraw-Hill, Pg. 54.

5_{Kren, Lawrence; "Big-Bang Bonding", Machine Design, November 16, 2000, Pg. 67.} 6_{Tench, D.M.; White, J.T.; Hillman, D.; Lucey, G.K.; Gher, T.; and Piekarski, B.;}

"Fluxless Wave Soldering", Circuit World, Volume 21, Number 2, January 1995, Pg. 45.