Lead-free solder systems : phase relations and microstructures Citation for published version (APA): Oberndorff, P. J. T. L. (2001). Lead-free solder systems : phase relations and microstructures Eindhoven: Technische Universiteit Eindhoven DOI: 10.6100/IR545496
DOI: 10.6100/IR545496 Document status and date: Published: 01/01/2001 Document Version: Publisher’s PDF, also known as Version of Record (includes final page, issue and volume numbers) Please check the document version of this publication: • A submitted manuscript is the version of the article upon submission and before peer-review. There can be important differences between the submitted version and the official published version of record. People interested in the research are advised to contact the author for the final version of the publication, or visit the DOI to the publisher's website. • The final author version and the galley proof are versions of the publication after peer review. • The final published version features the final layout of the paper including the volume, issue and page numbers. Link to publication
General rights Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights. • Users may download and print one copy of any publication from the public portal for the purpose of private study or research. • You may not further distribute the material or use it for any profit-making activity or commercial gain • You may freely distribute the URL identifying the publication in the public portal. If the publication is distributed under the terms of Article 25fa of the Dutch Copyright Act, indicated by the “Taverne” license above, please follow below link for the End User Agreement: www.tue.nl/taverne Take down policy If you believe that this document breaches copyright please contact us at:
[email protected] providing details and we will investigate your claim.
Download date: 09. Mar. 2019
Lead-free Solder Systems: Phase Relations and Microstructures
PROEFSCHRIFT
ter verkrijging van de graad van doctor aan de Technische Universiteit Eindhoven, op gezag van de Rector Magnificus, prof.dr. M. Rem, voor een Commissie aangewezen door het College voor Promoties in het openbaar te verdedigen op woensdag 27 juni 2001 om 16.00 uur
door
Pascal Johannes Theodorus Lambertus Oberndorff
geboren te Geleen
Dit proefschrift is goedgekeurd door de promotoren:
prof.dr. F.J.J. van Loo en prof.dr. J.K. Kivilahti
Druk: Universiteitsdrukkerij, Technische Universiteit Eindhoven
CIP-DATA LIBRARY TECHNISCHE UNIVERSITEIT EINDHOVEN Oberndorff, Pascal J.T.L. Lead-free solder systems : phase relations and microstructures / by Pascal J.T.L. Oberndorff. - Eindhoven : Technische Universiteit Eindhoven, 2001. Proefschrift. - ISBN 90-386-2862-5 NUGI 813 Trefwoorden: vastestofchemie / non-ferro metalen en legeringen / loodvrij soldeer / diffusie / fasediagrammen / microstructuur Subject headings: solid state chemistry / nonferrous metals and alloys / lead-free solder / diffusion / phase diagrams / microstructure
Voor Wim Ruers (14/7/1969-7/5/1981) Omdat jij nooit de kans hebt gekregen.
Contents Chapter 1:
Introduction
1
1.1 A Short History of Soldering
1
1.2 Lead in Solders
4
1.3 Possibility of Lead-free Solder in Electronics
7
1.4 Outline of This Work
9
References
10
Chapter 2:
Theoretical Framework
13
2.1 Phase Diagrams
13
2.1.1
Diffusion Path
14
2.1.2
Predictability of Diffusion Paths in Ternary Systems
18
2.1.3
Variations of the Diffusion Couple Technique
19
2.1.4
Error Sources Encountered in the Diffusion Couple Technique
2.2 Diffusion in Solids
21 23
2.2.1
Atom Movements
24
2.2.2
Reactive Diffusion
24
2.2.3
The Position of the Kirkendall Plane(s)
30
2.3 Diffusion in Liquids
32
References
33
Chapter 3:
Experimental Procedures
35
3.1 Equilibrated Alloys and Diffusion Couples
35
3.2 Annealing Procedures
37
3.3 Analysis
38
References
43
Chapter 4:
Phase Relations in the Sn-Ag-Sb System at 220 oC
4.1 Introduction
45 45
o
4.2 Intermediate Phases in the Binary Sn-Sb System at 220 C
46
4.3 Phase Relations in the Sn-Ag-Sb System at 220 oC
51
4.4 Concluding Remarks
55
References
56
Chapter 5:
Phase Relations in the Sn-Ni-Cu System at 235 oC
57
5.1 Introduction
57
5.2 The Binary Sub-systems
58
5.2.1 The Cu-Ni System
58
5.2.2 The Cu-Sn System
59
5.2.3 The Ni-Sn System
61
5.3 Equilibrated Alloys
61
5.4 The Diffusion Couple Technique
70
5.4.1 Solid-Solid Diffusion Couples
70
5.4.2 Solid-Liquid Diffusion Couples
71
5.5 The Isothermal Cross-section of the Sn-Ni-Cu System at 235 oC
75
5.6 Concluding Remarks
76
References
77
Chapter 6: Phase Relations in the Bi-Ni-Pd System
79
6.1 Introduction
81
6.2 Solid-state Diffusion in the Binary Bi-Pd System
82
6.3 The Binary Ni-Bi Phase Diagram
88
6.4 Isothermal Cross-section of the Bi-Ni-Pd System at 235oC
90
6.4.1 The Diffusion Couple Technique
90
6.4.2 Equilibrated Alloys
92
6.5 Concluding Remarks
95
References
96
Chapter 7:
Anomalous Behavior in Solid-liquid Systems
97
7.1 Introduction
97
7.2 Whisker Growth in the Sn-Cu System
99
7.3 Intermetallic Layer Formation in the Au-Sn System
109
7.4 Concluding Remarks
111
References
112
Chapter 8:
Epilogue
115
Summary
117
Samenvatting
119
Acknowledgements
121
Curriculum Vitae
123
‘We didn’t cook none of the pies in the wash pan, afraid the solder would melt; but uncle Silas he had a noble brass warming-pan which he thought considerable of, because it belonged to one of his ancestors with a long wooden handle that come over from England with William the Conqueror in the Mayflower or one of them early ships and was hid away up garret with a lot of other old pots and things that was valuable, not on account of being any account because they warn’t, but on account of them being relics, you know, and we snaked her out, private, and took her down there, but she failed on the first pies, because we didn’t know how, but she come up smiling on the last one.’ Mark Twain – The adventures of Huckleberry Finn, p. 247, 1885 Penguin Popular Classics, England
Chapter 1 Introduction
1.1
A Short History of Soldering Soldering is a technique known by most people around the world, but only few
realize that soldering has been used for thousands of years already. Early in history mankind discovered the importance of joining. One of the techniques developed for joining is soldering, not only for making tools, which is sometimes said to make the difference between animals and humans, but also for ornamental purposes. Ornaments played an important role in daily life and afterlife; often people were buried with their best jewelry. Since these burial sites were considered sacred, and therefore well preserved, they provide excellent research objects for archeologists. The finds made by these archeologists enable us to reconstruct and assess the level of technical skills in ancient times. In the following part a short summary of the most important developments in soldering throughout history will be given [1-6]. The discovery of soldering goes back several thousands of years. The reason for this early discovery is that the metals used can easily be molten on a wood fire available anywhere. The earliest traces of hard solder (i.e. liquidus above 300 oC) that
1
Chapter 1
can be found in Sumeria and Egypt date from about 5,500 years ago, see Fig. 1.1. In these times Au-Cu-, Ag-Cu- and Pb-Cu- alloys were used as solders. The soldering skills were advanced already, because on some of the recovered artifacts (mainly jewelry) it is very hard to see that these are really soldered joints.
Fig. 1.1: Oldest known picture of soldering. An Egyptian goldsmith soldering with a blowing pipe on a coalfire. Painting in an Egyptian tomb, Thebe, 1475 BC [1]. In soft soldering Sn is used in order to lower the melting point of the solder. About the starting of this technique one can only guess, because Sn will transform rather easily into Sn acids. Therefore, only a limited number of ancient artifacts using 2
Introduction
soft solder have been recovered. Another impeding factor is that Sn was unknown to Egyptian society before 2000 BC. Still another reason why so little is known about the early history of soldering is that soldering was considered a slave’s task. However, it is believed that soft soldering has been used for the passed 4,500 years. The Greeks were masters at soldering their jewelry, as can be seen at the ‘treasure of Priamos’ found during the excavations of Troje II (2600 BC). It is very strange to note that after 1400 BC soldering techniques seemed to be diminishing. Only around 350 BC the Greeks come back at their earlier level of soldering. At this time the Greeks are the first to use soldering for building machines. They mainly used copper, bronze and lead for that purpose. Several Greek writers, such as Hippokrates (430 BC), Theoprast (300 BC) and Poseidonius of Apamea (139 BC) mention solder (Chrysokolla) in their works. The exact composition of this “Chrysokolla’ is controversial [7]. The Romans introduced Sn-Pb solder, with as most famous example: the aqueducts. The Roman aqueducts are a marvel of architecture and engineering. Therefore, they are described in several ancient works (e.g. by Pliny, Vitruvius and Frontinus). The aqueducts often have lead sheets to convey the water. In order to connect these lead sheets together, Sn-Pb solder was used, as allegedly described by Pliny (Historia Naturalis liber XXXIII) [5]. The Romans also used Cu and Au alloys as solder as well as ‘Chrysokolla’. When Sn prices increased they started using pure lead as solder material, especially for water containing vessels. Except in these civilizations, soldering was also well-known in the rest of Europe. Evidence of this can be found in Irish ornaments, dating back to 1500 BC and archeological finds in Central Europe dating back to around 1200 BC. According to Pliny and Caesar, the Celts and Gauls were masters working with Sn alloys. The first technical descriptions of solder can be found in the Papyrus Graecus Holmiensis and the Papyrus Graecus Leidensis (about 150 AD). In the first couple of hundred years after Christ, Chrysokolla is used as solder but, as was said before, confusion exists about what material is actually meant with Chrysokolla. Agricola makes the distinction between ‘ancient Chrysokolla’ and Chrysokolla being borax, like it is used in his age [8]. In the Middle Ages Persian, Syrian and Arabian scrolls about soldering appear, making the soldering tradition more efficient in the next couple of hundred years. Around 1400 AD, the first legislation about soldering appears. This mostly deals with the gold solders being diluted with less noble metals. 3
Chapter 1
In the same era a lot of documents appear about soldering and in Europe the use Pb in joining materials is spread, as written by Leonardo da Vinci in his Codex Atlantico. Famous scientists like Galilei, Paracelsus and Descartes study soldering techniques. In 1727 Newton is the first to add Bi to a Sn-Pb alloy, lowering the melting point to below 100 oC. In the early 1900’s the use of Bi, Cd and Zn as alloying materials to lower the melting point of solder is becoming more and more common. The discovery of aluminum causes problems for soldering science, since none of the soldering materials known at that time wets aluminum. In the 19th century, the first scientific descriptions of the process and metallurgy of soldering appear (Thompson (1807), diffusion in liquids, Sorby (1864) etching and photographing metal samples, Tammann (1897), kinetics of crystallization). In the late 19th century the function of soldering became twofold; in addition to making a metallurgical bond, solder joints are also used for their electrical conductivity. The advance in electronics in the 20th century caused solder and soldering to gain in importance. E.g., in the 1990’s an estimated 1013 solder joints were manufactured annually in electronic industry [9].
1.2
Lead in Solders Although Sn-Pb solders have been used for centuries already, only in the
1990’s fundamental knowledge of soldering became more important for several reasons. The first and initial reason was that environmental awareness had been growing for several years and now solder and soldering were scrutinized also, since the most commonly used solder material is an alloy of tin and lead. Already when lead was first used it was known that it has malicious effects on health. Hippocrates was the first, around 400 BC, to recognize lead poisoning among miners. Nevertheless, the Romans used Pb extensively and even knew about it’s detrimental
4
Introduction
effects on human health in combination with water as is shown in the following passage of Vitruvius his books on Architecture [10]: ‘Etiamque multo salubrior est ex tubulis aqua quam per fistulas, quod plumbum videtur esse ideo vitiosum quod ex eo cerussa nascitur, heac autem dicitur esse nocens corporibus humanis. Itaque quod ex eo procreatur (si) id est vitiosum, non est dubium quin ipsum quoque non sit salubre.’ ‘Water conducted through earthen pipes is more wholesome than that through lead; indeed that conveyed in lead must be injurious, because from it white lead is obtained and this is said to be injurious to the human system. Hence, if what is generated from it is pernicious, there can be no doubt that itself cannot be a wholesome body.’ Some historians even claim that the decline of the Roman Empire was due to the use of lead utensils in combination with lead water pipes [6]. In 1845 Sir John Franklin led an expedition to the arctic with a food supply of three years. His expedition failed and all of the crew perished, not just because of extreme cold or hunger, but also due to lead contamination from badly soldered cans containing food [11]. In the 1990’s this aspect came back into full view due to the growing environmental awareness and the increasing insight in the precise effects of lead on the human body. For example, lead is known to accumulate in the human body, binding to proteins. In this way it can cause nervous and reproductive system disorders, delays in neural and physical development as well as reduced production of hemoglobin resulting in anemia and hypertension. Especially for children it has harmful effects on neurological development [12]. Therefore, the use of lead in plumbing, gasoline and paint industry became strictly regulated in the 1980’s. Although the main use of lead is in storage batteries, this is concentrated use and easy to recycle. The use of lead in solder materials amounts only to a couple of percents of the total use of lead but it is in small volumes, which makes it easy to leach into surface water and therefore, causing a major threat to health and environment. The precise mechanism of the leaching of lead remains unknown [13]. 5
Chapter 1
After recognizing this danger of spreading lead, legislation was proposed to minimize use of lead in electronic industry. In the US several bills were proposed to both the Senate as well as the House of Representatives. Until now none of the bills have been passed yet [14]. In Europe the European Committee has a Directive on Waste from Electrical and Electronic Equipment, which states that the use of lead, among other harmful materials, should be ‘phased out’ in all electronic products by 2008 [15]. In Japan legislation is being proposed that will prohibit lead from being sent to waste disposal sites. In response to the proposed legislation, industry formed research consortia and soldering centers to look for possible replacements. Several of these published research reports about their findings, but none came up with a ‘drop-in’ replacement for the eutectic or near eutectic Sn-Pb solder [15-19]. The concern about health and environment constitutes only one side of the drive to develop lead-free solders and microelectronics. Something that is at least as important is the growing concern that with new technologies and stricter demands conventional Sn-Pb solder will not satisfy all the demands on the level of performance and reliability due to the increased density and complexity of the circuitry. For example, the pitch in electronics is getting smaller and smaller, posing a challenge to Sn-Pb solder joints. This development will continue in the future meaning that Sn-Pb cannot be used anymore. Therefore, new solder materials have to be tried. Another example can be seen in the automotive industry; more electronics are used in modern cars than before and these electronics have to function under conditions too extreme for Sn-Pb solder (think about engines where temperatures reach far above 150 oC [20]). So there is a need for solders that have a higher melting point, but have at least the same reliability as Sn-Pb. With upcoming miniaturization there is also a problem with lead because of emission of alpha particles, resulting from radioactive decay within the element, which can cause errors in the chip circuitry [21]. The third and last reason for developing new solder materials can be found in the behavior of customers and, therefore, the economy. When Matsushita (National or Panasonic) started selling Minidisc players in Japan at the end of 1999 with a label of being lead-free, their market share rose with 20 % in the first months [22]. Apparently, environmentally friendly products sell better. Of course, this is a 6
Introduction
consequence of the growing environmental awareness in society but it will be an important driving force for electronic industry to fulfill consumers wishes.
1.3
Possibility of Lead-free Solder in Microelectronics Soldering will remain a very important technology, unless there will be a
sudden improvement in the adhesive joining technology, which anyway is unlikely to replace the soldering completely. It is safe to say that tin will remain the major component in solder since the melting temperature of tin is low enough (232 oC). If one has to substitute lead in solders with (an)other element(s), only few elements remain simply on the basis of interaction with Sn and the corresponding melting temperatures. Of course, melting temperature will not be the only criterion a substitute solder has to fulfill. Some other criteria for the substitute solder would be: -
Adequate wetting properties
-
Good joint strength
-
Physical properties equal to Sn-Pb
-
Good fatigue resistance
-
Compatibility with existing flux systems
-
Non-toxicity
-
Low cost
-
Low dross formation
Table 1.1 gives an overview of the elements, other than Pb, that have been used as alloying additions to Sn for soldering purposes [23-26]. The moderate availability in the case of Bi is caused by the fact that Bi is a byproduct from lead mining. Another disadvantage of Bi is that if used in larger quantity (5 wt.%) it can cause brittleness in a Sn-based solder joint [27]. Cd cannot be used due to its toxicity. The costs of Ga are too high, and besides that the supply is problematic and it can cause brittleness. In has similar problems; besides this, there is an ongoing discussion about the toxicity of In. This discussion also includes the use of Sb, which can, and has been, used as a minor additive to solders. Zn is sufficiently
7
Chapter 1
available, but it is well known that Zn causes problems with wettability and corrosion [28]. Ag is in adequate supply, if used as a minor additive, but the costs are high. Cu has been used before with success as minor additive to solder. Apart from a large supply and low costs it is soluble in Sn in low percentages. Au is very expensive and, therefore, not applicable. The toxicity of Hg makes the use of this element impossible.
Element
Toxicity
Availability
Costs
Bi
No
Moderate
Acceptable
Cd
High
Moderate
Acceptable
In
?
Rare
High
Zn
No
Ample
Low
Au
No
Low
High
Tl
High
Low
?
Ga
No
Low
High
Hg
High
Low
High
Ag
No
Moderate
High
Cu
No
Ample
Low
Sb
?
Ample
Acceptable
Table 1.1: Possible alloying additions to Sn for alternative solder with their toxicity, availability and costs. Considering these facts, it can be stated that the solder will contain Sn with additions of Ag, Cu, Bi, and possibly Zn and Sb. Au and In can only be used in very small amounts. Some lead-free solders are already in use. Eutectic Sn-Ag can be used like it is already used in plumbing industry. Eutectic Sn-Bi can be used for low temperature applications. Promising other candidates are Sn-Ag-Sb, Sn-Ag-Cu, SnAg-Bi and eutectic Sn-Cu.
8
Introduction
Except for making the replacement solder lead-free, it is also of importance that the substrate and components will be lead-free. The most common substrate material is Cu, but Cu is prone to oxidation and, therefore, the wettability is not sufficient. To improve this, and to diminish the surface roughness, a top layer of Ni is generally used, often with a coating of precious metal in order to further improve wettability, corrosion resistance and shelf life. Another reason for making the substrate lead-free is that some of the elements mentioned in table 1.1 form a low-melting phase with Pb. E.g., a eutectic Sn-Bi solder would react with a substrate or component material containing Pb forming a lowmelting point ternary SnBiPb phase (96 oC) [29]. The formation of such a low melting phase in the solder joint would deteriorate the reliability of the joint and is, therefore, not wanted. Several materials have been proposed as possible constituents of substrates, the most common being Cu and Ni with possible coatings of precious metals.
1.4
Outline of This Work As is shown in the previous paragraphs research for lead-free solder alloys is
broad. Therefore, it is not possible for a university laboratory like ours to study all the necessary parameters for a ‘drop-in’ replacement. Since the laboratory of Solid State and Materials Chemistry at the Eindhoven University of Technology is well suited for fundamental research it was decided to shed some light on the fundamental aspects of solder systems, especially on the phase relations and microstructures which can be of importance to soldering technology. This means that not only phase relations in solders are studied, but also possible interaction between constituents of the solder and the underlying substrate metallization. Therefore, this work deals with the construction of isothermal crosssections of phase diagrams and the microstructures that are observed when studying the necessary alloys and interaction layers with (electron) microscopy. However, for the research described in this thesis no actual soldering techniques have been used. The phase diagrams presented in this work can be important in the soldering process to predict the existence of intermetallic compounds. It is well known that the
9
Chapter 1
presence of a thin and continuous intermetallic layer between solder and substrate is an essential requirement for good wetting and bonding. With the help of ternary phase diagrams, especially activity-composition diagrams, diffusion paths predictions can be made about the presence of intermetallics and the sequence of formation [30, 31]. The temperature chosen in this work remains under 260 oC, because this temperature is considered the limit for soldering processes so that the epoxy substrate and components will not be damaged. Although solder only stays at such elevated temperatures for tens of seconds, the microstructural evolution at lower temperatures can span several years. Therefore, evaluation of the microstructures and phase relations after hours of annealing at elevated temperatures remains relevant. The elements used in this investigation can all be used in soldering technology as indicated in the previous paragraph. Although at the time of writing of this thesis several lead-free solders were in use already, a real drop-in replacement for (near) eutectic Sn-Pb has not been found. Fundamental knowledge about phase relations and microstructures will be of help in finding a feasible solution.
References 1
J. Wolters, ‘Zur Geschichte der Löttechnik’, Degussa Frankfurt, Germany (1978)
2
M. Kranzberg and C.W. Pursell, Jr., eds., ‘Technology in Western Civilization, vol. 1, New York Oxford University Press, USA, (1967) 165
3
C. Singer, E.J. Holmyard and A.R. Hall, eds., ‘A History of Technology’, vol. 1, Oxford University Press Oxford, UK, (1955) 649 - 653
4
J.G. Landels, “ Engineering in the Ancient World”, Chatto & Windus London, UK, (1978) 42
5
P.T. Vianco and D.R. Frear, J. of Met. 7 (1993) 14
6
K. Gilleo, Circuits Assembly, 10 (1994) 30
7
G. Demortier, Archeological Chemistry, (1989) 249
8
H.C. Hoover and L.H. Hoover translation from “De Re Metallica” by Georgius Agricola, Dover Publications, Inc. New York USA (1950) 221
10
Introduction
9
W.J. Plumbridge, J. Mat. Sc. (1996) 2501
10
Marcus Vitruvius Pollio, ‘De Architectura, liber VIII’ (www.ukans.edu/history/index/europe/ancient_rome/)
11
O. Beattie, J. Geiger and J.J. Geiger, ‘Frozen in Time’, Greystone Publishing Vancouver Canada (2000)
12
E.R. Monsalve, ‘Lead Ingestion Hazard in Hand Soldering Environments’, in Proceedings of the 8th Annual Soldering Technology and Product Assurance Seminar, Naval Weapons Center, China Lake, CA USA, February 1984
13
M. Abtew, G. Selvaduray, Mat. Sc. Eng. 27 (2000) 95
14
Draft Proposal for a European Parliament and Council Directive on Waste Electrical and Electronic Equipment (WEEE), Brussels Belgium (2000)
15
‘Lead-free Solder Project: Final Report’; NCMS, Ann Arbor Michigan USA (1997)
16
B.P. Richards, K. Nimmo et al, ‘Lead-free Soldering’, Department of Trade and Industry, UK, (1999 and update 2000) (www.lead-free.org/)
17
W. Peng, K. Zeng and J.K. Kivilahti, ‘A Literature Review on Potential Leadfree Solder Systems’, Internal Report, Helsinki University of Technology, Espoo Finland (2000)
18
R. Brandt, J. Knarren, G. Manders et al., “Lead-free Solders for Electronic Applications”, Internal Report, MDP Project Group “Lead-free Soldering”, Eindhoven University of Technology, Eindhoven the Netherlands (1999)
19
‘Improved Design Life and Environmentally Aware Manufacturing of Electronics Assemblies by Lead-Free Soldering’: ‘IDEALS’, (1999) (www.caswelltechnology.co.uk/tech/emtec.htm)
20
W.L. Winterbottom, J. of Met. 7 (1993) 20
21
G. Stix, Scientific American 12 (1999) 25
22
J.D. Raby, R.W. Johnson, Electronic Packaging &Production 8 (1999) 18
23
H. Steen, Electronic Packaging & Production 12 (1994) 32
24
N.C Lee, Circuits Assembly 4 (1998) 64
25
N.C. Lee, Soldering and Surface Mount Technology 26 (1997) 65
26
H.H. Manko, Electronic Packaging & Production 2 (1995) 70
27
T. Reinikainen and J.K. Kivilahti, Met. Mat. Trans. 30A (1999) 123
28
J. Kivilahti, IEEE Trans. Comp., Pack. Man. Tech. B, 18 (1995) 326
11
Chapter 1
29
R.R. Tummala, E.J. Rymaszewski, A.G. Klopfenstein, ‘Microelectronics Packaging Handbook, subsystem Packaging part III’, second edition Chapman & Hall, New York USA (1989) 221
30
K. Zeng, J.K. Kivilahti, J. Electron Mater. 30 (2001) 35
31
F.J.J. van Loo, Prog. Solid State Chem. 20 (1990) 47
Further reading Since this thesis does not deal with soldering directly, some references for information about the process of soldering are cited here: A.1
R.J. Klein Wassink, “Soldering in Electronics”, 2nd edition, Electrochemical Publications Limited Port Erin UK (1989)
A.2
J.S. Hwang, “Modern Solder Technology for Competitive Electronics Manufacturing”, McGraw-Hill New York USA (1996)
A.3
G. Humpston, D.M. Jacobsen, “Principles of Soldering and Brazing”, ASM International, Metals Park OH USA (1993)
A.4
E.E. Kluizenaar, Philips Electron Optics Bulletin 131 Dec (1991) 3
More background relating to the search for lead-free solders and difficulties in changing the solder process can be found in the following literature: B.1
S. Crum, Electronic Packaging & Production 1 (2000) 26
B.2
J.D. Raby, R.W. Johnson, Electronic Packaging & Production 8 (1999) 18
B.3
B. Gilbert, Surface Mount Technology 3 (1997) 66
B.4
J. Glazer, International Materials Review 40 2 (1995) 65 (review)
B.5
J. Glazer, J. Electron. Mater. 23 8 (1994) 693
B.6
B. Hampshire, Soldering and Surface Mount Technology 25 (1997) 11
B.7
S. Hwang, Z. Guo, Circuit World 20 4 (1994) 19
B.8
S. Hwang, Circuits Assembly 10 (1993) 32
B.9
J. Plumbridge, Soldering & Surface Mount Technology 12 1 (2000) 32
B.10
K. Seelig, Circuits Assembly 10 (1995) 46
B.11
J.H. Vincent, G. Humpston, Circuits Assembly 7 (1994) 38
B.12
M. Witt, Surface Mount Technology 10 (1996) 70
B.13
E.P. Wood, K.L. Nimmo, J. Electron. Mater. 23 8 (1994) 709
12
Chapter 2 Theoretical Framework
The work presented in this thesis deals mainly with the chemical interaction between solder- and substrate- material. Equilibrium thermodynamics and diffusion kinetics are used to rationalize the variation of chemical composition in the reaction zone between solder and substrate. Understanding of phase diagrams is essential for this approach and, therefore, this chapter will start with this aspect. By its generality this combined thermodynamic and diffusion kinetic approach is well suited to provide a uniform treatment for understanding chemical solder interactions with substrate materials.
2.1 Phase Diagrams In the present work phase diagrams (ternary isothermal cross-sections) will be extensively used to rationalize behavior in a solder system. A phase diagram is the visual representation of the thermodynamic relationship of phases under specified circumstances. Most often equilibrium phase diagrams are used, which can be
13
Chapter 2
constructed by means of calculation if knowledge of the Gibbs energy curves as a function of their variables is available. These data are often not available and then calculation can be used as a supplementary method to experimental determination of phase equilibria. In this thesis the equilibria are determined experimentally. The topology of phase diagrams is governed by the well-known phase rule, introduced by Gibbs in the late 19th century: F =C−P+2
(2.1)
where F is the number of variables that may be varied independently without changing the number of phases present. C is the number of components present in the system and P is the number of phases. The number 2 represent two variables, being temperature and pressure. In this investigation we often will make use of the so-called diffusion couple technique. A diffusion couple consists of two materials (end-members) brought into intimate contact. During diffusion annealing at a fixed temperature and pressure the materials will interdiffuse and form a diffusion or reaction zone, which can consist of several phases. In a binary diffusion couple at a fixed temperature and pressure, the activity of one element should also be a variable; the activity of the other component cannot vary independently because of the Gibbs-Duhem relationship. It follows that only singlephase regions can exist throughout the whole binary couple. In ternary couples, however, two-phase regions may exist. Examples will be given in the following sections. 2.1.1 Diffusion Path In the treatment of diffusion couples that follows, kinetic barriers for nucleation of new phases are neglected and local thermodynamic equilibria are assumed at the interfaces. This implies that the chemical potentials (activities) of the species change continuously within a phase layer and have the same value at both sides of an interface, which means that reactions are very rapid and that diffusion is rate limiting.
14
Theoretical Framework
In a ternary system it is possible to develop two-phase areas in the diffusion zone because of the extra degree of freedom. The diffusion zone morphology, which develops during solid-state interaction in a ternary couple, is defined by type, structure, number, shape and topological arrangement of the newly formed phases. The resulting microstructure of the reaction zone can be visualized with the aid of the so-called diffusion path. This is a line on the ternary isotherm, representing the locus of the average composition in planes parallel to the original interface throughout the diffusion zone. Naturally, the diffusion path in a ternary system must fulfill the law of conservation of mass. If no material is lost or created during the interaction, then the diffusion path is forced to cross the straight line between the end-members of the reaction couple (so-called mass balance line) at least once. If phases are separated by planar interfaces, the diffusion path crosses the twophase region on the isotherm parallel to a tie-line and local equilibrium can be assumed along the whole interface. However, this is not necessarily the case: regions of supersaturation can be formed near the interfaces, which will give rise to wavy interfaces or isolated precipitates. Kirkaldy and Brown [1] formulated a number of rules, which relate the composition and morphology of the reaction zone to the phase diagram. Later, these rules were conventionalized by Clark [2]. In recent years, the research in multicomponent diffusion took on a new lease of life, due primarily to the availability of computational power matching the complexity of the available algorithms. The reader seeking more information on this rather complex subject should look up the relevant work of the research groups of Kirkaldy [3], Morral [4] and Ågren [5]. Referring for details especially to Ref. [2] we will just summarize here the main ideas, using a hypothetical reaction couple of an A-B-C- ternary system shown in Fig. 2.1.
15
Chapter 2
Fig. 2.1: A reaction zone structure in a hypothetical couple A/Z of the A-B-C system (on the left) and the corresponding diffusion path plotted on the isotherm of the ternary diagram (on the right). The low-case letters relate the structure to the appropriate composition on the isotherm. (Note: all the paths in three-phase fields must be denoted by dashed lines, as a three-phase layer cannot form in a ternary diffusion couple.) For a given couple under conditions of chemical equilibrium the reaction path involves a time-independent sequence of intermediate layers. The plot gives information about the order of the product layers, their morphology and their compositions. For example, in the hypothetical system shown in Fig. 2.1, a solid line crossing a single phase field on the isothermal section (e.g. h-i) denotes an existing layer of that phase in the reaction zone of the couple A/Z. A dashed line parallel to a tie-line in a two-phase field (g-h) represents a straight interface between two single phases. A solid line crossing tie-line on the isotherm (b-c) represents a locally equilibrated two-phase zone (in fact, a wavy interface) in the couple. A solid line entering a two-phase field and returning to the same phase field (d-e-f) represents a region of isolated precipitates. A dashed line crossing a three-phase field (e.g. i-j or kl) implies an interface in the diffusion structure with equilibrium between three
16
Theoretical Framework
phases, either a two-phase layer adjacent to a layer consisting of a different phase (e.g. i/j interface) or adjacent two-phase layers with one common phase (e.g. k/l interface).
a)
b) Fig. 2.2: a) Microstructure of the diffusion zone in Si/Co 70 at.% Ni diffusion couple after reaction at 800 oC for 400 h (BEI) b) The corresponding isothermal cross-section through the ternary phase diagram Co-Ni-Si at 800 oC with corresponding diffusion path (black line) [6].
17
Chapter 2
In a practical example, e.g. the Ni-Co-Si system, this theory can be put to use. In Fig. 2.2a a Backscatter Electron Image (BEI) is shown of a Co30Ni70 vs. Si diffusion couple. Fig. 2.2b displays the corresponding isothermal cross-section plus the diffusion path as measured from the composition of the diffusion zone, using the notation of Fig. 2.1.
2.1.2 Predictability of Diffusion Paths in Ternary Systems Looking to the isothermal cross-section of the Co-Ni-Si system in Fig. 2.2b one could imagine hundreds of diffusion paths, which would all fulfill the massbalance requirements. Is it possible to predict why the specific path given in Fig. 2.2a is followed? Actually, a full prediction is not yet possible but one can exclude many of these paths using irreversible thermodynamics considerations [20]. The intrinsic diffusion fluxes of all species (see section 2.2.2) are directed towards the side where its thermodynamic potential or chemical activity is lowered. If one knows the thermodynamics of the system, then a plot can be made of the chemical activity of an element i as a function of the molar ratio of the other two elements. It is then simple to see whether the activity of the diffusing element i fulfills the requirement of continuous decreasing, starting from the end-member where its activity is highest. Often qualitative considerations can be enough to exclude diffusion paths. For instance, the path Ni70Co30 N Co/Co2Si/CoSi/Ni2Si/Ni3Si2/NiSi/NiSi2/Si is possible from mass-balance point of view, but clearly impossible from thermodynamic point of view, because the activity of Ni, that has to diffuse through Co and Co-silicides where its activity becomes virtually zero, will certainly raise when it enters Ni2Si. This is not allowed. In the present work the relevant thermodynamic data are not known well enough to make a priori predictions about the diffusion paths, mainly because of the occurrence of extended solid solubilities or ternary phases, which are described in this study for the first time.
18
Theoretical Framework
2.1.3 Variations of the Diffusion Couple Technique There are several variations of the diffusion couple method. In the first variant, the sample to be studied is a classical semi-infinite diffusion couple, which means that after the diffusion annealing the couple ends still have their original compositions. If volume diffusion in a semi-infinite couple is the rate-limiting step, local equilibrium is supposed to exist, in which case the rules described above can be used to relate the reaction zone morphology, developed during isothermal diffusion, to the phase diagram. The main feature of this variant of the diffusion couple method is that the phase composition of the reaction zone is independent of time and that the diffusion path is fixed. The versatility of this technique in constructing isothermal crosssections of ternary systems has been demonstrated repeatedly (see e.g. Refs. [7-10]).
a)
b)
Fig. 2.3: Determination of the phase equilibria and the diffusion path PQ on the isotherm of the ternary A-B-C system using two-phase alloys as end-members (a) and schematic view of a possible reaction zone in a hypothetical diffusion couple P/Q (b). It is often necessary to investigate couples with end-members of various compositions in order to get all the information needed to find the phase relations at the annealing temperature. However, the necessary number of samples can be decreased appreciably by using polyphase terminal materials in diffusion couples. In the case of the study of, for instance, a ternary system, the chance “to hit” interfaces at which three phases are in equilibrium, is much larger when two-phase alloys are used as end-members. Schematically this procedure is shown in Fig. 2.3. If a diffusion couple between P and Q is assembled, the reaction zone after annealing at a specified 19
Chapter 2
temperature might exhibit the morphology as indicated in Fig. 2.3b. In area 1, microprobe measurements will reveal the three-phase equilibrium α + µ + T, whereas from area 2 the equilibrium triangle β + γ + T existing on the isotherm can be found. A practical example of this method can be found in Chapter 4.
Fig. 2.4: Schematic view of the reaction zones and diffusion paths on an isotherm after increasing annealing times t1 to t4: a) initial “sandwich” sample; b) reaction zone morphology for different annealing times; c) diffusion paths for various annealing times. Further development of this technique for studying phase diagrams is connected with changing the “macrostructure” of the classical diffusion couple. A sample is prepared by joining two plane-parallel slices of metal (alloy) through a thin layer of the third metal (alloy) as is shown schematically in Fig. 2.4a. In such a layered system, in which the central part is eventually consumed, the diffusion path is not fixed as in the semi-infinite couple. The phase composition of the complex diffusion zone is changing continuously with time as a result of the overlapping of two quasi-equilibrated diffusion zones. To relate the morphology and composition of
20
Theoretical Framework
the reaction zone to the phase diagram, rules similar to those explained before still can be used. For instance, in Fig. 2.4 the composition found in the phases α1 and α2 at t = t3 and t = t4, respectively are the end points of two tie-lines in the two-phase region. Before ending this part of the discussion, it should be mentioned that within the past dozen years these variants of the diffusion couple technique have proven themselves as a valuable tool in phase diagram studies (see, for example, Refs. [1114]). The efficiency of this method is very high. By making only one finite “sandwich” sample, re-annealing and investigating after various annealing times, a great deal of information can be gained about the whole isotherm. As demonstrated in the preceding sections, at an elementary level, the diffusion couple technique is nothing more than a tool for establishing a correlation between the morphology developed in the diffusion zone of the couple and a certain type of phase relations in the system. However, it must be added immediately that like in the case of other seemingly simple methods used in materials science, the proper use of diffusion couples for determination of multicomponent phase diagrams is by no means the trivial procedure as it might look at first sight. A number of error sources may appear when multiphase diffusion experiments are used for establishing phase equilibria. Apart from these qualitative aspects, the quantitative investigation of diffusion couples provides valuable diffusion parameters, like diffusion coefficients and activation energies (see section 2.2.2).
2.1.4 Error Sources Encountered in the Diffusion Couple Technique The experimental results may contain errors directly attributable to the nature of the sample. One of the possible dangers of technical importance is a system in which the terminal compositions (end-members) are solids, but a liquid phase exists at the annealing temperature. It is then possible for the diffusion path to wander into this field, with disastrous results indeed! Poor adherence at the interfaces in the diffusion zone and accelerated reaction rates due to defects such as cracks, grain boundaries, etc., may also render the interpretation of diffusion couple experiments cumbersome. This problem is very likely to occur in soldering systems, since the homologous
21
Chapter 2
T temperature Thom = Tmelt
is usually close to one and, of course, liquid is present
while soldering. Another source of error is in the experimental measurements themselves. The difficulties connected with the accurate determination of the boundary concentrations in the reaction zone are a problem for both semi-infinite and finite diffusion couple techniques. Several items concerning the electron-beam microanalytical techniques used will be treated in Chapter 3 (see also [15]). Another group of problems arises from the formation of a quasi-equilibrated diffusion zone. When the diffusion couple techniques are used for phase diagram determination, it is of fundamental importance to be sure that equilibrium values are “really” attained at the interphase interfaces and that all equilibrium phases pertaining to the system under the conditions of the diffusion couple experiment, are formed in the reaction zone. Sometimes certain phases seem to be missing in a diffusion couple when investigated by microscopic or microprobe analysis. One of the reasons for the absence of an equilibrium phase might be the presence of a barrier layer at the interface, such as, for example, oxide films at the contact surface or the presence of impurities in the starting materials. In the latter case, the segregation of impurities, which may be present only in the ppm-range in one of the end-members can cause enrichment in the diffusion zone, making nucleation of a certain phase difficult. After short annealing times phases can be absent because of slow reaction which is then the rate-limiting step instead of diffusion (see p. 30). Even though the absence of certain phases might be due primarily to difficulties in nucleation, it is important to point out that the apparent absence of a particular intermediate phase in a diffusion zone cannot automatically be interpreted as the result of nucleation problems. It is possible that the phase is present in such a minute quantity that it cannot be determined easily by the experimental techniques available. In the overall context of a diffusion couple experiment, the question of the apparent absence of a certain equilibrium phase in a reaction zone or, conversely, the formation of metastable phases during interaction, becomes an important issue. For instance, in a “notorious” example of a Ti/Al diffusion couple only TiAl3 was found [16]. Yet,
22
Theoretical Framework
several other intermetallic phases should have been formed according to the phase diagram [17]. On the other hand, when incremental Ti/TiAl3 diffusion couples were annealed at 800 °C, all the possible intermetallic phases according to the equilibrium phase diagram, were indeed found in the reaction zone: Ti/Ti3Al/TiAl/TiAl2/TiAl3. However, when a layer of Al was joined on the outside of the TiAl3 layer of the above couple and annealed at 625 °C for 15 hrs, the compounds Ti3Al, TiAl and TiAl2 disappeared, resulting in the basic original configuration of Ti/TiAl3/Al. Clearly, in this case nucleation of Ti3Al, TiAl, TiAl2 cannot be the problem. Rather, the apparent absence of the other phases is due to the relatively slow diffusion in these compounds, i.e. due to slow growth kinetics. An additional reason for caution here is that occasionally a non-equilibrium phase, often stabilized by impurities present in terminal materials, can grow in a diffusion zone. Very telling examples of such situation are carbides like Mo5Si3C [18] or Mo6Ni6C [19], which might be confused with the purely binary phases Mo5Si3 and MoNi. The same is true for oxides, e.g. Ti4Ni2O and nitrides like Nb4Ni2N, which might be confused with the binary compound Ti2Ni or with an, in fact non-existing, binary “compound” Nb2Ni. An easy way to verify the purely binary character of an equilibrium phase layer is the use of incremental couples. The terminal compositions are then chosen quite close to the apparent phase in question, in such a way that only this phase might be formed. The end-members may be single-phase or two-phase alloys. A true equilibrium phase then grows parabolically with time as a relatively thick layer, whereas an impurity-stabilized phase ceases to grow after some time since the impurity from the end-member is totally consumed [20].
2.2 Diffusion in Solids Thermodynamics can, to a certain extent, predict which phases are stable under given temperature and pressure conditions in a solder system. The next step is to develop an understanding of how things happen and how fast. Therefore, understanding diffusion is essential.
23
Chapter 2
2.2.1 Atom movements If a dense reaction layer is formed in a solid-state diffusion couple, then diffusion of the species through that layer is necessary in order for the reaction to proceed. The diffusion is nothing else than movement of atoms or ions under the influence of a certain driving force. This force can have a thermal, chemical and/or mechanical origin. The fundamental mechanism by which atoms move through crystals depends on crystal structure, atomic sizes and the extent of defects in the crystals. Solid state diffusion can be divided into two major categories: volume or bulk diffusion and short-circuit diffusion. At elevated temperatures volume diffusion is normally predominant. At lower temperatures (T < 0.5 Tm for pure materials) short-circuit diffusion, often associated with grain boundaries, can become predominant.
2.2.2 Reactive Diffusion In a strictly binary system, like Co and Si shown in Fig. 2.5, only straight interfaces can develop. This is dictated by the phase rule mentioned in section 2.1. The fact that the layers are not completely straight can be explained in terms of anisotropic diffusion behavior through the product crystals or in terms of grain boundary diffusion, making the diffusion two-dimensional. Another explanation can be the presence of impurities, making the system in fact ternary. If the total volume remains constant we can express the interdiffusion flux of ~ atoms, J , of for example cobalt across any plane in a diffusion couple by Fick’s first law: ~ ~ ∂C J Co = − D( Co ) ∂x
(2.2)
where the gradient is taken parallel to the x-axis (m) and perpendicular to the ~ original interface between cobalt and silicon. This interdiffusion flux J Co is measured with respect to the position of the original contact interface (Matano plane), which has ~ a fixed position with respect to the ends of the diffusion couple. D is called the
24
Theoretical Framework
chemical or interdiffusion coefficient (m2/s) and CCo is the concentration of cobalt (mol/m3).
Fig. 2.5: Reaction zone of a Co/Si diffusion couple (BEI) annealed for 64 h at 800 oC. The label K indicates the position of the Kirkendall plane. Under “transient” or “unsteady-state” conditions, the concentration gradient and, therefore, the flux of cobalt atoms in Eq. 2.2 change with time. This can be expressed by Fick’s second law: ∂CCo ∂ ~ ∂C = [ D Co ] ∂t ∂x ∂x
(2.3)
~ In the case of constant total volume the interdiffusion coefficient D (Co*) at a * certain concentration CCo can be found from the penetration plot of CCo vs. x by the
Matano-Boltzmann solution of Fick’s second law: 1 dx ~ * D(CCo )=− ( * ) 2t dCCo
* C Co
∫ xdC
Co
(2.4)
− C Co
− where CCo is the concentration of Co in the left-hand end-member of the
couple and x is the position parameter with respect to the Matano plane, defined as: * C Co
∫ xdC
Co
=0
(2.5)
− C Co
25
Chapter 2
The relationship obtained in Eq. 2.4 takes into account the concentration dependence of the diffusion coefficient and is only valid if the parabolic time dependence of the growth of phases is preserved. The diffusion couple should also be semi-infinite, which means that the ends of both couple halves still have the original ~ composition. Utilization of the method for measuring D(C B* ) is illustrated schematically in Fig. 2.6. The position of the Matano interface (xo=0) can be found by making the vertically hatched areas in Fig 2.6 equal. The integral in Eq. 2.4 equals the horizontally hatched area. ~ The interdiffusion coefficient, D , can be described as an “overall” measure for the redistribution of the elements after diffusion has taken place. It can be used for the description of the combined diffusion behavior of both species (e.g. cobalt and silicon) in a certain phase. It contains no information on the relative diffusivities of the species involved in the diffusion. For that purpose, the intrinsic diffusion coefficients (DCo and DSi) have to be determined by measuring the intrinsic diffusion fluxes JCo and JSi relative to a set of markers, originally placed inside the diffusion zone. In practice, markers put at the original interface, also called Kirkendall plane, form a convenient plane of reference. Practice shows that the position of this plane can be revealed by “natural” markers such as small debris (as a result of the metallographic preparation of the starting materials prior to joining), pores, or by a sudden change of crystal morphology [20]. M +
CB
γ
↑ CB β
*
CB
CB
α
*
x xo=0 Fig. 2.6: Concentration of a binary diffusion couple between two alloys with starting composition CB+ and CB− . The plane M at xo = 0 is the Matano plane.
26
Theoretical Framework
When a vacancy mechanism is operative, the unequal diffusion fluxes of the components with respect to these inert markers in a phase are compensated by a flux of vacancies in the same direction as the flux of the slower component. The vacancies can coagulate and form voids (Kirkendall voiding or Frenkel defects). For example, in the Co/Si diffusion couple (Fig. 2.5) the Kirkendall plane is located Co2Si near to the Co/reaction product interface, which leads to the conclusion that cobalt is the most mobile species in the intermetallic compound Co2Si. The velocity of the Kirkendall frame of reference with respect to the Matano plane of reference is given by Eq. 2.6: v = VCo ( DCo − DSi )
∂CCo ∂x
(2.6)
where VCo is the partial molar volume of cobalt, and DCo, DSi are intrinsic diffusion coefficients. The relation between the interdiffusion and the intrinsic diffusion coefficients is given by the Darken equation: ~ D = CCoVCo DSi + CSiVSi DCo
(2.7)
In case of a narrow homogeneity range of phases, so called “line-compounds” (e.g. CoSi2 see Fig. 2.5), it is practically not possible to measure a concentration gradient and to determine the interdiffusion coefficient. Therefore, it is impossible to apply the conventional Matano-Boltzmann solution to Fick’s second law. To avoid this problem Wagner introduced the concept of the integrated diffusion coefficient, Dint [21]. This material constant is defined for a phase as the interdiffusion coefficient in this phase integrated over its (unknown) homogeneity limits N’and N’’ (molar fractions): N ''
~ Dint = ∫ DdN
(2.8)
N'
If the diffusion couple consists of only phases with a very narrow region of homogeneity, Dint of phase i can be found using the following equation:
27
Chapter 2
( N + − N (i ) ) ∆x (i ) 2 D = (N − N ) N+ − N− 2t (i ) v = i −1 v = n −1 ( i ) + Vm Vm − − v (v ) (i ) (i ) ( N − N ) ( N − N ) ∆ x + ( N − N ) ( N + − N v )∆x ( v ) ∑ ∑ (i ) (v) (v ) ∆x v = 2 Vm v = i +1 Vm + + − 2t N −N (2.9) (i ) int
(i )
−
Fig. 2.7: Schematic concentration profile of the diffusion couple Co/Si showing the formation of Co2Si, CoSi, CoSi2. No solid solution of silicon in cobalt is taken into account (due to negligible thickness). Co2Si and CoSi are considered as linecompounds. The phases are numbered sequentially by i, starting with i=1 for one endmember of the couple and finishing with i=n, v is the serial number of the product phase layers, N(i) is the mole fraction of one of the components in phase i. The superscripts + and – refer to the mole fraction of that component in the end-member of the diffusion couple. Vm is the molar volume, ∆x is the layer thickness, t the reaction time. The integrated diffusion coefficient for e.g. the CoSi phase can be calculated through [22]:
28
Theoretical Framework
2
∆x a * b (∆x(CoSi ) ) ( CoSi ) Dint = + ( CoSi ) 2t 2t a + b
VmCoSi VmCoSi + b * P * a * Q * VmCoSi 2 VmCo 2 Si a+b
(2.10)
with P and Q equal to the black areas in Fig 2.7, and a = N v − N − and b = N+Nv from Eq 2.9. Similarly, the integrated diffusion coefficients for the phases Co2Si and CoSi2 can be determined. Since the integrated diffusion coefficient is a material constant, its value does not depend on the starting materials in the reaction couple. This means that the value of Dint in the CoSi phase will be the same in a Co/Si and in a Co2Si/CoSi2 incremental diffusion couple, if annealed at the same temperature. For the CoSi phase, for ~ instance, we can estimate a value for the interdiffusion coefficient D from a modification of Eq. 2.8: ~ Dint ≈ Dav ∆N
(2.11)
~ where D is the average interdiffusion coefficient and ∆N the difference in mole fraction in the CoSi phase from the cobalt to the silicon side. The analysis can also be used in ternary diffusion couples as long as no solid solutions are formed and only straight bounded line-compounds are formed. The results are consistent with data found in the constituent binary systems as long as the activities at the phase interfaces are the same [22]. All the analysis to this point assumes that the overall reactions are controlled by the same diffusion process throughout the whole annealing procedure. Then the layer thickness ∆x and the annealing time t are related through the parabolic growth constant k p : (∆x) 2 = 2k pt
(2.12)
In practice, several deviations from parabolic growth may exist due to, for instance, impurities or oxide films present in the starting materials, and to a contribution of grain boundary diffusion changing with time. Further, the product layer growth can be linear with time as a result of a reaction-limited step as explained by Dybkov [23] and Philibert [24].
29
Chapter 2
Dybkov considers the flux of a compound through a reaction layer γ of thickness
∆x between pure A and B as consisting of two parts, viz. a diffusion
limited flux J D = D
(C1 − C2 ) (as treated before, with C1, 2 the actual concentration of ∆x
the fastest diffusing component A in γ at the interfaces γ/A and γ/B, respectively) and a reaction limited flux J R = K (C2 − C2eq ) . In that case, the reaction at interface γ/A is considered to be infinitely rapid, whereas at the γ/B interface it takes some time to reach the equilibrium concentration C2eq . So, in the beginning C2 > C2eq . In the quasi-stationary situation JD=JR=J, which leads to a combined growth equation 2
∆x (∆x) + • = t − to • 2D K
where D • = D
(C1 − C2eq ) (C − C2eq ) and K • = K 1 C2 C2
So, if D>>K∆x the growth will be reaction limited and linear in time, whereas for D21), like combinations of the elements Fe/Ni, Cu/Co, etc. In general, in order to avoid the problem of fluorescence (and absorption) “uncertainty”, a number of correction schemes have been introduced [1, 2]. One should, however, keep in mind that in all these correction procedures it is implicitly assumed that the primary and secondary production of X-ray radiation as well as subsequent absorption all take place in the same homogeneous (single phased) matrix. Apparently, these conditions are violated to increasing extent as the incident electron beam approaches an interface or when the size of particles decreases below a certain limit. Corrections for this effect are possible but difficult [3]. It is best to determine how large the effect is by using undiffused couples and to correct the data accordingly. In a binary diffusion couple, usually one of the elements does not suffer from the fluorescence effect and one could measure the concentration of that element by microprobe analysis to describe the distribution of elements across the reaction zone. If, for instance, pure Cu and Co are clamped together, without any diffusion taking place, CoKα characteristic X-ray radiation can apparently be measured at a distance of up to 40 µm from the interface in the pure Cu [3]. About 4.8 at.% of Co appears to be in Cu at the interface. Measuring CuKα radiation diminishes the error, but still an apparent concentration of 2 at.% is found in cobalt near the interface,
42
Experimental Procedures
whereas the Cu X-ray radiation can be detected at about 15 µm from the interface in cobalt. Obviously, in the case of a ternary system, the situation can be even more intricate. The reader interested in greater details concerning the correction procedures is invited to turn his attention to the references cited.
References 1
J.I. Goldstein, D.E. Newbury, P. Echlin, D.C. Joy, A.D. Romig, Jr., C.E. Lyman, C. Fiori and E. Lifshin, ‘Scanning Electron Microscopy and X-ray Microanalysis’, Plenum Press, New York USA (1992)
2
G.F. Bastin, J.M. Dijkstra and H.J.M. Heijligers , X-Ray Spectrom. 27 (1998) 3
3
G.F. Bastin, F.J.J. van Loo, P.J.C. Voster and J.W.G.A. Vrolijk, Scanning 5 (1983) 172
43
‘First he (van Helsing) took out a soldering iron and some plumbing solder, and then a small oil lamp, which gave out, when lit in a corner of the tomb, gas which burned at fierce heat with a blue flame; then his operating knives, which he placed to hand; and last a round wooden stake, some two and a half three inches thick and about three feet long.’ Bram Stoker - Dracula, p. 256, 1897 Penguin Popular Classics, England
44
Chapter 41 Phase Relations in the Sn-Ag-Sb system at 220 oC
4.1 Introduction Ternary Sn-Ag-Sb alloys are worthy of consideration as lead-free solders for applications in which interconnections might be subjected to large stress and strain or when a relatively high service temperature is required. Knowledge about the ternary phase diagram Sn-Ag-Sb is needed in order to predict the microstructural evolution of the solder itself and the reaction products at the interface with the substrate materials that are formed during fabrication and service. No information on this ternary diagram has been reported (despite the rather intensive use of the Sn-Ag-Sb alloys as commercial solders [1]). The binary Sn-Ag and Ag-Sb phase diagrams are known quite satisfactorily [2]. In contrast, a longstanding controversy exists over the binary Sn-Sb system, especially about the formation of the intermetallic compounds. This system has often been investigated [310], but the solid state equilibria have not been firmly established.
1
This chapter is based on the following paper: P.J.T.L. Oberndorff, A.A. Kodentsov, V. Vuorinen, J.K. Kivilahti and F.J.J. van Loo, Ber. Bunsenges. Phys. Chem. 102 (1998) 1321
45
Chapter 4
The tin-antimony phase diagram proposed by Predel et al. [3], mainly based on DTA work, shows three peritectic reactions, tin (Snss) and antimony (Sbss) terminal solid solutions and also two intermetallic phases. The Sn3Sb2-compound is only present between 242 and 324 oC, and the β-phase is present in a large nonstochiometric domain extending between 43 and 60 at.% of antimony. Recently, Vassiliev et al. [10], based on the results of e.m.f.-measurements and X-ray diffraction studies, proposed a new variant for the antimony rich part of the binary Sn-Sb diagram. They present evidence for the existence of four intermetallic phases in the system: β-SnSb, β’-Sn12Sb13, β’’-Sn2Sb3 and β’’’-SnSb2. According to the authors, these compounds are stable in the range from room temperature to 220 o
C. The results are astonishing since they differ so much from those of the earlier
investigations. However, no microstructural examination of the investigated alloys has been performed. Here phase relations between the binary Sn-Sb intermetallics in the solid state region of the phase diagram are investigated with the diffusion couple technique and the isothermal cross-section through the ternary Sn-Ag-Sb phase diagram at 220oC is constructed.
4.2 Intermediate phases in the binary Sn-Sb system at 220 oC Anticipating the specific results of the present study it seems worthwhile to give a rather detailed analysis of the literature data on the binary Sn-Sb intermetallics and compare these with our experimental observations. X-ray diffraction analysis of various alloys containing the “β-SnSb”-phase showed that the structure is always rhombohedral over the range from room temperature up to 220 oC and can be considered as a slightly distorted NaCl (B1)structure [8]. The unit dimensions and the degrees of deformation of the lattice vary with the composition. But, in spite of careful heat treatments, it was found that heterogeneities are always present in the sample [9]. One might expect that the antimony and “β“-phases exhibit many structural similarities. Both crystallographic cells are rhombohedral ( R 3 m ) and can be described as hexagonal packing by periodic stacking of atomic planes in the [001]-
46
Phase Relations in the Sn-Ag-Sb system at 220 oC
direction, each with hexagonal packing non-compact occupancy. In contrast to pure antimony, binary tin-antimony “β“-phases are ordered in the c-direction [001] with alternately, Sb- or Sn-atoms occupancy. It is clear that the stoichiometry of the phases can vary, introducing stacking faults in the c-direction. One can imagine a large variety of stacking sequences, even with aperiodic stacking faults. It is important to realize that the thermodynamic stability of stacking variants (with unequal stacking periods) is very close. As a result these stacking variants may exist in coalescence, especially when they nucleate inside a melt. Such biphase structures may make it difficult to obtain defined properties of an alloy at a given composition within the “β-phase” field. Another experimental problem originates from the fact that tin and antimony atoms have practically the same X-ray scattering factor [11]; therefore, the variation of the stacking period cannot be detected by X-ray diffraction. This is probably the main reason why these different biphase systems have previously been described in the literature as a continuous variation of stoichiometry inside the β-monophase region. In this respect, the study of the interaction between tin and antimony in diffusion couples provides a powerful framework for evaluating solid state equilibria in this binary system, as was explained in chapter 2. In our particular case, multiphase diffusion (reaction diffusion) experiments in combination with EPMA have an obvious advantage: it is not necessary to produce and equilibrate alloys which might still contain “heterogeneities” even after long heat treatments. In the diffusionreaction experiment the intermetallic compounds will grow during the experiments. As will be demonstrated later, this is of importance, especially where solid state equilibria involving thermodynamically very close phases (like the stacking variants within the “β-SnSb”-region) are concerned. From a fundamental point of view, ordered structures (intermetallic compounds) present special features with regard to the operation of a vacancy mechanism for diffusion. It should be mentioned that until now, no reaction diffusion studies on systems in which intermetallic phases may exist in coalescence have been reported. The microstructure of the reaction zone in the binary Sn/Sb diffusion couple after annealing at 220 oC for 264 h in vacuum is given in Fig. 4.1. A continuous layer
47
Chapter 4
of a phase exhibiting a very weak polarization effect in the optical microscope formed along the whole contact surface of the couple. On the tin-rich side, grains of the product phase had a more or less globular shape with an average size of about 20-25 µm. On the antimony-rich side, a 10-15 µm thick layer of relatively smaller grains formed.
Fig. 4.1: Backscatter Electron Image (BEI) of the transition zone in the binary Sn/Sb diffusion couple after heat treatment at 220 oC for 264 h in vacuum. K indicates the position of the Kirkendall plane. Fig. 4.2 shows the concentration profile belonging to the diffusion couple in Fig. 4.1 as determined by EPMA. Two distinct layers can easily be recognized. The average measured composition of the product phases are found to be close to the stoichiometric formulae Sn4Sb3 (more coarse grained layer) and Sn3Sb4, respectively.
48
Phase Relations in the Sn-Ag-Sb system at 220 oC
Fig. 4.2: Concentration profile across the diffusion zone of the binary Sn/Sb diffusion couple after annealing at 220 oC for 264 h in vacuum. However, the observed concentration jump could also be considered as a steep concentration gradient within a single phase. This should indicate that the diffusion coefficient as a function of the concentration within the homogeneity region of the intermetallic phase goes through a minimum, as over a small distance in the diffusion layer the magnitude of the diffusion coefficient is mainly governed by the magnitude of the reciprocal of the concentration gradient. The minimum in the diffusion coefficient could be related to the presence of some kind of ordering process within the crystal structure of the intermediate phase, leading to a minimum of structural defects for a certain composition. Indeed, in most cases of ordered alloys, the concentration dependence of the diffusion coefficient within the homogeneity range of the binary intermetallic exhibits a sharp minimum around the stoichiometric composition [13]. This corresponds to the existence of a maximum in the curve of the activation energy versus the composition. In other words, diffusion in ordered phases generally tends to be much slower than in disordered alloys.
49
Chapter 4
We believe that the measured concentration jump (Fig. 4.2) reflects the actual equilibrium between the two phases (stacking variants) Sn3Sb4 and Sn4Sb3. The presence of a continuous solid solution between these phases is in conflict with the different grain morphologies of Fig. 4.1, and the gradient seems too steep. The original interface (Kirkendall plane) is visible as a row of pores (or maybe grinding debris) and is located at the Sn4Sb3/tin- interface. Following the reasoning given in section 2.2.3 one may conclude, that for the Sn4Sb3-phase the ratio
DSb DSn
should be (very close to) zero, whereas for the Sn3Sb4-phase the absence of any marker indicates a ratio
DSb < 5. DSn
In an attempt to present further proof for the existence of the two intermetallics two incremental diffusion couples were made as explained in chapter 2 (Sn65Sb35 vs. Sn45Sb55 and Sn55Sb45 vs. Sn38Sb62). After annealing for 240 h at 220 oC a small intermetallic layer was visible in the first couple. This layer could be identified as Sn4Sb3. The second couple didn’t show any reaction layer, even after repeated annealing. Since one would expect thicker intermetallic layers in an incremental couple it is likely that contact problems are the reason for the absence or small thickness of the reaction layer. By analogy with another system, namely Sn-As [14], a Sn4Sb3 compound is to be expected in the tin-antimony system as is an equiatomic phase SnSb. Because of the larger difference in electronegativety between the components the binary intermetallics SnAs and Sn4As3 (“Sn3As2”) show up in the Sn-As system as thermodynamically very stable phases. This stability is indicated by the fact that SnAs has a strictly stoichiometric composition and that both SnAs and Sn4As3 are apparently congruently melting compounds [2]. Note that the compound SnAs has the same crystal symmetry ( R 3 m ) as the intermediate phases of the tin-antimony system. Therefore, it was very surprising to find a reaction product with the approximate composition of Sn3Sb4, but not the equiatomic SnSb in the binary tin-antimony diffusion couple. As mentioned earlier, all intermetallic phases in the binary tin-antimony system have a rhombohedral cell and can be understood as hexagonal packing by the periodic stacking of atomic planes in the [001]-direction. In order to explain the 50
Phase Relations in the Sn-Ag-Sb system at 220 oC
experimental results we propose two stacking variants each with 21 atomic layers parallel to the basal plane. The atomic planes are composed of three units of six alternating layers of Sn or Sb with hexagonal packing after which there is one repeating layer.
4.3. Phase relations in the Sn-Ag-Sb system at 220 oC The isothermal cross-section through the ternary diagram Sn-Ag-Sb at 220 oC was constructed by combining the traditional methods of equilibrated alloys and diffusion couples with two-phase end-members. More details concerning the use of this diffusion couple technique can be found in chapter 2 and [15].
Fig. 4.3: BEI of the reaction zone between Sb and a two-phase Sn-Ag alloy with the nominal composition Sn75Ag25 after anealing at 220 oC for 240 h in vacuum. K indicates the position of the Kirkendall plane.
51
Chapter 4
The microstructure of the reaction zone in the diffusion couple based on antimony and a two-phase alloy, with a nominal composition Sn75Ag25, consisting of ε-Ag3Sn and virtually pure tin, is shown in Fig. 4.3. Solid state interaction at 220 oC led to the formation of the intermetallic phases Sn4Sb3 and Sn3Sb4. The Kirkendall plane is clearly visible as a row of pores in the vicinity of the two-phase alloy/reaction product interface, implying that under the experimental conditions tin is by far the most mobile species in the Sn-rich reaction layer. It can be seen that in some areas of the diffusion zone particles of the Ag3Sn-phase, present in the initial two-phase alloy, are in direct contact with the newly formed Sn4Sb3-phase. The same is also true for the crystals of the tin-based solid solution. Two different grain morphologies can be seen within the reaction layer. EPMA gave positive identification of the formation of Sn4Sb3- and Sn3Sb4-phase layers within the transition zone. The resulting microstructure is somewhat similar to that observed in the binary Sn/Sb couples (Fig. 4.1). No solubility of silver was found in the tin-antimony intermetallics and no antimony was detected in the ε-Ag3Sn-phase. Analysis of the reaction zone morphology evolved in this diffusion couple shows the three-phase equilibrium Snss + ε-Ag3Sn + Sn4Sb3 on the Sn-Ag-Sb isotherm at 220 oC. When a two-phase alloy with a nominal composition Sn75Sb25 was used as an end-member of the diffusion couple, the interfacial reaction with silver at 220 oC for 230 h resulted in the formation of a continuous layer of the intermetallic compound εAg3Sn. Based on the mass balance and the fact that ε-Ag3Sn cannot be in equilibrium with Ag (see Fig. 4.5) one would expect a layer of ζ-Ag4(Sn,Sb) to be present between Ag and ε-Ag3Sn. The absence of this can presumably be explained by the fact that it can be too thin to be detected with SEM/EPMA. Most probably the Ag-rich composition (and thus higher homologous temperature) causes slower diffusion. Prior to making the diffusion couple, the initial binary alloy was annealed at 220 oC for 200 h in vacuum. After this heat treatment the alloy microstructure consists of very large faceted crystals of the Sn4Sb3-phase distributed inside the matrix of the tin-based solid solution containing up to 8 at.% of antimony. Actually, three phases were found in equilibrium within the diffusion zone: Sn-based solid solution (Snss), ε-Ag3Sn and Sn4Sb3. Such a morphology confirms the three-phase equilibrium Snss + ε-Ag3Sn + Sn4Sb3 in the tin-silver-antimony system at this temperature. No antimony was detected in the ε-Ag3Sn-layer and no solubility of silver in Snss and Sn4Sb3 was
52
Phase Relations in the Sn-Ag-Sb system at 220 oC
determined by EPMA. This is in agreement with the result of the Sn75Ag25/Sb diffusion couple, discussed earlier in this section. To verify the equilibrium determined with the diffusion couple technique a ternary alloy, with nominal composition Sn55Ag25Sb20, was investigated. A backscattered electron image of the alloy after heat treatment in an evacuated ampoule at 220 oC for 240 h and quenching is presented in Fig. 4.4, proving the existence of the earlier mentioned three-phase equilibrium. Approximately 8 at.% of antimony was found in the tin based solid solution and a maximum solubility of antimony in the εAg3Sn was estimated as 3 at.%.
Fig. 4.4: BEI of the Sn55Ag25Sb20 alloy after annealing at 220 oC for 240 h and quenching. Another three-phase alloy examined in the present study is Sn25Ag15Sb60, after heat treatment in an evacuated glass ampoule at 220 oC for 288 h and quenching in water. Three phases, namely Sbss, Sn3Sb4 and ζ-Ag4Sn were distinguished within the microstructure, underlining the existence of a three-phase equilibrium Sbss+ Sn3Sb4 + .-Ag4Sn on the ternary isotherm at this temperature. The maximum solubility of tin in the antimony-based solid solution was determined as ~11.5 at.%, and no silver was detected in the Sn3Sb4-phase. The average composition of the ζ-Ag4(Sn,Sb) phase present in the equilibrated alloy was found to be Ag74Sn15Sb11. 53
Chapter 4
To provide evidence that the binary Sn4Sb3 and Sn3Sb4 phases may exist in coalescence, EPMA was performed on an alloy with the nominal composition Sn30Ag40Sb30 (see Fig. 4.5), equilibrated at 220 oC for 240 h. Numerous precipitates of the ζ-Ag4(Sn,Sb) phase with the average composition Ag74.5Sn16.5Sb9 were observed within the microstructure. Although
no morphological features
indicating
heterogeneity of the matrix were found, most likely the matrix is a two-phase mixture containing the stacking variants Sn4Sb3 and Sn3Sb4. This biphase nature of the matrix was revealed by the results of microprobe measurements. Neither Sn4Sb3 nor Sn3Sb4 was detected in their pure form in the alloy, which might be due to the fact that the stacking variants Sn4Sb3 and Sn3Sb4, existing in coalescence, form rather fine crystallites. That makes it impossible to discern them within the two-phase mixture using EPMA, and the fact that tin and antimony have very close mass absorption coefficients for X-ray radiation adds even more complexity to the problem, as explained in chapter 2.
Fig. 4.5: Isothermal cross-section through the ternary phase diagram Sn-Ag-Sb at 220 o
C experimentally determined in the present study. The dots indicate equilibrated
alloys.
54
Phase Relations in the Sn-Ag-Sb system at 220 oC
In order to determine the phase boundaries on this isotherm more precisely, a number of other two-phase alloys were equilibrated at 220 oC and examined with EPMA and XRD-analysis. The results of this investigation are summarized in Table 4.1.
Alloy
Annealing time (h)
Phases present at 220 oC
Sn90Ag5Sb5
240
ε-Ag3Sn + Snss
Sn82Ag12Sb6
192
ε-Ag3Sn + Snss
Sn5Ag45Sb50
288
ε’-Ag3Sb + Sbss
Sn10Ag75Sb15
240
ζ-Ag4(Sn,Sb) + Sbss
Sn15Ag80Sb5
240
ζ-Ag4(Sn,Sb)
Table 4.1: Phases present in equilibrated alloys after annealing at 220 oC according to EPMA and X-ray diffraction analysis (the various phases are denoted by their binary formulae). Finally, the results from phase analysis and equilibrium concentration measurements in diffusion couples and equilibrated alloys led to the cross-section of the tin-silver-antimony diagram as presented in Fig. 4.5.
4.4 Concluding Remarks The experiments on binary Sn/Sb diffusion couples performed in the present work provided a clear-cut evidence for the existence of two binary phases (stacking variants), viz. Sn4Sb3 and Sn3Sb4, in this system at 220 oC. No “equiatomic” compound SnSb is present as an equilibrium phase at this temperature. The isothermal cross-section through the ternary phase diagram Sn-Ag-Sb at 220 oC was constructed by means of multiphase diffusion couples and equilibrated alloys. No ternary phases are formed in the system at this temperature. It was found that the intermetallic phases ζ-Ag4Sn and ζ-Ag4Sb, being isomorphous, form an
55
Chapter 4
extensive region of mutual solubility on the isotherm. Both intermetallic compounds Sn4Sb3 and Sn3Sb4 are in equilibrium with this solid solution. Approximately 8 at.% of antimony could dissolve in the Sn-based solid solution, and the maximum solubility of antimony in ε-Ag3Sn was estimated as ~3at.%. The solid state solubility of tin in antimony was determined as ~ 11.5 at.%, of Sn in ε’-Ag3Sb ~ 7at.% and no silver was detected in the Sn-Sb intermetallics. More generally, the diffusion couple technique proved to be a powerful research tool in establishing phase relations, especially when phases of very close thermodynamic stability are concerned.
References 1
J. Glazer, Int. Mater. Rev. 40 (1995) 65
2
T.B. Massalski, J.I. Murray, L.H. Bennett, H.Baker (eds.), ‘Binary Alloy Phase Diagrams’, ASM Metals Park OH USA (1986)
3
B. Predel and W. Schwermann, J. Inst. Met. 99 (1971) 169
4
K. Iwase, N. Aoki and O. Osawa, Sci. Rep. Tohoku Univ. 20 (1931) 353
5
R. Blondel and P. Lafitte, Compt. Rend. Acad. Sci. Paris 200 (1935) 1472
6
V.M. Goldschmidt, Skrifter Norske Videnskaps, Akad. Oslo, 8 (1927)
7
A. Osawa, Nature, 124 (1929) 14
8
E.G. Bowen and W. Morris-Jones, Phil. Mag. 12 (1931) 441
9
G. Hägg and A.G. Hybinette, Phil. Mag. 20 (1935) 913
10
V. Vassiliev, M. Lelaurain and J. Hertz, J. Alloy Comp. 247 (1997) 233
11
B.D. Cullity, ‘Elements of X-ray diffraction’, Addison-Wesley, UK, (1967)
12
M.J.H. van Dal, A.M. Gusak, C. Cserháti, A.A. Kodentsov and F.J.J. van Loo, Phys. Rev. Lett. 86 (2001) 3352
13
H. Mehrer, Mat. Trans. JIM 37 (1996) 1259
14
K. Schubert, ‘Kristallstrukturen zweikomponentiger Phasen’, Springer-Verlag Berlin Germany (1964)
15
J.A. van Beek, A.A. Kodentsov and F.J.J. van Loo, J. Alloy. Comp. 221 (1995) 108
56
Chapter 5 Phase Relations in the Sn-Ni-Cu system at 235 oC
5.1 Introduction Phase equilibria are not only important for designing possible solder systems, as discussed in the previous chapter, but also for understanding the interaction of solders with substrates. A common setup of a substrate used in the electronic industry is an epoxy substrate underlayer with Cu leads on top, which are coated with Ni, which is used as a protective layer. The main component of the solder system will probably remain Sn (melting point 232 oC), because it is a suitable element to alloy, lowering the melting point further, so it can be used for electronic systems without damaging the substrate. Another constituent of Pb-free solder can be Cu [1-4]. So Cu is very often present in the substrate as well as in the solder. The three binary phase diagrams of this system are well established [5]. Besides, there have also been many investigations on the diffusion and growth behavior of intermetallic phases in the sub-systems, probably because these systems form the cornerstone of modern electronic industry. Although Oh [6] reported a 57
Chapter 5
partial cross section at 200 oC and Gupta et al. [7] also reported partial cross-sections, mainly in the Cu-rich domain, a complete cross-section at such a low temperature has not been established [6-12], despite the importance. This is primarily due to a number of experimental difficulties in this system. Firstly, the temperature at which the soldering process is performed and which, therefore, is the temperature range relevant for a cross-section is very low, making the diffusion slow and equilibrium difficult to reach in the Cu-Ni rich alloys. Especially the low-temperature miscibility gap in the Cu-Ni system and its extension in the ternary system are virtually impossible to investigate experimentally. Another reason is that the standard analytical technique used, viz. EPMA, is very difficult to use owing to the fact that Cu and Ni are very similar in atomic weight and thus have approximately the same backscatter factor. Therefore, different phases with approximately the same amount of tin, but a variable amount of copper and nickel are difficult to distinguish with a backscatter electron image. In case of investigation with XRD, fluorescence has to be reckoned with. A third factor is that Sn is liquid at the temperature investigated here (235 oC), which means that during cooling solidification processes occur. Therefore, the phases present at the annealing temperature are often difficult to distinguish. In this chapter, these problems will be discussed and an isothermal crosssection is presented, which is determined with the help of equilibrated alloys and the diffusion couple technique.
5.2 The Binary Subsystems 5.2.1 The Cu-Ni System The binary subsystems are extensively investigated and a summary of these investigations and key results is given in this section. Copper and nickel show complete solubility in both the liquid and the solid state above 400 oC. Several researchers have concluded that a miscibility gap occurs at low temperature. However, there is disagreement about the temperature and composition range over which this gap occurs. Since equilibrium at such a low
58
Phase Relations in the Sn-Ni-Cu system at 235 oC
temperatures is difficult to establish statements about the extension of the miscibility gap in the solid solution remain speculative [13]. Chakrabarti et al. have reviewed the experimental studies and have discussed the controversial arguments in detail [14]. 5.2.2 The Cu-Sn System Although the Cu-Sn system is rather complex, it is well established. This is probably due to the fact that this system is of enormous importance for the industry, since Cu is the most commonly used substrate material and Sn is the main constituent of solder. Two intermetallics, ε-Cu3Sn and η-Cu6Sn5, occur in the temperature range, which is of interest to the electronic industry [5]. The first, ε-Cu3Sn, is orthorhombic, while the second, η-Cu6Sn5, is hexagonal. From 189 to 186 oC the η- phase transforms into the η’- phase [5,7]. About the morphology and microstructure of this compound more is told in Chapter 7. Numerous studies have been performed involving growth kinetics and diffusion, also including studies at low temperatures [15-37]. Distinction has to be made between solid state diffusion experiments [15-26] and experiments at temperatures higher than the melting point of Sn, i.e. solid-liquid experiments [25-37].
Fig. 5.1: Backscattered electron image (BEI) of a Cu/Sn diffusion couple after annealing for 800 h at 220oC.
59
Chapter 5
In solid state diffusion researchers generally agree that in bulk couples the formation of ε-Cu3Sn follows the parabolic law, i.e. is diffusion controlled [6, 15,17,19,21-26]. The microstructure of a Cu/Sn diffusion couple annealed for 800 h at 220oC is shown in Fig 5.1. This phase appears to grow at the expense of the η-Cu6Sn5, which is formed first [6, 20]. About the growth behavior of the η-Cu6Sn5 phase controversy exist [31-37]. The layer of η-Cu6Sn5 is not bound by straight interfaces but shows an irregular morphology. The probable reason is that the diffusion in the individual grains of η-Cu6Sn5 is not equally fast but depends on the crystal orientation. This makes the exact determination of diffusion rates more difficult and might account for other opinions about the growth behavior in thin films and at temperatures below 200 oC [20]. In solid state experiments above 200 oC parabolic growth behavior was observed [16, 18, 23]. In solid-liquid diffusion studies, important for the soldering process, controversy exists over the growth behavior of the intermetallic phases (for a micrograph of a diffusion couple of Cu vs. liquid Sn see Fig. 5.8). This can be explained by the fact that the liquid phase can penetrate easily at the substrate grain boundaries, resulting in a non-planar interface (also called scallop-like) [e.g. 28]. Apart from the scallops attached to the Cu, also scallops sometimes break off from the interface and float around in the Sn-matrix, resulting in peculiar microstructures. This could be caused by turbulences in the liquid state, but other explanations are possible. We will go into further detail about these phenomena in Chapter 7. The thickness of the Cu6Sn5 and Cu3Sn- layers found by us (Fig. 5.1) correspond to the values found by other researchers [6]. We found ThO2- markers, originally put at the contact interface, back inside the Cu6Sn5- layer. Applying the analysis given in section 2.2.3 it follows that for the Cu6Sn5- phase the ratio
DSn equals 1.4, in agreement with statements in the literature DCu
which claim Sn to be the fastest diffusing component in that phase [16, 18, 23]. On the other hand, it is claimed that for the Cu3Sn- phase Cu is the fastest diffusing element, in accordance with the Cu3Au –rule [6, 38].
60
Phase Relations in the Sn-Ni-Cu system at 235 oC
Since we did not find markers in Cu3Sn or at its interfaces, we conclude from D the analysis in section 2.2.3 that Cu < 3.5 . In order to be completely sure, DSn Cu 3 Sn dedicated marker experiments designed to answer this question should be carried out. 5.2.3. The Ni-Sn System The Ni-Sn system is also of importance for the electronic industry. Ni is often used as a layer on top of the Cu substrate to improve wettability. In this system three intermetallic compounds are stable at 235 oC: Ni3Sn, Ni3Sn2 and Ni3Sn4 [5]. In diffusion couples between Ni and liquid Sn at 580-800 oC, Ni3Sn and Ni3Sn2 grow parabolically while Ni3Sn4 grows very irregularly in the form of twinned crystals. It is also found that in the first two intermetallics Ni diffusion is faster, while it is speculated that Sn diffusion is faster in Ni3Sn4 [40]. At lower temperatures, it appears difficult to detect the growth of Ni3Sn and Ni3Sn2 [6, 30, 41, 42], which might be due to difficulties in nucleation. These studies concluded that only Ni3Sn4 originates during soldering and grows parabolically with time, while Ni3Sn and Ni3Sn2 are only formed afterwards. Besides nucleation difficulties this can also be due to slow kinetics at the lower temperature range, which means that the phases are too thin to be detected. Kang et al. also found parabolic growth for Ni3Sn4 when using liquid Sn at temperatures as low as 300 oC [41]. In recent work, at temperatures just above the melting point of Sn, parabolic growth for Ni3Sn4 at 235 oC has been observed with an activation energy of 28 kJ/mol, which corresponds to the value of the activation energy for diffusion of Ni in liquid Sn [42].
5.3 Equilibrated Alloys in the Ternary Sn-Ni-Cu System In order to get a general idea of phase equilibria present in the ternary system a number of equilibrated alloys were studied. Table 5.1 summarizes the alloys and the phases present after annealing according to EPMA analysis.
61
Chapter 5
Several alloys showed a three-phase microstructure, see Fig. 5.2. Analytical results showed a phase with a ternary composition around Sn44Cu27Ni29. Whether or not this is a separate ternary phase is not easy to determine.
T Sn Ni3Sn4(Cu)
Fig. 5.2: Backscatter electron image of the Sn75Cu10Ni15 alloy after annealing for 3300 h at 235 oC. T stands for Sn44Cu27Ni29. Since both Ni3Sn2 as well as η-Cu6Sn5 have the same crystal structure, see table 5.2, complete mutual solubility of these phases is expected. However, analyses gave indications for a different situation. In the alloy with a composition of Sn65Cu25Ni10 a phase with the composition of Sn44Cu27Ni29 could already be identified next to Ni- containing η-Cu6Sn5 and the Sn-phase, making this alloy a three-phase alloy. This would lead to the conclusion that Sn44Cu27Ni29 and η-Cu6Sn5 are different phases, implying that there is no complete mutual solubility between η-Cu6Sn5 and Ni3Sn2. The explanation given above is not as straightforward and simple as it looks, because both observation in optical microscopy as well as in EPMA do not unambiguously show a three-phase structure. In the case of optical microscopy this can be explained by the fact that the grains are small and dispersed over the complete sample, making it impossible to determine phase differences with the help of polarization microscopy.
62
Phase Relations in the Sn-Ni-Cu system at 235 oC
Alloy
Annealing time (h)
Phases present at 235 oC
Sn75Cu10Ni15
3300
Sn(Cu,Ni) + Ni3Sn4(Cu) + Sn44Cu27Ni29
Sn65Cu25Ni10
1700
Sn (Cu,Ni) + Cu6Sn5(Ni) + Sn44Cu27Ni29
Sn65Cu10Ni25
3300
Sn(Cu,Ni) + Ni3Sn4(Cu) + Sn44Cu27Ni29
Sn65Cu5Ni30
3300
Sn(Ni,Cu) + Ni3Sn4(Cu)
Sn60Cu10Ni30
3300
Sn(Cu,Ni) + Ni3Sn4(Cu) + Sn44Cu27Ni29
Sn50Cu15Ni35
840
Ni3Sn4(Cu) + Ni3Sn2(Cu) + Sn44Cu27Ni29
Sn50Cu5Ni45
1700
Ni3Sn4(Cu) + Ni3Sn2(Cu)
Sn45Cu43Ni12
1100
Cu6Sn5(Ni) + Sn44Cu27Ni29
Sn45Cu29Ni26
840
Sn44Cu27Ni29 + Sn
Sn45Cu20Ni35
1700
Ni3Sn4(Cu) + Ni3Sn2(Cu) + Sn44Cu27Ni29
Sn40Cu50Ni10
1700
Cu6Sn5(Ni) + Cu3Sn(Ni) + Sn44Cu27Ni29
Sn40Cu45Ni15
1100
Cu6Sn5(Ni) + Cu3Sn(Ni) + Sn44Cu27Ni29
Sn40Cu40Ni20
1100
Cu6Sn5(Ni) + Cu3Sn(Ni) + Sn44Cu27Ni29
Sn40Cu35Ni25
1100
Ni3Sn2(Cu) + Cu3Sn(Ni) + Sn44Cu27Ni29
Sn40Cu25Ni35
1100
Ni3Sn2(Cu) + Ni3Sn(Cu)
Sn35Cu20Ni45
3300
Ni3Sn2(Cu) + Ni3Sn(Cu)
Sn30Cu10Ni60
1700
Ni3Sn(Cu) + Ni3Sn2(Cu)
Sn30Cu45Ni25
840
Ni3Sn2(Cu) + Ni3Sn(Cu)
Sn20Ni40Cu40
1100
Ni3Sn(Cu) + Cu,Ni(Sn)
Sn15Cu65Ni20
840
Ni3Sn(Cu) + Cu,Ni(Sn)
Sn15Cu20Ni65
840
Ni3Sn(Cu) + Cu,Ni(Sn)
Sn10Cu45Ni45
840
Ni3Sn(Cu) + Cu,Ni(Sn)
Table 5.1: Phases present in equilibrated alloys after annealing at 235 oC according to EPMA. For clarity reasons the composition of the ternary phase is given as Sn44Cu27Ni29, although a homogeneity range in the Cu-Ni ratio is probably present. 63
Chapter 5
Phase
Pearson’s Symbol
Space Group
Lattice parameter (nm)
Sn
tI4
I41/amd
a = 0.58315 c = 0.31814
Ni
cF4
Fm3 m
a = 0.35241
Cu
cF4
Fm3 m
a = 0.36148
Ni3Sn (h)
cF16
Fm3 m
a = 0.598
Ni3Sn (l)
hP8
P63 / mmc
a = 0.5286 c = 0.4243
Cu3Sn (h)
cF16
Fm3 m
a = 0.61166
ε-Cu3Sn (l)
oC80 [43]
Cmcm
a = 0.5529 b = 4.775 c = 0.4323
Ni3Sn2
hP4
P63 / mmc
a = 0.4125 c = 0.5198
η-Cu6Sn5
hP4
P63 / mmc
a = 0.4190 c = 0.5086
η’-Cu6Sn5 (l)
..
..
a = 2.089 c = 2.515
Ni3Sn4
mC14
C2/m
a = 1.2223 b = 0.4061 c = 0.5187
Table 5.2: Phases present, relevant for this study, in the Sn-Ni-Cu -system [5,10,43]. (l) stands for low temperature variant, while (h) stands for high temperature variant. In EPMA/SEM the differentiation between the Sn44Cu27Ni29-phase and the ηCu6Sn5-phase is extremely difficult, since the Sn content in both phases is the same and the backscatter factor for Cu and Ni are also similar, resulting in a poor contrast
64
Phase Relations in the Sn-Ni-Cu system at 235 oC
between the two phases, making visual identification of these phases with help of a Backscatter Electron Image difficult. The Sn phase in the Sn65Cu25Ni10 alloy is easy to determine in both light microscopy as well as EPMA. The presence of this phase, would mean that if the alloy consists of two phases, the other phase would have a composition which is on a straight line from pure Sn through the original composition of Sn65Cu25Ni10 in order to fulfill the mass balance. However, the analyses of the EPMA resulted in compositions ranging from Cu6Sn5 with 4% of Ni to Sn44Cu27Ni29. This can be explained by presuming that the measurements are not taken in a single phase area, but also at phase boundaries between Sn44Cu27Ni29 and Cu6Sn5. This is also supported by the observation that the grains present are very small. Then there would be three phases present. Another possibility is that the Cu-Ni redistribution in the Cu6Sn5- phase is very slow (remember that in this phase Sn is the fastest diffusing element). In that case non-equilibrium compositions could remain in the alloy, and no conclusion about the presence of a ternary compound could be drawn. A second explanation for the existence of a phase with composition Sn44Cu27Ni29 could be an extended solubility of Cu in the Ni3Sn2-phase. This assumption is contradicted by the alloys with a composition Sn50Cu15Ni35 and Sn45Cu20Ni35, where an extended solubility of Cu in Ni3Sn2 could be found next to the ternary phase. A Backscatter Electron Image (BEI) of the sample with composition Sn45Cu20Ni35 is shown in Fig. 5.3. Following this line of thought only one explanation remains for the existence of the Sn44Cu27Ni29-phase: we are dealing with a ternary phase, different in structure from the known binary Ni-Sn and Cu-Sn phases. In order to confirm the existence of the ternary phase, two more alloys were molten: Sn45Cu29Ni26 and Sn45Cu43Ni12.
65
Chapter 5
Ni3Sn2 Ni3Sn4
Sn44Cu27Ni29
Fig. 5.3: BEI of the Sn45Cu20Ni35 alloy after annealing for 1700 h at 235 oC.
Cu6Sn5
Sn44Cu27Ni29
Fig. 5.4: BEI of the microstructure of the Sn45Cu43Ni12 alloy after annealing for 1100 h at 235 oC showing two-phases, by contrast difference. The darker areas are Sn44Cu27Ni29. Analyses of the latter alloy showed a two-phase matrix, which can only be explained by the presence of the ternary phase next to the Cu6Sn5(Ni) , see Fig. 5.4. The BEI gives contrast shades in the ternary phase, which can be caused by a different crystallographic orientation but can also be due to a broad homogeneity range of the ternary phase. 66
Phase Relations in the Sn-Ni-Cu system at 235 oC
The alloy with a composition of Sn45Cu29Ni26 was analyzed with the help of WDS with a linescan resulting in Fig. 5.5. Here it can be seen that over a length of 100 µm the Sn content stayed the same, while the Ni content varied between 21.21 at.% and 33.91, and the Cu content varied between 23.30 and 31.84 at.%. This supports the observation made in Fig. 5.4. These measured values could be the limits of the ternary phase, however, it is advisable to be cautious with these limits since the driving force for homogenization might be very low.
Fig. 5.5: Results of a linescan on the Sn45Cu29Ni26 alloy after annealing for 840 h at 235 oC, showing the homogeneity of the sample. Ni is shown by nand Cu by g. Also in the samples with 40 at.% Sn and up to 25 at.% Ni the Sn44Cu27Ni29phase was present together with η-Cu6Sn5. The microstructure of these samples was similar to that of Sn65Cu25Ni10, the only difference being that the third phase present was identified as ε-Cu3Sn with a solubility of approximately 3 at.% Ni.
67
Chapter 5
Another supporting factor for the existence of a ternary phase with the specified composition can be found in literature. Oh [6] already reported ternary phases ranging from η-Cu6Sn5 to ~33 at.% Ni with a narrow stochiometric range with respect to Sn at 200 oC. However, no convincing proof for this was shown. Also in the paper of Choi et al. [44] a phase is mentioned, which we believe is the same as the phase discussed above. However, an accurate description of this phase is also missing in this paper. The argument that the annealing time was too short to obtain equilibrium seems unlikely, considering the consistency of analyses results of the different alloys. Another attempt has been made in order to find out whether this phase is actually a ternary phase or an extended solubility of Ni in η-Cu6Sn5 by performing XRD of the several alloys. Although the X-ray peaks showed splitting, which is used by Markovski [45] to prove the existence of a ternary phase, it remains uncertain whether this is due to a ternary phase or a mixture of Cu6Sn5 and Ni3Sn2 with extended solubilities, since diffraction patterns of Cu6Sn5 and Ni3Sn2 are similar because both have the same crystal structure. XRD measurements of different samples gave inconclusive results, since only small deviations from known diffraction spectra could be found and only at a large value of 2 θ. The extension of the solubility of Cu in Ni3Sn2 was very difficult to establish precisely. In the two-phase sample with composition Sn30Cu45Ni25 the extension of the solubility of Cu in Ni3Sn2 was found to be around 17 at.%, which was confirmed in the three-phase alloy with a composition of Sn50Cu15Ni35. However, in the sample with a composition of Sn45Cu20Ni35 (see Fig. 5.3) this solubility only extended to approximately 11 at.%, which agreed with the sample Sn35Cu20Ni45, which is displayed in Fig. 5.6. An explanation for this difference might be that that the diffusion at this temperature is very sluggish, the more so when the composition is close to equilibrium where the driving force will be minimal.
68
Phase Relations in the Sn-Ni-Cu system at 235 oC
Ni3Sn2
Ni3Sn
Fig. 5.6: BEI of the Sn35Cu20Ni45 alloy after annealing for 3300 h at 235 oC. Similar reasons as mentioned above hold for the solubility range of Ni in ηCu6Sn5, which is particularly difficult to establish due to the same backscatter coefficient of Ni and Cu as explained in the previous part about the ternary phase. However, we believe this solubility to be around 4 at.%. The composition of the phases in the alloys Sn35Cu20Ni45 and Sn30Cu10Ni60 has been determined with EPMA and both show a large solubility of Cu3Sn in Ni3Sn. An argument in favor of this observation is that both Ni3Sn as well as Cu3Sn have the −
same crystal structure at higher temperature (cF16, Fm 3 m). Ni3Sn and Cu3Sn have a different structure at 235 oC and, therefore, full solubility of these phases is impossible. However, it is reasonable to assume that at lower temperatures the structures will still be quite similar, making it possible to dissolve a large amount of Cu in Ni3Sn. This has also been reported previously [22]. The resulting three-phase areas are represented with dashed lines because the limits of these areas have not been established with certainty; XRD analysis of the sample with composition Sn30Cu45Ni25 showed that the solubility of Cu3Sn in Ni3Sn extends to at least (Cu0.85Ni0.15)3Sn. It is very difficult to determine the phase relations for the Sn-poor area, i.e. the area below the range between of Cu3Sn and Ni3Sn, because here the homologous temperature is very low and, therefore, kinetics is sluggish. Nevertheless, we annealed
69
Chapter 5
four alloys in this area (Sn content < 30 at.%). Microscopic observation showed a two-phase microstructure as shown in Fig. 5.7. The composition of the phases were (Ni0.56Cu0.44)3Sn and Ni0.33Cu0.67(Sn), respectively, which agreed with the mass balance. The microstructure of the as-cast alloys (prior to annealing), however, is very similar, which means that the temperature for diffusion in these tin-poor alloys is possibly too low to achieve equilibrium in the specified annealing time. The same was true for the other three alloys. Therefore, it cannot be stated that these samples represent the actual phase equilibria present at 235 oC.
Fig. 5.7: BEI of the Sn20Ni40Cu40 alloy after annealing for 1700 h at 235 oC, showing a two-phase microstructure. The light phase is Ni3Sn(Cu) and the darker Cu,Ni(Sn).
5.4 The Diffusion Couple Technique 5.4.1 Solid-Solid Diffusion Couples In order to confirm the phase relations established with the equilibrated alloys diffusion couples were prepared. Three different ternary diffusion couples with solid state end-members were prepared and annealed for 400 h at 235 oC in vacuum. These diffusion couples consisted of: 70
Phase Relations in the Sn-Ni-Cu system at 235 oC
1. Ni/(Cu6Sn5+Cu3Sn) 2. Cu/(Ni3Sn4+Ni3Sn2) 3. (Cu6Sn5+Cu3Sn)/(Ni3Sn4+Ni3Sn2) Unfortunately, in none of the diffusion couples measurable reaction took place. Also after a closer look at the contact surfaces of the end-members separately after interaction, none of the end-members showed any signs of the presence of an intermetallic layer. This can simply be explained by bad contact between the endmembers, but it can also be a matter of difficult nucleation combined with very slow diffusion at such a low temperature. This statement is supported by earlier investigations such as the one by Oh [6] who reported an intermetallic layer of just 5 µm in diffusion couples of Cu/Ni3Sn4 and Cu6Sn5/Ni3Sn4 after annealing for close to 3000 h at 200 oC. The fact that in the binary systems the Sn-poor constituents also have difficulty in nucleating points in the same direction. Besides this, the intermetallics that can be formed are very brittle, making it possible that if a thin layer is formed it could break easily while handling the sample. 5.4.2 Solid-liquid Diffusion Couples A second set of diffusion couples was prepared with liquid Sn as one endmember and a CuNi alloy as the other. This means that the conventional experimental method for diffusion couples, where the end-members are clamped together, cannot be used. Instead, the solid end-member was polished as usual and on top of this alloy Sn grains were put. This ensemble was put in a porcelain crucible, which was encapsulated in an evacuated pyrex capsule. Since it is impossible to apply pressure in this way to the liquid phase it was unclear whether a reaction layer would form. In other words, the wettability of the alloys might not be sufficient, therefore, a flux was used in some cases. The composition of the end-members these diffusion couples is given below: 4. Cu/Sn 5. Ni/Sn 6. Cu25Ni75/Sn
71
Chapter 5
7. Cu50Ni50/Sn 8. Cu75Ni25/Sn with and without ZnCl2 -flux 9. Cu85Ni15/Sn with ZnCl2 -flux 10. Cu92Ni8/Sn with ZnCl2 -flux The first two couples were used as reference material. The numbers 4 to 8 were encapsulated and annealed without flux for a period of 400 hours. Numbers 8 to 10 were encapsulated and annealed with a ZnCl2 flux for just 24 hours. The resulting sample structure after annealing was astonishing. It appeared that the samples had reacted very fast. Some had even reacted to full extent, leaving nothing of the original alloy. Microstructures of some of the couples are shown in Fig. 5.8 –5.11.
Sn Cu6Sn5
Cu
Cu3Sn
Fig. 5.8: BEI of the diffusion couple Cu versus liquid Sn after annealing at 235 oC. Fig. 5.8 shows a relatively thin and straight Cu3Sn reaction layer with on top a scallop-like Cu6Sn5 layer as reported in literature (see above). In the original liquid Sn Cu6Sn5 crystals could be seen. Some had a hexagonal structure others had needle-like structures. In chapter 7 we will discuss further details about the morphology of the microstructure of these samples.
72
Phase Relations in the Sn-Ni-Cu system at 235 oC
Fig. 5.9 shows again an irregular interface, with Ni3Sn4 present as crystals throughout the liquid Sn matrix. The other phases (Ni3Sn and Ni3Sn2) could not be observed.
Sn Ni3Sn4
Ni
Fig. 5.9: BEI of the diffusion couple with Ni versus liquid Sn after annealing for 400 h at 235 oC without flux. Like in all diffusion couples with a (Cu,Ni)- alloy as end-member, as shown in Fig. 5.10 and 5.11, the solid substrate has reacted with liquid Sn to form the previously mentioned ternary phase. This can be seen in case of the Cu85Ni15 –alloy in Fig. 5.11, where also η-Cu6Sn5 is formed. In case of the couples where flux was used the microstructures were similar to the one shown in Fig. 5.11. The reaction zone consisted of fine grains of η-Cu6Sn5 and the ternary phase. The macrostructure of the diffusion couples after annealing was remarkable, since the intermetallic had risen and was in some cases even partly detached from the original interface. From comparison of the reaction zone between the Cu75Ni25/Sn couples with and without flux it was evident that the reaction without flux was impeded. This is probably due to an oxide layer on the substrate, since the couples were annealed under atmospheric conditions. Since the intermetallics formed were spread throughout the liquid Sn it is practically impossible to plot a diffusion path (as mentioned in chapter 2). Also
73
Chapter 5
statements about the mass balance are difficult to make because of the spreading of the reaction products. A fact that has to be remarked about kinetics is that the diffusion couples with an end-member of a (Cu,Ni)- alloy reacted faster than with pure Cu or Ni. The fastest reacting couple was the couple with Cu92Ni8 –alloy as end-member. This couple had completely reacted to a two-phase matrix of the ternary phase and η-Cu6Sn5.
CuNi
Sn44Cu27Ni29
Ni3Sn4
Sn
Fig. 5.10: BEI of the diffusion couple of Cu50Ni50 versus liquid Sn after annealing at 235 oC.
Cu6Sn5 + Sn44Cu27Ni29 Cu85Ni15
Fig. 5.11: BEI of a diffusion couple with Cu85Ni15 versus liquid Sn after annealing for 30 h at 235 oC with a ZnCl2 flux; the light phase is Sn.
74
Phase Relations in the Sn-Ni-Cu system at 235 oC
Another interesting phenomena is the absence of a (Cu,Ni)3Sn- phase in the diffusion couples (Cu,Ni)/liquid Sn. Since according to the constructed phase diagram this phase should be there it is thought either to be absent for kinetic reasons (p. 30) or it is too thin to be observed. In all cases the slow-growing Ni3Sn(Cu) phase is expected rather than the Cu3Sn(Ni) phase which grows faster (see Fig. 5.12). Apparent absence of this slow-growing phase can, indeed, be expected. In pure Cu/ liquid Sn the Cu3Sn-phase is present (see Fig. 5.8).
5.5 The Isothermal Cross-section of the Sn-Ni-Cu System at 235 oC Sn (liq)
Ni3Sn4 η-Cu6Sn5 Ni3Sn2
Ni3Sn (l)
Ni
ε-Cu3Sn
Cu
Fig. 5.12: The most likely isothermal cross-section of the Sn-Ni-Cu system based on the results from the equilibrated alloy measurements. Tielines which are measured with insufficient certainty are indicated by dashed lines; dotted lines indicate tielines which are estimated but not measured; dotted lines indicate tielines which are estimated but not measured.
75
Chapter 5
The iso-thermal cross-section presented in Fig. 5.12 is a cross-section, which describes in the best way the results obtained in this study. However, it has to be remarked that the limits of the homogeneity range of the ternary compound cannot be established with certainty. So the homogeneity range might be broader or smaller than depicted. The same holds true for the solubility range of the Ni3Sn2 compound, as described in the previous sections. In the Sn-poor part of the cross-section only (dashed) tielines are drawn of phase compositions actually measured by EPMA. Here it remains uncertain whether equilibrium was reached, therefore, no further conclusions are drawn for this area. Dashed lines represent uncertainties in the diagram; dotted lines represent estimated tielines. Studies performed by other investigators in this Sn-poor region report other ternary phases, but the experimental data originates from higher temperature ranges and these are extrapolated to lower temperature regions. The reported phase compositions and crystal structures are CuNi5Sn2 (oP8, Pmmn) and Cu2NiSn (cF4, Fm3 m ) [10, 46, 47]. These phases can also be denoted as (Cu0.17Ni0.83)3Sn and (Cu67Ni0.33)3Sn, respectively. In our Sn15Cu20Ni65 sample the (Cu0.17Ni0.83)3Sn composition is present, but no further proof of the ternary phase could be found. However, the existence of ternary phases cannot be completely ruled out because equilibrium is very difficult to attain in the (Cu,Ni)3Sn area at this low temperature.
5.6 Concluding Remarks The isothermal cross-section of the Sn-Cu-Ni system at 235
o
C was
constructed with the help of equilibrated alloys and the diffusion couple technique using liquid Sn end-members. In this system a ternary phase was found with an average composition of Sn44Cu27Ni29. Proof for the existence of this phase was presented by microstructural, XRD and EPMA analysis. The homogeneity range of the ternary phase could not be established but is thought to be quite wide. The solubility range of Cu in Ni3Sn2 could not be established precisely but is believed to be somewhere between 11 and 17 at.%. Approximately 7 at.% of Cu could dissolve in Ni3Sn4, about 4 at.% of Ni in Cu6Sn5 and about 3 at.% for Ni in Cu3Sn.
76
Phase Relations in the Sn-Ni-Cu system at 235 oC
The solubility of Cu in Ni3Sn probably extends to at least (Ni0.15Cu0.85)3Sn , which can be explained by the fact that Cu3Sn and Ni3Sn have similar crystal structures at 235 oC and even the same structure at higher temperatures. Although kinetics in the low Sn region are sluggish, an attempt has been made to establish the phase relations in this area, the results obtained have to be interpreted with caution, because it is not clear whether equilibrium was established due to the sluggishness of diffusion at such low temperature.
References 1
‘Lead-free Solder Project: Final Report’; NCMS, Ann Arbor Michigan USA, (1997)
2
B.P. Richards, K. Nimmo et al, ‘Lead-free Soldering’, Department of Trade and Industry, UK, (1999 and update 2000) (www.lead-free.org/)
3
W. Peng, K. Zeng and J.K. Kivilahti, ‘A Literature Review on Potential Leadfree Solder Systems’, Internal Report, Helsinki University of Technology, Espoo Finland(2000)
4
R. Brandt, J. Knarren, G. Manders et al., ‘Lead-free Solders for Electronic Applications’, Internal Report, MDP Project Group “Lead-free Soldering”, Eindhoven University of Technology Eindhoven the Netherlands (1999)
5
T.B. Massalski, J.I. Murray, L.H. Bennett, H.Baker (eds.), ‘Binary phase Diagrams’, ASM Metals Park OH USA (1986).
6
M.Oh, Ph.D. thesis Lehigh University, USA (1994)
7
K.P. Gupta, S.B. Rajendraprasad, D. Ramakrishna And A.K. Jena, J. Alloy. Ph. Diagrams 4 3 (1988) 160
8
B.D. Bastow, D.H. Kirkwood, J. Inst. Met. 99 (1971) 277
9
E. Wachtel, E. Bayer, Z. Metallk. 75 (1984) 61
10
K.P. Gupta, J. Phase Equil. 21 5 (2000) 479
11
T.M. Korhonen, P.Su, S.J. Hong, M.A. Korhonen and C.-Y. Li, J. Electron. Mater. 29 10 (2000) 1194
12
H. Pal, S. K. Prahdan and M. De, Jpn. J. Appl. Phys. 34 3 (1995) 1619
13
S. an Mey, Z. Metallkde. 78 (1987) 502
77
Chapter 5
14
D.J. Chakrabarti, D.E. Laughlin, S.W. Chen and Y.A. Chang, Binary alloy Phase diagrams (1991) 1442
15
E. Starke, H. Wever, Z. Metallkde. 55 (1964) 107
16
C. Bhedwar, K.K. Kay, S.D. Kulkarni and V. Balasubramanian, Scripta Met. 6 (1972) 919
17
D.A. Unsworth, C.A. Mackay, Trans. Inst. Met. Fin. 51 (1973) 85
18
M. Onishi, H. Fujibuchi, Trans. JIM. 16 (1975) 539
19
P.J. Kay, C.A. Mackay, Trans. Inst. Met. Fin. 54 (1976) 68
20
K.N. Tu, R.D. Thompson, Acta Metall. 30 (1982) 947
21
Q. Yiyu, F. Hongyuan, C. Dinghua, F. Fuhua and H. Lixia, Brazing & Soldering 13 (1987) 39
22
Z. Mei, A.J. Sunwoo and J.W. Morris Jr., Met. Trans. A 23A (1992) 857
23
Y. Wu, J.A. Sees, C. Pouraghabagher, L.A. Foster, J.L. Marshall, E.G. Jacobs and R.F. Pinizotti, J. Electron. Mater. 22 7 (1993) 769
24
P.T. Vianco, K.L Erickson and P.L. Hopkins, J. Electron. Mater. 23 8 (1994) 721
25
Z. Lubyová, P. Fellner and K. Matiašovsky, Z. Metallkde. 66 (1975) 179
26
E.K. Ohriner, Weld. J. Res. Suppl. 7 (1987) 191
27
H.H. Fidos, H. Schreiner, Z. Metallkde. 61 (1970) 225
28
J.O.G. Parent, D.D. Chung, I.M. Bernstein, J. Mat. Sc. 23 (1988) 2564
29
F. Bartels, J.W. Morris Jr, G. Dalke and W. Gust, J. Electron. Mater. 23 8 (1994) 787
30
S. Bader, W. Gust and H. Hieber, Acta Metall. Mater. 43 1 (1995) 329
31
H.K. Kim, K.N. Tu, Phys. Rev. B 53 23 (1996) 16027
32
C.R. Kao, Mat. Sc. Eng. A. 238 (1997) 196
33
A. Hayashi, C.R. Kao and Y.A. Chang, Scripta Mat. 37 4 (1997) 393
34
L.H. Su, Y.W. Yen and S.W. Chen, Metall. Mat. Trans. B 28 (1997) 927
35
M. Schaeffer, R.A. Fournelle and J. Liang, J. Electron. Mater. 27 11 (1998) 1167
36
S. Chada, W. Laub, R.A. Fournelle and D. Shangguan, J. Electron. Mater. 28 11 (1999) 1194
37
Y.G. Lee, J.G. Duh, J. Mat. Sc. Mat. Elec. 10 (1999) 33
38
F. d’Heurle, R. Ghez, Thin Solid Films 215 (1992) 26
78
Phase Relations in the Sn-Ni-Cu system at 235 oC
39
M.J.H. v. Dal, A.M. Gusak, C. Cserháti, A.A. Kodentsov and F.J.J. van Loo, Phys. Rev. Lett. 86 (2001) 3352
40
J.A. van Beek, S.A. Stolk and F.J.J. van Loo, Z. Metallkde. 73 (1982) 439
41
S. Kang, V. Ramachandran, Scripta Met. 14 (1980) 421
42
D. Gur, M. Bamberger, Acta Mater. 46 14 (1998) 4917
43
P Villars and L.D. Calvert, “Pearson’s Handbook of Crystallographic Data for Intermetallic Phases”, 2nd edition, ASM Metals Park OH USA (1991)
44
W.K. Choi and H.M. Lee, J. Electron. Mater. 28 11 (1999) 1251
45
S.L. Markovski, ‘Chemical Interaction between Metals and Compound Semiconductors’, Ph.D. thesis Eindhoven University of Technology, the Netherlands (1999) 100
46
N. Saunders and M.P. Miodownik, Bull. Alloy Phase Diagrams 11 3 (1990) 278
47
J.S. Lee Pak, K. Mukherjee, O.T. Inal and H.R. Pak, Mater. Sci Eng. 117A (1989) 167
79
‘Lead was eventually discovered to be too soft for the purpose of Bludger manufacture (any indentation left on a Bludger will affect its ability to fly straight).’ J.K. Rowling - Quidditch through the Ages, Bloomsbury Publishing London UK (2001) 22
80
Chapter 61 Phase Relations in the Bi-Ni-Pd System
6.1 Introduction In the previous chapters the ternary isothermal cross-section of a possible solder system and the cross-section of a diagram of the main constituent of a solder, i.e. Sn, with common substrate materials was discussed. In this chapter the interaction of a substrate material (Ni/Pd) with another possible constituent of lead-free solder (Bi) is discussed. Since wettability is important for the formation of solder joints, common substrate materials like Cu or Ni are often pre-plated in soldering. This can be done with Sn, which is prone to whisker formation (see chapter 7) and oxidation, but also with noble metals, one of which is Pd. Ni coated with Pd is a well-known substrate, 1
This chapter is based on the following papers: P.J.T.L. Oberndorff, M.G.A. van Vinken, A.A. Kodentsov and F.J.J. van Loo, “Solid State Diffusion in the Bi-Pd System”, accepted for publication in Defect and Diffusion Forum and P.J.T.L Oberndorff, M.G.A. van Vinken, A.A. Kodentsov and F.J.J. van Loo, “Phase Relations in the Bi-Ni-Pd System”, J. Phase Equil. 22 (2001)
81
Chapter 6
which has been investigated before [1]. Also studies have been done on the interaction of Sn [2] and SnPb [3-7] with Pd, but if Bi is used as element in a lead-free solder also the interaction of Bi with Pd and Ni is of importance. Some industries are already moving to Pd-Ni terminations for other reasons than lead-free compatibility. Although these elements will never be major alloying elements (because of costs of Pd and the fact that Bi causes brittleness; see section 1.3) the phase relations are still important and also interesting, especially from a fundamental point of view. In this chapter not only the isothermal cross-section will be presented, but also a study in diffusion and kinetics of the binary system Bi-Pd. This study was conducted since in this binary system the existence of intermediate phases remained uncertain. Another reason is to gain insight in the growth-rate of intermetallic layers at such low temperatures used in electronic industry for soldering. This low temperature is the main reason for the sluggishness of phase formation, making the study more complicated.
6.2 Solid State Diffusion in the Binary Bi-Pd System Earlier studies [8] in the binary Bi-Pd system show that still uncertainties remain about the stabilities of intermetallic compounds. Especially on the Pd-rich side of the phase diagram controversy exists. In previous work it is shown that in the temperature range of our interest (up to 250 oC) two intermetallic compounds (BiPd and Bi2Pd) are present with certainty [8]. Another compound, with a composition of BiPd3, is believed to be stable on the Pd-rich side. In order to determine diffusion characteristics it is of vital importance to know which intermetallic layers will be stable at the temperature range investigated. Since the Pd-rich side contains uncertainties an incremental couple was prepared, having Pd and BiPd as end-members. The couple was annealed for 400 hours at 250 oC. After investigation of this couple two intermetallic compounds were found, a layer at the Pd interface with a thickness of approximately 1-2 µm and one irregular layer, which grew up to 30 µm. The thin layer was identified by EPMA as being BiPd3. The other layer had an average composition of 62.4 at.% of Pd. This composition was not reported to be stable in earlier investigations [8]. However, at higher temperatures a
82
Phase Relations in the Bi-Ni-Pd System
phase was reported to be present with a composition of Bi3Pd5 (γ-phase), which is close to the composition analysis showed. Therefore, it is not unlikely that the intermetallic found by us is the γ-phase. It could be argued from the incremental couple that the γ-phase is a metastable phase. For that reason an alloy with composition of 62.5 at.% Pd was melted and annealed at 235 oC for 1000 hours. The microstructure is shown in Fig 6.1. EPMA analysis showed that this alloy consists of the γ-phase and plate-like crystals analyzed as BiPd3, proving the stability of the γ-phase.
γ BiPd3
Fig. 6.1: Backscattered Electron Image (BEI) of an equilibrated alloy with composition 62.5 at.% Pd and 37.5 at.% Bi. The previously described analysis made it clear that four intermetallics are to Pd be expected in the binary Bi-Pd system at the temperature range of 200-250 oC.
Diffusion couples were annealed at five different temperatures (200, 215, 230, 245 and 250 oC) and for times varying from 94 to 1000 hours. A typical microstructure of one of the diffusion couples, with corresponding concentration profile, is shown in Fig. 6.2.
83
Chapter 6
Heen
a)
γ
b) Fig. 6.2: a) BEI of a Bi/Pd couple annealed for 196 h at 215 oC. b) Concentration profile across the diffusion zone of the diffusion couple in Fig. 6.2a. After careful observation with EPMA it could be resolved that in a number of samples (especially after long annealing times) a thin layer of BiPd3 was present. However, this layer was too thin for extensive analysis, which is probably due to kinetic reasons as explained in chapter 2.
84
Phase Relations in the Bi-Ni-Pd System
The layers of the other intermetallics were found to be irregular in width. There can be several explanations for this phenomenon. It could be possible that specimen pre-treatment procedures result in thickness variation or perhaps the differences are merely manifestations of the crystallographic orientation of the grains of the reaction layer as claimed by Unsworth [9]. Another explanation could be that so-called Kirkendall pores are formed in the BiPd3-phase. Very likely Pd will diffuse more easily in BiPd3 than Bi (Cu3Au-rule [10]) which will lead to a Kirkendall plane near the Pd/BiPd3 interface, see section 2.2.3. This means that at this side impurities gather possibly causing a diffusion barrier, which might impede reaction layer formation. Although the thickness of the individual intermetallic compounds showed variation, the total layer thickness shows a parabolic relation with the annealing time indicating a diffusion-controlled mechanism. Deviations of some of these measurements can be explained by contact imperfections on the Pd side of the diffusion couple, which gives rise to the formation of one phase at the expense of another. The γ-phase appeared very irregular, sometimes not even being present as a continuous layer, which can be explained by the formation of pores near the Pd interface as well. Since the pores inhibit Pd transport, the BiPd layer will consume part of the γ-phase. At 200 oC the BiPd layer remained very thin (Fig. 6.3), as compared to the higher temperatures. This is probably due to the fact that BiPd undergoes a structure transformation from oC32 to mP16 above 210
o
C [8]. A somewhat similar
phenomenon has also been observed in the Pt2Si-phase of the Pt-Si system [11]. Despite this irregularity in thickness of the different intermetallic layers, it is possible to calculate the integrated diffusion coefficients for the γ-phase, BiPd and Bi2Pd. This has been done by determining an average layer thickness for each intermetallic layer in well-developed areas. Since the BiPd3 layer-thickness remains around 1 µm (independent of annealing time and temperature) one can only give a rough estimate for the integrated diffusion coefficient of the order of 8*10-18 m2/s at 215oC for this layer.
85
Chapter 6
BiPd Pd
γ
Bi2Pd
Bi
Fig. 6.3: BEI of a Bi/Pd diffusion couple annealed at 200 oC for 596 hours, showing a narrow BiPd layer. T [oC]
γ Dint [m2/s]
DintBiPd [m2/s]
DintBi2 Pd [m2/s]
200
5.8.10-17
9.7.10-18
1.3.10-16
215
7.6.10-17
6.4.10-16
2.3.10-16
230
1.3.10-16
2.4.10-15
4.8.10-16
245
-
3.1.10-15
7.9.10-16
250
2.9.10-16
3.6.10-15
6.2.10-16
Table 6.1: Experimentally determined integrated diffusion coefficients for solid state diffusion in the Bi-Pd system, in the temperature range from 200-250 oC. The value for DintBiPd 3 is of the order of 8*10-18 m2/s at 215 oC. With these data it is possible to make an Arrhenius plot for the temperature dependence of intermetallic growth (Fig. 6.4). An approximately linear relationship is present, indicating a reasonable consistency of the measurements. The datapoint for BiPd at 200 oC has been excluded on grounds of the phase transformation.
86
Phase Relations in the Bi-Ni-Pd System
γ
Fig. 6.4: Temperature dependence of the integrated diffusion coefficients for the γ-, BiPd- and Bi2Pd - phases in the temperature range 200 - 250 oC. From Fig 6.4. the apparent pre-exponential factor (D0) and the “activation energy” (Q) can be calculated for the integrated diffusion coefficient. The results are shown in Table 6.2. The low values indicate an important role of short-circuit diffusion. Especially when looking at the γ-phase this seems to be the case. As stated before this phase is not formed in all the diffusion couples, but when formed it can be observed to have thin columnar grains. This would mean that the ‘activation energy’ and the pre-exponential factor listed for this phase should be regarded with caution, since if grain boundary diffusion and volume diffusion change their contribution with time, other rules would apply. It should also be stressed that the apparent values for D0 and Q include possible variations in the homogeneity range of the phases as a function of temperature. In fact, these values have to be regarded as experimental parameters, which describe the observed phenomena. From the absence of any markers in the phases γ, BiPd and Bi2Pd one can conclude that the
DPd ratio in these phases is in the range: γ: > 150 or < 15; BiPd: > DBi
15 or < 0.3; Bi2Pd : > 0.3.
87
Chapter 6
γ
BiPd
Bi2Pd
D0 [m2/s]
1.38.10-9
5.44.10-8
1.25.10-8
Q [kJ/mol]
67.3
73.0
73.2
Table 6.2: The apparent pre-exponential factor and activation energy for different phases in the Bi-Pd system.
6.3 The Binary Ni-Bi phase diagram Apart from the uncertainties in the Bi-Pd binary diagram also some controversy in the Ni-Bi binary diagram existed about the presence of the intermetallic NiBi. In the Ni-Bi phase diagram at 235 oC two intermetallics exist: NiBi and NiBi3. However, earlier work of our group [12] showed that in semi-infinite diffusion couples at 250 oC only NiBi3 was formed after short annealing times (< 225 h, see Table 6.3). The NiBi-phase was missing, probably due to kinetic reasons [13,14]. To check whether NiBi and NiBi3 both form at 235 oC, a ‘sandwich’ couple of Ni/Bi/Ni with a relatively thin Bi foil of 200 µm was made and annealed for 400 h at this temperature in vacuum, with the same intention as the Ti/Al example in chapter 2.
Thickness (µm) NiBi3
kp (10-14 m2/s)
Dint (10-15 m2/s)
16
56
2.8
5.2
36
76
2.3
4.3
100
116
1.9
3.6
145
166
2.7
5.0
225
214
2.9
5.3
Time (h)
Table 6.3: Layer thickness and integrated diffusion coefficient for the NiBi3 phase at 250 oC at different annealing times.
88
Phase Relations in the Bi-Ni-Pd System
During the reaction Bi is consumed and then NiBi can start to grow between Ni and NiBi3. The microstructure showing that both NiBi3 as well as NiBi are formed is presented in Fig. 6.5. This experiment proves that NiBi is a real equilibrium phase in the system at 235 oC. Now we can calculate the thickness that the NiBi-layer should reach in a Ni/Bi couple annealed at that temperature during 225 h. The presence of the gap in Fig. 6.5, and the absence of any indication for a Kirkendall plane in the NiBi- phase, suggests that in both NiBi3 and NiBi, Bi is the only diffusing component.
Fig. 6.5: The BEI of the microstructure of the Ni/Bi/Ni diffusion couple annealed for 400 hours at 235 °C. From the layer thickness of ~ 200 µm after 225 h one can calculate how long it takes to consume the Bi-foil in the Ni/Bi/Ni diffusion couple. That is the case when the NiBi3-layer has reached a thickness of approximately 100 µm (at each side, see Fig. 6.5). that means that it takes roughly
(100) 2 * 225h = 50h to get a diffusion couple of (200) 2
the type Ni/NiBi3. Then the NiBi-layer starts growing, and does that during (40050)=350 hours, forming a 17 µm thick layer (Fig. 6.5). The integrated diffusion coefficient for NiBi can then be found as
89
Chapter 6
DintNiBi =
0.50 * 0.25 (17 *10 −6 ) 2 * = 1.9 * 10 −17 m 2 / s 0.75 2 * 350 * 3600 That means that, growing during 225 h between pure Ni and Bi, next to a 200
µm NiBi3 layer, its thickness can be calculated through: DintNiBi =
10 −12 0.5 * 0.25 * (∆x) 2 + ∆x * 0.5 * 0.25 * 200 = 1.9 * 10 −17 m 2 / s 2 * 225 * 3600 0.75
which results in ∆xNiBi≈ 1 µm. It is, therefore, very well possible that the NiBi-layer is present in this couple Ni/Bi after 225 h at 235oC but that it is too thin to be seen.
6.4 The Isothermal Cross-section of the Bi-Ni-Pd System at 235 oC 6.4.1 The Diffusion Couple Technique In order to establish the Bi-Ni-Pd isothermal cross-section at 235 °C a diffusion couple with BiPd and Ni as end-members and a diffusion couple of Ni54Pd46 and Bi were annealed at 235 °C for 400 hours. Besides, seven ternary alloys of different compositions were melted and annealed at 235 °C for over 1000 hours. The microstructure of the reaction zone of the diffusion couple with Ni54Pd46 and Bi as end-members is given in Fig. 6.6. In the reaction zone of the diffusion couple three different layers can be distinguished, a 3 µm one-phase layer at the Bi-side and a layer with a thickness of 330 µm, consisting of two phases. At the NiPd-side a very thin layer (< 0.5 µm) was found. The diffusion couple was analysed with EPMA and it was found that the layer at the Bi-side has an average composition of Bi66Pd30Ni4, the Bi2Pd phase with a solubility of 4% nickel. The formation of this layer next to the Bi end-member makes it clear that equilibrium exists between Bi2Pd(Ni) and Bi.
90
Phase Relations in the Bi-Ni-Pd System
Fig.6.6: The BEI microstructure of the Bi/Ni54Pd46 diffusion couple annealed at 235 °C for 400 hours. In the thick layer, consisting of two phases, one having a composition of Bi66Pd30Ni4 and the other a composition of Bi76Ni20.5Pd3.5, presumably the NiBi3 phase with a solubility of 3.5 at.% Pd. The thin layer at the NiPd-side was analyzed as NiBi. In Fig. 6.7 the microstructure of the NiPd-side of the diffusion couple is shown. In this figure it can be seen that the NiBi phase not only grows as a thin layer, but also seems to grow in a finger-like structure. Interfaces can be found at which both Bi2Pd and NiBi3 are in equilibrium with the NiBi ‘fingers’. This indicates the existence of a three-phase equilibrium between those phases. At this side of the sample a crack occurs, which runs along the whole sample. Close examination of this end of the diffusion couple showed another layer too thin to analyse (< 0.5 µm) between the NiBi and the Ni54Pd46. Because of these reasons it can not be stated with certainty which phases are in equilibrium with the Ni54Pd46 end-member of the diffusion couple. Since Pd can dissolve 17 at.% Bi, and the Ni-Pd system is characterised by a (Ni, Pd) solid solution, a (Ni, Pd) solid solution will exist between Pd with a maximum solubility of 17 at.% Bi and Ni. Because in the Bi/Ni54Pd46 diffusion couple no solubility of Bi in the Ni54Pd46 was found, it is expected that the (Ni, Pd) solid solution has almost no solubility for Bi at this composition.
91
Chapter 6
In a diffusion couple with BiPd and Ni as end-members no reaction has taken place. These phases are indeed in equilibrium, as will be shown in the next paragraph.
Fig. 6.7. The BEI microstructure of the NiPd-side of the Bi/NiPd diffusion couple annealed at 235 °C for 400 h.
6.4.2 Equilibrated Alloys Ternary Bi-Ni-Pd alloys were melted and annealed for over 1000 hours at 235 °C. They were analysed with WDS and the results are given in Table 6.4. The results of EDS-analysis made it clear that the solubility of Ni (< 2 at.% ) and Pd (0 at.%) in Bi at 235 °C were very small. The first two samples indicate a three-phase equilibrium Bi +NiBi3 + Bi2Pd in the Bi-Ni-Pd system at 235°C. After annealing of the Bi70Ni23Pd7 alloy, three phases could be distinguished (NiBi3, Bi2Pd and NiBi), indicating the existence of this three phase on the ternary isotherm at this temperature. This is in accordance with the results of the Bi/NiPd diffusion couple. Between those two three-phase equilibria, a two-phase equilibrium between Ni3Bi and Bi2Pd has to exist.
92
Phase Relations in the Bi-Ni-Pd System
Alloy
Annealing time (h)
Phases present at 235 °C
Bi85Ni7.5Pd7.5
1000
NiBi3(Pd), Bi2Pd(Ni), Bi
Bi70Ni7Pd23
1000
NiBi3(Pd), Bi2Pd(Ni), Bi
Bi70Ni23Pd7
1000
NiBi3(Pd), Bi2Pd(Ni), NiBi(Pd)
Bi25Ni60Pd15
1850
Ni(Pd), Bi2Pd(Ni), BiPd(Ni)
Bi40Ni35Pd25
1850
Ni(Pd), Bi2Pd(Ni), BiPd(Ni)
Bi40Ni15Pd45
1850
Ni(Pd), BiPd, Bi3Pd5(Ni)
Bi55Ni23Pd22
1850
Ni(Pd), Bi2Pd(Ni), NiBi(Pd)
Table 6.3: Phases present in equilibrated Bi-Ni-Pd alloys after annealing at 235 °C for over 1000 h. The alloys with a composition of Bi25Ni60Pd15 and Bi40Ni35Pd25 exhibited a microstructure with three phases as can be seen in Fig. 6.8. The phases were identified as Ni(Pd), Bi2Pd(Ni) and BiPd(Ni). This explains why the diffusion couple of BiPd against Ni did not result in the formation of intermetallic layers. The last alloy had a composition of Bi55Ni23Pd22 and showed that Ni(Pd) was also in equilibrium with Bi2Pd(Ni) and NiBi(Pd). These results seem to contradict the results obtained with the help of the diffusion couples, in which Ni54Pd46 seems to be in contact with NiBi. However, this is not completely true, since the diffusion couple technique cannot predict anything if too slow kinetics are involved. It is well known that the Bi-poor region of this ternary system has a much lower homologous temperature and, therefore, it can be assumed that the kinetics will be very slow at this temperature. Due to this fact it is very difficult to establish the Pd-rich corner of the ternary system. Identification of the thin layer at the Ni54Pd46- side could shed some more light on the existing equilibria.
93
Chapter 6
Fig. 6.8: BEI of the microstructure of the Bi40Pd25Ni35 alloy. Bi
NiBi3 Bi2Pd
NiBi
BiPd
Bi3Pd5 BiPd3
? Ni Fig. 6.9: Phase relations in the Bi-Ni-Pd system at 235 °C as experimentally determined in the present work. 94
Pd
Phase Relations in the Bi-Ni-Pd System
The results from the Bi/NiPd diffusion couple and equilibrated Bi-Ni-Pd alloys led to the following isothermal cross-section of the bismuth-nickel-palladium system represented in Fig. 6.9.
6.5 Concluding Remarks It was shown that a phase with an average composition of 62.4 at.% Pd and 37.6 at.% Bi (Bi3Pd5) was found to be stable at temperatures as low as 200 oC. Probably this is the phase that has been reported earlier to exist at a temperature range from around 400 oC to 683 oC. Another compound, BiPd3, causing controversy was found to be present in an equilibrated alloy and in diffusion couples from 200-250 oC, although growth kinetics are slow in this temperature interval. The integrated diffusion coefficients for the γ-, BiPd- and Bi2Pd were determined in the temperature range from 200-250 oC. A significant increase in the value of the integrated diffusion coefficient in BiPd at 215 oC can be attributed to the transformation in BiPd at 210 oC. The fact that the total layer thickness grows parabolically with time suggests a diffusion-controlled process, but probably shortcircuit mechanisms play an important role. It became clear that in the Ni-Bi system two intermetallics are stable, NiBi and NiBi3. It was shown that the fact that NiBi could not be found in earlier research is probably due to slow kinetics. Furthermore, the isothermal cross-section through the ternary phase diagram Bi-Ni-Pd at 235 oC was constructed by means of multiphase diffusion couples and equilibrated alloys. No ternary phases are formed in the system at this temperature. Pd showed a solubility of 5 at.% in NiBi3. Approximately 20 at.% of Pd could dissolve in NiBi. The maximum solubility of Ni in BiPd was determined as 20 at.%, while the solubility of Ni in Bi2Pd was estimated to be 6 at.%. Sluggishness of diffusion kinetics prevented investigation of the Pd-rich corner of the phase diagram.
95
Chapter 6
References 1
D.C. Abbott, R.M. Brook, N. McLelland and J.S. Wiley, IEEE Trans. CMTH 14 (1991) 567
2
M. Mathon, M. Gambino, E. Hayer, M. Gaune-Escard and J.P. Bros, J. Alloy. Comp. 285 (1999) 123
3
G. Ghosh, J. Electron. Mater. 27 11 (1998) 1154
4
S. Vaynman, G. Ghosh and M.E. Fine, J. Electron. Mater. 27 11 (1998) 1223.
5
G. Ghosh, Metall. Mater. Trans. A 30A (1999) 5
6
H. Tanaka, M. Tanimoto, A. Matsuda, T. Uno, M. Kurihara and S. Shiga, J. Electron. Mater. 28 11 (1999) 1216
7
G. Ghosh, J. Electron. Mater. 28 11 (1999) 1238
8
T.B. Massalski, J.I. Murray, L.H. Bennet, H. Baker (eds.), ‘Binary Alloy Phase Diagrams’, ASM Metal Park OH USA (1990)
9
D.A. Unsworth, C.A. Mackay, Trans. Institute of Metal Finishing 51 (1973) 90
10
F. d’Heurle, R. Ghez, Thin Solid Films 215 (1992) 26
11
M.R. Rijnders, “Periodic Layer Formation During Solid State Reactions”, Ph.D. Thesis Eindhoven University of Technology, Eindhoven the Netherlands, (1996) 76
12
I. Oomen, Master thesis, Eindhoven University of Technology, Eindhoven the Netherlands (1997) 44-50
13
V.I. Dybkov, O.V. Duchenko, J. Alloy. Comp. 234 (1996) 297
14
P. Feschotte, J.-M. Rosset, J. Less-Common Metals 143 (1988) 31
96
Chapter 7 Anomalous Behavior in Solid-liquid Systems During the study of the diffusion couples in the Sn-Ni-Cu system, some interesting phenomena occurred, as mentioned in chapter 5. This was clearly due to the use of a liquid end-member instead of the normal use of solid-solid diffusion couples. Since the presence of liquid is common practice in soldering these phenomena will be presented in some detail in this chapter and possible explanations for these anomalies will be given.
7.1 Introduction In literature, scallop-like formation of intermetallics has been reported repeatedly, not just in the Cu-Sn System [e.g.,1,2,3] but also in Ni-Sn [2,4,5], Bi-Ni [3], Au-Sn [6], Fe-Sn [5,7], Mg-Ni [8], Al-Ni [8], Cu-In [3] Au-In [9] and Ni-In [10] to name but a few. These irregular interfaces naturally also return in soldering [e.g. 11,12] and are a reason of concern for electronic industry because of the brittle nature of the intermetallic compounds.
97
Chapter 7
The appearance of these irregular interfaces in binary systems cannot be explained with the existing thermodynamics, and the fact that literature gives different explanations for the existence of these irregular interfaces form enough reason to take a closer look at them. Let us first look at explanations offered in literature. Assuming that the interface irregularity only has to do with the fact that a liquid is involved is not correct since it has also been reported in solid-solid systems (as also can be seen later in this chapter). However, the fact that liquid is involved makes the matter more complicated, because one should keep in mind that the rate of intermetallic formation in solid-liquid reactions is several orders of magnitude higher than in solid-solid reactions. The reason for this phenomenon is unclear, although many researchers report that this has to do with the dissolution of the solid substrate in the liquid [3,4,11], maybe in combination with grain boundary controlled diffusion [13]. The general idea is that the solid substrate and its intermetallics dissolve in the liquid along grain boundaries, most likely achieving super-saturation followed by precipitation. Kim et al. [11] postulate a three step mechanism to explain the scallop-like growth in the Cu-Sn system. This mechanism can be interpreted as follows: the etching rate (of the substrate by the liquid Sn) exceeds the rate of dissipation (via diffusion, if convection is prevented) of the dissolved substrate material. Thus (near the solid/liquid interface) the solubility limit will be exceeded locally. The solid grains of the intermetallic phase act as an additional diffusion barrier causing any further atoms of the substrate material to diffuse preferentially through the remaining, still liquid, passages. These soon will become supersaturated and a continuous, but irregularly thick layer of solid intermetallic (e.g. Cu6Sn5) will separate the substrate from the liquid. In the remaining annealing time at the solid/solid interface a “normal” diffusion couple with a wavy starting plane will develop and at the solid/liquid interface Ostwald ripening yields the observed “scallops”. Upon cooling the whiskers grow very rapidly due to supersaturation. The last step, as remarked, will only have a minor influence on the “scallops” since it was observed that the morphology of the Cu-Sn compounds change dramatically with time and, therefore, cannot be caused by solidification.
98
Anomalous Behavior in Solid-liquid Systems
Only a few studies show these irregular surfaces or scallops in a threedimensional view [2,4,11,14]. We were hoping that it would be possible to get more information about the mechanism of growth by selective etching and looking on top of the reacted surface. In the following section we will present some threedimensional views on the reaction layer between Cu and liquid Sn.
7.2 Whisker Growth in the Sn-Cu System Since the Cu-Sn system is one of the most important systems in the electronic industry, a lot of research has been done concerning this system as already noted in chapter 5. It was discovered early on during these investigations that the microstructure of samples involving liquid Sn or SnPb was completely different from the straight interfaces obtained with solid-solid diffusion couples. Looking at the normal cross-sections (two-dimensional), instead of straight interfaces in a two component system, wavy interfaces were found in Cu, wetted by liquid Sn. Scallop-like protrusions of Cu6Sn5 are seen, Fig. 7.1, while the Cu3Sn layer is relatively straight.
Fig. 7.1: BEI of the reaction zone between Cu and liquid Sn, after annealing for 24 h at 240 oC, showing scallop-like formation of the Cu6Sn5 with whiskers on top.
99
Chapter 7
Apart from these hillocks or scallops, ‘whiskers’ can be observed on top of them. To further investigate the formation of these whiskers, we did several experiments where a polished Cu disc was wetted with liquid Sn. Several set-ups for these experiments were tried, but it would go into too much detail to discuss them here. After careful quenching (preventing turbulence in the liquid, which would disturb the original structure) the Sn was carefully etched away with a FeCl3 etchant, see Table 7.1. Microscopic observations revealed peculiar microstructures. These microstructures were analyzed with EPMA in order to determine what intermetallic was observed. A comparison between two-dimensional microstructures and threedimensional etch structures is given in Fig. 7.2 a-d.
Fig. 7.2: a) BEI of a cross-sectional view of the reaction zone of a Cu-Sn diffusion couple after 35 minutes annealing at 280 oC. b) Three-dimensional view (after etching) of the couple displayed in a), showing the Cu6Sn5- scallops and some long, facetted Cu6Sn5-crystals.
100
Anomalous Behavior in Solid-liquid Systems
Fig. 7.2: c) BEI of the reaction zone of a Cu-Sn diffusion couple after annealing for 32 h at 280 oC. d) Three-dimensional view (after etching) of the Cu6Sn5 layer of the diffusion couple shown in c).
Components
Amount
FeCl3 p.a.
2g
HCl (smoking) p.a.
5 ml
Demineralized water
30 ml
Ethanol
60 ml
Table 7.1: Composition of the used etching agent.
101
Chapter 7
Apparently the Cu6Sn5 crystals cluster to form bigger scallops after longer annealing times. From this it can be concluded that the scallops are not formed during cooling but are a product of the solid-liquid interaction. In these micrographs no whiskers, such as displayed in Fig. 7.1 can be found. This could be the result of washing the etchant of the surface of the sample; Cu6Sn5 whiskers of just a couple of µm and a much longer length could break off easily, also due to the brittle nature of the intermetallic. After very careful washing of the etched surface a microstructure as displayed in Fig. 7.3 resulted.
Fig. 7.3: Secondary Electron Image (SEI) of a top down view of the reaction zone of a Cu-Sn diffusion couple after annealing for 35 minutes at 290 oC. In Fig 7.3 several different morphologies of the Cu6Sn5 phase can be observed: -
hair-like fibers
-
long, facetted crystals, further on referred to as “whiskers”
-
shorter hexagonal scallops
As shown in Fig. 7.4 the hair-like thin fibers can also be found after longer annealing times. Apparently these fibers do not grow during the annealing but are a result of the solidification process. This can be explained by the fact that a eutectic in the Cu-Sn system between Cu6Sn5 and pure Sn exists at 99.1 at.% Sn and 227 oC. In
102
Anomalous Behavior in Solid-liquid Systems
that case, the eutectic structure is formed by fibers of Cu6Sn5 in a Sn-matrix [15]. Since these Cu6Sn5 fibers have a diameter of far less than 1 µm, they cannot be observed in a back-scattered electron image (BEI). The resolution of BEI is limited since the back-scattered electrons are electrons with high energy that come from deeper regions in the sample and, therefore, have not such a high lateral resolution as obtained by secondary electrons. It is possible to observe this eutectic structure with optical microscopy. In this case it is best to take a polishing agent that etches selectively, like OPS (SiO2 polishing suspension VEM Metallurgy, Houten the Netherlands).
Fig. 7.4: SEI of Cu6Sn5 hair-like fibers with parts of Cu6Sn5 whiskers (larger facetted structures) in a Cu-Sn diffusion couple after annealing for 21 h at 280 oC. This eutectic can, however, not explain the appearance of the larger facetted whiskers, because by applying the lever rule on this eutectic it can be concluded that the fraction of Cu6Sn5 in the Sn-matrix is much too high. A closer look at the scallop-like structures offers new challenges. In Fig. 7.5 the surface of the scallops are shown and it appears that they are not only hexagonal, which could be expected, but on top there are small extensions. Fig. 7.6 shows that these extensions can serve as nuclei for whisker formation.
103
Chapter 7
From Fig. 7.7 it seems that if more than one of these extensions serves as a point for whisker growth, they cluster together to form a larger whisker. The hexagonal nature of the whisker can still be found in the angles of 120o.
Fig. 7.5: SEI of Cu6Sn5 scallops after etching and annealing for 25 h at 270 oC.
Fig. 7.6: SEI of a Cu6Sn5 scallop with a whisker of the same phase attached. In this sample the etching process was not completed and therefore the matrix is still Sn. The annealing time was 25 h at 270 oC.
104
Anomalous Behavior in Solid-liquid Systems
Furthermore it seems that these whiskers are hollow. A very clear example of a hollow whisker is given in Fig. 7.8. Since these whiskers cannot be a result of the solidification of the eutectic there has to be another explanation for their formation. One can immediately notice the repeating microstructure in these whiskers as well as angles of 120o, signifying a hexagonal structure. Chadwick reported that in crystals growing from constitutional supercooling often hexagonal structures are observed [16]. If this is the case, then the whiskers should have a considerable growth rate since it takes only a couple of minutes (maximum) till the sample reaches room temperature. According to Flemings et al. [17], constitutional supercooling can explain growth rates of up to 102 cm/s. Assuming that constitutional supercooling is responsible for the growth mechanism of the whiskers, it still does not explain the fact that these whiskers are hollow as is shown in Fig. 7.8. Simov discusses hollow crystals of II-VI compounds [18]. In this publication several explanations are given, like the presence of impurities or a supersaturation gradient. But it is noted that systematic studies considering this subject are few. Another explanation offered are dislocations with a large Burgers vector, but this remains controversial. In this case, however, one would expect a step like surface as shown by Simov [18] and not a smooth one as shown here.
a)
105
Chapter 7
b) Fig. 7.7: a) SEI of a whisker (ground off in order to speed the etching process) on top of a scallop, attached to two extensions. b) Close-up of the base of the whisker in a).
Fig 7.8: Hollow whisker of Cu6Sn5. The top has been ground off in order to speed up the etching process. Annealing time 25 h at 270oC. Another possible cause for the formation of hollow whiskers is diffusion induced stress as reported by Rabkin et al. [19]. In his paper Rabkin argues that dislocations can be responsible for hollow cores in whiskers, especially when the
106
Anomalous Behavior in Solid-liquid Systems
solubility of the liquid metal in the solid substrate is high. In this case the core of the whisker would be filled with liquid metal with another composition than outside this core due to solubility of the intermetallics. However, in our experiments we were not able to find Sn in the Cu-matrix, which would be necessary according to Rabkin’s explanation. We believe another mechanism is responsible for the growth of these hollow, well-facetted whiskers. In hexagonal crystals the plane with the highest packing density is (001). Thus this is, due to the surface energy, the slowest growing face and crystals will grow perpendicular to this face. Crystal growth on densely packed planes, however, requires a very high nucleation energy unless defects like dislocations, kinks, edges etc. are available. During annealing the development of scallops (“recrystallization”) heals most dislocations etc. and only crystal corners and edges remain as growth sites. Therefore, the crystals grow layer by layer from the corners and edges. As the growth rate accelerates, the time for nucleating the next layer becomes shorter than the time needed to complete the growth for the previous layer all the way to the center of the plane. The Cu atoms cannot reach the center of the (001) plane (or the ones reaching the center cannot solidify fast enough anymore). This results in hollow, but well facetted whiskers like shown in Fig. 7.8. At even higher rates, layers starting to grow at the corners cannot link up with each other anymore, not even along the edges. This results in (up to 6) separate ‘needles’ sitting on top of the “scallop” as shown in Fig. 7.5 and 7.6 (where most growth tips apparently were broken off). The effect obviously is cooling rate dependent and more dramatic in systems with strong temperature dependency of the solute [20]. Bader et al. [2] related the growth of whiskers during cooling to the solubility of the metal in the liquid and to a preferred growth direction of the whiskers. For that reason it was decided to investigate the orientation of the Cu6Sn5 scallops and whiskers with orientation imaging microscopy (OIM). A similar method has been used by Kurt et al. [21, 22] to determine an orientation relation between the scallops and the whiskers in the liquid in the systems Sn-Cu and Au-Ti. The samples were prepared for Orientation Imaging Microscopy by standard metallographic procedure (see Chapter 3). OIM uses the electron beam in the EPMA to get electron backscatter diffraction and produce Kikuchi lines. In line with the results of Kurt et al. it was 107
Chapter 7
discovered that the basic growth direction of the whiskers was the (001) direction. This is also mentioned by Maeda et al. for hexagonal structures [23]. Hereby it should also be kept in mind that the η-Cu6Sn5 undergoes a phase transition at 189 oC to η’-Cu6Sn5, also with a hexagonal, NiAs type, structure [24]. It might be that because of this transformation a certain crystallographic direction shows preferential growth. Anisotropy of this sort in combination with Ostwald ripening is e.g. also reported in Si3N4 [e.g. 25]. No proof for this could be found in our experiments. To further investigate the importance of orientation of the substrate on intermetallic and whisker growth we repeated the experiment with a single crystal of Cu. The microstructure is shown in Fig. 7.9. As shown in Fig. 7.9 both scallops as well as whiskers are present when a single crystal serves as substrate. However, the top of the scallops has a different microstructure than observed in earlier experiments. This microstructure might have been caused by Ostwald ripening.
Fig. 7.9: SEI of the reaction zone (after etching) of a Cu single crystal vs. liquid Sn. The reaction time was 21 h at 280 oC. Cu6Sn5 scallops can be seen and in the right corner a facetted whisker is present.
108
Anomalous Behavior in Solid-liquid Systems
7.3 Intermetallic Layer Formation in the Sn-Au System The Sn-Au system is of practical importance not only because Sn is the main constituent of solder and Au might be used as a protective layer against oxidation of the substrate, but also because of the use of Sn20Au80 solder in electronic packaging. To further investigate intermetallic growth in solid-liquid systems and the influence of the solubility of the substrate we chose to investigate the Sn-Au system. It has to be remarked here that solubility rate and extend of solubility are coupled. The extent of solubility of Sn in Au at 235 oC is approximately 10 at.% of Sn in Cu at the same temperature approximately 2 at.%, while the extent of solubility of Sn in Ni at 235 oC is less than 1 at.% [26]. The solubility rate of Au in Sn is larger than of Cu in Sn. In experiments where a Au wire was hot-dipped in liquid Sn, it was observed that the rate of formation of intermetallics was extremely fast, in many cases the Au-wire reacted completely within one minute. When Au-wire was still present a microstructure resulted as displayed in Fig. 7.10 and 7.11.
Au AuSn4
Sn
Fig. 7.10: BEI of a Au-wire reacted with liquid Sn. Reaction time 10 s at 245 oC. A thin AuSn2 –layer is present between AuSn4 and Au, but is too thin to be seen at this magnification.
109
Chapter 7
Fig. 7.11: BEI of the detail from the intermetallic layers formed in the sample shown in Fig. 7.10. The AuSn4 intermetallic forms as needles or plates and can be observed at considerable distance from the surface after just a couple of seconds. This together with a slower reaction of Ni with liquid Sn (as shown in Chapter 5) gives a strong indication that the solubility of the substrate plays an important role in intermetallic formation in solid-liquid systems. Around the original Au-wire AuSn2 was observed and most probably a very thin layer of AuSn is also present, see Fig 7.11. To compare the rate of formation of intermetallics also a solid-solid diffusion couple was prepared. The microstructure of this couple is shown in Fig. 7.12. Just like in the Cu-Sn system also here irregular interfaces can be observed and even an isolated second phase seems to form in a layer. However, this precipitate is probably connected to the original layer through the third dimension. An explanation for these phenomena cannot be given, but after comparing Fig. 7.10 and 7.12 it can be concluded that intermetallic formation in the solid-liquid system appears at a much higher rate than in solid-solid interaction, taking into account the very different annealing times. Especially the AuSn4-formation in solid-liquid couples must have been formed according to a completely different mechanism than in solidsolid couples. Penetration of liquid Sn through grain boundaries and other short110
Anomalous Behavior in Solid-liquid Systems
circuit paths must be involved. The decrease of reaction rate from Au-Sn via Cu-Sn to Ni-Sn supports the idea that the solubility rate of the substrate in the liquid is an important parameter.
Au
AuSn
AuSn4
Sn
AuSn2
Fig. 7.12: BEI of a diffusion couple of Au vs. Sn after annealing for 64 h at 220 oC.
7.4 Concluding Remarks It is reasonable to assume that, although the experiments mentioned in this chapter are reproducible, the obtained microstructure strongly depends on the cooling and etching methods used. The hair-like fibers are believed to be a result of the solidifying eutectic present in Cu-Sn, since they are present in the same volume ratio to the Sn matrix independent of the annealing time. The scallop formation in the Cu-Sn system is probably due to dissolution of Cu into the liquid with subsequent precipitation and Ostwald ripening. We presume that constitutional supercooling is responsible for the facetted whisker growth. It is believed that impurities serve as nuclei for the formation of whiskers. Questions remain about the growth mechanism of facetted whiskers; from the experiments it cannot be proven whether they grow from the base or the top.
111
Chapter 7
Study of these whiskers by hot-stage microscopy might provide an answer to this question. More ideas on the topic of intermetallic layer formation with liquid Sn and CuNi metallizations will be published in the near future by Kivilahti et al. [27]. Further investigation of the facetted whiskers by OIM and TEM might possibly give new directions for research. In the case of OIM sample preparation can best be done by Focussed Ion Beam etching in order to limit distortion of the structures by preparation. From comparison of our experiments in the Ni-Sn, Cu-Sn and Au-Sn systems we believe that the rate of intermetallic formation is coupled with the solubility rate of the substrate in liquid Sn.
References 1
H. Fidos, H. Schreiner, Z. Metallkde. 61 (1971) 225
2
S. Bader, W. Gust and H. Hieber, Acta metall. mater. 43 1 (1995) 329
3
C.R. Kao, Mater. Sci. Eng. A 238 (1997) 196
4
D. Gur, M. Bamberger, J. Mater. Sci. 35 (2000) 4601
5
J.A. van Beek, S.A. Stolk and F.J.J. van Loo, Z. Metallkde. 73 (1982) 439
6
P.G. Kim, K.N. Tu, Mater. Chem. Phys. 53 (1998) 165
7
T. Ishida, Trans. JIM 14 (1973) 37
8
C. Tsao, S. Chen, J. Mater. Sci. 30 (1995) 5215
9
Y.M. Liu, T.H. Chuang, J. Electron. Mater. 29 4 (2000) 405
10
Y.H. Tseng, M.S. Yeh and T.H. Chuang, J. Electron. Mater. 28 2 (1999) 105
11
H.K. Kim, H.K. Liou and K.N. Tu, Appl. Phys. Lett. 66 18 (1995) 2337
12
P.G. Kim, K.N. Tu, J. Appl. Phys. 80 7 (1996) 3822
13
M. Schaefer, R.A. Fournelle, J. Liang, J. Electron. Mater. 27 11 (1998) 1167
14
K.N. Tu, R.D. Thompson, Acta Met. 30 (1982) 947
15
Personal communication with M. Notis, Lehigh University.
16
G.A. Chadwick, ‘Metallography of Phase Transformations’, Butterworths London, UK , (1972), 94-99
17
M.C. Flemings, Y. Shiohara, Mat. Sc. Eng. 65 (1984) 157
18
S. Simov, J. Mater. Sci. 11 (1976) 2319
112
Anomalous Behavior in Solid-liquid Systems
19
E. Rabkin, B. Straumal and W.Gust, Defect and Diffusion Forum 129-130 (1996) 229
20
Personal communication with J.C. Schuster
21
S. Däbritz, W. Hauffe and R. Kurt, Mikrochim. Acta. 125 (1997) 3
22
R. Kurt, S. S. Däbritz, W. Hauffe, D. Bergner, K. Richter, Defect and Diffusion Forum 143-147 (1997) 609
23
M.Maeda, F. Goto and K. Miyata, Jap. J. Appl. Phys. 3 (1964) 426
24
K.P. Gupta, J. Phase Equil. 21 5 (2000) 479
25
M. Kitayama, K. Hirao, M. Toriyama and S. Kanzaki, J. Am. Ceram. Soc. 82 10 (1999) 2931
26
T.B. Massalski, J.I. Murray, L.H. Bennet, H. Baker (eds.), ‘Binary Alloy Phase Diagrams’, ASM Metal Park OH USA (1990)
27
Personal communication with J.K. Kivilahti
113
114
Chapter 8 Epilogue Although we realize that soldering in electronic industry will deal with shorter contact times than used in this study, it is believed that studying phase relations is relevant. This is because the shelf life of products can be years and in this time equilibrium can be established. Apart from this, many of the intermetallic compounds present in the discussed phase diagrams often appear already after a very short reaction time. As is shown in chapters 4 through 6 the diffusion couple technique is a very powerful tool to investigate phase relations. This also holds for solid-liquid interaction, although here more difficulties arise as to the practical side of the preparation of diffusion couples and the interpretation of the results. One of the main problems encountered in this study is the sluggishness of diffusion at relevant temperatures in solid state reactions. Since soldering will have to be done in this temperature range, no changes for this aspect are expected. This means that studying systems relevant for soldering has to be done with patience. Numerous systems remain for experimental analysis of the phase equilibria and stability of intermetallics, not forgetting the study of cross-sections at different temperature in order to investigate temperature dependence. In this study ternary
115
Epilogue
cross-sections of phase diagrams are presented, but it is obvious that quaternary or even more component systems will play an important role. One of the systems that is certainly of importance will be the Sn-Ag-Cu-Ni. A start for evaluation of this system has been made with the study of the Sn-Ni-Cu system in chapter 5 and with numerous investigations concerning the Sn-Ag-Cu system. By adding more components theory and experiments will become more complex. But in recent years, thermodynamic databases of solders and solder substrate interaction have been set up and with the help of thermodynamic modeling it is possible to calculate the phase relations in these complex systems. Apart from the phase equilibria it will be necessary to study the reliability of the solder joints. This will most certainly be one of the main topics of study in the near future, since it looks like lead-free solder will be implemented soon. This has to be done in close cooperation with the electronic industry. It is not only important to find out how and why failures in electronics occur, but especially how to prevent these failures. With upcoming miniaturization also the interest in the formation of fibers and whiskers will increase. This, of course, is closely connected to the earlier mentioned reliability. Further experiments to help uncover the origin of these structures, and in this way discovering how to prevent their growth, are needed.
116
Summary
In the last decade, environmental awareness grew and this has led to proposals in legislation concerning the use of toxic elements. One of these elements is Pb. Although Pb in electronic solder systems only accounts for a few percent of the total amount of lead used, it might have a major impact on the environment since it is so finely dispersed. Therefore, it has been proposed to look for alternatives for lead containing solder in electronic industry. Since not much data is known about the fundamental aspects of lead-free soldering and interaction with substrates and components, research in this aspect is very important. In this investigation fundamental properties of systems relevant to the soldering industry have been studied. This has been done with the help of equilibrated alloys and the diffusion couple technique. The samples have been analyzed with light microscopy, scanning electron microscopy (SEM), electron probe microanalysis (EPMA) and X-ray diffraction (XRD). After establishing the intermetallics present in the binary diagrams ternary isothermal cross-sections have been constructed with the above mentioned techniques. This has been done for the following systems: Sn-Ag-Sb at 220 oC, Sn-Ni-Cu at 235 o
C and Bi-Ni-Pd at 235 oC. Sn-Ag-Sb is a system that can be used when requiring solder with a higher
melting point. It was discovered that, contrary to the one phase area believed to exist in the binary Sn-Sb system, two intermetallics are stable at 220 oC: Sn3Sb4 and Sn4Sb3. Since alloys are nearly impossible to anneal to equilibrium at this temperature, the diffusion couple technique is, in fact, the only tool in establishing the phase relations. In the ternary Sn-Ni-Cu system at 235 oC a new ternary phase has been discovered. This can explain the fast growth rate of intermetallics found in this system when Cu is alloyed with Ni versus the growth rate of pure copper with liquid Sn.
117
The Bi-Ni-Pd system can be of importance for electronic industry because Bi is a possible replacement of Pb in solders, while Cu-conductors are often covered with Ni, which in turn can be coated with Pd for better oxidation resistance. Except for establishing the ternary cross-section of this system at 235 oC, also a study of solid state diffusion has been performed for the binary Bi-Pd and Bi-Ni systems. This showed that the γ-phase, Bi3Pd5, is stable at 200 oC and not only at temperatures higher than 400 oC as reported earlier. For the first time evidence is given that BiPd3 is stable at lower temperatures. Furthermore, the study established the growth behavior inclusive integrated and intrinsic diffusion coefficients of several intermetallics present. It was found that the growth of intermetallics during liquid-solid interactions was many times faster than in solid-solid interaction and that the morphology was completely different. After solid-liquid interaction a peculiar growth of ‘whiskers’ and fibers of Cu6Sn5 was discovered. In fact, three types of different morphologies were observed. First, hair-like fibers, believed to be caused by the solidification of the eutectic present in the Cu-Sn system. Then, a scallop-like layer morphology of Cu6Sn5 was formed which is believed to be the result of dissolution of the substrate and subsequent precipitation with additional Ostwald ripening. Finally facetted whiskers are observed, which probably originate during cooling as a result of constitutional supercooling.
118
Samenvatting In het laatste decennium is het milieubewustzijn gegroeid en dit heeft geleid tot wetsvoorstellen aangaande giftige elementen. Eén van deze elementen is lood. Hoewel het lood in de electronische industrie maar een klein percentage bedraagt van de totale loodconsumptie, zou dit lood van groot belang kunnen zijn in het milieu omdat het zo fijn is verdeeld in de micro-electronica. Daarom is er voorgesteld om naar alternatieven voor loodhoudend soldeer in de electronische industrie te zoeken. Omdat er niet veel gegevens bekend zijn over de fundamentele aspecten van solderen is onderzoek in deze richting zeer belangrijk. In dit onderzoek zijn systemen onderzocht die relevant zijn voor de soldeer industrie. Dit is gedaan met behulp van legeringen en diffusiekoppels. De monsters zijn geanalyseerd met lichtmicroscopie, raster electronenmicroscopie, electronen straal microanalysator en Röntgendiffractie. Nadat werd vastgesteld welke intermetallische verbindingen aanwezig zijn in de binaire diagrammen, zijn er ternaire isotherme doorsnedes bepaald met behulp van de bovengenoemde technieken. Dit is gedaan voor de volgende systemen: Sn-Ag-Sb bij 220 oC, Sn-Ni-Cu bij 235 oC en Bi-Ni-Pd bij 235 oC. Sn-Ag-Sb is een systeem dat gebruikt kan worden indien een soldeer met een hoger smeltpunt nodig is. In deze studie is ontdekt dat twee intermetallische fasen stabiel zijn in het Sn-Sb systeem; Sn3Sb4 en Sn4Sb3, in tegenstelling tot eerdere onderzoeken waar één fase gevonden werd. Omdat legeringen via verwarming bij deze
temperatuur
niet
tot
evenwicht
gebracht
kunnen
worden
is
de
diffusiekoppeltechniek de enige manier om faserelaties te bepalen. In het ternaire Sn-Ni-Cu systeem bij 235 oC is een nieuwe ternaire fase ontdekt. Dit zou een verklaring kunnen geven voor de hogere groeisnelheid van de gevonden intermetallische verbindingen in dit systeem wanneer Cu gelegeerd wordt met Ni in vergelijking met de groeisnelheid wanneer zuiver Cu reageert met vloeibaar Sn.
119
Het Bi-Ni-Pd systeem kan van belang zijn voor de electronische industrie omdat Bi een mogelijke vervanger voor Pb is in soldeermaterialen. Substraten zijn vaak bedekt met een laagje Ni, dat op zijn beurt weer gecoat is met Pd om een hogere oxidatieweerstand te verkrijgen. Behalve het vaststellen van de ternaire doorsnede van dit systeem bij 235 oC, is er ook onderzoek gedaan naar de vaste stof diffusie in de binaire Bi-Pd en Ni-Bi systemen. Dit heeft uitgewezen dat de γ- fase, Bi3Pd5, stabiel is bij 200 oC en niet alleen bij temperaturen hoger dan 400 oC, zoals eerdere studies uitwezen. Verder is er voor de eerste keer is er bewijs geleverd dat BiPd3 stabiel is bij lagere temperaturen. Het onderzoek heeft verder het groeigedrag van de aanwezige intermetallische lagen verduidelijkt, inclusief de waarde van de geïntegreerde en intrensieke diffusiecoëfficienten in diverse fasen. Tijdens het onderzoek werd abnormale ‘whisker’-groei geobserveerd in het Cu-Sn systeem bij vloeibaar -vast interacties. Dit gedrag is verder bestudeerd. Daarbij is gevonden dat de groei van de intermetallische verbindingen bij vloeibaar- vast interacties vele malen sneller gaat dan bij vast-vast interacties en dat de morfologie volledig verschillend is. Drie verschillende morfologien zijn waargenomen: Haarachtige vezels van Cu6Sn5, waarvan gedacht wordt dat ze gevormd worden door het stollen van het eutecticum aanwezig in het Cu-Sn systeem. Daarnaast wordt een laag van Cu6Sn5 met een heuvelachtige structuur gevormd, waarvan we aannemen dat ze ontstaat door het oplossen van het kopersubstraat met vervolgens precipitatie en “Ostwald ripening”. Ten slotte zijn Cu6Sn5- ‘whiskers’ gevonden, die waarschijnlijk ontstaan via het proces van ‘constitutionele onderkoeling’
120
Acknowledgement
Writing a thesis is a job which can never be done without the help, not just scientifically but also on a personal level, of many other people. I can only mention a few of them here (of course, all others that have supported me throughout the years are not forgotten): Als eerste natuurlijk mijn promotor, Prof.dr. Frans van Loo, die mij met raad en daad heeft bijgestaan, zelfs terwijl hij in het buitenland was. I would also like to express my gratitude to my second promotor Prof.dr. Jorma Kivilahti, for his advice and the nice discussions we had, often while enjoying a tasteful dinner. Also thanks for the opportunity to get a closer look at the Finnish culture. I had a great time in Finland, also thanks to all Finnish colleagues! Of course I thank Dr. Sasha Kodentsov for all his support and help to finish things up (including everything that is liquid!). De afstudeerders en stagiaires die een zeer essentiele bijdrage hebben geleverd aan dit werk: Ingrid Oomen, Mariëlle van Vinken, Patrick Kaas, Tom Jansen en de MDP-groep “lood-vrij solderen”. Gedurende de jaren in de groep heb ik meerdere kamergenoten gehad, die ik wil bedanken voor de gezellige sfeer op de kamer. In het bijzonder wil ik Stephan en Slobodan bedanken, samen waren we de eerste ‘bewoners’ van kamer STO 2.39. Al mijn collegas bij de groep SVM, met wie ik niet alleen tijdens werktijd met veel plezier heb samengewerkt, maar ook buiten werktijd. In het bijzonder wil ik bedanken: Han, Hans, Giel, Gerben, Huub, Linda, Csaba en Mark. De F.O.R.T., waar ik onder het genot van meer dan één pilsje de zorgen van me af kon schudden. Al mijn studievrienden, met name: Roger, Sander, Manfred, Edwin en Pascal. All my international friends, wherever they are, for making it fun to travel! Ola thank you for supporting me in the last, but difficult, period of my work. Kocham ci“ S»odka! Last but certainly not least: mijn ouders, omi, zus, neefjes, Bert, Henny en de rest van de familie.
121
122
Curriculum Vitae Pascal Johannes Theodorus Lambertus Oberndorff was born on the 28th of March 1972 in Geleen, the Netherlands. After finishing Gymnasium at the Bernardinuscollege in Heerlen in 1990, he decided to study Chemical Engineering at the Eindhoven University of Technology. He graduated there in August 1996 on the subject ‘EPMA Measurements with Non-perpendicular Incidence of the Electron Beam’. After staying 2 months in South Africa and working for 5 months as coordinator at the Science Shop for Chemistry and Chemical Engineering at the Eindhoven University of Technology, he started his Ph.D.-project in the laboratory of Solid State and Materials Chemistry, under supervision of Prof.dr. F.J.J. van Loo, which resulted in this thesis.
123
Stellingen behorende bij het proefschrift Lead-free Solder Systems: Phase Relations and Microstructures By Pascal Oberndorff 1. De diffusiekoppeltechniek is de beste experimentele methode om fasenrelaties bij een lage temperatuur te bepalen (Hfd. 4). The diffusion couple technique is the best experimental method to establish phase relations at low temperatures (Chapter 4).
2. De bewering van Choi et al. en Oh over het bestaan van een ternaire fase in het SnNi-Cu systeem worden niet met overtuigende feiten ondersteund (Hfd. 5). Statements made by Choi et al. and Oh about the existence of a ternary phase in the Sn-Ni-Cu system are not supported by convincing evidence (Chapter 5). W.K. Choi and H.M. Lee, J. Electron. Mater. 28 11 (1999) 1251 M.Oh, Ph.D. thesis Lehigh University, USA (1994) 126-135
3. Het niet voldoende op de hoogte zijn van de principes van de matrix correctie procedures in electronen microscopie leidt vaak tot een verkeerde interpretatie van de meetresultaten (Hfd. 2&3). Not knowing the principles of matrix correction programs in electron microscopy often leads to a wrong interpretation of the result of the analyses (Chapter 2 & 3).
4. ‘Whiskers’ in het Cu-Sn systeem ontstaan door ‘constitutionele onderkoeling’ (Hfd. 7). Facetted whiskers in the Cu-Sn system originate as a result of constitutional supercooling (Chapter 7).
5. Het bestaan van een Koningshuis is strijdig met artikel 1 van de Verklaring van de Universele Rechten van de Mens. The existence of Royalty is in conflict with article 1 of the Universal Declaration of Human Rights. Article 1: ‘All human beings are born free and equal in dignity and rights. They are endowed with reason and conscience and should act towards one another in a spirit of brotherhood’ Universal Declaration of Human Rights, United Nations New York USA (1948).
6. Trekken gaat makkelijker dan duwen. Pulling is easier than pushing. Manual garbage can, Heerlen the Netherlands (1992)
7. Het intelligentste dier op deze aarde, is dat dier dat niet door de mensheid wordt ontdekt. The smartest animal on this planet is that animal that will not be discovered by mankind.
8. Nare ervaringen in iemands jeugd zouden minder vaak gebruikt moeten worden als een excuus voor latere misstappen, maar zouden juist moeten dienen als motivatie om misstappen te voorkomen. Bad experiences in someone’s childhood should be used less frequently as an excuse for later misbehavior, but instead they should be used as motivation to avoid misbehavior.
9. Successierecht kan worden gezien als een vorm van lijkenroof. Inheritance tax can be seen as robbing a corpse.
10. De vraag of lood in ammunitie schadelijk is voor de gezondheid is overbodig. Questioning whether the use of lead in ammunition is unhealthy is unnecessary.
11. Carnaval is celebrated in February/March with greater vigour in Maastricht than almost anywhere else in Europe. The Netherlands, Lonely Planet, Footscray Australia (2001) 311
12. Geniet en drink met maten!