Harnessing heterogeneous nucleation to control tin orientations in electronic interconnections

While many aspects of electronics manufacturing are controlled with great precision, the nucleation of tin in solder joints is currently left to chance. This leads to a widely varying melt undercooling and different crystal orientations in each joint, which results in a different resistance to electromigration, thermomechanical fatigue, and other failure modes in each joint. Here we identify a family of nucleants for tin, prove their effectiveness using a novel droplet solidification technique, and demonstrate an approach to incorporate the nucleants into solder joints to control the orientation of the tin nucleation event. With this approach, it is possible to change tin nucleation from a stochastic to a deterministic process, and to generate single-crystal joints with their c-axis orientation tailored to best combat a selected failure mode.


Harnessing heterogeneous nucleation to control tin orientations in electronic
The reason for the positioning of the black square in Figure 2A is not apparent. p. 6: How were Sn droplets made and deposited on the substrate? This is not described. Were they deposited as balls, and then heated and cooled in a DSC? Also, did you use oriented single crystals of all three of the Stannide IMC materials? This is not clear. There should be enough information so that others can reproduce your work.
p. 8 Al on Al3Ti 44). However, importantly, the ORs reported in Table 2 for all 182 of the droplets examined, irrespective of which OR formed, the [001] of tetragonal βSn was always in the plane of the largest facet of the IMC. Thus, these ORs are not only reproducible but also useful since, by controlling the You describe facets of the IMC, but the reader has no way of know what the facet refers to. I thought you were using a polished single crystal, but reference to a facet makes this seem like there was a polycrystal. Please clarify what the condition of the substrate was? βSn grain structure and orientation control in solder joints. To incorporate a PtSn4, αCoSn3, or βIrSn4 seed crystal into ball-grid array (BGA) solder joints, they were first bonded to Cu pads using a form of transient liquid phase bonding (TLPB). An immersion tin coating was applied to Cu pads and the nucleant IMC wafer was laid flat with the main facet in the plane of the pad as shown in Figure 5 (A) using αCoSn3 as an example. The αCoSn3 was then TLPB to the pad by a reflow of 5 minutes at 240°C, Please clarify: while you used TLPB, did the little wafer of the IMC stay solid during the bonding process? Was there any flux used?
Regarding the prior comment, it appears that you sliced wafers of a IMC single crystal and that you refer to the flat surface as the facet. Is this correct? p. 9 The nucleation undercooling of 550 μm Cu/SAC305+IMC/Cu joints for each type of IMC seed crystal were measured and are plotted versus the lattice disregistry in Figure 4 (D?). There is a similar trend Practical question -did you put a board+chip with a single solder joint in the dsc? It is hard to imagine what you did to get the data in Figure 4D?
The methods section answers above questions adequately, except for the making of a bga joint in a DSC. In the main body of the text, clarifying how you made and placed the nucleant wafers would help the reader not get distracted from your story. p. 12 The Sn was melted on a hot plate with a peak temperature of 240 ℃ to ensure wetting and spreading on the facet, and then flux residues were removed. Thermal cycles were then performed 10 How were flux residues removed? chemical analysis mould for X-Ray Fluorescence (XRF) spectroscopy analysis. To prepare solder balls with a diameter of 550±25 μm, Sn-3.0Ag-0.5Cu and Sn-3.5Ag were rolled to 0.05 mm foils, punched to AE1.6 mm discs, and reflowed on an inert hotplate with a ROL-1 tacky flux (Nihon Superior Co.,20 what was the intert surface? p. 13 crystals were placed on the tin layer with their main facet near-parallel with the substrate at a 7 desired rotation angle, and the nucleants were transient liquid phase bonded to the substrate by 8 holding at 240 ℃ for 5 min (for the αCoSn3 case) or at 300 ℃ for 180 min (for the PtSn4 and IrSn4 9 cases). Since PtSn4, αCoSn3 and βIrSn4 are all solderable surfaces using a standard ROL-1 flux, these 10 nucleant-modified pads were then used in the same manner as a Cu-OSP substrate.
What about environment -in air or nitrogen?

Reviewer #2 (Remarks to the Author):
This is a very good paper, well planned, well implemented and well presented. However., I m not sure that it is suitable for publication in Nature. My understanding is that Nature papers should present significant new science and, while the implementation of the ideas presented here is new and definitely contribute to solder technology, I really don't see any new science. The basic ideas are well known, long discussed, and widely used in one form or another.
The paper is good and interesting, and I will be happy to recommend its publication in Nature Communications if the editors believe its scientific (as opposed to technological) novelty is sufficient to justify it.
Reviewer #3 (Remarks to the Author): Both academic and industrial people are eager to seek methods to effectively control the grain orientation of solder joints due to the anisotropy of β-Sn in mechanical and physical properties. The authors claim they have generated single crystal joints by using IMC nucleants as seed crystal. The authors are supposed to address the comments:

Response to comments:
Reviewer text is in blue italics, our response is in black.
We are grateful to the Reviewers for raising valuable comments and suggestions. Based on the review, we have made revisions (highlighted in yellow in the revised manuscript) that we feel have improved the manuscript.

Reviewer #1
This paper presents a major step forward for improving the reliability and design of solder joints. It could well be the start of a revolutionary change in the industry. The authors have recognized the importance of oriented nucleation and found a practical means to get this under control, with impressive results. There are minor editorial suggestions in the attached document, but the paper is publishable in NCOMMS.

p. 5 line 21 electronic fluxes, not electronics fluxes
The 'electronics fluxes' has been changed into 'electronic fluxes' on page 5 line 21.
The reason for the positioning of the black square in Figure 2A is not apparent. To clarify this, the following text has been added to the caption of Figure 2 on page 22: The black square on each net of Sn atoms indicates the projection of the corresponding unit cell, and its position is determined by the origin for the crystallographic settings in Table 1.
p. 6: How were Sn droplets made and deposited on the substrate? This is not described. Were they deposited as balls, and then heated and cooled in a DSC? Further details of the droplet nucleation experiments have been added to the Methods section on page 13: '99.9% purity Sn particles (balls) of between 1 and 50μm were placed on the largest facets of the as-extracted (without further processing) PtSn 4 , αCoSn 3 , and βIrSn 4 single crystals with a NH 4 Cl-ZnCl 2 based flux (Stay-Clean liquid flux, HARRIS). The Sn was melted on a hot plate with a peak temperature of 240 ℃ to ensure wetting and spreading on the facet, and then IMC crystals with Sn droplets spread on were ultrasonically bathed in ethanol to remove flux residues. Thermal cycles were then performed in a Mettler Toledo DSC 1 under a nitrogen atmosphere with a heating rate of 0.17 K/s, peak temperature 240 ℃, time above the eutectic temperature of 180 s and a cooling rate 0.33 K/s.' The following text is added on page 6 line 24 to alert the reader to the detailed methods at the end of the paper: Experimental details are as given in the Methods section. The following text has been added to page 7 line 23: …inert oxidised Al substrates.
Figure caption 4 on page 24 was also expanded with : All samples were measured on inert oxidised Al substrates.

Also, did you use oriented single crystals of all three of the Stannide IMC materials? This is not clear. There should be enough information so that others can reproduce your work.
All three stannide IMCs were used for droplet nucleation experiments and joints. More details have been added to clarify the whole experimental processes in the Methods section on page 13 line 8: ….the largest facets of the as-extracted (without further processing) PtSn 4 , αCoSn 3 , and βIrSn 4 single crystals with a NH 4 Cl-ZnCl 2 based flux (Stay-Clean liquid flux, HARRIS). The Sn was melted on a hot plate with a peak temperature of 240 ℃ to ensure wetting and spreading on the facet, and then IMC crystals with Sn droplets spread on were ultrasonically bathed in ethanol to remove flux residues.

p. 8
Al on Al 3 Ti 44 ). However, importantly, the ORs reported in Table 2  Since the natural growth shape of primary αCoSn 3 , PtSn 4 , and βIrSn 4 crystals are thin plates (Figure 3 (A)-(C)) whose main facet is the desired nucleation plane, seed crystals were obtained by dissolving the matrix βSn and using the largest natural growth facet as a seed crystal without the need to take slices from a wafer of IMC.
βSn grain structure and orientation control in solder joints. To incorporate a PtSn 4 , αCoSn 3 , or βIrSn 4 seed crystal into ball-grid array (BGA) solder joints, they were first bonded to Cu pads using a form of transient liquid phase bonding (TLPB). An immersion tin coating was applied to Cu pads and the nucleant IMC wafer was laid flat with the main facet in the plane of the pad as shown in Figure 5 (A) using αCoSn 3 as an example. The αCoSn 3 was then TLPB to the pad by a reflow of 5 minutes at 240°C, Please clarify: while you used TLPB, did the little wafer of the IMC stay solid during the bonding process? Was there any flux used? To clarify this, text was added on page 8 line 12: ….where the IMC seed crystal remained solid during the TLPB.
Text was also added in the Methods section on page 14 line 7 to clarify what flux is used: PtSn 4 , αCoSn 3 or βIrSn 4 crystals were placed on the tin layer with NH 4 Cl-ZnCl 2 based flux (Stay-Clean liquid flux, HARRIS) coated on…...

Regarding the prior comment, it appears that you sliced wafers of a IMC single crystal and that you refer to the flat surface as the facet. Is this correct?
Sorry for the lack of clarity here. To make this clear, we have added the following text to page 8 line 7: Since the natural growth shape of primary αCoSn 3 , PtSn 4 , and βIrSn 4 crystals are thin plates (Figure 3 (A)-(C)) whose main facet is the desired nucleation plane, seed crystals were obtained by dissolving the matrix βSn and using the largest natural growth facet as a seed crystal without the need to take slices from a wafer of IMC.

p. 9 The nucleation undercooling of 550 μm Cu/SAC305+IMC/Cu joints for each type of IMC seed crystal were measured and are plotted versus the lattice disregistry in Figure 4 (D?). There is a similar trend
Practical question -did you put a board+chip with a single solder joint in the dsc? It is hard to imagine what you did to get the data in Figure 4D?
The methods section answers above questions adequately, except for the making of a BGA joint in a DSC. To make this clear, additional text is added to the Methods section on page 14 line 13: All solder joints were firstly made by reflowing in the Tornado LFR400 reflow oven. Some joints were reflowed subsequently in the DSC to measure the nucleation undercooling. Substrates were cut to ~3x3mm for DSC measurements.
In the main body of the text, clarifying how you made and placed the nucleant wafers would help the reader not get distracted from your story. The paragraph on page 8 is expanded with the new yellow text to clarify the joint-making process: βSn grain structure and orientation control in solder joints. To incorporate a PtSn 4 , αCoSn 3 , or βIrSn 4 seed crystal into ball-grid array (BGA) solder joints, they were first bonded to Cu pads using a form of transient liquid phase bonding (TLPB), where the IMC seed crystal remained solid during the TLPB. An immersion tin coating was applied to Cu pads and the nucleant IMC was laid flat with the main facet in the plane of the pad as shown in Figure 5 (A) using αCoSn 3 as an example. The αCoSn 3 was then TLPB to the pad by a reflow of 5 minutes at 240°C, which resulted in the Cu/Cu 3 Sn/Cu 6 Sn 5 /αCoSn 3 layers shown in Figure 5 (B). Since PtSn 4 , αCoSn 3 and βIrSn 4 are all solderable surfaces using a standard ROL-1 flux, these nucleant-modified pads were then used in the same manner as a Cu-OSP substrate and solder joints were made following the procedure in the Methods section.

p. 12
The Sn was melted on a hot plate with a peak temperature of 240 ℃ to ensure wetting and spreading on the facet, and then flux residues were removed. Thermal cycles were then performed 10 How were flux residues removed? Sorry for the lack of this information here. Additional text has been added to the Methods section on page 13 line 10 to clarify this: …..and then IMC crystals with Sn droplets spread on were ultrasonically bathed in ethanol to remove flux residues. chemical analysis mould for X-Ray Fluorescence (XRF) spectroscopy analysis. To prepare solder balls with a diameter of 550±25 μm, Sn-3.0Ag-0.5Cu and Sn-3.5Ag were rolled to 0.05 mm foils, punched to 1.6 mm discs, and reflowed on an inert hotplate with a ROL-1 tacky flux (Nihon Superior Co.,20 what was the intert surface? Text has been added to page 13 line 21 and 22 to make this clear: …..and reflowed on a non-wetting highly-oxidised Ni sheet with a ROL-1 tacky flux (Nihon Superior Co., Ltd.) and peak temperature of 280 ℃.
p. 13 crystals were placed on the tin layer with their main facet near-parallel with the substrate at a desired rotation angle, and the nucleants were transient liquid phase bonded to the substrate by holding at 240 ℃ for 5 min (for the αCoSn 3 case) or at 300 ℃ for 180 min (for the PtSn 4 and IrSn 4 cases). Since PtSn 4 , αCoSn 3 and βIrSn 4 are all solderable surfaces using a standard ROL-1 flux, these nucleant-modified pads were then used in the same manner as a Cu-OSP substrate.
What about environment -in air or nitrogen? Text has been added to page 14 line 15 to make this clear: ….N 2 atmosphere was used in all reflowing cases.

Reviewer #3
Both academic and industrial people are eager to seek methods to effectively control the grain orientation of solder joints due to the anisotropy of β-Sn in mechanical and physical properties. The authors claim they have generated single crystal joints by using IMC nucleants as seed crystal. The authors are supposed to address the comments: 1. For Fig. 3, could the authors provide the EBSD map in X and Z directions? Similarly,the EBSD maps in X and Y directions for Figs. 5 and 6. Maps in the other directions have been added to the Supplementary Information, and are also given below. The IPF X, Y and Z maps of the new Supplementary Figures 3, 4, 5, and 6 all show joints with βSn orientated with its c-axis near-parallel with the substrate plane. In Supplementary Figure 4, note that the mechanical twin across one sample and minor orientations at the sample edges were caused by deformation during sample preparation.
We have also made all EBSD datasets available on the Zenodo repository for readers to explore if they wish. The following data availability statement has been added to the end of the Methods section: Data availability. The EBSD and DSC datasets generated during the current study are available in the Zenodo repository (doi: 10.5281/zenodo.884113).  We have added the following text to page 8 and 9 to provide more information on the influence of the Ag 3 Sn and Cu 6 Sn 5 IMCs: The microstructure consists of primary Cu 6 Sn 5 , βSn dendrites with <110> growth directions indicated by the blue arrows, βSn+Ag 3 Sn+Cu 6 Sn 5 interdendritic eutectic, and a bonded αCoSn 3 particle on the bottom substrate…. Note that, even though primary and interfacial Cu 6 Sn 5 are present prior to βSn nucleation, the nucleation of βSn always occurred on the seed crystal because they (αCoSn 3 , PtSn 4 , βIrSn 4 ) are more potent nucleants than Cu 6 Sn 5 . Primary Ag 3 Sn plates were not observed here because the seed crystals require only a relatively small undercooling for βSn nucleation We have added the following text to page 12 to provide more information about the role of Bi and In enrichment phases on the nucleation of βSn.
The seed crystals remain effective with these additions because they do not react with the seed crystals or introduce a more potent nucleant phase. For example, the Ni addition mostly influences the Cu 6 Sn 5 phase which is a less potent nucleant than the seed crystals; and these Bi and In additions introduce extra (Bi) and ζ(Ag,In) phases that form later during solidification and do not strongly affect the nucleation of βSn. The successful orientation control in these Ni, Bi and In-containing solders indicates that this orientation control method is likely to be applicable to solder compositions developed in the future.
We have also added further annotations to Supplementary Figure 7 to more clearly show the phases in each joint containing Ni, Bi and In.