Supporting Online Material for

Transcription

1 Supporting Online Material for The Role of a Bilayer Interfacial Phase on Liquid Metal Embrittlement Jian Luo,* Huikai Cheng, Kaveh Meshinchi Asl, Christopher J. Kiely, Martin P. Harmer* *To whom correspondence should be addressed. mph2@lehigh.edu (M.P.H.); jluo@alum.mit.edu (J.L.) This PDF file includes: Materials and Methods SOM Text Figs. S1 to S13 Published 23 September 2011, Science 333, 1730 (2011) DOI: /science

2 Supporting Online Material The Role of a Bilayer Interfacial Phase on Liquid Metal Embrittlement Jian Luo 1, *, Huikai Cheng 2, Kaveh Meshinchi Asl 1, Christopher J. Kiely 2, and Martin P. Harmer 2, * 1 School of Materials Science and Engineering; Center for Optical Materials Science and Engineering Technology, Clemson University, Clemson, SC 20634, USA 2 Department of Materials Science and Engineering; Center for Advanced Materials and Nanotechnology, Lehigh University, Bethlehem, PA, USA * To whom correspondence should be addressed. mph2@lehigh.edu (M.P.H.); jluo@alum.mit.edu (J.L.)

3 Materials and Methods: To prepare LME specimens, small pieces of Bi-Ni alloys of the liquidus compositions (26.5 at. % Ni at 700 C or 39 at. % Ni at 1100 C; thus, at either annealing temperature, the liquid metal was always in a chemical equilibrium with the Ni polycrystalline solid and there was virtually no dissolution of Ni into the liquid) were placed on the top of Ni foils ( %; grain size = ~ 75 μm), and the assemblies were annealed isothermally at 700 or 1100 C in a tube furnace with flowing Ar - 4 % H 2. All specimens were water quenched to preserve the high temperature structure. Quenched specimens were sectioned, polished and examined using scanning electron microscopy (SEM, Hitachi S- 4800). Transmission electron microscopy (TEM) specimens were prepared by the combination of a focused ion beam instrument (FIB; FEI Strata DB 235) and a careful low energy ion milling technique (Fischione 1010; for polishing the specimens to remove possible preparation artifacts caused by FIB) and characterized using an aberration corrected scanning TEM (STEM; JEOL 2200FS). All STEM micrographs reported here (except for Fig. S1(a) and S1(c)) are Z-contrast high-angle annular dark-field (HAADF) images. Magnifications of STEM HAADF images were calibrated by measuring silicon lattice spacings. Methodology: A merit of this study was that we have used a combination of various complementary characterization tools and a systematic experimental approach. First, we used SEM to extensively characterize the cross-sectional and fracture surfaces of many Ni-Bi specimens. Then, we targeted specific grain boundaries (GBs) of interest and used the FIB lift out technique to prepare TEM specimens of specific selected locations from cross sectional samples. These were then characterized with high-resolution transmission

4 electron microscopy (HRTEM), energy-dispersive X-ray (EDX) spectroscopy and aberration corrected HAADF STEM. We have imaged seven independent GBs in polycrystals (instead of special GBs in bicrystals with pre-determined orientation) in four separate specimens prepared using three different sets of conditions (one annealed at 700 C for 1 hour; one annealed at 700 C for 5 hours; and two annealed at 1100 C for 5 hours). Bilayers were observed in all but one GB. Subsequently, we used electron backscatter diffraction (EBSD) to show that the only GB that was free of a bilayer is in fact a low-angle special GB. In one specimen (annealed at 1100 C for 5 hours), we examined four GBs over a length of greater than one millimeter. Such experiments that were focused on general (i.e. high-angle and low symmetry) GBs require highly exacting sample preparation and microscope set-up conditions.

5 SOM Text Figs. S1-S13 (I) Conventional HRTEM imaging As shown in Fig. S1, the bilayer interfacial phase observed in this study cannot be clearly discerned in phase-contrast lattice imaging using conventional HRTEM or STEM bright-field imaging. Fig. S1(a) and Fig. S1(b) are a pair of bright-field and HAADF STEM images of exactly the same area. The conventional HRTEM image of the same GB was also taken; HRTEM and STEM images that were taken are for the same GB at close locations, and we have sampled over the whole GB and found that different areas along that GB exhibit similar character. The application of Z-contrast HAADF STEM enabled us to visualize this new interfacial structure, which is not clearly distinguished by conventional HRTEM or bright-field STEM. Furthermore, EDX spectroscopy confirmed that bismuth is enriched in these bilayers (Fig. S2), which gives the bright contrast in the HAADF images.

6 Fig. S1. Aberration-corrected STEM a) bright-field and b) HAADF images of the exactly same area of a GB in a specimen that was annealed at 700 C for 5 hours. c) A phase-contrast lattice image of the same GB using conventional HRTEM. The bilayer interfacial phase, which is clearly visible in the STEM HAADF image shown in (b), cannot be clearly resolved by conventional HRTEM or bright-field STEM imaging.

7 Fig. S2. EDX spectra of the areas a) on a GB (bilayer) and b) inside the adjacent grain indicate that bismuth (Bi) is present in the bilayer. The inset is a STEM HAADF image where the areas on which EDX analyses were performed are labeled. (II) The detailedd structure of the bilayer Careful extrapolation of the atomic planes in the grains indicates that both of the Bi columns line up with the adjacent atomic planes perfectly but not with the planes on the other side (Fig. S3). This suggests that the right column of the Bi atoms actually are adsorbed coherently on the right Ni grain surface (GB plane), and the left column of the Bi atoms are adsorbed coherently on the left Ni grain surface. However, there is no lattice matching relationship between the two adsorbed Bi layers (i.e., an incoherent Bi- Bi interface). This interfacial structure of the bilayer is schematically illustrated in Fig. 4(B) in the Report and Fig. S11(b).

8 Fig. S3. STEM HAADF micrograph showing the relationship between two columns of Bi atoms and the lattices of the adjacent grains. It shows that the right column of the Bi atoms matches with the blue lines but not with the red lines, while the left column of the Bi atoms matches well with the red lines but not with the blue lines. This indicates that the two columns of the Bi atoms are actually adsorbed (coherently) respectively on either side of the GB planes (grain surfaces) and there is no lattice matching relationship between the two adsorbed Bi layers. (III) Excluding imaging artifacts Significant efforts were made to exclude possible imaging artifacts. First, the analysis in Fig. S3 shows that the observed bi-layer adsorption is not a projected zigzag structure or atomic-level faceting. Second, a series of images were obtained at different defocus values (Fig. S4) to ensure that the observed bilayers are not monolayer adsorbates on an atomic step (which would look like a bilayer in a two-dimensional projection). Finally, since this bilayer interfacial phase was observed at several GBs from different specimens,

9 it is highly unlikely that each random GB exhibits a zigzag structure or has exactly one atomic step. Thus, the observed GB structure represents true bilayer adsorption. Fig. S4. Through-focus series of STEM HAADF micrographs showing no significant relative intensity change of the two columns of the Bi atoms with defocus (varying depth) values. The defocus values are labeled. (IV) Additional notes about STEM HAADF images STEM HAADF images are always two-dimensional projections of three-dimensional atomic arrangements. While the STEM images of bilayer adsorption Bi at Ni GBs in Fig. 2 in the Report all show two arrays of Bi columns, these are images after tilting of specimens so that the individual Bi columns can be resolved. When it is projected along a general direction (for a still edge-on GB), the arrays of Bi columns will be smeared to

10 be lines in the two-dimensional (projected) images. Some of such less periodic (but in fact more general) STEM images are shown in Fig. S5. We also want to note that the in-layer period of the adsorption of Bi depends on the orientation of the grain surface as well as the projected imaging direction (the measured in-layer periods range from to 0.48 nm for 16 2 individual arrays), and Fig. S3 shows that there is no matching relationship between the two adsorbed Bi column arrays. Fig. S5. STEM HAADF micrographs of less periodic (but more generally occurring) images of adsorbed Bi bilayers, where the specimens were not tilted to low-index directions to resolve individual adsorbed Bi columns. (V) Discussion of the bilayer terminology We refer to the observed interfacial phase as bilayers because the majority of adsorbed Bi atoms are located within two atomic layers at the GB core according to our aberration corrected STEM HAADF images. This does not necessarily imply that the GB excess of Bi is identically equivalent to 2 monolayers of pure Bi. In a series of prior quantitative studies on other interfacial systems (18-21) using an STEM-EDX area-scan method that was initially developed by Ikeda et al. (21), GB excesses were measured in atoms per nm 2 and subsequently converted to the equivalent number of monolayers assuming that the adsorbates have the same density as a pure bulk phase of the impurity. However, it should be noted that the geometrical interfacial width (measured by an imaging technique) of an interfacial phase is not necessarily equal to the equivalent chemical interfacial width (estimated indirectly from the measured interfacial excess

11 with simplifying assumptions that (i) the adsorbates follow the structure of a bulk phase and (ii) that there is no mixing). In the present case, (i) the GB excess of these bilayers is likely less than a 2 monolayer equivalent of a pure bulk phase of Bi (according the conventional definition of the monolayer coverage in prior studies (18-21)) and (ii) the GB excess of Bi should exhibit a significant boundary-to-boundary variation for the following two reasons: (1) Aberration-corrected STEM HAADF images (Figs. 2A and S3) show that each Bi layer in the bilayer is coherently adsorbed on the Ni grain surface, and that the density of the adsorption sites should therefore depend on the orientation/inclination of the specific grain surface (GB plane). For a general GB, the two grain surfaces do not generally exhibit low-index, close packed orientations. Thus, the densities of the adsorbed sites should be significantly lower than the density of Bi atoms in a bulk phase. Furthermore, the density of the adsorption sites is dictated by the specific surface orientation of the adjacent grain; thus, individual bilayers at general GBs of different crystallographic character should have different levels of GB excesses, even if they are all bilayers based on our determination of interfacial geometrical width. (2) Moreover, entropy-driven mixing is inevitable at high temperatures, which should further reduce the actual GB excesses of Bi in these bilayers. This is analogous to the case of Bi 2 O 3 -doped ZnO, where a carefully-conducted prior study (19) showed that GB excesses of Bi 2 O 3 in solid-state sintered specimens measured by the STEM-EDX area-scan method are only equivalent to < 1 monolayer of pure Bi 2 O 3 (i.e., the equivalent chemical interfacial width is less than 1 monolayer), while HRTEM showed the GBs to have ~ 1 nm thick Bi 2 O 3 -enriched adsorbed intergranular films (i.e. the actual geometrical interfacial width is greater than ~3 monolayers). This was explained to be due to a significant intermixing of ZnO and Bi 2 O 3 in these adsorbed films that leads to ZnO-rich binary film compositions (19).

12 (VI) Characterization of multiple GBs in one specimen Fig. S6 illustrates the characterization of multiple GBs in one specimen that was annealed at 1100 C for 5 hours and quenched. In this specimen, four independent GBs (located from near a liquid penetration tip to ~ 1 mm away from the liquid front) were extracted using a FIB and characterized using STEM HAADF. Bilayers were observed at three of the four GBs. The only boundary for which bilayer adsorption of Bi was not observed (GB #2 in Fig. S6) was found to be a special (low-angle) low-energy GB by EBSD mapping, as shown in Fig. S7. These observations are schematically illustrated in Fig. 3A in the Report. Furthermore, GB #4 in Fig. S6 is comprised of multiple facets; bilayers were observed at all facets that could be clearly imaged (six facets in total, four of which can be seen in Fig. S6). Enlarged images of bilayers on these facets are shown in Fig. S8.

13 Fig. S6. SEM micrograph of four GBs at different distances from the liquid front and the corresponding STEM HAADF micrographs of these GBs. Locations where TEM specimens were lifted using a FIB (the trenches) can be seen in the SEM micrograph, and the approximate distances to the liquid front are labeled. GB #2, which is virtually free of Bi adsorption, was found to be a low-angle GB by EBSD characterization (Fig. S7). GB #4 was faceted; yet bilayers were observed at all six (6) facets that can be clearly imaged (four facets are shown here; additional enlarged images are shown in Fig. S8). These observations are schematically illustrated in Fig. 3(A) in the Report.

14 Fig. S7. a) SEM micrograph of the GB # 2 shown in Fig. S6. The inset is the EBSD map showing that it is a low-angle GB with misorientation of 7.5 / [0, -13, -18]. b) STEM HAADF micrograph showing no observable Bi atoms along this GB. Fig. S8. Bilayers were observed at all facets along the GB #4 shown in Fig. S6.

15 (VII) Summary of observed bilayers at multiple GBs in multiple specimens We have characterized seven independent GBs in four different specimens that were processed with three sets of different conditions (2 specimens annealed at 1100 C for 5 hours, 1 specimen annealed at 700 C for 5 hours; and 1 specimen annealed at 700 C for 1 hour). Among all GBs that we characterized, only one GB was free of bilayer adsorption (i.e., the GB #2 in Fig. S6), and the subsequent EBSD analysis showed that this particular GB was a special (low-angle) GB (Fig. S7). Bilayers were observed in all of the other six GBs; representative STEM HAADF micrographs are shown in Fig. S9. Fig. S9. Representative STEM HAADF micrographs of the bilayers observed at six independent general GBs in four specimens: a) Specimen A (700 C 1 hour); b) Specimen B (700 C 5 hours); c) Specimen C (1100 C 5 hours); d) Specimen D (1100 C 5 hours; GB #1, near a tip); e) Specimen D (1100 C 5 hours; GB #3, ~661 μm away from the liquid front); and f) Specimen D (1100 C 5 hours; GB #4, ~983 μm away from the liquid front).

16 Moreover, in one case (GB #4 in Fig. S6), an independent GB comprised of several faceted segments (Fig. S8), which are also largely independent (since they are not connected, although they have the same misorientation). In this case, bilayers have been observed on all six faceted segments that can be clearly imaged. Several images are shown in Fig. S6 (GB #4), Fig. S8 and Fig. S9(f). Thus, we have effectively characterized 12 GB segments: with an exception of a low-angle special GB, bilayers have been observed on 11 independent GBs or faceted GB segments. We wish to make two further comments regarding our experiments. First, a distinct merit of this study is our interest in characterizing general (low-symmetry random ) GBs. While there is a significant bias in prior studies toward special Σ or low-angle GBs, general GBs are more important for real materials and they often exhibit different phase behaviors than those special GBs; special GBs are much less likely to exhibit structural (phase) transitions because they already have low energies. Second, preparing and characterizing TEM specimens for general GBs are much more challenging and time consuming. The number of observations made in this study represents a substantial amount of experimental work. In summary, the fact that we have (i) characterized seven randomly selected GBs (that comprise 12 independent GB segments ), (ii) found bilayers at all except one of them, and (iii) further showed by EBSD analysis that the only GB where a bilayer was not observed is a special (low-angle) GB enable us to draw a statistically significant conclusion that bilayers are representative for general GBs. (VIII) Observations of Bi precipitation and wetting Isolated Bi particles have precipitated at GBs, as shown by the SEM image of a fracture surface (Fig. S10(a)) and the STEM HAADF image (Fig. S10(c)). Such nanoscale Bi precipitates have also been observed in a prior study (3). Our SEM results also showed that triple line wetting likely occurred at 1100 C (Fig. S10(a) and (b)). Finally, it appears that the liquid film can jump at GBs, as shown in Fig. S10(d), although the two liquid films shown in Fig. S10(d) can be connected in 3-D. The observations of isolated Bi precipitates and apparent jumps indicate that Bi atoms can

17 diffuse rapidly along the bilayer interfacial phase. This is consistent with the suggestion that diffusion rates in bilayers should be generally greater than those in both GBs with Langmuir-McLean type adsorption and intrinsic GBs (7, 8). Fig. S10. a) and b) SEM micrographs of fractured surfaces indicating the precipitation of isolated small Bi particles and the occurrence of triple line wetting. c) STEM HAADF micrograph of a nanoscale Bi precipitate at a GB. d) Cross-sectional SEM micrograph showing a gap in the penetrating liquid Bi film along a Ni GB. In our experiments, we always started with binary Bi-Ni liquids with compositions on the liquidus line (thus the liquid was always in chemical equilibration with Ni at the specific annealing temperature). Thus, this is not a case of reactive wetting with respect to the bulk phase formation. However, the bilayer interfacial phase does form via diffusion and a reaction of Bi with Ni at GBs. In this regard, the formation of this bilayer interfacial phase has some phenomenological similarities to reactive wetting.

18 (IX) An explanation of the stabilization of the bilayer interfacial phase In the Report, we propose a process for the formation of this bilayer interfacial phase. In the presence of liquid Bi species, a general GB in Ni would be separated into two grain surfaces, each with a monolayer surface adsorption of Bi, which are only weakly bonded to each other. Such a bilayer interfacial phase can be stabilized if Bi atoms bond stronger to the Ni atoms on the adjacent Ni grain surface than to themselves (adsorbed Bi atoms). To support this argument, we performed a calculation using the Miedema model (using the free software developed by Dr. R. F. Zhang, available at and estimated the formation enthalpy for a Bi 0.5 Ni 0.5 alloy to be -3.7 kj/mol (i.e., E Ni-Bi > ½ E Ni-Ni + ½ E Bi-Bi ). This is also consistent with the formation a NiBi compound at low temperatures (<640 C) in this binary system. Since the strength of similar bonds scales with the melting temperature, Bi-Bi bonds should be very weak ( E Bi-Bi << E Ni-Ni ). Thus, Bi-Ni bonds are likely much stronger than the Bi-Bi bonds ( E Bi-Bi << E Ni-Bi ). As illustrated in Fig. S11, it is energetically more expensive to form a Bi monolayer or a Ni-Bi compound-like layer at a general GB because they both require the breaking of a substantial fraction of Ni-Bi bonds to form at least one incoherent Ni-Bi interface. First, if Bi-Ni bonds are strong, a monolayer adsorption is energetically unfavorable at a general GB, because it is impossible to maintain coherence of a Bi monolayer with both Ni grain surfaces (Fig. S11(a)). Second, a nanoscale Bi-Ni compound-like interfacial phase would also be difficult to form at general GBs because of the lattice mismatch (Fig. S11(c)). Finally, the stabilization of bilayers is schematically illustrated in Fig. S11(b). In this case, two Bi layers are coherently adsorbed on the adjacent Ni grain surfaces to form strong Ni-Bi bonds, and a mismatch can occur between the two coherently adsorbed Bi layers to break the weak Bi-Bi bonds (to form and incoherent Bi-Bi interface). Overall, this configuration is energetically favored. This proposed stabilization mechanism for bilayers (as illustrated in Fig. S11(b)) is firmly supported by (i) the lattice matching analysis of a STEM HAADF image in Fig. S3, which illustrates that each Bi layer is adsorbed on the surface of the adjacent Ni grain coherently, but there is no matching relationship between the two adsorbed Bi layers; and

19 (ii) the analysis of bond lengths in Fig. S13, which shows the lengthening and weakening (and presumably breaking) of the interlayer Bi-Bi bonds. Fig. S11. The stabilization of bilayers in this system can be rationalized by the fact that the formation of Ni-Bi is energetically favored ( E Ni-Bi > ½ E Ni-Ni + ½ E Bi-Bi ) and Ni-Bi bonds are likely much stronger than Bi-Bi bonds ( E Bi-Bi << E Ni-Bi ). Since there is no lattice matching relationship between the two grains for a general GB, a mismatch in lattice (incoherent interface) must occur somewhere. b) In the bilayer structure, two Bi adsorption layers are bonded to the Ni grain surfaces coherently to form strong Ni-Bi bonds, and a mismatch can occur between the two Bi layers to break (or lengthen) the weak Bi-Bi bonds. In contrast, formation of a) a monolayer adsorption of Bi or c) a NiBi compound-like layer will both require to break a substantial fraction of strong Ni-Bi bonds (to form at least one incoherent Ni-Bi interface), thereby being energetically unfavorable. While ordered structures are plotted in this schematic, we acknowledge the inevitable mixing and disordering in all interfacial phases due to entropic effects and the need to repair broken bonds (relaxation); the levels of mixing and disordering would likely increase with increasing interfacial width. It is interesting to further note that the proposed mechanism of stabilizing this bilayer interfacial phase (Fig. S11(b)) is somewhat analogous to those of stabilizing bilayers in biological and chemical/solution systems (such as the stabilization of lipids or vesicles).

20 (X) Observation of a trilayer We also made a single observation of a trilayer interfacial phase, which covered only a section of one GB (near a liquid tip) and was adjacent to a bilayer (Fig. S12). Throughfocus imaging indicates that this trilayer structure is not an artifact due to the interaction of an atomic step and a bilayer adsorption. Only a single observation of a trilayer was made, indicating that this trilayer likely represents a metastable interfacial phase (corresponding to a local minimum in free energy). The formation of this particular trilayer, which is adjacent to a penetrating liquid tip, could be related to a normal tensile stress generated by the adjacent penetrating liquid, which was recently demonstrated by Klinger and Rabkin (24). The co-existence of a bilayer and trilayer at one GB indicates the possible existence of a phase transition from trilayer to bilayer. More conclusive assessments must await further observations. Fig. S12. STEM HAADF micrograph showing the coexistence of bilayer and trilayer interfacial phases at a single GB, indicating a possible GB phase transition between them. A through-focus imaging series indicated that this trilayer image did not result from an atomic level step. (XI) Measurement of projected distances between neighboring Bi atom columns in the opposite layers Fig. S13 shows 250 individual measurements of the projected distances between two neighboring adsorbed Bi atom columns across two adsorbed layers. The average (projected) distances between two neighboring Bi columns from the opposite layers is

21 0.39 nm, being greater than the Bi-Bi metallic bond length of nm and close to twice of the Bi van der Waals radius (0.2 2 nm); the actual distance between two Bi atoms in three dimension should be equal to or even greater than the value of 0.39 nm. Fig. S13. Measured (projected) distances between two neighboring adsorbed Bi atom columns across two adsorbed layers.