Bottom-up Cu filling of annular through silicon vias: Microstructure and texture

Abstract

Microstructural and morphological evolution during bottom-up Cu filling of annular through silicon vias (TSV) in a CuSO₄−H₂SO₄−Cl⁻–poloxamine electrolyte is examined. Deposition proceeds in two distinct stages beginning with a passive-to-active state transition on the via sidewalls whose depth and ultimate thickness depends on the polymer flux. Growth is conformal or tapered with columnar grains whose width and texture differ between the outer and inner sidewalls of the annulus due to area reduction and expansion respectively. The outer sidewall and bottom surface have a preferred texture; 〈111〉//ND (via normal) and 〈110〉//CD (via circumferential). With time the sidewalls passivate while further deposition is localized to the via bottom. Bottom-up growth then fills the TSV with the formation and selective expansion of the 〈110〉//ND textured grains. At higher suppressor concentrations the initial onset and transient period of sidewall deposition is displaced to greater depth and the subsequent shift to bottom-up filling with large 〈110〉//ND texture grains occurs earlier. The dominant 〈110〉//ND texture during bottom-up filling is congruent with Cl⁻ stabilized texture development. The absence of suppression indicates that the polymeric suppressor does not adsorb on the active upward propagating surface.

1. Introduction

Electrodeposition of Cu is widely used in the fabrication of microelectronic interconnects. Electrolyte additives play a central role in guiding morphological evolution during deposition on non-planar metallized surfaces. Depending upon the length scale of the recessed features, e.g. trenches, vias, through holes, etc., different electrolyte additive packages and processing conditions are used for optimal filling. Various correlations and models detail the interactions between the additives and morphological evolution during feature filling. These range from the fabrication of submicrometer on-chip interconnects by the Curvature Enhanced Adsorbate Coverage mechanism [1] to the filling of larger scale through silicon vias (TSV) [2–6] and through holes [7] by the S-shaped Negative Differential Resistance (S-NDR) mechanism. Beyond feature filling, the additives also impact the as-deposited and post-deposition microstructure. Despite many reports that discuss Cu metallurgy relevant to on-chip Damascene metallization, TSV and other electroplating applications, much remains to be understood as to the relationship between additives and microstructural evolution [8–13].

X-ray diffraction investigations of the relationship between deposition rate and deviations from epitaxy go back to the 1950’s [14]. The importance of a low stacking fault energy and availability of relevant surface sites to the prevalence of growth twins has been widely discussed [14–20]. Likewise the connection to texture development during growth on single crystalline, polycrystalline or amorphous substrates has also been reported [14–21]. Controlling the Σ3 twin density has proven to be a particularly fruitful strategy in terms of enhanced mechanical properties without compromising the electrical conductivity of Cu [22–24]. Studies of the correlations between additive chemistry and microstructure have been reported but efforts utilizing the more powerful tools available for modern metallurgical analysis are more limited [25]. Amongst others, the application of analytical methods such as laser ablation mass spectrometry [26,27] and electron backscatter diffraction (EBSD) [11,12,28] promise to provide new insights into the interplay between additives and microstructural evolution. At the same time interpretation of earlier works is often hampered by limited specification or knowledge of the nature and/or purity of the electrolytes used. This is especially true with respect to halides that even at dilute concentrations are known to induce substantial anisotropy into the surface structure, i.e. steps and terraces, and influence the growth dynamics of Cu surfaces [29–31]. Indeed several recent galvanostatic Cu deposition studies indicate that a 〈110〉 growth texture is stabilized by sub-micromolar, and greater, additions of Cl⁻ or Br⁻ to CuSO₄−H₂SO₄ electrolytes [32–35]. Similarly, the role of additive incorporation and its effect on deposit structure and subsequent microstructural evolution is a subject of continuing interest. In the case of two-component halide-suppressor additive chemistries, significant negative impacts are evident in the forming of robust Cu–Sn solder joints that might be addressed with proper consideration of the processing conditions [36–38]. The present report details the microstructure of Cu TSV fabricated by extreme bottom-up filling in a prototypical two-component suppressor chemistry, with the characterization provided by EBSD.

The generic binary combination of Cl⁻ and polyethers has been shown to yield robust bottom-up filling in CuSO₄−H₂SO₄ electrolytes [2–7,39–42]. Suppression arises from co-adsorption of Cl⁻ and the polyether to form a bilayer where the polymer blocks access of Cu²⁺ to the surface [2,43–46]. The Cl⁻ layer serves multiple roles. Firstly, it displaces the sulfate anion from the surface and acts as a surfactant that guides step faceting during metal deposition [29,30,43–45]. Secondly, it alters the adjacent water structure to make the interface more hydrophobic, thereby facilitating the coadsorption of the polymer layer that blocks access of Cu²⁺ to the Cl⁻ covered Cu surface [45]. Formation of the co-adsorbed suppressor phase can be constrained by the available flux of Cl⁻ or polyether. Its disruption can be sustained by some combination of Cl⁻ incorporation within the growing deposit or preclusion of polymer adsorption on advancing surfaces that is possibly associated with release of the water of hydration following Cu²⁺ reduction [2–7,40,43–47].

Electroanalytical and feature filling studies have identified the additive concentration and potential regimes where bottom-up filling can occur [2–7,39–42]. When either component of the suppressor is dilute in the electrolyte TSV filling begins with a sharp transition from passive to active deposition at some depth within the via that is dependent on the concentration and applied potential [39–42]. Active deposition on the lower sidewall and bottom surface can be either conformal or tapered [39–42]. For certain conditions this stage of bifurcation is temporary as the active sidewalls subsequently passivate followed by the shift to bottom-up filling [39,40]. As the concentration of the additive increases the extent of transient sidewall deposition decreases with a more rapid shift to bottom-up filling [39,40]. In larger features and under certain conditions filling can require potential waveforms to progressively advance the location of the passive-active transition up the sidewall [40,42]. In this report the microstructure of filled annular TSV, shown in Fig. 1, and its relationship to the morphological transitions evident during feature filling under fixed applied potential are investigated as a function of the poloxamine concentration.

Fig. 1.

Schematic diagrams of the annular TSVs pattern along with a cross-sectioned TSV imaged by SEM after filling.

2. Experimental

The cross-sectioned TSV samples used in this study were previously examined by scanning electron microscopy (SEM) to follow the morphological evolution during feature filling [39]. Electrodeposition was performed at room temperature in 1 mol/L CuSO₄ - 0.5 mol/L H₂SO₄ - 1 mmol/L NaCl with suppression provided by co-adsorption of the halide with a poloxamine additive, Tetronic 701 (Aldrich¹). The Tetronic polyether (TET) contains an ethylenediamine core with four propoxylate-blocks capped with ethoxylate giving a molecular mass of 3600 g/mol. Deposition was performed at a fixed applied potential referenced to a Hg/Hg₂SO₄/saturated K₂SO₄ reference electrode (SSE). Working electrodes were rectangular wafer fragments patterned with ≈56 μm deep TSVs of annular cross-section (provided by IBM) with a 1 μm thick Cu seed on the field and a lesser amount on the sidewalls of the vias. Definition of the metal ion and additive transport during deposition was provided by rotating the wafer fragments about one end of a Pt spindle, like a helicopter blade, with the patterned surface facing up. Based on the ≈1 cm distance between the features and rotational axis for most of the imaged TSVs, the 100 rpm (200π rad/min) rotation rate corresponds to an estimated 10 cm/s flow rate over the surface. The wafer fragments were pre-wet with ethyl alcohol to displace air bubbles that were otherwise trapped in the TSVs during Cu deposition. Following immersion, the specimens were rotated for 2 min to mix the electrolyte with the alcohol in the TSVs; during this period the potential was maintained at −0.40 V_SSE (≈−30 mV relative to the measured open circuit potential of ≈ −0.37 V_SSE) to avoid corrosion of the Cu seed layer. With insufficient dwell times the delayed arrival of the electrolyte to the most recessed TSV region impacts the initial stages of feature filling. This effect has been seen to persist for at least 4 min [39]. In addition, significant Cu⁺ can be produced adjacent to the electrode during the dwell period that may also influence subsequent metal deposition. Following this mixing period, the potential was stepped to a specific value for a defined deposition time followed by removal, rinsing and drying. The applied potential was not corrected for uncompensated cell resistance.

The microstructure of the cross-sectioned TSV were examined by EBSD. The EBSD system (EDAX Pegasus with Hikari XP EBSD Camera), was mounted on a Schottky type field emission gun scanning electron microscope (FEI Inspect F50). The orientation imaging maps (OIM) of the cross-sectioned TSV were examined at locations over a two-dimensional array with 70 nm pitch. Each map took approximately 30 min to acquire, and no correction has been made for beam drift during data acquisition over the nominally rectangular region of each scan. Inverse pole figures providing quantitative analysis of crystallographic texture were obtained from the mapping results using EDAX-TSL’s OIM version 7.3.1 software.

3. Results

3.1. TSV filling: dilute polyether concentration and transient conformal deposition

The initial stage of Cu deposition after 4 min at −0.64 V_SSE in the presence of 6 μmol/L poloxamine is dominated by passive-active sidewall bifurcation that leads to conformal deposition in the bottom half of the representative annular via shown in Fig. 2a. EBSD analysis of this specimen was performed on both radial cuts of three annular vias. For the purposes of this communication, discussion and analysis is focused on the right side of the annular via highlighted by the yellow arrow. The crystalline texture is characterized by the crystal planes whose normal lie along the normal direction (ND) to the wafer surface, i.e., along the axis of the via, the circumferential direction (CD) that is the normal to the plane of the cross-section, and along the radial direction (RD) that is normal to the sidewall. These coordinate axes are inset in the lower left corner of Fig. 2a. The ND, CD and RD maps are shown in Fig. 2b along with a colorized stereographic triangle key for the grain orientations. The microstructure can be largely divided into three regions: the via inner sidewall, the via outer sidewall, and the via bottom. The inverse pole figures (IPF) for the ND, CD and RD maps, each including data from both sidewalls as well as the bottom surface segments, are shown in Fig. 2c. The IPFs convolve three different growth surfaces with three different elements of curvature, an issue that will be detailed below. Nevertheless, significant 〈111〉 texture in the ND direction is manifest in the 2.67 multiple of the population of grains in the 〈111〉 orientation relative to that in a random powder sample. In the CD direction, a measurable bias toward 〈111〉 and 〈110〉 textures is indicated while, for RD, a maximum near C321D is evident, congruent with the 〈111〉 and 〈110〉 textures in the orthogonal directions. The grains in the conformal deposits are on the whole columnar, consistent with epitaxial growth on the Cu seed layer. A map colorized by average binned grain size is shown on the far right of Fig. 2b. The average width of the columnar grains appears to be slightly larger on the inner (left side) convex wall of the radial cut in the annular via. The size (width) of the columnar grains on the bottom surface appears to fall between the larger value on the inner sidewall and smaller value on the outer sidewall. These observations can be understood to naturally arise from initially uniform grain size of the seed layer on the two sidewalls and bottom and the area change accompanying deposition on the respective surface segments, i.e. expansion versus shrinkage, naturally giving rise to changes in the average grain width in accord with the local curvature. One would anticipate the area change effect to be most evident in the horizontal cross section of the annular via, capturing the curvature components directly in the regions corresponding conformal columnar growth on the sidewalls. The grain width is significant in that the grain boundary density can play an important role in the development of growth stresses in the respective deposits [49,50].

Fig. 2.

Annular TSVs cross-sectioned after Cu deposition in electrolyte containing 6 μmol/L TET at −0.64 V_SSE for 4 min. (a) SEM image, (b) EBSD derived texture maps and map of grain size are indicated by color according to the respective legends and (c) IPF of the texture distributions in the ND, CD and RD that are defined in Fig. 2a. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

A more accurate IPF assessment of the microstructure of the same TSV is obtained by restricting the analysis to sections with a common growth normal (i.e., inner sidewall, outer sidewall and bottom) as shown in Fig. 3 and Fig. 4. With the sidewalls in Fig. 3 the RD maps are aligned with the direction of growth, while the ND and CD map show the texture in the orthogonal in-plane directions. The significant difference between the opposing sidewalls is immediately evident. The outer wall has an average texture in the (growth) normal direction (i.e., RD in Fig. 3c) centered close to 〈112〉, with 〈110〉 texture in the CD and 〈111〉 texture in the ND in plane directions. In contrast, the IPF of the inner wall exhibits a broader distribution along the (growth) normal direction (RD) centered near 〈134〉 with a bimodal texture centered around both 〈111〉 and 〈014〉 in the CD direction and 〈012〉 and 〈122〉 in the ND in-plane directions. In contrast to the sidewalls, maps of the bottom surface in Fig. 4 indicate a strong 〈111〉 texture in the (growth) normal direction (ND) with a band of 〈110〉 to 〈112〉 texture in the CD direction and an analogous band centered around 〈134〉 in the other in-plane direction RD, both of which are congruent with the predominance of 〈111〉 texture in the (growth) normal direction. The dominance of columnar growth on the sidewall and bottom surfaces is such that the orientation difference reflects either variation in the underlying seed layer or, possibly, the effect of area change and associated variation in growth stress. Quantification of the average grain size in the columnar grains shown in Fig. 3d indicates the grains are 20% smaller on the outer sidewall as compared to the inner sidewall, which is attributed to the impact of area change during growth on grain dimensions and selection. For comparison, note that the visible ≈1 μm thick Cu deposit yields an approximate −11% reduction of circumference from the initial 9.5 μm outer radius and a +25% increase of circumference from the initial 4 μm inner radius. Importantly, the conformal nature of the deposit on the feature length scale despite the significant variation the crystalline texture between the three surfaces indicates that such anisotropy does not play a major role in the deposition rate or morphological evolution at this stage in the filling process. That said, close inspection does reveal variations in the flatness associated with individual grains congruent with the known effect of Cl⁻ on step faceting on different Cu surfaces [14–18,29–31].

Fig. 3.

EBSD derived texture maps of inner (left) and outer (right) sidewalls of annular TSV cross-sectioned after Cu deposition in the electrolyte containing 6 μmol/L TET at −0.64 V_SSE for 4 min. EBSD maps and IPFs of the texture distributions in the (a) ND, (b) CD and (c) RD. (d) Maps of grain size indicated by color according to the defined scale with the average grain size on the inner and outer sidewalls indicated beneath the respective map. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

Fig. 4.

EBSD derived texture maps of the central bottom region of an annular TSV cross-sectioned after Cu deposition in the electrolyte containing 6 μmol/L TET at −0.64 V_SSE for 4 min. The EBSD maps and IPF results correspond to (a) ND, (b) CD and (c) RD.

Variations in texture between individual vias are seen, with a summary of EBSD results in the ND direction for three annular vias presented in Figure S-1 and S-2. Accordingly, the highlight of the preceding analysis and discussion is the distinctive nature of columnar growth on the sidewalls and bottom surface of the annular vias.

3.2. TSV filling: dilute polyether concentration and bottom-up filling

Deposition at −0.64 V_SSE beyond 4 min exhibits a major change in morphological evolution, switching from conformal growth in the active lower regions to bottom-up filling. This results in the lower third of the via being filled by 8 min as shown in Fig. 5. Comparison to Fig. 2 makes clear that deposition on previously active regions of the sidewall halt with the shift to bottom-up filling. Only near the via bottom do the EBSD map reveal a small region of extended sidewall growth that collides with the advancing bottom surface, marking the initiation of bottom-up filling. At this junction new grains form that are elongated in the bottom-up growth direction with a tendency to cluster near the 〈113〉//ND orientation that subsequently gives way to the formation of larger 〈110〉//ND grains. The 〈110〉 grains expand to fill the available cross-section while propagating upwards in the via. In some instances, faulting to yield equally wide 〈113〉//ND grains is evident followed by faulting back to 〈110〉//ND just before the filling experiment was terminated. Grain orientation maps binned within 15° of the 〈111〉, 〈113〉 and 〈110〉 in the ND are shown in Fig. 6. The general trend in microstructural evolution from 〈111〉 through 〈113〉 to 〈110〉 texture during bottom-up filling is evident for the six different sectioned vias sections. The evolution towards 〈110〉 texture in the growth direction is not unlike that reported in the literature for Cu electroplating in the absence of the polyether suppressor additive [14,16–18,21,23]. A schematic summary of the microstructural evolution during feature filling at −0.64 V_SSE and 6 μmol/L poloxamine is shown in Fig. 7.

Fig. 5.

Cross-sectioned annular TSVs after Cu deposition in electrolyte containing 6 μmol/L TET at −0.64 V_SSE for 8 min. (a) SEM image, (b) EBSD derived texture maps and map of grain size indicated by color according to the defined scale and (c) IPF images for ND, CD and RD. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

Fig. 6.

Binned maps of the ND grain orientation within 15° of the 〈111〉, 〈113〉 and 〈110〉 for six different Cu filled vias deposited at −0.64 V_SSE in 6 μmol/L TET for 8 min.

Fig. 7.

Schematic diagrams of cross-sectioned annular TSVs after Cu deposition in the electrolyte containing 6 μmol/L TET at −0.64 VSSE for (a) 4 min and (b) 8 min. Growth direction and grain size is reflected in grain boundaries sketched within the regions of sidewall deposit, initial bottom surface deposit and bottom-up growth region.

3.3. TSV filling: increased polyether concentration favors bottom-up filling

Deposition at the same applied potential but with double the poloxamine concentration, 12 μmol/L, leads to a much earlier shift to bottom-up filling with minimal transient deposition on the sidewalls. The resulting filling and microstructural evolution is shown for progressively longer deposition times in SEM images and EBSD maps in Fig. 8a–c and the accompanying filling schematic in Fig. 8d. Dominance of the 〈110〉 texture in the growth direction (ND) is obvious, with a single grain occupying ≈ 80% of the filled region after 12 min; texture development during feature filling still exhibits some variation across the patterned array as shown for a selection of vias after 12 min of deposition in Figure S-3. The bottom-up growth dynamic and 〈110〉 growth texture are already established by 4 min. Higher magnification maps of the microstructure at 4 min shown in Fig. 9 reveal limited columnar sidewall growth that is consistent with an initial period of active deposition on the lower sidewall. Unlike the conformal growth observed in the more dilute poloxamine solution, the sidewall deposits are thinner and tapered, suggesting a more rapid, progressive quenching of the metal deposition reaction due to the higher available flux of poloxamine. A slim region of columnar growth is also evident along the sidewalls of the filled regions in the 8 min and 12 min specimens shown in Fig. 8. Its tapered profile reflects the variation of growth time associated with the progressively longer delay in arrival and absorption of the suppressing polymer farther down the feature. It is also clear from Fig. 9b that deposition on the bottom surface initially exhibits a 〈111〉 texture along the growth normal (ND), as with the lower concentration (Fig. 4), followed by competition between the formation and selection of grains that yields the dominant 〈110〉 motif of the bottom-up growth within the first few minutes and micrometers of deposition. The corresponding EBSD maps of texture in CD and RD are shown in Fig. 9. The incorporation of adjacent grains with orientations creating Σ3 boundaries, i.e., twinned orientations, during the bottom-up deposition is flagged by identical coloring of grains related by such orientations in the grain size map in Fig. 9. Aside from twinning, once the 〈110〉 texture is established, as seen in Fig. 8, it is largely sustained during subsequent deposition.

Fig. 8.

Annular TSVs cross-sectioned after Cu deposition in the electrolyte containing 12 μmol/L TET at −0.64 V_SSE for (a) 4 min, (b) 8 min, (c) 12 min and (d) schematic diagram of the microstructure evolution.

Fig. 9.

Annular TSVs cross-sectioned after Cu deposition in the electrolyte containing 12 μmol/L TET at −0.64 V_SSE for 4 min. (a) SEM image, (b) EBSD maps of texture in the indicated directions and colorized grain size distribution and (c) IPF images for ND, CD and RD. The map of grain size on the far right of (b) considers adjacent grains with Σ3 orientations (i.e., twinned orientations) as single entities with the maximum dimensions of the resulting grains indicated by color according to the defined scale. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

Increasing the poloxamine concentration to 25 μmol/L further limits transient sidewall deposition and yields an earlier shift to bottom-up filling as shown in Fig. 10. The transition from 〈111〉 to the dominant 〈110〉 texture is also more rapid, being evident within the first few grains from the bottom (i.e., first micrometers of deposit). In the original study of Cu deposition in these TSV partially bottom-up filled features were sometimes observed adjacent to the fully filled ones on the same substrate [39]. As noted therein, the fractional filling reflects the convolution of electrolyte resistance and the two-state operation of the critical system as opposed to poor wetting by the electrolyte. Fortunately, improved uniformity of filling can be achieved by appropriate process control, such as a more negative applied potential or, alternatively, galvanostatic conditions [5,7,40–42].

Fig. 10.

Cross-sectioned annular TSVs after Cu deposition in electrolyte containing 25 μmol/L TET at −0.64 V_SSE for 15 min. (a) SEM image, (b) EBSD derived texture maps and map of grain size indicated by color according to the defined scale and (c) IPF images for ND, CD and RD. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

4. Discussion

EBSD analysis of TSV filled with Cu using a simple two component polyether-halide suppressor in a sulfate-based electrolyte reveals important insight into the deposition process. The filling experiments were performed with poloxamine concentration between 6 μmol/L to 25 μmol/L. Filling experiments were performed at a fixed applied potential; however, due to iR losses in the electrolyte the actual deposition overpotential was a function of the current flowing through the system at any moment in time. Two regimes of behavior are observed dependent on the polyether concentration. With dilute concentrations the initial stage of deposition is dominated by a passive-to-active transition on the sidewalls with conformal growth of columnar grains in the lower portion of the via. Columnar grain growth reflects the epitaxial constraint associated with 〈111〉 surface normal textured vapor deposited Cu seed layer. Nevertheless, microstructural development on the opposing curved sidewalls is measurably different. Other things being equal, area change during deposition results in the columnar grains becoming wider on the inner sidewall and narrower on the outer sidewall. The evolving grain boundary density and grain orientation may impact the residual stress, but, in the absence of related measurements, further discussion lies beyond the present work [48,49]. The deposit on the outer sidewall has preferred orientations in the in-plane direction of 〈111〉//ND and 〈110〉//CD while the deposit on the inner sidewall has a bimodal in-plane texture. Growth on the sidewalls occurs only during the first few minutes of deposition before shutting down as subsequent deposition becomes restricted to the bottom of the feature. Coincident with the shift of the filling evolution is a decrease in the electroactive area; the associated decrease in total deposition current decreases resistive losses, thereby increasing the overpotential that drives deposition [2–4,7,39,40]. This accounts for acceleration of the growth velocity associated with bottom-up deposition between 4 min, Fig. 2 and 8 min, Fig. 5. Coincident with the accelerated upward motion is development of a dominant 〈110〉 texture. When bottom-up filling begins the sidewalls are already passivated and it is not unreasonable to expect the adsorbed suppressor bilayer (comprised of carbon, oxygen, nitrogen and chloride) to be trapped and buried at the interface between the columnar sidewall grains and advancing bottom up growth front. Indeed, Auger electron spectroscopy observation of C-rich material distributed along the TSV sidewalls was recently reported although the chemistry of the bath and filling conditions were not disclosed [27]. In contrast, the faster growth and large elongated grains that accompany bottom-up filling suggest weakened interactions between the polymer suppressor and the bottom surface during propagation of the growth front from the bottom of the via. Importantly, in polymer-free systems, Cl⁻ is known to stabilize the 〈110〉 growth texture during Cu electrodeposition over a range of current density that span, at least, 7 mA/cm² to 45 mA/cm² [32–35]. Accordingly, the inability of the polymer to suppress metal deposition on the upward advancing deposit is consistent with prior suggestions that the metal deposition itself alters the ability of the polymer to assemble on the surface [2,6,45,46]. This is likely related to metal deposition induced changes in the hydrophilicity of the interface [2,45,46]. In particular, the polymer is subject to localized equilibrium attachment and detachment dynamics; the low current on the passive surface can be viewed as a reflection of the lifetime of such openings where Cu²⁺(H₂O)₅ can access the surface. Thus, as the overpotential increases, the release of water of hydration that accompanies acceleration of the deposition rate is expected to hinder polymer reattachment to the surface, thereby providing positive feedback for the deposition process. The possible role of lateral surface forces in conveying the polymer to the passive sidewalls is also worthy of consideration [50].

Higher poloxamine concentration decreases the initial stage of transient sidewall deposition as the increased flux of the polymer leads to more rapid formation of the polyether-halide suppressor bilayer on the sidewalls throughout the height of the via. There is a correspondingly more rapid shift to bottom-up filling and development of the 〈110〉 growth habit. As in Figs. 8 and 10 and Fig. S-3, the resulting 〈110〉//ND elongated grains can span the full width and majority of the length of the via. The transition from the 〈111〉//ND texture of the bottom surface to 〈110〉//ND occurs rapidly, often with no more than two or three intermediary grains.

The observation of 〈110〉 as a dominant texture during deposition in uninhibited CuSO₄–H₂SO₄ electrolytes has a long history [14–21]. Electrodeposits initially grow under the influence of the substrate, with subsequent deviations from substrate orientation mediated largely by twinning. More recent work indicates that the transition to 〈110〉 texturing, as well as the rate of its expression, is reinforced by the presence of Cl⁻ [32–35]. In the case of polycrystalline substrates the transition from isotropic deposition to 〈110〉 texture has been noted for Cl⁻ concentration as low 10⁻⁶ mol/ L. It is not unreasonable to speculate that many older reports of 〈110〉 growth texture were based on electrolytes with (unspecified or unknown) Cl⁻ levels above this value. The impact of adsorbed halide adlayers on texture development and sustained growth of Cu (110) oriented surfaces is related to the Cl⁻ adlayer structure. Specifically, it influences step faceting, terrace reconstruction and associated surface phase transitions that, in combination, greatly reduce the presence and lifetime, i.e. probability, of Cu (111) microfacets that provide the necessary pathway for twinning to occur.

The uninhibited nature of bottom-up growth combined with sustained growth of 〈110〉 oriented grains is consistent with the absence of suppressor on the growth surface. Since suppressor co-adsorption has been related to the hydrophobic character of the interface induced by adsorbed Cl⁻, its absence can be attributed to a change in the hydrophilicity of the moving interface arising from the combination of disruption of the adsorbed Cl⁻ layer and the release of water of hydration at the site of Cu²⁺/Cu⁺/Cu reduction.

5. Conclusions

Microstructural and morphological evolution during bottom-up Cu filling of TSV in a CuSO₄–H₂SO₄–Cl⁻–poloxamine electrolyte was examined. In dilute polyether electrolyte the morphology and microstructure develop in two well defined stages. Deposition first exhibits a passive to active transition on the via sidewalls. At depths below this transition a conformal layer with columnar grain that arise from epitaxy with the Cu seed layer is formed. The grain width and texture of the inner and outer sidewalls of the annular via differ, with the former consistent with area change accompanying growth on the curved surfaces. The deposit on the outer wall has the same texture as that on the bottom surface with respect to the ND and CD directions. Subsequently, passivation propagates down the sidewall, ultimately leaving deposition localized to the bottom surface. The transition to bottom-up filling is accompanied by the development of a strong 〈110〉 texture and elongated grains in the growth (ND) direction.

Higher polyether concentrations shorten the transient period of sidewall deposition, leaving a thinner, tapered profile that thickens toward the bottom of the via. The more rapid transition to bottomup filling is associated with more rapid evolution of the texture from the epitaxially defined 〈111〉//ND to the Cl⁻ mediated 〈110〉//ND within just one to two grains for the highest poloxamine concentration examined. It has been known for many years that 〈110〉 is the dominant texture developed during deposition in uninhibited CuSO₄–H₂SO₄ electrolyte. More recent work indicates that this is related to the presence of Cl⁻ in the electrolyte. The dominance and durability of the 〈110〉 texture reflects the lowered probability of developing growth twins due to faceting induced by Cl⁻ adsorption on (110) surfaces that limits the probability and size of 〈111〉 microfacets required for twin formation. The uninhibited nature of the bottom-up growth and the extended grains in the growth direction are consistent with the absence of the polymer suppressor on the growth surface. Because suppressor co-adsorption has been related to the hydrophobic character induced by adsorbed Cl⁻, its absence on the rapidly moving bottom interface is attributed to a change in the hydrophilicity of the interface related to the combination of disruption of the adsorbed Cl⁻ layer and perturbations related to released water of hydration at the site of Cu²⁺/Cu⁺/Cu reduction.