2024 Volume 92 Issue 12 Pages 127004
Electroless Ni–P plating films, used as the seed layers for the backside electrodes of gallium arsenide (GaAs) semiconductor devices, cause substrate warping (wafer warpage) during annealing, leading to substrate cracking and chipping. In this study, Ni–P films with different P concentrations (6, 17, and 20 at%) were deposited on a GaAs substrate, and the wafer warpage was analyzed after annealing at 240 °C for 1 h. The wafer warpage showed a reduction in size with an increase in P concentration in the as-deposited Ni–P film. Cross-sectional observations revealed the formation of a reaction layer at the interface between Ni–P and GaAs after annealing, and its thickness decreased with increasing P concentration. X-ray photoelectron spectroscopy revealed that the reaction layer was a Ni3GaAs alloy. The Ni atoms in the Ni–P films diffused to the GaAs substrate, and the P concentration in the Ni–P layers increased to a constant value (31–33 at%). X-ray diffraction measurements indicated that the Ni–P layers crystallized to Ni12P5 upon annealing because of the increase in P concentration owing to Ni diffusion. To evaluate the contribution of the reaction layer to the wafer warpage, the Ni–P layer was removed by ion milling, and the wafer warpage was measured. The estimated stress of the reaction layer was 580 MPa. From the calculation of the contribution of each layer to the wafer warpage, it was found that the Ni3GaAs layer largely contributed to the wafer warpage.
Gallium arsenide (GaAs) semiconductor devices possess high electron mobility and are widely used in high-frequency operating devices, such as mobile phones and radar devices in vehicles.1,2 An important issue in the wafer process of GaAs devices is wafer warpage because of backside film stress. Wafer warpage must be suppressed because it causes the failures of wafer transfer and stage chuck in the equipment and the deterioration of device characteristics.3–6 A schematic of the cross-sectional structure of a GaAs device is shown in Fig. 1. The device contains transistors and a ground electrode on the front and back surfaces, respectively. The surfaces are electrically connected using through-substrate electrodes (via). The ground electrode is formed by gold (Au) electroplating, which requires a seed layer. The seed layer contains metals such as titanium (Ti) or nickel (Ni), which have high adhesion to GaAs and high diffusion barrier properties for Au.7 An electroless Ni–P plating film is used as a seed layer for the backside electrode because of its high adhesion to the GaAs substrate attributed to Ni content and adequate coverage achieved through a simple solution process. Electroless Ni–P plating can be triggered by depositing catalytic metals on the GaAs substrate using galvanic substitution reactions.8–10 Especially, it is effective to use a PdCl2 solution to activate the GaAs substrate.8 The stress in the Ni–P film causes warping of the wafer (wafer warpage), leading to substrate cracking and chipping. Hence, the stress in the electrode film must be reduced. The increase in the stress is linked to the formation of a reaction layer at the interface between Ni–P and GaAs.11,12 The reaction layer is formed by the unidirectional diffusion of Ni into the GaAs substrate, and the Ni diffusion stops when the P concentration in the Ni–P film increases and crystallizes.13,14 However, the contribution of the reaction layer to the wafer warpage has not been clearly understood. In this study, we analyzed the stress induced by the reaction layer using Ni–P films with different P concentrations formed on GaAs substrates. The wafer warpage, interfacial reaction, and crystallinity of the Ni–P films were evaluated before and after annealing at 240 °C for 1 h by thin-film stress measurement, scanning transmission electron microscopy (STEM), X-ray photoelectron spectroscopy (XPS), and X-ray diffractometry (XRD). The annealing condition is the same as those performed at the end of the wafer process of the devices to stabilize the electrical characteristics.
Schematic cross-sectional structure of a GaAs device.
A polished four-inch undoped (100) GaAs wafer with a thickness of 625 µm was used as a substrate. Electroless Ni–P plating films with different P concentrations (low, medium, and high) were formed on one side of the GaAs substrate. The target thickness of the film was 0.3 µm. All samples were evaluated before and after annealing at 240 °C for 1 h. Annealing was performed in an electric furnace (Chuo Riken Co., Ltd.; custom-made). Nitrogen gas with a purity of 99.999 % was flown into the furnace at 10 L/min during the annealing. Low P (Samples 1 and 4), medium P (Samples 2 and 5), and high P (Samples 3 and 6) concentration samples were prepared for wafer warpage evaluation and film analysis (STEM, XPS, and XRD) before and after annealing. Note that the same sample was not used for both wafer warpage evaluation and film analysis because the former requires the entire wafer while the sample was cut out for the latter.
Samples 1, 2, and 3 were used to evaluate the wafer warpage using a thin-film stress measurement system (Toho Technology, FLX-2320-S). The wafer warpage was calculated from the reflection angle of the laser light irradiating the wafer surface. As shown in Fig. 2, the irradiation position was scanned in a direction parallel to the orientation flat through the center of the wafer. The wafer warpage is defined as the height difference (Δ) in the scanning range (R = 9 cm) as shown in Fig. 3. t is the GaAs substrate thickness, d is the Ni–P plating thickness, and r is the radius of curvature. Note that the wafer warpage is exaggerated in this figure, and in practice, the wafer warpage is concentric and Δ ≪ R. Applying the approximation to r2 = (r − Δ)2 + (R/2)2, the radius can be expressed as follows:
| \begin{equation} r = R^{2}/8\varDelta. \end{equation} | (1) |
Samples 4, 5, and 6 were analyzed by performing STEM, XPS, and XRD. Cross-sectional samples were prepared using a focused ion beam (FIB) mounted on a STEM (Thermo Fisher Scientific, Strata 400S). XPS (ULVAC-PHI, Quantum 2000) was conducted with AlKα radiation with a spot diameter of 100 µm. A 2-mm-square area was sputtered using a 4 keV Ar ion beam for depth profiling. XRD (Bruker, D8 DISCOVER) was performed using CuKα radiation.
Irradiation position on a four-inch wafer.
Schematic cross-sectional structure of wafer warpage caused by thin-film tensile stress. The wafer thickness is t, the thin film thickness is d, and the radius of curvature is r. The wafer warpage is defined as the height difference (Δ) in the scanning range (R).
Table 1 shows reagent, temperature, and treatment time of each process for Ni–P plating of low, medium, and high P concentration. The chemical products used for Ni–P plating were obtained from C. Uyemura & Co., Ltd., Japan. The process flow of electroless Ni–P plating with low P concentration is shown below. For rinsing, the GaAs substrate was immersed in 1.14 M (M: mol/L) hydrochloric acid (HCl)-containing solution (MCL-10) at 50 °C for 5 min, followed by Oxygen (O) plasma cleaning to remove organic contaminants. The substrate was immersed in a 0.34 mM PdCl2-containing solution (MCK-70) at 30 °C for 3 min. For Ni–P plating, the substrate was immersed in a sodium hypophosphite (NaH2PO2) solution containing 4.5 g/L of Ni ions (NLL-1) at 80 °C for 2.5 min.
| Process flow | Low P | Medium P | High P |
|---|---|---|---|
| 1. Rinse | HCl (MCL-10) 50 ± 1 °C 5 min |
HCl 23 ± 1 °C 5 min |
HCl (MCL-10) 50 ± 1 °C 5 min |
| 2. Activation | PdCl2 (MCK-70) 30 ± 1 °C 3 min |
PdCl2 23 ± 1 °C 4 min |
PdCl2 (MCK-70) 30 ± 1 °C 3 min |
| 3. Ni–P plating | NiSO4, NaH2PO2 (NLL-1) 80 ± 1 °C 2.5 min |
NiSO4, NaH2PO2 (KSB-18) 81 ± 1 °C 2.5 min |
NiSO4, NaH2PO2 (KRB) 82 ± 1 °C 2 min |
The process flow of electroless Ni–P plating with medium P concentration is shown below. For rinsing, the GaAs substrate was immersed in 0.57 M HCl solution at 23 °C for 5 min, followed by O plasma cleaning. The substrate was immersed in a solution containing 0.56 mM PdCl2 at 23 °C for 4 min as an activation treatment. For Ni–P plating, the substrate was immersed in a NaH2PO2 solution containing 6.5 g/L of Ni ions (KSB-18) at 81 °C for 2.5 min. The process flow of electroless Ni–P plating with high P concentration is shown below. For rinsing, the GaAs substrate was immersed in MCL-10 at 50 °C for 5 min followed by O plasma cleaning. The substrate was immersed in MCK-70 at 30 °C for 3 min. For Ni–P plating, the substrate was immersed in a NaH2PO2 solution containing 5.5 g/L of Ni ions (KRB) at 82 °C for 2 min.
Figure 4 shows the wafer warpage of Samples 1, 2, and 3 as-deposited and after annealing. The wafer warpage was obtained by subtracting the initial wafer warpage of a few micrometers before Ni–P film formation. The warpages of GaAs substrates with Ni–P films were 6–15 µm in the tensile direction, which increased to 18–42 µm after annealing. The wafer warpage after annealing decreased with increasing P concentration in the as-deposited Ni–P film.
Wafer warpages of low (Sample 1), medium (Sample 2), and high (Sample 3) P concentration as-deposited (circles) and after annealing at 240 °C for 1 h (squares).
Figures 5, 6, and 7 show the XPS depth profiles of as-deposited and annealed Samples 4, 5, and 6 with the deposition of Ni–P films of low, medium, and high P concentration. As-deposited samples were all composed of a Ni–P layer, a Pd–O layer, and a GaAs substrate. Pd was deposited on the GaAs substrate during the activation process. The composition of the Ni–P layer was uniform along its thickness. All annealed samples were composed of a Ni–P layer, a Pd–O layer, a Ni3GaAs layer, and a GaAs substrate. The Ni3GaAs layer was formed between the Pd–O layer and the GaAs substrate. The reaction layer is formed by the unidirectional diffusion of Ni into the GaAs substrate through the Pd-O layer.13 The compositions of the Ni–P and Ni3GaAs layers were uniform along their thickness. The P concentrations of each sample of as-deposited and after annealing are plotted in Fig. 8. The P concentrations of as-deposited Samples 4, 5, and 6 were 6, 17, and 20 at%, respectively. The P concentrations of the Ni–P layer of annealed Samples 4, 5, and 6 were 31–33 at%.
XPS depth profiles of sample of low P concentration (Sample 4) (a) as-deposited and (b) after annealing at 240 °C for 1 h.
XPS depth profiles of the sample of medium P concentration (Sample 5) (a) as-deposited and (b) after annealing at 240 °C for 1 h.
XPS depth profiles of sample of high P concentration (Sample 6) (a) as-deposited and (b) after annealing at 240 °C for 1 h.
P concentration in the Ni–P layers formed on GaAs substrates as-deposited (circles) and after annealing (squares). Ni–P films with low (Sample 4), medium (Sample 5), and high (Sample 6) P concentration were deposited on the substrates.
Figures 9 and 10 show the cross-sectional STEM images of as-deposited and annealed Samples 4, 5, and 6 with the deposition of Ni–P films of low, medium, and high P concentration, respectively. In the cross-sections in Figs. 9a, 9b, and 9c, two layers with different contrasts were observed. These two layers were identified as the Ni–P layer and the Pd–O layer by comparison with the XPS analysis of Figs. 5a, 6a, and 7a. The as-deposited Ni–P thickness was 0.31, 0.30, and 0.25 in Samples 1, 2, and 3, respectively. The film thickness varied by 11–14 % across the entire wafer. In Sample 3, the film thickness was slightly smaller than the target thickness. In the cross-sections in Figs. 10a, 10b, and 10c, two layers with different contrasts were observed. These two layers were identified as the Ni–P layer and Ni3GaAs layer by comparison with the XPS analysis of Figs. 5b, 6b, and 7b. Annealed Sample 6 was coated with platinum (Pt) to protect the surface before cross-section processing, so a Pt layer was observed in Fig. 10c. There were black dots indicating voids between the two layers of annealed Samples 4 and 6, but not in annealed Sample 5 (Fig. 10). This is probably due to the differences in the activation process, although the mechanism is still unclear.
Cross-sectional STEM images of as-deposited Ni–P films of (a) low (Sample 4), (b) medium (Sample 5), and (c) high (Sample 6) P concentration on GaAs substrates.
Cross-sectional STEM images of Ni–P films of (a) low (Sample 4), (b) medium (Sample 5), and (c) high (Sample 6) P concentration on GaAs substrates after annealing at 240 °C for 1 h.
Table 2 lists the layer thicknesses measured from the cross-sectional STEM images and the percentage thickness of the as-deposited Ni–P films. The ratio of the Ni–P thickness and Ni3GaAs thickness after annealing to the as-deposited Ni–P thickness is also presented in Table 2, which makes it possible to compare the Ni–P thickness and Ni3GaAs thickness after annealing between samples, without depending on the as-deposited thickness. The ratio of the Ni–P thickness after annealing to the as-deposited Ni–P thickness was 22 %, 60 %, and 64 % for Samples 4, 5, and 6, respectively. The thickness of the Ni–P layer after annealing increased with increasing P concentration of the as-deposited Ni–P film. The ratio of the Ni3GaAs thickness after annealing to the as-deposited Ni–P thickness was 164 %, 80 %, and 72 % for Samples 4, 5, and 6, respectively. The thickness of the Ni3GaAs layer after annealing decreased with increasing P concentration of the as-deposited Ni–P film.
| Sample | As-deposited | After annealing | |
|---|---|---|---|
| Ni–P (µm) |
Ni–P (µm) |
Ni3GaAs (µm) |
|
| 4 (Low P) |
0.31 | 0.07 (22 %) |
0.51 (164 %) |
| 5 (Medium P) |
0.30 | 0.18 (60 %) |
0.24 (80 %) |
| 6 (High P) |
0.25 | 0.16 (64 %) |
0.18 (72 %) |
Percentage thickness of the as-deposited Ni–P film is shown in parenthesis.
Figure 11 shows the XRD patterns of Samples 4, 5, and 6 before and after annealing. Before annealing, broadened Ni peaks were observed at approximately 45° in all the samples. After annealing, the broadened peaks disappeared, and other peaks appeared. These peaks were attributed to the presence of Ni12P5 (JCPDS ICDD card 22-1190). The positions and the intensity of Ni12P5 peaks changed depending on the samples, which can be because of the difference in crystal orientation. The Ni3GaAs peaks appeared at approximately 32° and 67° (JCPDS ICDD card 47-1146).15
XRD patterns of low (Sample 4), medium (Sample 5), and high (Sample 6) P concentration (a) as-deposited and (b) after annealing at 240 °C for 1 h.
Some authors have reported that the composition of the reaction layer is Ni2GaAs,16,17 but it can be identified as Ni3GaAs by volumetric analysis.18 The volumetric ratios VNixGaAs/VNi (x is the number of Ni atoms in the reaction layer) have been calculated19 as follows:
| \begin{equation} V_{\text{Ni2GaAs}}/V_{\text{Ni}} = 3 \end{equation} | (2) |
| \begin{equation} V_{\text{Ni3GaAs}}/V_{\text{Ni}} = 2. \end{equation} | (3) |
Table 2 presents the change in the layers thickness of Samples 4, 5, and 6 before and after annealing. The difference of the Ni-P layers thickness before and after annealing was 0.24 µm, 0.12 µm, and 0.09 µm, respectively. Assuming that the contribution of P to the change in the thickness is negligible, the thickness of the reaction layer is about twice the reduced thickness of the Ni-P layer. Thus, we concluded that the reaction layer is Ni3GaAs.
4.2 Crystallization of Ni–P filmsThe results of the XPS depth profiles (Figs. 5, 6, and 7) and the cross-sectional STEM (Figs. 9 and 10) confirmed the formation of a Ni3GaAs reaction layer between the Ni–P film and the GaAs substrate. The reaction layer is generated by the unidirectional diffusion of Ni onto the GaAs substrate.13 The P concentration in the Ni–P film increases with the progress of Ni diffusion. As shown in Fig. 8, the P concentrations in the as-deposited Ni–P films were 6, 17, and 20 at%, which increased to 31–33 at% after annealing. The P concentrations in the Ni–P films were almost the same after annealing. According to the XRD results (Fig. 11), as-deposited low P film (P concentration; 6 at%) is Ni microcrystals, while medium and high P films (P concentration; 17, and 20 at%) are amorphous, which is consistent with the results reported in the literature.20–22 The Ni–P layer of each sample crystallized to Ni12P5 after annealing at 240 °C for 1 h. The crystallization of the Ni–P films was induced by the increase in the P concentration owing to the Ni diffusion into the GaAs substrate. After crystallization, the Ni atoms in the Ni–P film became energetically stable, and the Ni diffusion stopped. The P concentration of the Ni–P film after annealing is 31–33 at%, which is higher than the P concentration of Ni12P5. It is assumed that P is segregated at the grain boundaries of Ni12P5.
4.3 Influence of Ni3GaAs on the wafer warpageAs shown in Fig. 4, as the P concentration in the as-deposited Ni–P film increased, the wafer warpage became smaller and the Ni3GaAs layer thinner. Particularly, the Ni3GaAs layer influences the wafer warpage. The contribution ratios of the Ni–P film and Ni3GaAs layer to the wafer warpage were investigated. Ion milling was performed on annealed Samples 1, 2, and 3 to remove the Ni–P layer and measure the wafer warpage caused by the Ni3GaAs layer. Figure 12 shows the relationship between the wafer warpage and the thickness of the Ni3GaAs layer. The thickness of the Ni3GaAs layer was measured using the STEM images. The wafer warpage showed a tendency to increase with the Ni3GaAs layer thickness. The stress of a thin film can be estimated from the wafer warpage and film thickness using the following equation:23
| \begin{equation} \sigma = (E_{\text{GaAs}}\times t_{\text{GaAs}}{}^{2})/(6(1 - \varLambda_{\text{GaAs}})\times r\times t), \end{equation} | (4) |
where σ is the internal stress of the film (Pa), EGaAs is the Young’s modulus of the GaAs substrate (Pa), tGaAs is the thickness of the GaAs substrate (m), ΛGaAs is the Poisson’s ratio of the GaAs substrate, r is the radius of curvature (m), and t is the thickness of the film (m). Using the Eq. 1,
| \begin{equation} \sigma = (E_{\text{GaAs}}\times t_{\text{GaAs}}{}^{2}\times 8\times \varDelta)/(6(1 - \varLambda_{\text{GaAs}})\times t\times R_{\text{GaAs}}{}^{2}). \end{equation} | (5) |
The stress-estimation results are listed in Table 3. Initially, we estimated the stress of the as-deposited Ni–P films. Using EGaAs = 82.68 GPa, tGaAs = 6.25 × 10−4 m, ΛGaAs = 0.31, and RGaAs = 0.09 m (scanning range), the stress values of Samples 1, 2, and 3 were found to be 410, 160, and 390 MPa, respectively, in the tensile direction. Subsequently, we estimated the stress in the Ni3GaAs layer formed after annealing. We assumed that the Ni3GaAs layer has a specific stress value. The slope (Δ/t in Eq. 1) was determined by fitting a linear function passing through the origin in Fig. 12. The stress value was determined to be 580 MPa. Finally, we estimated the stress in the Ni–P layer after annealing. We assumed that the wafer warpage could be expressed as the sum of the warpage because of the Ni–P and Ni3GaAs layers, as follows:
| \begin{equation} \varDelta = \varDelta_{\text{Ni–P}} + \varDelta_{\text{Ni3GaAs}}. \end{equation} | (6) |
The wafer warpage of Sample 1, 2, and 3 due to the Ni3GaAs layer (ΔNi3GaAs) was 31.1 µm, 18.4 µm, and 15.7 µm respectively, using Eq. 5, and that due to the Ni–P layer (ΔNi–P) was 11.2 µm, 9.7 µm, and 2.3 µm respectively, using Eq. 6. ΔNi3GaAs is more than 1.9 times higher than that of ΔNi–P, and thus, the formation of the Ni3GaAs layer largely contributes to the wafer warpage after annealing.
Relationship between the wafer warpage and the thickness of the Ni3GaAs layer of Samples 1, 2, and 3 after removing the Ni–P layer.
| Sample | Condition | Thickness (µm) | Wafer warpage (µm) | Stress (MPa) | ||||
|---|---|---|---|---|---|---|---|---|
| Ni–P | Ni3GaAs | Total | Due to Ni–P |
Due to Ni3GaAs |
Ni–P | Ni3GaAs | ||
| 1 | As-deposited | 0.27*1 | — | — | 14.5 | — | 410 | — |
| 2 | As-deposited | 0.31*1 | — | — | 6.3 | — | 160 | — |
| 3 | As-deposited | 0.30*1 | — | — | 15.0 | — | 390 | — |
| 1 | Annealed | 0.06*2 | 0.42*2 | 42.4 | 11.2*5 | 31.1*4 | — | — |
| 2 | Annealed | 0.18*2 | 0.25*2 | 28.2 | 9.7*5 | 18.4*4 | — | — |
| 3 | Annealed | 0.20*2 | 0.21*2 | 18.0 | 2.3*5 | 15.7*4 | — | — |
| 1 | Annealed and etched | — | 0.41*3 | — | — | 31.2 | — | 580 |
| 2 | Annealed and etched | — | 0.17*3 | — | — | 13.8 | — | |
| 3 | Annealed and etched | — | 0.14*3 | — | — | 13.0 | — | |
*1. Calculated from the weight difference of the samples before and after film formation.
*2. Calculated from the thickness ratio as-deposited and after annealing of Samples 4, 5, and 6 (Table 2).
*3. Measured from the cross-sectional STEM images.
*4. Calculated from the stress values and the thicknesses of the Ni3GaAs layers after annealing.
*5. Value obtained by subtracting the warpage because of Ni3GaAs from the total warpage.
Ni–P plating films with different P concentrations (6, 17, and 20 at%) were formed on a GaAs substrate, and the wafer warpage was evaluated before and after annealing at 240 °C for 1 h. As the P concentration in the as-deposited Ni–P film increased, the wafer warpage after annealing became smaller. Cross-sectional STEM observations revealed that a reaction layer was formed at the GaAs and Ni–P interface after annealing, and its thickness decreased with increasing P concentration in the as-deposited Ni–P film. XPS depth and XRD analysis indicated that the reaction layer was a Ni3GaAs alloy. Because the Ni atoms diffused into the GaAs substrate, the concentration of P in the Ni–P layer increased after annealing. Regardless of the P concentration in the as-deposited Ni–P film, the P concentration in the Ni–P layer after annealing was almost constant (31–33 at%). According to the XRD analysis, the Ni–P layers crystallized to Ni12P5 after annealing, which was attributed to the increase in P concentration because of Ni diffusion. Presumably, Ni diffusion stopped owing to the crystallization of the Ni–P layer. To evaluate the contributions of the Ni3GaAs and Ni–P layers to the wafer warpage, we removed the Ni–P layer after annealing using ion milling. The stress value of the Ni3GaAs layer on the GaAs substrate was estimated to be 580 MPa. From the calculation of the contribution of each layer to the wafer warpage, it was found that the Ni3GaAs layer largely contributed to the wafer warpage. It is effective to use a Ni-P film with a high P concentration to suppress the wafer warpage, leading to the production of reliable GaAs devices.
We thank Dr. Yukinori Oda, Uyemura & Co., Ltd., Japan, for his assistance with sample preparation.
Koichiro Nishizawa: Conceptualization (Lead), Data curation (Lead), Formal analysis (Equal), Investigation (Lead), Methodology (Lead), Writing – original draft (Lead)
Ayumu Matsumoto: Methodology (Equal), Supervision (Lead), Writing – review & editing (Lead)
Yasuyuki Nakagawa: Investigation (Lead), Methodology (Equal)
Hitoshi Sakuma: Resources (Lead), Writing – review & editing (Equal)
Yoshiki Kojima: Resources (Lead), Writing – review & editing (Supporting)
Naoki Fukumuro: Supervision (Lead), Writing – review & editing (Equal)
Shinji Yae: Supervision (Lead), Writing – review & editing (Equal)
The authors declare that they have no conflict of interest.
K. Nishizawa, A. Matsumoto, Y. Nakagawa, H. Sakuma, Y. Kojima, and S. Yae: ECSJ Active Members
