Influence of underlayer roughness on the properties of Sc_0.4Al_0.6N thin films prepared via sputter deposition

First, a 25 nm thick Mo layer was deposited as the bottom electrode on a 6-inch Si substrate using a DC sputtering system. Subsequently, a 30 nm thick SiN layer was deposited using plasma-enhanced CVD (AMAT P5000). The deposition conditions were as follows: SiH₄:36 sccm, NH₃:26 sccm, and N₂:2250 sccm were used as the introduced gas with a gas pressure of 688 Pa, substrate temperature of 333 °C, and RF power of 303 W.

Next, the SiN surface was etched using dry etching equipment (AMAT P5000). Ar:50 sccm was used as the introduced gas at a pressure of 2 Pa and an RF power of 50 W. The substrates were not heated. The film thicknesses before and after the Ar etching were measured using an ellipsometer, giving an etching rate of 0.17 nm s⁻¹. The roughness before and after etching was evaluated using a scanning probe microscope (SPM; Shimadzu SPM-9700) and a cantilever (Olympus OMCL-AC200TS-C3) in dynamic mode. A 500 nm square area was scanned with a resolution of 256 × 256 points, and the arithmetic mean roughness (Ra) was determined from the two-dimensional topographic data.

A 60 nm thick poly-Si layer was formed on the Si substrate using thermal CVD. SiH₄, PH₃, and N₂ gases were introduced, the temperature was set to 580 °C, and the gas pressure was set to 30 Pa. The poly-Si was doped with phosphorus to increase its conductivity, allowing it to function as a bottom electrode. The surface was treated with Ar etching in the same manner as the SiN underlayer.

A 70 nm thick BPSG underlayer was deposited directly on the Si substrate using ozone-CVD equipment (AMAT P5000). Si(OC₂H₅)₄, B(OC₂H₅)₃, C₆H₁₅O₄P, O₃, and N₂ were used as source gases. The pressure and the temperature of the substrate were set to be 2.7 × 104 Pa and 480 °C, respectively. The BPSG surface was flattened by a reflow process without plasma treatment. This allowed us to determine whether the suppression of AGG was due to the surface roughness of the underlayer or the plasma treatment. The reflow process was performed in a quartz tube furnace at 900 °C for 20 min under a nitrogen atmosphere.

A ScAlN thin film was deposited on each underlayer using a parallel-plate reactive DC sputtering system (ULVAC SME-200E). The target was a 300 mm diameter Sc_0.4Al_0.6 alloy fabricated by sintering. The deposition conditions were as follows: the introduced gas was N₂:26 sccm and Ar:19 sccm at a pressure of 0.3 Pa, a substrate temperature of 400 °C, DC power of 8 kW, and RF power applied to the substrate susceptor of 20 W. RF power was applied to improve the film crystallinity by increasing the energy of the ions incident on the film. The deposition thickness of the films was 500 nm. The composition ratio of Sc/(Sc+Al) in the ScAlN film was measured to be 40 ± 1% by Rutherford backscattering spectrometry (RBS).

The crystallinity of the ScAlN thin films deposited on each underlayer was evaluated using an X-ray diffractometer (Philips X'Pert MRD). The surface morphology was observed using scanning electron microscopy (SEM; JEOL JSM-IT510LV) to evaluate the density of abnormal grains.

To evaluate the electrical and piezoelectric properties, a 300 nm thick Al pattern with a diameter of 3 mm was fabricated on the ScAlN thin film as the topside electrode. A 20 nm thick Ti adhesion layer was utilized at the interface between the ScAlN thin film and the Al layer. In addition, Al/Ti layers with thicknesses of 700/20 nm were deposited on the back and sides of the chip to serve as terminals for the bottom electrode. The piezoelectric coefficient d₃₃, loss tangent, and leakage current versus voltage were measured using a d₃₃ meter (Piezotest PM2000), an impedance analyzer (Agilent 4195 A), and a semiconductor parameter analyzer (HP 4156 B), respectively.

To estimate the defects in the ScAlN thin films, cathodoluminescence (CL) measurements were performed at room temperature using a SEM (JEOL JSM-7800F) equipped with CL apparatus (HORIBA MP-32S). The acceleration voltage and probe current were set as 10 kV and 11 nA, respectively.

3.1.1. Effect of Ar etching on the SiN layer and analysis of the crystalline quality, orientation, and morphology of the ScAlN thin film

Figure 1 shows topographic images of the SiN surface before and after Ar-plasma etching. The as-deposited SiN surface exhibited a mound aggregation morphology with individual mounds—10–20 nm in diameter, as shown in Fig. 1(a). SiN deposited by the CVD process is reported to have such a morphology. ²⁷⁾ As shown in Figs. 1(b)–1(d), the contours of the individual mounds became unclear as the etching time increased. The R_a values in this area also decreased, as shown in Fig. 2. The SiN surface was planarized via Ar etching. In general, such planarization is attributed to the dependence of sputtering yield on the angle of ion incidence. This results in faster etching of the inclined planes on an uneven surface, with the convex areas becoming sharper and the concave areas flattened. In other words, the surface may be planarized by the lateral movement of the sloped surface region. ^27,28)

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Figure 3 shows X-ray diffraction (XRD) results for ScAlN thin films on SiN layers treated with different Ar etching times. The crystallinity of the ScAlN thin film improved with increasing etching time. In the 2θ-ω scan of Fig. 3(a), the peaks around 36.5°, which correspond to the ScAlN (0002) diffraction, were observed in the specimens treated with Ar etching. Thus, the treated films exhibit a preferential c-axis orientation. The peak positions shifted slightly to lower angles as the etching time increased. This shift might indicate that the film stress shifted toward the compressive side. As shown in Fig. 3(b), the peak intensity increased, and the FWHM of the rocking curve of the (0002) diffraction decreased with increasing etching time. However, the fluctuations became saturated at 90 s. The FWHM reached 1.7° after saturation. Therefore, ScAlN thin films with higher crystallinity and c-axis orientation tend to grow on SiN surfaces planarized by Ar plasma etching.

Zoom In Zoom Out Reset image size

Because the (0002) plane in the wurtzite-type structures of AlN and ScAlN is the most stable, the growth of a c-axis oriented thin film on the surface of such an amorphous material is reasonable. Interestingly, slight differences in the surface roughness significantly changed the crystallinity and orientation. The R_a values of the SiN surfaces were 1 nm and 0.3 nm before and after the 90-s etching, respectively. The former already had a mirror-like surface; however, even slight roughness inhibited the crystallization of ScAlN.

As shown in Fig. 4(a), abnormal grains were observed in the entire surface of the ScAlN thin film on the SiN layer before etching. However, the grain density decreased with increasing etching time [Figs. 4(b) and 4(c)]. This result is consistent with the XRD results.

Zoom In Zoom Out Reset image size

3.1.2. Measurement of the piezoelectricity, loss tangent and leakage current

Like the crystalline quality, the piezoelectric coefficient d₃₃ increased with increasing etching time. As shown in Fig. 5(a), d₃₃ reached approximately 22 pC N⁻¹ for Ar etching times of 40 s or longer. This value is comparable to that reported in Ref. 2. Comparing Figs. 3 and 5(a) shows that the etching time was shorter than the time at which the crystallinity was saturated. This indicates that piezoelectricity does not necessarily require extremely high crystallinity. In addition, the presence of a small number of abnormal grains, as shown in Fig. 4(c), did not significantly degrade the piezoelectricity. The abnormal grain densities estimated from the SEM images were 16 and 9 μm⁻² for etching times of 30 and 90 s, respectively. Therefore, abnormal grain densities of approximately 10 μm⁻² is considered to be a guideline for sufficient piezoelectricity.

Zoom In Zoom Out Reset image size

Further, the loss tangent also improved, i.e. lowered to approximately 0.2% after 30 s [Fig. 5(b)]. This value is smaller than that reported in a previous study at the same Sc concentration. ¹⁴⁾ The causal factors increasing the loss tangent are probably within or at the grain boundaries of the abnormal grains, considering that Ar etching is accompanied by a decrease in the density of abnormal grains.

Figure 6 plots the current–voltage curves of the ScAlN thin films with and without Ar etching. The leakage current was significantly reduced by Ar etching but those after 23 s and 90 s long etchings were almost equivalent. This result is consistent with the fact that the piezoelectric constant and loss tangent become nearly constant when the Ar etching time exceeds 23 s, as shown in Fig. 5.

Zoom In Zoom Out Reset image size

Therefore, the Ar etching of the SiN underlayer demonstrably improved the crystallinity, piezoelectricity, loss tangent, and leakage current of ScAlN thin films. These important properties influence the performance of specific piezoelectric MEMS devices.

Inserting a non-piezoelectric SiN layer between the ScAlN thin film and Mo bottom electrode could reduce the effective piezoelectricity of the ScAlN film. However, this should not be a problem in practice for the following reasons: The drive voltage applied between the top and bottom electrodes is distributed between the ScAlN thin film and SiN layer. Consequently, the effective d₃₃ value is smaller than the intrinsic d₃₃ value of the ScAlN thin film. In this case, each distributed voltage is determined from the ratio of the capacitances of the two layers. Accordingly, a smaller thickness and higher dielectric constant of the nonpiezoelectric underlayer are preferable for applying a large drive voltage to ScAlN thin films. As shown in this study, a SiN layer that is much thinner than the ScAlN thin film can be deposited on metal electrodes owing to the low-temperature process of plasma-enhanced CVD. In addition, SiN has relatively high permittivity. Therefore, the decrease in d₃₃ is not significant. SiO may also be another candidate for the underlayers.

3.1.3. Analysis by cathode luminescence (CL)

Figure 7 shows the emission spectrum of the CL used to estimate the defects in the ScAlN thin film. First, the band-edge emission peak derived from ScAlN was not observed at 270 nm. Note that this wavelength corresponds to approximately 4.5 eV as the band gap. ²⁹⁾ That may be due to the low crystallinity of the ScAlN thin film. In contrast, a broad peak consisting of four to five narrow peaks appears at wavelengths ranging from 200 to 800 nm. A similar broad peak is observed in other materials such as gallium nitride (GaN) crystals or thin films. In the case of GaN, such a broad peak is reported to be caused by defects in the material. ^30,31) GaN is a III–V compound semiconductor material with a wurtzite structure similar to that of ScAlN. Thus, the broad peak in the spectrum of ScAlN is also attributed to defects. Multiple thin peaks appeared owing to multiple reflection interferences at the top and bottom surfaces of the ScAlN thin film. ³⁰⁾

Zoom In Zoom Out Reset image size

The most important aspect of the spectra was the decrease in the intensity of the broad peak after Ar etching. This indicates that the number of defects in the film decreased. It is likely that defect reduction led to a reduction in the leakage current. In addition, increasing the Ar-etching time to 90 s did not lead to a further decrease in the intensity. Saturation behavior of the etching effect was observed, as was the case for the leakage current. Thus, the Ar-etched SiN underlayer contributes to the deposition of ScAlN with fewer defects.

Next, we investigated whether the above planarization effect occurs not only in SiN, but also in other underlying materials. The poly-Si surface was planarized by Ar etching in the same manner as the SiN described above. As shown in Fig. 8(a), the poly-Si layer had rounded grains of approximately 20 nm in diameter before Ar etching. After etching of 30 s, the contours of these grains disappear, leaving an uneven pattern with a smaller spatial frequency [Fig. 8(b)]. The R_a value decreased significantly from 1.14 to 0.37 nm as shown in Fig. 2.

Zoom In Zoom Out Reset image size

Figure 9 shows the XRD patterns of the ScAlN thin films deposited on poly-Si underlayers with and without Ar etching. The peak intensity of the (0002) plane in the ScAlN thin film deposited on the Ar-etched underlayer was approximately 60 times higher than that of the film deposited on the non-etched underlayer. The FWHM of the rocking curve of the (0002) plane also improved significantly from 5.5° to 1.7° after Ar etching. The former sample had many abnormal grains [Fig. 10(a)], whereas the latter had a smaller number of grains [Fig. 10(b)], as in the case of the SiN underlayer. The piezoelectric coefficient d₃₃ also improved from 16 to 25 pC N⁻¹ after Ar etching. The value of 25 pC N⁻¹ is higher than that for the SiN underlayer. This is perhaps due to the direct contact between the ScAlN thin film and the poly-Si bottom electrode layer without an insulation layer, unlike in the case of the SiN/Mo bottom electrode.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The abovementioned results demonstrate that the quality of ScAlN thin films is improved by Ar etching of both the SiN amorphous layer and poly-Si layer. In addition, to distinguish whether the improvement was due to surface roughness or another factor, a ScAlN thin film was deposited on a BPSG film with and without planarization by a reflow process based on thermal treatment.

Figure 11 shows topographic images of the surfaces of the BPSG thin films with and without the reflow processes. The surface of the as-prepared BPSG layer had a granular pattern with a diameter of less than several 10 nm. The R_a value was 0.42 nm and the surface was already mirror-like. After the reflow process, the granular pattern almost disappeared, and R_a decreased to 0.20 nm.

Zoom In Zoom Out Reset image size

ScAlN films were then deposited on the BPSG layers with and without the reflow process. As shown in Fig. 12, the flatter BPSG layer provided a much higher crystallinity to the deposited ScAlN thin film. The FWHM of the rocking curve of the (0002) plane also improved from 3.8° to 1.9° owing to the reflow. The number of abnormal grains was also significantly reduced, as shown in Fig. 13.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The R_a value of 0.42 nm without reflow corresponded to that of the SiN surface after Ar etching for more than 30 s (Fig. 2). Thus, if only a small R_a value contributes to the enhancement of the crystallinity of the deposited ScAlN thin film, even the non-reflowed BPSG layer should provide much higher crystallinity for ScAlN. However, this was not the case. Thus, not only R_a, but also other factors may contribute to the improvement in crystallinity.

The SPM images of the BPSG without reflow [Fig. 10(a)] and SiN with an etching time of 30 s [Fig. 1(c)] have similar R_a values. However, the former exhibits a smaller grain size and a clearer grain outline. This indicates that each microscopic protrusion on the surface has a steeper slope. The steepness of the slope was not reflected in the R_a value. Such steep slopes are likely the starting points for abnormal grains. Therefore, an improvement in the crystalline quality of the ScAlN thin film was confirmed in the case of the reflowed-BPSG-underlayer with a smooth surface.

In summary, the abovementioned results clearly indicate that the roughness of the underlayer is a critical factor in determining the abnormal grain density and performance of the ScAlN thin film. Excellent flatness with a sub-nanometer-level R_a is required to achieve sufficiently high crystallinity and performance. The planarization method does not need to be Ar etching. In addition, it is unlikely that the reflow process created defects on the BPSG surface or removed surface contamination. This also supports the idea that the primary effect of Ar etching the underlayer is planarization of the surface.