betaFeedback

WorksPhysics and Astronomy

Unprecedented Enhancement of Piezoelectricity in Wurtzite Nitride Semiconductors via Thermal Annealing

Abstract

The incorporation of rare-earth elements in wurtzite nitride semiconductors, e.g., scandium alloyed aluminum nitride (ScAlN), promises dramatically enhanced piezoelectric responses, critical to a broad range of acoustic, electronic, photonic, and quantum devices and applications. Experimentally, however, the measured piezoelectric responses of nitride semiconductors are far below what theory has predicted. Here, we show that the use of a simple, scalable, post-growth thermal annealing process can dramatically boost the piezoelectric response of ScAlN thin films. We achieve a remarkable 3.5-fold increase in the piezoelectric modulus, d33 for 30% Sc content ScAlN, from 12.3 pC/N in the as-grown state to 45.5 pC/N, which is eight times larger than that of AlN. The enhancement in piezoelectricity has been unambiguously confirmed by three separate measurement techniques. Such a dramatic enhancement of d33 has been shown to impact the effective electromechanical coupling coefficient kt2 : increasing it from 13.8% to 76.2%, which matches the highest reported values in millimeter thick lithium niobate films but is achieved in a 100 nm ScAlN with a 10,000 fold reduction in thickness, thus promising extreme frequency scaling opportunities for bulk acoustic wave resonators for beyond 5G applications. By utilizing a range of material characterization techniques, we have elucidated the underlying mechanisms for the dramatically enhanced piezoelectric responses, including improved structural quality at the macroscopic scale, more homogeneous and ordered distribution of domain structures at the mesoscopic scale, and the reduction of lattice parameter ratio (c/a) for the wurtzite crystal structure at the atomic scale. Overall, the findings present a simple yet highly effective pathway that can be extended to other material families to further enhance their piezo responses.

Introduction

Piezoelectricity is the generation of internal electric field in certain materials under applied mechanical stress.[1] In crystalline materials with a lack of inversion symmetry, piezoelectricity is produced due to electrochemical interaction between the electrical and mechanical states of the constituent atoms. The piezoelectric effect is also reversible, i.e., a piezoelectric material can generate mechanical strain when subject to an applied electric field.[2] Since the first demonstration of direct piezoelectricity effect by P. Curie and J. Curie in 1880, they have been widely utilized in applications such as high-voltage and high-power sources, sensors, actuators, time and frequency control, surgery, and ultrasound medical devices.[1, 3] Thin film piezoelectric materials have also been explored and intensively studied to advance a broad range of emerging devices and systems, including next-generation microelectromechanical systems (MEMS)[3, 4, 5, 6] and acoustic wave resonators and filters for future communication technologies.[7] They have already reshaped communication technologies like 5G and beyond and empowered the evolution of sensing applications such as electroacoustic and immunosensors, microphones, touch-sensitive screens, and activity monitors in smart wearables, to name just a few.[5, 7, 8, 9, 10] Additionally, as the demand for energy harvesting technologies grows, piezoelectric energy harvesting systems have gained prominence.[11]

To date, the commercial piezoelectric application space mostly relies on a few well-researched piezoelectric materials including ferroelectric ceramics such as Pb(ZrxTi1−x)O3\mathrm{Pb}(\mathrm{Zr}_{x}\mathrm{Ti}_{1-x})\mathrm{O}_{3} (PZT), barium titanate (BaTiO3\mathrm{BaTiO}_{3}), and other perovskite titanates.[3, 12] While these materials exhibit good piezoelectric properties, as measured from their longitudinal piezoelectric strain constant d33,fd_{33,f} (often interchangeably used as d33d_{33} in prior literature), they are incompatible with existing CMOS process, and their low Curie temperatures (generally below 400 °C) makes them prone to formation of oxygen vacancies, resulting in a sharp deterioration in their ferroelectric and piezoelectric properties under harsh conditions.[13, 14] Additionally, PZT poses environmental hazards due to its high toxicity and volatility at high operational temperatures.[15] Recently, aluminum nitride (AlN) has emerged as a promising platform for a wide range of applications, including MEMS/nanoelectromechanical system devices as well as high frequency resonators and filters, due to its high Curie temperature (∼1150 ∘C\sim1150\,^\circ\mathrm{C}), minimal acoustic and dielectric losses, elevated acoustic wave velocity, non-toxicity and compatibility with back-end CMOS integration requirements.[16, 17, 18] Yet, pure AlN thin films exhibit a relatively low piezoelectric modulus (d33∼4d_{33} \sim4 pC/N)[19, 20], restricting its potential application in state-of-the-art piezoelectric technologies.[21]

However, the introduction of one or more alloying elements in AlN can significantly enhance the piezoelectric performance.[21, 22] An exhaustive list of 25+ elements have been identified that can be alloyed within AlN to enhance its piezoelectric properties.[21, 23] One of the most promising approaches is to alloy AlN with Scandium (Sc) to form wurtzite ScxAl1−xN\mathrm{Sc}_{x}\mathrm{Al}_{1-x}\mathrm{N}, which significantly amplifies the piezoelectric constants of AlN.[19] Recent theoretical predictions reveal a maximum possible enhancement of d33d_{33} to ∼100\sim100 pC/N at nearly 64% Sc composition.[24] There exists discrepancies in the impact of internal strain, finite size of the super-cell and associated configurational disorder on the results, thus widely varying values have been reported for the theoretical limit of d33d_{33} in ScAlN.[25] The significantly enhanced piezoelectric response along the cc-axis in wurtzite ScAlN crystals, defined by the piezoelectric modulus, d33d_{33}, is primarily due to the energy landscape flattening as a result of competition between the layered hexagonal (hh-ScAlN, attained at ~50% Sc content) and the parent wurtzite phase with increasing Sc content.[19] The addition of Sc to AlN causes distortion in the wurtzite crystal structure, resulting in a decrease in the c/ac/a ratio and an increase in the internal uu parameter. This shift towards a planar hexagonal structure destabilizes the crystal and enhances the piezoelectric response.

Experimentally, maintaining high-quality crystal structures becomes extremely difficult with increasing Sc content, thereby constraining the realization of exceptional piezoelectric properties predicted by theory.[30] Moreover, a high Sc content also leads to a phase transition from the wurtzite to the nonpolar rock-salt phase, which can completely eliminate piezoelectric response from the material.[19] The highest d33d_{33} value experimentally measured in ScAlN is in the range of 25-32 pC/N, which is achieved at a Sc composition exceeding 40%, shown in Figure 1.[19, 27, 29] However, such high Sc-alloying often leads to material imperfections, limiting its practical application. Consequently, most of the ScAlN based acoustic devices reported to date are largely based on relatively lower Sc concentrations (~20 – 30%), with d33d_{33} merely in the range of 10-15 pC/N, that is only 2-3 times larger than AlN.[26, 27, 28]

Benchmarking of piezoelectric coefficient as a function of Sc composition

Figure 1. Benchmarking of piezoelectric coefficient, d33d_{33} (d33,fd_{33,\mathrm{f}}) with recent reports in literature as a function of varying Sc compositions.[19, 25, 27, 28, 29, 31, 32, 33, 34, 35, 36, 37]

Herein, we demonstrate that a simple, scalable, post-growth thermal annealing process can dramatically enhance the piezoelectric response of ScAlN. We measure a 250% (3.5×) enhancement in d33d_{33}, from 12.3 pC/N for the as-grown Sc0.3_{0.3}Al0.7_{0.7}N sample, to 45.5 pC/N for the sample with optimized annealing, shown in Figure 1. This is more than 700% (8×) higher than that of AlN and ~1.5 times higher than the highest value of d33d_{33} reported in literature for ScAlN across all Sc compositions. This could significantly enhance acoustic device performance, as a higher d33d_{33} value leads to more efficient mechanical-to-electrical energy conversion. This efficiency is quantified by the effective electromechanical coupling coefficient kt2k_t^2, which has been demonstrated to increase substantially from 13.8% to 76.2%, approaching the highest values reported in millimeter-thick lithium niobate films but in a ScAlN film that is merely 0.01% of the thickness. This unprecedented enhancement in the piezoelectric response of ScAlN thin films unlocks its transformative potential for frequency scaling in bulk acoustic wave resonators used in centimeter and millimeter-wave microwave spectral processing and beyond 5G-wireless communication technologies.

Results and Discussion

Thermal annealing for d33d_{33} enhancement

From the structural perspective, the piezoelectric response of wurtzite AlN depends on a plethora of factors, including but not limited to the orientation of the cc-axis, distribution of the constituent atoms, orientation of the polar grains, MM (metal) – NN (Nitrogen) bond length, strain in the film and the presence of defects within the films.[38, 39] Herein, we explore the potential of controlled atmosphere annealing to enhance the orientation along the cc-axis and the overall crystallinity of the as-grown ScAlN thin films. The evolution of d33d_{33} for various annealing conditions for ScAlN are outlined in Figure 2. For the as-grown sample, the d33d_{33} was measured to be 12.3 pC/N from piezoelectric force microscopy (PFM) measurement, which is consistent with other reports in the literature.[32, 33] Shown in Figure 2(a), we observe an enhancement in d33d_{33} with increasing annealing temperature, peaking at 45.5 pC/N at 700°C for 2 hours in vacuum, followed by a sharp decline. This marks a remarkable 3.5 times increase compared to the as-grown film. Similarly, when keeping the temperature constant at 700°C, varying the annealing duration from 1 to 8 hours reveals a gradual increase in piezo response, peaking at 45.5 pC/N after 2 hours of annealing, shown in Figure 2(b). Beyond the optimal temperature, d33d_{33} shows a sharp decline, whereas it remains reasonably stable with prolonged annealing durations. Based on the above findings, we infer that the increase in d33d_{33} is more sensitive to annealing temperature than time. The narrow annealing window explains the nominal enhancement in d33d_{33} achieved till date for wurtzite III-Ns despite some previous attempts.[26, 40, 41] Furthermore, to analyze the effect of ambient conditions on the piezo response, we varied the ambient conditions and repeated the annealing process at the optimized condition (700 °C, 2h) and the results are plotted in Figure 2(c). The observed d33d_{33} is similar for vacuum (d33d_{33} = 45.5 pC/N) and Ar ambience (d33d_{33} = 47 pC/N), with a reduction observed for N₂ ambience (d33d_{33} = 39.8 pC/N) and a sharp reduction for O₂ ambience (d33d_{33} = 34.5 pC/N).

Four-panel plots of $d_{33}$ versus annealing temperature, annealing time, ambient conditions, and measurement technique

Figure 2. (a) Plot of d33d_{33} versus annealing temperature showing a maximum enhancement at T = 700 °C for PVD-grown ScAlN. (b) Plot of d33d_{33} vs annealing time showing a maximum enhancement at a duration of 2 h, showing an enhanced d33 (max)d_{33\ (\mathrm{max})} = 45.5 pC/N at the optimized annealing condition from 12.3 pC/N for the as-grown sample. (c) Plot of d33d_{33} at the optimized annealing condition in varying ambient conditions (vacuum, Ar, N2_2 and O2_2). (d) Plot of d33d_{33} values, before and after annealing, showing good agreement across different measurement techniques (PFM and LDV).

Previous studies have shown that the d33d_{33} of AlN films is highly sensitive to surface oxidation.[42]

Due to the large oxygen affinity of both Sc and Al elements, and further exacerbated by the high-temperature annealing, a rapid surface oxidation process occurs when ScAlN films are annealed in an O2_2 ambience.[43] This leads to a sharp drop in the d33d_{33} compared to other annealing environments. Furthermore, annealing in N2_2 and O2_2 ambiences at a high temperature may introduce additional defects in the crystal lattice and can induce phase separations. For example, annealing AlN in excess nitrogen leads to the formation of aluminum oxynitride (Al–O–N) phase[40], which may result in a comparatively lower enhancement compared to Ar (inert) and vacuum conditions. Moreover, we carried out Laser Doppler Vibrometry (LDV) measurements on both the as-grown and optimally annealed samples. As shown in Figure 2(d), the LDV results revealed a d33d_{33} value of 12.7 pm/V for the as-grown sample and a peak d33d_{33} value of 42.1 pm/V for the sample optimized at 700°C for 2 hours. These findings are in excellent agreement with those obtained from the PFM measurements, providing further evidence to support the dramatic enhancement achieved through annealing.

To explore correlation of electromechanical coupling with annealing, displacement and capacitance loops of the two samples were measured using a Radiant PiezoMEMS Analyzer (Figure 3). For displacement loop extraction, triangular pulses at 60V, 2kHz are applied to capacitors implemented in both the samples and displacement is measured by a synchronized laser vibrometer. After annealing, the largest displacement increased almost 3.5 times, from 0.5 nm to 1.7 nm, corresponding well with PFM and LDV results. The displacement loop of both samples have broad linear regions, which were used to calculate d33,fd_{33,f} by fitting a line, as highlighted in Figure 3(a).

Displacement and capacitance loops of as-grown and annealed AlScN samples

Figure 3. (a) Converse piezoelectric effect as measured from the displacement loop of as grown and annealed AlScN sample, from which d33,fd_{33,f} could be calculated from the broad linear regions. (b) C-V loops of the as-grown and annealed film showing nearly identical response.

After annealing, an effective converse piezoelectric coefficient d33,fd_{33,f} of 37.9 pm/V and -41.9 pm/V was obtained in the positive and negative branch of the loop, respectively. These results further confirm the significant enhancement of piezoelectric properties after annealing process. To enable estimation of electromechanical coupling, capacitor loops (i.e., C-V) are also measured using triangular pulses at 25-27V, 1 MHz applied to 500 μm×500 μm500\,\mu\mathrm{m}\times500\,\mu\mathrm{m} capacitors implemented on the two samples (Figure 3(b)). A linear fit is used to extract relative permittivity (ε33\varepsilon_{33}) of the samples, resulting in 16.8 and 17.1 for as-grown and annealed samples, respectively.

The extracted d33,fd_{33,f} and ε33\varepsilon_{33} are used, along with cc-axis elasticity (c33c_{33}) of 241 GPa[25], to estimate kt2k_t^2 through:

kt2≈d33,f2c33d33,f2c33+ε0ε33(1)k_t^2 \approx\frac{d_{33,f}^2c_{33}}{d_{33,f}^2c_{33}+\varepsilon_0\varepsilon_{33}} \tag*{(1)}

The quadratic increase in d33,fd_{33,f} combined with the nearly constant permittivity, leads to an increase in kt2k_t^2 from 13.8% to 76.2% after annealing. The large kt2k_t^2 of annealed ScAlN sample is approaching the 85.5% value that is recently reported by Yao et. al.[44] for twisted bilayer Lithium Niobate structure in a 1 mm bulk layer. It is worth noting, the results reported herein are from a 100 nm ScAlN film that is 10,000× thinner, thereby promising extreme frequency scaling of bulk acoustic wave resonators for integrated microwave spectral processing in centimeter and millimeter-wave frequencies. The very large kt2k_t^2 of the annealed ScAlN film enables creation of ultra-wide bandpass filters that are excessively needed for beyond-5G wireless communication technologies.

Morphological and structural characterization

To achieve maximum piezoelectric response from a wurtzite crystal, it is important to obtain a highly ordered cc-axis orientation and columnar grains. ScAlN films deposited by radio-frequency (RF) sputtering often exhibits uneven orientations, resulting in degraded crystalline quality and piezoelectric response.[26] Furthermore, annealing can significantly alter the stress in the film, thereby affecting the crystalline quality and the piezoelectric response of the film.[26, 45] To analyze the effect of annealing in boosting crystallinity of the as-grown ScAlN films, we conducted X-ray diffraction (XRD) analysis on ScAlN samples annealed at various temperatures for a duration of 2 hours. The full width at half maximum (FWHM) of the (0002) plane ScAlN ω\omega-peak is 2.85∘2.85^\circ for the as-grown sample. In Figure 4(a), the trend reveals a consistent decrease in the FWHM as the annealing temperature rises, suggesting a progressive enhancement in crystalline quality. Specifically, the FWHM reaches its minimum value of 2.4∘2.4^\circ at an annealing temperature of 700 °C, aligning closely with the observed improvement in d33d_{33}. However, as the annealing temperature is raised to 800 °C, there is a sharp reversal in this trend, with the FWHM experiencing a notable increase. This notable rise in FWHM continues to escalate with increasing annealing temperature, rendering it unmeasurable at 900 °C, indicating a substantial loss of crystallinity at such elevated temperatures.

Plot of FWHM and AFM images of ScAlN samples

Figure 4. (a) Plot of full width at half maximum (FWHM) of the (0002) plane ScAlN ω\omega-peak showing a decreasing trend with increasing annealing temperature, with the lowest value of 2.4∘2.4^\circ at an annealing temperature of 700 °C, aligning closely with the observation in d33d_{33} enhancement. Comparison of surface morphology from AFM images obtained for (b) as-grown ScAlN sample showing an r.m.s. roughness of 0.15 nm and (c) sample annealed at 700 °C for 2h, showing an r.m.s roughness of 0.18 nm for a 1 µm × 1 µm scanning area.

Previous studies showed that high temperature annealing was often associated with a degradation of surface morphology, which can translate to degradation in device performance.[46] Thus, to analyze the effect of high temperature annealing on the surface morphology, we compared the atomic force microscopy (AFM) images of pristine and post-annealed samples, at the optimized condition (700 °C, 2 hours), shown in Figure 4(b) and (Figure 4c). No significant deterioration could be observed for samples annealed under optimized conditions, with the as-grown ScAlN sample showing an r.m.s. roughness of 0.15 nm and the sample annealed at 700 °C for 2h showing an r.m.s roughness of 0.18 nm for a 1 μm×1 μm1\,\mu\mathrm{m} \times1\,\mu\mathrm{m} scanning area.

The PFM data presented in Figure 5 showcases the comparison of ScAlN sputtered film before and after annealing at 700°C for 2 hours. The amplitude plot of as-grown film, which measures the out-of-plane piezo response, exhibits a relatively disordered domain structure with numerous small and randomly oriented unresponsive spots for the as-grown sample. The high density of dark spots suggests significant non-uniformity in piezoelectric activity, indicating that the domains are not well aligned along the cc-crystallographic direction. Post-annealing, the amplitude image shows a marked reduction in the density of dark spots. Moreover, the overall amplitude response is also higher than the as-grown sample, as can be observed in the scale bar of amplitude plots. This indicates an increased alignment of domains towards the cc-axis, resulting in the reduction of unresponsive grains and a more homogeneous distribution of piezoelectric activity. The thermal treatment induces beneficial modifications in the material’s microstructure by primarily reducing defects and dislocations within the crystalline structure.[47, 48] Additionally, annealing mitigates residual stresses introduced during the deposition process, thereby improving mechanical stability and optimizing the stress profile within the film.[49, 50, 51]

Piezoresponse Force Microscopy (PFM) amplitude and phase images of ScAlN sputtered films

Figure 5. Piezoresponse Force Microscopy (PFM) amplitude and phase images of ScAlN sputtered films. (a) As-grown and (b) after annealing at 700°C for 2 hours. The amplitude images show a transition from a disordered domain structure with randomly oriented domains in the as-grown film to a more consistent and aligned arrangement of domains after annealing. The phase images further support this observation, demonstrating a shift from a disordered to a homogeneous arrangement of domains.

The phase images further substantiate these findings by providing insights into the polarization direction and uniformity of domain orientation. The phase image of the as-grown film shows multiple spots with non-uniform phase contrasts, indicating a lack of uniform polarization direction among the domains, contributing to a lower piezoelectric response. Post-annealing, the phase image displays a more uniform phase contrast, with significantly reduced number of misoriented domains, indicating coherent orientation along the cc-axis. Furthermore, annealing stabilizes the desired wurtzite phase of ScAlN, ensuring a homogeneous distribution of Sc within the AlN matrix and reducing the presence of secondary phases that adversely affect piezoelectric performance.[52] Analogously, in more traditional oxide ferroelectrics like hafnium oxide, the material often crystallizes into multiple phases during the growth process that are not desirable for ferroelectric applications. To achieve the transition from centrosymmetric tetragonal phase to a non-centrosymmetric orthorhombic phase, the material is subjected to specific growth conditions or post-deposition treatments such as annealing.[53] The effect of annealing, as explored herein, can be extended beyond nitride ferroelectrics to have a profound impact on the piezoelectric properties of other materials.

In addition to the more homogeneous domain structures observed at a mesoscopic scale, there is often a strong correlation between piezoelectricity of a material and its microscopic structures.[22] Therefore, gaining a deeper insight into piezoelectric properties in wurtzite materials at the atomic scale could potentially establish design guidelines to boost the piezoelectric performance of wurtzite structures. This could allow for the identification of a manageable fundamental parameter that can be further engineered to enhance the wurtzite piezoelectricity, leveraging advanced approaches such as first-principles and machine learning techniques,[22, 23] which have garnered significant attention recently. To explore the structural origins of the enhancement in d33d_{33} after thermal annealing, we performed atomic resolution measurements of the lattice parameters using scanning transmission electron microscopy (STEM). The piezoelectric response along the cc-axis in wurtzite crystals is defined by the piezoelectric modulus d33d_{33} that can be approximated as[1, 54, 55]:

d33≈ε0e33χ33(2)d_{33} \approx\varepsilon_{0} e_{33} \chi_{33} \tag*{(2)}

Thus, the enhancement in the piezoelectric coefficient d33d_{33} can be achieved by increasing the polarization-strain coupling (e33e_{33}) and by increasing the polarization-electric field coupling (χ33\chi_{33}).

Prior studies have shown that the reduction of the lattice parameter ratio (c/ac/a) due to Sc substitution in AlN is strongly correlated with an increasing e33e_{33} and an increase χ33\chi_{33}.[55, 56] This behavior suggests that other factors, such as compressive in-plane strains or internal stresses that increase the c/ac/a ratio, ratio may be the potential limiting factors of piezoelectric response.[57]

Herein, atomic resolution differentiated differential phase contrast (dDPC)-STEM imaging is used to reliably perform bond length measurements as the images are linearly related to the Laplacian of the phase of the sample transmission function, providing reliable measurement of atomic-column locations of both relatively heavy and light elements.[58, 59] Using gaussian fitting of the metal and nitrogen positions, the average values of the aa parameter, cc parameter, and c/ac/a ratio were measured from atomic resolution dDPC-STEM images of ScAlN before and after annealing at the optimized condition of 700°C for 2 hours, as shown in Figure 6. These measurements revealed that the c/ac/a ratio is reduced from 1.542 to 1.522 after annealing. This lattice parameter ratio reduction results mainly from an increase in the aa parameter (3.230 to 3.265 Å) as the decrease in the cc parameter (4.980 to 4.968 Å) is less significant.

dDPC-STEM images of the (a) PVD as-grown and (b) the post-annealed sample

Figure 6. dDPC-STEM images of the (a) PVD as-grown and (b) the post-annealed sample at the optimized condition of 700 °C, 2 hours.

As the atomic resolution images are localized measurements from the center of the ScAlN thin film, we also performed four-dimensional (4D)-STEM measurements over a larger 550 nm wide area to further confirm the findings of the atomic-resolution dDPC-STEM imaging and provide a more statistically reliable analysis of the lattice parameter changes. Figure 7(a) is a virtual dark field image of the ScAlN thin film constructed by the integration of the 4D-STEM data showing columnar grains of 5-30 nm in diameter. Averaging all diffraction patterns from the ScAlN film gives the patterns as shown in Figure 7(b), which displays evidence of a highly ordered cc-axis growth orientation with a randomly oriented aa-axis. For reliable measurement of the aa and cc parameters, only the diffraction patterns which lie close to the [112ˉ0][11\bar{2}0] zone axis was used by selecting those with high intensity in the virtual dark field image constructed from integration of the (11ˉ0l)(1\bar{1}0l) and (1ˉ10l)(\bar{1}10l) diffraction spot positions. The values of aa and cc parameters were then measured from each of these electron diffraction patterns using the exit wave power cepstrum (EWPC) transform which provides a precise measurement of lattice parameter values and is more robust to small variations in sample tilt than direct measurement of the diffraction patterns.[60] An example EWPC pattern from the 4D-STEM dataset is given in Figure 7(c), where the spots in the pattern represent the real space lattice of the ScAlN crystal, with periodicities of aa and cc in the horizontal and vertical direction, respectively. The distributions of values obtained from the 4D-STEM EWPC analysis of both the as-grown and annealed samples are given in the histograms in Figure 7(d-Figure 7f). The EWPC measurements show that after annealing at the optimized condition of 700°C for 2 hours, the c/ac/a ratio is reduced from 1.560 to 1.553 with the aa parameter increasing from 3.185 to 3.201 Å and the cc parameter showing little change (4.967 to 4.970 Å). Moreover, the entire histogram for the aa parameter shows a shift towards higher values, representing an in-plane relaxation of the lattice, that is the driving factor for the c/ac/a decrease. The statistical analysis has been summarized for both measurement techniques (refer to Table I in Supplementary Material). Previous studies show using XRD and 4D-STEM measurements that the aa parameter increases and the cc parameter decreases with increasing film thickness for Sc0.28_{0.28}Al0.72_{0.72}N grown on GaN, which is attributed to stress relaxation[61]. Thus, our lattice parameter measurements suggest that the enhancement of d33d_{33} after thermal annealing is likely due, in part, to stress relaxation of the crystal lattice.

Virtual dark field image, diffraction patterns, EWPC, and lattice parameter histograms

Figure 7. (a) Virtual dark field image constructed from the 4D-STEM data by integration of the (11ˉ0l)(1\bar{1}0l) and (1ˉ10l)(\bar{1}10l) diffraction spot positions. (b) Average diffraction pattern from the ScAlN. (c) EWPC of a single 4D-STEM diffraction pattern. Histograms of the lattice parameter values measured via 4D-STEM EWPCs for the (d) aa lattice parameter, (e) cc lattice parameter, and (f) lattice parameter ratio (c/ac/a).

TechniqueSampleaa (Å)cc (Å)c/ac/a
dDPC-STEMAs-grown3.230 ± 0.0454.980 ± 0.0861.542 ± 0.033
dDPC-STEMTreated3.265 ± 0.0524.968 ± 0.0731.522 ± 0.036
4D-STEMAs-grown3.185 ± 0.0374.967 ± 0.0281.560 ± 0.022
4D-STEMTreated3.201 ± 0.0314.970 ± 0.0201.553 ± 0.018

Table I. Summary of lattice parameter measurement averages ± one standard deviation.

An important consideration herein is that the values reported in this work represent d33,fd_{33,f} and not pristine d33d_{33}, as has been commonly used interchangeably in literature. Although the enhancement observed in d33d_{33} is related to the in-plane relaxation and an increased aa lattice parameter of ScAlN upon annealing, we need to consider that it is still limited by the clamping effect from the substrate.

Thus, the true d33d_{33} value of the material may be much higher when released from the substrate, thus making it even more attractive for applications involving free-standing piezoelectric material systems. The simplicity of this approach can lead to the development of highly scalable and universal processes for enhancement of piezoelectricity across various material systems. Furthermore, this approach opens a promising path to attain even higher d33d_{33} values by optimizing the deposition process, increasing Sc composition, or the incorporation of other IIIB elements, e.g., Y and La in the wurtzite lattice, transforming piezoelectric and electroacoustic devices over a broad range of applications. In particular, the enhanced kt2k_t^2 enabled by the d33d_{33} boost paves the way for the creation of ultra-wide band filters for the next generations of wireless communication. The large coupling despite the small thickness of 100 nm enables piezoelectric transduction of acoustic waves in low-loss substrates (e.g., silicon, silicon carbide, diamond) with minimum perturbation of quality factor (QQ). This facilities creation of integrated high-QQ resonators and acoustic waveguides for ultra-stable clock generation, high resolution sensors, and emerging optomechanical and quantum-acoustic devices.

Conclusion

In summary, we have experimentally demonstrated a simple, scalable, and highly effective strategy to dramatically enhance the piezoelectric response of ScAlN by employing a high temperature annealing technique. The atomic level investigation not only advances our fundamental understanding of d33d_{33} enhancement but also paves the way for the rationale design and optimization of materials with enhanced piezoelectric performance. Moreover, the simplicity of the approach promises scaling accessibility across various material systems. The unprecedentedly high piezoelectric response achieved with ScAlN in this study, along with its potential for seamless compatibility with back-end CMOS integration, not only positions ScAlN but also other IIIBdoped nitride semiconductors as versatile materials for a broad spectrum of applications, spanning from electronics and acoustics to photonics, quantum technologies, and energy harvesting systems, driving progress for commercial adoption.

Experimental Section

Synthesis of ScAlN\mathrm{ScAlN}: A 100 nm Sc0.3Al0.7N\mathrm{Sc}_{0.3}\mathrm{Al}_{0.7}\mathrm{N} was grown by reactive magnetron sputtering on Si (001) substrates with a chuck temperature of 350 °C.

Annealing: Initially the samples underwent a thorough solvent cleaning procedure using acetone, isopropyl alcohol (IPA) and de-ionized (DI) water, followed by annealing under various ambient conditions using an Angstrom Engineering (AE) Furnace. At first, the sample was placed into the annealing tube, and then the tube was evacuated to eliminate remaining air. The temperature was subsequently increased. After the temperature reached a steady state, the sample was subjected to baking for the intended duration.

Piezoelectric modulus (d33d_{33}) and capacitance measurements: Conductive silver paste was used to connect the exposed ScAlN layer to a metal sheet, enhancing signal transmission and measurement precision. PFM investigations utilized vertical domain modes with voltage applied between the top and bottom electrodes. The piezoelectric coefficient d33d_{33} was determined from the amplitude of the piezoresponse signal using localized surface plasmon resonance (LSPR) amplitude plots. Bias voltage was sequentially applied at 0 V and 10 V, with phase and amplitude responses recorded and background noise subtracted from the 10 V amplitude response. The d33d_{33} measurements from PFM were further correlated and validated with measurements utilizing a LDV system with a helium-neon laser (λ\lambda=633 nm) from Polytec GmbH, coupled with a Precision Multiferroic Tester from Radiant Technologies, Inc., to electrically stimulate the films. To minimize ambient vibrational noise and ensure a high signal-to-noise ratio (SNR), the setup was placed on an active vibration-isolation platform within an acoustic enclosure. A custom electrode was patterned by standard photolithography on the ScAlN surface with a diameter of 50 μ\mum, and electrical excitation was achieved using tungsten probes with a 10 V amplitude at 20 kHz. The substrate's backside was grounded electrically, and surface displacement was recorded post-bias application. Bias-dependent measurements confirmed consistent displacement values, indicating negligible ambient noise impact on measurements.

Longitudinal displacement butterfly loop was measured using Polytech NLV-2500 Laser Vibrometer, connected to Radiant PiezoMEMS tester providing electric stimulus. Sample for this measurement is n-doped Si/ AlScN/ Mo capacitor with an area of 0.0025 cm2^2. Testing signals correspond to 60V, bipolar, 2kV triangle waves. To get rid of environmental noise and substrate bending effect, samples were placed on Thorlabs vibration isolation table and bonded to large mass. The results were post-processed by averaging five measurement cycles and correcting for drift.

Capacitor-voltage loops are measured by Radiant PiezoMEMS taster, using 500mm ×\times 500mm capacitors, driven by 1MHz triangular pulses with 27 V and 25 V amplitudes for as-grown and annealed samples, respectively.

Structural and morphological characterization: For structural characterization of the sample, X-ray rocking curve scans (XRC) were performed using a Rigaku SmartLab diffractometer with a Cu Kα1_{\alpha1} radiation x-ray source (1.5406 Å). The surface morphology of the films was analyzed using a Bruker ICON Atomic Force Microscopy (AFM) system. Cross sectional TEM samples of the ScAlN thin films were prepared using a FEI Helios G4 PFIB CXe Dual Beam FIB/SEM, with the final polish using an acceleration voltage of 5 kV. dDPC-STEM imaging was performed using a Themis Z (Thermo Fisher) Cs_{\mathrm{s}} probe corrected microscope with an acceleration voltage of 200kV. and a probe semi convergence angle of 25 mrad. dDPC-STEM imaging was performed using a pixel dwell time of 20 μs and a collection angle of 18 – 69 mrad on a segmented detector with 4 azimuthal segments (DF4). The dDPC images were post-filtered using a gaussian filter with a sigma value of 1.5 pixels. The 4D-STEM data was acquired with an electron microscope pixel array detector (Thermo Fisher) using a pixel dwell time of 1 ms, an acceleration voltage of 200kV, and a probe semi convergence angle of 0.9 mrad, resulting in a probe size with a full-width at half-maximum of approximately 1.5 nm.

Supporting Information

Additional information supporting the findings of this work can be found here:

Acknowledgements

S.M. and M.M.H.T contributed equally to this work. This work was supported by the Department of Defense Advanced Research Projects Agency (Award # HR00112390018), University of Florida and College of Engineering, University of Michigan. The authors acknowledge the Research Service Centers at the Herbert Wertheim College of Engineering at the University of Florida for the electron microscopy work.

Supplementary Information

The lattice parameter values obtained from both atomic resolution dDPC-STEM imaging and 4D-STEM EWPC analysis are summarized in Table I. The two measurement methods are in good agreement with each other and show a clear trend of a reduction in the c/ac/a ratio, primarily driven by an increase in the aa parameter while the c parameter is less significantly changed. Additionally, the 4D-STEM measurements show a reduction in the distribution of measured lattice parameter values after thermal annealing, consistent with the reduction in FWHM of the (0002) plane ScAlN ω\omega-peak measured with XRD indicating an improvement in the crystalline quality of the sample.

References

  1. [1]Richard M Martin, Physical Review B 5 (4), 1607 (1972).
  2. [2]D Bolten, U Böttger, and R Waser, Journal of the European Ceramic Society 24 (5), 725 (2004).
  3. [3]Kenji Uchino, in Advanced Piezoelectric Materials (Elsevier, 2017), pp. 1.
  4. [4]Haonan Jin, Xiangyu Gao, Kaile Ren, Jinfeng Liu, Liao Qiao, Mingzi Liu, Wei Chen, Yuhang He, Shuxiang Dong, and Zhuo Xu, IEEE Transactions on ultrasonics, ferroelectrics, and frequency control 69 (11), 3057 (2022).
  5. [5]Susmriti Das Mahapatra, Preetam Chandan Mohapatra, Adrianus Indrat Aria, Graham Christie, Yogendra Kumar Mishra, Stephan Hofmann, and Vijay Kumar Thakur, Advanced Science 8 (17), 2100864 (2021).
  6. [6]Faysal Hakim, Nicholas G Rudawski, Troy Tharpe, and Roozbeh Tabrizian, Nature Electronics 7 (2), 147 (2024).
  7. [7]Angelika Zaszczyńska, Arkadiusz Gradys, and Paweł Sajkiewicz, Polymers 12 (11), 2754 (2020).
  8. [8]Xianhao Le, Qiongfeng Shi, Philippe Vachon, Eldwin J Ng, and Chengkuo Lee, Journal of Micromechanics and Microengineering 32 (1), 014005 (2021).
  9. [9]Yongling Wu, Yulin Ma, Hongyu Zheng, and Seeram Ramakrishna, Materials & Design 211, 110164 (2021).
  10. [10]P Shivashankar and SJSM Gopalakrishnan, Smart Materials and Structures 29 (5), 053001 (2020).
  11. [11]Satyaranjan Bairagi, Mohammad Shahadat, Daniel M Mulvihill, and Wazed Ali, Nano Energy 111, 108414 (2023).
  12. [12]Madunuri Chandra Sekhar, Eshwarappa Veena, Nagasamudram Suresh Kumar, Kadivala Chandra Babu Naidu, Allam Mallikarjuna, and Dudekula Baba Basha, Crystal Research and Technology 58 (2), 2200130 (2023).
  13. [13]Z Gubinyi, C Batur, A Sayir, and F Dynys, Journal of Electroceramics 20 (2), 95 (2008).
  14. [14]Prasanna V Balachandran, Dezhen Xue, and Turab Lookman, Physical Review B 93 (14), 144111 (2016).
  15. [15]PK Panda, B Sahoo, TS Thejas, and M Krishna, Journal of Electronic Materials, 1 (2022).
  16. [16]Gianluca Piazza, Valeriy Felmetsger, Paul Muralt, Roy H Olsson III, and Richard Ruby, MRS bulletin 37 (11), 1051 (2012).
  17. [17]Chunlong Fei, Xiangli Liu, Benpeng Zhu, Di Li, Xiaofei Yang, Yintang Yang, and Qifa Zhou, Nano Energy 51, 146 (2018).
  18. [18]A Ababneh, Ulrich Schmid, J Hernando, JL Sánchez-Rojas, and Helmut Seidel, Materials Science and Engineering: B 172 (2), 253 (2010).
  19. [19]Morito Akiyama, Toshihiro Kamohara, Kazuhiko Kano, Akihiko Teshigahara, Yukihiro Takeuchi, and Nobuaki Kawahara, Advanced Materials 21 (5), 593 (2009).
  20. [20]Yong Qing Fu, JK Luo, Nam-Trung Nguyen, AJ Walton, Andrew John Flewitt, Xiao-Tao Zu, Yifan Li, Glen McHale, Allan Matthews, and Enrique Iborra, Progress in Materials Science 89, 31 (2017).
  21. [21]Jacob Startt, Mohammed Quazi, Pallavi Sharma, Irma Vazquez, Aseem Poudyal, Nathan Jackson, and Remi Dingreville, Advanced Electronic Materials 9 (4), 2201187 (2023).
  22. [22]Kenji Hirata, Hiroshi Yamada, Masato Uehara, Sri Ayu Anggraini, and Morito Akiyama, ACS omega 4 (12), 15081 (2019).
  23. [23]Huirong Jing, Chaohong Guan, Yu Yang, and Hong Zhu, Journal of Materials Chemistry A 11 (27), 14840 (2023).
  24. [24]Kevin R Talley, Samantha L Millican, John Mangum, Sebastian Siol, Charles B Musgrave, Brian Gorman, Aaron M Holder, Andriy Zakutayev, and Geoff L Brennecka, Physical Review Materials 2 (6), 063802 (2018).
  25. [25]Miguel A Caro, Siyuan Zhang, Tommi Riekkinen, Markku Ylilammi, Michelle A Moram, Olga Lopez-Acevedo, Jyrki Molarius, and Tomi Laurila, Journal of Physics: Condensed Matter 27 (24), 245901 (2015).
  26. [26]PM Mayrhofer, POÅ Persson, A Bittner, and U Schmid, Microsystem Technologies 22, 1679 (2016).
  27. [27]Yuan Lu, Markus Reusch, Nicolas Kurz, Anli Ding, Tim Christoph, Mario Prescher, Lutz Kirste, Oliver Ambacher, and Agnė Žukauskaitė, APL Materials 6 (7) (2018).
  28. [28]Stefan Mertin, Clemens Nyffeler, Tapani Makkonen, Bernd Heinz, Andrea Mazzalai, Thorsten Schmitz-Kempen, Stephan Tiedke, Tuomas Pensala, and Paul Muralt, presented at the 2019 IEEE International Ultrasonics Symposium (IUS), 2019 (unpublished).
  29. [29]Keiichi Umeda, H Kawai, A Honda, M Akiyama, T Kato, and T Fukura, presented at the 2013 IEEE 26th International Conference on Micro Electro Mechanical Systems (MEMS), 2013 (unpublished).
  30. [30]Ping Wang, Ding Wang, Nguyen M Vu, Tony Chiang, John T Heron, and Zetian Mi, Applied Physics Letters 118 (22) (2021).
  31. [31]Morito Akiyama, Kazuhiko Kano, and Akihiko Teshigahara, Applied Physics Letters 95 (16) (2009).
  32. [32]R Petrich, H Bartsch, K Tonisch, K Jaekel, S Barth, H Bartsch, D Glöß, A Delan, S Krischok, and S Strehle, presented at the 2019 22nd European Microelectronics and Packaging Conference & Exhibition (EMPC), 2019 (unpublished).
  33. [33]Stephan Barth, Hagen Bartzsch, Daniel Gloess, Peter Frach, Thomas Herzog, Susan Walter, and Henning Heuer, IEEE transactions on ultrasonics, ferroelectrics, and frequency control 61 (8), 1329 (2014).
  34. [34]Michele Pirro, Bernard Herrera, Meurvert Assylbekova, Gabriel Giribaldi, Luca Colombo, and Matteo Rinaldi, presented at the 2021 IEEE 34th International Conference on Micro Electro Mechanical Systems (MEMS), 2021 (unpublished).
  35. [35]PM Mayrhofer, E Wistrela, M Schneider, A Bittner, and U Schmid, Procedia Engineering 168, 876 (2016).
  36. [36]Yuri Kusano, Guo-Lun Luo, David Horsley, I Taru Ishii, and Akihiko Teshigahara, presented at the 2018 IEEE International Ultrasonics Symposium (IUS), 2018 (unpublished).
  37. [37]Joseph Casamento, Celesta S Chang, Yu-Tsun Shao, John Wright, David A Muller, Huili Grace Xing, and Debdeep Jena, Applied Physics Letters 117 (11) (2020).
  38. [38]Hongyan Liu, Fei Zeng, Guangsheng Tang, and Feng Pan, Applied surface science 270, 225 (2013).
  39. [39]Fredrik Engelmark, G Functes, IV Katardjiev, A Harsta, Ulf Smith, and Sören Berg, Journal of Vacuum Science & Technology A: Vacuum, Surfaces, and Films 18 (4), 1609 (2000).
  40. [40]Feby Jose, R Ramaseshan, S Dash, S Bera, AK Tyagi, and Baldev Raj, Journal of Physics D: Applied Physics 43 (7), 075304 (2010).
  41. [41]Sean Wu, Meng Ying Wu, Jow - Lay Huang, and Ding - Fwu Li1, International Journal of Applied Ceramic Technology 11 (5), 894 (2014).
  42. [42]Minghua Li, Huamao Lin, Kan Hu, and Yao Zhu, Applied Physics Letters 121 (11) (2022).
  43. [43]Danhao Wang, Ding Wang, Peng Zhou, Mingtao Hu, Jiangnan Liu, Shubham Mondal, Tao Ma, Ping Wang, and Zetian Mi, Applied Surface Science 628, 157337 (2023).
  44. [44]Hulin Yao, Pengcheng Zheng, Shibin Zhang, Chuanjie Hu, Xiaofeng Fang, Liping Zhang, Dan Ling, Huanyang Chen, and Xin Ou, Nature Communications 15 (1), 5002 (2024).
  45. [45]MX Wang, FJ Xu, N Xie, YH Sun, BY Liu, ZX Qin, XQ Wang, and B Shen, CrystEngComm 20 (41), 6613 (2018).
  46. [46]CB Vartuli, SJ Pearton, CR Abernathy, JD MacKenzie, ES Lambers, and JC Zolper, Journal of Vacuum Science & Technology B: Microelectronics and Nanometer Structures Processing, Measurement, and Phenomena 14 (6), 3523 (1996).
  47. [47]Chao Gao, Yao Cai, Yang Zou, Binghui Lin, Tingting Yang, Yaxin Wang, Yan Liu, Shishang Guo, and Chengliang Sun, Vacuum 219, 112669 (2024).
  48. [48]Minghua Li, Bangtao Chen, Jielin Xie, Wendong Song, and Yao Zhu, presented at the 2020 IEEE International Ultrasonics Symposium (IUS), 2020 (unpublished).
  49. [49]O Vohringer, Pergamon Press, Advances in Surface Treatments. Technology--Applications--Effects. 4, 367 (1987).
  50. [50]Yanpu Chen, Zishu Zhu, Haolei Zhou, Pingping Ma, Liren Lou, Wei Zhu, and Guanzhong Wang, Diamond and Related Materials 118, 108532 (2021).
  51. [51]F Brunner, L Cancellara, S Hagedorn, M Albrecht, and M Weyers, AIP Advances 10 (12) (2020).
  52. [52]Elmeri Österlund, Glenn Ross, Miguel A Caro, Mervi Paulasto-Kröckel, Andreas Hollmann, Manuela Klaus, Matthias Meixner, Christoph Genzel, Panu Koppinen, and Tuomas Pensala, Physical Review Materials 5 (3), 035001 (2021).
  53. [53]Zhaohao Zhang, Guoliang Tian, Jiali Huo, Fang Zhang, Qingzhu Zhang, Gaobo Xu, Zhenhua Wu, Yan Cheng, Yan Liu, and Huaxiang Yin, Science China Information Sciences 66 (10), 200405 (2023).
  54. [54]Mohammad Noor-A-Alam, Oskar Z. Olszewski, and Michael Nolan, ACS Applied Materials & Interfaces 11 (22), 20482 (2019).
  55. [55]Hiroyoshi Momida and Tamio Oguchi, Applied Physics Express 11 (4), 041201 (2018).
  56. [56]Oliver Ambacher, B Christian, N Feil, DF Urban, C Elsässer, M Prescher, and L Kirste, Journal of Applied Physics 130 (4) (2021).
  57. [57]Yijia Gu, Andrew C Meng, Aiden Ross, and Long-Qing Chen, Journal of Applied Physics 135 (9) (2024).
  58. [58]Sebastian Calderon, Stephen D Funni, and Elizabeth C Dickey, Microscopy and Microanalysis 28 (6), 2047 (2022).
  59. [59]Ivan Lazić, Eric GT Bosch, and Sorin Lazar, Ultramicroscopy 160, 265 (2016).
  60. [60]Elliot Padgett, Megan E Holtz, Paul Cueva, Yu-Tsun Shao, Eric Langenberg, Darrell G Schlom, and David A Muller, Ultramicroscopy 214, 112994 (2020).
  61. [61]Georg Schönweger, Adrian Petraru, Md Redwanul Islam, Niklas Wolff, Benedikt Haas, Adnan Hammud, Christoph Koch, Lorenz Kienle, Hermann Kohlstedt, and Simon Fichtner, Advanced Functional Materials 32 (21), 2109632 (2022).

Paper details

vlnc:W_djn8mhue
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents