Abstract
Polarization engineering has revolutionized the photonic and electronic landscape of III-nitride semiconductors over the past decades. However, recent revelations of giant ferroelectric polarization in wurtzite III-nitrides challenge the long-standing paradigms. Here, we experimentally elucidate the polarization, including its magnitude and orientation, and its relationship to lattice polarity in III-nitrides. Those experimentally determined polarizations exceeding 1 C/m2 with an upward orientation in metal-polar wurtzite nitride compounds align with recent theoretical predictions. To reconcile these findings, a unified polarization framework is established based on the centrosymmetric layered-hexagonal reference structure. This unified framework redefines the polarization landscape in contemporary GaN heterostructures, quantum structures, and ferroelectric heterostructures. Furthermore, we predict significant tunability and a dramatic increase in sheet carrier concentration in ferroelectric ScAlN/GaN heterostructures, heralding advancements in high-power, high-frequency, and reconfigurable transistors, and non-volatile memories. This work bridges the critical gap in the understanding of polarization in both conventional and ferroelectric wurtzite nitrides, offering fundamental insights and paving the way for next-generation photonic, electronic, and acoustic devices.
Subject terms: Semiconductors, Semiconductors, Ferroelectrics and multiferroics, Ferroelectrics and multiferroics
The authors provide experimental evidence for giant polarization in wurtzite III-nitrides, establishing a unified framework for understanding and engineering polarization in these materials and their functional device architectures.
Introduction
III-nitride semiconductors, characterized by their non-centrosymmetric wurtzite (WZ, P63mc) structure, inherently exhibit nonvanishing spontaneous polarization and piezoelectric polarization1,2. The pronounced polarization discontinuities at the III-nitride heterointerfaces result in the formation of bound sheet charges. These charges generate an internal polarization field, leading to phenomena such as the formation of two-dimensional electron or hole gases (2DEG or 2DHG)3–5, polarization-induced doping6, quantum-confined Stark effect7, etc. in III-nitride heterostructures. Over the past decades, polarization engineering has remarkably boosted the photonic and electronic ecosystem of III-nitrides8–10. However, recent discoveries of wurtzite nitride ferroelectrics, such as scandium aluminum nitride (ScAlN)11–15, which exhibit a giant spontaneous polarization compared to traditional expectations3,16, challenge the well-established paradigms. Therefore, refining and unifying the cognition of polarization within the III-nitride semiconductor family is crucial to ensuring continued progress in material and device development.
Experimentally determining the polarization of a solid, particularly the spontaneous polarization, faces significant challenges. In crystalline solids, people generally measure the changes in polarization rather than the absolute values17. Consequently, theoretically predicted spontaneous polarizations are often relied upon to simulate and interpret the experiments. Within the framework of the Modern Theory of Polarization (MTP), the spontaneous polarization of a crystal is defined as the difference in the formal polarization between the material of interest and a reference structure: , where and are the formal polarization of the material and its reference structure, respectively, is the spontaneous polarization of the material relative to the reference structure17–20. Normally, the reference structure should have vanished formal polarization. The reference structure for wurtzite crystals can be either a zinc-blende (ZB, ) or a h-BN like layered-hexagonal (LH, P63/mmc) structure, as shown in Fig. 1a and detailed in Supplementary Note 1. The polarization properties of WZ III-nitrides were firstly investigated by Bernardini et al. in 1997 using the MTP framework, with ZB serving as the reference structure1. Their predictions, with polarization values below 0.1 C/m2, have been extensively used to predict the net polarization bound charge at GaN-based heterointerfaces. In 2016, Dreyer et al. revised the polarization constants within WZ III-nitrides using a LH reference structure21, predicting giant spontaneous polarizations exceeding 1 C/m2—more than an order of magnitude larger than the initial ones with the ZB reference structure. This divergence garnered attention following the report of ferroelectric polarization switching in ScAlN films by Fichtner et al. in 2019, where remnant polarizations over 1 C/m2 were measured11,22–24. Nevertheless, the polarization values and their orientations in III-nitride semiconductors are still open questions, in particular, convincing experimental evidences remain elusive.
Fig. 1. Local polarization measurements of WZ III-nitrides.
a Crystal structures of III-nitrides, including WZ, ZB, and LH configurations (top), and atomic schematics illustrating the adiabatic and gap-preserving displacement between WZ and its ZB and LH reference structures along the c-axis (bottom). b Electron localization function (ELF) of WZ AlN projected along the zone-axis, showing both the ionic (Pion) and electronic (Pel) contributions to polarization. c Diagram of the local polarization measurement method based on a simulated ABF-STEM image of WZ AlN. Atomic displacement (Δr) and lattice constants (a and c) are measurable. The inset shows a quadrilateral prism representing the unit cell with volume of Ω. d–k ABF-STEM images of WZ M-polar (d) AlN, (e) GaN, (f) InN, and (g) ScAlN, and N-polar (h) AlN, (i) GaN, (j) InN, and (k) ScAlN, each overlaid with vector map of the measured local polarization.
In this work, we establish a unified perspective on the polarization of III-nitride semiconductors through systematic experimental examinations. The inherent correlation between the spontaneous polarization and lattice polarity are determined via comprehensive local and macroscopic polarization characterizations. These experimental findings corroborate that the understanding of the spontaneous polarization, ferroelectric polarization, and lattice polarity in WZ III-nitride family can be much more accurately unified by referencing the LH structure, rather than the long-standing ZB reference structure. Accordingly, the polarization landscape in the advanced AlGaN/GaN and ferroelectric ScAlN/GaN heterostructures has been reevaluated using the newly unified polarization framework. This approach not only reconciles the existing experimental observations, but also predicts significant tunability and a dramatic increase of sheet carrier concentration in the ferroelectric ScAlN/GaN heterostructures, providing new insights into the development of high-performance electronic and ferroelectric devices.
Results
Local polarization measurements
As illustrated in the top panel of Fig. 1a, III-nitride semiconductors exhibit a wurtzite ground state structure, featuring with an in-plane lattice a for the basal hexagon, an out-of-plane lattice c for the hexagonal prism, and an internal parameter u, which is defined as the ratio of the metal-nitrogen (M-N) bond length along the c-axis to the lattice constant c. Wurtzite is a non-centrosymmetric structure where metal and nitrogen atoms occupy two separate atomic planes. A sublattice displacement, , represents the smallest interplanar spacing between the metal and nitrogen atomic planes in unit of c. The lack of spatial inversion symmetry results in a non-zero spontaneous polarization along the c-axis. Two opposite crystallographic lattice polarities exist along this axis, defined as M-polarity and N-polarity, corresponding to the [0001] and directions, respectively3. In the M-polarity configuration, metal atoms are positioned above the nearest nitrogen atom plane, resulting in d > 0 (Fig. 1a), whereas in the N-polar one, the order is reversed. Although spontaneous polarization plays a crucial role in the development of III-nitride semiconductors, direct polarization measurements have been experimentally challenging. Therefore, the reported values of polarization are basically theoretically predicted ones1,2,21,25. The MTP modeling is usually used for such theoretical prediction17,20,26, where it suggests that both ionic and electronic contributions to polarization can be calculated from the atomic displacement (Δr) using the Born effective charges (Z *) (Fig. 1c and Supplementary Note 2). In this approach, Δr is the atomic displacement of a crystal compared to its centrosymmetric structure. The bottom panel of Fig. 1a shows the atomic chains for the WZ phase and the corresponding ZB and LH reference structures along their polar axes, specifically the , , and axes, respectively. Obviously, the centrosymmetric reference structure for WZ is the LH phase, where both metal and nitrogen atoms share the same atomic plane. Moreover, an adiabatic, gap-preserving atomic displacement path can only be established between the WZ and LH reference structure for III-nitrides.
In the WZ structures, as shown in Figs. 1a, c, the smallest atomic plane spacing along the c-axis is the atomic displacement used for polarization calculation, i.e., Δr = dc. This displacement is possible to be directly measured using atomically resolved scanning transmission electron microscope (STEM) by identifying the positions of metal and nitrogen atom columns27,28. Furthermore, for III-nitride semiconductors, Δr determines the polarization orientation21,25,28: positive (d > 0) and negative (d < 0) Δr correspond to upward and downward polarizations, which can be defined as +P and −P, respectively, as indicated in Fig. 1a. Therefore, local polarization measurements utilizing STEM provide a direct and accurate way for determining the intrinsic polarization at the unit cell scale. While STEM has been widely used to investigate microstructure and lattice polarity in III-nitrides15,29, its application for local polarization measurements remains limited.
In this work, we conducted comprehensive local polarization measurements on III-nitrides, including M-polar and N-polar AlN, GaN, InN, and ScAlN by using annular bright field STEM (ABF-STEM). Figure 1d–g displays ABF-STEM images of M-polar AlN, GaN, and InN films, respectively, which were grown on sapphire substrates. The lattice parameters of these films, measured by X-ray diffraction (XRD), are consistent with previous theoretical and experimental reports, showing negligible strain (Fig. S4 and Table SII). The spatial positions of atom columns were determined by using a two-dimensional Gaussian fitting process on the ABF-STEM images, enabling precise measurement of Δr across the entire image (Fig. S6). Polarization was then quantified using Born effective charges derived from the density-functional theories (DFT) (Supplementary Note 2 and Table SII). Vector maps of the calculated polarization per unit cell were superposed on the corresponding ABF-STEM images (Fig. 1d–g). The measured average polarization for M-polar AlN, GaN, and InN is +1.37 ± 0.11, +1.29 ± 0.11, and +1.04 ± 0.15 C/m2, respectively, all exhibiting upward-orientated spontaneous polarization. The error bar arises from the measurement uncertainty resulting from the atomic position fitting process.
We further conducted local polarization measurements on a M-polar ScAlN layer which was grown on GaN/sapphire template by molecular beam epitaxy (MBE)30. The ScAlN has a monocrystalline wurtzite structure with 18% Sc content, which is nearly lattice-matched to the underlying GaN, minimizing strain effect on the polarization measurements16,31,32. More details can be found in the Methods and Supplementary Fig. S5. Figure 1g shows the ABF-STEM image of the ScAlN, overlaid with the measured polarization vector map. We obtained an average polarization of +1.38 ± 0.17 C/m2. To further validate the accuracy and reliability of this approach and obtain a systematic understanding of polarization in WZ III-nitrides, the same measurements were performed on the N-polar counterparts of these materials. As illustrated in Fig. 1h–k, the measured average polarization for N-polar AlN, GaN, InN, and ScAlN is −1.34 ± 0.20, −1.28 ± 0.24, −1.09 ± 0.26, and −1.38 ± 0.21 C/m2, respectively, all showing downward-orientated spontaneous polarization.
Since ScAlN is a ferroelectric material, we then conducted macroscopic electrical measurements on the MBE-grown ferroelectric ScAlN/GaN heterostructure to determine its polarization value. Figure 2a shows the high angle annular dark field (HAADF) STEM image captured from the ScAlN/GaN sample, exhibiting atomically sharp interface. Lithographically patterned circular Ti/Au pads served as top electrodes, while the n+-GaN layer functioned as the bottom electrode, with voltages applied from the top electrode, as schematically shown in the inset of Fig. 2b. Figure 2b shows a typical polarization versus electric field (P-E) hysteresis loop measured from the ferroelectric ScAlN/GaN heterostructure, yielding a remanent polarization of approximately 1.48 C/m2, consistent with the locally measured polarization value of 1.38 C/m2. The slightly higher remanent polarization is attributed to overestimation due to electric leakage at high voltage. This issue can be partially mitigated by using positive-up-negative-down (PUND) measurements (Fig. S7). The remanent polarization obtained from PUND analysis using a triangular pulse sequence is 1.11 C/m2.
Fig. 2. Polarization characteristics of WZ ferroelectric ScAlN/GaN heterostructure.
a HAADF-STEM image of the as-grown ScAlN/GaN heterostructure, exhibiting an atomically sharp interface. b Typical P-E hysteresis loop recorded from a ScAlN capacitor at 1 kHz without subtracting electric leakage. The inset shows a schematic of the ScAlN metal-ferroelectric-semiconductor (MFS) capacitor structure. c, d ABF-STEM images of ScAlN, captured from (c) positively and (d) negatively biased MFS capacitors, as indicated in (b). The superimposed metal (pink) and nitrogen (blue) atoms highlight the atomic stacking sequence, revealing N- and M-polar lattices in (c) and (d), respectively. e Monopolar measurements executed on ScAlN MFS capacitors with opposite initial polarization states: (i, ii) fresh M-polar ScAlN, exhibiting upward spontaneous polarization, and (iii, iv) switched N-polar ScAlN, presenting downward spontaneous polarization. The insets depict the applied voltage pulses and the evolution of polarization orientation under an external electric field, with arrows in the ScAlN/GaN heterostructures indicating the polarization orientation.
The ferroelectric nature of ScAlN also enables local polarization measurement in III-nitrides after reversing the lattice polarity, which was previously unattainable. Figure 2c, d shows ABF-STEM images captured from two ScAlN devices with opposite polarization states, corresponding to the positive and negative biases indicated in Fig. 2b. By comparing the stacking order of metal (Sc/Al) and nitrogen atoms in these ABF images with the atomic model in Fig. 1a, the lattice polarity was confirmed as N-polarity for positively-biased state (Fig. 2c) and M-polarity for negatively-biased one (Fig. 2d). The measured average polarizations are −1.34 ± 0.14 and +1.43 ± 0.11 C/m2, respectively. The excellent agreement of these STEM quantified and electrically measured polarization magnitudes for as-deposited and lattice-reversed ScAlN films unambiguously identifies that this material possesses a polarization over 1 C/m2. Notably, recent studies have demonstrated ferroelectric polarization switching in sputter-deposited AlN films33,34, showing a remanent polarization of 1.5 C/m2, which aligns well with the polarization values obtained for AlN in Fig. 1. Given that conventional III-nitrides and ferroelectric III-nitrides share identical crystal symmetry, as well as comparable Born effective charges and atomic displacements (Table SII), we can infer, based on the MTP, that the polarization landscape of the WZ III-nitride family should be similar. In other words, conventional III-nitrides possess strong polarization effects akin to those observed in ferroelectric nitrides. Thus far, both local and macroscopic polarization measurements corroborate that WZ III-nitrides exhibit spontaneous polarization values exceeding 1 C/m2.
Identification of polarization orientation
Subsequently, we examined the intrinsic orientation of the spontaneous polarization in WZ III-nitrides, an ongoing topic of debate, through macroscopic electrical measurements. The wake-up-free nature and highly controllable, uniform lattice polarity of MBE-grown monocrystalline ferroelectric ScAlN allowed us to perform monopolar measurements on fresh devices without pre-cycling treatments. This is essential for examining intrinsic polarization orientation. In ferroelectrics, once the orientation of the spontaneous polarization is reversed by an external electric field, the redistribution of the polarization-compensated charges gives rise to an instantaneous current flow known as “displacement current”35,36. Displacement current, therefore, serves as an indicator of changes in polarization orientation. For monopolar measurements, a single positive or negative triangular pulse exceeding the coercive field was applied to the top electrodes of fresh ScAlN capacitors. The corresponding current density versus electric field (J-E) curves are plotted in Fig. 2e. A distinct displacement current is observed following the application of a positive pulse bias on a fresh M-polar ScAlN capacitor, while no displacement current is detected after executing a negative pulse bias, as displayed in the top panel of Fig. 2e. This behavior indicates that the orientation of the spontaneous polarization in M-polar ScAlN can be inverted by a downward external electric field (towards the ScAlN/GaN interface), but not by an upward one (towards the ScAlN surface), as depicted in the insets of Fig. 2e(i, ii).
The same measurements were conducted on N-polar ScAlN/GaN capacitors, shown in the bottom panel of Fig. 2e. Those N-polar ScAlN/GaN capacitors were obtained by converting the original M-polar to N-polar lattice via applying positive voltage pulses (Fig. 2c). In contrast to the M-polar capacitors, the N-polar ones exhibited negligible response to a positive voltage pulse, while a definite displacement current was generated under a negative pulse. This phenomenon suggests that it is an upward external electric filed—rather than a downward one—that induces inversion of the spontaneous polarization orientation in N-polar ScAlN, as illustrated in the insets of Fig. 2e(iii, iv). According to the theory of dielectric polarization, the spontaneous polarization of a ferroelectric aligns with the external electric field after polarization reversing17,36,37. Based on this, above monopolar analyses suggest that M-polar ScAlN exhibits an upward spontaneous polarization, while N-polar ScAlN has a downward one. These findings, when combined with the local polarization measurements shown in Fig. 1, provide unambiguous experimental confirmation that M-polar III-nitrides possess an upward spontaneous polarization (+P), while N-polar III-nitrides exhibit a downward one (−P).
Unified polarization framework
The spontaneous polarization of wurtzite ScAlN with Sc content in a range of 0 to 0.375 was calculated by using first-principles DFT (Fig. S3a). See Supplementary Note 1 for details. The theoretically predicted spontaneous polarizations align well with both local and macroscopic measurements, as well as the experimentally reported remanent polarization for ScAlN, in terms of both magnitude and evolution trend (Fig. S3b). Table 1 lists our results alongside previously reported experimental measurements34,38–40 and theoretical calculations1,16,21,25,41 of spontaneous polarizations in III-nitride semiconductors. Notably, while the experimentally determined local and macroscopic polarizations in our data are well consistent, they show significant discrepancies when compared to the widely adopted polarization values for III-nitride semiconductors over the past two decades, which were obtained using ZB reference structure1,16,21,38. These discrepancies are evident in two key aspects: (i) the experimentally determined spontaneous polarization values are an order of magnitude higher than the theoretical predictions using ZB reference structure; (ii) for the same lattice polarity, such as M-polar AlN, our experimental findings indicate an orientation that is exactly opposite to the previous theoretical predictions.
Table 1.
Comparison of spontaneous polarization reported for WZ III-nitride semiconductors
Materials | Measured polarization | DFT calculated polarization | ||
---|---|---|---|---|
Local polarization measurements |
Electrical |
Ref. ZB1,16,21,38 | Ref. LH21,25,41 | |
AlN | 1.37a | 1.40 | −0.081 to −0.090 | 1.339–1.351 |
GaN | 1.29a | −0.022 to −0.035 | 1.299–1.312 | |
InN | 1.04a | −0.032 to −0.053 | 1.026–1.032 | |
ScAlN | 0.5–1.5 | −0.089 to −0.296 | 1.35–1.00 | |
ScAlNa | 1.34–1.43 (i) | 1.48 (i) | 1.37–1.15 (ii) |
aThis work: the values are obtained from ScAlN with Sc content of (i) 0.18 and (ii) 0–0.375.
All listed values correspond to III-nitrides with an M-polar lattice configuration.
The reason for the significant discrepancy lies in the reference structure used when calculating the polarization within the MTP framework (Supplementary Note 1)17,26. In the MTP modeling, polarization is a periodic multiple-valued lattice, where only the polarization differences between two states, such as the material of interest and the reference, taken from the same branch, are uniquely defined17,18,20,21. The multivalued formal polarizations for WZ, ZB, and LH III-nitrides within the MTP modeling are illustrated in Fig. 3a. The spontaneous polarization relative to the ZB () and LH () references can be calculated by taking the difference in formal polarization from the same branch17,19,20,26. Obviously, both magnitude and orientation of the polarization are strongly linked to the reference structure. For example, as shown in Fig. 3b and Table 1, M-polar AlN has a spontaneous polarization of −0.090 C/m2 when referenced to the ZB structure, whereas a spontaneous polarization of +1.351 C/m2 is obtained using the LH reference structure. The latter one agrees well with our experimentally measured polarization of +1.37 C/m2.
Fig. 3. Spontaneous polarization of WZ III-nitride semiconductors within the MTP modeling.
a Multivalued formal polarization Pf, predicted by the MTP modeling as a function of the atomic displacement d. Each black dashed line represents a branch of multivalued function, while the difference between two branches is the modulo of the polarization quantum Pq. The blue, brown, and red spots correspond to the Pf values for a WZ structure and its ZB and LH reference structures, respectively. b Formal polarization of WZ, ZB and LH III-nitrides (AlN, GaN, InN and ScAlN) with M-polar lattice (d > 0) taking from the “0” branch in (a). Black arrows represent spontaneous polarizations of WZ III-nitrides relative to the ZB and LH reference structures, showing downward and upward orientations, respectively.
There is no inherent issue with utilizing either the ZB or LH reference structure to define the spontaneous polarization in WZ crystals. Nevertheless, when using the spontaneous polarization to evaluate the polarization difference at a WZ heterointerface, such as an AlN/GaN (), a correction term accounting for the polarization difference between the corresponding reference structures () must be incorporated21,25, as shown in Fig. 3b and Supplementary Note 1. For the AlN/GaN heterostructure, the correction term for ZB reference structure is clearly nonzero: , where e is the electronic charge, and are the in-plane lattice constant of WZ AlN and GaN, respectively. In contrast, the correction term for LH reference structures () is just zero because all LH reference structures exhibit vanished formal polarization. Therefore, the spontaneous polarization of WZ III-nitrides determined by using the LH reference structure can be directly used to calculate the polarization difference at heterointerfaces without requiring a correction term. Moreover, spontaneous polarizations derived using the ZB reference are not possible for understanding the reversible giant polarization of wurtzite ferroelectrics (Fig. 2 and Supplementary Note 1). It is thus clear that the discrepancies between theoretically predicted and experimentally determined spontaneous polarization are fairly addressed by using the LH reference structure.
Refined polarization understanding in III-nitride heterostructures
The above findings contribute to a refined understanding of polarization in GaN-based heterostructures, where strain induced piezoelectric polarization should also be considered. WZ III-nitrides are typically hetero-grown on substrates along the polar c-axis. Consequently, an in-plane biaxial strain (, where , with and being the in-plane lattice constants for the relaxed and strained layers, respectively) is present in the top layer, leading to a strain induced piezoelectric polarization ()3. is defined as the formal polarization difference between the strained and relaxed structure. It can be expressed in terms of the in-plane biaxial strain (), piezoelectric coefficients (, ), and elastic constants (, )3,16,21,42–44. See details in Supplementary Note 1.
We use a state-of-the-art M-polar AlGaN/GaN high electron mobility transistor (HEMT) structure as an example to assess the spontaneous, piezoelectric, total polarization, as well as the interface polarization difference (). Calculation details are discussed in Supplementary Note 3. To quantitively compare the difference arising from the divergent reference structures, we set the Al composition to 0.3 (Al0.3Ga0.7N), a commonly used barrier in AlGaN/GaN HEMT structures10. The biaxial strain effect in Al0.3Ga0.7N is evaluated by considering two extreme cases: fully relaxed () and fully strained () one. The left panel of Fig. 4a(i-iii) presents the polarization schematic in M-polar AlGaN/GaN heterostructures derived using the LH reference, exhibiting a significant difference compared to the conventional understanding based on the ZB reference, as shown in Fig. 4a(iv-vi).
Fig. 4. Polarizations, sheet carriers, and band diagrams in AlGaN/GaN heterostructures.
a Spontaneous polarization, piezoelectric polarization, interface polarization difference, and sheet carriers in (i, iv) relaxed, (ii, v) tensile-strained, and (iii, vi) compressive-strained AlGaN/GaN heterostructures. The schematics in (i-iii) and (iv-vi) are drawn with the LH and ZB reference structures, respectively. Black arrows indicate the polarization orientation. b–d Polarization charge distribution and corresponding band diagrams for (b) M-polar AlGaN/GaN, (c) M-polar GaN/AlGaN, and (d) N-polar GaN/AlGaN heterostructures, illustrating the formation of 2DEG or 2DHG at the interface.
For fully relaxed AlGaN/GaN heterostructures, a positive (+0.012 C/m2) is obtained, suggesting a negative net polarization charge at the interface. Therefore, as shown in Fig. 4a(i), the fully relaxed AlGaN/GaN interface tends to attract free holes, rather than electrons, to compensate the negative net polarization charge. However, the fully relaxed structure is a hypothetical scenario, as strain is typically present in real systems. For fully strained AlGaN/GaN heterostructures, a negative (−0.021 C/m2) is obtained, which is comparable with (−0.030 C/m2). The similarity between these two values results from the cancelation of the correction terms, which has been discussed in detail by Dreyer et al.21. As shown in Fig. 4a(ii), the negative polarization difference leads to a positive net polarization charge, which tends to accumulate free electrons at the tensile strained AlGaN/GaN heterointerface. This phenomenon underlies the formation of 2DEG in AlGaN/GaN HEMTs.
The polarization discontinuities at other M-polar and N-polar AlGaN/GaN heterostructures can be deduced using the same method. Figure 4a presents the polarization schematics for several representative heterostructures. Based on this understanding, the polarization charge distribution and band profiles for three typical III-nitride heterostructures, including M-polar AlGaN/GaN, M-polar GaN/AlGaN, and N-polar GaN/AlGaN, are redrawn in Fig. 4b–d. These heterostructures have attracted tremendous attention in high-frequency and high-power electronics due to their potential to form high-mobility 2DEG or 2DHG. Although there are no significant differences in band profiles between the two reference structures, the refined polarization distribution and formation mechanisms provide a deeper and more solid foundation for the understanding and engineering of polarization in advanced nitride heterostructures. A comprehensive comparison of our predictions with previously reported experimental results and theoretical studies for AlGaN/GaN HEMT structures is provided in Fig. S9 and further discussed in Supplementary Note 3. Referring to the pioneering explanation of III-nitride polarization and 2DEG/2DHG formation, this new understanding aligns well with existing experimental observations. Furthermore, by comparing the fully relaxed and strained AlGaN/GaN heterostructures within the unified polarization framework, we conclude that: (i) piezoelectric polarization is oriented opposite to spontaneous polarization under tensile strain while in the same orientation under compressive strain; and (ii) piezoelectric polarization plays a crucial role in the formation of 2DEG or 2DHG at the interface.
We then turn to ScAlN/GaN heterostructures, which are more attractive since ScAlN has a ferroelectric polarization, and the strain in ScAlN can be tuned from tensile to compressive through modifying the Sc content and the orientation of the piezoelectric polarization can thus be manipulated as well11–13,16,31,45. Figure 5a illustrates the polarization schematics of ferroelectric ScAlN/GaN heterostructures, where the ScAlN layer is in M- (left panel) and N-polarity (right panel), while the underlying GaN layer remains M-polarity in both cases. In Fig. 5a, the evolution of the in-plane lattice parameter a can be divided into three regimes with increasing Sc content from top to bottom: (panel i, tensile strain), (panel ii, fully relaxed), and (panel iii and iv, compressive strain). The spontaneous and piezoelectric polarizations, as well as the polarization differences, are indicated with black arrows in Fig. 5a. The net polarization charge at the ScAlN/GaN interface is plotted in Fig. 5b (solid curve), with the corresponding maximum sheet carrier concentration indicated on the right axis. Further details are provided in Supplementary Note 3.
Fig. 5. Polarizations and sheet carriers in ferroelectric ScAlN/GaN heterostructures.
a Spontaneous polarization, piezoelectric polarization, interface polarization difference, and sheet carriers in as-grown M-polar (left) and polarization switched N-polar (right) ferroelectric ScAlN/GaN heterostructures with varying Sc content. The strain transitions in ScAlN undergo three regimes along with Sc content increasing: (i) tensile strain, (ii) fully relaxed, and (iii, iv) compressive strain. Black arrows indicate the polarization orientation. b Net polarization charge and corresponding maximum sheet carrier density of ScAlN/GaN heterostructures as a function of Sc content. The red and blue curves represent sheet charge density before and after polarization switching, respectively. While solid and dashed curves correspond to the values calculated using LH and ZB reference structures, respectively. c Schematic of a ScAlN/GaN Fe-HEMT with opposite polarization orientations for ScAlN (left) and its typical transfer characteristic curves (right), showing counterclockwise ID–VGS hysteresis loops with tunable Vth.
As shown in Fig. 5a, the initial tensile strain in the ScAlN layer gradually evolves into compressive strain with increasing Sc content, leading to a reversed orientation of the piezoelectric polarization. Consequently, the sheet carriers at the heterointerface transform from 2DEG to 2DHG (Fig. 5a, b). As the Sc content increases from 0 to 0.25, the positive polarization bound charge initially rises slightly and then lowers down to zero, followed by a gradual increase in negative polarization bound charge. This evolution trend reflects the trade-off between spontaneous and piezoelectric polarizations. For comparison, the net polarization charge calculated using ZB reference structure is plotted as a blue dashed line in Fig. 5b, revealing a larger transition point at the Sc content of 0.4.
The polarization bound charge in the switched ScAlN/GaN heterostructure is also plotted in Fig. 5b (red curve). In this scenario, the top ScAlN layer exhibits N-polar lattice while the underlying GaN remains M-polar one, forming a head-to-head polarization domain wall at the ScAlN/GaN interface, as illustrated in the right panel of Fig. 5a. This configuration results in a significant increase in positive polarization sheet charge, at least 30 times higher than typically observed, which is completely contradict to the predictions by using ZB reference structure, indicated with red dashed curve in Fig. 5b. In idealized configurations, assuming sharp, clean polarization reversal interfaces without dead layer, dangling bonds, defects, alloy composition fluctuations, or charge compensation, this high positive polarization charge leads to pronounced downward band bending (Fig. S10). This facilitates the formation of 2DEG with ultra-high sheet carrier densities exceeding 1 × 1015 cm−2 in the polarization switched ScAlN/GaN heterostructures. Such high carrier density has never been observed at any semiconductor heterointerfaces. This unprecedented phenomenon reveals a significant potential for advancing high-frequency and high-power electronic applications.
Furthermore, since ScAlN is a ferroelectric semiconductor, the threshold voltage (Vth) of ScAlN/GaN ferroelectric HEMTs (Fe-HEMTs) can be tuned by ferroelectric polarization through either the quantity or sign of the polarization bound charge, resulting in a counterclockwise ID–VGS hysteresis loop, as shown in Fig. 5c. The initial net polarization charge at the ScAlN/GaN interface determines the highest Vth, while the giant positive polarization charge induced by ferroelectric switching governs the lowest Vth. Therefore, Vth can be precisely modulated by controlling the ferroelectric switching process. Additionally, a steep-slope ID–VGS curve with a reduced subthreshold swing is likely observed during backward (positive-to-negative) sweeps, benefiting from the negative capacitance effect caused by ferroelectric switching46,47. It should be noted that the substantial positive polarization charge induced by polarization reversal at the interface can significantly enhance vertical conductivity. This enhancement contributes to the low-resistance state observed in ferroelectric ScAlN/GaN memristors48,49. Such a phenomenon could also facilitate the formation of low-resistance drain and source Ohmic contacts in ScAlN/GaN Fe-HEMTs and photodetectors.
Discussion
Although the above analyses are promising, experimental demonstration and precise determination of the electron density and mobility of the 2DEG formed at the ferroelectric switched ScAlN/GaN interface remain significant challenges. The primary difficulty lies in the presence of ferroelectric dead layers, as well as dangling bonds, defects, and roughness at the polarization reversal interface in practical ferroelectric heterostructures15,29,50. In contrast, polarization modeling normally assumes an idealized polarization reversal interface. Those undesirables in practical systems significantly suppress the modulation capability of ferroelectric polarization on the ScAlN/GaN interface carriers, for instance, through mitigated field effects and partial charge compensation50, as discussed in detail in Supplementary Note 3. Nevertheless, the increased vertical conductivity48,49 and the negatively shifted pinch-off voltage (Supplementary Note 3) or Vth51 observed in ferroelectric ScAlN/GaN heterostructures after polarization reversal confirm the enhancement of interfacial carrier concentration. These experimental observations suggest that the potential applications of the extreme sheet charges predicted in Fig. 5b are already emerging. We believe that ongoing advancements in ferroelectric nitride technologies, such as mitigating or eliminating dead layers, achieving sharp switching, and creating clean polarization-reversal interfaces, will gradually bring experimental results closer to the theoretical predictions presented here.
In summary, the magnitude and orientation of polarization in wurtzite III-nitrides have been identified through quantitative local polarization analysis at the unit cell scale as well as macroscopic electrical measurements. Giant upward (downward) spontaneous polarizations exceeding 1 C/m2 have been experimentally determined in as-grown M-polar (N-polar) wurtzite III-nitride films, including conventional AlN, GaN, and InN, as well as the emerging ferroelectric ScAlN. These findings align well with theoretical predictions based on the centrosymmetric LH reference structure, establishing a unified perspective on the polarization of III-nitride semiconductors with strong support from both theoretical and experimental evidence. Furthermore, the polarization landscape and band profile in the cutting-edge AlGaN/GaN and ferroelectric ScAlN/GaN heterostructures have been reevaluated within this unified polarization framework. The potential for achieving 2DEG with ultra-high sheet carrier densities in ferroelectric ScAlN/GaN heterostructures has been highlighted. The proposed and validated new understanding not only bridges the knowledge gap in polarization engineering when integrating ferroelectric nitrides with contemporary GaN device architectures, but also remains consistent with existing experimental observations. This work provides essential insights into the understanding and modulation of polarization in III-nitride heterostructures, setting the stage for the development of advanced photonic, electronic, optoelectronic, and acoustic devices by incorporating ferroelectricity into the III-nitride ecosystem.
Methods
Material growth and device fabrication
M-polar InN and ScAlN films were grown on commercial GaN/sapphire templates, while N-polar AlN, GaN, InN, and ScAlN films were grown on C-face 4H-SiC substrates utilizing an SVTA MBE system. M-polar GaN and AlN films were prepared using metal organic chemical vapor deposition (MOCVD) on sapphire substrates. The MBE system was equipped with a Veeco Unibulb radio frequency (RF) N-plasma source, dual-filament Knudsen cells for the Al (purity 99.99995%), Ga (purity 99.99999%), and In (purity 99.99999%) sources, and a high-temperature Knudsen cell for the Sc source (purity 99.999%). The epitaxy conditions for MBE-grown AlN, GaN, and InN films have been previously reported52–54. For ScAlN growth, an N2 (purity 99.9999%) gas flow rate of 1 sccm with an RF power of 400 W was maintained throughout the growth process, corresponding to a deposition rate of 300 nm/h for GaN under slightly Ga-rich conditions. The GaN/sapphire template was degassed at 300 °C for 1 h in the MBE preparation chamber, followed by outgassing at 750 °C for 10 min in the growth chamber prior to deposition. A 100-nm-thick Si-doped n+-GaN (with an electron concentration of 5 × 1019 cm-3) was first grown, followed by the deposition of an 80-nm-thick ScAlN under N-rich growth conditions. In order to match the underlying GaN lattice, the Sc content was controlled to be about 18%, confirmed utilizing X-ray photoelectron spectroscopy (XPS) and electron dispersive spectroscopy (EDS) in STEM. For standard metal/ScAlN/GaN capacitor fabrication, 50/100 nm thick Ti/Au circular pads with a diameter of 20–100 μm were lithographically patterned on the ScAlN surface as top electrodes, with the n+-GaN layer serving as the bottom electrode.
Characterization
The surface morphology of ScAlN was characterized using a Dimension ICON atomic force microscope (AFM). XRD analysis was performed with a Philips PANalytical X’pert high-resolution XRD system equipped with a 1.54 Å Cu Κα1 X-ray source. Ferroelectric measurements were conducted with a Radiant Premier II Ferroelectric Tester. The P-E hysteresis loop was measured using a triangular wave with a frequency of 1 kHz. The sample lamellae used for STEM measurements were prepared with a ThermoFisher Helios G4 UX focused ion beam (FIB). HAADF- and ABF-STEM images were captured using a JEM-ARM300F2 aberration-corrected STEM at an acceleration voltage of 300 kV, with pixel integration times ranging from 0.5 to 2 μs.
Polarization calculations
See details in Supplementary Information.
Supplementary information
Source data
Acknowledgements
This work was supported by the National Key Research and Development Program of China (No. 2023YFB3610400 to P.W.), the National Natural Science Foundation of China (NSFC) (No. 62321004 to X.W.; 62374002 to P.W.; 62304008 to R.W.; 62374010 to T.W.; 12304218 to W.T.), the Natural Science Foundation of Beijing Municipality (No. Z230024 to P.W.), the Research Fund of Suzhou Laboratory (No. SK-1202-2024-012 to P.W.), the China Postdoctoral Science Foundation (No. 2023T160016 to R.W.), Shanghai Pujiang Program (No. 23PJ1402200 to C.D.), and the State Key Laboratory of Artificial Microstructure and Mesoscopic Physics at Peking University.
Author contributions
H.Y., P.W., and X.W. conceived the original concept and initiated the project. H.Y., P.W., R.F., and B.A. conducted the material growth and performed the AFM, XRD and XPS analyses. R.W. was responsible for the device fabrication and electrical measurements. J.W., X.X., and T.W. conducted the STEM measurements, with H.Y., P.W., J.W., X.X., and T.W. analyzing the STEM data. X.X. and H.Y. carried out the local polarization measurements. W.T. and C.D. performed the DFT calculations. H.Y., P.W., W.T., F.L., B. Sheng, W.M., H.L., Z.C., W.G., B. Shen, and X.W. contributed to the polarization analysis. H.Y., P.W., and X.W. drafted the manuscript, with all authors contributing to the discussions of the results and providing feedback on the manuscript at all stages. P.W. and X.W. supervised the research.
Peer review
Peer review information
Nature Communications thanks the anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.
Data availability
All data in this work are available within the main text and Supplementary Information files, as well as available from the corresponding author. Source data are provided in this paper. Source data are provided with this paper.
Competing interests
The authors declare no competing interests.
Footnotes
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Contributor Information
Ping Wang, Email: pingwang@pku.edu.cn.
Tao Wang, Email: cwwangtao@pku.edu.cn.
Xinqiang Wang, Email: wangshi@pku.edu.cn.
Supplementary information
The online version contains supplementary material available at 10.1038/s41467-025-58975-0.
References
- 1.Fiorentini, V., Vanderbilt, D. & Bernardini, F. Spontaneous polarization and piezoelectric constants of III-V nitrides. Phys. Rev. B.56, R10024–R10027 (1997). [Google Scholar]
- 2.Fiorentini, V., Vanderbilt, D. & Bernardini, F. Accurate calculation of polarization-related quantities in semiconductors. Phys. Rev. B.63, 193201 (2001). [Google Scholar]
- 3.Ambacher, O. et al. Two-dimensional electron gases induced by spontaneous and piezoelectric polarization charges in N- and Ga-face AlGaN/GaN heterostructures. J. Appl. Phys.85, 3222–3233 (1999). [Google Scholar]
- 4.Chaudhuri, R. et al. A polarization-induced 2D hole gas in undoped gallium nitride quantum wells. Science365, 1454–1457 (2019). [DOI] [PubMed] [Google Scholar]
- 5.van Deurzen, L. et al. Using both faces of polar semiconductor wafers for functional devices. Nature634, 334–340 (2024). [DOI] [PMC free article] [PubMed]
- 6.Simon, J., Protasenko, V., Lian, C., Xing, H. & Jena, D. Polarization-induced hole doping in wide–band-gap uniaxial semiconductor heterostructures. Science327, 60–64 (2010). [DOI] [PubMed] [Google Scholar]
- 7.Takeuchi, T. et al. Quantum-confined Stark effect due to piezoelectric fields in gainn strained quantum wells. Jpn. J. Appl. Phys.36, L382–L385 (1997). [Google Scholar]
- 8.Nakamura, S. Nobel lecture: background story of the invention of efficient blue InGaN light emitting diodes. Rev. Mod. Phys.87, 1139–1151 (2015). [DOI] [PubMed] [Google Scholar]
- 9.Amano, H. Nobel lecture: growth of GaN on sapphire via low-temperature deposited buffer layer and realization of p-type GaN by Mg doping followed by low-energy electron beam irradiation. Rev. Mod. Phys.87, 1133–1138 (2015). [DOI] [PubMed] [Google Scholar]
- 10.Amano, H. et al. The 2018 GaN power electronics roadmap. J. Phys. D: Appl. Phys.51, 163001 (2018). [Google Scholar]
- 11.Fichtner, S., Wolff, N., Lofink, F., Kienle, L. & Wagner, B. AlScN: A III-V semiconductor based ferroelectric. J. Appl. Phys.125, 114103 (2019). [Google Scholar]
- 12.Wang, P. et al. Fully epitaxial ferroelectric ScAlN grown by molecular beam epitaxy. Appl. Phys. Lett.118, 223504 (2021). [Google Scholar]
- 13.Wolff, N. et al. Demonstration and STEM analysis of ferroelectric switching in MOCVD-grown single crystalline Al0.85Sc0.15N. Adv. Phys. Res.3, 2300113 (2024). [Google Scholar]
- 14.Heron, J. & Mi, Z. Emerging Ferroelectric Materials and Devices (Elsevier, 2023).
- 15.Schönweger, G. et al. From fully strained to relaxed: epitaxial ferroelectric Al1-xScxN for III-N technology. Adv. Funct. Mater.32, 2109632 (2022). [Google Scholar]
- 16.Ambacher, O. et al. Wurtzite ScAlN, InAlN, and GaAlN crystals, a comparison of structural, elastic, dielectric, and piezoelectric properties. J. Appl. Phys.130, 45102 (2021). [Google Scholar]
- 17.Resta, R. & Vanderbilt, D. In: Ascheron, C. E. $ Skolaut, W., editor. Physics of Ferroelectrics: A Modern Perspective 31–68 (Springer Berlin Heidelberg, 2007).
- 18.Resta, R. Theory of the electric polarization in crystals. Ferroelectrics136, 51–55 (1992). [Google Scholar]
- 19.Resta, R. Modern theory of polarization in ferroelectrics. Ferroelectrics151, 49–58 (1994). [Google Scholar]
- 20.Spaldin, N. A. A beginner’s guide to the modern theory of polarization. J. Solid State Chem.195, 2–10 (2012). [Google Scholar]
- 21.Dreyer, C. E., Janotti, A., Van de Walle, C. G. & Vanderbilt, D. Correct implementation of polarization constants in wurtzite materials and impact on III-nitrides. Phys. Rev. X.6, 21038 (2016). [Google Scholar]
- 22.Yassine, M. et al. Modeling of polarization reversal-induced interface sheet charge in wurtzite-type AlScN/GaN heterostructures. J. Appl. Phys.135, 155702 (2024). [Google Scholar]
- 23.Wang, D., Wang, D., Yang, S. & Mi, Z. Rethinking polarization in wurtzite semiconductors. Appl. Phys. Lett.124, 263502 (2024). [Google Scholar]
- 24.Fichtner, S., Yassine, M., Van de Walle, C. G. & Ambacher, O. Clarification of the spontaneous polarization direction in crystals with wurtzite structure. Appl. Phys. Lett.125, 40501 (2024). [Google Scholar]
- 25.Benbedra, A., Meskine, S., Boukortt, A., Abbassa, H. & Abbes, E. H. Polarization properties of wurtzite III-nitride alloys using the hexagonal reference structure. ECS J. Solid State Sci. Technol.12, 103008 (2023). [Google Scholar]
- 26.King-Smith, R. D. & Vanderbilt, D. Electric polarization as a bulk quantity and its relation to surface charge. Phys. Rev. B.48, 4442–4455 (1993). [DOI] [PubMed] [Google Scholar]
- 27.Calderon, S., Funni, S. D. & Dickey, E. C. Accuracy of local polarization measurements by scanning transmission electron microscopy. Microsc. Microanal.28, 2047–2058 (2022). [DOI] [PubMed] [Google Scholar]
- 28.Calderon, S. et al. Atomic-scale polarization switching in wurtzite ferroelectrics. Science380, 1034–1038 (2023). [DOI] [PubMed] [Google Scholar]
- 29.Wolff, N. et al. Atomic scale confirmation of ferroelectric polarization inversion in wurtzite-type AlScN. J. Appl. Phys.129, 34103 (2021). [Google Scholar]
- 30.Ye, H. et al. ScAlInN/GaN heterostructures grown by molecular beam epitaxy. Appl. Phys. Lett.125, 122105 (2024). [Google Scholar]
- 31.Zhang, S., Fu, W. Y., Holec, D., Humphreys, C. J. & Moram, M. A. Elastic constants and critical thicknesses of ScGaN and ScAlN. J. Appl. Phys.114, 243516 (2013). [Google Scholar]
- 32.Motoki, K. et al. Improved crystallographic order of ScAlN/GaN heterostructures grown at low temperatures under metal rich surface conditions. J. Appl. Phys.135, 135105 (2024). [Google Scholar]
- 33.Zhu, W. et al. Strongly temperature dependent ferroelectric switching in AlN, Al1-xScxN, and Al1-xBxN thin films. Appl. Phys. Lett.119, 62901 (2021). [Google Scholar]
- 34.Hasegawa, K., Shimizu, T., Ohsawa, T., Sakaguchi, I. & Ohashi, N. Full polarization reversal at room temperature in unsubstituted AlN. Appl. Phys. Lett.123, 192903 (2023). [Google Scholar]
- 35.Devonshire, A. F. Theory of ferroelectrics. Adv. Phys.3, 85–130 (1954). [Google Scholar]
- 36.Rabe, K. M., Ahn, C. H. & Triscone, J. Physics of fErroelectrics: A Modern Perspective Vol. 105 (Springer Science & Business Media, 2007).
- 37.Rabe, K. M., Dawber, M., Lichtensteiger, C., Ahn, C. H. & Triscone, J. in Physics of Ferroelectrics: A Modern Perspective 1–30 (Springer Berlin Heidelberg, 2007).
- 38.Brandt, O. et al. Direct experimental determination of the spontaneous polarization of GaN. Phys. Rev. B.86, 81302 (2012). [Google Scholar]
- 39.Wang, P. et al. Dawn of nitride ferroelectric semiconductors: from materials to devices. Semicond. Sci. Technol.38, 43002 (2023). [Google Scholar]
- 40.Kim, K., Karpov, I., Olsson, R. H. & Jariwala, D. Wurtzite and fluorite ferroelectric materials for electronic memory. Nat. Nanotechnol.18, 422–441 (2023). [DOI] [PubMed] [Google Scholar]
- 41.Wang, H., Adamski, N., Mu, S. & Van de Walle, C. G. Piezoelectric effect and polarization switching in Al1-xScxN. J. Appl. Phys.130, 104101 (2021). [Google Scholar]
- 42.Martin, R. M. Piezoelectricity. Phys. Rev. B.5, 1607–1613 (1972). [Google Scholar]
- 43.Lax, M. & Nelson, D. F. Linear elasticity and piezoelectricity in pyroelectrics. Phys. Rev. B.13, 1785–1796 (1976). [Google Scholar]
- 44.Vanderbilt, D. Berry-phase theory of proper piezoelectric response. J. Phys. Chem. Solids61, 147–151 (2000). [Google Scholar]
- 45.Urban, D. F., Ambacher, O. & Elsässer, C. First-principles calculation of electroacoustic properties of wurtzite (Al,Sc)N. Phys. Rev. B.103, 115204 (2021). [Google Scholar]
- 46.Wang, D. et al. Fully epitaxial, monolithic ScAlN/AlGaN/GaN ferroelectric HEMT. Appl. Phys. Lett.122, 90601 (2023). [Google Scholar]
- 47.Íñiguez, J., Zubko, P., Luk Yanchuk, I. & Cano, A. Ferroelectric negative capacitance. Nat. Rev. Mater.4, 243–256 (2019). [Google Scholar]
- 48.Wang, R. et al. Composition-graded nitride ferroelectrics based multi-level non-volatile memory for neuromorphic computing. Adv. Mater. 37, e2414805 (2024). [DOI] [PubMed]
- 49.Wang, D. et al. Ultrathin nitride ferroic memory with large ON/OFF ratios for analog in-memory computing. Adv. Mater.35, 2210628 (2023). [DOI] [PubMed] [Google Scholar]
- 50.Wang, D. et al. Electric-field-induced domain walls in wurtzite ferroelectrics. Naturehttps://www.nature.com/articles/s41586-025-08812-7 (2025). [DOI] [PubMed]
- 51.Wang, D. et al. An epitaxial ferroelectric ScAlN/GaN heterostructure memory. Adv. Electron. Mater.8, 2200005 (2022). [Google Scholar]
- 52.Wang, P. et al. Interfacial modulated lattice-polarity-controlled epitaxy of iii-nitride heterostructures on Si(111). Acs Appl. Mater. Interfaces14, 15747–15755 (2022). [DOI] [PubMed] [Google Scholar]
- 53.Hu, M. et al. Heteroepitaxy of N-polar AlN on C-face 4H-SiC: structural and optical properties. Apl. Mater.11, 121111 (2023). [Google Scholar]
- 54.Wang, T. et al. High-mobility two-dimensional electron gas at InGaN/InN heterointerface grown by molecular beam epitaxy. Adv. Sci.5, 1800844 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
All data in this work are available within the main text and Supplementary Information files, as well as available from the corresponding author. Source data are provided in this paper. Source data are provided with this paper.