Semiconducting nonperovskite ferroelectric oxynitride designed ab initio

TL;DR

This paper reports the discovery of new nonperovskite ferroelectric oxynitrides, TaON and NbON, with promising electronic and piezoelectric properties suitable for multifunctional applications, guided by ab initio calculations.

Contribution

It introduces a novel class of ferroelectric oxynitrides with unique properties, expanding the design space for ferroelectric semiconductors using ab initio methods.

Findings

TaON and NbON exhibit large remnant polarization.

They have low switching barriers and negative piezoelectric effects.

Both materials can absorb visible light and have high charge mobility.

Abstract

Recent discovery of HfO2-based and nitride-based ferroelectrics that are compatible to the semiconductor manufacturing process have revitalized the field of ferroelectric-based nanoelectronics. Guided by a simple design principle of charge compensation and density functional theory calculations, we discover HfO2-like mixed-anion materials, TaON and NbON, can crystallize in the polar Pca21 phase with a strong thermodynamic driving force to adopt anion ordering spontaneously. Both oxynitrides possess large remnant polarization, low switching barriers, and unconventional negative piezoelectric effect, making them promising piezoelectrics and ferroelectrics. Distinct from HfO2 that has a wide band gap, both TaON and NbON can absorb visible light and have high charge carrier mobilities, suitable for ferroelectric photovoltaic and photocatalytic applications. This new class of multifunctional nonperovskite oxynitride containing economical and environmentally benign elements offer a platform to design and optimize high-performing ferroelectric semiconductors for integrated systems.

Tables

Table 1. Table 1: Energy ( Δ ​ E Δ 𝐸 \Delta E in meV/atom) and lattice constants (in Å) of TaON and NbON computed with PBE. The energy of P ​ c ​ a ​ 2 1 𝑃 𝑐 𝑎 subscript 2 1 Pca2_{1} -N np is chosen as the reference. The polar axis in P ​ c ​ a ​ 2 1 𝑃 𝑐 𝑎 subscript 2 1 Pca2_{1} is along the c 𝑐 c -axis. As a comparison, the energy of P ​ 2 1 / c 𝑃 subscript 2 1 𝑐 P2_{1}/c and P ​ b ​ c ​ a 𝑃 𝑏 𝑐 𝑎 Pbca of HfO 2 is − 28 28 -28 and − 4 4 -4 meV/atom, respectively, relative to P ​ c ​ a ​ 2 1 𝑃 𝑐 𝑎 subscript 2 1 Pca2_{1} . Experimental lattice parameters for the P ​ 2 1 / c 𝑃 subscript 2 1 𝑐 P2_{1}/c -N np configuration are reported in the parentheses  63 , 64 .

TaON	$Pca{2}_1$-Nnp	$Pca{2}_1$-Np	$Pca{2}_1$-Ndiag	$Pca{2}_1$-Ntetra	$P{2}_1/c$-Nnp	$Pbca$-Nnp
$\Delta E$	0	28	54	91	$-63$	$-1$
$a$	5.16	5.81	5.28	5.18	5.23 (5.18)	5.14
$b$	4.91	4.80	5.12	4.92	5.01 (4.96)	9.78
$c$	5.03	5.11	5.01	5.01	5.07 (5.03)	5.06
$\alpha$	90	90	90	90	90 (90)	90
$\beta$	90	90	99.8	90	90 (90)	90
$\gamma$	90	90	90	89.0	99.7 (99.6)	90
NbON	$Pca{2}_1$-Nnp	$Pca{2}_1$-Np	$Pca{2}_1$-Ndiag	$Pca{2}_1$-Ntetra	$P{2}_1/c$-Nnp	$Pbca$-Nnp
$\Delta E$	0	unstable	48	92	$-52$	5
$a$	5.19	-	5.28	5.18	5.22 (5.19)	5.14
$b$	4.90	-	5.11	4.88	5.00 (4.97)	9.74
$c$	5.02	-	5.01	5.03	5.06 (5.03)	5.05
$\alpha$	90	-	90	90	90 (90)	90
$\beta$	90	-	100.0	90	90 (90)	90
$\gamma$	90	-	90	90.7	99.8 (100.2)	90

Table 2. Table 2: Spontaneous polarization along the c 𝑐 c -axis ( P s subscript 𝑃 𝑠 P_{s} in μ 𝜇 \mu C/cm 2 ), piezoelectric coefficients ( d 33 subscript 𝑑 33 d_{33} in pm/V and e 33 subscript 𝑒 33 e_{33} in C/m 2 ), electron and hole effective masses ( m e ∗ superscript subscript 𝑚 𝑒 m_{e}^{*} and m h ∗ superscript subscript 𝑚 ℎ m_{h}^{*} in electron mass m 0 subscript 𝑚 0 m_{0} ), band gaps computed with PBE and eACBN0 ( E g PBE superscript subscript 𝐸 𝑔 PBE E_{g}^{\rm PBE} and E g eACBN0 superscript subscript 𝐸 𝑔 eACBN0 E_{g}^{\rm eACBN0} in eV), and Born effective charge tensors ( Z 𝑍 Z ) for O, N, and transition metal ( M 𝑀 M ) in ferroelectric P ​ c ​ a ​ 2 1 𝑃 𝑐 𝑎 subscript 2 1 Pca2_{1} -N np M 𝑀 M ON ( M 𝑀 M =Ta, Nb).

	$P_s$		$d_{33}$	$e_{33}$	$m_e^{\ast }$	$m_h^{\ast }$	$E_g^{\mathrm{PBE}}$	$E_g^{\mathrm{eACBN0}}$
TaON	51.7		$-$7.6	$-$1.70	0.435	$-$0.414	1.86	2.31
NbON	54.0		$-$5.9	$-$0.85	0.425	$-$0.432	1.32	1.75
	$Z_N$			$Z_O$			$Z_M$
	$Z_{11}$	$Z_{22}$	$Z_{33}$	$Z_{11}$	$Z_{22}$	$Z_{33}$	$Z_{11}$	$Z_{22}$	$Z_{33}$
TaON	$-$4.11	$-$2.56	$-$3.84	$-$2.77	$-$3.97	$-$2.11	6.87	6.53	5.94
NbON	$-$4.02	$-$2.71	$-$4.13	$-$2.50	$-$4.36	$-$2.18	6.51	7.07	6.30

Full text

\altaffiliation

Contributed equally to this work

\altaffiliationContributed equally to this work

\alsoaffiliationInstitute of Natural Sciences, Westlake Institute for Advanced Study, Hangzhou, Zhejiang 310024, China

\alsoaffiliationInstitute of Natural Sciences, Westlake Institute for Advanced Study, Hangzhou, Zhejiang 310024, China

The discovery of ferroelectricity in HfO2-based thin films 1 is revolutionizing the research and development for ferroelectric-based electronic devices. The demonstrated nanoscale ferroelectricity 2 and excellent compatibility with the modern complementary metal oxide semiconductor (CMOS) technology make HfO2-based ferroelectrics the leading candidate material for next-generation nonvolatile information storage technology 3, 4. The realization of ferroelectricity in this bindary oxide also offers several conceptual breakthroughs. First, despite the admittedly importance of perovskite ferroelectrics that often contain 3$d$ elements such as Ti, nonperovskite oxides containing 5$d$ elements like Hf could afford great potential for new silicon-compatible ferroelectrics 3. The finding that the polar orthorhombic phase (PO) of HfO2 (space group $Pca2_{1}$) responsible for the ferroelectricity in thin films is actually higher in energy than the bulk monoclinic (M) phase (space group $P2_{1}/c$) 5, 6 implies the importance of exploring metastable ferroelectric materials.

Another progress in the search of ferroelectrics for semiconductor process integration is the successful synthesis of nitride ferroelectrics represented by nitride perovskites 7 and doped III-V piezoelectrics 8, 9. Though many nitride perovskites have long been predicted to be thermodynamically stable based on first-principles density functional theory (DFT) calculations 10, 11, it was only until very recently that high-quality polycrystalline thin films of pure nitride perovskite, lanthanum tungsten nitride (LaWN3), was synthesized with physical vapor deposition 7. Previous DFT calculations predicted a switchable polarization of 61 $\mu$C/cm2 in LaWN3 10. The synthesized LaWN3 indeed has a polar symmetry and exhibits a piezoelectric response stronger than other known nitrides. Different from the strategy of stabilizing the polar phase in HfO2-based ferroelectrics and LaWN3 by fine tuning synthesis parameters, the occurrence of ferroelectricity in AlN-based ferroelectrics is to destabilize AlN that is already in a polar phase 8, 9. For example, by doping Sc into AlN, a switchable polarization (80–110 $\mu$C/cm2) has been realized in Al1-xScxN 8. Taking advantage of decades of advances in nitride semiconductor technology, it is envisioned that the integration of nitride ferroelectrics with the mainstream semiconductor industry could enable novel device types for widespread applications such as nonvolatile memory, neuromorphic electronics, and negative capacitance transistors 12.

Both hafnia-based and nitride-based ferroelectrics open up opportunities for incorporating ferroelectric functionalities into integrated circuits. However, there are few of them. This leads us to the question of how to obtain more CMOS-compatible ferroelectrics. DFT-based materials modeling offers a great platform to discover and design new ferroelectric materials with desired properties. Recent theoretical studies have predicted a few oxynitride-based ferroelectric materials, such as Ca3Nb2N2O5 13, YGeO2N, LaSiO2N, and LaGeO2N 14. The combination of predictive high-throughput computations and the tools of data science further speeds up the material discovery 15, 16, 17. After a promising material is predicted in theory, an immediate question is how to synthesize it in experiments; it often took months if not years of efforts to figure out the synthesizing protocol for a completely new compound. In this work, we aim to design new ferroelectrics that have structures similar to the polar phase of HfO2 based on experimentally synthesized compounds. Guided by a simple design principle of charge compensation and using the experimentally synthesized ferroelectric phase of HfO2 as the template, we discover that the nonpolar monoclinic TaON and NbON that have already been synthesized can crystallize in the polar $Pca2_{1}$ phase that is dynamically stable, as supported by phonon spectrum calculations and ab initio molecular dynamic (AIMD) simulations. Based on first-principles DFT calculations, we demonstrate that $Pca2_{1}$ TaON and NbON have low switching barriers of $\approx$0.25 eV, large spontaneous polarization of $\approx$50 $\mu$C/cm2, and direct band gaps of $\approx$2 eV, making them intrinsic semiconducting ferroelectrics with nontoxic elements. Owing to the dispersed 4$d$ and 5$d$ orbitals, both oxynitrides acquire light electron and hole effective masses and high carrier mobility of $\approx$400 cm2V*-1s-1*, competing favorably with superstar halide perovskites such as CH3NH3PbI3 18, 19. These highly polarized and visible photoresponsive oxynitrides have great technological potential for use in ferroelectric photovoltaics 20, 21, 22, 23, multienergy harvester, and photocatalyst applications.

We use the ferroelectric $Pca2_{1}$ phase of HfO2 as the template structure. In a simple ionic picture, after replacing all Hf4+ cations with Ta5+, an element sitting right next to hafnium in the periodic table, we need to replace half of the anions (O2-) with more-negatively-charged anions such as N3- in order to maintain the charge neutrality. This leads to a ternary mixed-anion oxynitride, TaON. Similarly, NbON is chosen as another candidate as TaON and NbON possess the same polymorphs. Before performing any calculations, we present a few evidences justifying this simple design principle. First, the bulk phase of HfO2 adopts the $P2_{1}/c$ monoclinic phase, while TaON and NbON in the same space group have already been synthesized in experiments and extensively studied because of high photocatalytic activity under visible light 24, 25, 26. It is thus worthy to explore whether TaON and NbON could crystallize in the polar $Pca2_{1}$ phase, similar to that in HfO2. Additionally, the $Pca2_{1}$ phase of HfO2 contains two symmetry-inequivalent oxygen atoms: a three-fold coordinated oxygen atom that has polar local displacement (Op) and a fourfold-coordinated nonpolar oxygen (Onp), as illustrated in Fig. 1a. Since N3- anions are more negatively charged than O2-, they expectedly will occupy fourfold-coordinated Onp sites, leading to a spontaneous ordering of N3- anions without affecting polar oxygen atoms. We computed the energy difference between a unit cell of $Pca2_{1}$ TaO1.75N0.25 (NbO1.75N0.25) that has one N atom occupying the Op site and another configuration with one N atom occupying the Onp site. We found that the Onp-site N substitution is 0.95 eV lower in energy than the Op-site N substitution for TaO1.75N0.25, and 2.04 eV for NbO1.75N0.25. This confirms that N atoms have a strong tendency to replace the $np$-site O atoms. Such anion ordering in monoclinic TaON has been suggested based on neutron and synchrotron powder diffraction and DFT studies 25.

First-principles DFT calculations are performed using QUANTUM ESPRESSO  27, 28 package with Garrity-Bennett-Rabe-Vanderbilt (GBRV) ultrasoft pseudopotentials 29. The exchange-correlation function is Perdew-Burke-Ernzerhof (PBE) functional of generalized gradient approximation. The lattice constants and atomic positions are optimized with a plane-wave cutoff of 50 Ry and a charge density cutoff of 250 Ry. A $4\times 4\times 4$ $k$-point grid centered on $\Gamma$ point is used for Brillouin zone sampling. The total energy and force convergence threshold is set to 10*-5* Ry and 10*-4* Ry/Bohr, respectively. The phonon spectra are calculated using the finite displacement method implemented in PHONOPY 30 code in conjunction with QUANTUM ESPRESSO. The structural stability at finite temperatures is gauged by $ab~{}initio$ molecular dynamics (AIMD) simulations implemented in Vienna $ab~{}initio$ simulation package (VASP) 31, 32. We construct a $3\times 3\times 3$ supercell for AIMD simulations using a $\Gamma$-point sampling, an energy cutoff of 350 eV and an energy convergence threshold of 10*-5* eV. The Nosé-Hoover thermostat is applied for temperature controlling. The minimum energy paths (MEPs) of polarization reversal are determined using the variable-cell nudged elastic band (NEB) technique implemented in USPEX code 33, 34, 35. Noted that VCNBE allows strain relaxation during the polarization switching process 36. The MEP is considered to reach convergence when the root-mean-square forces are lower than 0.03 eV/Å on images or the energy barrier remains unchanged for successive 10 steps. The carrier mobility is estimated using semiclassical Boltzmann transport using Boltztrap 37 within the constant relaxation time ($\tau$=0.2 ps) approximation,.

Different from the oxygen atoms in perovksite oxides that often possess similar local chemical environments, the Op  and Onp sites have distinct local bonding characteristics. We investigate the effect of anion ordering on the relative thermodynamic stability of different atomic configurations of $Pca2_{1}$ TaON. For a 12-atom unit cell, there are five possible configurations, Nnp, Np, Ndiag, Nplane, and Ntetra, that are named based on the ordering of N atoms (Fig. 1b-f). As expected, the N atom has a strong tendency to occupy the fourfold-coordinated Onp site. The $Pca2_{1}$-Nnp configuration that has all N atoms occupying Onp sites is the most stable, followed by the $Pca2_{1}$-Np configuration ($\Delta E$= 28 meV/atom) that has N atoms located at Op sites, whereas all the other configurations are much higher in energy (Table 1). For example, $Pca2_{1}$-Nplane is unstable such that it spontaneously evolves to the $P2_{1}/c$ phase with all N atoms being fourfold-coordinated to tantalum after the structural optimization. Such strong tendency of O/N anion ordering is beneficial for the emergence of ferroelectricity as the polar oxygen atom in $Pca2_{1}$-Nnp TaON acquires a similar local environment to that in ferroelectric HfO2. The spontaneous O/N anion ordering also presents in NbON (Table 1). We note that the monoclinic $P2_{1}/c$-Nnp configuration remains the most stable thermodynamically for TaON and NbON. As we discussed above, an important lesson learned from the discovery of ferroelectric HfO2 is that stabilizing a high-energy metastable phase in thin films is feasible experimentally. Therefore, we argue that it is worthwhile to attempt synthesizing $Pca2_{1}$-Nnp TaON and NbON (if they are dynamically stable, see discussions below).

For HfO2, the antiferroelectric-like $Pbca$ phase is the second most stable 5 and is 4 meV/atom lower in energy than $Pca2_{1}$ 38. In comparison, the $Pbca$-Nnp and $Pca2_{1}$-Nnp phases in TaON have comparable energies (difference within 1 meV/atom). The $Pca2_{1}$-Nnp phase of NbON becomes more stable than $Pbca$-Nnp. This is probably helpful for the stabilization of the metastable $Pca2_{1}$-Nnp phase in NbON. Hereafter, unless otherwise specified, all results are for configuration $Pca2_{1}$-Nnp.

The structural stability of the $Pca2_{1}$ phase of TaON and NbON is examined by computing the phonon spectrum. A (meta)stable material situating at a local minimum of the potential energy surface will have all positive phonon frequencies. The phonon spectra of $Pca2_{1}$ TaON and NbON reveal no imaginary frequencies in the whole Brillouin zone (Fig. 2a-b), confirming the dynamical stability of both compounds. This is similar to HfO2 where several metastable phases such as $P2_{1}/c$, $Pbca$, $P4_{2}/nmc$ that are higher in energy than the $P2_{1}/c$ phase are all dynamically stable as confirmed by phonon spectra exhibiting no imaginary modes 39. We further perform AIMD simulations at elevated temperatures to check whether both compounds will be stable against larger atomic distortions due to thermal fluctuations. Figure 2c plots the fluctuations of the total energy for $Pca2_{1}$ TaON (NbON) at 400, 600, and 800 K during AIMD simulations, showing no sign of structural destruction or reconstruction to other phases. This serves as a strong evidence supporting the room-temperature stability of $Pca2_{1}$ TaON and NbON. Additional AIMD simulations reveal that TaON becomes nonpolar when the temperature is above 2000 K. We analyze the distribution of local displacements ($\Delta z$) of oxygen atoms along the polar axis relative to the nearest Hf atomic plane using configurations sampled from a 10 ps equilibrium trajectory. With increasing temperature, the peak position shifts toward lower values, indicating the polar oxygen atoms move closer to their nearest Hf plan es.

The switching barriers are computed using the VCNEB method that takes into account the strain relaxation effect during the switching 36. There are two possible switching pathways in ferroelectric HfO2, the shift-inside (SI) pathway that has Op atoms moving inside two Hf atomic planes (Fig. 3a) and the shift-across (SA) pathway that has Op atoms moving across the Hf atomic plane (Fig. 3b) 40, 41. Consistent with previous studies 40, the SI barrier in HfO2 is 0.38 eV, much lower than the SA barrier of 0.80 eV. Interestingly, the opposite trend is found in TaON and NbON: the SA pathway becomes kinetically favored over the SI pathway. This is consistent with AIMD simulations in which polar oxygen atoms tend to move toward their nearest Hf atomic planes with increasing temperatures. Importantly, both oxynitrides are switchable as the SA barrier is $\approx$0.25 eV, comparable to the switching barrier of prototypical ferroelectric like PbTiO3 (0.17 eV) 36. These results demonstrate the presence of switchable polarization in $Pca2_{1}$ TaON and NbON, a hallmark feature of ferroelectricity.

We determine the magnitude of the spontaneous polarization with the Berry-phase approach 42, 43 by tracking the change in Berry phase during the SA pathway. As shown in Fig. 3c, both TaON and NbON have large and nearly identical magnitudes of polarization, $\approx$50 $\mu$C/m2. The Born effective charge tensors ($Z$) of transition metal, oxygen, and nitrogen atoms are reported in Table 2. An interesting observation is that for the $Z_{33}$ component, transition metal ($M$) and nitrogen atoms have effective charges (\approx$$+6 for $M$ and \approx$$-4 for N) larger than their nominal ionic valence ($+5$ for $M$ and $-3$ for N), while the O ion acquires a value of $\approx-2.1$, close to the nominal value. In comparison, the $Z_{33}$ value of O in perovskite ferroelectric BaTiO3 is anomalously large ($-5.7$) 44, a signature of strong charge transfer and covalency. We suggest that the $M$-N bond is likely more covalent that the $M$-O bond along the polar axis.

Additionally, the SA pathway is associated with an unconventional polarization-strain coupling that has the strain along the polar axis ($\eta_{c}$) increasing with reducing polarization, hinting at negative longitudinal piezoelectric effect 45. Indeed, TaON has $e_{33}$ of $-1.70$ C/m2 and $d_{33}$ of $-7.63$ pm/V, while NbON has $e_{33}$ of $-0.85$ C/m2 and $d_{33}$ of $-5.92$ pm/V, all possessing a negative sign and a decent magnitude that compares well to those of commercial piezoelectrics such as Al0.92Sc0.08N ($\approx$10 pm/V) and LiNbO3 ($\approx$10 pm/V) 7. The piezoelectric coefficient can be decomposed into the clamped-ion contribution (fixed ion with vary strains) and the internal-strain contribution arising from ion relaxations 46. Our calculations suggest that the clamped-ion ($-0.17$ C/m2 for NbON and $-1.24$ C/m2 for TaON) and internal-strain contributions ($-0.48$ C/m2 for NbON and $-0.62$ C/m2 for TaON) are both negative, thus leading to a macroscopic negative piezoelectric response. By combining piezoelectrics possessing negative piezoelectric effect with conventional piezoelectrics, it is possible to achieve on-demand design of $e_{33}$ and precise control of the electromechanical response 45.

The PBE band structures of TaON and NbON in the $Pca2_{1}$ phase are shown in Fig. 4a-b, revealing a direct band gap of 1.86 and 1.32 eV, respectively, at the X point. The projected density of states (PDOS) indicate that bands near the valence band maximum (VBM) take mostly N-$2p$ character with some contributions from O-$2p$ and Ta-$5d$ (Nb-$4d$) characters while those near the conduction band minimum (CBM) mainly consist of $d$-orbitals of the transition metal. We estimate the effective masses of electrons ($e$) and holes ($h$) based on the dispersion relationship along X-S with results reported in Table 2. Owing to the spatially more dispersed $5d$ ($4d$) orbitals, the effective electron and hole masses are almost identically light in ferroelectric TaON (NbON), both at a value of $\approx$0.4. This resembles to that in tetragonal CH3NH3PbI3 47 and is beneficial for balanced transport between the electrons and holes. The theoretical spatially averaged charge carrier mobility, $\mu$, is plotted in Fig. 4c as a function of charge carrier concentration ($\rho_{e}$ and $\rho_{h}$). At a low carrier concentration ($\rho<10^{19}$/cm3), the electron and hole mobilities ($\mu_{e}$ and $\mu_{h}$) are insensitive to $n$. Notably, we predict $\mu_{h}$ of 455 and $\mu_{e}$ of 332 cm2V*-1s-1* in TaON, and $\mu_{h}$ of 464 and $\mu_{e}$ of 301 cm2V*-1s-1* in NbON. These values are at least one order of magnitude higher than those ($<10$ cm2V*-1s-1*) in CH3NH3PbI3 estimated with the same method that assumes a longer carrier relaxation time of 1 ps 48.

It is well known that PBE underestimates the band gap due to the the remnant self-interaction error in the approximation to the exchange-correlation functional. To address this issue, we further compute the band structures using the newly developed pseudohybrid Hubbard density functional, extend Agapito–Cuetarolo–Buongiorno Nardelli (eACBN0) 49, 50, 51. The eACBN0 function is essentially a DFT+$U$+$V$ method with self-consistently computed Hubbard $U$ ($V$) parameters that account for the onsite (intersite) Coulomb interaction. Particularly for covalent semiconductors (e. g., Si and GaAs), the band gaps predicted with eACBN0 are much better than PBE values, and are comparable with advanced methods such as the Heyd-Scuseria-Ernzerhof (HSE) hybrid density functional and $GW$ approximations but at a PBE-level computational cost 52, 50, 53. As shown in Fig. 4a-b, eACBN0 predicts a larger band gap of 2.31 eV for TaON and 1.75 eV for NbON than PBE, while both functionals predict similar band dispersion relationships (and thus effective masses and carrier mobilities). Therefore, TaON and NbON are indeed intrinsic ferroelectric semiconductors suited to absorb visible light. Additionally, we found that the structural parameters obtained with PBE and eACBN0 are comparable. Interestingly, PBE and eACBN0 predict nearly identical $e_{33}$ for NaON, while eACBN0 predicts a lower magnitude of $e_{33}$ ($-0.80$ C/m2) than PBE ($-1.70$ C/m2) for TaON.

The direct-band-gap nature of proposed oxynitrides makes them better solar absorbers than Si which is known to have poor absorption of low-energy photons below the direct band gap 54. We compare the PBE absorption spectra of TaON, NbON, Si, and CH3NH3PbI3 in Fig. 4d. It is evident that both TaON and NbON possess superior absorption performances than Si in the visible region, and NbON competes favorably with CH3NH3PbI3. Utilizing well-developed epitaxy techniques, TaON and NbON could be made as thin-film solar absorbers for efficient light absorption. Compared with conventional perovskite ferroelectric materials whose band gaps are around 3 eV, the band gaps of TaON and NbON are relatively small and may be associated with larger leakage currents at high electric fields. We suggest that these two ferroelectric semiconductors are probably more suitable for photovoltaic-related applications. Further studies are needed to resolve the leakage current issue.

Finally, compared with defect-engineered semiconducting ferroelectric perovskite oxides such as (KNbO3)1-x(BaNi0.5Nb0.5O${}_{3-\delta})_{x}$ 55 and (Na0.5Ba0.5TiO3)1-x(BaTi0.5Ni0.5O${}_{3-\delta})_{x}$ 56, TaON and NbON are intrinsic semiconducting photoferroics functioning in the visible range with much simpler compositions. Given that tantalum oxides/nitrides 57, 58 and niobium oxides/nitrides 59, 60 have all demonstrated applications in the production of microelectronics, both oxynitrides are expected to have CMOS compatibility.

Using experimentally synthesized ferroelectric HfO2 as the template, we design a new class of nonperovskite oxynitrides represented by TaON and NbON in the space group of $Pca2_{1}$ that have promising CMOS compatibility. Results from first-principle calculations demonstrate that both oxynitrides are multifunctional ferroelectric semiconductors, with key material parameters such as magnitude of spontaneous polarization, piezoelectric coefficient, carrier mobility, and light absorption being competitive with many state-of-the-art materials such as Si and halide perovskites. Combined with the enhanced electron-hole separation in ferroelectrics 55, 61, 62, the visible-light-responsive oxynitrides with simple chemical composition and nontoxic elements could act as efficient components in photovoltaic and photocatalytic devices. The design strategy in this work also highlights the importance of nonperovskite structure and $4d$/$5d$ elements for the discovery and development of ferroelectric semiconductors for next-generation energy and information technology.

This work is supported by National Key R&D Program of China (2021YFA1202100), National Natural Science Foundation of China (12074319), Natural Science Foundation of Zhejiang Province (2022XHSJJ006), and Westlake Education Foundation. The computational resource is provided by Westlake HPC Center. We acknowledge useful discussions with Dr. Fan Zheng.