betaFeedback

WorksPhysics and Astronomy

Piezoelectric effect and polarization switching in Al1−x_{1-x}Scx_xN

Abstract

Aluminum nitride is piezoelectric and exhibits spontaneous polarization along the cc-axis, but the polarization cannot be switched by applying an electric field. Adding Sc to AlN enhances the piezoelectric properties, and can make the alloy ferroelectric. We perform a detailed first-principles analysis of spontaneous and piezoelectric polarization. Comparisons between explicit supercell calculations show that the virtual crystal approximation produces accurate results for polarization, but falls short in describing the phase stability of the alloy. We relate the behavior of the piezoelectric constant e33e_{33} to the microscopic behavior of the internal displacement parameter uu, finding that the internal strain contribution dominates in the Sc-induced enhancement. The value of uu increases with scandium concentration, bringing the alloy locally closer to a layered hexagonal structure. Our approach allows us to calculate the ferroelectric switching barrier, which we analyze as a function of Sc concentration and temperature based on Ginzburg-Landau theory.

Introduction

The low symmetry of the wurtzite (wz) structure allows AlN to be piezoelectric and also to exhibit spontaneous polarization. The spontaneous polarization of AlN has a specific orientation along the cc axis, and cannot be switched. Switchable polarization would make the material ferroelectric, but in AlN this cannot be accomplished with an applied electric field below its dielectric breakdown limit. Adding Sc to AlN qualitatively changes the properties. It had already been found that ScAlN alloys exhibit enhanced piezoelectricity,[1] with a piezoelectric constant e33e_{33} larger than piezoelectric materials that are presently in use.[2] Excitingly, Fichtner et al.3 demonstrated ferroelectric switching and reported a high paraelectric transition temperature of 600°C for Al0.64Sc0.36N\mathrm{Al}_{0.64}\mathrm{Sc}_{0.36}\mathrm{N}. Very recently, the electric-field-induced formation of Al-polar domains in originally N-polar ScAlN thin films has been directly observed using scanning transmission electron microscopy,[4] providing direct evidence for ferroelectric switching in ScAlN alloys at the atomic scale.

Unlike AlN, ScN stabilizes in the rocksalt structure, however, experimentally,[1, 3] it has been found that Al1−xScxN\mathrm{Al}_{1-x}\mathrm{Sc}_{x}\mathrm{N} alloys can maintain the wz structure up to about x=0.43x=0.43. To analyze the behavior of an alloy, the zeroth order approximation is to interpolate properties between the end points; for Al1−xScxN\mathrm{Al}_{1-x}\mathrm{Sc}_{x}\mathrm{N}, this would mean interpolating between AlN and ScN, both in the wz structure. Interestingly, ScN cannot be stabilized in the wz structure (space group P63mcP6_{3}mc); calculations[5] show that, when initialized in the wz structure, ScN spontaneously relaxes to a layered hexagonal (h) structure (P63/mmcP6_{3}/mmc) in which the anions and cations lie in the same plane. This can be characterized by the internal displacement parameter uu illustrated in the insets of Fig. 1(a): uu is the ratio of the bond length along the cc axis to the cc lattice parameter; uu is close to 0.375 in the wz structure and equal to 0.5 in the layered hexagonal structure.

We note that the layered hexagonal structure has inversion symmetry and hence cannot exhibit polarization;[6] this renders the enhancement in piezoelectricity upon addition of Sc all the more intriguing. However, the expected increase in uu, based on interpolation between wz-AlN and h-ScN, renders ferroelectric switching more feasible. Switching indeed involves moving anions and cations past each other (e.g., from u=0.375u=0.375 to u=0.625u=0.625), with the midpoint being the u=0.5u=0.5 structure. The closer the uu value of the alloy is to 0.5, the lower the switching barrier will be.

In this paper we report first-principles calculations aimed at elucidating the microscopic mechanisms underlying the enhancement of the piezoelectric response and the potential for ferroelectric switching in Al1−xScxN\mathrm{Al}_{1-x}\mathrm{Sc}_{x}\mathrm{N} alloys. The calculations are based on density functional theory (DFT), and alloys are treated in two distinct ways: by performing explicit supercell calculations, and by using the virtual crystal approximation (VCA). We will demonstrate that the combination and comparison of these approaches provides valuable insight into the underlying physics. We have investigated the phase stability, and the variation of the uu parameter as a function of Sc concentration xx. We find that, as the Sc concentration increases, the spontaneous polarization of the alloy decreases, and so does the energy barrier for switching the polarization direction. Our results show that the enhancement of the piezoelectric constant e33e_{33} of the alloy, which can be attributed to the large Born effective charge Z33∗Z^{*}_{33} and the sensitivity of the internal parameter uu to external strain,[2] is reliably described using VCA. The ability to use VCA to describe polarization in alloys offers the prospect of performing large-scale simulations that would otherwise be prohibitively expensive, for instance to model domain-wall motion.

Computational Details

The DFT calculations were performed using projector augmented wave (PAW) potentials,[7, 8] as implemented in the Vienna Ab-initio Simulation Package (VASP),[9] with a plane-wave energy cutoff of 500 eV. The generalized gradient approximation (GGA) by Perdew, Burke, and Ernzerhof (PBE)[10] was employed to account for exchange and correlation. Berry phase calculations[11, 12] and density functional perturbation theory were used to compute polarization, Born effective charges and piezoelectric tensors.

To investigate alloying effects we employ both the supercell method and the VCA.[13] For the supercell approach, we used a 108-atom cell, which is a 3×3×33 \times3 \times3 multiple of the four-atom wz unit cell. We consider both supercells in which Sc and Al atoms are randomly placed on cation sites, as well as supercells based on special quasi-random structures (SQS), which minimize pairwise short-range order up to fourth-nearest cation-cation pairs.[14, 15] The calculated physical observables were configurationally averaged over different supercells.

In the VCA method, virtual cations interpolate between Al and Sc elements, thus mimicking arbitrary Sc concentrations in Al1−x_{1-x}Scx_xN solid solutions. The benefit of VCA lies in its computational efficiency, since calculations can be performed in a 4-atom primitive cell without explicit configurational averaging. It has been demonstrated that the VCA approach can be integrated with the Berry phase calculations, reproducing reliable Born effective charges and piezoelectric coefficients.[13]

Hellmann-Feynman forces on each atom were converged to 10 meV/Å. Brillouin-zone integration for the primitive cell were performed using a 8×8×58 \times8 \times5 kk-point mesh; for the supercells the Γ point was used as a single kk point.

Results and Discussions

Structural parameters

In Figure 1 we plot the internal displacement parameter uu, lattice parameter aa and the c/ac/a ratio as a function of Sc concentration. The value of the internal displacement parameter uu is 0.381 in wz-AlN and 0.500 in layered hexagonal ScN. When Sc is added to wz-AlN, uu will increase. We studied the evolution of uu by gradually increasing the Sc concentration and determining the most stable structure for each concentration. As we will discuss in Sec. III B, in the VCA calculations the wz structure can be stabilized up to a point between 25% and 26% Sc. We note that our calculations do not address a possible phase transition to the rocksalt structure. At higher Sc concentrations, rocksalt should be most stable. However, since we start with the wz structure, there are not enough degrees of freedom during the structural optimization to transition to a rocksalt structure. Still, our calculations address the regime that is most relevant for exploring the polarization properties.

Internal displacement parameter and lattice parameters as a function of Sc concentration

Figure 1. (a) Internal displacement parameter uu and (b) lattice parameter aa and c/ac/a ratio as a function of Sc concentration xx, as obtained using VCA and from supercell calculations. In (a), uu values from supercell calculations are shown for individual cations, along with the weighted average. The inserts define uu and depict the structures of Al1−x_{1-x}Scx_xN near the phase transition point: x=0.25x=0.25 corresponds to the wurtzite phase, x=0.26x=0.26 to the layered hexagonal phase.

We first analyze the results obtained using the VCA. Twenty-one different Sc concentrations were calculated. uu and aa increases monotonically with Sc concentration xx, up to a point between 25% and 26% Sc where a discontinuity occurs and uu jumps to 0.5. At 25% the crystal structure can still be stabilized in a wz phase. At 26% the system can no longer be stabilized in a wz phase, and the structure spontaneously relaxes to a hexagonal phase with u=0.5u=0.5. This layered hexagonal phase (P63/mmcP6_{3}/mmc) has a larger aa and a smaller c/ac/a value than wz. Our VCA result for the concentration at which this transition occurs is lower than the reported experimental value of ∼\sim43%,[1, 3] and from our calculations using supercells, which indicate the transition occurs near 50%∼\sim60%; the latter result is consistent with the calculations of Zhang et al.[16] who identified the transition at 56%. A similar shortcoming of VCA in predicting relative stability of different phases was observed in calculations of pairs.[14, 15] The calculated physical observables were configurationally averaged over different supercells.

In the VCA method, virtual cations interpolate between Al and Sc elements, thus mimicking arbitrary Sc concentrations in Al1−x_{1-x}Scx_xN solid solutions. The benefit of VCA lies in its computational efficiency, since calculations can be performed in a 4-atom primitive cell without explicit configurational averaging. It has been demonstrated that the VCA approach can be integrated with the Berry phase calculations, reproducing reliable Born effective charges and piezoelectric coefficients.[13]

Hellmann-Feynman forces on each atom were converged to 10 meV/Å. Brillouin-zone integration for the primitive cell were performed using a 8×8×58 \times8 \times5 kk-point mesh; for the supercells the Γ point was used as a single kk point.

Results and Discussions

Structural parameters

In Figure 1 we plot the internal displacement parameter uu, lattice parameter aa and the c/ac/a ratio as a function of Sc concentration. The value of the internal displacement parameter uu is 0.381 in wz-AlN and 0.500 in layered hexagonal ScN. When Sc is added to wz-AlN, uu will increase. We studied the evolution of uu by gradually increasing the Sc concentration and determining the most stable structure for each concentration. As we will discuss in Sec. III B, in the VCA calculations the wz structure can be stabilized up to a point between 25% and 26% Sc. We note that our calculations do not address a possible phase transition to the rocksalt structure. At higher Sc concentrations, rocksalt should be most stable. However, since we start with the wz structure, there are not enough degrees of freedom during the structural optimization to transition to a rocksalt structure. Still, our calculations address the regime that is most relevant for exploring the polarization properties.

We first analyze the results obtained using the VCA. Twenty-one different Sc concentrations were calculated. uu and aa increases monotonically with Sc concentration xx, up to a point between 25% and 26% Sc where a discontinuity occurs and uu jumps to 0.5. At 25% the crystal structure can still be stabilized in a wz phase. At 26% the system can no longer be stabilized in a wz phase, and the structure spontaneously relaxes to a hexagonal phase with u=0.5u=0.5. This layered hexagonal phase (P63/mmcP6_{3}/mmc) has a larger aa and a smaller c/ac/a value than wz. Our VCA result for the concentration at which this transition occurs is lower than the reported experimental value of ∼\sim43%,[1, 3] and from our calculations using supercells, which indicate the transition occurs near 50%∼\sim60%; the latter result is consistent with the calculations of Zhang et al.[16] who identified the transition at 56%. A similar shortcoming of VCA in predicting relative stability of different phases was observed in calculations of Pb(Zr0.5_{0.5}Ti0.5_{0.5})O3_3 alloys. [17, 18] Figure 1 also shows results obtained from supercell calculations. At low Sc concentrations, the supercell values agree reasonably well with the VCA results. However, deviations occur when xx approaches 25%, related to the lack of stability of the wurtzite phase in VCA calculations when x>25%x>25\%.

Energy profiles

Using VCA, we calculated the energy profiles of Al1−x_{1-x}Scx_xN alloys for Sc concentrations ranging from 5% to 25%. Figure 2 illustrates the energy profile as a function of the lattice parameters aa and c/ac/a for x=10%x=10\% [Figure 2(a)] and x=25%x=25\% [Figure 2(b)]. We note that the calculated energy profiles in both cases are flat over a wide range of lattice parameters. The flattening of the energy landscape due to a competition between the wz and the hexagonal phase contributes to the enhancement of the piezoelectric coefficient in Al1−x_{1-x}Scx_xN alloys. [2]

Figure 2. Energy profiles [in eV per formula unit (f.u.)] from VCA calculations for Al1−x_{1-x}Scx_xN with (a) x=10%x=10\% and (b) x=25%x=25\%. Green circles denote the global minimum while the blue diamonds denote metastable states.

For 10% Sc, the global minimum is found to be a wz structure with internal displacement parameter u=0.385u=0.385 [see Figure 2(a)]. For x≥10%x\ge10\%, the hexagonal structure becomes the ground state, but a wz structure can be stabilized as long as x≤25%x\le25\% [see Figure 2(b)]. For x=25%x=25\%, the global minimum corresponds to a hexagonal phase with u=0.500u=0.500, but we can also stabilize a metastable state, which has the wz phase with u=0.408u=0.408. When x>25%x>25\%, the wz structure can no longer be stabilized. Figure 1 and Figure 2 show that the hexagonal structure has a larger aa lattice parameter and a smaller c/ac/a ratio than the wz structure.

Spontaneous polarization

Figure 3 shows the results of our calculations of spontaneous polarization PspP_{\mathrm{sp}} in Al1−x_{1-x}Scx_xN alloys with concentrations of Sc up to 25%, i.e., the range over which the wz phase is stable (or metastable). We first examined the polarization of pure AlN and found the spontaneous polarization PspP_{\mathrm{sp}} (expressed per unit cell area) to be extremely close to the result of Ref. 19, which was calculated using the hybrid functional of Heyd, Scuseria and Ernzerhof. [20] At each concentration xx, we followed the approach of Ref. 19: the structure was optimized; the reference structure for calculating PspP_{\mathrm{sp}} was chosen to correspond to the layered hexagonal structure (which has zero polarization) at the same lattice parameters; and to ensure we stay on a specific branch of the multivalued polarization, we constructed an array of interpolated structures with fixed lattice parameters but different internal displacement parameters uu. The resulting values of energy and polarization as a function of uu will also serve to map out the switching barrier, as described in Sec. III E.

Figure 3. Spontaneous polarization PspP_{\mathrm{sp}} of Al1−x_{1-x}Scx_xN as a function of Sc concentration xx. Purple circles represent VCA results while triangles are supercell results. The data point at x=0x=0 was calculated for pure AlN. Experimental data points are extracted from the work of Fichtner et al. [3] and represented as brown diamonds.

Over the range from 0 to 25% Sc, the polarization decreases monotonically with increasing scandium concentration. Figure 3 displays results obtained with VCA and with supercells. Up to x=0.2x=0.2, the VCA results are approximately linear as a function of xx. For x=0.125x=0.125, the agreement between VCA and supercell results is very good, but a deviation is evident at 25%. The supercell result seems to follow the roughly linear decrease with xx, while the VCA result is much lower. The deviation in the VCA result is related to the lack of stability of the wz structure in VCA, as discussed in Sec. III A; right above x=0.25x=0.25, the layered hexagonal structure, which has zero polarization, becomes more favorable.

Piezoelectric polarization

The piezoelectric constants eije_{ij} can be decomposed into a clamped-ion term and internal-strain contributions:[21, 22]

eij=eijclamped+eijint=∂Pi∂εj∣u+∑α,k∂Pi∂uα,k∣ε∂uα,k∂εj.(1)\begin{aligned} e_{ij}=e_{ij}^{\mathrm{clamped}}+e_{ij}^{\mathrm{int}} \\ &=\left.\frac{\partial P_i}{\partial\varepsilon_j}\right|_u+\left.\sum_{\alpha,k}\frac{\partial P_i}{\partial u_{\alpha,k}}\right|_{\varepsilon}\frac{\partial u_{\alpha,k}}{\partial\varepsilon_j}. \tag*{(1)} \end{aligned}

PiP_i is the polarization along the ii direction and εj\varepsilon_j is the strain along the jj direction. uα,ku_{\alpha,k} is the internal atomic coordinate along the kk direction for atom α\alpha in the cell. Note that uα,ku_{\alpha,k} is distinct from the wurtzite internal displacement parameter uu as defined in Figure 1, but they are related by, e.g.,

uSc=uSc,3−uN,3.(2)u_{\mathrm{Sc}}=u_{\mathrm{Sc},3}-u_{\mathrm{N},3}. \tag*{(2)}

The clamped-ion term arises from the contributions of electrons when the ions are frozen at their zero-strain equilibrium internal atomic coordinates uα,ku_{\alpha,k}, i.e., the ionic coordinates follow the strain-induced deformation of the lattice vectors. The internal-strain contributions term reflects the distortion of the ionic coordinates at fixed strain. We note that the first factor in the internal-strain term can be rewritten in terms of the Born effective charge Z∗Z^*, which is defined as

Zα,ij∗=Vqe∂Pi∂rα,j.(3)Z_{\alpha,ij}^{*}=\frac{V}{q_e}\frac{\partial P_i}{\partial r_{\alpha,j}}. \tag*{(3)}

rα,jr_{\alpha,j} is the Cartesian atomic coordinate along the jj direction of atom α\alpha, VV is the volume of the unit cell, and qeq_e is the electron charge.

Here we focus on piezoelectric polarization along the zz axis and hence examine e33e_{33}:

e33=∂P3∂ε3∣u+∑αqecZα,33∗V∂uα,3∂ε3=∂P3∂ε3∣u+4qeZ33∗3a2∂u∂ε3.(4)\begin{aligned} e_{33}=\left.\frac{\partial P_3}{\partial\varepsilon_3}\right|_u+\sum_{\alpha}\frac{q_e c Z_{\alpha,33}^{*}}{V}\frac{\partial u_{\alpha,3}}{\partial\varepsilon_3} \\ &=\left.\frac{\partial P_3}{\partial\varepsilon_3}\right|_u+\frac{4q_e Z_{33}^{*}}{\sqrt{3}a^2}\frac{\partial u}{\partial\varepsilon_3}. \tag*{(4)} \end{aligned}

The transition from internal atomic coordinates uα,3u_{\alpha,3} to Cartesian coordinates rα,3r_{\alpha,3} in the definition of Z∗Z^* has introduced the factor cc (lattice parameter). In the second line of Eq. (4) the internal-strain contribution is rewritten in terms of the wurtzite internal displacement parameter uu and Z33∗Z_{33}^{*}, the Born effective charge for the virtual cation. The factor a2a^2 in the denominator arises from expressing the volume of the wz unit cell VV in terms of the lattice parameters.

Figure 4 displays our results for e33e_{33} and its various contributions calculated using VCA. We compare with results obtained using supercells in Ref. 2. du/dε3du/d\varepsilon_3 is shown in Figure 4(a). The VCA results show a striking dependence on xx: du/dε3du/d\varepsilon_3 increases markedly with xx. After averaging over the Sc and Al values, the supercell results (green triangles) are close to the VCA results in the low-xx limit but again deviate as xx increases towards 0.25.

Figure 4. Contributions to the piezoelectric coefficient e33e_{33} as a function of Sc concentration calculated with VCA and with supercells; supercell results are adapted from Ref. 2. (a) Response of the wurtzite internal displacement parameter uαu_{\alpha} for species α\alpha [as defined in Eq. (2)] to strain ε3\varepsilon_{3}. VCA results are shown in light blue, and results of supercells are distinguished by using yellow triangles for Sc, grey triangles for Al and green triangles for averaged cation. (b) Born effective charge of cations as a function of Sc concentration obtained using VCA (colored circles) and the supercell method (blue triangles). (c) Different contributions to e33e_{33}: clamped-ion (purple), internal strain (orange), and total (brown). The VCA and supercell results are distinguished using solid and dashed lines.

The much larger values of du/dε3du/d\varepsilon_3 for Sc in the supercells can be explained as follows: We know that ScN prefers the (hexagonal) u=0.5u=0.5 structure, in which the Sc and N atoms are relaxed fully coordinated, and in which the aa lattice parameter is significantly larger than in the AlN wz structure. In Al1−x_{1-x}Scx_xN alloys with low Sc concentrations, aa will be close to the value in wz AlN, and hence the position of the Sc atom is constrained; the corresponding uu value will be farther from 0.5 than Sc prefers. When strain is applied, the atoms will eagerly take advantage of the distortion to locally adopt a structure closer to their preferred environment, leading to a large and positive du/dεdu/d\varepsilon value. For Al atoms, uu is closer to its preferred value, and hence Al atoms do not experience such a strong driving force to change their uu parameter, explaining the smaller magnitude of the du/dεdu/d\varepsilon values. In spite of the distinct structural mechanisms experienced by Al and Sc atoms, the VCA does a decent job reproducing the weighted average of du/dε3du/d\varepsilon_3 for the cations in the supercells. The deviation as xx approaches 25% is again due to the exaggerated stability of the hexagonal phase in VCA.

For the Born effective charge Z33∗Z_{33}^{*} [Figure 4(b)], we see that the dependence on Sc concentration is roughly linear. We found the average value of the cation Born effective charges calculated in VCA (green dots) to be in excellent agreement with the supercell results (blue triangles). This consistency highlights the power and efficiency of the VCA in accurately predicting of the Born effective charges.

Z33∗Z_{33}^{*} and du/dεdu/d\varepsilon are combined to calculate e33inte_{33}^{\mathrm{int}}, which is plotted as a function of xx in Figure 4(c), along with e33clampede_{33}^{\mathrm{clamped}} as well as the total e33e_{33}. The clamped-ion term e33clampede_{33}^{\mathrm{clamped}} is negative and significantly smaller in magnitude than e33inte_{33}^{\mathrm{int}}. The supercell results for e33clampede_{33}^{\mathrm{clamped}} show virtually no variation with xx. The VCA results are reasonably close to the supercell results, but show a slight decrease in the magnitude of e33clampede_{33}^{\mathrm{clamped}} as xx increases towards 0.25, the concentration at which (within VCA) the polarization disappears. e33inte_{33}^{\mathrm{int}} is both larger in magnitude and varies more quickly with xx than e33clampede_{33}^{\mathrm{clamped}}. The increase in e33inte_{33}^{\mathrm{int}} is overestimated in VCA, as compared to supercell results. This can be directly attributed to the overestimate of du/dεdu/d\varepsilon in VCA [Figure 4(a)] discussed above.

Ferroelectric switching barrier

For a wz structure, energy minima occur for values of the internal displacement parameters uu and 1−u1-u; these represent degenerate and symmetric states with opposite polarization, corresponding to the minima of a double-well potential. Our calculated double-well potentials as a function of polarization (proportional to uu) for different Sc concentration are shown in Figure 5(a). As each Sc concentration, we freeze the lattice parameters aa and cc and perform VCA calculations for total energy and polarization as a function of uu. The symmetric structure at u=0.5u=0.5, which has zero polarization, constitutes the barrier; the well depth thus corresponds to the energy barrier to switch the polarization.

Figure 5. (a) Energy per formula unit (f.u.) of Al1−xScxN\mathrm{Al}_{1-x}\mathrm{Sc}_{x}\mathrm{N} at T=0T=0 as a function of polarization for various Sc concentrations. The data points (shown only for positive polarization) at each Sc concentration are based on VCA calculations as described in the text. The curves are fits to Eq. (5). (b) Energy per f.u. as a function of polarization predicted by phenomenological modeling according to Eq. (5) at T=300T=300 K, for different Sc concentrations. (c) Free energy per f.u. as a function of polarization according to Eq. (5) for Al0.8Sc0.2N\mathrm{Al}_{0.8}\mathrm{Sc}_{0.2}\mathrm{N}, for different temperatures. In all plots the energy at the point of zero polarization has been set to zero.

The switching path we are considering here is restricted in the sense that all the dipoles in the material are switched simultaneously and uniformly. In real systems, ferroelectric switching usually involves initial nucleation of a reversed ferroelectric domain followed by gradual switching of the global polarization via domain-wall motion. The switching barrier for this non-uniform process is usually much lower than the barrier calculated here. In addition, our calculations are performed assuming the lattice parameters aa and cc are kept fixed during the switching. This lack of structural flexibility will yield a higher switching barrier. Therefore, our calculated barrier represents an upper bound on the actual barrier. However, our calculations provide insight into how the Sc concentration affects the ability to switch polarization.

Further insight into the switching process at finite temperatures can be obtained by employing the Ginzburg-Landau theory for ferroelectric-paraelectric phase transitions.[23] The free energy of the system E(T,P)E(T,P) as a function of polarization is expressed as:

E(T,P)=E0(T)+12a(T)P2+14b(T)P4+⋯ .(5)E(T,P)=E_{0}(T)+\frac{1}{2}a(T)P^{2}+\frac{1}{4}b(T)P^{4}+\cdots. \tag*{(5)}

PP is the polarization along cc direction, and E0(T)E_{0}(T) is the free energy of system with zero polarization. a(T)a(T) and b(T)b(T) are the coefficients of the second and fourth order terms.

We use Eq. (5) to fit our calculated first-principles energies at T=0T=0 as a function of polarization, allowing us to extract values for a(0)a(0) and b(0)b(0) at each Sc concentration xx. The curves shown in Figure 5(a) are a result of these fits. The energy barrier between the polarized state and the zero-polarization reference state decreases with increasing scandium concentration; for pure AlN the calculated energy barrier is 0.51 eV/f.u., while at 20% Sc the barrier is decreased to 0.35 eV/f.u. (see Figure 6). As noted above, these values constitute an upper bound because we keep the lattice parameters fixed. Some insight into the effect of allowing lattice relaxation can be obtained by comparing with the results of Ye et al., [24] who employed the solid-state nudge elastic band method (ss-NEB)[25] that allows for cell relaxation. For pure AlN the ss-NEB result for the switching barrier is 0.23 eV/f.u., about half of our value. For Al0.8Sc0.2N\mathrm{Al}_{0.8}\mathrm{Sc}_{0.2}\mathrm{N}, the ss-NEB result for the barrier is 0.12 eV/f.u. Still, this barrier is much too high to allow switching.

Figure 6. Ferroelectric switching barrier (eV/f.u.) for Al1−xScxN\mathrm{Al}_{1-x}\mathrm{Sc}_{x}\mathrm{N} at T=0T=0 K and at 300 K as a function of Sc concentration.

To assess the temperature effect on the switching barrier, we follow the Landau theory and assume the parameter a(T)a(T) can be written as

a(T)=a0T−TcTc(6)a(T) = a_0 \frac{T-T_c}{T_c} \tag*{(6)}

where a0a_0 is a positive constant and TcT_c is the paraelectric transition temperature. In the absence of specific information, we adopt the lower-limit TcT_c value for Al0.64Sc0.36N\mathrm{Al}_{0.64}\mathrm{Sc}_{0.36}\mathrm{N}, which is 600°C (Ref. 3), for all considered Al1−xScxN\mathrm{Al}_{1-x}\mathrm{Sc}_{x}\mathrm{N} alloys.

Figure 5(b) shows the energy barriers at 300 K as a function of Sc concentration. Again, we see that the ferroelectric switching barrier of Al1−xScxN\mathrm{Al}_{1-x}\mathrm{Sc}_{x}\mathrm{N} decreases with increasing scandium concentration. Figure 5(c) shows the energy barriers for a 20% scandium concentration at different temperatures. The energy barrier is reduced from 0.35 eV/f.u. at 0 K to 0.15 eV/f.u. at room temperature. The switching barriers at T=300T=300 K are included in Figure 6 and compared with the T=0T=0 barriers.

Based on our calculations, we can estimate the switching electric field as follows. If E⃗\vec{E} is the applied electric field necessary for switching, then ω=12E⃗⋅P⃗\omega= \frac{1}{2}\vec{E}\cdot\vec{P} is the electric field energy density,[26] which we can obtain from the calculated energy EbE_b at the switching barrier: ω=EbVf.u.\omega= \frac{E_b}{V_{\mathrm{f.u.}}}. The static field part 12ϵ0E2\frac{1}{2}\epsilon_0 E^2 is ignored. At room temperature, the polarization PP of Al0.8Sc0.2N\mathrm{Al}_{0.8}\mathrm{Sc}_{0.2}\mathrm{N} is 0.85 C/m2^2 [Figure 5(b)], and the corresponding energy barrier EbE_b is 0.15 eV/f.u. (Figure 6). We thus obtain a switching electric field of 26.5 MV/cm.

Experiments have demonstrated ferroelectric switching in Al1−xScxN\mathrm{Al}_{1-x}\mathrm{Sc}_{x}\mathrm{N} starting at Sc concentrations of x=27%x=27\%.[3] For x=27%x=27\%, the polarization was estimated to be 100 μ\muC/cm2=1^2=1 C/m2^2, and the switching electric field was estimated to be close to 4.5 MV/cm. The estimate for the switching electric field based on our calculations is thus almost 6 times larger than the experimental one. The overestimate is expected, based on our assumption of uniform coherent switching and the lack of cell relaxation when computing the double-well potential in Figure 5.

Conclusions

We studied the properties of Al1−xScxN\mathrm{Al}_{1-x}\mathrm{Sc}_{x}\mathrm{N} alloys with the aim of elucidating the underlying mechanisms of the enhancement in piezoelectricity and the observation of ferroelectricity. We investigated the internal parameter uu and mapped energy profiles as a function of scandium concentration. By decomposing the piezoelectric constant e33e_{33} into its contributions, we found that the internal strain contribution dominates, which can be attributed to larger displacements associated with Sc ions. Our comparison between VCA and supercell results indicated that VCA can produce reliable results for polarization in Al1−x_{1-x}Scx_xN alloys, offering the prospect of performing large-scale simulations that would otherwise be prohibitively expensive. Our calculations of polarization as a function of internal displacement allowed us to calculate values for the ferroelectric switching barrier at T=0T=0. We also employed Ginzburg-Landau theory to estimate barrier heights at finite temperatures. While the barrier heights are overestimated (due to the assumption of uniform coherent switching and the absence of cell relaxation), our results offer insight into the variation of barrier height with Sc concentration.

Acknowledgments

The authors appreciate fruitful discussions with Xie Zhang, Haidong Lu and Simon Fichtner. This work was supported by the Air Force Office of Scientific Research under Award No. FA9550-18-1-0237. Use was made of computational facilities purchased with funds from the National Science Foundation (NSF) (No. CNS-1725797) and administered by the Center for Scientific Computing (CSC). The CSC is supported by the California NanoSystems Institute and the Materials Research Science and Engineering Center (NSF DMR 1720256) at UC Santa Barbara. This work also used the Extreme Science and Engineering Discovery Environment (XSEDE), which is supported by NSF under Grant No. ACI-1548562.

Data Availability

The data that support the findings of this study are available from the corresponding author upon reasonable request.

References

  1. [1]M. Akiyama, T. Kamohara, K. Kano, A. Teshigahara, Y. Takeuchi, and N. Kawahara, Adv. Mater. 21, 593 (2009).
  2. [2]F. Tasnádi, B. Alling, C. Höglund, G. Wingqvist, J. Birch, L. Hultman, and I. A. Abrikosov, Phys. Rev. Lett. 104, 137601 (2010).
  3. [3]S. Fichtner, N. Wolff, F. Lofink, L. Kienle, and B. Wagner, J. Appl. Phys. 125, 114103 (2019).
  4. [4]N. Wolff, S. Fichtner, B. Haas, M. R. Islam, F. Nickiel, M. Kessel, O. Ambacher, C. Koch, B. Wagner, F. Lofink, and L. Kienle, J. Appl. Phys. 129, 034103 (2021).
  5. [5]N. Farrer and L. Bellaiche, Phys. Rev. B 66, 201203 (2002).
  6. [6]S. Biswas and B. Saha, Phys. Rev. Mater. 3, 020301 (2019).
  7. [7]P. E. Blöchl, Phys. Rev. B 50, 17953 (1994).
  8. [8]G. Kresse and D. Joubert, Phys. Rev. B 59, 1758 (1999).
  9. [9]G. Kresse and J. Furthmüller, Phys. Rev. B 54, 11169 (1996).
  10. [10]J. P. Perdew, K. Burke, and M. Ernzerhof, Phys. Rev. Lett. 77, 3865 (1996).
  11. [11]R. Resta, Rev. Mod. Phys. 66, 899 (1994).
  12. [12]D. Vanderbilt, J. Phys. Chem. Solids 61, 147 (2000).
  13. [13]L. Bellaiche and D. Vanderbilt, Phys. Rev. B 61, 7877 (2000).
  14. [14]A. Zunger, S.-H. Wei, L. G. Ferreira, and J. E. Bernard, Phys. Rev. Lett. 65, 353 (1990).
  15. [15]L. Chen, D. Holec, Y. Du, and P. H. Mayrhofer, Thin Solid Films 519, 5503 (2011).
  16. [16]S. Zhang, D. Holec, W. Y. Fu, C. J. Humphreys, and M. A. Moram, J. Appl. Phys. 114, 133510 (2013).
  17. [17]N. J. Ramer and A. M. Rappe, J. Phys. Chem. Solids 61, 315 (2000).
  18. [18]N. J. Ramer and A. M. Rappe, Phys. Rev. B 62, R743 (2000).
  19. [20]J. Heyd, G. E. Scuseria, and M. Ernzerhof, J. Chem. Phys. 118, 8207 (2003).
  20. [21]A. Dal Corso, M. Posternak, R. Resta, and A. Baldereschi, Phys. Rev. B 50, 10715 (1994).
  21. [22]H. Momida, A. Teshigahara, and T. Oguchi, AIP Adv. 6, 065006 (2016).
  22. [23]L. Landau, Nature 138, 840 (1936).
  23. [24]K. Han, Y. Ge, G. Han, I. W. Yeu, C. S. Hwang, and J.-H. Choi, Phys. Status Solidi RRL , 2100009 (2021).
  24. [25]D. Sheppard, P. Xiao, W. Chemelewski, D. D. Johnson, and G. Henkelman, J. Chem. Phys 136, 074103 (2012).
  25. [26]D. J. Griffiths, Introduction to electrodynamics (Pearson, 2013).

Paper details

vlnc:W_v4n3jnwu
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 2

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