HfO₂/ZrO₂ Superlattices
- HfO₂/ZrO₂ superlattices are ordered multilayer heterostructures that exploit fluorite-derived polymorphism to tailor ferroelectric and dielectric behaviors.
- They use epitaxial constraints and periodic interface engineering to control phase transitions, enhancing polarization and cycling endurance.
- Fabrication via methods like pulsed laser deposition, ALD, and solution processing enables CMOS compatibility and nanoscale domain control.
HfO/ZrO superlattices are ordered multilayer heterostructures composed of alternating hafnium-oxide and zirconium-oxide building blocks, or closely related stacks in which HfZrO sublayers alternate with pure ZrO. Within fluorite-derived oxides, these architectures are used to manipulate the competition among monoclinic, tetragonal, orthorhombic, and rhombohedral polymorphs; to impose epitaxial and interfacial constraints unavailable in homogeneous solid solutions; and to tune ferroelectric, antiferroelectric, dielectric, endurance, and gate-stack characteristics through periodicity, layer sequence, and interface density. Recent work places these systems at the intersection of CMOS-compatible ferroelectrics, phase-boundary engineering, and atomically confined polar-texture physics (Li et al., 1 Jul 2025, Gent et al., 7 Jul 2025, Mukherjee et al., 2024).
1. Structural definition and phase landscape
The structural basis of HfO/ZrO superlattices is the fluorite-derived polymorphism shared by HfO and ZrO. In both oxides, the monoclinic phase is the ground state, while technologically relevant polar behavior emerges from metastable fluorite-derived phases. Several distinct polar and nonpolar reference structures recur across the literature. The ferroelectric orthorhombic phase is 0, the nonpolar tetragonal reference is 1, and a nonpolar orthorhombic reference also discussed in this context is 2. In the 3 structure, the unit cell along the stacking direction alternates between nonpolar and polar half-cells, and the polarization originates from off-centering of the O4 sublattice relative to the surrounding cation cages. In another experimentally important regime, epitaxial Hf5Zr6O7/ZrO8 superlattices are dominated by the rhombohedral polar phase, described as 9 (Afroze et al., 25 Jul 2025, Gent et al., 7 Jul 2025).
This phase landscape is highly sensitive to superlattice periodicity. In epitaxial 0 superlattices, smaller periodicity produces a higher fraction of the orthorhombic polar phase, whereas larger periodicity increases the monoclinic nonpolar phase. For the 1 nm series with 2, 3 gives the strongest ferroelectric response and the highest proportion of the orthorhombic polar phase; when 4, monoclinic reflections near 5 and 6 increase. Pole-figure XRD for 7 shows 12 distinct diffraction spots consistent with four crystallographic domains, and atomic-resolution STEM/EDXS resolves distinct HfO8/ZrO9 interfaces with only single-atomic-layer-scale interdiffusion and zig-zag cation arrangements characteristic of orthorhombic 0 (Li et al., 1 Jul 2025).
First-principles work further shows that fully polar HfO1/ZrO2 superlattices are not exceptional outliers but part of a broader HfO3-based superlattice design space. For equal-thickness 2/2 and 4/4 cation-sublayer stacks, layered along pseudo-cubic [100], [010], or [001], the Zr/Hf system remains fully polar in both layers. In 4/4 Zr/Hf superlattices, the reported energy penalties are 56 and 58 meV/cation and the polarization is about 4–5, indicating a competitive polar state but not the strongest thermodynamic stabilization among all HfO6-based partner oxides (Mukherjee et al., 2024).
2. Architectures and fabrication routes
Experimentally, the term “HfO7/ZrO8 superlattice” spans several architectures rather than a single canonical stack. The most studied families differ in layer thickness, chemical ordering, substrate, and intended operating mode.
| Architecture | Representative stack | Reported context |
|---|---|---|
| Epitaxial periodic superlattice | 9 | Ferroelectric stabilization on LSMO-buffered STO(001) |
| Rhombohedral HZ–Z superlattice | 0-T | Hf1Zr2O3/ZrO4 with 1:1 thickness ratio |
| FE/AFE nanolaminate | W / [Hf5Zr6O7 (1 nm) / ZrO8 (1 nm)] 9 / W | MPB-like dielectric enhancement |
| Thick multilayer film | Ten alternating 5 nm La:HfO0 and ZrO1 layers | Solution-processed 50 nm ferroelectric multilayer |
| BEOL MFIM superlattice | SL5 and SL3 HfO2:ZrO3 sequences | Low-voltage ferroelectric switching |
| Gate-stack laminate | HZH, ZHZ, HZHA | High-4, low-EOT CMOS gate dielectric |
Epitaxial 5 superlattices are grown on LSMO-buffered STO(001) by pulsed laser deposition, with periodicity 6 given in unit cells and total thicknesses spanning 7 nm series and 8 films from 4 to 100 nm. A related epitaxial family, 9-T, alternates Hf0Zr1O2 and pure ZrO3 sublayers, usually at a 1:1 thickness ratio, with 4, 5, or 6. In this family, sequence matters: HZ–Z stacks outperform inverse Z–HZ stacks because the bottom LSMO/Hf-containing interface appears important for stabilizing the ferroelectric phase, and the ZrO7 sublayers are kept below 5 nm to avoid relaxation into the tetragonal phase (Li et al., 1 Jul 2025, Gent et al., 7 Jul 2025).
Silicon-compatible nanolaminates are realized by ALD and by solution processing. One ALD implementation uses a 10 nm W/[Hf8Zr9O0 (1 nm)/ZrO1 (1 nm)]2/W metal-insulator-metal capacitor fabricated on SiO3/Si, with deposition at 4 and anneals between 5 and 6. A solution-processed multilayer uses ten 5 nm-thick layers of La:HfO7 and ZrO8, alternately deposited to form a 50 nm stack on platinized Si; STEM-EDS confirms the alternating chemistry, and HRTEM shows grain continuation across interfaces rather than complete breakup at every layer (Kashir et al., 2021, Mandal et al., 2024).
Two additional variants extend the architecture beyond conventional capacitor ferroelectrics. BEOL-compatible MFIM capacitors use about 10 nm total ferroelectric thickness, with SL5 defined as a 5:5 HfO9:ZrO0 sublayer sequence repeated over 11 supercycles and SL3 as a 3:3 sequence repeated over 19 supercycles. In logic-oriented high-1 stacks, the same layering principle appears in HZH = HfO2/ZrO3/HfO4, ZHZ = ZrO5/HfO6/ZrO7, and HZHA, where a 3 Å Al8O9 dipole layer is embedded within the Hf/Zr laminate (Li et al., 8 Sep 2025, Song et al., 27 Dec 2025).
3. Stabilization mechanisms and phase engineering
The central mechanism of HfO0/ZrO1 superlattice engineering is the replacement of bulk-like relaxation by interface-dominated phase selection. In the epitaxial 2 system, smaller 3 means more interfaces, stronger structural constraint, and a higher fraction of the polar orthorhombic phase. Leakage decreases as the number of interfaces increases, and the authors interpret this as suppression of charged-defect migration. The theoretical analysis emphasizes kinetics and interfacial formation energy rather than simple bulk thermodynamics: for ZrO4 adjacent to the HfO5 orthorhombic lattice, the transition barrier from tetragonal to monoclinic increases by about 6, while the barrier to the orthorhombic phase remains nearly unchanged. The reported 7-8 interfacial formation energies are 9 in a comparative HZO superlattice and 00 in the HfO01/ZrO02 superlattice, yielding derived critical thicknesses of 03 and 04 nm, respectively (Li et al., 1 Jul 2025).
In the Hf05Zr06O07/ZrO08 superlattices, the reported design rules are explicit. The ZrO09 layers act as a booster for the total remnant polarization 10, but the mechanism is not a trivial linear addition of ZrO11 polarization. Instead, the architecture preserves strain and polarization within each thin sublayer, and the enhancement is attributed to two coupled effects: the intrinsic increase in polarization with higher Zr content, since rhombohedral ZrO12 is more polar than rhombohedral HfO13; and the suppression of thickness relaxation, which maintains an elongated out-of-plane 14. The formal superlattice formation energy is written as
15
and the calculations indicate that the rhombohedral phase can be stabilized by epitaxial strain alone. The resulting optimization rules are to maximize Zr content within the Hf16Zr17O18 sublayers, keep both sublayers thin, use more repetitions and therefore more interfaces, prefer HZ–Z ordering, and avoid excessively thick ZrO19 sublayers (Gent et al., 7 Jul 2025).
A broader thermodynamic perspective comes from HfO20-based superlattice theory. For Zr/Hf, the elastic mismatch is small, so the formation energies are described as almost entirely bulk-driven; this explains why the system is robustly polar but not dramatically stabilized relative to bulk ferroelectric HfO21. By contrast, partner oxides lacking a monoclinic ground state can drive stronger thermodynamic stabilization. This suggests that ZrO22 is a natural and experimentally mature partner for HfO23, but not the only route to polar phase stabilization in fluorite-derived oxide superlattices (Mukherjee et al., 2024).
4. Electrical, dielectric, and endurance characteristics
The ferroelectric performance of HfO24/ZrO25 superlattices spans several distinct regimes. In epitaxial 26, the optimized 27 periodicity gives a maximum polarization of 28, remanent polarization of 29, and coercive field of about 30–31 for the 32 nm periodicity series. Ferroelectricity is retained from 4 to 100 nm total thickness, and at 100 nm the coercive field decreases to 33 while the film maintains a polar phase ratio exceeding 80%. In the 6 nm 34 superlattice, the remanent polarization remains near 35 with less than 15% variation over 36 switching cycles, whereas the comparison HZO film breaks down after 37 cycles. The response is stable from 500 Hz to 50 kHz and from 77 K to 473 K, and the near frequency independence is interpreted as evidence of nucleation-limited switching (Li et al., 1 Jul 2025).
The highest reported ferroelectric polarization in the surveyed Hf38Zr39O40/ZrO41 epitaxial superlattices occurs at very high Zr fraction. A superlattice with total 87.5% ZrO42 content exhibits record polarization with 43 and can be cycled 44 times while maintaining 45. The best member of the HfO46-based family, 47, reaches a remnant polarization of about 48, and the best member of the Hf49Zr50O51-ZrO52 family reaches 53. Cyclability is described as mostly dependent on the number of interfaces, which redistribute oxygen vacancies and slow the formation of conductive filaments (Gent et al., 7 Jul 2025).
Multilayering also changes wake-up kinetics and phase-boundary behavior. In a 50 nm solution-processed La:HfO54/ZrO55 multilayer, ferroelectric switching is observed even though the pure ZrO56 comparison film is paraelectric. After wake-up, both the multilayer and the La:HfO57 reference show a positive remanent polarization of 58, with coercive fields of 59 and 60, respectively. Under direct high-field cycling, the multilayer reaches a maximum 61 of 62 after 1000 cycles, and saturation occurs about ten times faster than in the La:HfO63 film. Separately, the ALD W/[Hf64Zr65O66 (1 nm)/ZrO67 (1 nm)]68/W nanolaminate shows a double hysteresis loop with finite remanence, 69, 70 at 71, and a highest reported dielectric constant of 72 after annealing at 73 for 30 s, with equivalent oxide thickness reduced to about 6 Å. The latter behavior is attributed to FE/AFE phase coexistence and an MPB-like interfacial landscape (Mandal et al., 2024, Kashir et al., 2021).
BEOL-compatible superlattice capacitors extend this property space to low-voltage and analog operation. In the SL5-2 device, proper tuning of the FE/AFE balance yields 74 at only 75, essentially imprint-free operation with 76, a 3 MV/cm potentiation/depression bias window, and an on/off ratio of 20. SL5-2 shows only 5% degradation after 77 cycles, while SL3-2 provides the longest endurance in the reported device set; recoverable fatigue is demonstrated by 78 V recovery pulsing at 1 kHz (Li et al., 8 Sep 2025).
Hf/Zr superlattice concepts also appear in non-ferroelectric gate dielectrics. After a 79, 30 s, N80 anneal, HZH achieves an EOT of 7.3 Å, lower than HfO81-only stacks at 8.5 Å, while HZHA combines an 8.4 Å EOT with a flatband-voltage shift greater than 200 mV. Under 82 V negative-bias temperature stress at 83 for 100 s, HZHA and HA show comparable flatband-voltage drifts of 87 and 97 mV, respectively. This places Hf/Zr layering within both ferroelectric and advanced high-84 logic-device trajectories (Song et al., 27 Dec 2025).
5. Atomic-scale polar textures and domain-wall hierarchy
An important recent development is the recognition that HfO85/ZrO86-type layered ferroelectrics are not adequately described as simple stacks of uniformly polarized planes. In fluorite oxides such as ZrO87 and HfO88, the material already self-organizes into alternating stacks of two-dimensional polar and nonpolar half-unit-cell layers. In plan-view ABF-STEM of a 10 nm ZrO89 film on Si, a 90-oriented grain shows polar columns separated by nonpolar columns, and within each polar column the dipole direction reverses every 2–6 unit cells along the in-plane polar axis. The smallest and largest observed domain lengths are 2 and 4.5 unit cells, with lateral footprint below 91, making these among the smallest stable polar domains reported in any polar material (Afroze et al., 25 Jul 2025).
Every such reversal produces a charged 92 domain wall. The observed walls alternate between head-to-head, 93, and tail-to-tail, 94, configurations, and this alternation recurs across 38 imaged domains. The two wall types are structurally distinct. The head-to-head wall contains a 95-like slice in which O96 atoms shift into the cation plane, whereas the tail-to-tail wall locally resembles a distorted tetragonal 97 half-cell with the O98 sublattice near a centrosymmetric position. Multislice simulations reproduce the ABF contrast for both wall types, supporting these assignments. The work identifies these interfaces as the first experimental observation of antipolar ferroic ordering via strongly charged domain walls nested within the self-organized polar/nonpolar layering (Afroze et al., 25 Jul 2025).
The microscopic explanation is tied to lattice dynamics. For 99 and 00, the low-frequency longitudinal-optical modes 01 and 02 along 03 have very flat dispersions, about 04 wide, and the wall width is summarized by
05
where 06 is the LO-band dispersion. In fully relaxed DFT models, the charged walls shrink to about 07, whereas cubic PbTiO08 has much broader charged walls exceeding 09. The paper argues that the flat transverse-optical dispersion along the stacking axis decouples neighboring dipolar planes, so each sheet can host its own internal in-plane ordering; as a result, walls do not line up from one polar sheet to the next, and domain lengths vary within and between sheets. This suggests that, in HfO10/ZrO11 superlattices, the effective ferroelectric response may depend not only on the fraction of polar material but also on how each atomically thin polar sheet internally self-organizes (Afroze et al., 25 Jul 2025).
6. Electronic-structure constraints and broader technological context
A recurring misconception is that HfO12/ZrO13 superlattices are necessarily superior to homogeneous Hf14Zr15O16 in every device-relevant metric. Electronic-structure calculations show a more complicated picture. In orthorhombic ferroelectric HfO17/ZrO18 superlattices built from alternating HfO19 and ZrO20 monolayers, denoted H21Z22-SL, the band gap decreases as the period 23 increases, and H12Z12-SL and H24Z24-SL have gaps below 24, i.e. below bulk orthorhombic ZrO25. This is qualitatively different from the Hf26Zr27O28 solid-solution trend, whose gap remains bounded by the two end members. The reduction is attributed to asymmetric HfO29/ZrO30 interfaces: oxygen anions closer to Hf are more negatively charged than those closer to Zr, producing an internal electric field, a Hartree-potential slope across each layer, and downward spreading of the conduction manifold. The practical implication is a possible penalty in charge injection, leakage, dielectric breakdown, and cycling reliability for large-period superlattices (Huang et al., 2023).
A second qualification concerns structural softness. First-principles strain studies identify a strain-driven ferroelastic instability of ferroelectric HfO31, with the antipolar 32 oVIII phase as the competing polymorph. In that framework, a 2/2 Hf/Zr superlattice behaves very similarly to pure HfO33, whereas Hf/Pb and Hf/Sn substantially lower the critical tensile strain needed to access enhanced response. This indicates that Hf/Zr layering is compatible with ferroelectric hafnia-like behavior but is not the most effective route for reducing the strain threshold for structural softness (Mukherjee et al., 2024).
The broader design lesson is therefore twofold. First, HfO34/ZrO35 superlattices are a robust and experimentally mature fluorite-oxide platform that can support fully polar states, MPB-like FE/AFE coexistence, high endurance, low-voltage switching, and even high-36 gate-dielectric functionality (Mukherjee et al., 2024, Song et al., 27 Dec 2025). Second, their behavior is governed by a nontrivial combination of periodicity, interface chemistry, internal fields, sublayer thickness, and microscopic polar texture rather than by composition alone. A plausible implication is that future optimization will increasingly treat HfO37/ZrO38 superlattices not merely as compositionally modulated films, but as hierarchical ferroic systems in which phase stability, domain topology, electronic structure, and defect transport are jointly engineered.