materials.yaml

# FeCIM Lattice Tools - Physics Configuration
# =============================================================================
#
# This file contains all physical parameters for ferroelectric CIM simulation.
# Edit this file to experiment with different materials and hardware configurations.
#
# FIELD REFERENCE (all materials):
# ---------------------------------------------------------------------
# Polarization:
#   ec_v_m          Coercive field (V/m). Field required to switch polarization.
#   ps_uC_cm2       Saturation polarization (µC/cm²). Max P at high field.
#   pr_uC_cm2       Remanent polarization (µC/cm²). Residual P at zero field.
#
# Geometry & Dielectrics:
#   thickness_m      Film thickness (meters). Affects field scaling E = V/d.
#   epsilon_hf      High-frequency permittivity (dimensionless).
#   epsilon_lf      Low-frequency permittivity (dimensionless).
#   loss_tangent    Dielectric loss tanδ (dimensionless).
#
# Temperature:
#   curie_temp_k    Curie temperature (K). Phase transition point.
#   temp_coeff_ec   dEc/dT (V/m/K). Coercive field temperature dependence.
#   temp_coeff_pr   dPr/dT (C/m²/K). Remanence temperature dependence.
#
# Switching Dynamics:
#   tau_s           Characteristic switching time (seconds). KAI model.
#   activation_energy_ev  Activation energy (eV). Thermal barrier.
#   kai_exponent    KAI exponent (dimensionless). Domain growth dimensionality.
#
# Reliability:
#   endurance_cycles Max cycling before degradation (cycles).
#   retention_time_s Data retention time at operating temp (seconds).
#   imprint_field_v_m DC-bias-induced asymmetry (V/m).
#   depolarization_factor_vm_c  Depolarization field coefficient (V·m/C).
#
# Landau Coefficients (thermodynamics):
#   alpha_landau    α coefficient (J·m/C²). Curvature at origin.
#   beta_landau     β coefficient (J·m⁵/C⁴). Double-well depth.
#   gamma_landau    γ coefficient (J·m⁹/C⁶). High-P stability.
#   rho_viscosity   Viscosity ρ (Ω·m). Damping coefficient.
#
# Conductance Mapping:
#   gmin_s          Minimum conductance (S). High-resistance state.
#   gmax_s          Maximum conductance (S). Low-resistance state.
#
# Circuit:
#   series_resistance_ohm  Series resistance (Ω). IR drop modeling.
# ---------------------------------------------------------------------
#
# =============================================================================
# REFERENCES (Updated 2026-02-02)
# =============================================================================
#
# Core Material (HfO2-ZrO2 Superlattice):
#   [1] Park et al., Adv. Mater. 27, 1811 (2015) - HZO ferroelectricity discovery
#   [2] Cheema et al., Nature 580, 478 (2020) - Superlattice enhancement
#   [3] Nature Commun. 2025, doi:10.1038/s41467-025-61758-2 - Epitaxial superlattice stability
#   [4] ACS Nano 2024, doi:10.1021/acsnano.4c01992 - 4nm HZO, 10^10 endurance at 1.2V
#   [5] ACS Appl. Nano Mater. 2024, doi:10.1021/acsanm.4c04974 - 50µC/cm² Pr, 10^10 endurance
#   [6] ACS Omega 2025, doi:10.1021/acsomega.4c10603 - Multilayer reduced wake-up
#   [7] APL Materials 2023, doi:10.1063/5.0148068 - HZO roadmap
#
# Switching Dynamics:
#   [8] Purdue Thesis 2024 - Sub-ns switching (925ps), 360ps record in sub-µm arrays
#   [9] Nature Commun. 2025, doi:10.1038/s41467-025-66807-4 - Domain switching holography
#
# Cryogenic Operation:
#   [10] Adv. Electron. Mater. 2024, doi:10.1002/aelm.202300879 - 75µC/cm² Pr at 4K
#   [11] Adv. Electron. Mater. 2025, doi:10.1002/aelm.202400840 - Compact model to 4K
#
# CIM & Energy:
#   [12] Samsung Nature 2025 - 25-100x energy vs NAND
#   [13] Nature Electron. 2025, doi:10.1038/s41928-025-01477-0 - Analog matrix solving
#   [14] DAC 2024, doi:10.1145/3649329.3658472 - 160 GOPS/mm², 25.5 TOPS/W
#
# FeFET Crossbar:
#   [15] Nature Commun. 2023, doi:10.1038/s41467-023-42110-y - MLC FeFET crossbar demo
#   [16] IEEE TED 2022, doi:10.1109/TED.2022.3216973 - 28nm current-limited FeFET crossbar, 97% MNIST with limiter vs 9.8% without
#
# Conference / process reports:
#   [17] COSM 2025 conference presentation - 30 discrete states, 87% MNIST (unverified)
#   [18] Adv. Electron. Mater. 2024, doi:10.1002/aelm.202400603 - In2Se3 FWF synthesis
#
# Endurance & Reliability:
#   [19] IEEE IRPS 2022 - 10^9 cycle endurance verified
#   [20] ResearchGate 2024 - 10^10 endurance with AFE ZrO2 seed layer
#
# Multi-Level Demonstrations:
#   [21] Oh et al., IEEE EDL 38(6), 732 (2017) - 32 analog states HZO FeFET
#   [22] Song et al., Adv. Science 2024, doi:10.1002/advs.202308588 - 2D FeFET >7-bit (140 states)
#   [23] ACS Appl. Mater. Interfaces 2022, doi:10.1021/acsami.2c04441 - High-speed FTJ multilevel
#
# Alternative Ferroelectrics:
#   [24] Nature Commun. 2025, doi:10.1038/s41467-025-62904-6 - AlScN decoupled Pr/Ec, 15V window
#   [25] APL Materials 2023, doi:10.1063/5.0148068 - AlScN 100-170 µC/cm² Pr, 2-6 MV/cm Ec
#
# In₂Se₃ 2D Ferroelectric (flash-within-flash synthesis reports):
#   [26] Nature Commun. 2017, doi:10.1038/ncomms14956 - In2Se3 intrinsic 2D ferroelectricity (0.11 eÅ dipole)
#   [27] InfoMat 2022, doi:10.1002/inf2.12341 - In2Se3 for ferroelectric data storage
#   [28] AIP Appl. Phys. Rev. 2024, doi:10.1063/5.0190692 - In2Se3 phase transitions & devices
#   [39] Nature Commun. 2024, doi:10.1038/s41467-024-54841-7 - Stacking-selected polarization switching
#   [40] Nano Research 2020, doi:10.1007/s12274-020-2640-0 - Thickness-dependent Ec, CVD α-In2Se3
#   [41] Science Advances 2022, doi:10.1126/sciadv.abo0773 - Phase/polarization modulation TEM
#   [42] JAP 137, 124102 (2025), doi:10.1063/5.0236192 - AI-assisted phase field, PSO-fitted Landau (PAYWALLED)
#   [43] arXiv:2007.02752 - FeSMJ 8th-order Landau potential for In2Se3
#   [44] UT Tyler thesis - ρ < 0.1 Ω·m requirement for GHz 2D ferroelectrics
#   [45] Nano Letters 2023, arXiv:2308.11221 - In2Se3 Arrhenius switching, curved domains
#
# Landau-Khalatnikov / Thermodynamics:
#   [29] Hoffmann et al., J. Appl. Phys. 118, 072006 (2015). DOI: 10.1063/1.4927805
#   [30] Starschich et al., Appl. Phys. Lett. 108, 032903 (2016). DOI: 10.1063/1.4940370
#   [31] Park et al., J. Appl. Phys. 117, 074103 (2015) - Electrostriction coefficients (Q11/Q12)
#
# Depolarization / Negative Capacitance:
#   [32] Siannas et al., Commun. Phys. 5, 178 (2022). DOI: 10.1038/s42005-022-00951-x
#   [33] Zhou et al., Sci. Adv. 8, eadd5953 (2022). DOI: 10.1126/sciadv.add5953
#   [34] Salahuddin & Datta, Nano Lett. 8, 405–410 (2008). DOI: 10.1021/nl071804g
#
# Series Resistance / Compact Models:
#   [35] Chatterjee et al., UC Berkeley EECS-2018-131 (2018)
#   [36] Tung et al., UC Berkeley EECS-2025-13 (2025)
#   [37] Tung et al., IEEE Trans. Electron Devices 69, 4761–4764 (2022). DOI: 10.1109/TED.2022.3181573
#
# Analog/MLC FeFET:
#   [38] Jerry et al., arXiv:1708.04729 (2017)


# =============================================================================
# MATERIALS
# =============================================================================
# Three-tier system:
#   default_hzo: Baseline standard HZO from literature
#   fecim_hzo: FeCIM HZO conference-baseline values (conservative)
#   literature_superlattice: Best academic results (Cheema 2020)

materials:
  default_hzo:
    name: "HZO (Si-doped, Park 2015 midpoint)"  # Human-readable material/profile name.
    description: "Standard Si-doped HfO2-ZrO2 from literature"  # Short physical context and intended use-case.
    reference: "Park et al., Adv. Mater. 27, 1811 (2015) [1]"  # Primary literature/source backing this parameter set.
    analog_states: 30           # Standard HZO achieves ~30 discrete states
    target_range_frac: 0.90    # Back-off from saturation for level mapping

    # Polarization parameters [1][4] [claim: hzo-remanent-polarization-range]
    pr_c_m2: 0.245              # Remanent polarization (C/m²) = 24.5 µC/cm² midpoint of Park et al. 2015 range
    ps_c_m2: 0.30               # Saturation polarization (C/m²) = 30 µC/cm²

    # Field parameters [1][7]
    ec_v_m: 1.2e8               # Coercive field (V/m) = 1.2 MV/cm

    # Dielectric properties [7]
    epsilon_hf: 30              # High-frequency permittivity
    epsilon_lf: 38              # Low-frequency permittivity
    loss_tangent: 0.02          # Dielectric loss (tan δ)

    # Film geometry
    thickness_m: 10.0e-9        # Film thickness = 10 nm
    area_m2: 100.0e-12          # Cell area = 100 nm² (10nm x 10nm)

    # Switching dynamics [8]
    tau_s: 1.0e-9               # Switching time constant = 1 ns
    tau0_s: 1.0e-13             # Attempt frequency inverse
    activation_energy_ev: 0.7   # Activation energy for switching
    kai_exponent: 2.0           # KAI model exponent (2D domain growth)

    # Temperature properties (empirical Preisach temperature-scaling fit)
    # NOTE: 723K is the empirical tuned value used in Go presets for DefaultHZO Preisach
    # temperature scaling. Materlik 2015 LGD value (598K) is used in MaterlikHfO2() preset.
    curie_temp_k: 723           # Curie temperature (K) — empirical fit for Preisach T-scaling
    temp_coeff_ec: -2.0e5       # dEc/dT (V/m/K)
    temp_coeff_pr: -5.0e-5      # dPr/dT (C/m²/K)

    # Reliability [4][19][20]
    endurance_cycles: 1.0e10    # Endurance limit (verified)
    retention_time_s: 3.15e9    # 100 years at 85°C
    imprint_field_v_m: 1.0e6    # Imprint field shift

    # Depolarization field (polycrystalline slant) [32][33][34]
    # Physics rule: E_dep = -k_dep * P; for stack-based derivation use
    # k_dep ≈ (epsilon_fe/epsilon_dead) * (d_dead/d_fe).
    depolarization_factor_vm_c: 2.5e8  # Educational assumed value; typical HZO stack-model range ~1-5×10^8 V·m/C (e.g., Hoffmann et al., Adv. Funct. Mater., 2016)

    # Landau-Khalatnikov parameters (Materlik et al., J. Appl. Phys. 117, 134109 (2015), doi:10.1063/1.4916229) for β/γ; ρ kept as calibrated damping [claim: materlik-lgd-coefficients]
    thermodynamics:
      beta_landau: -6.720e8     # First-order barrier (Negative) [J m^5 / C^4]
      gamma_landau: 1.950e10    # Stability (Positive) [J m^9 / C^6]
      rho_viscosity: 0.05       # Damping/Viscosity (Ohm·m), tuned for ~1ns switching
      curie_const_k: 1.5e5      # Curie constant (K)

    # Electrostriction & stress coupling [31][30]
    coupling:
      q11_electrostriction: 0.089  # Longitudinal electrostriction (m^4 / C^2)
      q12_electrostriction: -0.026 # Transverse electrostriction (m^4 / C^2)
      stress_gpa: 1.0              # TiN capping stress (typical)

    # Circuit parasitics (series resistance) [35][36][37]
    circuit:
      series_resistance_ohm: 50    # Model default (literature uses 100Ω-scale fits)

    # Nucleation-Limited Switching (Merz law)
    # Literature anchors for τ₀/Ea-style NLS fits: Müller et al., IEEE Trans. Electron Devices (2013); Trentzsch et al., IEDM (2016).
    nls:
      activation_field_v_m: 1.2e9  # 12 MV/cm (model default)
      tau_inf_s: 1.0e-10           # 100 ps attempt time (model default)

    # Conductance transfer (FeFET mapping)
    # Educational placeholder: demo windows often 10:1-100:1; production FeFET ON/OFF can span ~10^2-10^5 (e.g., Jerry et al., IEDM 2017; Ni et al., IEEE EDL 2018).
    conductance:
      gmin_s: 1.0e-6               # High-resistance state (HRS), placeholder
      gmax_s: 100.0e-6             # Low-resistance state (LRS), placeholder

  fecim_hzo:
    name: "FeCIM HZO"  # Human-readable material/profile name.
    description: "FeCIM HZO - conference-baseline values only"  # Short physical context and intended use-case.
    reference: "COSM 2025 conference presentation [17]"  # Primary literature/source backing this parameter set.
    analog_states: 30           # 30 discrete states - conference baseline (unverified)
    target_range_frac: 0.90    # Back-off from saturation for level mapping

    # Depolarization field (polycrystalline slant) [32][33][34]
    depolarization_factor_vm_c: 2.5e8  # k_dep for E_dep = -k_dep * P

    # Polarization (estimated - not disclosed) [17][18]
    pr_c_m2: 0.30               # 30 µC/cm² (estimated from In2Se3 work)
    ps_c_m2: 0.35               # 35 µC/cm² (estimated)

    # Field parameters (estimated) [17]
    ec_v_m: 1.0e8               # 1.0 MV/cm (estimated)

    # Dielectric properties (estimated)
    epsilon_hf: 32  # Relative permittivity at high frequency (dimensionless, typically 8–40).
    epsilon_lf: 40  # Relative permittivity at low frequency (dimensionless, typically >= epsilon_hf).
    loss_tangent: 0.01          # Low loss

    # Film geometry [17]
    thickness_m: 10.0e-9        # 10 nm
    area_m2: 2.025e-15          # 45nm pitch = 45nm x 45nm
    cell_pitch_nm: 45           # Cell pitch

    # Switching dynamics [17]
    tau_s: 10.0e-9              # 10 ns (conference slide estimate, not 1ns)
    tau0_s: 1.0e-13  # Attempt time constant in seconds for thermally activated switching.
    activation_energy_ev: 0.6  # Activation barrier for switching in eV (typ. ~0.4–1.2 eV).
    kai_exponent: 2.0  # KAI/Avrami exponent (domain-growth dimensionality, typ. 1–3).

    # Temperature properties [7]
    curie_temp_k: 723  # Curie transition temperature in Kelvin.
    temp_coeff_ec: -1.8e5  # dEc/dT in (V/m)/K; usually negative near room temperature.
    temp_coeff_pr: -4.0e-5  # dPr/dT in (C/m²)/K; usually negative with increasing T.

    # Reliability - DEMONSTRATED values only [17][19]
    endurance_cycles: 1.0e9     # 10^9 DEMONSTRATED (10^12 is TARGET)
    retention_time_s: 1.0e7     # ~116 days DEMONSTRATED
    imprint_field_v_m: 1.0e6  # Built-in imprint/bias field in V/m shifting hysteresis center.

    # =========================================================================
    # UNIFIED THEORY PARAMETERS (L-K + Electrostriction) [Gemini Compendium]
    # =========================================================================
    thermodynamics:
      beta_landau: -6.720e8       # First-order barrier (Negative) [J m^5 / C^4]
      gamma_landau: 1.950e10      # Stability (Positive) [J m^9 / C^6]
      rho_viscosity: 0.05         # Damping/Viscosity (Ohm·m), tuned for ~1ns switching
      curie_const_k: 1.5e5        # Curie constant (K)
    
    coupling:
      q11_electrostriction: 0.089  # Longitudinal electrostriction (m^4 / C^2)
      q12_electrostriction: -0.026 # Tensile stability [m^4 / C^2]
      stress_gpa: 1.0              # TiN Capping Stress [GPa]

    # Circuit parasitics (series resistance) [35][36][37]
    circuit:
      series_resistance_ohm: 50    # Model default (literature uses 100Ω-scale fits)

    # Nucleation-Limited Switching (Merz law) [30]
    nls:
      activation_field_v_m: 1.2e9  # 12 MV/cm (within 13–19 MV/cm HZO range)
      tau_inf_s: 1.0e-10           # 100 ps attempt time (HZO typical)

    # Conductance transfer (FeFET mapping) [15]
    conductance:
      gmin_s: 1.0e-6               # High-resistance state (HRS)
      gmax_s: 100.0e-6             # Low-resistance state (LRS)

  fecim_hzo_target:
    name: "FeCIM HZO (TARGET)"  # Human-readable material/profile name.
    description: "FeCIM HZO - aspirational conference targets (not demonstrated)"  # Short physical context and intended use-case.
    reference: "COSM 2025 conference presentation"  # Primary literature/source backing this parameter set.
    analog_states: 30           # Same as FeCIM demonstrated
    target_range_frac: 0.90    # Back-off from saturation for level mapping

    # Depolarization field (polycrystalline slant) [32][33][34]
    depolarization_factor_vm_c: 2.5e8  # k_dep for E_dep = -k_dep * P

    # Same as fecim_hzo except reliability targets
    pr_c_m2: 0.30  # Remanent polarization Pr in C/m² (typ. HZO ~0.2–0.8 C/m²).
    ps_c_m2: 0.35  # Saturation polarization Ps in C/m² (must be >= Pr).
    ec_v_m: 1.0e8  # Coercive field Ec in V/m (typ. 0.8e8–6e8 V/m by material).
    epsilon_hf: 32  # Relative permittivity at high frequency (dimensionless, typically 8–40).
    epsilon_lf: 40  # Relative permittivity at low frequency (dimensionless, typically >= epsilon_hf).
    loss_tangent: 0.01  # Dielectric loss tangent tanδ (dimensionless, often 0.005–0.05).
    thickness_m: 10.0e-9  # Ferroelectric film thickness in meters (typ. 4e-9–20e-9 m).
    area_m2: 2.025e-15  # Active device/cell area in m² used for charge/energy scaling.
    cell_pitch_nm: 45  # Cell pitch in nm (layout spacing; process-dependent).
    tau_s: 1.0e-9               # TARGET: 1ns (demonstrated ~10ns)
    tau0_s: 1.0e-13  # Attempt time constant in seconds for thermally activated switching.
    activation_energy_ev: 0.6  # Activation barrier for switching in eV (typ. ~0.4–1.2 eV).
    kai_exponent: 2.0  # KAI/Avrami exponent (domain-growth dimensionality, typ. 1–3).
    curie_temp_k: 723  # Curie transition temperature in Kelvin.
    temp_coeff_ec: -1.8e5  # dEc/dT in (V/m)/K; usually negative near room temperature.
    temp_coeff_pr: -4.0e-5  # dPr/dT in (C/m²)/K; usually negative with increasing T.
    endurance_cycles: 1.0e12    # TARGET - not yet achieved
    retention_time_s: 3.15e9    # TARGET - 100 years (not demonstrated)
    imprint_field_v_m: 1.0e6  # Built-in imprint/bias field in V/m shifting hysteresis center.

    # =========================================================================
    # UNIFIED THEORY PARAMETERS (L-K + Electrostriction) [29][30]
    # =========================================================================
    thermodynamics:
      beta_landau: -6.720e8     # First-order barrier (Negative) [J m^5 / C^4]
      gamma_landau: 1.950e10    # Stability (Positive) [J m^9 / C^6]
      rho_viscosity: 0.05       # Damping/Viscosity (Ohm·m), tuned for ~1ns switching
      curie_const_k: 1.5e5      # Curie constant (K)

    coupling:
      q11_electrostriction: 0.089  # Longitudinal electrostriction (m^4 / C^2)
      q12_electrostriction: -0.026 # Tensile stability [m^4 / C^2]
      stress_gpa: 1.0              # TiN capping stress (typical)

    # Circuit parasitics (series resistance) [35][36][37]
    circuit:
      series_resistance_ohm: 50    # Model default (literature uses 100Ω-scale fits)

    # Nucleation-Limited Switching (Merz law) [30]
    nls:
      activation_field_v_m: 1.0e9  # 10 MV/cm target (lower barrier than demonstrated 12 MV/cm)
      tau_inf_s: 0.5e-10           # 50 ps target (faster than demonstrated 100 ps)

    # Conductance transfer (FeFET mapping) [15]
    conductance:
      gmin_s: 1.0e-6               # High-resistance state (HRS)
      gmax_s: 100.0e-6             # Low-resistance state (LRS)

  literature_superlattice:
    name: "Literature Superlattice"  # Human-readable material/profile name.
    description: "Best academic HfO2/ZrO2 superlattice results"  # Short physical context and intended use-case.
    reference: "Cheema et al., Nature 580, 478 (2020) [2]; Nature Commun. 2025 [3]"  # Primary literature/source backing this parameter set.
    # analog_states: 64           # Enhanced superlattice supports more states
    analog_states: 30           # Enhanced superlattice supports more states
    target_range_frac: 0.90    # Back-off from saturation for level mapping

    # Depolarization field (polycrystalline slant) [32][33][34]
    depolarization_factor_vm_c: 2.5e8  # k_dep for E_dep = -k_dep * P

    # Enhanced polarization — IEEE 10787441 2024: 2Pr=43.32; revised from Cheema 2020 (50 µC/cm²)
    pr_c_m2: 0.22               # 22 µC/cm² (IEEE 10787441 2024; PMC 12254504 2025)
    ps_c_m2: 0.27               # 27 µC/cm²

    # Reduced coercive field [3]
    ec_v_m: 0.85e8              # 0.85 MV/cm (Nature Commun. 2025)

    # Dielectric properties [2][7]
    epsilon_hf: 35  # Relative permittivity at high frequency (dimensionless, typically 8–40).
    epsilon_lf: 50  # Relative permittivity at low frequency (dimensionless, typically >= epsilon_hf).
    loss_tangent: 0.015  # Dielectric loss tangent tanδ (dimensionless, often 0.005–0.05).

    # Film geometry [3]
    thickness_m: 10.0e-9  # Ferroelectric film thickness in meters (typ. 4e-9–20e-9 m).
    area_m2: 100.0e-12  # Active device/cell area in m² used for charge/energy scaling.

    # Faster switching [8]
    tau_s: 0.36e-9              # 360 ps RECORD (Purdue sub-µm arrays)
    tau0_s: 1.0e-13  # Attempt time constant in seconds for thermally activated switching.
    activation_energy_ev: 0.5   # Lower barrier
    kai_exponent: 2.5  # KAI/Avrami exponent (domain-growth dimensionality, typ. 1–3).

    # Higher thermal stability [3][7]
    curie_temp_k: 773           # ~500°C
    temp_coeff_ec: -1.5e5  # dEc/dT in (V/m)/K; usually negative near room temperature.
    temp_coeff_pr: -3.0e-5  # dPr/dT in (C/m²)/K; usually negative with increasing T.

    # Literature best-case reliability [3][4][20]
    endurance_cycles: 1.0e10    # 10^10 verified [3][4], 10^12 claimed but rare
    retention_time_s: 3.15e8    # 10-year at 85°C verified
    imprint_field_v_m: 0.5e6  # Built-in imprint/bias field in V/m shifting hysteresis center.

    # =========================================================================
    # UNIFIED THEORY PARAMETERS (L-K + Electrostriction)
    # =========================================================================
    thermodynamics:
      beta_landau: -3.8e8         # Deeper wells for Superlattice
      gamma_landau: 2.1e10        # Higher stability
      rho_viscosity: 0.02         # Lower viscosity (faster switching: 360ps)
      curie_const_k: 1.5e5        # Curie constant (K), assumed HZO-like
    
    coupling:
      q11_electrostriction: 0.089  # Longitudinal electrostriction (m^4 / C^2)
      q12_electrostriction: -0.035 # Stronger coupling in superlattice
      stress_gpa: 1.5              # Higher epitaxial stress

    # Circuit parasitics (series resistance) [35][36][37]
    circuit:
      series_resistance_ohm: 50    # Model default (literature uses 100Ω-scale fits)

    # Nucleation-Limited Switching (Merz law) [30]
    nls:
      activation_field_v_m: 1.0e9  # 10 MV/cm (superlattice lower barrier)
      tau_inf_s: 0.5e-10           # 50 ps attempt time (faster switching)

    # Conductance transfer (FeFET mapping) [15]
    conductance:
      gmin_s: 0.5e-6               # Lower leakage (HRS)
      gmax_s: 150.0e-6             # Higher LRS due to higher Pr

  # NEW: Cryogenic operation profile (for quantum computing integration)
  cryogenic_hzo:
    name: "Cryogenic HZO"  # Human-readable material/profile name.
    description: "HZO performance at 4K for quantum computing"  # Short physical context and intended use-case.
    reference: "Adv. Electron. Mater. 2024 [10]; 2025 [11]"  # Primary literature/source backing this parameter set.
    analog_states: 48           # Enhanced polarization at cryo enables more states
    target_range_frac: 0.90    # Back-off from saturation for level mapping

    # Depolarization field (polycrystalline slant) [32][33][34]
    depolarization_factor_vm_c: 2.5e8  # k_dep for E_dep = -k_dep * P

    # ENHANCED polarization at cryogenic temps [10]
    pr_c_m2: 0.75               # 75 µC/cm² at 4K (RECORD - improved from RT!)
    ps_c_m2: 0.80               # 80 µC/cm²

    # Field parameters at 4K [10][11]
    ec_v_m: 1.5e8               # Higher Ec at cryo (~1.5 MV/cm)
    memory_window_v: 7.0        # 6-8V memory window at cryo

    # Dielectric properties [10]
    epsilon_hf: 28              # Slightly reduced at cryo
    epsilon_lf: 35  # Relative permittivity at low frequency (dimensionless, typically >= epsilon_hf).
    loss_tangent: 0.008         # Lower loss at cryo

    # Film geometry
    thickness_m: 10.0e-9  # Ferroelectric film thickness in meters (typ. 4e-9–20e-9 m).
    area_m2: 100.0e-12  # Active device/cell area in m² used for charge/energy scaling.

    # Switching at cryo [10]
    tau_s: 1.0e-9               # ~1 ns maintained
    tau0_s: 1.0e-13  # Attempt time constant in seconds for thermally activated switching.
    activation_energy_ev: 0.7  # Activation barrier for switching in eV (typ. ~0.4–1.2 eV).
    kai_exponent: 2.0  # KAI/Avrami exponent (domain-growth dimensionality, typ. 1–3).

    # Temperature [10][11]
    curie_temp_k: 723           # Curie transition temperature in Kelvin
    temp_coeff_ec: -2.0e5       # dEc/dT (V/m/K) - matches CryogenicHZO() preset
    temp_coeff_pr: -5.0e-5      # dPr/dT (C/m²/K) - matches CryogenicHZO() preset
    operating_temp_k: 4         # 4 Kelvin
    linearity_improvement: true # Better linearity below 100K

    # Reliability at cryo [10]
    endurance_cycles: 1.0e9     # 10^9 at ±5V verified at cryo
    retention_time_s: 3.15e10   # >10 years at cryo (improved)
    imprint_field_v_m: 0.3e6    # Reduced imprint at cryo

    # =========================================================================
    # UNIFIED THEORY PARAMETERS (L-K + Electrostriction)
    # Cryo-specific LGD coefficients are not widely published; defaults match 10nm HZO [29][30]
    # =========================================================================
    thermodynamics:
      beta_landau: -6.720e8     # HZO default (room-temp reference)
      gamma_landau: 1.950e10    # HZO default (room-temp reference)
      rho_viscosity: 0.05       # Damping/Viscosity (Ohm·m), model default
      curie_const_k: 1.5e5      # Curie constant (K), model default

    coupling:
      q11_electrostriction: 0.089  # Longitudinal electrostriction (m^4 / C^2)
      q12_electrostriction: -0.026 # Transverse electrostriction (m^4 / C^2)
      stress_gpa: 1.0              # TiN capping stress (typical)

    # Circuit parasitics (series resistance) [35][36][37]
    circuit:
      series_resistance_ohm: 50    # Model default (literature uses 100Ω-scale fits)

    # Nucleation-Limited Switching (Merz law) [30]
    nls:
      activation_field_v_m: 1.5e9  # 15 MV/cm (higher barrier at cryo)
      tau_inf_s: 2.0e-10           # 200 ps attempt time (slower at cryo)

    # Conductance transfer (FeFET mapping) [15]
    conductance:
      gmin_s: 0.5e-6               # Lower leakage at cryo (HRS)
      gmax_s: 200.0e-6             # Higher LRS due to higher Pr

  # ---------------------------------------------------------------------------
  # MULTI-LEVEL DEMONSTRATIONS (Peer-Reviewed)
  # ---------------------------------------------------------------------------

  hzo_standard_32:
    name: "HZO Standard (32 states)"  # Human-readable material/profile name.
    description: "Standard HZO FeFET demonstrating 32 analog states"  # Short physical context and intended use-case.
    reference: "Oh et al., IEEE EDL 38(6), 732 (2017) [21]"  # Primary literature/source backing this parameter set.
    analog_states: 32           # PEER-REVIEWED: 32 discrete states
    target_range_frac: 0.90    # Back-off from saturation for level mapping

    # Depolarization field (polycrystalline slant) [32][33][34]
    depolarization_factor_vm_c: 2.5e8  # k_dep for E_dep = -k_dep * P

    # Polarization parameters [21]
    pr_c_m2: 0.20               # 20 µC/cm²
    ps_c_m2: 0.25               # 25 µC/cm²

    # Field parameters [21]
    ec_v_m: 1.0e8               # 1.0 MV/cm

    # Dielectric properties
    epsilon_hf: 28  # Relative permittivity at high frequency (dimensionless, typically 8–40).
    epsilon_lf: 35  # Relative permittivity at low frequency (dimensionless, typically >= epsilon_hf).
    loss_tangent: 0.02  # Dielectric loss tangent tanδ (dimensionless, often 0.005–0.05).

    # Film geometry
    thickness_m: 10.0e-9        # 10 nm
    area_m2: 100.0e-12  # Active device/cell area in m² used for charge/energy scaling.

    # Switching dynamics
    tau_s: 10.0e-9              # 10 ns
    tau0_s: 1.0e-13  # Attempt time constant in seconds for thermally activated switching.
    activation_energy_ev: 0.7  # Activation barrier for switching in eV (typ. ~0.4–1.2 eV).
    kai_exponent: 2.0  # KAI/Avrami exponent (domain-growth dimensionality, typ. 1–3).

    # Temperature
    curie_temp_k: 723  # Curie transition temperature in Kelvin.
    temp_coeff_ec: -2.0e5       # dEc/dT (V/m/K) - matches HZOStandard32() preset
    temp_coeff_pr: -5.0e-5      # dPr/dT (C/m²/K) - matches HZOStandard32() preset

    # Reliability [21]
    endurance_cycles: 1.0e8     # 10^8 cycles (2017 era)
    retention_time_s: 3.15e8    # 10 years projected
    imprint_field_v_m: 1.0e6    # DC-bias-induced asymmetry - matches HZOStandard32() preset

    # =========================================================================
    # UNIFIED THEORY PARAMETERS (L-K + Electrostriction) [29][30]
    # =========================================================================
    thermodynamics:
      beta_landau: -6.720e8     # First-order barrier (Negative) [J m^5 / C^4]
      gamma_landau: 1.950e10    # Stability (Positive) [J m^9 / C^6]
      rho_viscosity: 0.05       # Damping/Viscosity (Ohm·m), tuned for ~1ns switching
      curie_const_k: 1.5e5      # Curie constant (K)

    coupling:
      q11_electrostriction: 0.089  # Longitudinal electrostriction (m^4 / C^2)
      q12_electrostriction: -0.026 # Transverse electrostriction (m^4 / C^2)
      stress_gpa: 1.0              # TiN capping stress (typical)

    # Circuit parasitics (series resistance) [35][36][37]
    circuit:
      series_resistance_ohm: 50    # Model default (literature uses 100Ω-scale fits)

    # Nucleation-Limited Switching (Merz law) [30]
    nls:
      activation_field_v_m: 1.2e9  # 12 MV/cm (within 13–19 MV/cm HZO range)
      tau_inf_s: 1.0e-10           # 100 ps attempt time (HZO typical)

    # Conductance transfer (FeFET mapping) [21]
    conductance:
      gmin_s: 2.0e-6               # Older HZO FeFET HRS
      gmax_s: 80.0e-6              # Older HZO FeFET LRS

  hzo_ftj_140:
    name: "HZO FTJ (140 states)"  # Human-readable material/profile name.
    description: "HZO Ferroelectric Tunnel Junction with >7-bit operation"  # Short physical context and intended use-case.
    reference: "Song et al., Adv. Science 2024 [22]; ACS AMI 2022 [23]"  # Primary literature/source backing this parameter set.
    analog_states: 140          # PEER-REVIEWED: >7-bit = 128-140 states
    target_range_frac: 0.90    # Back-off from saturation for level mapping

    # Depolarization field (polycrystalline slant) [32][33][34]
    depolarization_factor_vm_c: 2.5e8  # k_dep for E_dep = -k_dep * P (model default)

    # Polarization parameters [22][23]
    pr_c_m2: 0.25               # 25 µC/cm²
    ps_c_m2: 0.30               # 30 µC/cm²

    # Field parameters - FTJ requires thinner films
    ec_v_m: 1.2e8               # 1.2 MV/cm

    # Dielectric properties
    epsilon_hf: 30  # Relative permittivity at high frequency (dimensionless, typically 8–40).
    epsilon_lf: 40  # Relative permittivity at low frequency (dimensionless, typically >= epsilon_hf).
    loss_tangent: 0.015  # Dielectric loss tangent tanδ (dimensionless, often 0.005–0.05).

    # Film geometry - ULTRATHIN for tunneling [22][23]
    thickness_m: 4.5e-9         # 4.5 nm (tunnel barrier)
    area_m2: 100.0e-12  # Active device/cell area in m² used for charge/energy scaling.

    # Switching dynamics - FTJ is fast [23]
    tau_s: 20.0e-9              # 20 ns switching
    tau0_s: 1.0e-13  # Attempt time constant in seconds for thermally activated switching.
    activation_energy_ev: 0.5   # Lower barrier for tunneling
    kai_exponent: 2.0  # KAI/Avrami exponent (domain-growth dimensionality, typ. 1–3).

    # FTJ-specific parameters [22][23]
    ter_ratio: 834              # Tunneling electroresistance ratio
    gmax_gmin_ratio: 1.0e5      # >10^5 conductance ratio

    # Temperature
    curie_temp_k: 723  # Curie transition temperature in Kelvin.
    temp_coeff_ec: -2.0e5       # dEc/dT (V/m/K) - matches HZOFJT140() preset
    temp_coeff_pr: -5.0e-5      # dPr/dT (C/m²/K) - matches HZOFJT140() preset

    # Reliability [22]
    endurance_cycles: 1.0e7     # 10^7 cycles demonstrated
    retention_time_s: 3.15e8    # 10 years extrapolated
    imprint_field_v_m: 1.0e6    # DC-bias-induced asymmetry - matches HZOFJT140() preset

    # =========================================================================
    # UNIFIED THEORY PARAMETERS (L-K + Electrostriction) [29][30]
    # =========================================================================
    thermodynamics:
      beta_landau: -6.720e8     # First-order barrier (Negative) [J m^5 / C^4]
      gamma_landau: 1.950e10    # Stability (Positive) [J m^9 / C^6]
      rho_viscosity: 0.05       # Damping/Viscosity (Ohm·m), tuned for ~1ns switching
      curie_const_k: 1.5e5      # Curie constant (K)

    coupling:
      q11_electrostriction: 0.089  # Longitudinal electrostriction (m^4 / C^2)
      q12_electrostriction: -0.026 # Transverse electrostriction (m^4 / C^2)
      stress_gpa: 1.0              # TiN capping stress (typical)

    # Circuit parasitics (series resistance) [35][36][37]
    circuit:
      series_resistance_ohm: 50    # Model default (literature uses 100Ω-scale fits)

    # Nucleation-Limited Switching (Merz law) [30]
    nls:
      activation_field_v_m: 1.0e9  # 10 MV/cm (lower barrier for FTJ)
      tau_inf_s: 2.0e-10           # 200 ps attempt time (FTJ tunneling)

    # Conductance transfer (FTJ mapping) [22][23]
    conductance:
      gmin_s: 0.1e-6               # FTJ HRS (lower absolute current)
      gmax_s: 10.0e-6              # FTJ LRS (higher TER)

  # ---------------------------------------------------------------------------
  # ALTERNATIVE FERROELECTRICS (Lower TRL for 30-state applications)
  # ---------------------------------------------------------------------------

  alscn:
    name: "AlScN (8-16 states)"  # Human-readable material/profile name.
    description: "Aluminum Scandium Nitride - high Pr but high Ec limits granularity"  # Short physical context and intended use-case.
    reference: "Nature Commun. 2025 [24]; APL Materials 2023 [25]"  # Primary literature/source backing this parameter set.
    analog_states: 12           # Conservative midpoint (8-16 range) - matches AlScN() preset
    target_range_frac: 0.90    # Back-off from saturation for level mapping
    trl_level: 4                # Lower technology readiness

    # Polarization - VERY HIGH [24][25]
    pr_c_m2: 1.20               # 120 µC/cm² (range: 100-172 µC/cm²)
    ps_c_m2: 1.50               # 150 µC/cm²

    # Field parameters - VERY HIGH Ec limits state granularity [24][25]
    ec_v_m: 5.0e8               # 5.0 MV/cm (range: 2-6 MV/cm)
    memory_window_v: 15.0       # 15V memory window [24]

    # Dielectric properties [25]
    epsilon_hf: 10              # Lower permittivity than HZO
    epsilon_lf: 12  # Relative permittivity at low frequency (dimensionless, typically >= epsilon_hf).
    loss_tangent: 0.01  # Dielectric loss tangent tanδ (dimensionless, often 0.005–0.05).

    # Film geometry
    thickness_m: 20.0e-9        # 20 nm typical
    area_m2: 100.0e-12  # Active device/cell area in m² used for charge/energy scaling.

    # Switching dynamics
    tau_s: 10.0e-9              # ~10 ns
    tau0_s: 1.0e-13  # Attempt time constant in seconds for thermally activated switching.
    activation_energy_ev: 1.0   # Higher barrier
    kai_exponent: 2.0  # KAI/Avrami exponent (domain-growth dimensionality, typ. 1–3).

    # Temperature - VERY HIGH Curie temp [25]
    curie_temp_k: 1273          # >1000°C - excellent thermal stability

    # Composition (typical)
    sc_fraction: 0.36           # Al0.64Sc0.36N optimal

    # Reliability [24]
    endurance_cycles: 1.0e9     # 10^9 cycles
    retention_time_s: 3.15e8  # Retention horizon in seconds at stated condition (often extrapolated).

    # Conductance transfer - matches AlScN() preset
    conductance:
      gmin_s: 0.2e-6               # 0.2 µS HRS
      gmax_s: 300.0e-6             # 300 µS LRS

    # Limitations
    # - High Ec (5 MV/cm) requires high voltages, limits analog granularity
    # - Best for binary/few-level rather than 30-state applications

  # ---------------------------------------------------------------------------
  # 2D FERROELECTRIC: In₂Se₃ (flash-within-flash synthesis)
  # ---------------------------------------------------------------------------

  in2se3:
    name: "α-In₂Se₃"  # Human-readable material/profile name.
    description: "2D van der Waals ferroelectric semiconductor - flash-within-flash synthesis"  # Short physical context and intended use-case.
    reference: "Shin et al., Adv. Electron. Mater. 2025 [18]; Nature Commun. 2017 [26]"  # Primary literature/source backing this parameter set.
    analog_states: 30           # 30-state conference baseline
    target_range_frac: 0.90    # Back-off from saturation for level mapping
    trl_level: 4                # Research stage, not yet commercialized

    # =========================================================================
    # POLARIZATION PARAMETERS
    # In₂Se₃ has BOTH out-of-plane AND in-plane ferroelectricity
    # Values are lower than HZO but sufficient for synaptic devices
    # =========================================================================

    # Out-of-plane polarization [26][28]
    # Calculated dipole: 0.11 eÅ/unit cell → ~3-5 µC/cm²
    pr_c_m2: 0.04               # Remanent polarization = 4 µC/cm² (typical for 2D FE)
    ps_c_m2: 0.06               # Saturation polarization = 6 µC/cm²

    # In-plane polarization component (unique to In₂Se₃)
    pr_inplane_c_m2: 0.02       # In-plane Pr ~2 µC/cm²

    # =========================================================================
    # COERCIVE FIELD
    # Thickness-dependent: increases drastically for ultrathin films
    # =========================================================================
    ec_v_m: 3.0e8               # Coercive field = 3.0 MV/cm (thin film, >3 MV/cm reported)
    ec_thin_v_m: 5.0e8          # Ec for ultrathin (<5nm): ~5 MV/cm
    ec_thick_v_m: 1.0e8         # Ec for thicker films (>20nm): ~1 MV/cm

    # =========================================================================
    # DIELECTRIC & ELECTRONIC PROPERTIES
    # In₂Se₃ is a SEMICONDUCTOR (unlike insulating HZO)
    # =========================================================================
    epsilon_hf: 9               # High-frequency permittivity (2D material, lower than HZO)
    epsilon_lf: 12              # Low-frequency permittivity
    loss_tangent: 0.01          # Low dielectric loss
    bandgap_ev: 1.3             # Direct bandgap ~1.3 eV (semiconductor!)

    # =========================================================================
    # FILM GEOMETRY
    # Van der Waals layered structure: quintuple layers (QL)
    # Each QL is ~1nm thick (Se-In-Se-In-Se)
    # =========================================================================
    thickness_m: 10.0e-9        # 10 nm (~10 quintuple layers)
    min_thickness_m: 1.0e-9     # Single QL ferroelectric (unique to 2D FE!)
    quintuple_layer_nm: 1.0     # Each QL ~1 nm
    area_m2: 100.0e-12          # Cell area = 100 nm²

    # =========================================================================
    # SWITCHING DYNAMICS
    # 2D materials can have fast domain wall motion
    # =========================================================================
    tau_s: 10.0e-9              # Switching time ~10 ns (conference demo estimate)
    tau0_s: 1.0e-13             # Attempt frequency inverse
    activation_energy_ev: 0.5   # Lower barrier than HZO (2D switching)
    kai_exponent: 2.0           # Domain growth exponent

    # =========================================================================
    # TEMPERATURE PROPERTIES
    # HIGH Curie temperature - room temperature ferroelectric
    # =========================================================================
    curie_temp_k: 700           # Tc ~700K (verified, room-temp stable) [26][28]
    temp_coeff_ec: -1.0e5       # dEc/dT (V/m/K)
    temp_coeff_pr: -2.0e-5      # dPr/dT (C/m²/K)

    # =========================================================================
    # RELIABILITY
    # 2D materials have excellent endurance potential
    # =========================================================================
    endurance_cycles: 1.0e8     # Estimated ~10^8 cycles (research stage)
    retention_time_s: 3.15e7    # ~1 year demonstrated
    imprint_field_v_m: 0.5e6    # Lower imprint than HZO

    # =========================================================================
    # 2D-SPECIFIC PROPERTIES
    # Van der Waals material with unique characteristics
    # =========================================================================
    vdw_material: true          # Van der Waals layered structure
    stacking: "3R"              # 3R stacking (large hysteresis) vs 2H (small hysteresis)
    phase: "alpha"              # α-In₂Se₃ (ferroelectric) vs β' (antiferroelectric)

    # Phase transition
    alpha_to_beta_temp_k: 473   # α→β' phase transition ~200°C

    # =========================================================================
    # SYNAPTIC DEVICE PARAMETERS (from the flash-within-flash synthesis paper)
    # =========================================================================
    synaptic:
      potentiation_pulses: 100  # Pulses for LTP
      depression_pulses: 100    # Pulses for LTD
      pulse_width_s: 1.0e-6     # 1 µs pulse width
      pulse_voltage_v: 3.0      # Programming voltage
      nonlinearity_ltp: 0.8     # LTP nonlinearity factor
      nonlinearity_ltd: 0.9     # LTD nonlinearity factor

    # Conductance transfer (FeFET mapping) [18]
    conductance:
      gmin_s: 0.5e-6            # High-resistance state (HRS)
      gmax_s: 50.0e-6           # Low-resistance state (LRS)
      on_off_ratio: 100         # Gmax/Gmin ratio

    # =========================================================================
    # SYNTHESIS METHOD (Flash-within-Flash)
    # Unique to the flash-within-flash process - enables gram-scale production
    # =========================================================================
    synthesis:
      method: "FWF"             # Flash-within-Flash Joule heating
      scale: "gram"             # Gram-scale synthesis demonstrated
      precursors: ["In", "Se"]  # Elemental precursors
      synthesis_time_s: 1.0     # ~1 second synthesis
      energy_reduction: 0.5     # 50% energy reduction vs conventional

    # =========================================================================
    # LANDAU-KHALATNIKOV PARAMETERS (for Frankenstein equation)
    # =========================================================================
    # STATUS: ⚠️ DERIVED ESTIMATES - No explicit peer-reviewed L-K fits exist
    #
    # Derivation methodology: [42][43]
    #   - β estimated from: β ≈ 4·Pr³/(27·Ec²·ε₀) → ~3×10⁹ J·m⁵/C⁴ [42]
    #   - γ scaled from BaTiO₃ proportional to 1/Ps⁴
    #   - ρ constrained by: ρ < 0.1 Ω·m for GHz operation [44]
    #   - Energy barrier constraint: 66 meV/unit cell [26]
    #
    # RECOMMENDED: Request supplementary materials from JAP 2025 [42] for
    #              PSO-fitted coefficients specific to α-In₂Se₃
    # =========================================================================

    # Depolarization factor k_dep for E_dep = -k_dep * P
    # Geometry-dependent, not material-specific [32][33]
    # In₂Se₃ has exceptional depolarization resistance due to interlocked polarization [26]
    depolarization_factor_vm_c: 1.5e8  # DERIVED: scaled from HZO by Pr/ε ratio

    # Landau-Devonshire thermodynamic coefficients [42][43]
    thermodynamics:
      # β > 0 for first-order (note: sign convention varies in literature)
      # DERIVED: β ≈ 4·Pr³/(27·Ec²·ε₀) with Pr=0.04 C/m², Ec=3×10⁸ V/m
      beta_landau: 3.0e9        # DERIVED [42]: First-order barrier [J m^5 / C^4]
      # γ > 0 ensures stability at high polarization
      # DERIVED: scaled from BaTiO₃ (α₁₁₁ ≈ 1.3×10⁹) by (Ps_BTO/Ps_In2Se3)⁴
      gamma_landau: 5.0e10      # DERIVED: Stability [J m^9 / C^6]
      # Viscosity determines switching dynamics: ρ_eff = ρ + R*A/d
      # CONSTRAINT: ρ < 0.1 Ω·m required for GHz operation [44]
      # In₂Se₃ exhibits "ultralow damping" in domain wall motion [39][45]
      rho_viscosity: 0.05       # DERIVED [44]: Damping (Ohm·m), ≤0.1 for 10ns switching
      # Curie constant for α calculation via Curie-Weiss law
      # DERIVED: estimated from Tc=700K, no Curie-Weiss analysis published
      curie_const_k: 1.0e5      # DERIVED: Curie constant (K)

    # Electromechanical coupling [31]
    coupling:
      # Q12 transverse electrostriction coefficient
      # DERIVED: DFT-calculated but not published [42]; scaled from oxides
      # 2D vdW materials expected to have weaker coupling than bulk
      q12_electrostriction: -0.01  # DERIVED: Transverse electrostriction (m^4 / C^2)

    # Circuit model parameters [35][36][37]
    circuit:
      # Series resistance affects ρ_eff and RC dynamics
      # 2D materials typically have higher contact resistance
      series_resistance_ohm: 100   # ESTIMATED: 2D contact resistance

    # Nucleation-Limited Switching (NLS) parameters [45]
    nls:
      # Activation field for domain nucleation
      # In₂Se₃ switching follows Arrhenius dynamics [45]
      # Ea ≈ 66 meV → Ea/kT at 300K ≈ 2.5; Ea_field ≈ Ea/(q·d) ≈ 6.6×10⁷ V/m
      activation_field_v_m: 6.6e7  # DERIVED [26][45]: From 66 meV barrier
      # Intrinsic attempt time (domain nucleation frequency)
      tau_inf_s: 1.0e-12           # ESTIMATED: Attempt period (~1 ps)

    # =========================================================================
    # NOTES & LIMITATIONS
    # =========================================================================
    # Advantages:
    # - 2D material: atomically thin, flexible, stackable
    # - Semiconductor: enables FeFET without separate channel
    # - Both in-plane and out-of-plane polarization
    # - High Tc (700K) - room temperature stable
    # - Gram-scale FWF synthesis
    #
    # Limitations:
    # - Lower Pr than HZO (~4 µC/cm² vs 25-50 µC/cm²)
    # - High Ec for thin films (>3 MV/cm)
    # - Phase stability concerns (α↔β' transition at 200°C)
    # - Lower TRL than HZO (research stage)
    # - Limited endurance data