Hydrogenated NdNiO₃ Junction Devices

• Hydrogenated NdNiO₃ junction devices are solid-state systems that exploit proton doping to trigger metal–insulator transitions and enable tunable electronic, ferroelectric, and neuromorphic functions.
• Controlled hydrogen incorporation via Pd-catalyzed spillover induces structural distortions and resistive switching, achieving multi-level memory and field-tunable polarization including negative differential capacitance.
• Device optimization through anneal parameters, substrate strain, and electrode geometry demonstrates potential for steep-slope transistors, neuromorphic synapses, and reconfigurable logic circuits.

Hydrogenated NdNiO₃ junction devices are solid-state structures that leverage hydrogen (proton) doping in epitaxial NdNiO₃ (NNO) thin films to enable tunable electrical, ferroelectric, and neuromorphic functionalities. By incorporating interstitial protons via catalytic spillover from Pd electrodes, the metallic phase of NNO is converted into a polarizable insulating phase with rich ionic and dipolar mechanisms that can be engineered for multi-level switching, negative differential capacitance, and fast, spatially-coupled memory architectures. These devices exhibit field-dependent polarization, space-charge effects, resistive switching via ionic redistribution, and structural phase distortion, positioning hydrogenated NNO as a core material for next-generation neuromorphic, steep-slope, and intelligent hardware systems (Yuan et al., 2023, Zhou et al., 27 Dec 2025, Gamage et al., 2023).

1. Materials Synthesis and Hydrogen Incorporation

Hydrogenation of NdNiO₃ is achieved via annealing thin epitaxial films (typical thicknesses 50–150 nm) in forming gas (5% H₂, 95% inert carrier such as Ar or N₂) at moderate temperatures (115–200°C). Pd top electrodes, patterned via lift-off or e-beam lithography, catalyze H₂ dissociation into protons and electrons, which are driven into the NNO lattice:

• H₂ → 2H_i^+ + 2e⁻ (spillover at Pd)
• Each interstitial H_i^+ donates its electron to Ni–O hybrid states, converting local Ni³⁺ to Ni²⁺.
• Elastic recoil detection and nano-FTIR confirm proton concentrations (up to ~10²² cm⁻³) and vibrational signatures (transverse/longitudinal OH stretches at ~1064 and ~3000 cm⁻¹) (Gamage et al., 2023).
• X-ray diffraction identifies an out-of-plane c-axis expansion up to 6.7% in heavily doped regions, indicating strong local octahedral distortion (Yuan et al., 2023).

Device platforms typically use NdNiO₃ films grown on perovskite substrates (LaAlO₃, Nb:SrTiO₃) or SiO₂/Si for CMOS compatibility, with electrode configuration and annealing parameters crucial to tailoring H gradients and ionic profiles (Zhou et al., 27 Dec 2025).

2. Electronic Structure Modification and Metal–Insulator Transition

Hydrogen doping produces a pronounced metal–insulator transition:

• Pristine NNO displays metallic transport (ρ < 0.01 Ω·cm), with the Fermi level crossing Ni 3d–O 2p hybrid bands.
• H‐doped NNO ("H-NNO") becomes insulating (ρ ≈ 6 × 10³ Ω·cm) at room temperature due to electron localization and reduction of Ni³⁺ to Ni²⁺.
• DFT+U calculations report a wide gap (>2.5 eV) in H-NNO, with PDOS showing complete absence of H states near E_F and clear Ni²⁺ band splitting (Yuan et al., 2023).
• Carrier density decreases dramatically (n_e < 10¹⁶ cm⁻³), and local conductivity σ scales exponentially with H content: σ(n_H,T) = σ₀ exp[−E_a / (k_B T)], E_a in the 0.3–0.8 eV range as inferred from DFT/MD (Gamage et al., 2023).

Local hydrogen distribution can induce nanoscale modulation of conductivity, giving rise to alternating metallic/insulating stripes perpendicular to bias direction in operando, with stripe periodicity ~200–500 nm (Gamage et al., 2023).

3. Polarization Mechanisms and Field-Tunable Capacitance

Hydrogen in NNO creates composite polarization effects:

• Dipolar polarization arises from polar structural distortion, as interstitial H breaks inversion symmetry; DFT finds Berry-phase polarizations ranging from several μC/cm² to tens of μC/cm², depending on H configuration.
• Space-charge polarization dominates at high fields (E > E_th ≈ 243 kV/cm): long-range proton drift and trapping yield ionic polarization (P_space) up to ~158 μC/cm² (Yuan et al., 2023).
• The total field-dependent polarization P(E) follows an ionic-hopping model:

$$P(E) ≃ P_0 \exp \left[ \frac{q d E}{2 k_B T} - \frac{\Phi_B}{k_B T} \right]$$

where q = proton charge, d ≈ hop distance, Φ_B = barrier.

• Thin-film capacitors exhibit transient negative differential capacitance (NDC): during rapid voltage steps, charge increases while voltage across the film "snaps back," observable as dQ/dV < 0 (Yuan et al., 2023).

Relaxation kinetics show stretched-exponential decay:

$\Delta P(t) = \Delta P_0 \exp[ -(t/\tau)^\beta ]$

with τ ≈ 8 ms–1 s; after 1 s, polarization decays to ~20%, mimicking leaky integrate-and-fire dynamics in neuromorphic systems (Yuan et al., 2023).

4. Junction Device Architectures and Ionic Dynamics

Hydrogenated NNO junctions are engineered in symmetric and asymmetric formats:

• Symmetric Pd–Pd junctions: Both electrodes catalyze H insertion; under bias, proton clouds expand/shrink, yielding volatile, short-term fading memory (nanosecond dynamics).
• Asymmetric Pd–Au junctions: Only the Pd contact drives H doping; the Au side remains undoped, enabling non-volatile multilevel switching (long-term memory element) (Zhou et al., 27 Dec 2025).
• Band diagrams show Schottky barriers at Pd/H-NNO interfaces, modulated by local proton concentration.
• Transport characteristics:
  • I–V sweeps reveal hysteretic resistive switching, with 16 programmable resistance states over ~40 kΩ–350 kΩ, retention >10⁴ s, and sub-nJ programming energy.
  • Current decay after pulse stimuli follows double-exponential behavior set by integration and proton diffusion timescales:

  $$I(t) = I_0 [1 - \exp(-Nt/\tau_{int})] \exp(-t/\tau_{pd})$$

  with τ_pd = L²/D_H, where L ≈ 3–5 μm and D_H the proton diffusivity.

5. Nanoscale Ionic Patterns and Device-Level Switching

Operando nanoimaging and high-resolution XRD elucidate atomic-scale switching processes:

• Stripe-phase formation under applied E-field manifests as periodic metallic (low-H) and insulating (high-H) domains perpendicular to current, with resistive switching and multi-level conductance behavior linked directly to stripe nucleation and remelting (Gamage et al., 2023).
• Out-of-plane lattice expansion (Δc/c₀ up to +6%) yields volumetric strain, raising the migration barrier for protons and stabilizing patterns against back-diffusion; DFT predicts octahedral volume expansion from ~6.7% to >21% at high H concentration (Gamage et al., 2023).
• Device metrics: on/off ratio ≈ 10, switching dynamics saturating after ~200 s, with volatile to quasi-nonvolatile retention determined by strain and H distribution.

Optimization involves controlling anneal parameters, substrate choice (strain engineering), electrode geometry, and channel patterning to tune H gradients, switching energy, and endurance (Gamage et al., 2023).

6. Neuromorphic and Intelligent Hardware Applications

Hydrogenated NdNiO₃ junction arrays enable spatiotemporally rich and efficient neuromorphic computation:

• Spatiotemporal processing using symmetric Pd–Pd networks: Each node's conductance is dynamically modulated by proton migration and global coupling through substrate-mediated potentials. The node states x(t) evolve as

$x(t) = f(W_{dyn}(t-1)x(t-1) + u(t))$

with W_{dyn} representing input-dependent coupling between nodes.

• Programmable output layers using asymmetric Pd–Au junctions: Static weight matrices W_out are realized by programming distinct resistance levels; final output is

$y(t) = W_{out} x(t)$

Performance metrics:

• Spoken-digit classification reaches 95.3% accuracy (AudioMNIST), energy cost ≈ 0.2 nJ per node per pulse, sub-μs operation (Zhou et al., 27 Dec 2025).
• Early seizure detection achieves up to 85% accuracy at 3 s post-onset (CHB-MIT EEG).
• CMOS compatibility: H-NNO on SiO₂/Si shows analogous switching; device density, pad spacing, and endurance (>10⁷ cycles) are lithographically tunable.

A plausible implication is that intrinsic fast ionic and dipolar relaxation in H-NNO junctions mimics essential features of biological synapses and neurons, including leaky integration, multilevel memory, and spatially-coupled computation, with scalability to full wafer arrays (Zhou et al., 27 Dec 2025).

7. Engineering Considerations and Future Directions

Design optimization of hydrogenated NdNiO₃ junctions focuses on:

• Controlled hydrogen concentration via anneal time/temperature for targeted on/off ratios and switching time.
• Strain engineering through substrate selection to modulate diffusion barriers and retention.
• Electrode geometry (symmetry, material) for precise control of junction polarity and proton profiles.
• Nanopatterning (channels, gaps) for lower switching thresholds and higher device density.
• Co-doping (oxygen vacancies) and thermal management to fine-tune activation energies and endurance.

Collectively, these approaches aim to harness the full potential of field-tunable polarization, negative differential capacitance, and resistive switching enabled by hydrogenated NNO for advanced applications in steep-slope transistors, neuromorphic synapses, and reconfigurable logic circuits (Yuan et al., 2023, Zhou et al., 27 Dec 2025, Gamage et al., 2023).