Interface-Dipole Engineering

• Interface-dipole engineering is the deliberate manipulation of charge redistribution at material interfaces to create potential steps that control electronic properties.
• It applies to a variety of systems—oxides, organics, 2D materials, and ferroelectrics—to optimize phenomena like rectification, tunneling barriers, and carrier mobility.
• Experimental and computational techniques such as DFT, spectroscopy, and STEM are used to validate the impact of interface dipoles on band alignment and device performance.

Searching arXiv for the specified paper and closely related interface-dipole engineering literature for citation support. Searching for the 2025 MoS2 homojunction paper and other interface-dipole papers across oxides, organics, 2D contacts, ferroelectric junctions, and van der Waals systems. Interface-dipole engineering is the deliberate control of a net dipolar charge distribution localized at an interface that produces an internal electric field and a potential step across the junction. Across oxide heterostructures, organic donor–acceptor contacts, ferroelectric tunnel junctions, metal/semiconductor contacts, graphene devices, and fully two-dimensional lateral homojunctions, this potential step shifts vacuum levels and work functions, modifies Schottky barrier heights, tunneling barriers, and charge injection energetics, and can thereby control rectification, tunneling electroresistance, Fermi-level pinning, carrier mobility, and interlayer hybridization (Eckmann et al., 22 Sep 2025, Borisevich et al., 2011, Sai et al., 2012).

1. Electrostatic definition and governing relations

At an interface, the dipole moment per unit area is commonly expressed through the charge-density redistribution normal to the junction. For organic donor–acceptor interfaces, the plane-averaged form is written as

$$p/A = \int_{-\infty}^{\infty} z\,\Delta \rho(z)\,dz,$$

with the associated vacuum-level shift

$\Delta \Phi = -\frac{p/A}{\epsilon_0}.$

In a discrete molecular picture, this becomes

$\Delta \Phi = -\frac{\mu_\perp n}{\epsilon_0},$

where $\mu_\perp$ is the dipole component per molecule normal to the interface and $n$ is the surface molecular areal density (Sai et al., 2012).

For metal–semiconductor contacts, the same electrostatic step enters directly into the barrier formula,

$\Phi_{Bn} = \Phi_M - \chi_s - \Delta V,$

so that the dipole corrects the ideal Schottky–Mott alignment. In the unified bond-dipole theory, the potential step is written as

$$\Delta V = \frac{e\,a_p\,\sigma_s\,d}{\epsilon_0\epsilon_r} = c\,a_p,$$

with $\sigma_s$ the surface density of available dangling-bond orbitals, $a_p$ the bond polarity, $d$ an effective dipole length, and $\Delta \Phi = -\frac{p/A}{\epsilon_0}.$0 an effective interfacial dielectric constant (Xiang et al., 26 Nov 2025).

In lateral 1T/1H/1T–MoS$\Delta \Phi = -\frac{p/A}{\epsilon_0}.$1 homojunctions, the same electrostatic language is recast as a built-in drop across a tunnel barrier. Using identical 1T–MoS$\Delta \Phi = -\frac{p/A}{\epsilon_0}.$2 electrodes with work function $\Delta \Phi = -\frac{p/A}{\epsilon_0}.$3 and a 1H–MoS$\Delta \Phi = -\frac{p/A}{\epsilon_0}.$4 barrier of electron affinity $\Delta \Phi = -\frac{p/A}{\epsilon_0}.$5, the zero-bias conduction-band edge is

$\Delta \Phi = -\frac{p/A}{\epsilon_0}.$6

which yields interface barrier heights

$\Delta \Phi = -\frac{p/A}{\epsilon_0}.$7

The resulting zero-bias barrier is trapezoidal rather than rectangular when asymmetric interface dipoles are present (Eckmann et al., 22 Sep 2025).

2. Microscopic origins

A recurring microscopic origin is local interfacial charge transfer. In lateral Gr/BN/Gr and 1T/1H/1T–MoS$\Delta \Phi = -\frac{p/A}{\epsilon_0}.$8 junctions, electron difference density and electrostatic difference potential show that asymmetric terminations produce equal-magnitude, opposite-sign dipole steps at the two interfaces, giving a net built-in potential drop across the barrier. Symmetric armchair terminations preserve inversion symmetry, yield no built-in field, and show no rectification (Eckmann et al., 22 Sep 2025).

Localized defects can generate the same effect. At the SrRuO$\Delta \Phi = -\frac{p/A}{\epsilon_0}.$9/La$\Delta \Phi = -\frac{\mu_\perp n}{\epsilon_0},$0Sr$\Delta \Phi = -\frac{\mu_\perp n}{\epsilon_0},$1MnO$\Delta \Phi = -\frac{\mu_\perp n}{\epsilon_0},$2 interface, cation displacements indicate a dipole-like electric field even though both materials are nominally metallic. Density-functional calculations reproduce the observed displacement profile only when oxygen vacancies are present in the near-interface LSMO layers. Two vacancies in near-interface LSMO give a displacement difference between extremes of $\Delta \Phi = -\frac{\mu_\perp n}{\epsilon_0},$3 Å, matching the experimental $\Delta \Phi = -\frac{\mu_\perp n}{\epsilon_0},$4 Å, and generate a dipolar potential difference of order $\Delta \Phi = -\frac{\mu_\perp n}{\epsilon_0},$5–$\Delta \Phi = -\frac{\mu_\perp n}{\epsilon_0},$6 V (Borisevich et al., 2011).

Ground-state integer charge transfer is not a necessary condition. At the CuPc/C$\Delta \Phi = -\frac{\mu_\perp n}{\epsilon_0},$7 donor–acceptor interface, Bader analysis shows a total interfacial charge transfer of $\Delta \Phi = -\frac{\mu_\perp n}{\epsilon_0},$8 e for the face-on interface, far too small to account for the observed dipole barrier. The interface dipole is instead attributed to anisotropic polarization associated with the molecular quadrupole of CuPc, with the face-on geometry yielding $\Delta \Phi = -\frac{\mu_\perp n}{\epsilon_0},$9 D per C$\mu_\perp$0 and a scaled vacuum-level shift of $\mu_\perp$1 eV, whereas the edge-on geometry gives $\mu_\perp$2 D and $\mu_\perp$3 eV (Sai et al., 2012).

At metal–semiconductor contacts, the unified bond dipole theory places the origin in localized bonding between semiconductor surface dangling bonds and metal orbitals. Within the two-level Hamiltonian

$\mu_\perp$4

the bond polarity is

$\mu_\perp$5

and metal-induced gap states, dangling-bond-induced surface states, and bonding states embedded in the valence band are treated as different outcomes of the same underlying interface bonding mechanism (Xiang et al., 26 Nov 2025).

A separate class of dipoles arises from surface electrochemistry and interfacial water. In graphene field-effect transistors on hydrophilic SiO$\mu_\perp$6, silanol groups and an interfacial water layer create a dipole layer oriented with negative charge at SiO$\mu_\perp$7 and positive charge in graphene, producing hole doping. On hydrophobic SiO$\mu_\perp$8, siloxane termination suppresses water adsorption and no significant interfacial dipole forms (Nagamura et al., 2019). This suggests that interface-dipole engineering is not a single mechanism but a family of electrostatic phenomena linked by the same potential-step formalism.

3. Device manifestations across material platforms

In lateral 2D tunnel diodes, interface dipoles can create rectification without dissimilar contacts. First-principles electronic structure and quantum transport calculations show that asymmetric zigzag-type 1T/1H interfaces in monolayer 1T/1H/1T–MoS$\mu_\perp$9 generate a zero-bias trapezoidal barrier, whereas symmetric armchair interfaces yield strictly antisymmetric $n$0–$n$1. For the asymmetric device with $n$2 nm, the rectification ratio at $n$3 V is $n$4, and in the thickness range $n$5–$n$6 nm the asymmetry rises from $n$7 at $n$8 nm to nearly $n$9 at $\Phi_{Bn} = \Phi_M - \chi_s - \Delta V,$0 V for $\Phi_{Bn} = \Phi_M - \chi_s - \Delta V,$1 nm. In the minimal Gr/BN/Gr analogue, rectification reaches $\Phi_{Bn} = \Phi_M - \chi_s - \Delta V,$2 at $\Phi_{Bn} = \Phi_M - \chi_s - \Delta V,$3 V for $\Phi_{Bn} = \Phi_M - \chi_s - \Delta V,$4 nm, confirming that interface-induced dipoles, rather than work-function difference, enable the effect (Eckmann et al., 22 Sep 2025).

In ferroelectric tunnel junctions, asymmetric interfaces can be generated even with symmetric electrodes when one interface dipole is pinned and the other is switchable. With

$\Phi_{Bn} = \Phi_M - \chi_s - \Delta V,$5

the pinned dipole imposes a fixed potential step, and a finite tunneling electroresistance appears because the electrostatic profile differs for opposite ferroelectric polarization states. A large tunneling electroresistance is achieved when the pinned polarization points to the ferroelectric film and the interface dielectric constants are low; under the studied parameters, decreasing $\Phi_{Bn} = \Phi_M - \chi_s - \Delta V,$6 can raise tunneling electroresistance by about three orders of magnitude, whereas decreasing $\Phi_{Bn} = \Phi_M - \chi_s - \Delta V,$7 raises it by about one order of magnitude (Wu, 2014).

In mixed-dimensional p–n heterojunctions, the dipole coexists with semiconductor band bending. For monolayer MoS$\Phi_{Bn} = \Phi_M - \chi_s - \Delta V,$8 on p-GaN(0001), ARPES and HR-XPS give a valence-band offset of $\Phi_{Bn} = \Phi_M - \chi_s - \Delta V,$9 eV, a conduction-band offset of $$\Delta V = \frac{e\,a_p\,\sigma_s\,d}{\epsilon_0\epsilon_r} = c\,a_p,$$0 eV, and an interface dipole magnitude of $$\Delta V = \frac{e\,a_p\,\sigma_s\,d}{\epsilon_0\epsilon_r} = c\,a_p,$$1 eV. The GaN surface downward band bending after MoS$$\Delta V = \frac{e\,a_p\,\sigma_s\,d}{\epsilon_0\epsilon_r} = c\,a_p,$$2 transfer is $$\Delta V = \frac{e\,a_p\,\sigma_s\,d}{\epsilon_0\epsilon_r} = c\,a_p,$$3 eV, and its reduction relative to pristine GaN is $$\Delta V = \frac{e\,a_p\,\sigma_s\,d}{\epsilon_0\epsilon_r} = c\,a_p,$$4 eV (Henck et al., 2018).

By contrast, interface-dipole engineering can also aim at elimination rather than amplification. In homologous MoSi$$\Delta V = \frac{e\,a_p\,\sigma_s\,d}{\epsilon_0\epsilon_r} = c\,a_p,$$5N$$\Delta V = \frac{e\,a_p\,\sigma_s\,d}{\epsilon_0\epsilon_r} = c\,a_p,$$6/MoSi$$\Delta V = \frac{e\,a_p\,\sigma_s\,d}{\epsilon_0\epsilon_r} = c\,a_p,$$7N$$\Delta V = \frac{e\,a_p\,\sigma_s\,d}{\epsilon_0\epsilon_r} = c\,a_p,$$8(MoN)$$\Delta V = \frac{e\,a_p\,\sigma_s\,d}{\epsilon_0\epsilon_r} = c\,a_p,$$9 van der Waals heterostructures, the identical SiN outer sublayers on both sides produce nearly symmetric charge redistribution, Bader net transfer below $\sigma_s$0 e across the interface, and $\sigma_s$1. The resulting “zero-dipole” contact recovers near-ideal Schottky–Mott behavior even in the extreme close-contact regime (Tho et al., 2024).

4. Experimental and computational diagnosis

The characteristic observables of interface dipoles are charge redistribution, potential steps, and level shifts. In first-principles analyses of lateral tunnel junctions, electron difference density

$\sigma_s$2

and the electrostatic difference potential are used to identify depletion and accumulation at opposite interfaces. Transport is then evaluated with DFT–NEGF through the Landauer–Büttiker current,

$\sigma_s$3

The same workflow distinguishes symmetric interfaces, which produce no net dipole, from asymmetric interfaces, which generate a built-in drop and polarity-dependent tunneling (Eckmann et al., 22 Sep 2025).

In oxide heterostructures, aberration-corrected HAADF-STEM provides quantitative position mapping of B-site cation displacements, while EELS establishes termination and compositional sharpness. In SRO/LSMO, the displacement-derived polarization was reconstructed through

$\sigma_s$4

and the electrostatic potential obtained from

$\sigma_s$5

This structural route was more sensitive than O-K edge analysis because large intrinsic O-K differences between the oxides masked subtle vacancy signatures (Borisevich et al., 2011).

At buried organic interfaces, ultraviolet photoemission spectroscopy measures the secondary-electron cutoff and valence offsets, while TOF-SIMS resolves intermixing. In CuPc/C$\sigma_s$6, UPS measured a $\sigma_s$7 eV dipole barrier for C$\sigma_s$8 on CuPc and a barrier equivalent to $\sigma_s$9 eV when referenced CuPc$a_p$0C$a_p$1 for CuPc on C$a_p$2, whereas TOF-SIMS gave mixed-layer thicknesses of $a_p$3 nm and $a_p$4 nm for the two deposition sequences (Sai et al., 2012).

For microwave nanocomposites, high-frequency dielectric spectroscopy captures interface-dominated relaxation. In MWCNT/silicone elastomer, a distinct relaxation peak in the $a_p$5–$a_p$6 GHz window was fit with the Havriliak–Negami function,

$a_p$7

and the evolution of $a_p$8 and peak sharpness was used to infer interface modification, dispersion quality, and cyclic interface reconstruction (Li et al., 2017).

A methodological caution emerges from symbolic regression. High-throughput DFT and symbolic regression recovered the Topping form for free-standing dipolar SAMs with RMSE $a_p$9 meV, but in charge-transfer metal–organic interfaces highly accurate correlations were found that were clearly unphysical. This established that low RMSE alone is insufficient when dimensional analysis and electrostatic scaling are not enforced (Cartus et al., 2021).

5. Principal design levers

The most direct lever is interface structure and chemistry. In lateral 2D tunnel diodes, zigzag versus armchair termination, and which atom bonds at the edge, set the sign and magnitude of the dipole; electronegative bonding such as C–N versus electropositive bonding such as C–B reverses the built-in field. Barrier band gap, complex band, interface sharpness, and thickness then determine how the dipole translates into current: current decays exponentially with $d$0, while rectification ratio $d$1 increases with $d$2 and $d$3 (Eckmann et al., 22 Sep 2025).

Defect chemistry is equally important in oxides. Oxygen partial pressure during growth, post-growth oxygen annealing or vacuum annealing, capping layers that set oxygen chemical potential, and electrochemical or ionic gating all tune oxygen-vacancy concentration and therefore the interfacial dipole. The SRO/LSMO case identified near-interface oxygen vacancies in LSMO as the defect scenario that reproduces both the measured displacement profile and the dipolar potential step (Borisevich et al., 2011).

Morphology and orientation are dominant in molecular interfaces. In CuPc/C$d$4, promoting face-on CuPc relative to C$d$5 increases the dipole from the edge-on value of $d$6 D and $d$7 eV to the face-on value of $d$8 D and $d$9 eV. Intermixing is not incidental in this system; it creates local regions that include face-on CuPc relative to C$\Delta \Phi = -\frac{p/A}{\epsilon_0}.$00 and explains why UPS detects a sizable dipole consistent with the face-on prediction (Sai et al., 2012).

Surface passivation can be used either to create or to suppress dipoles. In GFETs, hydrophobic SiO$\Delta \Phi = -\frac{p/A}{\epsilon_0}.$01 obtained by high-temperature reoxidation suppresses the interfacial water-mediated dipole and minimizes hysteresis, whereas hydrophilic SiO$\Delta \Phi = -\frac{p/A}{\epsilon_0}.$02 prepared by O$\Delta \Phi = -\frac{p/A}{\epsilon_0}.$03 plasma creates a strong interfacial dipole and a built-in hole density of approximately $\Delta \Phi = -\frac{p/A}{\epsilon_0}.$04 (Nagamura et al., 2019).

Energy-space tuning by dipolar overlayers extends the concept beyond barrier formation. In van der Waals bilayers, a dipolar overlayer such as double-layer Ih-ice changes the onsite energy difference $\Delta \Phi = -\frac{p/A}{\epsilon_0}.$05 between neighboring-layer states and thereby the interlayer hybridization admixture ratio

$\Delta \Phi = -\frac{p/A}{\epsilon_0}.$06

In WS$\Delta \Phi = -\frac{p/A}{\epsilon_0}.$07/MoS$\Delta \Phi = -\frac{p/A}{\epsilon_0}.$08, $\Delta \Phi = -\frac{p/A}{\epsilon_0}.$09 is reduced from $\Delta \Phi = -\frac{p/A}{\epsilon_0}.$10 meV to $\Delta \Phi = -\frac{p/A}{\epsilon_0}.$11 meV and $\Delta \Phi = -\frac{p/A}{\epsilon_0}.$12 increases from $\Delta \Phi = -\frac{p/A}{\epsilon_0}.$13 to $\Delta \Phi = -\frac{p/A}{\epsilon_0}.$14; in MoS$\Delta \Phi = -\frac{p/A}{\epsilon_0}.$15/MoS$\Delta \Phi = -\frac{p/A}{\epsilon_0}.$16, $\Delta \Phi = -\frac{p/A}{\epsilon_0}.$17 increases from $\Delta \Phi = -\frac{p/A}{\epsilon_0}.$18 meV to $\Delta \Phi = -\frac{p/A}{\epsilon_0}.$19 meV and $\Delta \Phi = -\frac{p/A}{\epsilon_0}.$20 drops from $\Delta \Phi = -\frac{p/A}{\epsilon_0}.$21 to $\Delta \Phi = -\frac{p/A}{\epsilon_0}.$22 (Shao et al., 2022).

For self-assembled monolayers, the practical rule is the Topping form rediscovered by symbolic regression:

$\Delta \Phi = -\frac{p/A}{\epsilon_0}.$23

Large $\Delta \Phi = -\frac{p/A}{\epsilon_0}.$24, moderate packing density $\Delta \Phi = -\frac{p/A}{\epsilon_0}.$25, and reduced depolarization through smaller $\Delta \Phi = -\frac{p/A}{\epsilon_0}.$26 maximize the attainable work-function shift (Cartus et al., 2021).

6. Applications, controversies, and limits

Interface-dipole engineering has device relevance wherever electrostatic alignment controls transport or optical coupling. The applications explicitly identified in the cited work include ultrathin in-plane diodes and rectifiers, THz/infrared detectors, energy-harvesting tunnel devices and rectennas, pressure sensing through dipole-free tunneling contacts, optimization of donor–acceptor offsets in organic photovoltaics, microwave functionality in CNT/elastomer nanocomposites, and graphene field-effect transistor optimization (Eckmann et al., 22 Sep 2025, Tho et al., 2024, Sai et al., 2012, Li et al., 2017, Nagamura et al., 2019).

Several common simplifications are incorrect. Interface dipoles are not reducible to bulk work-function differences: lateral Gr/BN/Gr and 1T/1H/1T–MoS$\Delta \Phi = -\frac{p/A}{\epsilon_0}.$27 rectify with identical electrodes because local interface-induced dipoles, rather than work-function difference, set the trapezoidal barrier (Eckmann et al., 22 Sep 2025). Nor do they always require integer charge transfer: CuPc/C$\Delta \Phi = -\frac{p/A}{\epsilon_0}.$28 exhibits a sizable interface dipole while Bader analysis gives only $\Delta \Phi = -\frac{p/A}{\epsilon_0}.$29 e net transfer (Sai et al., 2012). Conversely, weak metal-induced gap states do not guarantee zero dipole; conventional van der Waals contacts can still deviate from Schottky–Mott behavior because of push-back and electronegativity differences, whereas homologous MoSi$\Delta \Phi = -\frac{p/A}{\epsilon_0}.$30N$\Delta \Phi = -\frac{p/A}{\epsilon_0}.$31/MoSi$\Delta \Phi = -\frac{p/A}{\epsilon_0}.$32N$\Delta \Phi = -\frac{p/A}{\epsilon_0}.$33(MoN)$\Delta \Phi = -\frac{p/A}{\epsilon_0}.$34 contacts are exceptional in achieving $\Delta \Phi = -\frac{p/A}{\epsilon_0}.$35 (Tho et al., 2024).

The main limits are equally consistent across platforms. Rectification in fully 2D tunnel diodes is highly sensitive to symmetry; symmetric terminations yield zero $\Delta \Phi = -\frac{p/A}{\epsilon_0}.$36 and antisymmetric $\Delta \Phi = -\frac{p/A}{\epsilon_0}.$37–$\Delta \Phi = -\frac{p/A}{\epsilon_0}.$38 (Eckmann et al., 22 Sep 2025). Strong metal-induced gap states can increase current but reduce apparent barrier integrity, as in 1T/1H/1T–MoS$\Delta \Phi = -\frac{p/A}{\epsilon_0}.$39 where the apparent gap reduction is $\Delta \Phi = -\frac{p/A}{\epsilon_0}.$40 (Eckmann et al., 22 Sep 2025). Excess oxygen vacancies can reduce carrier mobility, alter magnetic order, destabilize the perovskite structure, and lead to long-term drift (Borisevich et al., 2011). In SiC MOSFETs, a high density of neutral interfacial dipoles is not useful but harmful: dipole scattering with fitted $\Delta \Phi = -\frac{p/A}{\epsilon_0}.$41 and $\Delta \Phi = -\frac{p/A}{\epsilon_0}.$42 with $\Delta \Phi = -\frac{p/A}{\epsilon_0}.$43 nm reproduces the observed low inversion-layer mobility (Hatakeyama et al., 2022). Data-driven prediction also has a clear limit: highly accurate symbolic-regression expressions for charge-transfer interfaces can be unphysical unless constrained by electrostatics and dimensional analysis (Cartus et al., 2021).

Taken together, the literature defines interface-dipole engineering as a general electrostatic strategy rather than a material-specific trick. It operates by controlling the location, sign, and screening of interfacial charge redistribution—through termination, vacancies, molecular orientation, dipolar overlayers, dielectric environment, or homologous contact design—to impose, suppress, or reshape the potential step that governs band alignment and transport.