Specular Andreev Reflection in Graphene

• Specular Andreev reflection is an electron–hole conversion process where an incident electron is reflected as a hole with the same forward momentum, unlike retroreflection.
• The phenomenon hinges on band topology and energy criteria, with true specular reflection emerging when the Fermi energy is below the superconducting gap, |E_F| < Δ.
• Experimental setups in graphene–superconductor devices use conductance suppressions and oscillation signatures to distinguish specular reflection from its retro counterpart.

Specular Andreev reflection is an electron–hole conversion process at a normal–superconductor interface in which an incident electron is reflected as a hole that does not retrace the incoming trajectory. In the graphene setting that made the phenomenon canonical, the process occurs when an electron excited above the Fermi level in the conduction band is converted into a hole below the Fermi level in the valence band; the hole carries opposite charge, retains the same forward momentum component parallel to the interface, and is reflected “specularly” rather than retro. The distinction from ordinary Andreev retroreflection is tied to band topology, the relation between group velocity and wave vector, and, in many materials, to whether the reflected hole belongs to the same band as the incident electron or to the opposite band (Wang et al., 2017).

1. Physical definition and kinematic distinction

In a conventional Andreev reflection, an electron with energy $E$ above the Fermi level $E_F$ in the normal region is reflected at the superconductor interface as a hole of almost the same momentum but opposite velocity. Both electron and hole live in the same band, so they have the same sign of effective mass and the hole retraces the incoming path. This is the retroreflection regime (Wang et al., 2017).

In graphene and related Dirac materials, the gapless spectrum allows a distinct process when the Fermi energy is tuned close to the Dirac point. An electron in the conduction band can be converted into a hole in the valence band. The reflected hole then has reversed group velocity but the same forward momentum component parallel to the interface, so the scattering is specular rather than retro. The conduction and valence bands have opposite curvature of dispersion, hence opposite sign of effective mass; this sign change underpins the reversal of the transverse velocity in graphene-based descriptions (Wang et al., 2017).

A complementary statement, used across several material classes, is that the essential distinction between retro-Andreev reflection and specular-Andreev reflection is the relative sign of wave vector $k$ and group velocity $v$ for the reflected hole compared to those of the incident electron. In retroreflection the incident electron has $v\!\cdot\!k>0$ but the Andreev-reflected hole has $v\!\cdot\!k<0$; in specular reflection the hole satisfies $v\!\cdot\!k>0$, just like the incident electron (Cheng et al., 2021). In graphene rings this same dichotomy is expressed as $v_y^{(e)}\!\cdot\!v_y^{(h)}<0$ for retroconfiguration and $v_y^{(e)}\!\cdot\!v_y^{(h)}>0$ for specular configuration (Schelter et al., 2011).

The operational energy criterion is material-dependent in detail but takes a particularly simple form in Dirac systems. In graphene-based junctions, retroreflection occurs for $E<|E_F|$, while specular reflection occurs for $E_F$0 provided the excitation remains subgap, $E_F$1 (Wang et al., 2017). In single-layer graphene, when $E_F$2, Andreev reflection is an intraband process and remains retro. As one tunes $E_F$3 toward the Dirac point, the radius of the constant-energy contour for the reflected hole shrinks, and beyond a critical incidence angle the reflected hole state no longer exists; the process becomes non-retro, or the onset of specular Andreev reflection. True interband specular reflection requires $E_F$4 (Sahu et al., 2016).

2. Theoretical formulations

Two closely related theoretical descriptions recur in the literature: Bogoliubov–de Gennes equations adapted to Dirac or tight-binding normal-state Hamiltonians, and scattering formulations of the Blonder–Tinkham–Klapwijk or Landauer–Büttiker type.

For graphene in a nearest-neighbor tight-binding description with a single $E_F$5 orbital per site, the normal-state Hamiltonian is written as

$E_F$6

with $E_F$7 and gate-tunable $E_F$8. Introducing the Nambu spinor $E_F$9, the interface to a superconductor is described by the BdG equation

$k$0

Here $k$1 is the superconducting pair potential, with $k$2 in typical Nb or Al hybrids (Wang et al., 2017).

In single-layer graphene, the continuum Dirac–BdG form uses

$k$3

and the BdG matrix equation

$k$4

with continuity of the four-component Dirac–BdG spinor at the interface and $k$5 on the superconducting side (Sahu et al., 2016).

The non-equilibrium Green-function formulation expresses the retarded Green function in Nambu space for the central region as

$k$6

with self-energies from graphene leads and the superconductor. The zero-temperature Andreev reflection probability is then

$k$7

This formulation is used directly in finite-size Y-shaped graphene–superconductor devices and in graphene junctions with magnetic barriers (Wang et al., 2017).

For transport observables, a standard zero-temperature differential-conductance expression is

$k$8

where $k$9 and $v$0 are normal and Andreev reflection amplitudes between modes in the injecting lead (Wang et al., 2017). In single-layer graphene the BTK expression can be written as

$v$1

with the critical angle

$v$2

For $v$3, $v$4; as $v$5, $v$6, which directly encodes the angular blocking of Andreev reflection near the Dirac point (Sahu et al., 2016).

3. Graphene as the canonical platform

Graphene provides the best-developed setting for specular Andreev reflection because the Fermi energy can be tuned close to the Dirac point. In this regime, ordinary intraband retroreflection and interband specular reflection can be switched by bias and gating, with the boundary between the two regimes at $v$7 in the subgap window (Wang et al., 2017).

Single-layer graphene experiments at graphene–NbSe$v$8 van der Waals interfaces observed that the normalized conductance becomes suppressed as the Dirac cone is crossed by tuning Fermi level and bias energy. The suppression indicates blockage of Andreev reflection beyond a critical angle of the incident electron with respect to the normal. The measured feature is therefore a transition from retro to non-retro type Andreev reflection, rather than a demonstration that full interband specular reflection has already been reached. In the reported devices, $v$9 while $v\!\cdot\!k>0$0, so full specular reflection was not reached; the stated condition for true specular reflection is $v\!\cdot\!k>0$1 together with disorder broadening $v\!\cdot\!k>0$2 (Sahu et al., 2016).

Bilayer graphene is emphasized as an experimentally favorable platform because the Fermi energy broadening near the Dirac point is much weaker than in monolayer graphene. The theoretical crossover from intraband retroreflection to interband specular reflection yields a characteristic suppression of interfacial conductance when the excitation energy satisfies $v\!\cdot\!k>0$3. The sharpness of these conductance dips depends strongly on the potential step $v\!\cdot\!k>0$4 at the bilayer-graphene–superconductor interface (Efetov et al., 2016). A related theoretical analysis with a Zeeman field on the normal side predicts a central bias–chemical-potential “diamond” in which only specular reflection is allowed for one spin species even at zero bias, together with enhanced zero-bias conductance at $v\!\cdot\!k>0$5 not present when the Zeeman field vanishes (Soori et al., 2018).

Finite-size and edge effects are not ancillary details in graphene devices. In finite flakes, boundary scattering and intervalley mixing can blur the clean separation between retro and specular reflection. Zigzag edges support edge-localized zero modes and suffer weaker intervalley coupling than armchair edges, making them preferable for studying specular Andreev reflection with minimal mixing (Wang et al., 2017). In a Y-shaped graphene–superconductor device, zigzag terminals are identified as the better choice for detecting Andreev reflection without external field, while transport conductance alone does not permit a quantitative distinction between retroreflection and specular reflection in finite-size devices unless additional structure is imposed (Wang et al., 2017).

A further refinement arises when the charge-neutrality point varies spatially near a partially covered bilayer-graphene interface. In that case a smooth $v\!\cdot\!k>0$6 junction can form in the uncovered region, and when the Fermi level is close to the charge-neutrality point away from the interface, Andreev reflection takes place in a specular manner owing to diffraction of a reflected hole at that $v\!\cdot\!k>0$7 junction. The differential conductance then becomes asymmetric in bias, and when the Fermi level lies below the charge-neutrality point, quasi-bound states near the interface produce resonant peaks in the conductance (Takane et al., 2017).

4. Experimental signatures and detection strategies

A recurring theme in the literature is that conductance suppression near charge neutrality is an important but not exclusive signature. Several more discriminating probes have been proposed and analyzed.

In a graphene ring with one normal and one superconducting lead, the dominant Aharonov–Bohm oscillation period in subgap transport distinguishes the two Andreev configurations. In retroconfiguration the dominant oscillation period is $v\!\cdot\!k>0$8, whereas in specular configuration it is $v\!\cdot\!k>0$9. The interpretation is that in retroreflection the electron–hole pair carries net charge $v\!\cdot\!k<0$0 around the loop, while in specular reflection the hole exits through the opposite arm and interfering loops carry net charge $v\!\cdot\!k<0$1 (Schelter et al., 2011). The same study reports robustness against disorder and moderate changes to the system, making the oscillation period a clear signature for distinguishing the two processes (Schelter et al., 2011).

A second strategy is directional separation by a perpendicular magnetic field in a Y-shaped graphene device. A perpendicular field enforces cyclotron motion and chiral edge states so that incident electrons propagate along one edge toward the superconducting interface. Because retroreflected holes and specularly reflected holes have opposite effective masses, their cyclotron bending is opposite, and in a Y-shaped three-terminal flake the reflected holes are routed into different graphene arms. In this setting the external terminal receiving the Andreev-reflected hole depends entirely on the kind of Andreev reflection, and specular reflection can be clearly distinguished from retroreflection even for finite-size devices (Wang et al., 2017).

A third, more material-independent proposal is based on quasiclassical quantization in a semimetal–superconductor double junction. There the intrinsic character of the specularly reflected hole, having the same sign relation of wave vector and group velocity as the incident electron, produces periodic oscillation of conductance with the length of the middle semimetal. This oscillation is absent for retroreflection, for which the hole has the opposite sign relation and the length-dependent phase cancels at low energy. For normal incidence and zero bias, the predicted conductance peaks occur at

$v\!\cdot\!k<0$2

for $v\!\cdot\!k<0$3-wave or $v\!\cdot\!k<0$4 pairing, and at

$v\!\cdot\!k<0$5

for $v\!\cdot\!k<0$6 or chiral $v\!\cdot\!k<0$7-wave pairing, with universal spacing

$v\!\cdot\!k<0$8

The paper presents this as a universal detection method irrespective of the details of the materials (Cheng et al., 2021).

Interference can also encode a symmetry property of the reflection amplitude itself. In zigzag graphene ribbons, the parity of the Andreev reflection amplitude under mirror operation is even for Andreev retroreflection and odd for specular Andreev reflection. In four-terminal graphene–superconductor interferometers this odd parity can produce destructive interference of specular Andreev reflection even when the system is asymmetric, whereas the interference condition for retroreflection is the opposite (Xing et al., 2010). This does not merely offer another detection scheme; it identifies a symmetry distinction between intraband and interband Andreev processes in graphene.

5. Material generalizations beyond graphene

Although graphene supplied the original and most developed context, specular Andreev reflection is not confined to Dirac cones in monolayers. The literature extends it to bilayer graphene, topological-insulator thin films, Weyl semimetals, HgTe quantum wells, InAs/GaSb double quantum wells, pseudospin-1 fermion systems, Ising-superconductor junctions, and altermagnet–superconductor interfaces.

In inversion-symmetric Weyl semimetals, specular Andreev reflection is predicted with two unusual features. For $v\!\cdot\!k<0$9-wave BCS pairing, the Andreev conductance is anisotropic and depends on the angle between the line connecting a pair of Weyl points and the normal of the junction, with an effective gap $v\!\cdot\!k>0$0. For FFLO pairing states, the Andreev reflection spectrum is isotropic and independent of the finite momentum of the Cooper pairs (Chen et al., 2013).

In thin films of topological insulators subject to a gate electric field, specular Andreev reflection probability increases with electric field, and perfect specular Andreev reflection can occur for all angles of incidence at $v\!\cdot\!k>0$1. The same work associates the regime with intraband specular Andreev reflection and reports an exponential temperature dependence of thermal conductance (Majidi et al., 2015).

In doped HgTe/CdTe quantum wells, both specular Andreev reflection and retroreflection can be realized even in the presence of a large mismatch between the Fermi wavelengths on the two sides of the junction. There the Andreev conductance probes the dynamics of massive $v\!\cdot\!k>0$2 Dirac fermions, and the transition between retro, a forbidden intermediate window, and specular reflection is controlled by the relation among excitation energy, chemical potential, and the Dirac mass term (Guigou et al., 2010).

In the edge transport of an inverted InAs/GaSb double quantum well, a finite low-bias anomaly with full width $v\!\cdot\!k>0$3 is interpreted as a crossover from retroreflection at low bias to specular reflection at higher subgap bias. The threshold $v\!\cdot\!k>0$4 is tied to the inversion gap, and the anomaly persists across different interface-transparency regimes, although its sign in differential resistance depends on the barrier strength (Kononov et al., 2016).

For massive pseudospin-1 Dirac fermions, the distinction between retro and specular reflection survives but is modified by the mass term. In $v\!\cdot\!k>0$5-type systems, Andreev reflection probability at oblique incidence can exceed that at normal incidence, and the enhancement occurs in different doping configurations for retro and specular reflection. In the $v\!\cdot\!k>0$6-type case, an all-angle unit-efficiency Andreev reflection is predicted at $v\!\cdot\!k>0$7 and $v\!\cdot\!k>0$8 (Zeng et al., 2021).

A more recent extension concerns graphene coupled to an Ising superconductor with mirage gaps induced by an exchange field. There the chemical-potential range in graphene supporting specular-Andreev reflection lies between the two mirage gaps and is about twice the Ising spin-orbit coupling strength, which is stated to enhance resilience against graphene potential fluctuations. The same setting supports Andreev interference modulated not only by superconducting phase difference and chemical potential, but also by the relative orientation of exchange fields through the finite contribution of spin-triplet pairing (Li et al., 2 May 2025).

Altermagnets provide a different mechanism. In a single-band model with spin-split open Fermi surfaces, specular Andreev reflection occurs stably at altermagnet–superconductor interfaces without the fine chemical-potential tuning associated with Dirac or Weyl materials. The outgoing hole current is fully spin-polarized, and positive nonlocal conductance together with positive noise cross-correlation are identified as unambiguous signatures in a multiterminal device (Nagae et al., 2024). A microscopically motivated six-orbital description of V$v\!\cdot\!k>0$9O-based altermagnets reaches a similar conclusion and proposes a multiterminal geometry in which positive nonlocal conductance detects spin-polarized specular reflection (Nagae et al., 14 Apr 2026).

6. Constraints, ambiguities, and broader implications

Several recurring caveats shape the interpretation of experiments and model calculations. The first is that non-retro behavior, conductance suppression near the Dirac point, and true interband specular reflection are not identical statements. In single-layer graphene, the observed conductance suppression near charge neutrality was interpreted as a transition from retro to non-retro Andreev reflection, while the same analysis explicitly states that true specular reflection requires $v_y^{(e)}\!\cdot\!v_y^{(h)}<0$0 and disorder broadening smaller than $v_y^{(e)}\!\cdot\!v_y^{(h)}<0$1 (Sahu et al., 2016). This distinction has become central in evaluating experimental claims.

The second caveat is geometric and mesoscopic: finite-size scattering, edge termination, interface barriers, and smooth electrostatic profiles can either obscure or reshape the retro-to-specular crossover. In finite graphene flakes, intervalley mixing and interference can make quantitative distinction through conductance alone impossible; magnetic steering or interferometric observables then become decisive (Wang et al., 2017). In bilayer graphene, the interface potential step $v_y^{(e)}\!\cdot\!v_y^{(h)}<0$2 strongly controls the sharpness of conductance dips at the crossover (Efetov et al., 2016).

A third issue is that specular Andreev reflection can encode additional quantum numbers beyond charge conversion. In zigzag graphene ribbons it carries an odd mirror parity under reflection symmetry (Xing et al., 2010); in altermagnets it can be spin-polarized (Nagae et al., 2024); in Ising-superconductor junctions its accessible chemical-potential window is modified by mirage gaps and spin-orbit coupling (Li et al., 2 May 2025). This suggests that specular reflection is best regarded not as a single graphene-specific curiosity, but as a broader interband Andreev phenomenon whose concrete manifestation depends on dispersion, symmetry, and Fermi-surface geometry.

Finally, some recent work links specular Andreev reflection to emergent edge physics. On the surface of a three-dimensional topological insulator, when the Fermi level of massless surface electrons is at the Dirac point, specular Andreev reflection is identified as the origin of a helical Majorana edge mode at a simple normal–superconductor boundary, without a magnetic insulator. The resulting mode lacks the topological protection of a chiral counterpart, but the proposal presents specular reflection as the microscopic mechanism binding the charge-neutral edge excitation (Beenakker, 2024). This suggests a broader implication: specular Andreev reflection is not only a scattering signature, but can also reorganize low-energy boundary states in hybrid superconducting structures.