Retro Andreev Reflection Fundamentals

• Retro Andreev reflection is defined as the electron-to-hole conversion at NS interfaces, where the reflected hole follows the incident electron’s reversed trajectory.
• The BTK framework and diffusive scattering analysis show that interface transparency crucially controls the efficiency of Andreev conversion.
• Materials like graphene, HgTe quantum wells, and Weyl semimetals use retroreflection as a benchmark to probe band topology and unconventional superconductivity.

Searching arXiv for relevant papers on retro Andreev reflection. Searching arXiv for core references and related developments. Retro Andreev reflection is the conventional subgap electron–hole conversion process at a normal-metal–superconductor interface in which an incident electron is reflected as a hole that propagates back along nearly the same trajectory, while a Cooper pair is transferred into the superconducting condensate. In the standard metal/superconductor picture, energy and momentum conservation make the hole retrace the incoming path “along almost the same path taken by the incident electron,” and this time-reversed character is the defining geometric feature of the process. Across the literature, retro Andreev reflection is the reference limit against which specular, nonlocal, double-channel, and anomalous-trajectory Andreev processes are defined in graphene, bilayer graphene, HgTe quantum wells, semi-Dirac systems, Weyl and nodal-line semimetals, and related proximitized structures (Gloos et al., 2014, Bhandari et al., 2020).

1. Conventional definition and kinematic content

At a normal–superconductor interface, a quasiparticle with energy below the superconducting gap cannot enter the superconducting side as an isolated electron. Instead, it is converted into a Cooper-pair transfer process accompanied by reflection of a hole into the normal region. In the retro limit, the reflected hole is the time-reversed partner of the incident electron: its group velocity is reversed so that the hole returns approximately along the incoming path. In conventional three-dimensional metals this is the case in which all velocity components reverse, whereas specular reflection reverses only the component normal to the interface (Luo et al., 2020).

For a straight interface, conservation of momentum parallel to the boundary imposes

$p_e \sin \theta_{inc} = p_h \sin \theta_{ref}.$

In ordinary metals, where the Fermi energy is much larger than the superconducting gap, the electron and the Andreev-reflected hole lie in the same band sector, so the reflected trajectory is retro-like, with $\theta_{ref} \approx -\theta_{inc}$ (Efetov et al., 2015). The same intraband logic underlies the heavily doped limit of doped HgTe/CdTe quantum wells, where $|C_N| \gg \Delta_0$ and the reflected hole remains in the conduction band; the paper explicitly identifies this as standard retroreflection (Guigou et al., 2010).

The contrast with specular Andreev reflection is band-structural rather than merely geometric. In graphene and related Dirac materials, the reflected hole can instead occupy the opposite band when the Fermi level approaches charge neutrality. Then the hole no longer retraces the incoming path, and the reflection becomes mirror-like rather than retrograde (Efetov et al., 2015). In low-doped HgTe/CdTe, the separation between the retro, forbidden, and specular regimes is explicitly bias dependent: $0<eV<|C_N|-|M| \quad \text{retro},$

$|C_N|-|M|<eV<|C_N|+|M| \quad \text{suppressed},$

$eV>|C_N|+|M| \quad \text{specular}.$

That structure makes clear that retroreflection is the intraband branch of a broader Andreev-scattering problem rather than a universal property of all normal–superconductor interfaces (Guigou et al., 2010).

2. BTK description and the role of normal reflection

In conventional Andreev-reflection spectroscopy, retroreflection is described within the Blonder–Tinkham–Klapwijk framework through the superconducting gap $\Delta$, the Dynes broadening $\Gamma$, and especially the normal-reflection parameter $Z$. The interface transmission is

$\tau = \frac{1}{1+Z^2}.$

A small $\theta_{ref} \approx -\theta_{inc}$0 corresponds to a transparent interface and efficient retro Andreev reflection; a large $\theta_{ref} \approx -\theta_{inc}$1 corresponds to stronger normal reflection and weaker Andreev conversion (Gloos et al., 2014).

The hallmark spectroscopic consequence is the familiar double-minimum structure in differential-resistance spectra. In the interpretation given for elemental superconductor/normal-metal point contacts, this structure arises because retro Andreev reflection competes with ordinary electron reflection at the same interface. Increasing $\theta_{ref} \approx -\theta_{inc}$2 suppresses the Andreev signal and reduces interface transparency. The observed contacts yielded $\theta_{ref} \approx -\theta_{inc}$3–$\theta_{ref} \approx -\theta_{inc}$4 across many material pairs, a remarkably narrow interval for nominally different interfaces (Gloos et al., 2014).

Three mechanisms for this normal reflection were analyzed in detail. A dielectric tunneling barrier is modeled by

$\theta_{ref} \approx -\theta_{inc}$5

but this mechanism would be expected to fluctuate strongly from contact to contact, contrary to the observed clustering of $\theta_{ref} \approx -\theta_{inc}$6. Fermi-surface or Fermi-velocity mismatch was also examined through BTK-type extensions, yet the inferred velocity ratios from different metal pairings were not mutually self-consistent. The same work therefore concluded that Fermi-surface mismatch is not the dominant mechanism and may even be completely absent in the Andreev-reflection process as reflected in spectral line shape (Gloos et al., 2014).

The remaining explanation is diffusive elastic scattering in or near the contact region. An ideal long diffusive junction gives an effective

$\theta_{ref} \approx -\theta_{inc}$7

which is close to the experimentally observed $\theta_{ref} \approx -\theta_{inc}$8–$\theta_{ref} \approx -\theta_{inc}$9. On that basis, diffusive transport was identified as the most probable origin of the normal-reflection background and hence of the moderate retroreflection efficiency of typical metal/superconductor point contacts. A major consequence is methodological: Andreev-reflection spectroscopy is not a reliable tool for extracting relative Fermi velocities of the electrodes, because the spectral background is controlled mainly by interface scattering rather than simple Fermi-surface mismatch (Gloos et al., 2014).

3. Retro–specular crossover in graphene and bilayer graphene

Graphene and bilayer graphene supply the canonical setting in which retro Andreev reflection is continuously tunable into specular Andreev reflection. In bilayer graphene/superconductor theory, retroreflection is the intraband regime

$|C_N| \gg \Delta_0$0

while specular reflection is the interband regime

$|C_N| \gg \Delta_0$1

The crossover occurs at

$|C_N| \gg \Delta_0$2

where the critical angle collapses to zero and Andreev reflection is completely suppressed (Efetov et al., 2016).

This suppression has a direct transport signature. In bilayer graphene/NbSe$|C_N| \gg \Delta_0$3 van der Waals interfaces, gate tuning toward the charge neutrality point produced a characteristic conductance suppression when the Fermi energy became smaller than the superconducting gap, identified as a hallmark of the transition between intraband retro- and interband specular Andreev reflection. In that formulation, retroreflection is the regime $|C_N| \gg \Delta_0$4, while specular reflection is the regime $|C_N| \gg \Delta_0$5; the conductance minimum tracks the line $|C_N| \gg \Delta_0$6 because the critical angle vanishes there (Efetov et al., 2015).

A related bilayer-graphene proposal applied a Zeeman field only on the normal side of an NM–SC junction. In the zero-field case, the criterion is

$|C_N| \gg \Delta_0$7

$|C_N| \gg \Delta_0$8

With finite Zeeman splitting, the $|C_N| \gg \Delta_0$9 plane develops a diamond-shaped specular region and additional triangular specular regions, while Andreev reflection is retro outside those domains (Soori et al., 2018).

Gapped bilayer graphene adds another control parameter: opposite displacement fields $0<eV<|C_N|-|M| \quad \text{retro},$0 on the two layers. In the small-$0<eV<|C_N|-|M| \quad \text{retro},$1 regime, the standard condition remains $0<eV<|C_N|-|M| \quad \text{retro},$2 for retroreflection and $0<eV<|C_N|-|M| \quad \text{retro},$3 for specular reflection, but the single crossover point broadens into the forbidden window

$0<eV<|C_N|-|M| \quad \text{retro},$4

within which the relevant hole bands are absent. For $0<eV<|C_N|-|M| \quad \text{retro},$5, the Mexican-hat dispersion generates additional scattering branches, and the junction can simultaneously support specular normal reflection, retro normal reflection, specular Andreev reflection, and retro Andreev reflection (Ram et al., 2023).

4. Material-specific generalizations and anisotropic retroreflection

Beyond graphene, retro Andreev reflection has been used as a probe of nontrivial band topology, anisotropy, pseudospin texture, and multiband kinematics. In HgTe/CdTe quantum wells described by the BHZ massive Dirac Hamiltonian, heavy doping yields conventional retroreflection, while low doping reveals a bias-controlled transition among intraband retroreflection, a gap window with no Andreev reflection, and interband specular reflection. The same work emphasizes that HgTe/CdTe interpolates between graphene-like and ordinary-metal-like behavior because linear Dirac and quadratic Schrödinger terms coexist, so Andreev spectroscopy probes the dynamics of massive $0<eV<|C_N|-|M| \quad \text{retro},$6-dimensional Dirac fermions (Guigou et al., 2010).

In semi-Dirac materials, the outcome depends strongly on transport orientation. Along the linear-dispersion direction, the differential conductance exhibits a clear crossover from retro to specular Andreev reflection with increasing bias, and the conductance oscillates without a decaying profile when the interfacial barrier strength increases. Along the quadratic-dispersion direction, the boundary between retro and specular reflection is ambiguous, and the conductance decays with increasing momentum mismatch or barrier strength. The microscopic explanation is given in terms of anisotropic pseudo-spin textures (Li et al., 2021).

Other pseudospin systems show different retroreflection enhancements. In the $0<eV<|C_N|-|M| \quad \text{retro},$7-$0<eV<|C_N|-|M| \quad \text{retro},$8 lattice, the retro-reflection probability increases with the interpolation parameter $0<eV<|C_N|-|M| \quad \text{retro},$9, and when $|C_N|-|M|<eV<|C_N|+|M| \quad \text{suppressed},$0, the Andreev reflection approaches approximate all-angle perfect transmission for any $|C_N|-|M|<eV<|C_N|+|M| \quad \text{suppressed},$1 in the retro regime (Zhou, 2021). For massive pseudospin-1 fermions in the $|C_N|-|M|<eV<|C_N|+|M| \quad \text{suppressed},$2 lattice, the retro-reflection probability can be larger at oblique incidence than at normal incidence in the $|C_N|-|M|<eV<|C_N|+|M| \quad \text{suppressed},$3-doped $|C_N|-|M|<eV<|C_N|+|M| \quad \text{suppressed},$4-type and $|C_N|-|M|<eV<|C_N|+|M| \quad \text{suppressed},$5-doped $|C_N|-|M|<eV<|C_N|+|M| \quad \text{suppressed},$6-type cases, while the $|C_N|-|M|<eV<|C_N|+|M| \quad \text{suppressed},$7-type mass term supports all-angle unit Andreev efficiency at $|C_N|-|M|<eV<|C_N|+|M| \quad \text{suppressed},$8 and $|C_N|-|M|<eV<|C_N|+|M| \quad \text{suppressed},$9 (Zeng et al., 2021).

In overtilted semimetals, retroreflection coexists with additional Andreev channels. Type-II Weyl semimetal–superconductor junctions can support one retro and one specular Andreev reflection for a single incident electron because the hyperboloidal Fermi surface provides two distinct hole solutions at fixed conserved transverse momentum; changing the interface orientation can transform the channel structure through double retroreflection and eventually into retroreflection plus normal reflection (Hou et al., 2017). Nodal-line semimetal junctions display an even richer “quadruple reflection” structure—specular and retro normal reflection together with specular and retro Andreev reflection—when the nodal line is perpendicular to the interface (Cheng et al., 2020). A torus-shaped Fermi surface in doped nodal-line semimetals further permits anomalous-trajectory Andreev reflection, distinct from both retro and specular reflection; in that framework, retroreflection dominates in the heavy-doping regime (Luo et al., 2020).

Retroreflection also appears in nonlocal form. In topological bilayer exciton-condensate junctions built from top and bottom surfaces of a three-dimensional topological insulator, electrical gating can tune the Andreev process from purely retro to purely specular. Opposite carrier polarity on the two electrodes produces purely retro nonlocal Andreev reflection, whereas the same carrier polarity produces purely specular reflection (Veldhorst et al., 2014).

5. Imaging, interference, and noise diagnostics

Retro Andreev reflection is no longer inferred only from line-shape analysis; it can be imaged and phase-resolved. In ballistic graphene, a liquid-helium-cooled scanning gate microscope operating in the magnetic-focusing regime directly visualized Andreev-reflected holes. The measured transresistance,

$eV>|C_N|+|M| \quad \text{specular}.$0

changes sign according to the charge of the arriving carrier: negative signals correspond to electrons, positive signals to holes. In the superconducting focusing configuration, the second focusing peak reverses sign, indicating that electrons entering the superconducting contact are reflected as holes that continue along the same cyclotron orbit geometry. Destroying Andreev reflection by large current injection or heating above $eV>|C_N|+|M| \quad \text{specular}.$1 flips the reflected carriers back from holes to electrons (Bhandari et al., 2020).

Interferometry provides a complementary distinction between retro and specular geometries. In a graphene ring with one normal and one superconducting lead, the dominant Aharonov–Bohm oscillation period is $eV>|C_N|+|M| \quad \text{specular}.$2 in retroconfiguration and $eV>|C_N|+|M| \quad \text{specular}.$3 in specular configuration. The interpretation is path based: in retroconfiguration the hole follows the same arm as the electron, whereas in specular configuration the hole traverses the opposite arm. The resulting oscillation period is robust against disorder and moderate system changes (Schelter et al., 2011).

In semiconductor–superconductor hybrids, Aharonov–Bohm interferometry was proposed as a direct probe of the spatial reach of retroreflection into a proximitized region. The key scale is the quasiparticle coherence length,

$eV>|C_N|+|M| \quad \text{specular}.$4

which enters the effective electron–hole path length as

$eV>|C_N|+|M| \quad \text{specular}.$5

Below the superconducting gap, the extra phase accumulated in the proximitized segment shortens the Aharonov–Bohm oscillation period, allowing $eV>|C_N|+|M| \quad \text{specular}.$6 to be extracted from magnetoconductance (Nowak et al., 2018).

Finite graphene devices require additional care because conductance alone may not cleanly separate retro from specular reflection. In a Y-shaped graphene–superconductor device, boundary scattering in the finite central region makes the two processes difficult to distinguish in zero field, but a sufficiently strong perpendicular magnetic field routes retroreflected and specularly reflected holes into different graphene terminals. In the strong-field regime, nonzero $eV>|C_N|+|M| \quad \text{specular}.$7 indicates retroreflection, while nonzero $eV>|C_N|+|M| \quad \text{specular}.$8 indicates specular reflection (Wang et al., 2017). A more recent proposal uses shot noise as the discriminator: in graphene-based superconducting heterojunctions, retro Andreev reflection suppresses shot noise, while specular Andreev reflection enhances it, so the Fano factor becomes a direct fingerprint of the reflection type (Salim et al., 30 Aug 2025).

6. Conceptual status and recurrent misconceptions

A persistent misconception is that Andreev reflection is generically retro. The modern literature treats retroreflection as the conventional intraband limit, but not as the only possible subgap electron–hole conversion. Graphene, bilayer graphene, HgTe, semi-Dirac materials, type-II Weyl semimetals, nodal-line semimetals, toroidal Fermi surfaces, and nonlocal exciton-condensate interfaces all admit systematic departures from the retro limit, including specular, double-channel, anomalous-trajectory, and nonlocal variants (Efetov et al., 2015, Guigou et al., 2010, Hou et al., 2017, Luo et al., 2020, Veldhorst et al., 2014).

A second misconception is that standard conductance or differential-resistance spectroscopy always determines the microscopic origin of interface reflection. For conventional metal interfaces, the detailed BTK analysis of elemental contacts argues that the observed normal-reflection background is controlled mainly by diffusive scattering near the contact, with at most a minor contribution from a dielectric barrier, and that Fermi-surface mismatch is not the dominant mechanism and may be completely absent in the Andreev-reflection process as reflected in spectral shape. Real interfaces therefore usually exhibit only moderate transparency and moderate retroreflection efficiency, with $eV>|C_N|+|M| \quad \text{specular}.$9 values clustered near $\Delta$0 (Gloos et al., 2014).

A third misconception is that transport conductance by itself always distinguishes retro from specular reflection in mesoscopic devices. Finite-size scattering, intervalley mixing, interface smoothness, and multichannel interference can obscure the distinction, as shown explicitly in finite graphene geometries. For that reason, the field has increasingly relied on auxiliary observables—carrier-sign imaging, magnetic focusing, Aharonov–Bohm periods, terminal routing under magnetic field, and shot noise—to identify retroreflection in a manner that is less model dependent than conductance alone (Wang et al., 2017, Bhandari et al., 2020, Schelter et al., 2011, Salim et al., 30 Aug 2025).

Understood in this broader context, retro Andreev reflection is both a classical benchmark and a diagnostic limit. In the simplest metal/superconductor contact it is the standard electron–hole retracing process; in contemporary quantum materials it becomes a tunable scattering channel whose survival, suppression, or coexistence with competing channels reveals interface transparency, disorder, band inversion, pseudospin structure, Fermi-surface geometry, coherence length, and nonlocal pairing architecture (Gloos et al., 2014, Guigou et al., 2010).