WorksPhysics and Astronomy
Asymmetric Little–Parks Oscillations as a Probe of Coupled Pairing Channels in Surface-Superconducting \texorpdfstring{PtBi}{PtBi2}
Abstract
Motivated by the recent discovery of unconventional surface superconductivity in PtBi, we theoretically study its signatures in the Little-Parks oscillations of a narrow ring made from the same material. Using a Ginzburg-Landau model with three coupled pairing channels allowed by the trigonal crystal symmetry, we show that a fixed in-plane magnetic field can make the mean-field transition temperature asymmetric under reversal of the normal flux through the ring, while preserving periodicity in the superconducting flux quantum. The leading weak-field contribution to this asymmetry is cubic in the in-plane field strength. The flux-odd component respects the threefold crystal symmetry and changes sign when the in-plane field is rotated by at fixed magnitude. Within the homogeneous ring model, this asymmetry is absent for a single pairing channel and arises from the coupling between the three channels. Our calculations further show that increasing the in-plane magnetic field can enhance the superconducting nucleation temperature through mixing of pairing channels, illustrating magnetic control of multicomponent superconductivity in mesoscopic rings. Little-Parks interferometry with independent control of flux and field direction thus offers a mesoscopic probe of the pairing state, complementary to spectroscopic studies of the gap structure.
Introduction
The Little–Parks (LP) effect [1] is one of the clearest manifestations of macroscopic phase coherence in a multiply connected superconductor. In the conventional single-component Ginzburg–Landau (GL) description, fluxoid quantization makes the transition temperature periodic in the superconducting flux quantum , with a reciprocal response built from parabolic winding branches [2], , with . Moreover, LP interferometry is a phase-sensitive probe of unconventional and multicomponent superconductivity. A familiar example is the -shifted LP pattern, with extrema displaced from integer to half-integer flux, long discussed in connection with nontrivial pairing symmetry and multicomponent order [3, 4, 5, 6]. In two-dimensional rings, vortices near the Berezinskii–Kosterlitz–Thouless (BKT) transition can produce half-quantum shifts even without unconventional pairing [7], and half-quantum fluxoid features provide a further class of unconventional LP signatures [8, 9]. Unusual LP oscillations need not, however, imply a different flux period: multicomponent superconductors can preserve strict periodicity while reshaping the branch structure, shifting avoided crossings, or changing the relative weight of winding sectors [10].
Trigonal PtBi, in which superconductivity confined to the topological surface Fermi arcs has been observed [11, 12] together with signatures of two-dimensional superconductivity [13], provides a natural setting for unconventional LP oscillations. A multicomponent surface GL functional constrained by the point group , with superconducting sectors , , and and several symmetry-allowed magnetoelectric and Lifshitz invariants, has recently been proposed [14]. It builds on phenomenological and microscopic studies of Lifshitz invariants and magnetoelectric effects in noncentrosymmetric superconductors [15, 16, 17] and on the theory of polar superconductors [18]. Of particular interest are field-dependent gradient coupling terms that mix the one-component or order parameter with spatial variations of the two-component order parameter. In a ring, these terms couple the order-parameter modes with different winding numbers, describing different phase windings around the circumference. Together with the other field-induced couplings, they modify the leading superconducting instability and can make the LP transition curve asymmetric under reversal of the normal flux at a fixed in-plane magnetic field.
The normal flux and the in-plane field are independent control parameters, where and is measured from the positive axis in the ring plane. Throughout this work is performed at fixed “in-plane” values of and . A nonreciprocal Little–Parks response means
This asymmetry is compatible with magnetic time reversal, which reverses both the normal flux and the in-plane magnetic field:
Although motivated by PtBi, the proposed ring-based approach may also be applied to other layered noncentrosymmetric superconductors with suitable crystal symmetries and magnetoelectric couplings. Such a nonreciprocal response does not require a nodal superconducting gap.
We stress that the nonreciprocity under consideration concerns the equilibrium transition temperature under normal-flux reversal, not a current reversal. It is also distinct from the superconducting diode effect [19, 20] and from the nonreciprocal transport associated with LP oscillations in chiral superconducting nanotubes [21]. A superconducting diode effect has been observed in trigonal PtBi for both in-plane and perpendicular field orientations [22]. After an in-plane field is applied and removed, the effect persists at zero applied field, with its polarity controlled by the preceding field direction.
Other sources of LP asymmetry have been identified. Equilibrium LP phase shifts are predicted for rings partially covered by a ferromagnetic insulator, where the spatially nonuniform magnetoelectric coupling supplies a phase bias [23]. Magnetic coupling to chiral order can reshape the LP boundary but without normal-state time-reversal-symmetry breaking this modulation is symmetric under field reversal [24]. Field misalignment provides a geometrical source of asymmetric patterns [25].
The mechanism studied here is different: in a planar ring with spatially uniform GL coefficients, intercomponent winding coupling generate a flux-odd response that is cubic in the in-plane field at weak field and whose angular dependence is locked to the trigonal crystal axes.
We find that the LP oscillations remain strictly -periodic but acquire a flux-odd component. This asymmetry arises from the combined action of field-induced couplings that mix the pairing channels and couple modes with different winding numbers. Its leading weak-field contribution is proportional to the cube of the in-plane field strength, . The angular dependence reflects the threefold crystal symmetry: rotating the in-plane field by at fixed magnitude reverses the sign of the flux-odd component. In an ideal homogeneous ring, a single pairing channel yields a strictly reciprocal LP curve; consequently, the observed asymmetry indicates an interaction between the pairing channels. Finite flux can also raise the transition temperature above its zero-flux value at the same in-plane field. For the representative parameters studied here, this enhancement is generally smaller than the amplitude of the flux-odd component and varies strongly with the field and ring radius. The resulting signatures are compatible with the proposed i-wave pairing in PtBi [12, 14] without singling it out and complement probes of its nodal gap structure.
The paper is organized as follows: § 2 summarizes the quadratic multicomponent GL functional, § 3 derives its ring projection and winding-space structure, and Sec. IV formulates the generalized eigenvalue problem for . Numerical results are presented in § 5, followed by a discussion in § 6 and conclusions in § 7. Appendix § 9 discusses the choice of quadratic invariants and boundary contributions. Appendices B and C provide the numerical implementation and the analytical symmetry and perturbative analysis, respectively.
Ginzburg–Landau Free Energy
We start from the multicomponent Ginzburg–Landau functional
with gauge-covariant derivative
where is the Cooper-pair charge, is the elementary charge, and is the vector potential. The two-component -sector order parameter is
and the order parameters and transform as the and irreducible representations of , respectively. These order parameters label crystal-symmetry channels, not spin-singlet or spin-triplet sectors. In particular, the channel can contain both singlet and triplet contributions while producing the nodal -wave surface gap discussed in Ref. [12]. Following Ref. [14], we use an effective order parameter for each representation and do not resolve its microscopic singlet–triplet composition.
Near the superconducting transition, the normal state
loses stability when the lowest eigenvalue of the linearized GL operator passes through zero upon cooling. We denote the corresponding mean-field transition temperature by . In a finite, effectively one-dimensional ring, thermal amplitude and phase fluctuations broaden the superconducting onset, while phase slips can produce resistance below . A possible BKT transition instead concerns vortex–antivortex unbinding in an effectively two-dimensional system and is not assumed in the present one-dimensional model. Consequently, the measured resistance need not be described by a transition curve that shifts rigidly with the temperature .
We parametrize the temperature dependence of phenomenological coefficients as
and treat all remaining GL coefficients as temperature-independent constants. Here, is the transition temperature of the uncoupled, spatially uniform channel , and is its coefficient. The actual LP transition temperature is obtained from the coupled problem and need not coincide with any .
To isolate a concrete mechanism, we preserve the quadratic invariants listed below and set additional intersector cross-gradient terms and higher-order couplings to zero. The quadratic free-energy density is then given by [14]
The individual terms in Eq. (8) are specified below. All phenomenological coefficients are real. We introduce the following combinations of derivatives of the components:
Here, and are the divergence- and curl-like combinations, while and describe the symmetric traceless gradient combinations.
The quadratic mass and gradient terms are
The additional symmetry-allowed gradient term is set to zero here. Its relation to the effective tangential stiffnesses is discussed in Appendix § 9.
The zero-field mixed Lifshitz terms couple the scalar channels to derivatives of the components:
Writing the in-plane field as , the field-linear couplings without derivatives are
The field-dependent helical Lifshitz terms in the scalar sectors are
Here and below, c.c. denotes the complex conjugate of the preceding expression inside the same brackets.
The corresponding terms within the channel are
Finally, the tensorial Lifshitz terms couple the scalar channels to the combinations and :
These terms involve the symmetric traceless gradient combinations and generate couplings between winding indices differing by two in the ring model of § 3.
We use the expressions above as written. Their relation to alternative forms obtained by integration by parts, and the associated boundary contributions, are discussed in Appendix § 9.
Ring-geometry reduction
We consider a narrow superconducting ring of radius in the plane, parametrized by the polar angle . We treat it as an effectively one-dimensional system: the order-parameter fields depend only on , and we retain only covariant derivatives along the circumference in the quadratic GL terms. Their coefficients are phenomenological parameters of the ring and need not coincide with those of an extended superconducting surface. A complete derivation from a finite-width annulus with specified edge conditions is not attempted here.
For a normal (perpendicular) flux , the covariant derivative along the circumference is
where is the arc-length coordinate and . We distinguish the normal flux threading the ring from the local perpendicular magnetic field in the superconducting annulus. In the ideal flux-threading geometry, the latter vanishes, while the vector potential has circulation . Thus, the flux enters through , whereas explicit local- couplings are omitted. For a uniform perpendicular-field sweep, neglecting these couplings is an additional approximation. In this one-dimensional model, the Cartesian derivatives in the GL terms are replaced by
corresponding to differentiation along the local tangent .
For the doublet, it is convenient to use the components along the local radial direction and the tangent. The corresponding amplitudes and are related to the Cartesian components by
When differentiating these expressions, the angular dependence of the basis must be preserved.
The order-parameter components are single-valued, , and can be expanded in a series of winding harmonics:
Here is the amplitude of the th winding harmonic of the component in the chosen ring basis. Acting on a harmonic, the covariant derivative gives
Thus, the normal flux shifts the wave number entering the gradient energy. We collect the amplitudes with the same winding index into
Although the phenomenological coefficients are constant in the fixed crystal frame, the direction of the ring tangent changes with . This introduces angular factors into the one-dimensional GL terms. For example, the derivative combination in the scalar helical Lifshitz terms becomes
The cosine factor couples the winding harmonics with those indices differing by one, since a multiplication of by produces . This angular dependence reflects our ring geometry, not a spatial variation of the material coefficients.
The tensorial Lifshitz terms involve the gradient combinations and . Using the derivative replacements and the local-basis transformation above gives
Thus, the factors make these terms couple modes whose winding indices differ by two, . The curvature terms proportional to arise from differentiating the local basis. For example, the differential operator in the first expression acts on a winding harmonic as
Thus, multiplication by changes the winding index from to , while the differential operator supplies the coefficient . Further details of the gradient reduction are given in Appendix § 10.
The quadratic GL free energy in Eq. (8) gives the linearized GL equations , where . In this component basis, the differential operator is
All entries are scalar differential operators. The diagonal entries, with , are
The couplings within the scalar and sectors are
The scalar–doublet couplings are
All derivatives act on the order-parameter components to their right. The dagger denotes the operator adjoint with respect to the inner product on the ring,
The remaining matrix entries follow by taking the corresponding operator adjoints.
The factors and couple winding indices differing by one, whereas and couple terms with indices differing by two. The remaining, angle-independent terms preserve the winding index.
Substituting the winding expansion Eq. (28) into the quadratic free energy and integrating over gives the GL matrix in the winding basis. Its entries are
Here label the order-parameter components. The differential operator acts on before the integration, which extracts the th Fourier coefficient. Each matrix element therefore describes the coupling between the amplitudes and .
The angular factors in the operator above determine which winding modes are coupled. We define , where and are the incoming and outgoing winding indices, respectively. For the quadratic GL terms retained here, only occur. The quadratic free energy takes the form
where is the ring width and denotes the ring circumference. The overall factor does not affect we are looking for. The angle-independent terms contribute to the blocks. The factors and in the field-dependent couplings generate blocks, whereas the second harmonics and in the tensorial Lifshitz couplings generate blocks. The latter provide the additional winding-mode mixing that is absent in the conventional scalar LP problem.
The resulting GL matrix satisfies
where reverses the winding index, , and acts trivially in the component space. Eq. (45) implies the magnetic time-reversal relation for the transition temperature, Eq. (2).
Eigenvalue Problem for
Truncating the winding expansion to and stacking the sector vectors into , the GL free energy becomes
For the given functional, the matrix is block-pentadiagonal because the projection produces only matrix elements. We write
where is assembled from coefficients appearing in § 2, and the superscript denotes the temperature-independent part of the GL matrix (it does not denote zero magnetic field).
The normal state becomes unstable when the lowest eigenvalue of crosses zero. Equivalently, the transition temperature is the largest generalized eigenvalue of
Thus, the critical temperature is given by
where are the real finite generalized eigenvalues of Eq. (48). The maximization over all branches is responsible for the conventional LP behavior. Increasing the flux by one flux quantum can be compensated by decreasing every winding number by one, since . This relabeling leaves the GL eigenvalue spectrum unchanged, so is exactly -periodic. In the numerical calculations, we keep sufficiently many winding modes to reproduce this periodicity to numerical accuracy.
To describe the LP effect, we introduce three physical characteristics. The finite-flux enhancement is
the nonreciprocal odd-in-flux amplitude is
and, at a fixed reference flux , a signed odd response is
Here, is a representative value of the reduced flux , fixed while and are varied. It is chosen in the interval since the odd combination vanishes both at and, by periodicity, at . Here we use . Unlike , which takes the magnitude maximized over the flux, preserves the sign and is useful for displaying the sign-changing angular structure of the nonreciprocal response. All maximizations over the flux (for and ) are performed over one period, i.e. .
The angular dependence of the flux-odd response is constrained by the three mirror planes of . Their intersections with the ring plane define mirror lines at angles , where in our coordinates and . When the in-plane field is perpendicular to a mirror line (), the reflection reverses the normal flux but leaves the in-plane field unchanged. Consequently, , and both and vanish. For a field parallel to a mirror line, reflection reverses both magnetic controls and does not require reciprocity at fixed in-plane field. In our calculations, we choose , parallel to a mirror line.
Numerical Results
All numerical results in this section are obtained from the directly projected GL matrix of § 3. The truncation, quadrature, grids, and convergence checks are given in Appendix § 10.
Throughout, dimensionless units are used: temperatures are measured in units of the largest bare critical temperature, , and lengths in units of
with and taken from Table I, where both equal unity. The in-plane field is measured in units of a reference scale that is absorbed into the field-linear coefficients. The dimensionless GL coefficients used for all figures in this section are collected in Table I. They are identical for every figure.
| Sector / coupling | Symbol | Value |
| Bare critical temperatures | 0.90, 1, 0.75 | |
| Mass prefactors | 1, 1, 1 | |
| Gradient stiffnesses | 1, 0.9, 0.6, 0.45, 0 | |
| Zero-field mixing terms | 0, 0 | |
| Field-linear non-derivative terms | ||
| Scalar-sector helical Lifshitz terms | ||
| -sector Lifshitz terms | ||
| Tensorial Lifshitz terms |
Table I. Representative dimensionless GL coefficients used in the numerical figures. They are illustrative parameter choices, not fitted material constants due to the missing experimental realization of the proposed rings, so far. The geometric and field-control parameters , , and are specified separately in each figure caption.
In the surface-pairing picture of [12, 14], the channel corresponds to the proposed nodal -wave gap profile. The parameter hierarchy in Table I selects an instability at zero field and zero flux. Thus, for the parameters used here, the weak-field nonreciprocal response originates from an (-wave) parent state coupled to the subdominant and pairing channels. However, neither the cubic dependence on the in-plane field strength nor the trigonal angular pattern is unique to pairing. Both are symmetry allowed for other parent states. At higher fields, the leading instability can also change its dominant pairing-channel character.
We first plot the LP shift
defined in such a way that positive denotes the usual flux-induced suppression of , whereas negative values indicate a finite-flux enhancement.
Figure 1 illustrates the main effect at two in-plane magnetic fields. The LP oscillations remain periodic in and, for the representative parameter set, acquire a finite odd-in-flux component. At [Figure 1(a) and Figure 1(b)], it becomes largest close to a half-integer flux. At [Figure 1(c) and Figure 1(d)], the odd component has the opposite sign and is about four times larger.

Figure 1. (a), (c) LP shift and (b), (d) flux-odd component of at and for (a), (b) and (c), (d) . The red dashed horizontal lines in panels (b)–(d) mark the zero level of the respective plotted quantities.
This sign change accompanies a rearrangement of the leading instability, that is, a change in its component character. Writing the leading eigenvector of Eq. (48) as , with given by Eq. (30), we define the component weights
where so that with . The weight does not depend on the in-plane basis chosen for the doublet.
Figure 2 shows these weights at . At weak fields, the leading instability is dominated by the channel, with a field-induced admixture of the components. As the in-plane field increases, a mode dominated by the and components acquires a higher transition temperature and becomes the leading instability between and . At both and , the two leading eigenvalues approach each other without crossing, resulting in an avoided crossing. The field at which depends weakly on flux. The zero of at , marked by the dotted line in Figure 2, lies near the upper end of this crossover. At the mirror-preserving orientations , the situation is different: at zero flux, the two leading modes have opposite mirror parities and cross at . A finite flux breaks this mirror symmetry and turns the crossing into an avoided one.

Figure 2. Component content of the leading instability. Weights , , and , Eq. (55), vs. in-plane field at (solid) and (dashed), for and . The weight decreases from unity to a few percent across the mode-reorganization region, while the and components take over. The dotted line marks the zero of the response at (see Figure 3, center). Phenomenological coefficients are given in Table I.
For the coefficients in Table I, the odd-in-flux response grows as at sufficiently weak in-plane fields and originates from an -dominated instability. At higher fields, the larger odd amplitudes are associated with the field-reorganized mode dominated by the and components. The two field values used in Figure 1 illustrate this evolution. At and , the weight varies between 0.72 and 0.86 over the flux period at , whereas it remains below 0.08 at . Thus, the larger nonreciprocal response at characterizes the field-reorganized instability rather than the nearly pure parent state.
We now examine the field and angular dependence of the LP response in Figure 3. The three panels show the absolute odd amplitude , Eq. (51), the signed response , Eq. (52), and the finite-flux enhancement , Eq. (50), respectively.

Figure 3. Field-angle dependence of the nonreciprocal LP response. Left: absolute odd-in-flux amplitude . Center: signed odd response at , displaying alternating angular lobes and a polarity reversal in the mode-reorganization region near . Right: finite-flux enhancement , which is weak below the mode-reorganization region, shows a narrow ridge within it, and grows beyond it. The radius is fixed to , the other coefficients are given in Table I.
The map of is one of the central numerical results.
It exhibits clear angular lobes and an overall growth with increasing in-plane field, with a suppression in the mode-reorganization region. The signed response reverses sign near , consistent with the change in the leading instability shown in Figure 2. The finite-flux enhancement remains small at lower fields and exhibits a narrow ridge near –, associated with the reorganization of the leading modes at zero flux.
The finite-flux enhancement remains small at low fields but becomes more pronounced near and beyond the mode-reorganization region. A narrow enhancement ridge appears around –. At the mirror-preserving orientation , this feature originates from a crossing of the leading modes at zero flux. A small normal flux mixes these modes, producing two symmetric maxima of at finite flux. This enhancement is reciprocal and occurs even when the flux-odd response vanishes. It provides an additional signature of mode reorganization, distinct from the nonreciprocal LP response.
The periodicity follows from gauge invariance, while the trigonal crystal symmetry constrains the angular dependence of the flux-odd response. Its magnitude and field dependence depend on the GL coefficients. In the present model, the leading cubic contribution arises from three field-linear couplings connecting winding modes, for example, . The first two steps change the winding number by one, while the last involves a tensorial Lifshitz coupling that changes it by two in the opposite direction. Since each coupling is proportional to , their combined contribution is proportional to . Different sequences can contribute with opposite signs, so their contributions may partially or completely cancel.
For the parameters in Table I, setting reduces below for all scanned fields and orientations. This confirms that the tensorial couplings are essential for the nonreciprocal response within the retained model. The resulting cubic weak-field onset is demonstrated numerically in Figure 4, while its perturbative derivation is given in Appendix § 11.

Figure 4. Weak-field cubic onset of the nonreciprocal LP response. at and , shown over two decades in in-plane field on a log-log scale. Circles connected by a solid line show the numerical results and the dashed line is a guide. The response approaches the cubic dependence at weak fields and deviates from it toward the upper end of the plotted interval. The suppression associated with mode reorganization occurs at larger fields, outside this interval (see Figure 3).
The angular dependence is clarified by the signed quantity , shown in the center panel of Figure 3, which avoids the absolute value in and exposes the sign structure of the nonreciprocal response directly. The computed signed response obeys the expected magnetic-time-reversal property,
and is dominated by a single trigonal harmonic,
whose phase is set by the mirror-line angle : is largest for fields parallel to a mirror line and has opposite signs for adjacent mirror lines. Here, is the signed coefficient of the angular harmonic at the chosen reference flux and radius. The omitted terms contain the higher allowed angular harmonics. In fact, the angular dependence is constrained exactly by symmetry (see Appendix § 11): invariance, , together with the field-reversal sign change, , restricts the Fourier content to odd multiples of , . A discrete Fourier analysis of shows that the term dom-
inates: over the displayed field range the amplitude stays below 1.4% of the largest amplitude. Only near the polarity reversal at , where changes sign, does the term become comparable to the term. The same two symmetry relations give , and the analogous relation holds at every flux. Consequently, the positive amplitude has an exact 60° periodicity, whereas the signed response has period 120°.
Finally, we study the radius dependence at for the two fields of Figure 1, which lie on either side of the mode reorganization. The odd response and the finite-flux enhancement for , inside the -dominated regime, are shown in Figure 5(a), while the same quantities for , where the leading mode is dominated by the and components, are shown in Figure 5(b). At both fields, the absolute odd response decreases monotonically over the plotted range so that the absolute nonreciprocity is largest at the smallest radii considered. It decays faster at and falls below the value for . The two fields differ qualitatively in . At , see Figure 5(a), remains at least three orders of magnitude smaller than . At , see Figure 5(b), has its maximum near , where it is about an order of magnitude below , and the maximizing flux lies on the opposite side of integer flux from that at , consistent with the sign reversal between the two fields. Together with the maps in Figure 3, this shows that, for the representative parameters, appreciable finite-flux enhancement occurs near and beyond the mode reorganization.

Figure 5. Radius dependence of the absolute odd response (blue, left axis) and the finite-flux enhancement (red, right axis) at for (a) and (b) .
To assess the dependence on the coefficients, we have repeated the calculation for modified parameter sets that change the coefficients of Table I by , added the channel-dependent pair breaking
or combined both modifications, (for details see Appendix § 10). The real coefficients describe channel-resolved pair breaking in the dimensionless normalization used here. For all modifications, the leading instability at weak field remains dominated by the component and the odd response persists, while its amplitude varies by about an order of magnitude. The field of the mode reorganization also depends on the coefficients. For some sets the dominance of is already lost at .
For the parameters given in Table I, the weak-field response has the form . The cubic coefficient contains products of three field-linear couplings: one non-derivative, one tensorial, and one helical Lifshitz coupling. For example, a contribution can be proportional to . Different contributions can have opposite signs and may cancel, so neither the magnitude nor the sign of is universal.
Such cancellations can suppress the cubic term without eliminating the response at higher orders. In contrast, the entire odd response vanishes at mirror-preserving field orientations or, for the retained functional with , when one of these three groups of couplings is absent. Further details are given in Appendix § 11.
The parameters in Table I do not include pair breaking by the in-plane field. Consequently, the field-linear couplings can increase by mixing pairing channels. Our results near and beyond the mode reorganization therefore assume that superconductivity survives up to these fields.
For the equal temperature slopes used here, the channel-independent pair breaking, described by Eq. (58) with , shifts downward without changing its flux dependence at fixed fields (see Appendix § 10).
If the field-linear couplings originate from the Zeeman effect, their natural field scale corresponds to Zeeman energies of the order of . For a reference field of this order, the mode reorganization may occur near the Pauli limit, while the largest fields considered here may exceed it. Strong spin-momentum locking can weaken the paramagnetic pair breaking and favor helical superconductivity under an in-plane field [26, 27].
Importantly, the cubic weak-field response does not require superconductivity to persist up to the larger fields where the modes reorganize.
Discussion
Our numerical results are based on the representative parameter set in Table I and the variations discussed above. We have not attempted an exhaustive exploration of the GL parameter space. Our calculation determines the nucleation temperature and the corresponding order-parameter mode. Determining the superconducting phases below this temperature would require inclusion of the quartic GL terms.
Field-induced competition between pairing channels has also been investigated in the related layered compound -PdBi, where tunnelling measurements provide evidence for a change from fully gapped to nodal superconductivity under an in-plane field. This behaviour was interpreted in terms of competing - and -wave states [28]. The proposed transition occurs within the superconducting state and is distinct from the field evolution of the nucleating mode studied here.
The LP response constrains the pairing state without identifying it. In the homogeneous ideal ring considered here, a single-component order parameter gives an exactly reciprocal LP curve, regardless of its representation and singlet-triplet content (see Appendix § 11). Within this setting, the flux-odd response requires coupled pairing channels. In the present functional, these are a scalar channel and the doublet channel, coupled by the tensorial invariants. Additional boundary or spatial structure of a real ring can relax this requirement. The LP signatures do not uniquely identify an parent state since an or an parent state with the same couplings gives a response of comparable or larger (see Appendix § 11). Pair breaking of the form of Eq. (58), through which Zeeman effects on singlet and triplet components differ, leaves the coefficient of the cubic dependence of the odd response unchanged. The symmetry of the response is fixed by the crystal structure, whereas its sign and magnitude depend on the coefficients and on edge contributions (see Appendix § 10).
While symmetry constrains the allowed nonreciprocal LP response, its magnitude, sign, and characteristic field scales depend on the GL coefficients and the ring radius. Quantitative comparison with future experiments on PtBi rings will require device-specific parameters. We predict a signed, field-orientation-controlled, flux-odd transition-temperature component generated by the retained multicomponent couplings. Finite-flux enhancement and shifted extrema are additional, more parameter-dependent features. Taken together, these features constitute a testable scenario of the proposed multicomponent magnetoelectric mechanism. They are compatible with an (-wave) parent state in PtBi but do not by themselves uniquely identify the order-parameter symmetry. The calculated LP response does not determine the relative spin-singlet and spin-triplet contributions within the pairing channels. Since the GL functional is formulated in terms of these representation amplitudes the singlet and triplet contributions enter only through the values of its coefficients and the present transition-temperature calculations do not determine their relative weight.
LP interferometry would complement existing probes of surface superconductivity in PtBi. Angle-resolved photoemission constrains the superconducting Fermi arcs and their momentum-resolved gap structure [11, 12], while tunnelling spectroscopy, vortex imaging, and Josephson measurements provide local spectral information and evidence of phase coherence [29]. The proposed LP experiment addresses a different aspect: the response of the superconducting instability to phase winding and independently controlled magnetic fields. The combined flux, field-angle, and radius dependencies could constrain field-induced coupling between pairing channels, including channels that remain subdominant at zero field. For the parameter studies studied here, the nonreciprocal response tends to zero as when the in-plane field vanishes. The larger amplitudes found at higher fields are associated with a leading instability dominated by the and components rather than the zero-field parent. This information is complementary to identifying the momentum-space nodes or gap-sign structure. Provided the superconducting state of the patterned device is independently characterized, comparison with spectroscopic constraints could test whether a common multicomponent description accounts for both the gap structure and the mesoscopic magnetic response. We stress that the calculated enhancement of superconductivity at weak and intermediate in-plane fields is accompanied by an increased admixture of the and components to the initial parent state. To the best of our knowledge, such mesoscopic tuning of an unconventional multicomponent superconductor by external magnetic fields, accompanied by the nonreciprocal LP response described here, has not been reported previously.
Temperature-dependent scanning SQUID microscopy could provide a contactless route to investigate the flux-dependent superconducting onset at fixed in-plane field, complementing transport measurements [30]. Scanning SQUID measurements have already detected a diamagnetic response in PtBi [31]. Relating the measured magnetic response to the calculated mean-field would require accounting for superconducting fluctuations, which can be significant in narrow rings near the transition [32].
For the representative parameters studied here, the in-plane magnetic field can raise the calculated nucleation temperature through mixing of pairing channels [33]. This mechanism, explained above in terms of field-tuned multicomponent superconductivity in the mesoscopic rings, differs from the exchange-field compensation underlying the Jaccarino–Peter effect [34]. Whether a net increase occurs in an experimental device, particularly at intermediate and high in-plane magnetic fields, depends on the competition between the enhancement considered here and orbital and paramagnetic suppression of superconductivity by the same field. Although Figure 2 shows that the leading instability becomes dominated by the and channels, this does not establish spin-triplet dominance or survival beyond the conventional Pauli limit.
Related GL studies of Rashba superconductors illustrate how spin–orbit coupling can modify the magnetic-field response [35] with respect to the upper critical field peculiarities in case of bulk superconductivity. Neither surface superconductivity nor the LP effect have been addressed in the literature to the best of our knowledge, so far. Anyhow, despite the still necessary increase of the already large number of parameters, a useful extension of the present approach could explicitly resolve the spin-singlet and spin-triplet contributions within the pairing channels, respecting the symmetry of the PtBi surface. Such a treatment could clarify their influence on the magnetic response of the proposed (-wave) state.
A self-consistent treatment of surface and bulk superconductivity in time-reversal-symmetric Weyl semimetals shows how the penetration of Fermi-arc states into the bulk influences the surface gap and its temperature dependence [36]. Microscopic proposals for surface superconductivity include phonon-mediated pairing on Fermi arcs [37]. For PtBi, an -wave instability has also been obtained within an electronic Kohn–Luttinger approach [38], while anisotropic electron–phonon coupling combined with statically screened Coulomb repulsion provides another proposed route to nodal pairing [39]. In the latter study, five distinct gap functions emerge in different regions of the interaction-parameter space, including the nodal -wave profile considered here. Our phenomenological GL analysis does not select between these microscopic mechanisms. It studies the LP response of coupled symmetry-allowed pairing channels.
Conclusions
We have developed a multicomponent GL theory for the LP effect for a narrow ring with surface order-parameter sectors, motivated by the symmetry structure proposed for PtBi. The resulting LP response remains strictly -periodic but at fixed in-plane field it can be nonreciprocal in the flux through the ring. At weak in-plane fields, the leading symmetry-allowed flux-odd response is proportional to the cube of the field strength. It follows the threefold crystal symmetry and changes sign when the field direction is rotated by . Its periodic flux dependence can contain several sine harmonics. The GL functional produces a nonreciprocal response through the combined action of couplings between modes whose winding numbers differ by one or two, with the latter arising from tensorial Lifshitz terms. Because processes of opposite sign compete, the sign and size of the cubic contribution depend on the GL coefficients. For the representative coefficients, the in-plane field also changes the character of the leading instability, from predominantly at weak field to predominantly and , where the larger odd amplitudes obtained here belong to the field-reorganized mode rather than to the low-field state.
Finite-flux enhancement is possible but is more parameter dependent. For the representative parameters, appreciable enhancement occurs near and beyond the mode reorganization. Furthermore, we find that the odd-in-flux response decreases with increasing radius for both
the field values we considered. Hence, the calculation does not suggest an optimal intermediate radius for the nonreciprocal response.
An experimental test should combine flux reversal, in-plane field reversal, and crystallographically referenced angular scans with independent calibration of field misalignment and flux offsets. Reversal tests alone would not uniquely distinguish the proposed mechanism from geometrical asymmetry [25].
Because the coefficients used here are representative rather than fitted, a quantitative comparison with PtBi devices requires estimates of the surface GL coefficients, which set the magnitude of the odd response and the field scale. It also requires relating the calculated , the mean-field temperature at which superconductivity nucleates, to the measured resistance. Converting resistance oscillations into oscillations through the temperature slope of the resistance assumes that the resistive transition shifts rigidly with . However, PtBi flakes show a BKT transition [13] and in such a two-dimensional superconductor the resistance between the BKT and mean-field transitions arises from thermally excited free vortices. Near the BKT transition, its flux dependence can follow the free-vortex density rather than , and, depending on the ring width, can even be shifted by half a flux quantum [7].
Nevertheless, our results provide a concrete and timely motivation for experimental LP studies of mesoscopic PtBi rings with independently controlled out-of-plane flux and in-plane magnetic field. The predicted flux asymmetry, cubic weak-field onset, and crystallographically locked angular dependence offer testable consequences of multicomponent magnetoelectric coupling. In the representative parameter regime, the field-induced sign reversal and the distinct radius dependences of nonreciprocity and finite-flux enhancement provide further tests of the evolution of the superconducting instability. Such measurements could constrain the coupling between competing pairing channels and complement spectroscopic investigations of the proposed nodal surface state. The thin-ring LP approach developed here for the -wave scenario in -PtBi may also be used to investigate other unconventional surface pairing states within multicomponent GL models. A systematic comparative study of the corresponding GL functionals is left for future work.
Acknowledgments
We thank Marieke Altena, Flavio S. Nogueira, Ulrich Rößler and Sergey Borisenko for fruitful discussions. We acknowledge financial support by the Deutsche Forschungsgemeinschaft (DFG) through the Collaborative Research Center SFB 1143, projects A04 and A05, Project ID 247310070, and the Würzburg-Dresden Cluster of Excellence on Complexity, Topology and Dynamics in Quantum Matter—ctd.qmat, EXC 2147, Project ID
Additional remarks on the quadratic functional
Here, we discuss the choice of quadratic terms in Sec. II, their relation to alternative invariant forms, and the boundary contributions relevant to the effective ring model.
The symmetry also permits the symmetric traceless -sector gradient term . Under the tangential derivative replacement used in Sec. III, the identity
holds pointwise. This term can be absorbed into the effective ring stiffnesses through
We set after incorporating its contribution into and . This redundancy is specific to the tangential model and does not establish equivalence between general finite-width annuli with different boundary conditions.
The zero-field mixed Lifshitz terms in Eqs. (16) and (17) are understood under the spatial integral. They differ from the real parts of the alternative forms in Ref. [14], and its counterpart, only by total derivatives in two dimensions. Integration by parts leaves the functional unchanged when the resulting boundary contributions are included. For a finite-width annulus, azimuthal periodicity removes boundary contributions along the circumference, but not those at the inner and outer edges. Applying the tangential derivative replacement to different Cartesian forms without retaining the corresponding boundary terms can produce different curvature couplings of order in the effective ring model.
For example, Eqs. (16) and (17) give the winding-preserving matrix elements
whereas applying the same tangential prescription to the real parts of the alternative forms gives half these values. We use the expressions specified in Sec. II as written. Since in all calculations reported here, this particular difference does not affect the numerical results.
A similar dependence on the boundary prescription arises for the intercomponent contributions to the field-linear gradient terms in Eqs. (22)–(25). These couplings are nonzero in the numerical calculations. Changing their Cartesian forms without including the accompanying boundary contributions can modify the magnitude and polarity of the flux-odd response, as discussed in Appendix § 10. Such changes correspond to different effective ring functionals, rather than equivalent rewritings of the same complete functional.
Ring reduction and numerical implementation
Here, we describe how the Cartesian GL functional is reduced to the effective one-dimensional ring model and represented in the Fourier basis. We assume that the order parameters depend only on the angular coordinate and express the Cartesian derivatives in terms of tangential derivatives. The resulting differential operators are projected onto the winding modes to construct the matrix used in the numerical calculations. The starting point is the quadratic functional specified in Sec. II. For each angle , we express the components through using Eq. (27). The angular derivatives of the basis vectors are retained throughout. The GL matrix is obtained from the Fourier matrix element
where . In practice this integral can be evaluated either analytically or by high-resolution Fourier quadrature. The latter is particularly useful because it avoids premature simplification of the tensorial Lifshitz terms.
Carrying out the angular reduction explicitly, the local basis vectors and satisfy and , so that the divergence- and curl-like -sector gradients reduce to
while the symmetric traceless combinations and follow from
with and . The factor is the origin of the blocks, and the curvature-shifted operator reduces on winding sector to the real factor of Eq. (31).
The field-dependent invariants have the generic form
Both terms in Eq. (66) are retained. If denotes the Fourier projection of , the projection of is , and for real the contribution to the GL matrix is
Projecting the complete term and adding its Hermitian conjugate, including the factor , is equivalent to projecting the full expression. For invariants that couple different components forms related by integration by parts differ on the narrow ring by contributions from its inner and outer edges, which are of order , like the curvature terms that produce the odd response. We use the invariant forms specified in § 2, with the boundary prescription discussed in Appendix § 9. Other choices leave the symmetry of the response unchanged but can change its magnitude and reverse its weak-field polarity.
The resulting matrix has only blocks:
We denote
The full truncated matrix is reconstructed from these blocks as
where the notation indicates the corresponding block placement in winding space.
For the reference projection used to validate the implementation, the GL matrix is Hermitian by construction, so the nontrivial check is magnetic time reversal. Building the matrix independently at and at and comparing them, we find that Eq. (45) is satisfied with internal action to machine precision. This is the expected result because in the original GL basis the order parameters transform to their complex conjugates under time reversal. In the real Cartesian and radial–tangential component bases used here, no additional internal sign matrix is required. A different internal basis can have a nontrivial representation of time reversal and the representation must be transformed consistently with the GL matrix.
Unless stated otherwise, the calculations in § 5 use the winding truncation , which is sufficient for the displayed parameter sets, and evaluate the angular Fourier integrals with uniformly
spaced quadrature points. For the cubic-scaling results in Figure 4, we use , , and 401 flux points over . The field-angle maps use with 801 points, with 61 points, and with 121 points. The radius scans at and use with 301 points each and 1001 flux points. In the radius scans, is obtained by refining the maximum of with a bounded one-dimensional search between the grid points adjacent to the grid maximum. This matters at weak field: at and the maximizing flux lies closer to integer flux than half the grid spacing , so that the grid value of vanishes there although the itself is small but finite. In these calculations the relative residual of Eq. (45) stays below , and the -periodicity error is negligible on the scale of the plotted data.
Winding-cutoff and angular-quadrature convergence were checked separately from the resolution of the maximum over flux.
The sensitivity test of § 5 uses 56 parameter sets. Twelve sets change one group of coefficients (non-derivative, helical, tensorial, or stiffness) or one bare channel separation by . Twelve sets add the pair breaking of Eq. (58) with running over the permutations of and . The remaining 32 sets combine independent changes of the nonzero field-linear coefficients, stiffnesses, and channel separations with . The bare transition temperature and the mass coefficients are kept at unity, and the coefficients that vanish in Table I remain zero. All sets are evaluated at and . At , all of them keep over the whole flux period, and ranges from to of the LP amplitude, compared with for Table I. At , 14 sets lose the majority somewhere in the flux period. At , the twelve pair-breaking sets keep over the whole flux period, whereas channel-dependent pair breaking suppresses to zero at this field for .
Symmetry, mechanism, and scale of the nonreciprocal Little–Parks response
The weak-field power counting used below assumes an analytic expansion of an isolated instability eigenvalue at fixed flux. At winding-sector or internal-mode degeneracies, degenerate perturbation theory or direct diagonalization is required. In particular, the fixed-flux expansion need not be uniform near a branch degeneracy.
The transition temperature is a real scalar and must be invariant under the full magnetic symmetry of the surface: the unitary operations of , supplemented by time reversal. The normal flux transforms like the perpendicular field component , independently of the field profile: it is invariant under the threefold rotation and odd under a vertical mirror, and therefore transforms as . The in-plane field transforms as an doublet. The nonreciprocal part of the LP response is odd under , and hence its leading local dependence on flux transforms as . To obtain an invariant scalar, this factor must multiply an polynomial of the in-plane field.
Writing the in-plane field as , its transformation under the generators of is fixed. The threefold rotation acts as ,
while, in a frame whose axis lies along a mirror line (), the corresponding vertical mirror reverses the chirality of the in-plane field,
(an arbitrary mirror orientation is restored by the shift ). Since the flux-odd factor transforms as , we need a polynomial in the field that is invariant under Eq. (71) and changes sign under Eq. (72); enumerating the symmetric powers of locates the lowest one.
At first order, transforms as the doublet. We denote by the th symmetric tensor power of this representation. Here, this means the representation carried by all homogeneous polynomials of degree in the field components , or equivalently . Only symmetric products contribute because the components of the applied magnetic field commute: changing the order of the factors does not produce a distinct polynomial. At second order, these products decompose as
where is the rotation- and mirror-invariant modulus. There is no : the antisymmetric pairing that would carry it vanishes identically for a single field. At third order,
and the cubic pair finally supplies a one-dimensional . Under a cubic monomial accumulates a full rotation and is invariant, ; under it maps to . Writing , the real and imaginary parts therefore transform as
since the mirror sends while leaving invariant. The lowest-degree field polynomial transforming as is thus cubic, and for a general mirror orientation it reads
Thus, the leading smooth harmonic of the odd LP response has the form
where
More generally, each flux harmonic is separately odd in and also transforms as , so flux periodicity permits the same cubic angular factor to multiply any of them,
where the real coefficients label flux harmonics and depend on the radius and the GL coefficients. Eq. (78) should be understood as the leading smooth harmonic rather than as an exact representation of the branch-selected LP curve. Magnetic time reversal is automatic: and , so both factors in Eq. (78) change sign and their product is invariant. This gives
while still allowing
at fixed in-plane field.
As a result the angular harmonic is symmetry enforced at leading order. Higher powers of can renormalize the same harmonic through additional even powers of . Genuine higher angular harmonics compatible with the same magnetic symmetry occur only at odd multiples of three, . This explains why the signed numerical map is dominated by a trigonal sign-changing pattern, whereas the absolute-value map displays six lobes.
Symmetry fixes the form of Eq. (78), but not the coefficient . The mechanism producing a nonzero coefficient within the retained GL model is the winding-space structure of its ring-projected GL matrix. In the minimal cubic process, the field invariant is produced by three field couplings of the same angular chirality, giving . In the ring projection each winding step carries a definite field chirality: the field-linear and helical processes carry , while the tensorial field-dependent Lifshitz invariants and generate steps carrying . Consider three field-induced couplings that are all proportional to or all to . To return to the initial winding sector, their winding changes must sum to zero, . In the present model, the possible combinations are
up to permutations of the three steps. Each of these triangular processes contains a tensorial step. Let denote the truncated GL matrix with . Within the given functional, removing both tensorial coefficients eliminates the flux-odd response exactly, not only at cubic order. Indeed, the remaining field-dependent matrix elements connect only neighboring winding sectors. With
and , the generalized temperature spectrum is even in the in-plane field. Magnetic time reversal implies . Additional zero-field winding couplings need not preserve this extra relation, so this result is specific to the retained functional. Conversely, a nonzero isolated loop amplitude is not sufficient: the observable response depends on the leading mode and the sum over all contributing paths.
For , as in Table I, two further operations act in the same way. The sign reversal of the doublet, , maps to when the helical couplings vanish, and does so when the non-derivative couplings vanish. Each group of field-linear couplings is required to all orders in in this case; zero-field mixing lifts the last two constraints. Without a symmetry reason, can also vanish: increasing by 18% gives , and starts as there but at still reaches about 60% of its value for Table I.
Under the homogeneous ideal-ring assumptions used here, the parity of Eq. (85) holds for any single-component order parameter, whatever its representation and spin content. Time reversal makes the zero-field coefficients even in momentum, , so zero-field terms change the winding number by multiples of six and terms odd in the field by odd amounts. As a result maps to to all orders, and the LP curve is reciprocal. Accordingly, the , , or channel alone, or and without coupling to , gives no flux-odd response within numerical accuracy. Boundary or spatial structure beyond these assumptions, for example trigonal anisotropy of the ring edges, can lift this restriction.
The LP signatures also do not single out the parent representation: interchanging the bare transition temperature of the channel with that of the or the channel, at fixed couplings, gives a flux-odd response of comparable or larger size and for these couplings the parent yields the weakest response. The substitution , maps each bulk invariant onto an invariant of the same type, because , , and . With the degree prescription of Appendix § 9 the mixed scalar invariant Eq. (22) is not closed under this substitution. For , the ring inner product gives , so the substitution generates an additional curvature term proportional to . Only when this symmetry-allowed term is included does the -parent model map onto an -parent model with identical .
To calculate the contribution proportional to , we keep the full dependence of the coupling matrix elements on the winding wave numbers and define the Hermitian temperature operator
Since is positive definite, this transformation preserves the generalized eigenvalues of Eq. (48), where the largest one is . Let and denote the eigenvalues and normalized eigenvectors of , with the nondegenerate leading eigenvalue, and define . The zero-field operator is diagonal in winding, whereas changes winding by or , so . The third-order correction is given by
All couplings and eigenvalue separations preserve their actual flux and radius dependence. The flux-odd contribution follows by antisymmetrizing this correction between and . Each nonzero product represents a closed three-step winding process. The three-step sequences in Eq. (83) generate the cubic trigonal factor. The full sum determines its coefficient and permits cancellations between different paths. The pair-breaking term in Eq. (58) adds to the temperature operator in Eq. (86) so that
The field-independent operator is
The minus sign follows from the definition of the temperature operator: positive pair-breaking coefficients increase the quadratic GL energy and lower the transition temperature. In the orthonormal eigenbasis of , we define
The operator preserves the winding number, although it need not be diagonal in the eigenmode index. Since is diagonal in winding, its eigenvectors can be chosen with definite winding number. The operator connects sectors differing by , whereas connects only states in the same sector. Consequently, their mixed third-order contribution vanishes:
Such terms leave the cubic coefficient unchanged. Since the flux-odd part of is odd in , they first modify it at order . General symmetry-allowed field-quadratic invariants can mix winding sectors and change the cubic coefficient, but they are not included here.
The expansion applies when field-induced mixing is small compared with the relevant zero-field eigenvalue separations. Near a mode reorganization, the full diagonalization is required. Both the coupling matrix elements and the eigenvalue differences in the denominators depend on the ring radius. Therefore, this expression does not guarantee that has a maximum at a finite radius. The distinct radius dependences of and are illustrated in Figure 5.
For the Table I coefficients, , and , an independent implementation of the same Cartesian projection at the Figure 4 resolution gives with from fields between and . At these fields the true maximum lies within of half-integer flux, where the fixed-flux expansion is not uniform. Locally refined maxima give . Evaluating Eq. (87) at the fixed reference flux gives, independently of a field fit,
The fixed-flux coefficient provides a separate check that does not require maximizing over a discrete flux grid. The cubic asymptote should be distinguished from the finite-field amplitude and polarity changes in Figure 3.
References
- [1]W. A. Little and R. D. Parks, Observation of quantum periodicity in the transition temperature of a superconducting cylinder, Phys. Rev. Lett. 9, 9 (1962).
- [2]M. Tinkham, Introduction to Superconductivity, 2nd ed. (McGraw-Hill, New York, 1996).
- [3]V. B. Geshkenbein, A. I. Larkin, and A. Barone, Vortices with half magnetic flux quanta in “heavy-fermion” superconductors, Phys. Rev. B 36, 235 (1987).pubmed.ncbi.nlm.nih.gov/9942041
- [4]C. Kallin and J. Berlinsky, Chiral superconductors, Rep. Prog. Phys. 79, 054502 (2016).DOI
- [5]Y. Li, X. Xu, M.-H. Lee, M.-W. Chu, and C. L. Chien, Observation of half-quantum flux in the unconventional superconductor β-Bi2Pd, Science 366, 238 (2019).DOI
- [6]A. Almoalem et al., The observation of π-shifts in the Little–Parks effect in 4Hb-TaS2, Nat. Commun. 15, 4623 (2024).
- [7]Y.-M. Xie and N. Nagaosa, Theory of Little-Parks oscillations by vortices in two-dimensional superconductors (2026), arXiv:2601.23050.arxiv.org/abs/2601.23050
- [8]Y. Yasui et al., Little-Parks oscillations with half-quantum fluxoid features in Sr2RuO4 microrings, Phys. Rev. B 96, 180507(R) (2017).DOI
- [9]C. Hua, E. D. Tetra, M. Andriescu, and G. B. Halász, Theory of the Little-Parks effect in spin-triplet superconductors, Phys. Rev. B 107, 214503 (2023).DOI
- [10]Y. S. Yerin, S. V. Kuplevakhksii, and A. N. Omel’yanchuk, Little-Parks effect for two-band superconductors, Low Temp. Phys. 34, 891 (2008).
- [11]A. Kuibarov, O. Suvorov, R. Vocatuoro, A. Fedorov, R. Lou, L. Merkwitz, V. Voroshin, J. I. Facio, K. Koepernik, A. Yaresko, G. Shipunov, S. Aswartham, J. van den Brink, B. Büchner, and S. Borisenko, Evidence of superconducting Fermi arcs, Nature 626, 294 (2024).DOI
- [12]S. Changdar, O. Suvorov, A. Kuibarov, S. Thirupathaiah, G. Shipunov, S. Aswartham, S. Wurmehl, I. Kauparkh, K. Koepernik, C. Timm, B. Büchner, I. C. Fulga, S. Borisenko, and J. van den Brink, Topological nodal i-wave superconductivity in PtBi2, Nature 647, 613 (2025).DOI
- [13]A. Veyrat et al., Berezinskii–Kosterlitz–Thouless transition in the type-I Weyl semimetal PtBi2, Nano Lett. 23, 1229 (2023).DOI
- [14]H. Waje, F. Jakubczyk, J. van den Brink, and C. Timm, Ginzburg–Landau theory for unconventional surface superconductivity in PtBi2, Phys. Rev. B 112, 144519 (2025).DOI
- [15]V. P. Mineev and K. V. Samokhin, Helical phases in superconductors, JETP 78, 401 (1994).
- [16]V. P. Mineev and K. V. Samokhin, Nonuniform states in noncentrosymmetric superconductors: Derivation of lifshitz invariants from microscopic theory, Phys. Rev. B 78, 144503 (2008).DOI
- [17]Y. A. Bychkov and E. L. Rashba, Properties of a 2d electron gas with lifted spectral degeneracy, JETP Lett. 39, 78 (1984).
- [18]V. M. Edelstein, The Ginzburg–Landau equation for superconductors of polar symmetry, J. Phys.: Condens. Matter 8, 339 (1996).DOI
- [19]F. Ando, Y. Miyasaka, T. Li, J. Ishizuka, T. Arakawa, Y. Shiota, T. Moriyama, Y. Yanase, and T. Ono, Superconducting diode effect, Nature 584, 373 (2020).
- [20]M. Nadeem, M. S. Fuhrer, and X. Wang, The superconducting diode effect, Nat. Rev. Phys. 5, 558 (2023).DOI
- [21]F. Qin, W. Shi, T. Ideue, M. Yoshida, A. Zak, R. Tenne, T. Kikitsu, D. Inoue, D. Hashizume, and Y. Iwasa, Superconductivity in a chiral nanotube, Nature Communications 8, 14465 (2017).DOI
- [22]N. Jiang, M. Maeda, Y. Yamaguchi, M. Watanabe, M. Tokuda, K. Takaki, S. Masaki, T. Ikushima, T. Koyanagi, M. Matsubara, K. Kudo, and Y. Niimi, Zero-field superconducting diode effect induced by magnetic flux in the van der Waals trigonal superconductor PtBi2, Phys. Rev. B 112, 25313 (2025).DOI
- [23]J. W. A. Robinson, A. V. Samokhvalov, and A. I. Buzdin, Chirality-controlled spontaneous currents in spin-orbit coupled superconducting rings, Phys. Rev. B 99, 180501(R) (2019).DOI
- [24]X. Xu, W. Qin, Y. Shen, Z. Huang, Z. Zhou, Z. Wang, and Y. Li, Signature of chiral superconducting order parameter evidenced in mesoscopic superconductors, Phys. Rev. Lett. 137, 086002 (2026).DOI
- [25]A. Vekris, J. C. Estrada Saldaña, J. de Bruijckere, S. Lorić, T. Kanne, M. Marnauza, D. Ol- steins, J. Nygård, and K. Grove-Rasmussen, Asymmetric Little-Parks oscillations in full shell double nanowires, Scientific Reports 11, 19034 (2021).
- [26]O. Dimitrova and M. V. Feigel’man, Theory of a two-dimensional superconductor with broken inversion symmetry, Phys. Rev. B 76, 014522 (2007).DOI
- [27]M. Smidman, M. B. Salamon, H. Q. Yuan, and D. F. Agterberg, Superconductivity and spin-orbit coupling in non-centrosymmetric materials: a review, Rep. Prog. Phys. 80, 036501 (2017).DOI
- [28]L. Powell, W. Huang, G. Hawkins-Pottier, R. Jalil, J. Birkbeck, Z. Jiang, M. Kim, Y. Zou, S. Komrakova, S. Haigh, I. Timokhin, G. Balakrishnan, A. K. Geim, N. Walet, A. Principi, and I. V. Grigorieva, Multiphase superconductivity in PdBi₂, Nat. Commun. 16, 291 (2025).DOI
- [29]J. A. Moreno, P. García Talavera, E. Herrera, S. López Valle, Z. Li, L.-L. Wang, S. Bud’ko, A. I. Buzdin, I. Guillamón, P. C. Canfield, and H. Suderow, Robust two-dimensional surface superconductivity and vortex lattice in the Weyl semimetal γ-PtBi₂, Phys. Rev. Lett. 137, 086001 (2026).
- [30]N. C. Koshinick, H. Bluhm, M. E. Huber, and K. A. Moler, Fluctuation superconductivity in mesoscopic aluminum rings, Science 318, 1440 (2007).
- [31]S. Schimmel, Y. Fasano, S. Hoffmann, J. Besproswanny, L. T. Corredor Bohorquez, J. Puig, B.-C. Elsahlem, B. Kalisky, G. Shipunov, D. Baumann, S. Aswartham, B. Büchner, and C. Hess, Surface superconductivity in the topological Weyl semimetal t-PtBi₂, Nat. Commun. 15, 9895 (2024).DOI
- [32]J. A. Bert, N. C. Koshnick, H. Bluhm, and K. A. Moler, Fluxoid fluctuations in mesoscopic superconducting rings, Phys. Rev. B 84, 134523 (2011).
- [33]The field-dependent STM/STS spectra of PtBi₂ reported in Ref. [31] provide additional motivation to investigate magnetic and electric effects on the surface superconductivity. A connection with the channel-mixing mechanism considered here remains to be established: those measurements used an out-of-plane field, and the authors attributed the apparent significant gap increase from ambient field to 5 T primarily to spatial inhomogeneity with a typical mesoscopic size comparable to that of the optimal rings considered here.
- [34]V. Jaccarino and M. Peter, Ultra-high-field superconductivity, Phys. Rev. Lett. 9, 290 (1962).DOI
- [35]F. V. Kusmartsev, A. Kusmartseva, J. C. M. Hughes, F. H. Alkallas, A. B. G. Trabelsi, Q. Chen, and K. Jin, Ginzburg-Landau theory of Rashba superconductors: Concave upper critical field, Phys. Rev. B 113, 179601 (2026).
- [36]M. Trama, V. Könyé, I. C. Fulga, and J. van den Brink, Self-consistent surface superconductivity in time-reversal symmetric Weyl semimetals, Phys. Rev. B 112, 064514 (2025).DOI
- [37]K. Mæland, M. Bahari, and B. Trauzettel, Phonon-mediated intrinsic topological superconductivity in Fermi arcs, Phys. Rev. B 112, 104507 (2025).DOI
- [38]R. Dsouza, N. Parthenios, B. M. Andersen, and M. H. Christensen, Kohn–Luttinger superconductivity of Weyl Fermi arcs in PtBi₂ (2026), arXiv:2605.31501 [cond-mat.supr-con].arxiv.org/abs/2605.31501
- [39]K. Mæland, G. Sangiovanni, and B. Trauzettel, Mechanism for nodal topological superconductivity on PtBi₂ surface, Phys. Rev. Lett. 137, 056001 (2026).DOI