betaFeedback

WorksPhysics and Astronomy

Multiple superconducting phases and order-parameter evolution in pressurized UTe2_2

Abstract

The recently discovered heavy-fermion spin-triplet superconductor candidate UTe2_2 provides a rich platform for unconventional pairing and topological phenomena. However, limited has been known about its superconducting order parameters and their evolution with control parameters, largely due to the lack of appropriate symmetry-sensitive detections. Here, we report comprehensive point-contact spectroscopy measurements of pressurized UTe2_2 on the (0~0~1) surface. The observation of Andreev bound states strongly suggests the presence of a pzp_z component in the superconducting order parameters. Quantitative analysis based on an extended Blonder-Tinkham-Klapwijk model unveils the superconducting order parameters with a finite odd-kzk_z component (e.g. B2uB_{2u} or B3uB_{3u}) for both ambient and pressurized UTe2_2. Remarkably, the multiple superconducting phases can be distinguished by a single parameter ⟨Δz⟩/⟨Δx(y)⟩\langle Δ_{z}\rangle/\langleΔ_{x(y)}\rangle, the relative weight between the pzp_z-wave and px(y)p_{x(y)}-wave pairings. These findings place stringent constraints on the pairing symmetry and provide essential spectroscopic signatures for distinguishing pressure-induced multiple superconducting phases in UTe2_2.

INTRODUCTION

The classification of superconductors is fundamentally governed by the pairing symmetry of their order parameters. Unlike spin-singlet superconductivity (SC) (total spin S=0S = 0, even parity), which includes both conventional ss-wave and many unconventional (e.g., dd-wave) cases, spin-triplet pairing (S=1S = 1, odd parity) where electrons with parallel spins form Cooper pairs represents a major frontier in condensed matter physics. This distinction is profound because pairing symmetry not only dictates the microscopic mechanism of SC, but also constraints on their physical properties and hence application prospects. Crucially, spin-triplet superconductors may host Majorana zero modes (a kind of non-Abelian quasiparticles localized at topological defects or edges) that constitute a critical resource for fault-tolerant topological quantum computation [1, 2, 3]. A central but often unresolved challenge, however, is definitive identification of the order parameter of spin-triplet superconductors (or candidates), many of which (e.g. Sr2_2RuO4_4, UPt3_3, etc) have been in long-standing controversy [4, 5, 6, 7]. In particular, analogous to the superfluid ³He [8, 9], the interplay between spin and orbital degrees of freedom can stabilize multiple superconducting phases under different conditions [10–13]. This renders the determination of SC order parameters even more challenging in that many routine experimental probes may lose their efficacy at the extremes of ultra-low temperature, high pressure (pp), or strong magnetic field (μ0H\mu_0 H).

UTe2_2, standing out as a promising candidate of spin-triplet superconductor, has attracted tremendous recent attention [14–22]. Albeit with a relatively low critical transition temperature TcT_c (≈2\approx2 K), it exhibits exotic superconducting properties, including upper critical fields far exceeding the Pauli limit along all principal axes [14], halo-structured reentrance of SC under ultra-high magnetic field [23–25], possible chirality and time-reversal symmetry breaking [18, 26, 27], pair density wave [28], strange metallicity [29, 30], etc. The pressure-induced phase diagram of UTe2_2 is intriguing, too. Three SC phases denoted by SC1-3 were reported under pressure [29, 31, 32]. Although the boundaries among these SC phases seem to be clear, whether these states are mutually independent, what their pairing symmetries are, and how their order parameters evolve with temperature and pressure remain fundamental open questions. Unfortunately, little is known about these issues due to the lack of symmetry-sensitive experiments under pressure to date [33].

A natural approach to addressing this challenge is to look into the gap structure and the associated in-gap Andreev bound state (ABS) on specific sample surface. As illustrated in Figure 1, for an ABS to form at a surface with specular reflection condition, there must be a phase change between the incident and reflected quasiparticle wavefunctions to incur a destructive interference. This can be realized if one or all components of the gap function Δ(k)\Delta(\mathbf{k}) changes sign upon going from k\mathbf{k} to k′\mathbf{k}'. Such an ABS can be detected through point-contact spectroscopy (PCS) a technique well-suited for extreme conditions by measuring the differential conductance (GNSG_{\mathrm{NS}}) with an appropriate geometric alignment between tunneling current I\mathbf{I} and inter-nodal scattering wavevector qN\mathbf{q}_{\mathrm{N}}, seeing Figure 1(c-Figure 1d). In this way, one assumes that the four possible spin-triplet pairings in UTe2_2 (cf. Table 1) can be distinguished, in part, because zero-energy ABS (ZABS) can show up on the (0 0 1) surface only in the B2uB_{2u} and B3uB_{3u} irreducible representations (IRs) whose order parameters contain the kzk_z component; AuA_u may also exhibit in-gap ABS on this surface, however, due to the fully-gapped dispersion, these ABSs are off-center.

Gap structure and Andreev bound state in unconventional superconductors

Figure 1. Gap structure and Andreev bound state (ABS) in unconventional superconductors. a A cartoon illustration of the reflection of Cooper pairs at the boundary (characterized by a barrier potential ZZ) of normal metal (NM) and superconductor (SC) that changes the scattering wavevector k=(kx,ky,kz)\mathbf{k}=(k_x,k_y,k_z) into k′=(kx,ky,−kz)\mathbf{k}'=(k_x,k_y,-k_z). If certain component of the SC gap function [Δ(k)][\Delta(\mathbf{k})] changes sign upon taking kzk_z to −kz-k_z, an ABS can be trapped near the boundary. b Nodal gap structure with maximum Δ0\Delta_0. The inter-nodal scattering is denoted by the wavevector qN\mathbf{q}_{\mathrm{N}}. c For tunneling current I⊥qN\mathbf{I}\perp\mathbf{q}_{\mathrm{N}}, the sign change in Δ(kz)→Δ(−kz)\Delta(k_z)\to\Delta(-k_z) leads to the formation of zero-energy ABS (ZABS) and hence a sharp zero-bias conductance peak (ZBCP) in the differential conductance (GNSG_{\mathrm{NS}}) that readily exceeds 2 in the large-ZZ limit. d Such a ZABS is absent if the tunneling current is along qN\mathbf{q}_{\mathrm{N}}. For an arbitrary I\mathbf{I} with a finite component perpendicular to qN\mathbf{q}_{\mathrm{N}}, ZABS persists, albeit with reduced intensity. e Sketch of a fully-gapped structure, e.g. AuA_u representation. f GNSG_{\mathrm{NS}} does not exhibit a ZABS in this case; instead, a pronounced dip is expected at zero bias at large ZZ.

SampleΔx(y)\Delta_{x(y)} (meV)Δz\Delta_z (meV)ZZΓ\Gamma (meV)
S11.18(7)0.72(4)1.67(9)0.39(2)
S21.15(1)0.62(1)0.76(1)0.40(1)
S31.15(1)0.60(1)1.12(1)0.16(1)

Table 1. Fitting parameters of PCS at 0.3 K to B2uB_{2u} (or B3uB_{3u}) representation, for samples S1-S3 at p=0p = 0 and B=0B = 0.

Herein, by systematic PCS investigations of UTe2_2 under hydrostatic pressure, signatures of ZABS are observed on the (0 0 1) surface, unambiguously suggesting the presence of pzp_z component in the gap function. Fitting the results to an extended Blonder-TinkhamKlapwijk (BTK) model, we further find that the SC pairings of UTe2_2 most likely fall in the B2uB_{2u} or B3uB_{3u} representation, and the different SC phases under pressure can be distinguished by a single parameter ⟨Δz⟩⟨Δx(y)⟩\frac{\langle\Delta_z\rangle}{\langle\Delta_{x(y)}\rangle}, the relative weight between the pzp_z-wave and px(y)p_{x(y)}-wave pairings. Our study therefore offers an important clue to pinpoint the pairing symmetry in this compound, and also provides crucial spectroscopic evidence for the existence of multiple SC phases therein.

RESULTS

PCS is a powerful probe for the density of states and hence the gap structure of superconducting pairing [34–36]. At the interface between a normal metal (NM) and a superconductor, an electron incident from the NM is either specularly reflected or retro-reflected as a hole by creating a Cooper pair in the superconductor, the latter of which is the well-known Andreev reflection, cf. Supplementary Figure 5 in Supplementary Information (SI). A schematic set-up for the PCS measurements of UTe2_2 under hydrostatic pressure is illustrated in Figure 2(a). To avoid Ag-Te reaction, we used Au epoxy instead. The contact points were located on the (0 0 1) surface, with the magnetic field applied along the a-axis. To verify the sample quality, the standard resistivity measurements were also conducted on the same sample, with electrical current along a. More details about experimental method can be found in the Methods section.

Point-contact spectroscopy of UTe₂ at ambient pressure

Figure 2. Point-contact spectroscopy (PCS) of UTe₂ at ambient pressure. a A sketch of point contact spectroscopy measurement under pressure. The junction is made on the (0 0 1) plane. b Phase diagram of UTe₂ constructed by a contour plot of ρ(p,T)\rho(p,T). The abbreviations are: SC superconductor, MO magnetic order. The dashed lines are reproduced from previous specific heat and magnetic susceptibility studies [29, 31, 32]. These dashed lines divide the SC region into three possible phases, SC1-3. c The differential conductance (GNSG_{\mathrm{NS}}) of sample S1 at various temperatures. The curves are shifted vertically for clarity. d-f GNSG_{\mathrm{NS}} at 0.3 K for samples S1, S2 and S3, respectively. The red lines are fittings of the experimental data (open circles) to the extended BTK model in B2uB_{2u} (or B3uB_{3u}) representation. The obtained fitting parameters are summarized in Supplementary Table 1.

Schematics of Andreev reflection at the interface of normal metal and superconductor

Supplementary Figure 5. Schematics of Andreev reflection at the interface of normal metal (NM) and superconductor (SC). a Sketch of the NM-SC junction. In this work, the junction is made by Au-UTe2_2, on the x-y crystallographic plane of UTe2_2. b Andreev reflection (red arrows, aαβa_{\alpha\beta}) for spin-triplet SC. For an electron (blue circle) incident at the interface, a hole (open circle) is retro-reflected by creating a Cooper pair in the SC. Note that the spin state of the reflected hole relies on the specific configuration of the Cooper pair. The green arrows denote the normal specular reflection (bαβb_{\alpha\beta}) that preserves spin degree of freedom.

Supplementary Figure 1(a) shows that the ρ(T)\rho(T) profile of our sample S1 at ambient pressure is in good agreement with previously reported results for ultra-clean single crystals grown by the molten soft flux method [37, 38], exhibiting a high residual resistivity ratio (RRR) up to 410 and high superconducting transition temperatures Tco=2.1T_{\mathrm{c}}^{\mathrm{o}} = 2.1 K and Tcon=2.6T_{\mathrm{c}}^{\mathrm{on}} = 2.6 K. The low-TT part of ρ(T)\rho(T) under various pressures are displayed in Supplementary Fig. 1(b-c), and a phase diagram is constructed by the false-color contour plot of ρ(p,T)\rho(p,T) in Figure 2(b). At low pressure in the regime denoted SC1 (p<0.3p < 0.3 GPa), TcT_{\mathrm{c}} is slightly suppressed with increasing pressure. Beyond this threshold, the resistive TcT_{\mathrm{c}} turns up, and this yields a dome structure in the intermediate pressure regime 0.3<p≤1.40.3 < p \leq1.4 GPa; however, earlier bulk measurements including AC calorimetry and AC susceptibility measurements both revealed a dual-transition regime: a high-temperature phase (SC2) and a low-temperature phase (SC3) [29, 31, 32]. The boundaries among these SC phases are depicted by the dashed lines. For pressure above 1.4 GPa, two anomalies denoted by Tm1T_{\mathrm{m1}} and Tm2T_{\mathrm{m2}} appear just above TcT_{\mathrm{c}}, characteristic of the magnetically ordered phases MO1 and MO2, respectively [see Supplementary Figure 1]. Tm1T_{\mathrm{m1}} increases rapidly with pressure, while Tm2T_{\mathrm{m2}} is weakly pressure dependent. Overall, the as-obtained phase diagram resembles those reported in the literature [29, 32, 33].

Supplementary Figure 1. Supplementary Figure 1 | Resistivity data of UTe2_2 (sample S1) under pressure. Measured with electrical current I∥a\mathbf{I} \parallel a. a Main frame, electrical resistivity (ρ\rho) as a function of TT at ambient pressure. Inset, the zoom-in plot to show the superconducting transition at low temperature. The sample is of good quality with residual resistance ratio (RRR) ∼410\sim410, and SC transition temperatures Tcon=2.6T_{\mathrm{c}}^{\mathrm{on}} = 2.6 K and Tc0=2.1T_{\mathrm{c}}^{0} = 2.1 K. b and c ρ(T)\rho(T) profiles for under pressure. The curves are vertically shifted for clarity. The black triangles and diamonds denote the characteristic temperatures (Tm1T_{\mathrm{m1}} and Tm2T_{\mathrm{m2}}) of the two magnetically ordered phases MO1 and MO2, respectively.

Turning now to the PCS results of UTe2\mathrm{UTe}_{2}. We start with the data at ambient pressure, as shown in Figure 2(c). The normal-state spectrum is nearly flat. At 2.5 K, just below TconT_{\mathrm{c}}^{\mathrm{on}}, a slight hump appears in the differential conductance spectrum. As temperature decreases further, the hump stabilizes, forming a zero-bias conductance peak (ZBCP). Before relating such ZBCP to the feature of unconventional SC states [39], it is necessary to rule out other extrinsic origins. First, spurious ZBCP can be caused by heating effect in the thermal regime [40–42], in the regard that an increasing bias voltage causes the local current to exceed the critical current, and consequently GNSG_{\mathrm{NS}} decreases by forming a peak centering about V=0V = 0. In this case, it is speculated that the amplitude of the peak shall be nearly temperature independent, while the width of the peak shall shrink rapidly with magnetic field. Here, we rule out this possibility because the shape of the peak degrades gradually with increasing TT, and an application of a small field (<2< 2 T) only weakly affects its width, as shown in Supplementary Figure 2. More straightforwardly, the ratio l/d∼8l/d \sim8 can be estimated [43–45], strongly pointing to a ballistic regime in our measurements [38, 41, 46] (where ll is mean free path, and dd is effective microscopic contact diameter). Second, magnetic scattering may also induce conductance peaks at zero bias [47, 48]. However, this is not relevant here because UTe2\mathrm{UTe}_{2} does not exhibit magnetic ordering at ambient pressure [14]. In addition, ZBCP are also observed in different samples (S2 and S3) with various contact sizes and resistances, seeing Figure 2(e,Figure 2f). The presence of stable ZBCP in various samples strongly suggests the existence of zero-energy Andreev bound state, which further implies a high probability of gap function with odd kzk_z component.

Supplementary Figure 2. Additional Point-contact Andreev reflection spectroscopy of UTe2_2 at ambient pressure. a Isothermal GNS(V)G_{\mathrm{NS}}(V) under different magnetic field for fixed temperature T=0.5T = 0.5 K. b Representative fitting of the dI/dVdI/dV spectrum to the extended Blonder-Tinkham-Klapwijk (BTK) model, taking T=0.3T = 0.3 K as an example. c The obtained gap parameters Δx(y)\Delta_{x(y)} and Δz\Delta_z as functions of TT. They apparently deviates from the conventional Bardeen-Cooper-Schrieffer (BCS) behavior, strongly suggesting unconventional superconductivity.

The results of PCS under pressure (measured with sample S1) are displayed in Figure 3 and Supplementary Figure 3. At low pressure (e.g. 0.2 GPa), the spectrum remains largely similar to that at ambient pressure, cf. Figure 3(a), indicating that the SC order parameter likely remains unchanged. Remarkable differences are seen above the first critical pressure ∼0.3\sim0.3 GPa. As shown in Figure 3(b), at 0.6 GPa, besides the ZBCP that is present for low pressure, dips appear on both sides of the peak at low temperature. These dips become more pronounced upon cooling, accompanied by the appearance of additional humps at slightly higher bias. As we will further explain below, the appearance of these side dips suggests that these ZBCPs are intrinsic and associated with the formation of ZABS. Further increasing pressure to 1.0 GPa, the main features of the spectrum remain the same. However, a new minor peak appears inside the dip at low temperature, forming a double-dip structure, as can be seen in Figure 3(c). Similar behavior can also be identified at 1.2 GPa [Supplementary Figure 3]. Finally, at 1.4 GPa that is on the verge between SC and magnetically ordered phases, the central peak is substantially broadened with its top being nearly flat. Side-dip structure is also present at this pressure. To better visualize the evolution of PCS with pressure, we present contour plots of GNS(V,T)G_{\mathrm{NS}}(V,T) in the bottom panels of Figure 3.

Figure 3. Point-contact spectroscopy (PCS) of UTe2_2 under pressure. Upper panels, differential conductance (GNSG_{\mathrm{NS}}) at selected pressures, a 0.2 GPa; b 0.6 GPa; c 1.0 GPa; and d 1.4 GPa. The lower panels display the false-color contour plots of GNS(V,T)G_{\mathrm{NS}}(V,T).

Supplementary Figure 3. Additional Point-contact Andreev reflection spectroscopy data of UTe₂ under pressure. The differential conductance (GNSG_{\mathrm{NS}}) at various temperatures. a 0.3 GPa. b 0.8 GPa. c 1.2 GPa. The curves are shifted vertically for clarity. Note that these data were collected after releasing 1.4 GPa.

Discussion

To elucidate the evolution of the PCS of UTe2_2 with applied pressure, we analyzed the data in two steps. First, we theoretically simulated the GNS(V)G_{\mathrm{NS}}(V) in the framework of an extended BTK model, aiming to qualitatively screen out the potential pairing symmetry. After that, numerical fittings were conducted to extract the parameters such as SC gap Δ\Delta, the Dynes broadening factor Γ\Gamma, and the junction barrier strength ZZ. In accordance with the experimental setup, we consider a planar NM-SC junction whose interface is perpendicular to the crystallographic c-axis of UTe2_2. For spin triplet SCs, the Andreev reflection is a more complex process as compared with that for conventional spin-singlet ones, in the sense that the reflection can take place through multiple channels, Φ1=∣↑↑⟩\Phi_1 = |\uparrow\uparrow\rangle, Φ0=(∣↑↓⟩+∣↓↑⟩)/2\Phi_0 = (|\uparrow\downarrow\rangle+ |\downarrow\uparrow\rangle)/\sqrt{2}, and Φ−1=∣↓↓⟩\Phi_{-1} = |\downarrow\downarrow\rangle. A cartoon illustration is presented in Supplementary Figure 5.

In short, the differential conductance for PCS takes the following form,

GNS(V,T)=G0∫−∞+∞∂f(E−eV)∂V∑k∥{1+12∑α,β[∣aαβ(E+iΓ)∣2−∣bαβ(E+iΓ)∣2]} dE(1)G_{\mathrm{NS}}(V,T)=G_0\int_{-\infty}^{+\infty}\frac{\partial f(E-eV)}{\partial V}\sum_{\mathbf{k}\parallel}\left\{1+\frac{1}{2}\sum_{\alpha,\beta}\left[|a_{\alpha\beta}(E+i\Gamma)|^2-|b_{\alpha\beta}(E+i\Gamma)|^2\right]\right\}\,dE \tag*{(1)}

where G0G_0 represents the conductance of the junction in the normal state; aαβa_{\alpha\beta} and bαβb_{\alpha\beta} are functions of ZZ, Γ\Gamma and Δ\Delta, standing for the Andreev reflection and normal reflection coefficients, respectively; α\alpha and β\beta denote the spin states. Details of the calculation are provided in the SI. We first only consider the representative cases with pure pxp_x, pyp_y and pzp_z-wave gap functions. The numerical results are displayed in Fig. 4 and Supplementary Figure 6. In the limit of weak junction barrier (i.e. small ZZ) of our planar junction, the conductance spectra of all three states are essentially the same. The differential conductance exhibits a peak whose value is obviously less than 2 regardless of the choice of other parameters in Fig. 4(b) (i.e. Γ\Gamma, TT and Δ\Delta). This is similar to the behavior of other unconventional pairings, such as the dx2−y2d_{x^{2}-y^{2}}-wave pairing in the cases of various junction surface orientations [39, 49]. As ZZ increases, the conductance of the px(y)p_{x(y)}- and pzp_z-wave pairings behave in fundamentally different manners. For the px(y)p_{x(y)} pairing, the conductance peak diminishes with increasing ZZ and eventually forms a gapped spectrum at large ZZ. In contrast, the peak for the pzp_z-wave sharpens as ZZ increases, reaching a value that readily exceeds 2 at large ZZ. These pronounced ZBCPs originate from the zero-energy Andreev bound states at the junction interface (Figure 1), and are robust against FS anisotropy (Supplementary Figure 14) and multiplicity [50]. Their formation is closely related to the sign change of the gap function Δk\Delta_{\mathbf{k}} between the incident wavevector k=(kx,ky,kz)\mathbf{k} = (k_x, k_y, k_z) and the reflected quasiparticles, and thus allows for localized sub-gap states to develop. No such sign change occurs for pure pxp_x and pyp_y pairings, and hence no bound state formation therein. Notably, bound states can still form when pzp_z and px(y)p_{x(y)} pairings mix, as long as the sign-changing pzp_z component is finite. However, depending on the details of mixing, the sub-gap conductance may exhibit more complicated structures, as can be seen in e.g. Supplementary Figure 7. As a side effect, at large ZZ the presence of surface bound states tends to deplete the tunneling probability of the bulk continuum, which may then induce conductance side dips around the bias voltage V=±∣Δ∣V = \pm|\Delta|. These side dips, however, are not robust against pair-breaking scattering and may be smeared if the Dynes broadening Γ\Gamma is large, seeing below.

Supplementary Figure 6. Simulated PCS at different temperatures. The NM-SC junction is on the x-y facet. a px(y)p_{x(y)}, Z=0.2Z = 0.2, Γ/Δ=0.2\Gamma/\Delta= 0.2. b px(y)p_{x(y)}, Z=1Z = 1, Γ/Δ=0.2\Gamma/\Delta= 0.2. c pzp_z, Z=0.2Z = 0.2, Γ/Δ=0.2\Gamma/\Delta= 0.2. d pzp_z, Z=1Z = 1, Γ/Δ=0.2\Gamma/\Delta= 0.2. It is found that the behavior of ZBCP with a side-dip is present in pzp_z-wave pairing but absent in px(y)p_{x(y)}-wave pairing symmetry.

Simulated PCS for various irreducible representations

Supplementary Figure 7. Simulated PCS for various irreducible representations. a AuA_u. b B1uB_{1u}. c-d B2uB_{2u} or B3uB_{3u}. The simulations are made at T=0.3T = 0.3 K. Note that for the junction within the x-y plane, the results of B2uB_{2u} and B3uB_{3u} are the same. For AuA_u, a zero-bias dip appears for all Z>0Z > 0, and a side dip appears for large ZZ. For B1uB_{1u}, a zero-bias dip appears for large ZZ, but no side dip can be obtained. For B2uB_{2u} and B3uB_{3u}, the ZBCP persists for all ZZ, and a dip on each side appears for large ZZ. As Γ\Gamma decreases, the shoulder may evolves into a satellite peak.

Image not extracted

Supplementary Figure 14. Simulations of point-contact spectra with varying Fermi surface anisotropy. The ratio rv≡vFc/vFabr_v \equiv v_F^c/v_F^{ab}, with vFabv_F^{ab} and vFcv_F^c being the in-plane and out-of-plane Fermi velocities, respectively.

We then consider pairings in various IRs as proposed for UTe2_2 (see Table 1) [16, 51–53], with representative results displayed in Supplementary Figure 7. Firstly, the conductance peak of a generic AuA_u pairing typically splits at finite $Z [{xref}Supplementary Fig. 7(a)], unless $\Gamma$ is sufficiently large to smear out the splitting [{xref}Supplementary Fig. 8]. Such behavior does not agree with the definitive ZBCPs seen in our samples [{xref}Fig. 2(d-f)]. Secondly, since the $B_{1u}$ pairing lacks a component that changes sign upon taking $k_z$ to $-k_z$, it cannot exhibit conductance side dips at any values of $Z$ and $\Gamma$; this situation cannot be improved even after taking into account the high-order $k_xk_yk_z$ term in $B_{1u}$ ({xref}Supplementary Fig. 13). Lastly, the $B_{2u}$ and $B_{3u}$ pairings are indistinguishable in this junction, and they both can qualitatively capture all main features of the experimental spectra in {xref}Figs. 2 and 3, and {xref}Supplementary`

Fig. 3.

With these in mind, we then compare and fit the experimental spectra mainly in the framework of B2uB_{2u} (or B3uB_{3u}). The fittings for the ambient-pressure data for three samples are provided in Figure 2(d-Figure 2f), and the obtained parameters are listed in Supplementary Tab. 1. Although the three samples exhibit distinct profiles of PCS, a common feature is that the peak values of the GNS(V)G_{\mathrm{NS}}(V) are all less than 1.2, significantly lower than 2. This suggests that the px(y)p_{x(y)} pairing component plays a more dominant role than the pzp_z pairing. Albeit of the different values of ZZ and Γ\Gamma obtained considering junction variation, all three samples yield a similar set of gap amplitudes Δx(y)\Delta_{x(y)} and Δz\Delta_z with the ratio Δz/Δx(y)≈0.57(5)\Delta_z/\Delta_{x(y)} \approx0.57(5), which bolsters our confidence in these fittings. The temperature dependence of Δx(y)\Delta_{x(y)} and Δz\Delta_z of sample S1 at ambient pressure is presented in Supplementary Figure 2; they both strongly violate the BCS’s prediction, which again demonstrates the unconventional nature of SC in UTe2_2. We note that the gap amplitudes extracted from the BTK fittings may not necessarily coincide strictly with the thermodynamic gap values obtained from other probes; for instance, in RbCr3_3As3_3, another candidate of spin-triplet superconductor (Tc∼7T_{\mathrm{c}} \sim7 K), while PCS revealed SC gap as large as ∼2.5\sim2.5 meV, the STM measurements reported only a small gap ∼0.8\sim0.8 meV [54]. The inherent mechanism for such discrepancy is not clear and requires further investigations. Nevertheless, since our PCS measurements are performed in the ballistic regime, these fitted amplitudes provide effective spectroscopic energy scales for characterizing the main spectral features and tracking the pressure evolution of the relative pairing components. It is also worthwhile to mention here that an earlier PCS experiments on (0 0 1) and (0.4 0.6 0.7) facets at ambient condition suggested dominant-pyp_y wave gap function [55]; a recent zero-energy surface state visualization by STM measurements on (0 −1 1) facet most favors B3uB_{3u} IR [53]; their conclusions are qualitatively consistent with ours for ambient pressure.

Figure 5 summarizes the fitting in different regions of the phase diagram. Representative fitting curves are displayed in Supplementary Figure 9-Supplementary Figure 11. For small pressure p=0.2p = 0.2 GPa, the results are quite similar to those at ambient pressure, and the px(y)p_{x(y)} pairing is still found to dominate with Δz/Δx(y)≈0.64(2)\Delta_z/\Delta_{x(y)} \approx0.64(2). For p=0.6p = 0.6 GPa, however, the amplitude of ZBCP increases to ∼1.70\sim1.70, and most importantly, dips emerge symmetrically on the two sides of the central peak. These dips can only be reproduced by increasing the weight of pzp_z component, as shown in Figure 5(c). The fitting yields Δz/Δx(y)≈1.03\Delta_z/\Delta_{x(y)} \approx1.03. The situation at 0.8 GPa is similar to that at 0.6 GPa, and the amplitude of the ZBCP is further increased to 2.08, cf Figure 5(d).

Blonder-Tinkham-Klapwijk fittings of the Point-contact spectroscopy in different regions

Figure 5. Blonder-Tinkham-Klapwijk fittings of the Point-contact spectroscopy (PCS) in different regions. a Schematic phase diagram of UTe2UTe_2 under pressure. The abbreviations are: NM normal metal, SC superconductor, MO magnetic order. The boundaries of the phase diagram are generally reproduced from Figure 2(b). The dashed line is a guide to eyes to mark the transition between SC2 and SC3 determined by earlier AC calorimetry and AC magnetic susceptibility measurements [29, 32]. The red open circles represent the critical temperatures defined as the minimum of iso-pressure temperature dependent ⟨Δz⟩/⟨Δx(y)⟩\langle\Delta_z\rangle/\langle\Delta_{x(y)}\rangle, as exemplified by the results at 0.8 GPa in panel f. The inset presents ⟨Δz⟩/⟨Δx(y)⟩\langle\Delta_z\rangle/\langle\Delta_{x(y)}\rangle at 0.3 K as a function of pp, which reveals an abrupt increase in the weight of pzp_z-wave pairing from SC1 to SC3. b Fitting for p=0.2p = 0.2 GPa and T=0.3T = 0.3 K. c Fitting for p=0.6p = 0.6 GPa and T=0.3T = 0.3 K. d Fitting for p=0.8p = 0.8 GPa and T=0.3T = 0.3 K. e Fitting for p=1.2p = 1.2 GPa and T=0.3T = 0.3 K. f ⟨Δz⟩/⟨Δx(y)⟩\langle\Delta_z\rangle/\langle\Delta_{x(y)}\rangle as a function of TT at fixed pressure p=0.8p = 0.8 GPa.

Six panels showing fits of ambient-pressure PCS data at different temperatures

Supplementary Figure 9. Supplementary Figure 9 | Fitting of the ambient-pressure PCS for sample S1. a 0.3 K. b 0.5 K. c 0.7 K. d 1.1 K. e 1.5 K. f 1.85 K. The fittings were carried out with B2uB_{2u} (or B3uB_{3u}). The fitting parameters are listed in each panels.

Image not extracted

Supplementary Figure 11. Fitting of the point-contact spectra at 1.2 GPa for sample S1. Taking the data of 0.3 K as an example. It is found that the single-channel BTK model can not capture the double-side-dip feature at this pressure.

At even higher pressure, the standard fitting procedure can hardly capture the double-side-dip feature (Supplementary Figure 11). The exact origin of the double side dips remains elusive. One may conjecture that it arises from multiple gaps. An alternative possibility is due to pressure inhomogeneity, in the regard that the NM-SC junction may cover several SC domains with different gap parameters. For both cases, a linear superposition model to the total differential conductance (GNStotalG_{\mathrm{NS}}^{\mathrm{total}}) are applicable:

GNStotal(V)=∑iwiGNSi(V),(2)G_{\mathrm{NS}}^{\mathrm{total}}(V) = \sum_i w_i G_{\mathrm{NS}}^i(V), \tag*{(2)}

where GNSi(V)G_{\mathrm{NS}}^i(V) is the differential conductance for each channel, and wi∈[0,1]w_i \in[0,1] characterizes the fractional contribution. This model can fit the multi-dip behavior rather well as shown in Figure 5(e). Although this model does not account for the complex inter-band couplings inherent to a rigorous multi-band framework, it provides an effective estimate to the energy scale of SC order parameters [36, 56]. The ratio between the pzp_z and px(y)p_{x(y)} components is thus defined as

⟨Δz⟩⟨Δx(y)⟩≡∑iwiΔzi∑iwiΔx(y)i,(3)\frac{\langle\Delta_z\rangle}{\langle\Delta_{x(y)}\rangle} \equiv\frac{\sum_i w_i\Delta_z^i}{\sum_i w_i\Delta_{x(y)}^i}, \tag*{(3)}

the pressure dependence of which at 0.3 K is presented in the inset of Figure 5(a). It is clear that the weight of pzp_z wave pairing increases drastically going from the SC1 to SC3, and becomes dominant in the latter region. For p=1.2p = 1.2 GPa, ⟨Δz⟩⟨Δx(y)⟩\frac{\langle\Delta_z\rangle}{\langle\Delta_{x(y)}\rangle} reaches ∼1.30\sim1.30.

To gain insight into the transition from SC2 to SC3, we analyzed the temperature dependent PCS at fixed p=0.8p = 0.8 GPa, as shown in Figure 5(f). Remarkably, with decreasing temperature, the ratio ⟨Δz⟩⟨Δx(y)⟩\frac{\langle\Delta_z\rangle}{\langle\Delta_{x(y)}\rangle} first reduces, and then increases, with a minimum near 1.5 K. The results for p=0.6p = 0.6 and 1.2 GPa are supplemented in Supplementary Figure 12, where similar temperature dependence can be observed, with minima occurring at about 1.6 and 0.7 K, respectively. In contrast, this ratio is nearly temperature-independent at p=0p = 0 [Supplementary Figure 12]. It is particularly interesting to note that the critical temperatures defined in association with the minimum of ⟨Δz⟩⟨Δx(y)⟩\frac{\langle\Delta_z\rangle}{\langle\Delta_{x(y)}\rangle} match well with the boundary between SC2-SC3 that was determined through AC calorimetry and AC magnetic susceptibility [29, 32], as shown in Figure 5(a) (red open circles). The peculiar non-monotonic temperature dependence of ⟨Δz⟩⟨Δx(y)⟩\frac{\langle\Delta_z\rangle}{\langle\Delta_{x(y)}\rangle} is more likely indicative of an underlying phase transition between the SC2 and SC3 states than a mere crossover.

Image not extracted

Supplementary Figure 12. The obtained superconducting gap parameters for sample S1. a-c Δα\Delta_{\alpha} [α=x(y),z\alpha= x(y), z] as a function of temperature for p=0p = 0, 0.60.6 and 1.21.2 GPa. d-f ⟨Δz⟩/⟨Δx(y)⟩\langle\Delta_z\rangle/\langle\Delta_{x(y)}\rangle.

Additional remarks in order:

(1) Is AuA_u pairing possible for ambient-pressure UTe2_2? While B2uB_{2u} (or B3uB_{3u}) model provides a compelling fit to our data, we should emphasize that our results cannot entirely exclude this possibility. In fact, AuA_u was supported by an ambient-pressure 125^{125}Te NMR experiment where Knight shift - proportional to the spin susceptibility - was found to decrease below TcT_c for magnetic field along all principal axes [51], as well as thermal conductivity measurements that pointed to a fully gapped pairing state [57]. These works suggested that the SC order parameter d\mathbf{d} may contain non-negligible components along all x^\hat{\mathbf{x}}, y^\hat{\mathbf{y}} and z^\hat{\mathbf{z}} axes. However, we should also note that the change of Knight shift is rather anisotropic, meaning that a certain component of d\mathbf{d} may be rather small. If either the x^\hat{\mathbf{x}} or y^\hat{\mathbf{y}} component is negligible, the tunneling spectrum will be similar to those of B2uB_{2u} and B3uB_{3u} states. In this case, PCS lacks the necessary resolution to tell them apart.

(2) What can the order parameters be for pressurized UTe2_2? From our PCS measurements, it is fairly likely that SC1-3 are mutually independent phases. First of all, SC3 differs from SC1 strikingly in that the former is pzp_z-dominant while the latter is more dominated by px(y)p_{x(y)} pairing. The peculiar temperature dependence of ⟨Δz⟩⟨Δx(y)⟩\frac{\langle\Delta_z\rangle}{\langle\Delta_{x(y)}\rangle} also hints a phase transition taking place at the boundary between SC2-SC3. It should be pointed out that an earlier 125^{125}Te NMR experiment at 1.2 GPa [33] manifested that the Knight shift remains constant when it enters the SC2 phase from the normal state, but starts to drop at the boundary of SC2-SC3. This indicates that the y^\hat{\mathbf{y}}-component of d\mathbf{d} should be vanishingly small for SC2 but considerable for SC3. In this sense, and also taking into account our PCS experiment, B2uB_{2u} appears the most promising for SC2. Below the SC2-SC3 boundary, a secondary SC order parameter starts to condense, probably yielding a superposition of B2uB_{2u} and B3uB_{3u} pairings in SC3. If so, the transition at SC2-SC3 would be accompanied with additional symmetry breaking. However, we must admit that before a more systematic PCS investigation is made on other surfaces of UTe2_2, it is premature to draw a precise conclusion.

(3) Does a quantum critical region exist in UTe2_2 under pressure? Notably, the conductance spectrum near 1.4 GPa exhibits an anomalous broadening with a nearly flat top, making it challenging to adequately describe within a simple single-component BTK framework. This drastic spectral evolution may suggest an underlying change in the system’s ground state, rather than a mere enlargement of the superconducting gap. Given that previous pressure studies have reported a possible magnetic quantum critical point (QCP) nearby [29], we speculate that the anomalously broadened spectrum could be related to this criticality. Specifically, the proliferated critical fluctuations can lead to a significant enhancement of scattering rate, the latter of which can increase the Dynes broadening factor Γ\Gamma. While this scenario offers a plausible explanation for the 1.4 GPa data, we cautiously note that further experimental evidence is required to draw a definitive conclusion regarding its connection to a QCP.

In summary, we systematically investigated the pressure-evolution of point-contact spectroscopy in UTe2_2. Clear and reproducible ZBCP is observed on the (0 0 1) surface, and the profile of the ZBCP is strongly pressure dependent with pronounced side-dip structure appearing for pressures above 0.3 GPa, likely indicating multiple SC phases whose pairing gap functions contain kzk_z-odd components. Quantitative analysis based on an extended BTK model suggests B2uB_{2u} (or B3uB_{3u}) as the most conceivable irreducible representation for both ambient and pressurized cases, with the ratio ⟨Δz⟩⟨Δx(y)⟩\frac{\langle\Delta_z\rangle}{\langle\Delta_{x(y)}\rangle} serving as a useful parameter to differentiate the SC1-3 phases. It is found that the weight of the pzp_z-wave component increases drastically from SC1 to SC3, and the SC2-SC3 transition may involve the emergence of an additional superconducting order-parameter component. Our work not only provides spectroscopic evidence for the existence of multiple SC phases in pressurized UTe2_2, but also places critical constraints on its superconducting order parameter, thus collectively establishing UTe2_2 as a prototypical system for studying intertwined electronic orders upon pressure tuning.

Methods

Single crystals preparation and characterization

High-quality single crystals of UTe2_2 were grown by a molten soft flux method as described elsewhere [37, 38]. We employed a mixture of NaCl and KCl as a flux. High-purity U (99.999%), Te (99.9999%), NaCl (99.99%) and KCl (99.999%) were mixed in a molar ratio of 1:1.71:60:60. To prevent contamination or oxidation of the starting materials, the entire process was carried out in an Ar-filled glove box. The mixture was transferred into a carbon crucible and sealed in quartz tube which was then heated up to 473 K and dwelt for 12 hours. The temperature was then raised to 723 K723\ \mathrm{K} in 24 hours, held at this temperature for an additional 24 hours, and subsequently increased to 1223 K1223\ \mathrm{K} at a rate of 0.35 K/min0.35\ \mathrm{K}/\mathrm{min} and maintained for another 24 hours. After that, the temperature was gradually reduced to 923 K923\ \mathrm{K} at a rate of 0.03 K/min0.03\ \mathrm{K}/\mathrm{min}, kept stable for 24 hours, and finally cooled to room temperature over the next 24 hours. This controlled thermal profile ensured the synthesis of high-quality crystalline materials.

Measurements under pressure

Electrical transport measurements of UTe2\mathrm{UTe}_{2} (sample S1) were performed in an 8 T superconducting magnet system, using a 3He{}^{3}\mathrm{He} refrigerator (Oxford Instruments). Prior to junction fabrication, the surfaces of the sample were fine-polished to obtain a fresh crystallographic (0 0 1)(0\ 0\ 1) plane. A soft point-contact junction was then prepared by depositing a minimal droplet of Au epoxy (54L-2210) to glue the gold wires onto the fresh surface. The small Au particles in the epoxy are solutes in the organic suspension, enabling the construction of a point contact with a small effective contact area. The Au epoxy was heated at 423 K423\ \mathrm{K} for 30 minutes before the sample was transferred into a pressure cell. All these processes were conducted in an argon-filled glove box. A hybrid piston-clamp type cell was used to achieve quasi-hydrostatic pressures up to ∼1.8 GPa\sim1.8\ \mathrm{GPa} with Daphne 7373 oil serving as the pressure-transmitting medium. This rigorous sample preparation protocol preserves the intrinsic surface integrity of UTe2\mathrm{UTe}_{2} and provides a stable electrical interface that is essential for the accurate probing of quasi-particle dynamics. The pressure sequence was as follows: 0→0.2→0.6→1.0→1.4→0.3→0.8→1.2 GPa0 \to0.2 \to0.6 \to1.0 \to1.4 \to0.3 \to0.8 \to1.2\ \mathrm{GPa}. To check the reproducibility of the results, two additional samples (S2 and S3) were measured at ambient pressure, and pressure-cycling test was performed on sample S1.

Data availability

All data that support the findings of this study are available within the paper and the Supplementary Information. Raw data generated during the current study are available from the corresponding authors upon request.

References

ACKNOWLEDGEMENTS

A portion of this work was carried out at the Synergetic Extreme Condition User Facility (SECUF, https://cstr.cn/31123.02.SECUF). A.G.E. acknowledges support from the Henry Royce Institute for Advanced Materials through the Equipment Access Scheme enabling access to the Advanced Materials Characterization Suite at Cambridge (EP/P024947/1, EP/M000524/1, and EP/R00661X/1).

FUNDING STATEMENT

This work is supported by the National Key R&D Program of China (2023YFA1609600, 2022YFA1602602, and 2022YFA1402200), National Natural Science Foundation of China (U23A20580, 12374042, 12574147, 52588101, W2511006 and 12494592), Beijing National Laboratory for Condensed Matter Physics (2024BNLCMPKF004), and the New Cornerstone Science Foundation (Grants No. NCI202509). W.H. acknowledges support from a startup fund at Great Bay University and the Dongguan Key Laboratory of Artificial Intelligence Design for Advanced Materials.

AUTHOR CONTRIBUTIONS

Y.L., X.L. and H.Y. conceived and designed the experiments; Z.W., A.C., M.V., N.L., B.Z., A.G.E. and C.Z. grew and characterized high-quality single crystalline samples; S.Z., F.S. performed most of the experiments with the aids from Y.Z., W.Q., Z.W., H.Z., Y.M., Z.Y., K.Y., J.Z., Y.Z., K.L., G.L. and J.L; S.Z. and Z.Q. carried out numerical simulations and fittings with the aids from J.L.Z. and W.H; S.Z., Z.Q., W.H., X.L., H.Y. and Y.L. discussed the data, interpreted the results, and wrote the paper with input from all the authors.

COMPETING INTERESTS

The authors declare no competing interests.

ADDITIONAL INFORMATION

Supplementary information The online version contains supplementary material available at ********.

Correspondence and requests for materials should be addressed to Wen Huang, Huiqiu Yuan, Xin Lu or Yongkang Luo.

Reprints and permission information is available at http:\www.nature.com/reprints.

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

IRBasis functionsSpin componentZABS on (0 0 1)(0\ 0\ 1)?
AuA_{u}Δxkxx^+Δykyy^+Δzkzz^\Delta_{x}k_{x}\hat{x}+\Delta_{y}k_{y}\hat{y}+\Delta_{z}k_{z}\hat{z}No
B1uB_{1u}Δykyx^+Δxkxy^\Delta_{y}k_{y}\hat{x}+\Delta_{x}k_{x}\hat{y}z^\hat{z}No
B2uB_{2u}Δzkzx^+Δxkxz^\Delta_{z}k_{z}\hat{x}+\Delta_{x}k_{x}\hat{z}y^\hat{y}Yes
B3uB_{3u}Δzkzy^+Δykyz^\Delta_{z}k_{z}\hat{y}+\Delta_{y}k_{y}\hat{z}x^\hat{x}Yes

Table I. Possible pairing symmetries and the corresponding forms of the gap functions of UTe2_2 classified according to the D2h_{2h} point group [16, 51–53]. Note that we here ignore high-order terms beyond linear order of kik_i (i=x,y,zi=x,y,z), seeing more details in Supplementary Figure 13. Zero-energy Andreev bound state (ZABS) are expected on the (0 0 1)(0\ 0\ 1) surface for B2uB_{2u} or B3uB_{3u} representation.

Figure 4. Simulation of Point-contact Andreev reflection spectroscopy based on the extended Blonder-Tinkham-Klapwijk (BTK) model. The abbreviations are: NM normal metal, SC superconductor. a-b px(y)p_{x(y)} pairing symmetry. c-d pzp_z pairing symmetry. The simulations were performed with Δ=1\Delta= 1 meV, kBT/Δ=0.01k_BT/\Delta= 0.01, Γ/Δ=0.2\Gamma/\Delta= 0.2, and variable ZZ listed in the figures.

SI I: Resistivity results of UTe2_2 under pressure

Additional PCS results at ambient pressure

Additional PCS results under pressure

Figure 4 (PDF p. 28). Isothermal GNS(V)G_{\mathrm{NS}}(V) under different magnetic field for various pressures. a 0.6 GPa. b 0.8 GPa. c 1.0 GPa. The curves are shifted vertically for clarity.

SI IV: Extended BTK model for spin-triplet superconductivity

In this section, we provide the details to deduce the expression of differential conductance (GNS=dI/dVG_{\mathrm{NS}} = \mathrm{d}I/\mathrm{d}V) of Andreev reflection PCS for a junction between normal metal and superconductor (NM-SC), as sketched in Supplementary Figure 5. Andreev reflection refers to a scattering process that an electron incident on the interface from the NM is retro-reflected as a hole by creating a Cooper pair in the SC [1]. For spin-triplet SCs, the process of Andreev reflection is more complex compared with that of conventional spin-singlet counterparts due to the increase of the reflection channels. For instance, if the incident particle is a spin-up electron, the reflected particle could be either a spin-up hole through the channel of equal-spin Cooper pair Φ+1=∣↑↑⟩\Phi_{+1} = \lvert\uparrow\uparrow\rangle, or a spin-down hole through the channel of opposite-spin Cooper pair Φ0=(∣↑↓⟩+∣↓↑⟩)/2\Phi_0 = (\lvert\uparrow\downarrow\rangle+ \lvert\downarrow\uparrow\rangle)/\sqrt{2}, as depicted in Supplementary Figure 5. The Blonder-Tinkham-Klapwijk (BTK) model provides a theoretical framework for analyzing the Andreev reflection PCS [2], which is based on the Bogoliubov-de Gennes (BdG) method.

In this work, the NM-SC junction lies in the x-y plane (or a-b crystallographic plane), with a potential barrier V0δ(z)V_{0}\delta(z) at the interface z=0z = 0. The NM sits at the region z<0z < 0, while the spin-triplet SC resides in z>0z > 0.

The BdG equation that governs the system is given by [3]

H~BTK(r)Φ(r)=EΦ(r),(4)\widetilde{H}_{\mathrm{BTK}}(\mathbf{r})\Phi(\mathbf{r}) = E\Phi(\mathbf{r}), \tag*{(4)}

where H~BTK(r)=[ξ(r)σ^0Δ^(r)Δ^†(r)−ξ(r)σ^0]\widetilde{H}_{\mathrm{BTK}}(\mathbf{r}) = \begin{bmatrix} \xi(\mathbf{r})\hat{\sigma}_{0} & \hat{\Delta}(\mathbf{r}) \\ \hat{\Delta}^{\dagger}(\mathbf{r}) & -\xi(\mathbf{r})\hat{\sigma}_{0} \end{bmatrix}, ξ(r)=−ℏ2∇22m+V0δ(z)−εF\xi(\mathbf{r}) = -\frac{\hbar^{2}\nabla^{2}}{2m} + V_{0}\delta(z) - \varepsilon_{F}, Δ^(r)=Δ^eiφΘ(z)\hat{\Delta}(\mathbf{r}) = \hat{\Delta}e^{i\varphi}\Theta(z). εF\varepsilon_{F} is the Fermi energy of the system, and σ^0\hat{\sigma}_{0} is the 2×22 \times2 identity matrix. The gap matrix for a spin-triplet SC is defined as Δ^(k)=id(k)⋅σ^σ^2\hat{\Delta}(\mathbf{k}) = i\mathbf{d}(\mathbf{k}) \cdot\hat{\boldsymbol{\sigma}}\hat{\sigma}_{2}, where d=(d1,d2,d3)\mathbf{d} = (d_{1}, d_{2}, d_{3}) is the vector that describes the pairing states of electrons, and σ^=(σ^1,σ^2,σ^3)\hat{\boldsymbol{\sigma}} = (\hat{\sigma}_{1},\hat{\sigma}_{2},\hat{\sigma}_{3}) with σ^i\hat{\sigma}_{i} (i=1,2,3i = 1, 2, 3) being the Pauli matrix. We here assume d×d∗=0\mathbf{d} \times\mathbf{d}^{*} = 0, which denotes that the gap matrix is unitary [3].

The wave function in the NM region can be expressed as [4]

ΦNα(r)=∑∣k∥∣<kF[(δα↑δα↓00)eikzz+(bα↑bα↓00)e−ikzz+(00aα↑aα↓)eikzz]⋅fk∥(ρ),(5)\Phi_{\mathrm{N}\alpha}(\mathbf{r}) = \sum_{|\mathbf{k}_{\parallel}|<k_{F}} \left[ \begin{pmatrix} \delta_{\alpha\uparrow} \\ \delta_{\alpha\downarrow} \\ 0 \\ 0 \end{pmatrix} e^{ik_{z}z} + \begin{pmatrix} b_{\alpha\uparrow} \\ b_{\alpha\downarrow} \\ 0 \\ 0 \end{pmatrix} e^{-ik_{z}z} + \begin{pmatrix} 0 \\ 0 \\ a_{\alpha\uparrow} \\ a_{\alpha\downarrow} \end{pmatrix} e^{ik_{z}z} \right] \cdot f_{\mathbf{k}_{\parallel}}(\boldsymbol{\rho}), \tag*{(5)}

where δαβ\delta_{\alpha\beta} is the Kronecker symbol, while aαβa_{\alpha\beta} and bαβb_{\alpha\beta} stand for the Andreev reflection and normal reflection coefficients, respectively [Supplementary Fig. 5(b)]. The subscript indices α\alpha and β\beta denote the spin state. The spatial wave function parallel to the junction interface is expressed as fk∥(ρ)=eik∥⋅ρ/Sf_{\mathbf{k}_{\parallel}}(\boldsymbol{\rho}) = e^{i\mathbf{k}_{\parallel}\cdot\boldsymbol{\rho}}/\sqrt{S}, where SS is area of the interface. The displacement vector ρ=(x, y, 0)\boldsymbol{\rho} = (x,\ y,\ 0) lies within the interfacial plane and the wave vector component parallel to the junction interface is given by k∥=(kx, ky, 0)\mathbf{k}_{\parallel} = (k_{x},\ k_{y},\ 0). For those particles with energy E≪εFE \ll\varepsilon_{F}, the wave vector components obey the approximate relation kz2+∣k∥∣2≈kF2=2mεFℏ2k_{z}^{2} + |\mathbf{k}_{\parallel}|^{2} \approx k_{F}^{2} = \sqrt{\frac{2m\varepsilon_{F}}{\hbar^{2}}}.

The wave function in the SC is given by [3]

ΦSα(r)=∑∣k∥∣<kF[cα1(e)(u+0−d~1,+∗−id~2,+∗d~3,+∗)eikzz+cα2(e)(0u+d~3,+∗d~1,+∗−id~2,+∗)eikzz+cα1(h)(−d~1,−+id~2,−d~3,−u−0)e−ikzz+cα2(h)(d~3,−d~1,−+id~2,−0u−)e−ikzz]⋅fk∥(ρ)(6)\Phi_{S\alpha}(\mathbf{r})=\sum_{\lvert\mathbf{k}_{\parallel}\rvert<k_{F}}\left[ c_{\alpha1}^{(e)} \begin{pmatrix} u_{+}\\ 0\\ -\tilde{d}_{1,+}^{*}-i\tilde{d}_{2,+}^{*}\\ \tilde{d}_{3,+}^{*} \end{pmatrix}e^{ik_{z}z} +c_{\alpha2}^{(e)} \begin{pmatrix} 0\\ u_{+}\\ \tilde{d}_{3,+}^{*}\\ \tilde{d}_{1,+}^{*}-i\tilde{d}_{2,+}^{*} \end{pmatrix}e^{ik_{z}z} +c_{\alpha1}^{(h)} \begin{pmatrix} -\tilde{d}_{1,-}+i\tilde{d}_{2,-}\\ \tilde{d}_{3,-}\\ u_{-}\\ 0 \end{pmatrix}e^{-ik_{z}z} +c_{\alpha2}^{(h)} \begin{pmatrix} \tilde{d}_{3,-}\\ \tilde{d}_{1,-}+i\tilde{d}_{2,-}\\ 0\\ u_{-} \end{pmatrix}e^{-ik_{z}z} \right]\cdot f_{\mathbf{k}_{\parallel}}(\boldsymbol{\rho}) \tag*{(6)}

where

u±=E+ξ±(E+ξ±)2+∣d±∣2,d~i,±=di,±(E+ξ±)2+∣d±∣2,u_{\pm}=\frac{E+\xi_{\pm}}{\sqrt{(E+\xi_{\pm})^{2}+\lvert\mathbf{d}_{\pm}\rvert^{2}}},\qquad \tilde{d}_{i,\pm}=\frac{d_{i,\pm}}{\sqrt{(E+\xi_{\pm})^{2}+\lvert\mathbf{d}_{\pm}\rvert^{2}}},
ξ±=E2−∣d±∣2,d±=d(±kz,k∥)=(d1,±,d2,±,d3,±),\xi_{\pm}=\sqrt{E^{2}-\lvert\mathbf{d}_{\pm}\rvert^{2}},\qquad \mathbf{d}_{\pm}=\mathbf{d}(\pm k_{z},\mathbf{k}_{\parallel})=(d_{1,\pm},d_{2,\pm},d_{3,\pm}),

and cαi(e(h))c_{\alpha i}^{(e(h))} represents the transmission coefficient for an electron-like (hole-like) quasiparticle. Notably, the kinetic energy of quasiparticle, ξ=E2−∣d(k)∣2\xi=\sqrt{E^{2}-\lvert\mathbf{d}(\mathbf{k})\rvert^{2}}, satisfies the following relation, due to the causality of the transport coefficients [5]

ξ={E2−∣d(k)∣2,(E≥∣d(k)∣)i∣d(k)∣2−E2,(−∣d(k)∣<E<∣d(k)∣)−E2−∣d(k)∣2,(E≤−∣d(k)∣)(7)\xi= \begin{cases} \sqrt{E^{2}-\lvert\mathbf{d}(\mathbf{k})\rvert^{2}}, & (E\geq\lvert\mathbf{d}(\mathbf{k})\rvert)\\ i\sqrt{\lvert\mathbf{d}(\mathbf{k})\rvert^{2}-E^{2}}, & (-\lvert\mathbf{d}(\mathbf{k})\rvert<E<\lvert\mathbf{d}(\mathbf{k})\rvert)\\ -\sqrt{E^{2}-\lvert\mathbf{d}(\mathbf{k})\rvert^{2}}, & (E\leq-\lvert\mathbf{d}(\mathbf{k})\rvert) \end{cases} \tag*{(7)}

The boundary conditions for the wave functions are [5]

ΦSα(r)∣z=0+=ΦNα(r)∣z=0−,(8)\left.\Phi_{S\alpha}(\mathbf{r})\right|_{z=0^{+}}=\left.\Phi_{N\alpha}(\mathbf{r})\right|_{z=0^{-}}, \tag*{(8)}
−ℏ22m[∂zΦSα(r)∣z=0+−∂zΦNα(r)∣z=0−]+V0ΦSα(r)∣z=0+=0,(9)-\frac{\hbar^{2}}{2m}\left[\left.\partial_{z}\Phi_{S\alpha}(\mathbf{r})\right|_{z=0^{+}}-\left.\partial_{z}\Phi_{N\alpha}(\mathbf{r})\right|_{z=0^{-}}\right]+V_{0}\left.\Phi_{S\alpha}(\mathbf{r})\right|_{z=0^{+}}=0, \tag*{(9)}

where mm is the mass of the electron. The boundary conditions are derived from the continuity of the wave function and the conservation of the probability current density, respectively. Substituting the wave functions Eqs. (5) and (S3) into two boundary conditions, we can derive the following equations:

\begin{pmatrix} \hat{\sigma}_0 \\ \hat{R}_h \end{pmatrix} = \tilde{U} \begin{pmatrix} \left(1 + i\frac{Z}{k}\right)\hat{T}_e \\ i\frac{Z}{k}\hat{T}_h \end{pmatrix}, \tag*{(10)}\end{pmatrix}
\begin{pmatrix} \hat{R}_e \\ 0 \end{pmatrix} = \tilde{U} \begin{pmatrix} -i\frac{Z}{k}\hat{T}_e \\ \left(1-i\frac{Z}{k}\right)\hat{T}_h \end{pmatrix}, \tag*{(11)}\end{pmatrix}

where

U~=[u++σ^0v^−v^+u−−σ^0],v^+=Δ^+†(E+ξ+)2+∣d+∣2,\tilde{U} = \begin{bmatrix} u_+ + \hat{\sigma}_0 & \hat{v}_- \\ \hat{v}_+ & u_- - \hat{\sigma}_0 \end{bmatrix}, \qquad\hat{v}_+ = \frac{\hat{\Delta}_+^\dagger}{\sqrt{(E+\xi_+)^2+|\mathbf{d}_+|^2}},
v^−=Δ^−(E+ξ−)2+∣d−∣2,Δ^±=Δ^(±kz,k∥),Z=mV0ℏ2kF,k~=kzkF.\hat{v}_- = \frac{\hat{\Delta}_-}{\sqrt{(E+\xi_-)^2+|\mathbf{d}_-|^2}}, \qquad\hat{\Delta}_\pm= \hat{\Delta}(\pm k_z,\mathbf{k}_{\parallel}), \qquad Z = \frac{mV_0}{\hbar^2 k_F}, \qquad\tilde{k} = \frac{k_z}{k_F}.
R^h=[a↑↑a↓↑a↑↓a↓↓]\hat{R}_h = \begin{bmatrix} a_{\uparrow\uparrow} & a_{\downarrow\uparrow} \\ a_{\uparrow\downarrow} & a_{\downarrow\downarrow} \end{bmatrix}

is the Andreev reflection matrix,

R^e=[b↑↑b↓↑b↑↓b↓↓]\hat{R}_e = \begin{bmatrix} b_{\uparrow\uparrow} & b_{\downarrow\uparrow} \\ b_{\uparrow\downarrow} & b_{\downarrow\downarrow} \end{bmatrix}

is the normal reflection matrix, and

T^e(h)=[c1e(h)c1e(h)c2e(h)c2e(h)]\hat{T}_{e(h)} = \begin{bmatrix} c_{1}^{e(h)} & c_{1}^{e(h)} \\ c_{2}^{e(h)} & c_{2}^{e(h)} \end{bmatrix}

is the transmission matrix of electron-like (hole-like) quasiparticles, respectively.

By solving Eqs. (10) and (11), we can get solutions for the reflection matrix:

R^e=rn(1−v^−v^+u+u−)A^−1,R^h=∣tn∣2v^+u+A^−1(12)\hat{R}_e = r_n\left(1-\frac{\hat{v}_-\hat{v}_+}{u_+u_-}\right)\hat{A}^{-1}, \qquad\hat{R}_h = |t_n|^2\frac{\hat{v}_+}{u_+}\hat{A}^{-1} \tag*{(12)}

where rn=−iZk+iZr_n=\frac{-iZ}{k+iZ}, tn=k~k+iZt_n=\frac{\tilde{k}}{k+iZ}, A^=1−∣rn∣2v^−v^+u+u−\hat{A}=1-|r_n|^2\frac{\hat{v}_-\hat{v}_+}{u_+u_-}.

Total current across the NM-SC junction deduced from the Landauer-Buttiker expression is given by [6]

INS=I0∫−∞+∞[fL(E)−fR(E)]∑k∥{1+12∑α,β[∣aαβ(E)∣2−∣bαβ(E)∣2]} dE,(13)I_{\mathrm{NS}} = I_0 \int_{-\infty}^{+\infty} [f_L(E)-f_R(E)]\sum_{\mathbf{k}_{\parallel}}\left\{1+\frac{1}{2}\sum_{\alpha,\beta}\left[|a_{\alpha\beta}(E)|^2-|b_{\alpha\beta}(E)|^2\right]\right\}\,dE, \tag*{(13)}

where fL(R)(E)=(1+eE−μL(R)/kBT)−1f_{L(R)}(E)=\left(1+e^{E-\mu_{L(R)}/k_BT}\right)^{-1} is the Fermi-Dirac distribution function of the left (right) reservoirs with the chemical potentials μL=ϵF+eV\mu_L=\epsilon_F+eV and μR=ϵF\mu_R=\epsilon_F, respectively.

The reflection coefficients satisfy

∑α,β∣bαβ(E)∣2=Tr⁡(R^eR^e†),∑α,β∣aαβ(E)∣2=Tr⁡(R^hR^h†).(14)\sum_{\alpha,\beta} |b_{\alpha\beta}(E)|^2 = \operatorname{Tr}(\hat{R}_e\hat{R}_e^\dagger), \qquad\sum_{\alpha,\beta} |a_{\alpha\beta}(E)|^2 = \operatorname{Tr}(\hat{R}_h\hat{R}_h^\dagger). \tag*{(14)}

Then, with the Andreev reflection and normal reflection coefficients, the differential conductance (as a function of VV and TT) is given by

GNS(V,T)=ddVINS=G0∫−∞+∞∂f(E−eV)∂V∑k∥{1+12∑α,β[∣aαβ(E)∣2−∣bαβ(E)∣2]}dE.(15)G_{\mathrm{NS}}(V,T) = \frac{\mathrm{d}}{\mathrm{d}V} I_{\mathrm{NS}} = G_0 \int_{-\infty}^{+\infty} \frac{\partial f(E-eV)}{\partial V} \sum_{k_{\parallel}} \left\{1+\frac{1}{2}\sum_{\alpha,\beta}\left[|a_{\alpha\beta}(E)|^2-|b_{\alpha\beta}(E)|^2\right]\right\}\mathrm{d}E. \tag*{(15)}

Specially, at zero temperature, the differential conductance is obtained,

GNS(V)∣T=0=G0{1+12∑α,β[∣aαβ(eV)∣2−∣bαβ(eV)∣2]}.(16)G_{\mathrm{NS}}(V)\big|_{T=0} = G_0 \left\{1+\frac{1}{2}\sum_{\alpha,\beta}\left[|a_{\alpha\beta}(eV)|^2-|b_{\alpha\beta}(eV)|^2\right]\right\}. \tag*{(16)}

A detailed solution of the BdG equations in the presence of inelastic scattering [7] shows that the impact of finite quasiparticle lifetime can be effectively modeled through the substitution E→E+iΓE \to E+i\Gamma. This leads us to the final expression for the differential conductance in this model,

GNS(V,T)=G0∫−∞+∞∂f(E−eV)∂V∑k∥{1+12∑α,β[∣aαβ(E+iΓ)∣2−∣bαβ(E+iΓ)∣2]}dE.(17)G_{\mathrm{NS}}(V,T) = G_0 \int_{-\infty}^{+\infty} \frac{\partial f(E-eV)}{\partial V} \sum_{k_{\parallel}} \left\{1+\frac{1}{2}\sum_{\alpha,\beta}\left[|a_{\alpha\beta}(E+i\Gamma)|^2-|b_{\alpha\beta}(E+i\Gamma)|^2\right]\right\}\mathrm{d}E. \tag*{(17)}

Simulations of Andreev reflection PCS for spin-triplet superconductivity

Supplementary Figure 6 shows the simulated PCS at different temperatures, based on px(y)p_{x(y)}- (left column) and pzp_z- (right column) wave pairings. For comparison, all the simulations were performed with fixed Γ/Δ=0.2\Gamma/\Delta= 0.2. For small Z=0.2Z = 0.2, both px(y)p_{x(y)} and pzp_z waves show peak-like feature at zero bias, as shown in Supplementary Figure 6,Supplementary Figure 6). For large Z=1Z = 1, they differ to each other substantially. For px(y)p_{x(y)} wave, a zero-bias dip develops [Supplementary Figure 6]. In contrast, for pzp_z wave, the ZBCP is robust against varying ZZ, but meanwhile a dip develops on each side of the ZBCP, cf. Supplementary Figure 6. Another major difference between px(y)p_{x(y)} and pzp_z waves lies in the amplitude of the ZBCP, with the former obviously smaller than the latter, as is mentioned in the main text.

We then simulate the PCS for different representations, viz AuA_u, B1uB_{1u}, B2uB_{2u} and B3uB_{3u}, and the results are displayed in Supplementary Figure 7. The basis functions of these representations have been listed in Table 1. Note that here we have ignored the high-order (kxkykzk_xk_yk_z) terms whose effect on PCS turns out to be negligible, cf Supplementary Figure 13. The main parameters used are listed in the figures. For AuA_u presentation, a “flattened” ZBCP shows up for Z=0Z = 0, but once Z>0Z > 0, the peak gradually splits into two peaks; a dip appears at ∣V∣>0|V| > 0 when Z>1Z > 1, as can be seen in Supplementary Figure 7. The sensitivity of the ZBCP to the values of ZZ and Γ\Gamma is depicted by the contour plot of d2GNS(0)/dV2\mathrm{d}^{2}G_{\mathrm{NS}}(0)/\mathrm{d}V^{2} as shown in Supplementary Figure 8. For B1uB_{1u} presentation, a ZBCP is also visible for small ZZ, but as ZZ increases, it gradually evolves into a zero-bias dip [Supplementary Fig. 7(b)]. As is shown in Supplementary Fig. 7(c,d), the ZBCP in the B2uB_{2u} presentation is robust to the variations of ZZ and Γ\Gamma. In contrast, the finite-bias conductance is highly susceptible. For fixed Γ=0.05\Gamma= 0.05 meV, side dips emerge as ZZ exceeds 0.8. Furthermore, at Z=1Z = 1, a large Γ\Gamma of 0.40 meV produces only a shallow, shoulder-like dip. This dip deepens and sharpens into well-defined satellite peaks as Γ\Gamma is reduced. The B3uB_{3u} behaves similarly to B2uB_{2u}. We shall see that B2uB_{2u} and B3uB_{3u} capture all the main features of the experimental PCS of UTe2_2.

Contour plot and conductance simulations for $A_u$

Supplementary Figure 8. Supplementary Figure 8 | Additional simulations for AuA_u. The parameters used are: Δx=Δy=Δz=1\Delta_x = \Delta_y = \Delta_z = 1 meV, and T=0.3T = 0.3 K. a Contour plot of d2GNS(0)/dV2\mathrm{d}^{2}G_{\mathrm{NS}}(0)/\mathrm{d}V^{2} as a function of ZZ and Γ\Gamma. The black line is the boundary of the regions where GNS(V)G_{\mathrm{NS}}(V) does and doesnt split at zero bias. For any reasonably large Γ\Gamma, GNS(V)G_{\mathrm{NS}}(V) either splits or shows a flat and broad central peak at V=0V = 0, as exemplified by the simulations with fixed Γ=0.2\Gamma= 0.2 meV and variable ZZ in b. These features are qualitatively inconsistent with our experiments.

Image not extracted

Supplementary Figure 13. Comparison of point-contact spectra for the B1uB_{1u} pairing with and without the high-order kxkykzk_xk_yk_z term. kF=1k_F = 1 in these calculations. For simplified B1uB_{1u}, d=Δ(ky,kx,0)\mathbf{d} = \Delta(k_y, k_x, 0), and the results are represented by open circles; for high-order imposed B1uB_{1u}, d=Δ(ky,kx,kxkykz)\mathbf{d} = \Delta(k_y, k_x, k_xk_yk_z), and the results are depicted by solid lines. a Simulations with fixed Γ/Δ=0.1\Gamma/\Delta= 0.1, and varying Z=0Z = 0, 0.50.5 and 2.02.0. b Simulations with fixed Z=1.0Z = 1.0, and varying Γ/Δ=0\Gamma/\Delta= 0, 0.20.2 and 0.50.5.

Fitting of the experimental data

Image not extracted

Supplementary Figure 10. Fitting of the point-contact spectra at 0.8 GPa for sample S1. a 0.5 K. b 0.7 K. c 1.1 K. d 1.5 K. e 1.8 K. f 2.1 K. The fittings were carried out with B2uB_{2u} (or B3uB_{3u}).

Consideration of high-order term in SC order parameter

Symmetry allows higher-order basis functions – e.g., the kxkykzk_xk_yk_z term – in the B1uB_{1u} pairing gap, whose inclusion will also cause a sign change in Δ(k)\Delta(\mathbf{k}) upon going from kzk_z to −kz-k_z. To estimate how this term affects the PCS for tunneling current I∥c\mathbf{I} \parallel c, we carried out additional calculations for the B1uB_{1u} state. Assuming the high-order component has a large amplitude comparable to the linear-order ones, we set d=Δ(ky,kx,kxkykz)\mathbf{d} = \Delta(k_y, k_x, k_xk_yk_z) and compared with the simplified B1uB_{1u} form d=Δ(ky,kx,0)\mathbf{d} = \Delta(k_y, k_x, 0). kF=1k_F = 1 is used in the simulations. As shown in Supplementary Fig. 13, the inclusion of the kxkykzk_xk_yk_z term produces only minor modifications to the calculated spectra. In particular, this extended B1uB_{1u} model still cannot account for the prominent side-dip/three-peak features seen in the experiments.

Consideration of anisotropic Fermi surface

In this section, we performed some simulations aiming at comparing the conductance spectra for isotropic and anisotropic Fermi surfaces, seeing Supplementary Figure 14. The extent of FS anisotropy is controlled by the parameter rv≡vFc/vFabr_v \equiv v_F^c/v_F^{ab}, where vFabv_F^{ab} and vFcv_F^c are the in-plane and out-of-plane Fermi velocities, respectively. While the lineshape of GNS(V)G_{\mathrm{NS}}(V) changes with rvr_v, the behavior near the gap edge remains essentially unchanged.

References

References

  1. [1]Andreev, A. F. Thermal conductivity of the intermediate state of superconductors. J. Exp. Theor. Phys. 46, 1823-1828 (1964).
  2. [2]Blonder, G. E., Tinkham, M. & Klapwijk, T. M. Transition from metallic to tunneling regimes in superconducting microconstrictions: excess current, charge imbalance, and supercurrent conversion. Phys. Rev. B 25, 4515-4532 (1982).DOI
  3. [3]Sigrist, M. & Ueda, K. Phenomenological theory of unconventional superconductivity. Rev. Mod. Phys. 63, 239-311 (1991).DOI
  4. [4]Setiawan, F. et al. Conductance spectroscopy of topological superconductor wire junctions. Phys. Rev. B 91, 214513 (2015).DOI
  5. [5]Asano, Y. Andreev Reflection in Superconducting Junctions (Springer, Singapore, 2021).DOI
  6. [6]Datta, S. Electronic Transport in Mesoscopic Systems Ch. 2 (Cambridge Univ. Press, Cambridge, 1995).
  7. [7]Pleceník, A., Grajcar, M., Beňačka, Š., Seidel, P. & Pfuch, A. Finite-quasiparticle-lifetime effects in the differential conductance of Bi₂Sr₂CaCu₂O₇ / Au junctions. Phys. Rev. B 49, 10016-10019 (1994).

Paper details

vlnc:W_aag8brzg
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents