High transparency induced superconductivity in field effect two-dimensional electron gases in undoped InAs/AlGaSb surface quantum wells

Abstract

We report on transport characteristics of field effect two-dimensional electron gases (2DEG) in 24 nm wide indium arsenide surface quantum wells. High quality single-subband magnetotransport with clear quantized integer quantum Hall plateaus are observed to filling factor $\nu=2$ in magnetic fields of up to $B=18$ T, at electron densities up to 8$\times 10^{11}$ /cm². Peak mobility is 11,000 cm²/Vs at 2$\times 10^{12}$ /cm². Large Rashba spin-orbit coefficients up to 124 meV$\cdot$Å are obtained through weak anti-localization (WAL) measurements. Proximitized superconductivity is demonstrated in Nb-based superconductor-normal-superconductor (SNS) junctions, yielding 78$-$99 % interface transparencies from superconducting contacts fabricated ex-situ (post-growth), using two commonly-used experimental techniques for measuring transparencies. These transparencies are on a par with those reported for epitaxially-grown superconductors. These SNS junctions show characteristic voltages $I_{c}R_{\textsc{n}}$ up to 870 $\mu$V and critical current densities up to 9.6 $\mu$A/$\mu$m, among the largest values reported for Nb-InAs SNS devices.

The last decade has seen spectacular progress in InAs/AlGaSb two-dimensional electron gases (2DEGs).[1, 2, 3, 4, 5, 6] This material system has joined the small club where the fractional quantum Hall effect can be routinely observed,[7, 8] in addition to 2DEGs in GaAs/AlGaAs,[9, 10] in AlAs/AlGaAs,[11] in graphene,[12, 13] in Si/SiGe,[14, 15], in Ge/SiGe,[16] in CdTe,[17, 18] and in ZnO/MgZnO.[19, 20] The highest mobility reported in InAs/AlGaSb is 2.4$\times$10⁶ cm²/Vs,[21, 22] only exceeded by GaAs/AlGaAs,[23, 24] Ge/SiGe,[25] and AlAs/AlGaAs[11] material systems. The combination of high mobilities, strong spin orbit interactions (SOI), pinning of the Fermi level in the conduction band, small effective mass, and large Landé g-factor could make InAs/AlGaSb a strong candidate material system for topological quantum computing with Majorana zero modes.[26, 27, 28]

In the last decade, most efforts towards realizing Majorana fermions in a scalable platform have focused on surface quantum wells in the In(Ga)As/In_0.8Al_0.2As material system, where mobilities have significantly improved from 1$\times 10^{4}$ cm²/Vs initially[26] to more than 1$\times 10^{5}$ cm²/Vs recently.[29] However, in the context of topological quantum computing, the InAs/Al_0.8Ga_0.2Sb material system could offer possible advantages over the In(Ga)As/In_0.8Al_0.2As system, including better strain engineering, higher electron densities, and higher mobilities.

Furthermore, most efforts in this field have also centered on semiconductor-superconductor hybrid devices proximitized by “epitaxial aluminum,” grown directly on In(Ga)As/InAlAs heterostructures in the same growth chamber. This approach appeared to be the only way to reliably generate strong, “hard” superconducting gaps in the semiconductor,[33] as opposed to smaller, “soft” gaps, typically produced by superconductor contacts deposited ex-situ, post-growth.

In this Letter, we demonstrate gated 2DEGs in InAs/Al_0.8Ga_0.2Sb surface quantum wells, without parallel conduction from another conducting layer in magnetic fields up to 18 T and at electron densities up to $3\times 10^{12}$ cm^-2. Single-subband operation is demonstrated at lower 2DEG densities. Using SiO₂ as dielectric yielded stable and reproducible gating operations all the way down to pinch-off. Rashba spin-orbit coefficients up to 124 meV$\cdot$Å are obtained through weak anti-localization (WAL) measurements. Proximitized superconductivity is demonstrated in Nb-based superconductor-normal-superconductor (SNS) Josephson junctions, yielding deep gaps with up to unity transparencies from superconducting contacts fabricated entirely ex-situ (post-growth).

Two nominally identical heterostructures (G743 and G782) were grown by molecular beam epitaxy (MBE), with the following sequence of layers, starting from a GaSb (001) substrate: a 25 nm GaSb nucleation layer, a 800 nm Al_0.80Ga_0.20Sb_0.93As_0.07 quaternary buffer, a 20 nm Al_0.8Ga_0.2Sb bottom barrier, a 24 nm InAs quantum well, and a 6 nm In_0.75Ga_0.25As cap layer. There is no intentional do** anywhere in the heterostructure. Section I in the supplementary material provides additional details about MBE growth. Figure S1 from Section II in the supplementary material shows nextnano™ self-consistent simulations[34, 35] of bandstructure profiles, showing the extent of the 2DEG wavefunction within the InAs quantum well, and compares it to an In(Ga)As quantum well profile.

Hall bars were fabricated using standard optical lithography and wet-etching techniques, kee** all processes at or below a temperature of 150^∘C to prevent the deterioration of device characteristics.[36, 37, 38] Ti/Au Ohmic contacts were deposited directly on the InGaAs cap layer, with typical resistances of 400$-$500 $\Omega$ at magnetic field $B=0$, and 10 k$\Omega$ at $B=18$ T. Finally, 60 nm thick silicon dioxide (SiO₂) or hafnium dioxide (HfO₂) was deposited by atomic layer deposition (ALD) at 150^∘C, followed by the deposition of a Ti/Au global top-gate. See Section III.A of the supplementary material for more details on Hall bar fabrication. Using standard four-terminal ac lock-in measurement techniques, all transport experiments were performed in either a pumped-⁴He cryostat or ³He/⁴He dilution refrigerator, with a base temperature of 1.6 K and 11 mK respectively.

The electron density at top-gate voltage V${}_{g}=0$ in all gated Hall bars was significantly larger than the as-grown electron densities in ungated Hall bars. Figure 1 shows the pinch-off characteristics of Hall bars with different gate dielectrics. With SiO₂, the pinch-off curves are stable and reproducible, overlap** perfectly when $V_{g}$ is swept in the same direction, but showing some hysteresis when $V_{g}$ is swept in opposite directions [see Fig. 1(a)]. After pinch-off, the 2DEG does not turn itself back on with time.[22] With HfO₂ however, gating characteristics become unstable and non-reproducible near pinch-off [see Fig. 1(b)].

Figure 1: Typical differential conductance $G(V_{g})=dI/dV_{sd}$ (using 100 $\mu$V ac excitation) showing the turn-on voltage of a gated Hall bar with gate dielectric (a) SiO2 and (b) HfO2. In panel (a), eight traces are shown, four while increasing $V_{g}$ (black) and four while decreasing $V_{g}$ (gray). Only four traces are shown in panel (b), two going up and two going down in $V_{g}$. (c) Typical $n_{\text{2D}}(V_{g})$ of a Hall bar with SiO2. The 2DEG density increases linearly with $V_{g}$ in all samples, and is reproducible along the linear traces. (d) Electron mobility in the same device as in (c).

At high conductances ($G>100$ $\mu$S), both SiO₂ and HfO₂ produce stable and reproducible gating behavior. Figure 1(c) shows a typical electron density function $n_{\text{2D}}(V_{g})$ with SiO₂. Section IV in the supplementary material shows representative gating and mobility characteristics from four gated Hall bars with SiO₂ and HfO₂. Figure 1(d) shows a peak in the transport mobility $\mu=11\times 10^{3}$ cm²/Vs near 2DEG density $n_{\text{2D}}=2.0\times 10^{12}$ cm^-2. At lower densities, ionized impurity scattering is most likely limiting mobility. The much higher 2DEG carrier density at $V_{g}=0$ (by a factor of 2$-$3) in gated Hall bars (i.e., after the dielectric deposition) relative to the as-grown 2DEG densities in ungated Hall bars strongly hints at a high density of charge traps forming at the semiconductor-oxide interface. To eliminate the risk of parallel conduction, often observed in similar structures when electrons are supplied by remote do**, the AlGaSb/InAs interface was engineered to have an AlAs character. Such interfaces are known to have high concentrations of As antisites, which act as double donors, supplying carriers to the 2DEG in the InAs quantum well.[39] The resulting high density of As²⁺ ions at the interface subjects the 2DEG to strong Coulomb scattering, reducing its mobility. However, with the top quantum well barrier being InGaAs $–$ a necessary element of our design $–$ such a tradeoff is well justified. Indeed, Lee et al.[5] demonstrated a severe reduction of mobility, from $650\text{\times}{10}^{3}$ cm²/Vs down to $24\text{\times}{10}^{3}$ cm²/Vs, a value comparable with our results, by only replacing the AlGaSb top barrier layer with InGaAs in otherwise identical heterostructures in near-surface InAs quantum wells. Such reduction of mobility is likely due to the strong extent of the 2DEG wavefunction into the InGaAs barrier, with a lower conduction band offset than AlGaSb. At carrier densities above the mobility peak in Fig. 1(d), the observed decline in mobility with increasing 2DEG density can be attributed to interface roughness scattering and/or alloy scattering. This is likely because the electron wavefunction is drawn closer to the surface by the stronger electric field from the top-gate.[40, 41]

Figure 2: (a) SdH and Hall traces for $n_{\text{2D}}=8.57\times 10^{11}$ cm^-2 ($V_{g}=-8.0$ V). (b) Landau fan of SdH oscillations with densities ranging from $n_{\text{2D}}=7.3\times 10^{11}$ cm^-2 to $n_{\text{2D}}=1.8\times 10^{12}$ cm^-2. White labels identify corresponding effective filling factors (directly obtained from the Hall resistance with $\nu=h/e^{2}R_{H}$, regardless of subband occupation), and dashed white lines identify populating Landau levels from the second 2D subband. (c) Weak anti-localization peak at $n_{\text{2D}}=(0.5;0.9;1.4;1.9)\times 10^{12}$ cm^-2. Colored circles are experimental points, black lines are fits to the ILP model, and $n_{\text{2D}}$ is indicated in the upper right corner of each plot. (d) Spin orbit coefficient $\alpha_{so}$ as a function of $n_{\text{2D}}$, extracted from fits of the WAL data to the ILP and HLN models. The dashed vertical line indicates where the 2${}^{\text{nd}}$ subband populates.

In a Hall bar with SiO₂, Figure 2(a) shows Shubnikov-de-Haas (SdH) oscillations in the longitudinal resistivity $\rho_{xx}$ and well-defined quantized quantum Hall plateaus in the Hall resistance $R_{\textsc{h}}$. Quantum Hall plateaus occur at specific resistance values $R_{\textsc{h}}=h/\nu e^{2}$ at filling factors $\nu=hn_{2D}/eB$, where $h$ is the Planck constant and $e$ is the single electron charge. The presence of quantized Hall plateaus at $\nu=2,4$ confirms the formation of a 2DEG. We note SdH oscillations are not, by themselves, proof of the presence of a 2DEG, since they are also observed in 3D conductors,[42] albeit with much smaller amplitudes. Despite a large Landé $g$-factor ($g^{*}\sim$ 15), the spin-split $\nu=3$ Hall plateau is only starting to be resolved at $B\approx 10$ T, because of disorder. The visibility of spin splitting is dictated by $(g^{*}\mu_{\textsc{b}}B-\Gamma)>k_{\textsc{b}}T$, where $\mu_{\textsc{b}}$ is the Bohr magneton, $\Gamma$ is disorder associated with Landau level broadening, and $k_{\textsc{b}}$ is the Boltzmann constant. Thus the late onset in field of spin splitting implies $\Gamma\sim 9$ meV in our samples. This same disorder $\Gamma\sim 9$ meV is also responsible for the very late onset of SdH oscillations ($B\approx 3$ T), whose visibility is determined by $(\hbar eB/m^{*}-\Gamma)>k_{\textsc{b}}T$.

For $B>10$ T, the SdH oscillation minimum at $\nu=2$ reaches $\rho_{xx}=0$ in Fig. 2(a), implying no parallel conduction from another conductive layer. This remains true until at least $n_{\text{2D}}=2.8\times 10^{12}$ cm^-2 ($V_{g}=-0.5$ V). However, this does not exclude the possibility of a second subband populating the 2DEG. Indeed, at $n_{\text{2D}}=7.3\times 10^{11}$ cm^-2 ($V_{g}=-8.5$ V), there is a small mismatch ($<$3%) between the classical Hall density $n_{total}=B/eR_{\textsc{h}}$ and the 2DEG density determined from the periodicity of Shubnikov-de-Haas oscillations versus inverse field, given by $n_{SdH}=\frac{2e}{h}\left(\frac{1}{B_{\nu+1}}-\frac{1}{B_{\nu}}\right)^{-1}$. This mismatch grows significantly as the 2DEG density increases (until $n_{total}\approx 2n_{SdH}$ near $V_{g}=0$). Energy level crossings in the Landau fan from Figure 2(b) confirm the presence of another 2D subband, most likely corresponding to the two spin-split branches of the lowest Landau level. We estimate the second subband populates near $n_{\text{2D}}\approx 8\times 10^{11}$ cm^-2, which is consistent with similar reports of populating second subbands in 24 nm wide InAs/AlGaSb quantum wells.[21, 22] The “knee” observed in the pinch-off characteristics in Figure 1(a) is consistent with the population of a second subband in the 2DEG. We note that the second subband’s branch separating the regions labeled $\nu=4$ and $\nu=5$ at $B\sim 10$ T in Fig. 2(b) does not appear to cross the $\nu=3$ branch and does not extend into the $\nu=2$ region. This is reminiscent of similar occurrences in the ring-like structures from Landau fans with Landau level crossings between the first and second subbands of GaAs/AlGaAs 2DEGs[43, 44, 45] and InAs/AlGaSb 2DEGs.[21] Nonetheless, our Landau fan, obtained by swee** the top-gate at magnetic field increments, showcases the reproducibility and stability of gating characteristics with dielectric SiO₂.

Figure 3: (a) Two overlapped I-V traces (solid) in a SNS device, where $I_{sd}$ is the source-drain dc current and $V_{dc}$ is the four-terminal dc voltage drop across the SNS junction. The dashed lines are the extrapolated normal resistance $R_{\textsc{n}}$ and their intercepts at $V_{dc}=0$ are the excess currents $I_{ex}$. (upper inset) Zoom-in of the superconducting transition, showing critical current $I_{c}$. Two overlap** traces are shown (black, grey), swee** $I_{sd}$ in opposite directions. (lower inset) Sharp, deep gap in the four-terminal differential resistance of device SNS-4. (b) Normalized four-terminal differential conductance $dI/dV$, identifying MAR peaks in a SNS device. (c) Temperature dependence of $I_{c}$ of a SNS device, where crosses are experimental data and dashed lines are fits to Eqn. (3) for different values of $\mathcal{T}$.

The curvature of the energy level associated with $\nu=2$ near $B=17$ T at high densities in Fig. 2(b) is consistent with strong spin orbit interactions in the system. To quantify the strength of spin-orbit interactions, the Rashba coefficient $\alpha_{so}$ was determined from fits to the weak antilocalization (WAL) conductivity peak $\Delta\sigma_{xx}$ [see Fig. 2(c)] using the Hikami-Larkin-Nagaoka (HLN) and Iordanskii-Lyanda-Geller-Pikus (ILP) models,[46, 47] where $\Delta\sigma_{xx}$ = $\sigma_{xx}(B)-\sigma_{xx}(0)$, $\sigma_{xx}(B)$ is the field-dependent conductivity, and $\sigma_{xx}(0)$ is a constant background conductivity. Figure 2(d) summarizes the resulting fit values of $\alpha_{so}$ with both models, ranging from 16 to 124 meV$\cdot$Å, roughly linear with $n_{2D}$. These are consistent with published values for 2DEGs in InAs/AlGaSb[2, 4, 22] and In(Ga)As/InAlAs.[48, 29, 49, 50] Some of these literature values (and ours) span both the single subband and multi-subband regimes. Figure S6 from Section IV in the supplementary material presents a detailed comparison of our $\alpha_{so}(n_{\text{2D}})$ with literature. Both the HLN and ILP models are only applicable at magnetic fields where the mean free path $\ell_{e}$ is smaller than the magnetic length $\ell_{m}$, leading to the condition $B<B_{tr}$ where $B_{tr}=\hbar/2e\ell_{e}^{2}$ is the transport field. In our sample, one of the distinctive features of the conductivity used for fitting (the minima on either side of the WAL peak) occurs well beyond $B_{tr}$ for the data point at $n_{2D}=$1.9\text{\times}{10}^{12}$$ cm^-2, and the latter should therefore not be considered reliable.

Device	width	gap	$I_{c}$	$I_{ex}$	$R_{\textsc{n}}$	$\Delta_{\textsc{mar}}$	$\mathcal{T}$
ID	($\mu$m)	(nm)	($\mu$A)	($\mu$A)	($\Omega$)	(meV)	(%)
SNS-1	10	200	82	158	8.8	1.13	78 $\pm$ 5
SNS-2	10	200	89	183	8.6	1.16	81 $\pm$ 4
SNS-3	3	120	29	58	30	1.13	85 $\pm$ 2
SNS-4	3	120	25	54	33	1.05	87 $\pm$ 3
SNS-5	3	400	5	24	53	0.91	82 $\pm$ 2
SNS-6	3	400	2	26	55	0.94	84 $\pm$ 3

Table 1: List of SNS devices.

Three nominally identical pairs of superconductor-normal-superconductor (SNS) devices [see Table 1] were fabricated from wafers G743 and G782, using standard optical lithography, electron beam lithography, and wet-etching techniques, kee** all processes at or below a temperature of 180^∘C to prevent the deterioration of device characteristics.[36, 37, 38] Ti/Nb (2/80 nm) ohmic contacts were deposited directly on the InGaAs cap layer by sputtering, after Ar ion milling the contact areas for 6.5 minutes at 50 Watts. Immediately prior to loading samples in the Nb deposition system, the contact areas were treated with sulfur passivation.[51] The latter is designed to etch away native oxides, prevent further surface oxidation during transfer in air ($\sim$ 30 s) from the wetbench to the deposition chamber, and possibly dope the surface.[52, 53, 54] Four devices (SNS-3, … , SNS-6) were 3 $\mu$m wide, of which two (two) were fabricated with a gap $L=120$ nm ($L=400$ nm) between the Ti/Nb contacts. Devices SNS-1 and SNS-2 were fabricated on wafer G743 rather than G782, with a width $W=10$ $\mu$m and a gap of $L=200$ nm. Note the gap region is not gated in any device. See Section III.B of the supplementary material for more details on sample fabrication.

The sputter-deposited Nb had a critical temperature $T_{c}=8.1$ K, yielding the superconducting energy gap $\Delta_{\text{Nb}}=1.76\,k_{\textsc{b}}T_{c}=1.23$ meV. Using this value, the coherence length in the proximitized semiconductor of our SNS junctions is $\xi_{b}=\hbar v_{\textsc{f}}/2\Delta_{\text{Nb}}=220$ nm (246 nm) in the ballistic regime for wafer G782 (G743). Using the as-grown 2DEG density and mobility, the mean free path is $\ell_{e}=\hbar\mu\sqrt{2\pi n_{\text{2D}}}/e=244$ nm (175 nm) for wafer G782 (G743). Devices SNS-1 and SNS-2 are in the diffusive, short junction regime, since $\ell_{e}<L<\xi_{b}$. Devices SNS-3 and SNS-4 are in the quasi-ballistic, short junction regime, since $L<\ell_{e},\xi_{b}$. Devices SNS-5 and SNS-6 are in the diffusive, long junction regime, since $L>\ell_{e},\xi_{b}$. All devices are in the dirty regime, since $\ell_{e}\lessapprox\xi_{b}$.

All six SNS devices demonstrated a supercurrent. Figure 3(a) shows typical four-terminal dc I-V traces in a quasi-ballistic device. Remarkably, the I-V traces do not display any hysterectic behavior, as emphasized in the upper inset of Fig. 3(a); this is true for all six SNS devices reported here. All devices are well thermalized and do not adversely suffer from local Joule heating effects,[55, 56] due to the dilution refrigerator (Kelvinox TLM from Oxford Instruments) used in the experiments, where samples and wiring are completely immersed in the ³He/⁴He mixture. The lower inset in Fig. 3(a) demonstrates a deep, sharply-defined superconductivity gap.

Figure 3(b) plots the four-terminal differential conductance $G=dI/dV$ normalized by the conductance in the normal regime $G_{\textsc{n}}=1/R_{\textsc{n}}$, with periodic peaks arising from multiple Andreev reflections (MAR). The peak periodicity in $V_{dc}$ is described by $eV_{n}=2\Delta_{\textsc{mar}}/n$ where $n$ is an integer. Up to six peaks are observed in Fig. 3(b), with the first three ($n=1,...,3$) following the MAR sequence with $\Delta_{\textsc{mar}}=1.05$ meV. This value is only slightly smaller than $\Delta_{\text{Nb}}=1.23$ meV, and is consistent with a high-quality SNS junction. Section V in the supplementary material shows I-V traces and MAR analysis for all six SNS devices reported here (Figs. S8$-$S10).

The ratio $eI_{c}R_{\textsc{n}}/\Delta_{\textsc{mar}}$ and critical current density $I_{c}/W$ are often used as a figure of merit for SNS junctions, with typical ranges 0.02$-$0.15 and 0.3$-$1.6 $\mu$A/$\mu$m respectively for planar In(Ga)As quantum wells proximitized to Nb.[57, 58, 59, 60] For the four devices in the short junction regime (SNS-1, …, SNS-4 in Table 1), the ratio $eI_{c}R_{\textsc{n}}/\Delta_{\textsc{mar}}$ ranges from 0.64 to 0.79, and the critical current density $I_{c}/W$ ranges from 8.2 to 9.6 $\mu$A/$\mu$m. Both set of numbers indicate a very strong proximity effect from the Nb parent superconductor to the proximitized InAs 2DEG in the gap region.

A high-quality SNS junction is characterized by a high interface transparency $\mathcal{T}$ at the SN interfaces, which corresponds to a high probability of Andreev reflection.[61, 62, 63, 64] It can be analytically calculated with the generalized Octavion-Tinkham–Blonder-Klapwijk (OTBK) model:[65]

$$\frac{eI_{ex}R_{\textsc{n}}}{\Delta_{\textsc{mar}}} = \frac{2(1+2Z^{2})\tanh^{-1}(2\tilde{Z})}{Z\sqrt{(1+Z^{2})(1+6Z^{2}+4Z^{4})}}-\frac{4}{3}\qquad$$ (1)

$$\text{and~{}~{}}\mathcal{T} = (1+Z^{2})^{-1}$$ (2)

where $\tilde{Z}=Z\sqrt{(1+Z^{2})/(1+6Z^{2}+4Z^{4})}$, $I_{ex}$ is the excess current obtained from an I-V trace [see Fig. 3(a)], $\Delta_{\textsc{mar}}$ is the superconducting gap determined from the MAR periodicity [see Fig. 3(b)], and $Z$ is a dimensionless scattering parameter related to the barrier height at the SN interface. The formalism above explicitly assumes that both SN interfaces in the SNS junction are symmetric. To calculate the SN transparency, $Z$ is first used as a free variable in Eq. (1) to fit the experimental data ($I_{ex}$, $R_{\textsc{n}}$, $\Delta_{\textsc{mar}}$). Once a value for $Z$ is found, it is then used in Eq. (2) to calculate $\mathcal{T}$.

Using only the data at base temperature and Eqns. (1)$-$(2), all six SNS devices in Table 1 show consistently high transparencies $\mathcal{T}$ ranging in values from 78% to 87%. For context, reported interface transparencies of epitaxial aluminum to In(Ga)As surface quantum wells range from 75% to 97%,[66, 67, 68] when using the same experimental measurement method.

Another method for measuring $\mathcal{T}$ involves the temperature dependence of $I_{c}$ to be fit to the generalized Kulik-Omelyanchuk (KO) model:[69]

$$I_{c}(T,\phi,\mathcal{T}) = \frac{\pi\Delta(T)\sin(\phi)}{2eR_{\textsc{n}}\sqrt{1-\mathcal{T}\sin^{2}(\phi/2)}}\qquad\qquad\qquad\qquad$$ (3)

$$\times\tanh\left(\frac{\Delta(T)\sqrt{1-\mathcal{T}\sin^{2}(\phi/2)}}{2k_{\textsc{b}}T}\right)$$

where $\phi$ is the superconducting phase picked up from Andreev reflections at the semiconductor-superconductor interface, and $\Delta(T)$ is the temperature-dependent superconducting gap, which we model with the BCS theory relation $\Delta(T)=\Delta_{\textsc{mar}}\sqrt{1-(T/T_{c})^{2}}$ with $T_{c}=6.0$ K. The conventional fitting procedure involves finding the value of $\phi$ that maximizes $I_{c}$ for each temperature.[70, 71, 72, 73]

Figure 3(c) shows the temperature dependence and fit of $I_{c}$ on device SNS-1, where the experimental data is normalized by the maximum current for $\mathcal{T}=100$% predicted by Eqn. (3). The experiment yielded a transparency of $\mathcal{T}=97.8$%. The same experiment was also performed on device SNS-3 (shown in Fig. S11 of the supplementary material), which yielded $\mathcal{T}=99.5$%. Both values are significantly larger than the transparencies obtained by the MAR analysis ($\mathcal{T}=78\pm 5$ %, $85\pm 2$%) in the same sample [see Table 1].

Regardless of which experimental method is used for measuring $\mathcal{T}$, our main result is that semiconductor-superconductor interface transparencies achieved in samples with a superconductor deposited post-growth can be comparable to those from “epitaxial” superconductors. This could dramatically expand the repertoire of possible superconductors available for realizing semiconductor-superconductor hybrid devices, potentially impacting the fields of topological quantum computing, superconducting qubits (via the gatemon design), and superconducting logic circuits.

In conclusion, we presented Josephson SNS junctions fabricated with ex-situ sputtered Nb contacts to 2DEGs hosted in InAs/AlGaSb surface quantum wells. We observed consistent and highly transparent interfaces with values of $\mathcal{T}$ ranging 78$-$99%. Post-growth superconducting contacts to InAs quantum wells can be a viable method on a par with epitaxial aluminum systems, and do not depend on unreasonably stringent fabrication parameters in the InAs material system.

supplementary material

The supplementary material contains additional information on MBE growth, bandstructure profiles, sample fabrication, characterization of Hall bars, and I-V/MAR traces of SNS junctions.

E.A.B., F.S., and A.E. contributed equally to this paper. The authors thank Kaveh Gharavi, Sean Walker, and Christine Nicoll for useful discussions. E.A.B. acknowledges support from a Mike and Ophelia Lazaridis Fellowship. This research was undertaken thanks in part to funding from the Canada First Research Excellence Fund (Transformative Quantum Technologies) and the Natural Sciences and Engineering Research Council (NSERC) of Canada. The University of Waterloo’s QNFCF facility was used for this work. This infrastructure would not be possible without the significant contributions of CFREF-TQT, CFI, ISED, the Ontario Ministry of Research and Innovation, and Mike and Ophelia Lazaridis. Their support is gratefully acknowledged.