Abstract
Superconductor/semiconductor hybrid devices have attracted increasing interest in the past years. Superconducting electronics aims to complement semiconductor technology, while hybrid architectures are at the forefront of new ideas such as topological superconductivity and protected qubits. In this work, we engineer the induced superconductivity in twodimensional germanium hole gas by varying the distance between the quantum well and the aluminum. We demonstrate a hard superconducting gap and realize an electrically and flux tunable superconducting diode using a superconducting quantum interference device (SQUID). This allows to tune the current phase relation (CPR), to a regime where single Cooper pair tunneling is suppressed, creating a \(\sin \left(2\varphi \right)\) CPR. Shapiro experiments complement this interpretation and the microwave drive allows to create a diode with ≈ 100% efficiency. The reported results open up the path towards integration of spin qubit devices, microwave resonators and (protected) superconducting qubits on the same silicon technology compatible platform.
Similar content being viewed by others
Introduction
III–V semiconductors have become the materials of choice for realizing highquality hybrid devices, due to the possibility of growing epitaxial Al on top of them^{1}. Gatetunable superconducting and Andreev spin qubits^{2,3,4,5,6}, parametric amplifiers^{7}, highly efficient Cooper pair splitters^{8,9,10} and a minimal Kitaev chain^{11} are prominent examples of what has been achieved in the past decade. In addition, nonreciprocal devices, such as superconducting diodes have attracted a lot of interest^{12}, especially in Josephson junctions in the presence^{13,14,15} or absence^{16,17,18} of a Zeeman field and in multiterminal devices^{19,20}. Diodes can be also realized in a superconducting quantum interference device (SQUID) geometry by exploiting a magnetic flux to achieve timereversal breaking^{21,22,23}. Such SQUIDs can be also used as a building block to create a protected superconducting qubit by engineering a \(\sin \left(2\varphi \right)\) current phase relation (CPR)^{24,25,26,27,28,29}.
One drawback of III–V materials is their nonzero nuclear spin, which, through hyperfine interaction, drastically reduces the electron spin coherence time, limiting therefore the use of hybrid devices in combination with the spin degree of freedom^{5,6}. Germanium, on the other hand, is a material which allows proximity induced superconductivity and has shown great potential for spin qubit devices^{30}. Induced superconductivity in germanium was first demonstrated in 0D and 1D systems^{31,32}. A few years later, superconductivity was also induced in a twodimensional Ge hole gas^{33,34}. Recent works demonstrated how induced superconductivity can be improved in planar germanium, either by using a double superconducting stack^{35} or by annealing platinum contacts^{36}. Here, using a shallow quantum well (QW) we establish Ge/SiGe heterostructures as an alternative platform to III–V materials for hybrid devices and microwave experiments, opening therefore the path to the coexistence of semiconductor and superconducting qubits.
Results
Material characterization and Josephson junctions
Compressively strained Ge QWs have been deposited on relaxed, linearly graded buffers with 70% Ge content. The 18 nm thick QWs are separated from the top surface by a spacer of thickness D. The builtin inplane compressive strain leads to charge confinement in the heavyhole band (see Fig. 1a). Mobility (μ_{h}) and mean free path (l_{h}) as a function of the carrier density n_{h} are displayed in Fig. 1b for a QW with ≈ 5 nm Si_{0.3}Ge_{0.7} spacer (D5  red) and ≈ 8 nm (D8  blue), respectively. At high density, D5 (D8) shows μ_{h} ≈ 10,000 cm^{2}/Vs (30,000 cm^{2}/Vs) and l_{h} ≈ 250 nm (700 nm).
For creating hybrid superconductorsemiconductor devices a thin film of aluminum ( ≈ 8 − 10 nm) is deposited ex situ and at low temperature on top of the Si_{0.3}Ge_{0.7} spacer (see Methods for the growth, the aluminum deposition and fabrication details). The Al has a polymorphic structure and it is not grown epitaxially on top of the Si_{0.3}Ge_{0.7} spacer. Importantly, energy dispersive Xray (EDX) data do not reveal interdiffusion of Al inside the Si_{0.3}Ge_{0.7} spacer and Ge QW, see Fig. 1c.
In order to check if the superconducting properties can leak into the Ge hole gas, Josephson field effect transistors (JoFETs) were fabricated (Fig. 1d). Representative V_{JJ} vs I_{JJ} traces, measured in a fourterminal currentbiased configuration (Fig. 1e), for D5 [red, upper plot] and D8 [blue, lower plot] are shown in Fig. 1f. The devices switch from superconducting to the dissipative regime at the gate tunable switching current I_{sw}. A common figure of merit used for estimating the quality of the proximity effect is the product between I_{sw} and the normal state resistance R_{N}. Figure 1g reports this product as a function of the gate voltage V_{g} for 6 different junctions with different dimensions. At high negative values of gate voltages, I_{sw}R_{N} spans from slightly below 100 μeV to above 400 μeV depending on D and on the JoFET dimensions. These values are favorably compared to previous results obtained with Ge heterostructures hybridized by Al^{33,34} and they are on par with more mature material systems^{37}.
In sample D5, the proximity effect is expected to be more effective because the Al is closer to the Ge hole gas. It is then surprising that the measured I_{sw}R_{N} product shown in Fig. 1g is significantly smaller for sample D5. One factor that could play a role is the fact that D8, especially at high density, is in the short ballistic regime (L < l_{h}, ξ_{N}), where I_{c}R_{N} is expected to be equal to πΔ/e^{38}; I_{c} is the critical current and \({\xi }_{{{{{{{{\rm{N}}}}}}}}}=\frac{{\hslash }^{2}\sqrt{2\pi {n}_{{{{{{{{\rm{h}}}}}}}}}}}{2{m}_{{{{{{{{\rm{eff}}}}}}}}}{{\Delta }}}\) is the superconducting coherence length in the quantum well with m_{eff} being the effective mass. Using n_{h} = 10^{12}cm^{−2}, Δ = 200 μeV and m_{eff} to be around 10% of the electron mass^{39}, we estimated ξ_{N} ≈ 500 nm. On the contrary, samples D5 have L ≈ l_{h}, which implies a smaller I_{c}R_{N}^{38}, making it challenging to compare the I_{c}R_{N} of the D8 and D5 devices directly. Moreover, the variations of I_{sw}R_{N} in D5 and D8 as a function of the JoFET dimensions is not fully understood. For these reasons, the I_{sw}R_{N} of such JoFET devices is not sufficient to characterize the quality of the proximity effect, especially because the switching current probability distribution is rather broad at low temperatures^{40}.
Tunability of the induced superconducting gap
The above reported results demonstrate that proximity induced superconductivity can be achieved in Ge without direct contact with the superconductor, therefore avoiding metallization issues^{41}. However, it is not clear to what extent the Si_{0.3}Ge_{0.7} spacer of thickness D is influencing the value of the induced superconducting gap Δ^{*} and the subgap density of states. We expect that Δ^{*} depends on the coupling t between the Al and the QW, see the sketch of Fig. 2a. Since Si_{0.3}Ge_{0.7} acts as a tunnel barrier, t should be strongly dependent on D. In other words, if D is very thin, we expect Δ^{*} to be similar to the parent gap of Al Δ; whereas if D is very thick the two layers will be very decoupled (small t) and Δ^{*} will be quenched.
In order to investigate the dependence of Δ^{*} on D, tunneling spectroscopy experiments were performed to estimate the local density of states of the hybridized Ge QW, see Fig. 2b, c for the experimental layout.
The differential conductance dI/dV, plotted in logarithmic scale, as a function of sourcedrain bias V and fingergates voltage V_{dg} for D8 is shown in Fig. 2d. dI/dV is suppressed symmetrically around V = 0 independently on V_{dg} (Fig. 2e); this signals the presence of a superconducting gap in the hybridized Ge hole gas. Interestingly, a double peak structure is revealed. The first peak appears at V ≈ 240μeV and the second at V ≈ 80 μeV. We interpret them as the parent gap of Al, Δ, and the induced gap in the Ge hole gas Δ^{*}. Their presence is more evident if dI/dV is plotted in a linear scale, like in Fig. 2f where the linecuts have been shifted vertically for the sake of clarity. The subgap conductance is suppressed by one order of magnitude compared to the abovegap value.
We now turn our attention to the dI/dV of sample D5 (Fig. 2g). The difference between Fig. 2g, d is striking. The region of suppressed conductance around V = 0 is larger in Fig. 2g, demonstrating that the induced gap is bigger. The linecuts, see Fig. 2h, showcase a difference between the normalstate conductance and the conductance at V = 0 of about two orders of magnitude, indicative of a hard gap^{1}. Also for this device the double peak structure is observed (Fig. 2h, i). The parent gap appears at a similar value, namely Δ ≈ 230μeV, while Δ^{*} ≈ 150μeV.
Superconducting diode effect
Having demonstrated a hard superconducting gap, we use the hybrid Al/Ge platform, to build a SQUID which acts as a gate/flux tunable superconducting diode^{12} and as a generator of nonsinusoidal currentphase relations (CPRs). The superconducting diode effect (SDE) can appear in a simple SQUID either if its inductance L is significant^{42,43} or if the CPRs of the single junctions have higher order contributions, arising from Andreev bound states in a semiconducting junction^{21,44} or from junctions in the dirty limit^{45,46}.
Figure 3a shows the schematics of a SQUID, with the underlying 4probe currentbiased electrical circuit. I_{sq} is the current passing through the SQUID, J is the current circulating in the SQUID and V_{sq} is the measured voltage drop across the device. In the following, we always sweep I_{sq} from positive/negative values to I_{sq} = 0, such that the retrapping current (I_{sq,+()}) is recorded for both branches. The use of the retrapping current avoids the challenges arising from the stochastic nature of the switching current^{40}. Top gate voltages V_{g1} and V_{g2} are used to tune the retrapping current I_{ret1} of JJ1 and I_{ret2} of JJ2. We first tune the device to be slightly unbalanced, i.e., I_{ret1} = 46.5nA ≠ I_{ret2} = 83.5 nA, see methods for understanding how I_{ret1} and I_{ret2} have been determined for the SQUID geometry. Figure 3b represents a SQUID measurement for such configuration, where V_{sq} is recorded as a function of I_{sq} and Φ. I_{sq,+()} is periodically modulated by Φ, as can be clearly seen in Fig. 3c. However, two features are observed, which are not expected for a negligible inductance SQUID composed by tunnel junctions. First, the retrapping current at \({{\Phi }}=\frac{{{{\Phi }}}_{0}}{2}\) is expected to be ∣I_{ret1} − I_{ret2}∣ = 37nA (see brown horizontal line in Fig. 3b), instead the measured value is around 52nA. Moreover, the SQUID pattern is not symmetric with respect to \({{\Phi }}=\frac{{{{\Phi }}}_{0}}{2}\), see green arrows in Fig. 3b. This asymmetry gives rise to a finite SDE. The diode efficiency, defined as \(\eta=\frac{{I}_{{{{{{\rm{sq}}}}}},+} {I}_{{{{{{\rm{sq}}}}}},} }{{I}_{{{{{{\rm{sq}}}}}},+}+ {I}_{{{{{{\rm{sq}}}}}},} }\), is shown in Fig. 3d. In particular, η = 0 at integer (Φ = nΦ_{0}) and it changes its sign around \({{\Phi }}=\frac{n}{2}{{{\Phi }}}_{0}\). The maximum value observed for this device is around 15%.
In order to understand these results, we solve the static equation of the system:
I_{JJ1} [I_{JJ2}] is the current flowing through JJ1 [JJ2] which depends on the phase difference across the junction φ_{1} [φ_{2}]. The phase drops are related to the fluxoid quantization:
For a given Φ, I_{sq,+} [I_{sq,}] is obtained by finding the maximum [minimum] I_{sq} with respect to φ_{1}.
First, we attempt to understand our results assuming standard sinusoidal CPRs, i.e., \({I}_{{{{{{{{\rm{JJ1}}}}}}}}}\left({\varphi }_{1}\right)={I}_{{{{{{{{\rm{ret1}}}}}}}}}\sin \left({\varphi }_{1}\right)\) and \({I}_{{{{{{{{\rm{JJ2}}}}}}}}}\left({\varphi }_{1}\right)={I}_{{{{{{{{\rm{ret2}}}}}}}}}\sin \left({\varphi }_{2}\right)\), and adding the inductive contribution. L is composed of two terms, a geometric one L_{geo} and a kinetic one L_{kin}; we extracted L = 110pH, see methods for details. The orange traces in Fig. 3c, d represent the theoretical prediction. It is clear that the mere addition of a realistic L does not capture the full picture, especially around \(\frac{{{\Phi }}}{{{{\Phi }}}_{0}}=0.5\), where I_{sq,+} and ∣I_{sq,}∣ are greatly underestimated, see Fig. S2.
Therefore, it is necessary to consider higher order harmonics for explaining our results, namely we assume that our single junction CPRs are given by:
where K_{n} is the relative contribution of the nth harmonic and we assumed that the harmonics’ contribution is the same for both junctions. α_{1} [α_{2}] is a dimensionless parameter which is adjusted such that \(\max {I}_{{{{{{{{\rm{JJ1}}}}}}}}}={I}_{{{{{{{{\rm{ret1}}}}}}}}}\) [\(\max {I}_{{{{{{{{\rm{JJ2}}}}}}}}}={I}_{{{{{{{{\rm{ret2}}}}}}}}}\)].
The red traces in Fig. 3c, d are the outcome of a numerical fit using up to eight harmonic contributions. It is found that K_{1} = 0.66, K_{2} = 0.12 and K_{3} = 0.10, while higher order terms are smaller than 10%, see Fig. S2 to understand the effect of higher order terms. We point out that also asymmetric cases would give qualitatively similar results (Fig. S3), but the amount of free parameters of the fit would increase considerably.
We now turn our attention to the gate dependence of the SDE. Our measurements show that the SDE can be tuned by the gate voltages V_{g1} and V_{g2}. In Fig. 3e, we fix V_{g1} = − 1.5V and we study the behavior of η while varying Φ and V_{g2}. When \({V}_{{{{{{{{\rm{g}}}}}}}}2}\, > \, {V}_{{{{{{{{\rm{g}}}}}}}}1}\), η > 0 [η < 0] for \(\frac{{{\Phi }}}{{{{\Phi }}}_{0}} \, > \, 0.5\) [\(\frac{{{\Phi }}}{{{{\Phi }}}_{0}} \, < \, 0.5\)], while the trend is opposite if \({V}_{{{{{{{{\rm{g}}}}}}}}2}\, < \, {V}_{{{{{{{{\rm{g}}}}}}}}1}\). In other words, we have an inversion of the diode polarity going from one regime to the other and most importantly, the SDE completely vanishes independently of Φ when the two junctions are fully balanced (V_{g1} ≈ V_{g2}, i.e., I_{ret1} = I_{ret2} = I_{ret}).
I_{sq,+}, plotted in Fig. 3f, does not vanish even at half flux quantum and for balanced junctions, see the pink cross in Fig. 3f; we refer to this condition as the sweet spot. This is a crucial aspect, because at the sweet spot, the first harmonic of I_{sq} (\(\propto \sin \left(\varphi \right)\)) is completely suppressed but not the higherorder terms. We can understand this from considering Eqs. (1) at the sweet spot,
where for simplicity the inductance and higher order terms are neglected and α_{1} = α_{2} = α. Therefore the CPR of the SQUID would be \({I}_{{{{{{{{\rm{sq}}}}}}}}}\left({\varphi }_{1}\right)=\!2\alpha {K}_{2}{I}_{{{{{{{{\rm{ret}}}}}}}}}\sin \left(2{\varphi }_{1}\right)\). This CPR corresponds to transport through the SQUID being governed by pairs of Cooper pairs, while the exchange of single pairs is quenched. This is the condition needed for creating a certain type protected qubit^{24}.
This behavior can be further understood by solving Eqs. (1), (2) and assuming to have JJs with higher order contributions. Figure 3g [h] represents the theoretical calculation of η [I_{sq,+}] for a SQUID with the parameters extracted from the fit of Fig. 3c, d. The simulation results agree well on a qualitative level with the measurements. From the theoretical calculation it is possible to calculate the CPR of the SQUID and express it as a Fourier expansion:
where b_{n} and a_{n} represents the nth harmonic contribution and φ is the phase drop across the SQUID. The values of the first harmonic terms b_{1}, a_{1} and second harmonic term b_{2} obtained from numerical simulations are shown in Fig. 3i–k.
Next, we show the theoretical prediction of the ratio between the second and first harmonic, i.e., \(K=\frac{ {b}_{2}+ {a}_{2} }{ {b}_{1}+ {a}_{1} }\), see Fig. 3l. The red trace depicts the points where the first and second harmonics equally contribute to the SQUID CPR, whereas the ratio diverges close to the sweet spot.
We note that, different CPRs of the single Josephson junctions would give slightly different outcomes. However, it would not change the main conclusion that b_{1} and a_{1} can be completely suppressed. Moreover, the first harmonic contribution can be suppressed over a broad range of gate space, which also allows to tune the second harmonic contribution (Fig. S4).
Finally, we note that the first harmonic can be quenched by just having a high inductance and the possibility of tuning the critical currents, see Figs. S5,S6 and S7.
Halfinteger Shapiro steps and ideal SDE
The good qualitative match between the experiment and the theoretical prediction of Fig. 3 makes us confident in our interpretation of the diode data, but the CPR of the SQUID was not directly probed. The AC Josephson effect would help to further elucidate the CPR periodicity. In fact, for a standard sinusoidal CPR under microwave irradiation, the currentvoltage characteristics develop voltage steps when \(V=s\frac{h{f}_{{{{{{{{\rm{ac}}}}}}}}}}{2e}\), the socalled Shapiro steps, where s = 0, 1, 2, … and f_{ac} is the external applied frequency. On the contrary, if the CPR becomes \(\propto \sin \left(2\varphi \right)\), signaling tunneling of pairs of Cooper pairs, steps at halfinteger values also appear, i.e., s = 0, 0.5, 1, … ^{47,48,49}. In our case, we expect the ratio between the second and first harmonic to be maximized when the SQUID is balanced and \({{\Phi }} \, \approx \, \frac{{{{\Phi }}}_{0}}{2}\), see Fig. 3l as an example. Therefore, we would expect to observe halfinteger Shapiro steps when approaching the sweet spot^{21}.
In order to avoid flux generated by inductive effects which might lead to similar results^{50}, we present results of a 30 nmthick aluminum SQUID, which has a much smaller inductance (L < 15pH). Furthermore, a shunt resistor R_{shunt} of 1050 Ω, see Fig. 4a, was added in order to create overdamped junctions, allowing therefore to measure Shapiro steps at small external frequencies, avoiding issues related to LandauZener transitions^{51}.
In the following, we study a SQUID in a balanced configuration (I_{ret1} = I_{ret2}) subjected to an external drive at f_{ac} = 500 MHz. Figure 4b shows the differential resistance of the SQUID dV_{sq}/dI_{sq} as a function of the microwave drive power P and I_{sq} at \(\frac{{{\Phi }}}{{{{\Phi }}}_{0}}=0.58\). If P is high enough, dips corresponding to the integer Shapiro steps s = 1, 2, 3 appear. Similar results are obtained at \(\frac{{{\Phi }}}{{{{\Phi }}}_{0}}=0.42\), see Fig. 4c. However, the situation is different if \(\frac{{{\Phi }}}{{{{\Phi }}}_{0}}=0.5\) (Fig. 4d) a condition for which the first harmonic term should vanish. For this situation, the first halfinteger steps appears as theoretically expected (see Fig. S10 for the identification of the Shapiro steps).
This behavior is summarized in Fig. 4e where we fix P = 9dBm and we display dV_{sq}/dI_{sq} as a function of I_{sq} and Φ. Far from half flux quantum, dips corresponding to integer Shapiro steps are observed; while close to \(\frac{{{\Phi }}}{{{{\Phi }}}_{0}}=0.5\) the integer steps fade and the s = 0.5 step becomes pronounced, see white numbers. In order to further investigate the range over which the halfinteger Shapiro step is visible, we fix \(\frac{{{\Phi }}}{{{{\Phi }}}_{0}}=0.5\) and we vary I_{ret2} with the gate voltage V_{g2}. When the SQUID is close to the balanced regime (V_{g2} ≈ 0.1V) the first halfinteger step is evident (white numbers); however it fades away if the SQUID becomes unbalanced, i.e., V_{g2} > 0.2V. As expected from the previous analysis, the halfinteger step appears only if the device is close to the balanced position and close to half flux quantum when \({I}_{{{{{{{{\rm{sq}}}}}}}}}\left({\varphi }_{1}\right)={b}_{2}\sin \left(2{\varphi }_{1}\right)+{b}_{4}\sin \left(4{\varphi }_{1}\right)+\ldots \,\). Importantly, also a second device investigated under microwave irradiation showed the same behavior, see Fig. S10.
The SDE indicates that the symmetry I_{sq,+} = − I_{sq,} is broken in our system, which also implies that the widths ΔI_{±1} of the two first Shapiro steps, which eventually define I_{sq,+} and I_{sq,}, can be different^{21}. As the position and the width of the plateaus depend on the microwave drive, one can envision tuning to a situation in which the first negative plateau would start at zero current (I_{sq,} = 0) while the first positive one at a finite current (I_{sq,+} ≠ 0). At this particular strength of the ac driving, the SQUID is expected to become an ideal SDE, i.e., η ≈ 1, see theoretical analysis in ref. ^{52}.
In order to investigate this possibility a similar SQUID, with R_{shunt} = 50Ω, at Φ = 0.39Φ_{0} for different drive powers P was investigated (Fig. 4g). For small P, I_{sq,+} > ∣I_{sq,}∣ and η ≈ 0.18, see blue trace. When P increases both I_{sq,+} and ∣I_{sq,}∣ decrease, see orange trace. Eventually when P is high enough (green trace), I_{sq,} drops to zero, whereas I_{sq,+} ≠ 0, yielding a diode efficiency equal to 1. Moreover, we show that our device is nonvolatile, namely we can switch several times from the normalstate to the superconducting branch by changing the current direction, see Fig. 4h.
Discussion
In the past few years, planar germanium has established itself as a promising platform for spinqubit arrays^{30}. Here, we demonstrate its potential also for hybrid semiconductorsuperconductor quantum devices. Inspired by more mature technologies^{37}, we introduced a reliable way to induce superconductivity by using shallow QWs and, to the best of our knowledge, we have realized the largest hard gap in Ge. Our method does not rely on the precise etching of the QW and/or surface treatments^{35} and does not require insitu deposition of the superconductor. Furthermore, it minimizes the Fermi velocity mismatch due to the direct contact between Ge and proximitized Ge, enhancing Andreev reflection over normal reflection.
While the shallow QWs reported in this work are of limited mobility and have a larger charge noise, which can be a challenge for the realization of scalable spin qubits, possible mitigation strategies of this problem could include a careful engineering of the semiconductor/dielectric interface^{53}, including the use of Ge caps^{54}, or growing the QWs on Ge instead of Si wafers^{55}. A further solution could be to have a thin spacer in the areas where superconductivity should be induced and a thicker one in the areas where the spin qubits will be formed.
The reported large superconducting hard gap on a group IV material will enable spin qubit coupling via coherent tunneling and cotunneling processes that involve (crossed) Andreev reflection^{56,57}. In addition, the realized gate and fluxtunable superconducting diode can suppress the first harmonic term, making it therefore an interesting building block for creating protected superconducting qubits with semiconductor materials^{24,25,26,27,28}. In order to realize such qubits, superconducting resonators are key elements. A λ/4 notchtype resonator is shown in (Fig. 5a) (see methods for details). The upperleft inset of Fig. 5a depicts the crosssection of the resonator, pointing out that the Ge QW has been completely etched away prior to the resonator fabrication. Figure 5b shows the transmission amplitude \(\left{S}_{21}\right\) as a function of the probe frequency f. The internal quality factors Q_{i} were extracted^{58} and found to be around 7000 [20000] for 〈n_{ph}〉 ≈ 1 [100]; demonstrating the microwave compatibility of the used Ge/SiGe heterostructures. Interestingly, just slightly smaller Q_{i} values were extracted also for superconducting resonators fabricated on Ge/SiGe heterostructures where the Ge QW has been removed just in the gap between the central conductor and the ground plane, showing that the proximitized Ge does not lead to significant losses. The above demonstrated microwave compatibility of the used Ge/SiGe heterostructures opens a path towards spinphoton experiments^{59}, gate tunable transmon qubits^{2,3,4} and superconducting spin qubits in group IV materials^{6} and allow us to envision the transfer of quantum information between different types of qubits, all realizable on planar Ge.
After submission of our manuscript we became aware of similar works dealing with the superconducting diode effect in interferometer devices^{60,61,62}.
Methods
Growth and Al deposition
Strained Ge QW structures were grown by lowenergy plasmaenhanced chemical vapor deposition on forwardgraded buffers^{63} with Si_{0.3}Ge_{0.7} caps of 5 and 8 nm above the 18 nm Ge QW. These nominal QW and cap thicknesses vary across the wafer due to the intensity profile of the focused plasma. Thicknesses were verified by comparing highresolution xray diffraction ω–2θ scans with dynamical simulations based on a smoothed QW profile^{63}. This same composition profile was constructed within the NextNano 1d PoissonSchrödinger solver, along with a dieletric layer and top Schottky contact, in order to generate the band profile and wavefunction density shown in Fig. 1a. The Ge/SiGe heterostructures are cut in pieces of 6x6 mm^{2}. A \(3\min\) buffered oxide etch (7:1 ratio) removes the native surface oxide of the diced samples, after which they are transferred to a molecular beam epitaxy chamber. The samples are cooled down to 110K by active liquidnitrogen cooling. Subsequently, Al is deposited at a growth rate of 5.5 Å/min. Immediately after growth, samples are transferred insitu to a chamber equipped with an ultrahighpurity O_{2} source where they are exposed to 10^{−4}mbar of O_{2} for \(15\min\). The formed oxide layer prevents subsequent retraction of the metal film as the sample warms up to room temperature under ultrahigh vacuum conditions.
Sample fabrication
10 nmthick, cold deposited aluminum samples
A mesa of around 60nm depth is obtained by first removing Al with Transene D and then by etching the heterostructure with a SF_{6}O_{2}CHF_{3} reactive ion etching process. In a second step, Al is selectively etched away using Transene D in order to create the Josephson junction or tunneling spectroscopy devices. Then, for tunneling spectroscopy devices, normal metal ohmic contacts are created by argon milling the SiGe spacer followed by a deposition of 60 nm platinum at an angle of 5^{∘}. Finally, 9–18 nm plasma assisted aluminum oxide is deposited on top of all the sample at 150 ^{∘}C and then Ti/Pd gates are evaporated. For some devices, two layers of topgates were needed.
30 nmthick, roomtemperature deposited aluminum samples
A mesa of around 60nm depth is obtained by etching the heterostructure with a SF_{6}O_{2}CHF_{3} reactive ion etching process. The sample is then submerged for 15s in buffered HF and, subsequently, the 30nm Al film is deposited. Gates are patterned like for the 10nmthick sample. Importantly, this technique allows to fabricate devices without the need of wet etching for removing the superconductor.
CPW resonator, without QW
The QW is removed by a SF_{6}O_{2}CHF_{3} reactive ion etching process. Subsequently, the CPW resonator, the feedline and the ground plane are written by electron beam lithography followed by a 25 nmthick Al deposition at room temperature.
CPW resonator, with QW
Electron beam lithography is performed on a sample with low temperature deposited aluminum. The area between the ground plane and the signal line is exposed and after development the Al is removed by transene D etching. Finally, and before removing the resist the Ge QW is etched away by a SF_{6}O_{2}CHF_{3} reactive ion etching process.
Inductance estimation
The total inductance L of the measured SQUIDs has two contributions: the geometric one (L_{geo}) and the kinetic one (L_{kin}). For the L_{geo} we approximated the device to a loop with a radius R and with the wire diameter d
where μ_{0} is the magnetic vacuum permeability and we assumed the relative magnetic permeability μ_{r} = 1. Typical values of our SQUID geometry are R ≈ 1.25 μm and d ≈ 0.7 μm, which gives L_{geo} ≈ 2pH. However, in order not to underestimate this contribution, L_{geo} is assumed to be as large as ≈ 5pH. As regards the kinetic inductance per square L_{kin,□}, it was estimated from the values of the superconducting gap and the square normalstate resistance R_{□}^{64}:
where Δ is estimated from the critical temperature T_{c}, i.e., Δ = 1.76k_{B}T_{c} with k_{B} being the Boltzmann constant. The results are summarized in Table 1.
Estimation of the retrapping currents in the SQUID geometry
We note that the sum of the retrapping currents measured in isolation (I_{ret1,iso} and I_{ret2,iso}), i.e., with the other junction pinched off, in the absence of a shunt resistor is always smaller than the retrapping current of the squid at Φ = 0. This difference is attributed to the fact that the SQUID has a smaller resistance, which leads to lower dissipation^{65} and, as a result, a higher retrapping current. Therefore, we assume an even redistribution of retrapping currents such that \(\frac{{I}_{{{{{{{\rm{ret1}}}}}},iso}}}{{I}_{{{{{{{{\rm{ret2,iso}}}}}}}}}}=\frac{{I}_{{{{{{{{\rm{ret1}}}}}}}}}}{{I}_{{{{{{{{\rm{ret2}}}}}}}}}}\) and \({I}_{{{{{{{{\rm{ret}}}}}{1}}}}}+{I}_{{{{{{{{\rm{ret}}}}}{2}}}}}={I}_{{{{{{\rm{sq}}}}}},+}\left({{\Phi }}=0\right)\). This approach was used to estimate the retrapping currents for Fig. 3.
Data availability
All experimental data included in this work are available at https://zenodo.org/records/10119346.
References
Krogstrup, P. et al. Epitaxy of semiconductor–superconductor nanowires. Nat. Mater. 14, 400–406 (2015).
Larsen, T. W. et al. Semiconductornanowirebased superconducting qubit. Phys. Rev. Lett. 115, 127001 (2015).
De Lange, G. et al. Realization of microwave quantum circuits using hybrid superconductingsemiconducting nanowire Josephson elements. Phys. Rev. Lett. 115, 127002 (2015).
Casparis, L. et al. Superconducting gatemon qubit based on a proximitized twodimensional electron gas. Nat. Nanotechnol. 13, 915–919 (2018).
Hays, M. et al. Coherent manipulation of an Andreev spin qubit. Science 373, 430–433 (2021).
PitaVidal, M. et al. Direct manipulation of a superconducting spin qubit strongly coupled to a transmon qubit. Nat. Phys. 1–6 (2023).
Phan, D. et al. Gatetunable superconductorsemiconductor parametric amplifier. Phys. Rev. Appl. 19, 064032 (2023).
Wang, G. et al. Singlet and triplet Cooper pair splitting in hybrid superconducting nanowires. Nature 612, 448–453 (2022).
Bordoloi, A., Zannier, V., Sorba, L., Schönenberger, C. & Baumgartner, A. Spin crosscorrelation experiments in an electron entangler. Nature 612, 454–458 (2022).
Wang, Q. et al. Triplet Cooper pair splitting in a twodimensional electron gas. arXiv preprint arXiv:2211.05763 (2022).
Dvir, T. et al. Realization of a minimal Kitaev chain in coupled quantum dots. Nature 614, 445–450 (2023).
Ando, F. et al. Observation of superconducting diode effect. Nature 584, 373–376 (2020).
Baumgartner, C. et al. Supercurrent rectification and magnetochiral effects in symmetric Josephson junctions. Nat. Nanotechnol. 17, 39–44 (2022).
Pal, B. et al. Josephson diode effect from Cooper pair momentum in a topological semimetal. Nat. Phys. 18, 1228–1233 (2022).
Mazur, G. et al. The gatetunable Josephson diode. arXiv preprint arXiv:2211.14283 (2022).
Steiner, J. F., Melischek, L., Trahms, M., Franke, K. J. & von Oppen, F. Diode effects in currentbiased Josephson junctions. Phys. Rev. Lett. 130, 177002 (2023).
Wu, H. et al. The fieldfree Josephson diode in a van der Waals heterostructure. Nature 604, 653–656 (2022).
Trahms, M. et al. Diode effect in Josephson junctions with a single magnetic atom. Nature 615, 628–633 (2023).
Gupta, M. et al. Gatetunable superconducting diode effect in a threeterminal Josephson device. Nat. Commun. 14, 3078 (2023).
Chiles, J. et al. Nonreciprocal supercurrents in a fieldfree graphene Josephson triode. Nano Lett. 23, 5257–5263 (2023).
Souto, R. S., Leijnse, M. & Schrade, C. Josephson diode effect in supercurrent interferometers. Phys. Rev. Lett. 129, 267702 (2022).
Fominov, Y. V. & Mikhailov, D. Asymmetric higherharmonic squid as a Josephson diode. Phys. Rev. B 106, 134514 (2022).
Ciaccia, C. et al. Gatetunable Josephson diode in proximitized InAs supercurrent interferometers. Phys. Rev. Res. 5, 033131 (2023).
Smith, W., Kou, A., Xiao, X., Vool, U. & Devoret, M. Superconducting circuit protected by twoCooperpair tunneling. npj Quant. Inf. 6, 8 (2020).
Brooks, P., Kitaev, A. & Preskill, J. Protected gates for superconducting qubits. Phys. Rev. A 87, 052306 (2013).
Gyenis, A. et al. Experimental realization of a protected superconducting circuit derived from the 0–π qubit. PRX Quantum 2, 010339 (2021).
Larsen, T. W. et al. Parityprotected superconductorsemiconductor qubit. Phys. Rev. Lett. 125, 056801 (2020).
Schrade, C., Marcus, C. M. & Gyenis, A. Protected hybrid superconducting qubit in an array of gatetunable Josephson interferometers. PRX Quantum 3, 030303 (2022).
Maiani, A., Kjaergaard, M. & Schrade, C. Entangling transmons with lowfrequency protected superconducting qubits. PRX Quantum 3, 030329 (2022).
Scappucci, G. et al. The germanium quantum information route. Nat. Rev. Mater. 6, 926–943 (2021).
Xiang, J., Vidan, A., Tinkham, M., Westervelt, R. M. & Lieber, C. M. Ge/Si nanowire mesoscopic Josephson junctions. Nat. Nanotechnol. 1, 208–213 (2006).
Katsaros, G. et al. Hybrid superconductor–semiconductor devices made from selfassembled SiGe nanocrystals on silicon. Nat. Nanotechnol. 5, 458–464 (2010).
Hendrickx, N. W. et al. Gatecontrolled quantum dots and superconductivity in planar germanium. Nat. Commun. 9, 2835 (2018).
Vigneau, F. et al. Germanium quantumwell Josephson fieldeffect transistors and interferometers. Nano Lett. 19, 1023–1027 (2019).
Aggarwal, K. et al. Enhancement of proximityinduced superconductivity in a planar Ge hole gas. Phys. Rev. Res. 3, L022005 (2021).
Tosato, A. et al. Hard superconducting gap in germanium. Commun. Mater. 4, 23 (2023).
Mayer, W. et al. Superconducting proximity effect in epitaxial AlInAs heterostructures. Appl. Phys. Lett. 114, 103104 (2019).
Likharev, K. Superconducting weak links. Rev. Mod. Phys. 51, 101 (1979).
Rössner, B., Isella, G. & von Känel, H. Effective mass in remotely doped Ge quantum wells. Appl. Phys. Lett. 82, 754–756 (2003).
Haxell, D. Z. et al. Measurements of phase dynamics in planar Josephson junctions and squids. Phys. Rev. Lett. 130, 087002 (2023).
Reeg, C., Loss, D. & Klinovaja, J. Metallization of a Rashba wire by a superconducting layer in the strongproximity regime. Phys. Rev. B 97, 165425 (2018).
Fulton, T., Dunkleberger, L. & Dynes, R. Quantum interference properties of double Josephson junctions. Phys. Rev. B 6, 855 (1972).
Paolucci, F., De Simoni, G. & Giazotto, F. A gateand fluxcontrolled supercurrent diode effect. Appl. Phys. Lett. 122, 042601 (2023).
Legg, H. F., Laubscher, K., Loss, D. & Klinovaja, J. Parity protected superconducting diode effect in topological Josephson junctions. arXiv preprint arXiv:2301.13740 (2023).
Lambert, C., Raimondi, R., Sweeney, V. & Volkov, A. Boundary conditions for quasiclassical equations in the theory of superconductivity. Phys. Rev. B 55, 6015 (1997).
Galaktionov, A. & Ryu, C.M. The second Josephson harmonic in the dirty limit. J. Phys.: Condens. Matter 12, 1351 (2000).
Ueda, K. et al. Evidence of halfinteger Shapiro steps originated from nonsinusoidal current phase relation in a short ballistic InAs nanowire Josephson junction. Phys. Rev. Res. 2, 033435 (2020).
Zhang, P. et al. Large secondorder Josephson effect in planar superconductorsemiconductor junctions. arXiv preprint arXiv:2211.07119 (2022).
Iorio, A. et al. Halfinteger Shapiro steps in highly transmissive InSb nanoflag Josephson junctions. Phys. Rev. Res. 5, 033015 (2023).
Vanneste, C. et al. Shapiro steps on currentvoltage curves of dc squids. J. Appl. Phys. 64, 242–245 (1988).
Dartiailh, M. C. et al. Missing Shapiro steps in topologically trivial Josephson junction on InAs quantum well. Nat. Commun. 12, 78 (2021).
Souto, R. S. et al. Tuning the Josephson diode response with an ac current. arXiv preprint arXiv:2312.09204. (2023).
Degli Esposti, D. et al. Waferscale lowdisorder 2deg in 28Si/SiGe without an epitaxial Si cap. Appl. Phys. Lett. 120, 184003 (2022).
Su, Y.H., Chuang, Y., Liu, C.Y., Li, J.Y. & Lu, T.M. Effects of surface tunneling of twodimensional hole gases in undoped Ge/GeSi heterostructures. Phys. Rev. Mater. 1, 044601 (2017).
Stehouwer, L. E. A. et al. Germanium wafers for strained quantum wells with low disorder. Appl. Phys. Lett. 123, 092101 (2023).
Leijnse, M. & Flensberg, K. Coupling spin qubits via superconductors. Phys. Rev. Lett. 111, 060501 (2013).
Spethmann, M., Bosco, S., Hofmann, A., Klinovaja, J. & Loss, D. Highfidelity twoqubit gates of hybrid superconductingsemiconducting singlettriplet qubits. arXiv preprint arXiv:2304.05086 (2023).
Probst, S., Song, F., Bushev, P. A., Ustinov, A. V. & Weides, M. Efficient and robust analysis of complex scattering data under noise in microwave resonators. Rev. Sci. Instrum. 86, 024706 (2015).
Burkard, G., Gullans, M. J., Mi, X. & Petta, J. R. Superconductor–semiconductor hybridcircuit quantum electrodynamics. Nat. Rev. Phys. 2, 129–140 (2020).
Ciaccia, C. et al. Charge4e supercurrent in an InAsAl superconductorsemiconductor heterostructure. arXiv preprint arXiv:2306.05467 (2023).
Li, Y. et al. Interfering Josephson diode effect and magnetochiral anisotropy in Ta2Pd3Te5 asymmetric edge interferometer. arXiv preprint arXiv:2306.08478 (2023).
Matsuo, S. et al. Josephson diode effect derived from shortrange coherent coupling. Nat. Phys. 19, 1636–1641 (2023).
Jirovec, D. et al. A singlettriplet hole spin qubit in planar Ge. Nat. Mater. 20, 1106–1112 (2021).
Annunziata, A. J. et al. Tunable superconducting nanoinductors. Nanotechnology 21, 445202 (2010).
Courtois, H., Meschke, M., Peltonen, J. & Pekola, J. P. Origin of hysteresis in a proximity Josephson junction. Phys. Rev. Lett. 101, 067002 (2008).
Acknowledgements
We acknowledge Alexander Brinkmann, Alessandro Crippa, Francesco Giazotto, Andrew Higginbotham, Andrea Iorio, Giordano Scappucci, Christian Schonenberger, and Lukas Splitthoff for helpful discussions. We thank Marcel Verheijen for the support in the TEM analysis. This research and related results were made possible with the support of the NOMIS Foundation. It was supported by the Scientific Service Units of ISTA through resources provided by the MIBA Machine Shop and the nanofabrication facility, the European Union’s Horizon 2020 research and innovation programme under Grant Agreement No 862046, the HORIZONRIA 101069515 project, the European Innovation Council Pathfinder grant no. 101115315 (QuKiT), and the FWF Projects #P32235, #P36507 and #F8606. For the purpose of open access, the authors have applied a CC BY public copyright licence to any Author Accepted Manuscript version arising from this submission. R.S.S. acknowledges Spanish CM “Talento Program" Project No. 2022T1/IND24070. J.J. acknowledges European Research Council TOCINA 834290.
Author information
Authors and Affiliations
Contributions
M.V. and T.G. fabricated the transport devices and performed the measurements under the supervision of G.K. M.V. did the data analysis under the supervision of G.K. L.B. and O.S. fabricated and measured the CPW. M.J. and O.S. developed the microwave technology for the Ge/SiGe heterostructures. K.A. contributed to the transport measurements and the device fabrication. J.A.S. fabricated the Hall bars. T. A. made the Shapiro measurements possible. M.V. performed the simulations with input from C.S. and R.S.S. S.C., A.B., D.C., and G.I. were responsible for the Ge QW growth, Hall bar measurements, and NextNano simulations. J.J. and E.B. were responsible for the growth of lowtemperature Al and TEM data. M.L. and J.D. contributed to the interpretation of the experimental results. M.V. and G.K. wrote the manuscript with input from all the coauthors.
Corresponding authors
Ethics declarations
Competing interests
The authors declare no competing interests.
Peer review
Peer review information
Nature Communications thanks Nicola Paradiso, and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.
Additional information
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Supplementary information
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Valentini, M., Sagi, O., Baghumyan, L. et al. Parityconserving Cooperpair transport and ideal superconducting diode in planar germanium. Nat Commun 15, 169 (2024). https://doi.org/10.1038/s41467023441140
Received:
Accepted:
Published:
DOI: https://doi.org/10.1038/s41467023441140
This article is cited by

Strong tunable coupling between two distant superconducting spin qubits
Nature Physics (2024)
Comments
By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.