betaFeedback

WorksPhysics and Astronomy

Noise-Protected Superconducting Qudits from Higher-Harmonic Josephson Junctions

Abstract

Qudits, which generalize qubits to dd levels, store more information per circuit element than qubits. As with qubits, however, their sensitivity to noise must be overcome before they can be used for fault-tolerant quantum computing. Here we show that, by exploiting the second harmonic of the current–phase relation, a superconducting loop of three Josephson junctions can host qutrits strongly protected against relaxation, while dephasing limits their coherence. When the second harmonic is positive, a simple loop with a small asymmetry between the junctions provides a tunable, symmetry-protected qutrit. When it is negative, a protected qutrit can be realized by adding a small symmetry-breaking term, such as a weak Josephson diode effect. The two designs involve different trade-offs: the first has an exact charge selection rule, whereas the second is nearly insensitive to charge noise and its quasiparticle errors can be corrected. The preferable choice therefore depends on the materials used. Both require a sizable second harmonic to begin with, and we identify altermagnetic Josephson junctions as a candidate platform. Finally, microscopic calculations show that a gate voltage selects the sign of the second harmonic in these junctions, although gate-voltage noise then becomes the leading source of dephasing. These findings establish the sign of the second harmonic as a design principle for protected superconducting qudits.

Superconducting circuits are among the leading candidates for implementing qubits, with Josephson junctions providing the anharmonicity needed to address them. They combine this strong nonlinearity with lithographic scalability and fast electrical control [1, 2, 3]. However, their coherence is limited by dielectric loss, quasiparticles, and charge and flux fluctuations. One route toward fault tolerance is to build noise suppression into the circuit Hamiltonian, so that the computational subspace couples weakly to the dominant noise operators while remaining addressable [4, 5, 6, 7]. An example is a Josephson element whose current–phase relation (CPR) carries a sizable second harmonic, which transfers pairs of Cooper pairs coherently [8, 9]. Such elements are obtained by engineered interference [10, 11, 12, 13, 14], or natively in twisted dd-wave junctions near a 45∘45^\circ [15, 16, 17, 18, 19, 20, 21] and in altermagnetic junctions near a 00–π\pi transition [22, 23, 24, 25]. Devices built on them have so far encoded qubits [8, 12, 13, 26, 27, 28, 29, 30, 31], and their protection has been argued from charge and quasiparticle matrix elements between the computational states. Extending protection to qudits increases the information encoded per circuit element and enriches the routes to error correction [32, 33, 34]. Here we show that a three-junction loop of such elements hosts protected qutrits (d=3d=3 qudits) in both positive and negative sign classes of the second CPR harmonic, with the first harmonic taken positive. In the positive class, the junctions need only a sizable second harmonic and a small asymmetry between them. In the negative class, the junctions must also carry small, unequal energy terms odd in the phase, as in a Josephson diode. These can be supplied by intrinsically nonreciprocal junctions [21, 35, 36, 37] or auxiliary side junctions. We find that dephasing rather than relaxation limits both classes, and that the sign class decides whether quasiparticle errors can be corrected.

We take the harmonics of each junction as given. A junction with phase δ\delta across it has the energy U(δ)=−∑m≥1Emcos⁡(mδ)U(\delta)=-\sum_{m\geq1}E_m\cos(m\delta) with harmonic vector (E1,E2,…)(E_1,E_2,\ldots), and its sign class is set by the ratio

η≡E2∣E1∣=I22∣I1∣,(1)\eta\equiv\frac{E_2}{\lvert E_1\rvert}=\frac{I_2}{2\lvert I_1\rvert}, \tag*{(1)}

with ImI_m the CPR harmonics, η>0\eta>0 the positive class and η<0\eta<0 the negative one. The first harmonic enters through its magnitude because shifting δ\delta by π\pi changes the sign of E1E_1 but not E2E_2. We measure energies in units of ∣E2∣\lvert E_2\rvert.

**Circuit.—**The circuit is a loop of three branches, each a Josephson element shunted by a capacitor of charging energy ECE_C [Fig. 1(b)]. The phases across the elements are δ1=ϕ1\delta_1=\phi_1, δ2=ϕ2−ϕ1\delta_2=\phi_2-\phi_1 and δ3=ϕext−ϕ2\delta_3=\phi_{\mathrm{ext}}-\phi_2, with node phases ϕ1,2\phi_{1,2} and reduced flux ϕext=2πΦext/Φ0\phi_{\mathrm{ext}}=2\pi\Phi_{\mathrm{ext}}/\Phi_0 [38, 39, 40]. The Hamiltonian is

H^=4(n^−ng)TEC(n^−ng)+∑j=13Uj(δ^j),(2)\hat{H}=4(\hat{\mathbf{n}}-\mathbf{n}_g)^T E_C(\hat{\mathbf{n}}-\mathbf{n}_g)+\sum_{j=1}^{3}U_j(\hat{\delta}_j), \tag*{(2)}

where EC=(e2/2)C−1E_C=(e^2/2)C^{-1} is the charging-energy matrix, CC the capacitance matrix, n^\hat{\mathbf{n}} are the node charges and ng\mathbf{n}_g the offset charges. The eigenvalues are ϵk\epsilon_k, and the three lowest states are the computational states. With two harmonics, a junction has two wells only when ∣η∣>1/4\lvert\eta\rvert>1/4 [Fig. 1(a)]. In the positive class, η>1/4\eta>1/4, the wells lie at δ=0\delta=0 and π\pi and differ in energy by 2E12E_1. In the negative class, η<−1/4\eta<-1/4, they lie at ±δ⋆\pm\delta_\star with cos⁡δ⋆=1/(4∣η∣)\cos\delta_\star=1/(4\lvert\eta\rvert). The loop then fixes the operating flux, at which its lowest wells are degenerate [Fig. 1(c,d)]: ϕext=π\phi_{\mathrm{ext}}=\pi, half a flux quantum, for η>0\eta>0, and ϕext=δ⋆\phi_{\mathrm{ext}}=\delta_\star for an η<0\eta<0 qutrit or π−δ⋆\pi-\delta_\star for a four-level qudit [41].

Three exact symmetries.—S1, branch reflection: when the CPR is odd, U(δ)U(\delta) is even, so the η<0\eta<0 wells ±δ⋆\pm\delta_\star are degenerate for any harmonic content. S2, branch permutation: identical branches make the computational states a C3C_3 multiplet at every flux, with equal loop currents. S3, loop reflection at half flux: at φext=π\varphi_{\mathrm{ext}}=\pi the reflection (φ1,φ2)→(−φ1,−φ2)(\varphi_1,\varphi_2)\to(-\varphi_1,-\varphi_2) reverses every δj\delta_j and leaves Eq. (2) unchanged at ng=0n_g=0, while the node charges change sign. For η>0\eta>0 the wells lie at δj=0\delta_j=0 or π\pi, the fixed points of this reflection. Each (π,0,0)(\pi,0,0) configuration then maps to itself, the three computational states have the same parity, and

⟨i∣n^α∣j⟩=0identically,i,j∈{0,1,2},(3)\langle i|\hat n_\alpha|j\rangle=0 \qquad\text{identically},\quad i,j\in\{0,1,2\}, \tag*{(3)}

for any ECE_C, harmonic vector or branch asymmetry. The projected charge norm NQ=(∑α∥Pn^αP∥2)1/2N_Q=\left(\sum_\alpha\|P\hat n_\alpha P\|^2\right)^{1/2}, where PP projects on the computational subspace, stays at the numerical floor for every ∣E2∣/EC|E_2|/E_C computed [41]. Offset charge relaxes the rule linearly in ngn_g. The loop current is odd under S3, so its expectation value vanishes and the computational states have no first-order flux sensitivity.

The odd term.—The η<0\eta<0 class has no such rule. Its three lowest wells at φext=δ⋆\varphi_{\mathrm{ext}}=\delta_\star differ only in which branch sits at −δ⋆-\delta_\star, and S1 keeps them degenerate. A perturbation that keeps the branch potentials even (higher harmonics, branch asymmetry, unequal capacitors) therefore cannot split them at first order. What splits them is a term odd in δ\delta. Adding Ajsin⁡δjA_j\sin\delta_j shifts well kk, in which branch kk sits at −δ⋆-\delta_\star, by

ΔEk=sin⁡δ⋆(∑jAj−2Ak),(4)\Delta E_k=\sin\delta_\star\left(\sum_j A_j-2A_k\right), \tag*{(4)}

which splits the wells only if the AjA_j differ. The choice A=(Ea,−Ea/2,0)A=(E_a,-E_a/2,0) keeps the two transitions distinct. With Ea=0.05E_a=0.05 a sin⁡δ\sin\delta term splits the wells by about 0.10.1 and a sin⁡3δ\sin3\delta term by a third of that after zero-point averaging, while cos⁡δ\cos\delta or cos⁡3δ\cos3\delta terms act only at second order, about a hundred times weaker [41]. Equivalently, the first harmonic is phase-shifted relative to the second by a few degrees, which is the condition for a Josephson diode [35, 42, 43, 44], here about one percent efficiency. Ordinary junctions supply the odd term. A small junction of Josephson energy EaE_a across branch jj, closing a sub-loop that holds Φ0/4\Phi_0/4 or 3Φ0/43\Phi_0/4, adds −Eacos⁡(δj±π/2)=±Easin⁡δj-E_a\cos(\delta_j\pm\pi/2)=\pm E_a\sin\delta_j exactly [Figure 1(b)], as in the asymmetrically threaded SQUID of cat qubits [45]. At that bias the sine amplitude is stationary, and each sub-loop is about as sensitive to flux as the main loop. One coil can bias those three loops when their areas are in the ratio of about 1:3:51:3:5, the main loop holding one extra flux quantum. Percent-level area errors then move the levels by about 10−310^{-3} of the splitting [41].

Figure 1. The two sign classes. (a) Junction energy U(δ)/∣E2∣U(\delta)/|E_2| for three values of η\eta; the dotted curve at η=0.2\eta=0.2 lies below the threshold ∣η∣=1/4|\eta|=1/4. (b) The circuit. Three higher-harmonic elements JjJ_j, each shunted by a capacitor (not drawn), form a loop threaded by Φext\Phi_{\mathrm{ext}}; this is the whole η>0\eta>0 circuit. For η<0\eta<0, side junctions EaE_a and Ea/2E_a/2 (blue) outside the loop, biased by the same field, close sub-loops holding Φ0/4\Phi_0/4 and 3Φ0/43\Phi_0/4 and so add +Easin⁡δ1+E_a\sin\delta_1 and −(Ea/2)sin⁡δ2-(E_a/2)\sin\delta_2, the odd term of Eq. (4). (c) Potential of the loop for η=+0.8\eta=+0.8 at φext=π\varphi_{\mathrm{ext}}=\pi: the three (π,0,0)(\pi,0,0) permutations (white) are degenerate, and (π,π,π)(\pi,\pi,\pi) (gray) lies 4E14E_1 higher. (d) The same for η=−0.8\eta=-0.8 at φext=δ⋆\varphi_{\mathrm{ext}}=\delta_\star, before the odd term: the three (δ⋆,δ⋆,−δ⋆)(\delta_\star,\delta_\star,-\delta_\star) permutations (white) are degenerate, and a fourth well (gray) lies higher.

Figures of merit.—Both classes meet the matrix-element criteria used for single-junction qubits [27, 31], but these criteria say little here. The element ∣⟨0∣n^∣0⟩−⟨1∣n^∣1⟩∣2|\langle0|\hat n|0\rangle-\langle1|\hat n|1\rangle|^2 vanishes at ng=0n_g=0 for every design, a transmon included, and the quasiparticle criterion assumes the thermal quasiparticle density, which is negligible beside the measured fraction xqp∼10−9x_{\mathrm{qp}}\sim10^{-9}–10−610^{-6} [46]. We therefore evaluate every noise channel. Dielectric loss with quality factor QQ relaxes computational state ii at the rate

Γ1=∑j=13∑k<i∣⟨k∣ℏδ˙j∣i⟩∣24ECjℏQ,(5)\Gamma_1=\sum_{j=1}^{3}\sum_{k<i}\frac{|\langle k|\hbar\dot{\delta}_j|i\rangle|^2}{4E_{Cj}\hbar Q}, \tag*{(5)}

where ECjE_{Cj} are the branch charging energies and ℏφ˙α=8∑βEC,αβ(n^β−ngβ)\hbar\dot{\varphi}_\alpha=8\sum_\beta E_{C,\alpha\beta}(\hat n_\beta-n_{g\beta}). The rate is thus set by the charge matrix elements ∣⟨i∣n^α∣j⟩∣=(∑α∣⟨i∣n^α∣j⟩∣2)1/2|\langle i|\hat n_\alpha|j\rangle|=(\sum_\alpha|\langle i|\hat n_\alpha|j\rangle|^2)^{1/2} [41]. Quasistatic noise of amplitude σ\sigma dephases a qubit transition at the rate ℏΓφ=sσ\hbar\Gamma_\varphi=s\sigma, where ss is the sensitivity of its energy [1, 47]. For charge noise we take s=Ws=W, the charge dispersion [1], the peak-to-peak variation of a qutrit transition energy with ngn_g, and σng=10−3\sigma_{n_g}=10^{-3}. For critical-current noise we take s=qs=q, the largest slope of a qutrit splitting per unit fractional change uu of each junction’s harmonic vector, the junctions fluctuating independently, and σu=10−6\sigma_u=10^{-6}, its level in tunnel junctions [48]. A quasiparticle tunneling across junction jj is an event that acts through Mfi=⟨f′∣sin⁡(δ^j/2)∣i⟩M_{fi}=\langle f'|\sin(\hat{\delta}_j/2)|i\rangle, with ∣f′⟩|f'\rangle in the charge-parity sector the event switches to. Its rate is ∣Mfi∣2Sqp(ωif)|M_{fi}|^2S_{\mathrm{qp}}(\omega_{if}), where Sqp∝xqpS_{\mathrm{qp}}\propto x_{\mathrm{qp}} is the quasiparticle spectral density [49]. We take xqp=10−8x_{\mathrm{qp}}=10^{-8}, a superconducting gap Δ=5\Delta=5, kBT=0.04k_B T=0.04 and a junction energy EJ=1E_J=1 in the prefactor of SqpS_{\mathrm{qp}}. Because SqpS_{\mathrm{qp}} diverges for events that exchange no energy, we cut it off at frequencies between 10−610^{-6} and 10−310^{-3} of ∣E2∣/h|E_2|/h and show the spread as bands [41]. Such events dephase states kk and ll at the rate 12∣Mkk−Mll∣2Sqp\frac{1}{2}|M_{kk}-M_{ll}|^2S_{\mathrm{qp}}, summed over junctions, or 12(∣Mkk∣−∣Mll∣)2Sqp\frac{1}{2}(|M_{kk}|-|M_{ll}|)^2S_{\mathrm{qp}} when each event is detected and its sign flip corrected, which we call tracked.

The η>0\eta> 0 qutrit.—With identical branches the computational states form a C3C_3 multiplet, split only by tunneling. An asymmetry ϵ\epsilon that scales the three harmonic vectors by 1+ϵ1+\epsilon, 11 and 1−ϵ1-\epsilon splits them linearly Figure 2(a). It does not break S3S_3, which maps each branch onto itself. We take the altermagnet vector of Eq. (7) at ∣E2∣/EC=10|E_2|/E_C=10 and ϵ=0.12\epsilon=0.12, which puts the computational levels 0.10.1 apart beneath a gap of about 11. Relaxation vanishes at ϕext=π\phi_{\mathrm{ext}}=\pi and ng=0n_g=0 and returns away from either point Figure 2(b). Even at the worst offset, ℏQΓ1\hbar Q\Gamma_1 is a few times 10−410^{-4}. The dephasing rate ℏΓϕ\hbar\Gamma_\phi depends on what produces the splitting Figure 2(c). The splittings are differences of junction energies, so critical-current noise enters with q≈1q\approx1 for every ϵ\epsilon, and quasiparticles dephase at a comparable rate. Charge noise is ten

Figure 2. The η>0\eta> 0 qutrit: the harmonic vector of Eq. (7) on branches scaled by 1+ϵ1+\epsilon, 11 and 1−ϵ1-\epsilon, at ∣E2∣/EC=10|E_2|/E_C=10. (a) Level spacings versus ϵ\epsilon at ϕext=π\phi_{\mathrm{ext}}=\pi; E3−E2E_3-E_2 is the gap above the computational states. (b) Relaxation rate of the faster-decaying state, Eq. (5), at ng=0n_g=0 and maximized over ngn_g, and the charge dispersion WW, versus flux at ϵ=0.12\epsilon=0.12; at ϕext=π\phi_{\mathrm{ext}}=\pi the ng=0n_g=0 rate vanishes to machine precision and is drawn at the floor. (c) Dephasing rates at ϕext=π\phi_{\mathrm{ext}}=\pi from charge noise (σng=10−3\sigma_{n_g}=10^{-3}), critical-current noise (σu=10−6\sigma_u=10^{-6}) and quasiparticles (xqp=10−8x_{\mathrm{qp}}=10^{-8}; shaded bands show the spread over the cutoff of SqpS_{\mathrm{qp}}); tracked and untracked rates coincide. (d) ∣⟨i∣n^∣j⟩∣|\langle i|\hat n|j\rangle| for the nine lowest states at ϵ=0.12\epsilon=0.12, ϕext=π\phi_{\mathrm{ext}}=\pi, ng=0n_g=0: the computational block (outlined) is empty. Dotted lines: operating point.

times weaker at the operating point and fades as ϵ\epsilon localizes the wells. The computational block of the charge operator is empty Figure 2(d), so the qutrit is driven through a plasmon level above the computational states, shared by all three. This intermediate level survives the asymmetry only in shallow wells, which is why we choose ∣E2∣/EC=10|E_2|/E_C=10. A shaped two-tone pulse through it transfers ∣0⟩→∣2⟩|0\rangle\to|2\rangle fast and with high fidelity [41, 50, 51].

The η<0\eta< 0 qutrit.—With A=(Ea,−Ea/2,0)A=(E_a,-E_a/2,0) the level spacings approach 2Easin⁡δ∗2E_a\sin\delta_* and Easin⁡δ∗E_a\sin\delta_* once EaE_a exceeds the tunneling splitting Figure 3(a). We take ∣η∣=0.8|\eta|=0.8, ∣E2∣/EC=20|E_2|/E_C=20 and Ea=0.05E_a=0.05, which puts the computational levels 0.090.09 and 0.050.05 apart beneath a gap of 0.750.75. No symmetry forbids relaxation here. The charge elements between computational states are suppressed only by tunneling between deep wells, so Γ1\Gamma_1 is finite and insensitive to flux and offset charge Figure 3(b), with ℏQΓ1≈2×10−3\hbar Q\Gamma_1\approx2\times10^{-3}. These elements make the qutrit directly drivable Figure 3(d). Flat-top π\pi pulses of a charge drive reach high fidelity within a few tens of periods of the qutrit transitions [41]. The charge dispersion is negligible: W≈10−8W\approx10^{-8}, four orders of magnitude below the η>0\eta>0 device. The reason is topological. Offset charge enters only through tunneling paths that wind a node phase by 2π2\pi. Neighboring η<0\eta<0 wells are joined through the barrier at δj=0\delta_j=0 without winding, while winding must cross δj=π\delta_j=\pi, where the barrier is more than three times higher. Critical-current noise of the main junctions shifts all three wells together, by S1. It therefore acts through the side junctions and grows with EaE_a [Figure 3(c)].

Figure 3. The η<0\eta< 0 qutrit: two harmonics with ∣η∣=0.8|\eta|=0.8 at ∣E2∣/EC=20|E_2|/E_C=20 and ϕext=δ∗\phi_{\mathrm{ext}}=\delta_*, with side junctions supplying the odd cut term A=(Ea,−Ea/2,0)A=(E_a,-E_a/2,0). (a) Level spacings versus EaE_a. (b) As Figure 2(b), versus the main-loop flux at Ea=0.05E_a=0.05, sub-loop biases fixed; the two rates coincide. (c) As Figure 2(c), versus EaE_a; parity tracking lowers the quasiparticle rate. (d) As Figure 2(d), at Ea=0.05E_a=0.05: the computational block carries tunneling-suppressed elements.

Quasiparticles.— Each quasiparticle tunneling event acts through the computational block of MM. If the moduli of its diagonal elements are unequal, the event measures the state. If the block is proportional to a unitary, the event causes a known error, which can be corrected once the event is detected. The sign class decides which case applies. For η>0\eta> 0 the wells sit at the fixed points 0 and π\pi, where ∣sin⁡(δ/2)∣|\sin(\delta/2)| is 0 or 1. Each event therefore measures which branch sits at π\pi, an error that cannot be corrected [Figure 2(c)]. The property that makes the selection rule exact, wells at the fixed points, thus makes every event a measurement. For η<0\eta< 0 the mirror wells share ∣sin⁡(δ⋆/2)∣|\sin(\delta_\star/2)|, and an event across junction jj only flips the sign of the state in which branch jj sits at −δ⋆-\delta_\star. The charge parities of the two nodes identify jj. Tracking these parities [52] with a separate detector reveals each event, so the flip can be corrected and the dephasing rate falls by more than an order of magnitude [Figure 3(c)]. Side junctions reopen a weak measurement channel, because across a junction biased at ±π/2\pm\pi/2 the mirror wells give unequal ∣Mii∣|M_{ii}|. It grows with EaE_a, while at small EaE_a the wells are not yet localized, so the tracked rate is lowest near Ea≈0.01E_a \approx0.01–0.020.02. The same model classifies single-junction designs: each event measures the 2ϕ2\phi qubit of Ref. [31], which stores its states in the 0 and π\pi wells, but only flips a sign in the mirror-well flowermon [27].

An altermagnetic platform.— Altermagnets are compensated collinear magnets with spin-split bands and no net magnetization [53, 54, 55, 56]. They are being explored for spin torques [57, 58] and spin qubits [59, 60], and Josephson junctions of altermagnetic CrSb have been reported [61, 62]. In a superconductor/altermagnet/superconductor junction, tuning through a 0–π\pi transition suppresses the first harmonic and leaves a pronounced second [22, 23, 24, 25]. We use the continuum model of Ref. [24], an altermagnet between two ss-wave superconducting leads [Figure 4(a)]. The Hamiltonian of the altermagnet is

Figure 4. The altermagnetic platform. (a) A gate voltage VgV_g sets the chemical potential μ+δμ\mu+\delta\mu of the altermagnet (AM) between superconducting leads (S); its two spin species have Fermi surfaces elongated along orthogonal axes. (b) Junction energy Uˉ\bar{U}, normalized to its range, at three gate voltages; circles mark the wells. (c) First and second harmonics per transverse channel versus δμ/μ\delta\mu/\mu, gauged to E1≥0E_1 \ge0; shading marks the η>0\eta> 0 (orange) and η<0\eta< 0 (blue) classes. (d) Critical-current noise σu\sigma_u produced by chemical-potential noise σμ\sigma_\mu at μ=100\mu= 100 meV; the axis starts at σu=10−6\sigma_u = 10^{-6}, the value used in Figure 2 and Figure 3. Exchange J/μ=0.2J/\mu= 0.2, barrier Z=0.5Z = 0.5.

H=μ(k2kF2−1)−δμ±2JkF2kxky(6)H=\mu\left(\frac{k^2}{k_F^2}-1\right)-\delta\mu\pm\frac{2J}{k_F^2}k_xk_y \tag*{(6)}

Here k\mathbf{k} is the wave vector, and the two signs are for spin up and down. The leads have chemical potential μ\mu and Fermi wave vector kFk_F. The gate voltage shifts the chemical potential of the altermagnet by δμ\delta\mu. The last term is the altermagnetic exchange splitting, which has strength JJ and dd-wave symmetry. The current flows along xx, so the splitting is largest at 45° to the current. The altermagnet has no disorder and is 20/kF20/k_F long. The leads have a superconducting gap of 0.01μ0.01\mu. Each interface has a delta-function barrier of strength 2Zμ/kF2Z\mu/k_F. We compute the CPR with Kwant [63] from a Matsubara Green-function sum with semi-infinite leads. The temperature is 0.025 of the critical temperature of the leads. We validate the calculation against published results [41]. The junction energy U(δ)U(\delta) takes all three shapes of Figure 1(a) [Figure 4(b)], and the gate voltage selects the class [Figure 4(c)]. About a third of the gate voltages give the η>0\eta> 0 class, a quarter the η<0\eta< 0 class and the rest a single well, in interleaved windows a few meV wide. At δμ/μ=−0.6\delta\mu/\mu=-0.6 the harmonic vector is

Em/∣E2∣≈(0.41, 1, 0.06, −0.14, −0.02, 0.05)(7)E_m/|E_2|\approx(0.41,\,1,\,0.06,\,-0.14,\,-0.02,\,0.05) \tag*{(7)}

with η≈2.5\eta\approx2.5. With this vector on all branches at ∣E2∣/EC=200|E_2|/E_C=200, the computational states are degenerate with vanishing NQ\mathcal{N}_Q, and the next state lies about ∣E2∣/3|E_2|/3 above. Figure 2 uses the same vector at ∣E2∣/EC=10|E_2|/E_C=10. With Rashba coupling α\alpha the supercurrent depends on the Néel-vector direction [58]. The junction then acquires the odd term only where symmetry allows: for a Néel-vector component along the current, or for unequal barriers with an out-of-plane Néel vector. At the η<0\eta< 0 point δμ/μ=0.3\delta\mu/\mu=0.3, ∣A∣≃20(αRkF/μ)∣E2∣|A| \simeq20(\alpha_R k_F/\mu)|E_2|. The qubit’s odd term (Ea=0.05E_a=0.05) therefore needs only αRkF/μ≈0.0025\alpha_R k_F/\mu\approx0.0025, or αR≈0.15\alpha_R\approx0.15 meV nm at μ=100\mu=100 meV and the free-electron mass. However, identical junctions with one Néel vector acquire the same AA, which cannot split the qutrit. The intrinsic route needs a Néel-vector orientation or Rashba coupling that differs between branches. We have not modeled such a loop, so we use side junctions. For η>0\eta>0 the odd term breaks S3S_3, so the Néel vector must stay out of plane and the barriers symmetric [41].

**Gate-voltage noise.—**The gate voltage that selects the class also moves the harmonics, so its noise σμ\sigma_\mu acts as critical-current noise, σu=λσμ/μ\sigma_u = \lambda\sigma_\mu/\mu, where λ=dln⁡E/d(δμ/μ)\lambda= d\ln E/d(\delta\mu/\mu) and EE is the energy that carries the qutrit splitting [Figure 4(d)]. For η>0\eta> 0 that energy is Eodd=∑m oddEmE_{\mathrm{odd}} = \sum_{m\ \mathrm{odd}} E_m, half the splitting of the 0 and π\pi wells, and λ≈400\lambda\approx400. Then σu=10−6\sigma_u = 10^{-6} needs sub-neV stability of μ=100\mu= 100 meV, and a realistic 0.1 μ\mueV gives a dephasing time of about 50 ns per branch for ∣E2∣/h=10\lvert E_2\rvert/h = 10 GHz. The large slope is structural. The class windows are narrow because the first harmonic sweeps fast, their width scaling as 8∣E2∣/∣dE1/d(δμ/μ)∣8\lvert E_2\rvert/\lvert dE_1/d(\delta\mu/\mu)\rvert. The derivative that makes a window a few meV wide therefore makes the device limited by gate-voltage noise. Gate-charge noise already limits gatemons [12]. The η<0\eta< 0 class is less exposed, since its splitting comes from the side junctions and S1 protects the main ones. At the η<0\eta< 0 point λ≈60\lambda\approx60 for ∣E1∣\lvert E_1\rvert and 15 for ∣E2∣\lvert E_2\rvert, and 0.1 μ\mueV gives microseconds per branch, or longer when the branches share one gate [41].

**Conclusion.—**We have shown that a three-junction loop hosts protected qutrits in both sign classes of the second harmonic, whichever platform supplies the junctions. We estimate the coherence for ∣E2∣/h=10\lvert E_2\rvert/h = 10 GHz, Q=106Q = 10^6 and 20 mK, with the noise amplitudes above and 1 μΦ0\mu\Phi_0 of flux noise in every loop. Relaxation is where both classes gain over a transmon, by two to three orders of magnitude. A transmon qutrit [1, 64] modeled with the same QQ and σu\sigma_u has T1≈20T_1 \approx20–30 μ\mus and, from critical-current noise, Tϕ≈30T_\phi\approx30 μ\mus. The η>0\eta> 0 device has T1≳50T_1 \gtrsim50 ms at the worst offset charge (unbounded at ng=0n_g = 0), and the η<0\eta< 0 qutrit has T1≈10T_1 \approx10 ms. Dephasing therefore sets the coherence time. The η>0\eta> 0 device has a Ramsey time T2∗≈5T_2^* \approx5–10 μ\mus, set by critical-current noise and quasiparticles. For the η<0\eta< 0 qutrit, quasiparticles limit T2∗T_2^* to 5–15 μ\mus without tracking. Parity tracking raises it to about 0.1 ms, limited by the side junctions’ quasiparticle channel and by critical-current noise, with charge noise (seconds) and flux noise (a millisecond) far behind. A Hahn echo refocuses critical-current, charge and flux noise, but not quasiparticle events. The computational levels are 0.5–1 GHz apart, so 10–35% of the population is thermally excited at 20–40 mK. Both classes must therefore be initialized by active reset rather than by waiting.

Which class to use depends on what limits the device. The η>0\eta> 0 class has the exact charge selection rule and the faster gates, under 10 ns against a few tens of nanoseconds. The η<0\eta< 0 class is insensitive to charge noise, can be driven directly, and its quasiparticle errors can be corrected. For alternamgetic junctions, gate-voltage noise would dominate the estimates above. Interference-based elements realize the η>0\eta> 0 class [8, 10, 11, 12, 13, 14, 26], and the selection rule is exact when their two arms are balanced. An arm imbalance adds the odd term, which breaks S3, and the flux through the element’s own loop tilts the 0 and π\pi wells at first order [41]. Twisted dd-wave junctions supply the η<0\eta< 0 class natively [15, 18, 19, 20, 21], with a twist angle fixed at fabrication and a bath of nodal quasiparticles. With side junctions they too host the η<0\eta< 0 qutrit. The remaining priorities are a fast parity detector, the stability of small junctions and gate voltages, and quasiparticle physics beyond the tunnel-junction model for transparent weak links.

**Acknowledgments.—**This work was supported by the National Science Foundation/EPSCoR RII Track-1: Emergent Quantum Materials and Technologies (EQUATE), Award OIA-2044049 and by the UNL Grand Challenges catalyst award entitled Quantum Approaches addressing Global Threats. The calculations and writing of the paper were reviewed by Claude Opus 5.5.

References

References

  1. [1]J. Koch, T. M. Yu, J. Gambetta, A. A. Houck, D. I. Schuster, J. Majer, A. Blais, M. H. Devoret, S. M. Girvin, and R. J. Schoelkopf, Charge-insensitive qubit design derived from the Cooper pair box, Physical Review A 76, 042319 (2007).DOI
  2. [2]M. H. Devoret and R. J. Schoelkopf, Superconducting circuits for quantum information: An outlook, Science 339, 1169 (2013).DOI
  3. [3]A. Blais, A. L. Grimsmo, S. M. Girvin, and A. Wallraff, Circuit quantum electrodynamics, Rev. Mod. Phys. 93, 025005 (2021).DOI
  4. [4]A. Kitaev, Protected qubit based on a superconducting current mirror, arXiv preprint cond-mat/0609441 (2006).DOI
  5. [5]P. Brooks, A. Kitaev, and J. Preskill, Protected gates for superconducting qubits, Physical Review A 87, 052306 (2013).DOI
  6. [6]V. E. Manucharyan, J. Koch, L. I. Glazman, and M. H. Devoret, Fluxonium: single Cooper pair circuit free of charge offsets, Science 326, 113 (2009).DOI
  7. [7]A. Gyenis, P. S. Mundada, A. Di Paolo, T. M. Hazard, X. You, D. I. Schuster, J. Koch, A. Blais, and A. A. Houck, Experimental realization of a protected superconducting circuit derived from the 0–π qubit, PRX Quantum 2, 010339 (2021).
  8. [8]W. C. Smith, A. Kou, U. Xiao, U. Vool, and M. H. Devoret, Superconducting circuit protected by two-Cooper-pair tunneling, npj Quantum Information 6, 8 (2020).DOI
  9. [9]B. Douçot and L. B. Ioffe, Physical implementation of protected qubits, Rep. Prog. Phys. 75, 072001 (2012).DOI
  10. [10]S. Gladchenko, D. Olaya, E. Dupont-Ferrier, B. Douçot, L. B. Ioffe, and M. E. Gershenson, Superconducting nanocircuits for topologically protected qubits, Nature Physics 5, 48 (2009).DOI
  11. [11]M. T. Bell, J. Paramanandam, L. B. Ioffe, and M. E. Gershenson, Protected Josephson rhombus chains, Phys. Rev. Lett. 112, 167001 (2014).pubmed.ncbi.nlm.nih.gov/24815663
  12. [12]T. W. Larsen, M. E. Gershenson, L. Casparis, A. Kringhøj, N. J. Pearson, R. P. G. McNeil, F. K.ue.meth, P. Krogstrup, K. D. Petersson, and C. M. Marcus, Parity-protected superconductor-semiconductor qubit, Phys. Rev. Lett. 125, 056801 (2020).DOI
  13. [13]E. Royerch, A. Borgognoni, M. Villiers, K. Gerashchenko, W. C. Smith, C. Wilson, B. Douçot, A. Petrescu, P. Campagne-Ibarcq, and Z. Leghtas, Experimental realization of a cos(2φ) transmon qubit, arXiv preprint (2026), arXiv:2603.13114 [quant-ph].
  14. [14]N. K. Zhurbina, S. Singh, L. J. Splitthoff, E. Y. Huang, F. Yılmaz, A. M. Bozkurt, and C. K. Andersen, Coherence limitations of a Fourier-engineered cos(2φ) transmon qubit, arXiv preprint (2026), arXiv:2605.06372 [quant-ph].arxiv.org/abs/2605.06372
  15. [15]M. Sigrist and T. M. Rice, Unusual paramagnetic phenomena in granular high-temperature superconductors—A consequence of d-wave pairing?, Rev. Mod. Phys. 67, 503 (1995).DOI
  16. [16]Y. Tanaka and S. Kashiwaya, Theory of Josephson effects in anisotropic superconductors, Phys. Rev. B 56, 892 (1997).DOI
  17. [17]T. Löfwander, V. S. Shumeiko, and G. Wendin, Andreev bound states in high-Tc superconducting junctions, Supercond. Sci. Technol. 14, R53 (2001).DOI
  18. [18]O. Can et al., High-temperature topological superconductivity in twisted double-layer copper oxides, Nature Phys. 17, 519 (2021).DOI
  19. [19]S. Y. F. Zhao et al., Time-reversal symmetry breaking superconductivity between twisted cuprate superconductors, Science 382, 1422 (2023).DOI
  20. [20]T. Tummuru, S. Plugge, and M. Franz, Josephson effects in twisted cuprate bilayers, Phys. Rev. B 105, 064501 (2022).DOI
  21. [21]P. A. Volkov, É. Lantagne-Hurtubise, T. Tummuru, S. Plugge, J. H. Pixley, and M. Franz, Josephson diode effects in twisted nodal superconductors, Phys. Rev. B 109, 094518 (2024).DOI
  22. [22]J. A. Ouassou, A. Brataas, and J. Linder, dc Josephson effect in altermagnets, Phys. Rev. Lett. 131, 076003 (2023).DOI
  23. [23]C. W. J. Beenakker and T. Vakhtel, Phase-shifted Andreev levels in an altermagnet Josephson junction, Phys. Rev. B 108, 075425 (2023).DOI
  24. [24]B. Lu, K. Maeda, H. Ito, K. Yada, and Y. Tanaka, φ Josephson junction induced by altermagnetism, Phys. Rev. Lett. 133, 226002 (2024).DOI
  25. [25]H.-P. Sun, S.-S.-B. Zhang, C.-A. Li, and B. Trauzettel, Tunable second harmonic in altermagnetic Josephson junctions, Phys. Rev. B 111, 165406 (2025).
  26. [26]S. Messelot, A. Leblanc, J.-S. Tétel, F. Lefloch, Q. Ficheux, J. Renard, and É. Dumur, Coherence limits in interference-based cos(2φ) qubits, Phys. Rev. Appl. 26, 044021 (2026).DOI
  27. [27]V. Brosco et al., Superconducting qubit based on twisted cuprate van der Waals heterostructures, Phys. Rev. Lett. 132, 017003 (2024).DOI
  28. [28]I. Vsoughi-nia and M. P. Nowak, Magnetic-field-free parity-protected qubit based on a narrow altermagnet Josephson junction, Phys. Rev. B 114, 225408 (2026).
  29. [29]J. B. Tjernshaugen, M. Amundsen, and M. I. Jender, Superconducting qubits with altermagnetic Josephson junctions, arXiv preprint (2026), arXiv:2606.02761 [quant-ph].arxiv.org/abs/2606.02761
  30. [30]G.-L. Guo, Full gate-voltage control of a parity-protected superconducting qubit with an altermagnetic Josephson junction, arXiv preprint (2026), arXiv:2607.04097 [cond-mat.supr-con].arxiv.org/abs/2607.04097
  31. [31]X.-F. Pan, X.-L. Hei, F. Nori, and P.-B. Li, Superconducting qubit based on altermagnets, arXiv preprint (2026), arXiv:2609.24759 [quant-ph].arxiv.org/abs/2609.24759
  32. [32]M. Neeley, M. Ansmann, R. C. Bialczak, M. Hofheinz, E. Lucero, A. D. O’Connell, D. Sank, H. Wang, J. Wenner, A. N. Cleland, J. M. Geller, and J. M. Martinis, Emulation of a quantum spin with a superconducting phase qudit, Science 325, 722 (2009).DOI
  33. [33]L. Chirolli and J. E. Moore, Enhanced coherence in superconducting circuits via band engineering, Physical Review Letters 126, 187701 (2021).DOI
  34. [34]L. Chirolli, V. Brosco, U. Vool, G. Catelani, and L. Amico, Fraxonium: Fractional fluxon states for qudit encoding, arXiv preprint (2026), arXiv:2605.14586 [quant-ph].arxiv.org/abs/2605.14586
  35. [35]H. Vakili, M. Ali, and A. A. Kovalev, Field-free Josephson diode effect in a d-wave superconductor heterostructure, Phys. Rev. B 110, 104518 (2024).arxiv.org/abs/2406.11127
  36. [36]Q. Cheng, Y. Mao, and Q.-F. Sun, Field-free Josephson diode effect in altermagnet/normal metal/altermagnet junctions, Phys. Rev. B 110, 014518 (2024).DOI
  37. [37]B. K. Sahoo and A. Saori, Giant field-free transverse Josephson diode effect in altermagnets, Phys. Rev. B 113, 245433 (2026).DOI
  38. [38]T. P. Orlando, J. E. Mooij, L. Tian, C. H. van der Wal, L. S. Levitov, S. Lloyd, and J. J. Mazo, Superconducting persistent-current qubit, Physical Review B 60, 15398 (1999).DOI
  39. [39]M. H. Devoret, Quantum fluctuations in electrical circuits, in Quantum Fluctuations, Les Houches, Session LXIII (1995), edited by S. Reynaud, E. Giacobino, and J. Zinn-Justin (Elsevier, 1997).
  40. [40]U. Vool and M. H. Devoret, Introduction to quantum electromagnetic circuits, Int. J. Circ. Theor. Appl. 45, 897 (2017).DOI
  41. [41]See Supplemental Material for the circuit quantization, branch well structure and the d = 4 manifold, the full statements of the three symmetries, the charge-suppression scaling, the Bogoliubov–de Gennes method and its validation, the barrier and gate scans, the coherence budget, and the pulse design rules and speed budget.
  42. [42]R. S. Souto, M. Leijnse, and C. Schrade, Josephson diode effect in superconducting interferometers, Phys. Rev. Lett. 129, 267702 (2022).DOI
  43. [43]C. Baumgartner, L. Fuchs, A. Costa, S. Reinhardt, S. Gronin, G. C. Gardner, T. Lindemann, M. J. Manfra, P. E. Faria Junior, D. Kochan, J. Fabian, N. Paradiso, and C. Strunk, Supercurrent rectification and magnetochiral effects in symmetric Josephson junctions, Nat. Nanotechnol. 17, 39 (2022).DOI
  44. [44]M. Coraiola, A. E. Stegovetov, D. Z. Haxell, D. Sabonis, M. Hinderling, S. C. ten Kate, E. Cheah, F. Krizek, R. Schott, W. Wegscheider, J. C. Cuevas, W. Belzig, and F. Nichele, Flux-tunable Josephson diode effect in a hybrid four-terminal Josephson junction, ACS Nano 18, 9221 (2024).
  45. [45]R. Lescanne, M. Villiers, T. Peronnin, A. Sarlette, M. Delbecq, B. Huard, T. Kontos, M. Mirrahimi, and Z. Leghtas, Exponential suppression of bit-flips in a qubit encoded in an oscillator, Nat. Phys. 16, 509 (2020).DOI
  46. [46]K. Serniak, M. Hays, G. de Lange, S. Dimand, S. Shankar, L. D. Burkhart, L. Frunzio, M. Houzet, and M. H. Devoret, Not nonequilibrium quasiparticles in transmon qubits, Phys. Rev. Lett. 121, 157701 (2018).
  47. [47]G. Ithier, E. Collin, P. Joyez, P. J. Meeson, D. Vion, D. Esteve, F. Chiarello, A. Shnirman, Y. Makhlin, J. Schriefl, and G. Schön, Decoherence in a superconducting quantum bit circuit, Phys. Rev. B 72, 134519 (2005).DOI
  48. [48]D. J. Van Harlingen, T. L. Robertson, B. L. T. Plourde, P. A. Reichardt, T. A. Crane, and J. Clarke, Decoherence in Josephson-junction qubits due to critical-current fluctuations, Phys. Rev. B 70, 064517 (2004).DOI
  49. [49]G. Catelani, R. J. Schoelkopf, M. H. Devoret, and L. I. Glazman, Relaxation and frequency shifts induced by quasiparticles in superconducting qubits, Phys. Rev. B 84, 064517 (2011).DOI
  50. [50]T. Caneva, T. Calarco, and S. Montangero, Chopped random-basis quantum optimization, Physical Review A 84, 022326 (2011).DOI
  51. [51]F. Motzoi, J. M. Gambetta, P. Rebentrost, and F. K. Wilhelm, Simple pulses for elimination of leakage in weakly nonlinear qubits, Physical Review Letters 103, 110501 (2009).DOI
  52. [52]K. Serniak, S. Diamond, M. Hays, V. Fatemi, S. Shankar, L. Frunzio, R. J. Schoelkopf, and M. H. Devoret, Direct dispersive monitoring of charge parity in offset-charge-sensitive transmons, Phys. Rev. Appl. 12, 014052 (2019).DOI
  53. [53]L. Šmejkal, J. Sinova, and T. Jungwirth, Beyond conventional ferromagnetism and antiferromagnetism: A phase with nonrelativistic spin and crystal rotation symmetry, Phys. Rev. X 12, 031042 (2022).DOI
  54. [54]L. Šmejkal, J. Sinova, and T. Jungwirth, Emerging research landscape of altermagnetism, Phys. Rev. X 12, 040501 (2022).arxiv.org/abs/2204.10844
  55. [55]P. Wadley, J. Sinova, D. Kriegner, J. Minár, J. H. Dil, and T. Jungwirth, Altermagnetic lifting of Kramers spin degeneracy, Nature 626, 517 (2024).DOI
  56. [56]S. Reimers, L. Odenbreit, L. Šmejkal, et al., Direct observation of altermagnetic band splitting in CrSb thin films, Nat. Commun. 15, 2116 (2024).
  57. [57]H. Vakili, E. Schwartz, and A. A. Kovalev, Spin-transfer torque in altermagnets with magnetic textures, Phys. Rev. Lett. 134, 176401 (2025).DOI
  58. [58]H. Vakili, M. Ali, I. Žutić, and A. A. Kovalev, Supercurrent-driven Néel torque in superconduct-or/altermagnet hybrids, Phys. Rev. B 114, L200502 (2026).DOI
  59. [59]H. Vakili, Gate-controlled spin qubits in confined altermagnets, arXiv preprint (2026), arXiv:2606.24150 [cond-mat.mes-hall].arxiv.org/abs/2606.24150
  60. [60]J. C. Abadillo-Uriel, A. Majani, A. Cortijo, R. Aguado, and R. Seoane Souto, All-electrical dephasing-protected spin qubits in altermagnets, arXiv preprint (2026), arXiv:2606.26066 [cond-mat.mes-hall].
  61. [61]V. D. Esin, D. Y. Kazmin, Y. S. Barash, A. V. Timonina, N. N. Kolesnikov, and E. V. Deviatov, Josephson diode and spin-valve effects on the surface of altermagnet CrSb, JETP Lett. 123, 556 (2026).DOI
  62. [62]Y. Gao, J. Yan, J. Wei, et al., Strain-engineered long-distance supercurrent in an altermagnetic CrSb based Josephson junction, arXiv preprint (2026), arXiv:2610.11341 [cond-mat.supr-con].arxiv.org/abs/2610.11341
  63. [63]C. W. Groth, M. Wimmer, A. R. Akhmerov, and X. Waintal, Kwant: a software package for quantum transport, New J. Phys. 16, 063065 (2014).
  64. [64]M. J. Peterer, S. J. Bader, X. Jin, F. Yan, A. Kamal, T. J. Gudmundsen, P. J. Leek, T. P. Orlando, W. D. Oliver, and S. Gustavsson, Coherence and decay of higher energy levels of a superconducting transmon qubit, Phys. Rev. Lett. 114, 010501 (2015).pubmed.ncbi.nlm.nih.gov/25615454

Paper details

vlnc:W_7dfnbh6b
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 11

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