Qutrit T-State in Quantum Computation

• Qutrit T-state is a non-stabilizer magic resource state in three-level quantum systems, essential for implementing non-Clifford operations.
• It arises from applying the qutrit T gate to stabilizer states, with distinct conventions that reveal a rich underlying algebraic structure.
• The state supports protocols like gate injection and magic state distillation, advancing universal qutrit gate sets and experimental quantum networks.

Qutrit T-state denotes a non-stabilizer resource state for three-level quantum computation, but the term is not uniform across the literature. In current work it appears in at least three closely related senses: as the magic state obtained from a specific qutrit $T$ gate by acting on a stabilizer state, as a distinguished single-qutrit Clifford eigenstate such as the qutrit strange state, and, more speculatively, as a T-type or magic resource sought inside encoded bosonic qutrits. The common role is non-Clifford resource supply: these states are used to inject gates beyond the qutrit Clifford group, to support magic-state distillation, and to organize the structure of universal qutrit gate sets (Glaudell et al., 2022, Jain et al., 2020, Denys et al., 2022).

1. Terminology and algebraic setting

For a single qutrit with computational basis $\{\ket{0},\ket{1},\ket{2}\}$, the generalized Pauli operators are

$$X\ket{k}=\ket{k+1 \bmod 3},\qquad Z\ket{k}=\omega^k\ket{k},\qquad \omega=e^{2\pi i/3}.$$

The single-qutrit Clifford group is the normalizer of the qutrit Pauli group, and the qutrit $T$ gate is a non-Clifford diagonal gate lying in the third level of the qutrit Clifford hierarchy (Glaudell et al., 2022, Amaro-Alcalá et al., 2023).

Two concrete $T$-gate conventions appear in the recent literature. One defines

$$T=\operatorname{diag}(1,\zeta,\zeta^8),\qquad \zeta=e^{2\pi i/9},$$

while another uses

$$T=\operatorname{diag}(1,\omega_9^8,\omega_9),\qquad \omega_9=e^{2\pi i/9}.$$

Both are presented as qutrit $T$ matrices, both are diagonal, and both introduce ninth-root phases unavailable inside the qutrit Clifford group (Glaudell et al., 2022, Amaro-Alcalá et al., 2023).

In the narrow gate-injection sense, a qutrit T-state is the magic state associated with such a gate, typically of the form $T\ket{+}$. In a broader resource-theoretic sense, the term is also used for particularly important non-stabilizer qutrit states that are eigenstates of Clifford operators, especially the strange state $\ket{S}$, which is distinguished by symmetry and Wigner negativity rather than by direct definition from a specific $\{\ket{0},\ket{1},\ket{2}\}$0 matrix (Jain et al., 2020).

2. The gate-associated qutrit T-state

The most direct analogue of the qubit $\{\ket{0},\ket{1},\ket{2}\}$1 state is obtained by applying the qutrit $\{\ket{0},\ket{1},\ket{2}\}$2 gate to a qutrit $\{\ket{0},\ket{1},\ket{2}\}$3-basis stabilizer state. In the formulation using

$\{\ket{0},\ket{1},\ket{2}\}$4

a natural qutrit T magic state is

$\{\ket{0},\ket{1},\ket{2}\}$5

where $\{\ket{0},\ket{1},\ket{2}\}$6 is the $\{\ket{0},\ket{1},\ket{2}\}$7-basis (Glaudell et al., 2022). In the benchmarking-oriented convention,

$\{\ket{0},\ket{1},\ket{2}\}$8

and the associated state is

$\{\ket{0},\ket{1},\ket{2}\}$9

for

$$X\ket{k}=\ket{k+1 \bmod 3},\qquad Z\ket{k}=\omega^k\ket{k},\qquad \omega=e^{2\pi i/3}.$$0

(Amaro-Alcalá et al., 2023).

This gate-associated T-state inherits the defining properties of the underlying $$X\ket{k}=\ket{k+1 \bmod 3},\qquad Z\ket{k}=\omega^k\ket{k},\qquad \omega=e^{2\pi i/3}.$$1 gate. The qutrit $$X\ket{k}=\ket{k+1 \bmod 3},\qquad Z\ket{k}=\omega^k\ket{k},\qquad \omega=e^{2\pi i/3}.$$2 gate lies in the third level $$X\ket{k}=\ket{k+1 \bmod 3},\qquad Z\ket{k}=\omega^k\ket{k},\qquad \omega=e^{2\pi i/3}.$$3 of the qutrit Clifford hierarchy, is non-Clifford, and is injectable by a magic state; its magic states also support magic state distillation (Glaudell et al., 2022). The gate has finite order,

$$X\ket{k}=\ket{k+1 \bmod 3},\qquad Z\ket{k}=\omega^k\ket{k},\qquad \omega=e^{2\pi i/3}.$$4

yet Clifford+$$X\ket{k}=\ket{k+1 \bmod 3},\qquad Z\ket{k}=\omega^k\ket{k},\qquad \omega=e^{2\pi i/3}.$$5 remains approximately universal on qutrits (Glaudell et al., 2022).

A useful exact-synthesis distinction accompanies this state definition. Clifford matrices live in the ring $$X\ket{k}=\ket{k+1 \bmod 3},\qquad Z\ket{k}=\omega^k\ket{k},\qquad \omega=e^{2\pi i/3}.$$6, whereas Clifford+$$X\ket{k}=\ket{k+1 \bmod 3},\qquad Z\ket{k}=\omega^k\ket{k},\qquad \omega=e^{2\pi i/3}.$$7 matrices live in the larger ring $$X\ket{k}=\ket{k+1 \bmod 3},\qquad Z\ket{k}=\omega^k\ket{k},\qquad \omega=e^{2\pi i/3}.$$8, and $$X\ket{k}=\ket{k+1 \bmod 3},\qquad Z\ket{k}=\omega^k\ket{k},\qquad \omega=e^{2\pi i/3}.$$9. Consequently, the qutrit T-state is not merely a stabilizer state with a Clifford phase attached; it is tied to genuinely new algebraic structure (Glaudell et al., 2022).

3. Clifford-eigenstate notions of a qutrit T-state

A broader strand of the literature classifies qutrit magic states as non-stabilizer eigenstates of single-qutrit Clifford operators. In this classification there are four inequivalent non-degenerate qutrit Clifford eigenstates and two degenerate families. The four non-degenerate representatives are the strange state $T$0, the symmetric Hadamard eigenstate $T$1, the Norell state $T$2, and an equatorial state $T$3 (Jain et al., 2020).

The strange state is

$T$4

and it is singled out by several extremal properties. Its discrete Wigner function has one negative entry,

$T$5

and eight positive entries,

$T$6

Its sum negativity is $T$7, and its mana is

$T$8

The strange state is also a simultaneous eigenvector of all symplectic rotations $T$9, so its orbit under symplectic rotations has size $T$0, while its orbit under the full single-qutrit Clifford group has size $T$1 (Jain et al., 2020).

The Norell state,

$T$2

has the same global maximum mana $T$3, but later work cited there gives it smaller thauma than $T$4. The state

$T$5

is a less magical but still distinguished Hadamard eigenstate. The equatorial state has representative

$T$6

and plays a direct state-injection role (Jain et al., 2020).

This classification yields two distinct “T-like” interpretations. The strange state is the most natural qutrit T-state in the resource-theoretic sense: it is “the most magic qutrit state and the most symmetric qutrit state.” By contrast, the equatorial state is the cleanest gate-injection resource: a single pure copy can implement the qutrit analogue of the qubit $T$7 gate with no error, assuming ideal Clifford operations (Jain et al., 2020).

A further structural fact is dimension-specificity. No analogue of the qutrit strange state—no simultaneous eigenvector of all symplectic rotations—exists for odd prime dimensions $T$8. In that precise sense, the qutrit case is exceptional (Jain et al., 2020).

4. Injection, distillation, and expressive power

The operational meaning of a qutrit T-state is clearest in gate injection. Like the qubit $T$9 gate, the qutrit $$T=\operatorname{diag}(1,\zeta,\zeta^8),\qquad \zeta=e^{2\pi i/9},$$0 gate can be injected into a circuit using magic states, and its magic states can be distilled by magic-state distillation (Glaudell et al., 2022). The literature summarized there treats this as the qutrit counterpart of the standard qubit Clifford+$$T=\operatorname{diag}(1,\zeta,\zeta^8),\qquad \zeta=e^{2\pi i/9},$$1 architecture: a stabilizer backbone supplemented by consumable non-stabilizer states.

This perspective also sharpens the comparison with the metaplectic gate

$$T=\operatorname{diag}(1,\zeta,\zeta^8),\qquad \zeta=e^{2\pi i/9},$$2

Although $$T=\operatorname{diag}(1,\zeta,\zeta^8),\qquad \zeta=e^{2\pi i/9},$$3 had been proposed as a practical alternative non-Clifford resource, the exact-synthesis comparison shows that, when at least two qutrits are available, Clifford+$$T=\operatorname{diag}(1,\zeta,\zeta^8),\qquad \zeta=e^{2\pi i/9},$$4 is a strict subset of Clifford+$$T=\operatorname{diag}(1,\zeta,\zeta^8),\qquad \zeta=e^{2\pi i/9},$$5. The inclusion is constructive: $$T=\operatorname{diag}(1,\zeta,\zeta^8),\qquad \zeta=e^{2\pi i/9},$$6 admits a Clifford+$$T=\operatorname{diag}(1,\zeta,\zeta^8),\qquad \zeta=e^{2\pi i/9},$$7 implementation with T-count $$T=\operatorname{diag}(1,\zeta,\zeta^8),\qquad \zeta=e^{2\pi i/9},$$8 and one borrowed ancilla. The strictness is exact: the $$T=\operatorname{diag}(1,\zeta,\zeta^8),\qquad \zeta=e^{2\pi i/9},$$9 gate cannot be synthesized in Clifford+$$T=\operatorname{diag}(1,\omega_9^8,\omega_9),\qquad \omega_9=e^{2\pi i/9}.$$0 (Glaudell et al., 2022).

At the level of state resources, this implies an asymmetry between T-type and R-type magic states. T magic states are at least as powerful as R magic states, because any protocol using $$T=\operatorname{diag}(1,\omega_9^8,\omega_9),\qquad \omega_9=e^{2\pi i/9}.$$1 can be recast in Clifford+$$T=\operatorname{diag}(1,\omega_9^8,\omega_9),\qquad \omega_9=e^{2\pi i/9}.$$2; the converse fails, since $$T=\operatorname{diag}(1,\omega_9^8,\omega_9),\qquad \omega_9=e^{2\pi i/9}.$$3 itself is not in Clifford+$$T=\operatorname{diag}(1,\omega_9^8,\omega_9),\qquad \omega_9=e^{2\pi i/9}.$$4 (Glaudell et al., 2022). For encyclopedia purposes, this is the cleanest statement of the qutrit T-state’s expressive advantage among currently compared single-qutrit non-Clifford resources.

5. Encoded and bosonic T-type resources in the $$T=\operatorname{diag}(1,\omega_9^8,\omega_9),\qquad \omega_9=e^{2\pi i/9}.$$5-qutrit

The paper on the $$T=\operatorname{diag}(1,\omega_9^8,\omega_9),\qquad \omega_9=e^{2\pi i/9}.$$6-qutrit does not explicitly define a “qutrit T-state,” does not use the terminology “magic state,” and does not exhibit a non-Gaussian logical gate acting as a qutrit non-Clifford resource. What it does provide is a structured bosonic qutrit encoding in two modes, built from $$T=\operatorname{diag}(1,\omega_9^8,\omega_9),\qquad \omega_9=e^{2\pi i/9}.$$7 coherent states indexed by the binary tetrahedral group

$$T=\operatorname{diag}(1,\omega_9^8,\omega_9),\qquad \omega_9=e^{2\pi i/9}.$$8

with the three cosets $$T=\operatorname{diag}(1,\omega_9^8,\omega_9),\qquad \omega_9=e^{2\pi i/9}.$$9, $T$0, and $T$1 furnishing the logical qutrit labels (Denys et al., 2022).

The logical basis arises from coset superpositions

$T$2

and an orthonormal basis $T$3 obtained by discrete Fourier transform over the coset index. The $T$4 generator $T$5 is implemented by a passive Gaussian unitary $T$6 acting as logical

$T$7

Another passive Gaussian operation,

$T$8

permutes $T$9 and $T\ket{+}$0 while fixing $T\ket{+}$1 (Denys et al., 2022).

Inside the qutrit lies an embedded $T\ket{+}$2-qubit,

$T\ket{+}$3

on which $T\ket{+}$4 acts as logical Pauli-$T\ket{+}$5, while $T\ket{+}$6 acts as a phase gate of angle $T\ket{+}$7. A particularly simple state in this subspace is

$T\ket{+}$8

which factorizes into two single-mode 4-component cat-qubit states (Denys et al., 2022).

The code is stabilized by engineered jump operators

$T\ket{+}$9

together with number-parity and SWAP symmetries. Against pure loss, in the low-loss regime and moderate amplitudes such as $\ket{S}$0, the $\ket{S}$1-qutrit sits at a local optimum of the biconvex optimization, and random encodings in the same $\ket{S}$2-state space do not perform significantly better. Compared to single-mode cat qutrits, the $\ket{S}$3-qutrit has higher entanglement fidelity for reasonable $\ket{S}$4, particularly in the low-loss regime (Denys et al., 2022).

These facts do not produce an explicit qutrit T-state, but they do motivate one. The paper itself states that the $\ket{S}$5-qutrit is a natural platform for qutrit magic-state or T-type constructions, because logical $\ket{S}$6 and related permutation operations are already built in as passive Gaussian transformations, while the non-Abelian $\ket{S}$7 structure suggests a route toward richer logical gates once suitable non-Gaussian operations are added (Denys et al., 2022).

6. Characterization, benchmarking, and experimental realization

The recent benchmarking literature treats the qutrit T-state primarily through the quality of the underlying $\ket{S}$8 gate. A randomized benchmarking scheme based on the hyperdihedral group

$\ket{S}$9

uses the unitary representation

$\{\ket{0},\ket{1},\ket{2}\}$00

whose image is $\{\ket{0},\ket{1},\ket{2}\}$01. After quotienting by global phase, this yields $\{\ket{0},\ket{1},\ket{2}\}$02 physically distinct gates. The associated twirl decomposes the Choi space into one trivial irrep and four nontrivial irreps, with only two distinct nontrivial eigenvalues, $\{\ket{0},\ket{1},\ket{2}\}$03 and $\{\ket{0},\ket{1},\ket{2}\}$04, governing the decay. Combining this T-gate benchmarking procedure with known qutrit Clifford randomized benchmarking gives complete characterization of a universal qutrit gate set (Amaro-Alcalá et al., 2023).

This benchmarking perspective matters for T-states because the fidelity of a prepared magic state is limited by the fidelity of the $\{\ket{0},\ket{1},\ket{2}\}$05 gate and the surrounding Clifford operations. The same paper emphasizes that non-Clifford gates like $\{\ket{0},\ket{1},\ket{2}\}$06 are attractive because of established magic-state distillation procedures, so characterization of the $\{\ket{0},\ket{1},\ket{2}\}$07 gate is effectively characterization of the native T-state resource pipeline (Amaro-Alcalá et al., 2023).

On the hardware side, a superconducting microwave experiment has now demonstrated arbitrary qutrit-state transfer between remote nodes with a mean transferred-state fidelity of $\{\ket{0},\ket{1},\ket{2}\}$08 and a qutrit process fidelity of $\{\ket{0},\ket{1},\ket{2}\}$09, both beyond effective-qubit benchmarks. Using partial-transfer operations, the same platform reconstructed a remote two-qutrit state with negativity $\{\ket{0},\ket{1},\ket{2}\}$10, dense-coding capacity $\{\ket{0},\ket{1},\ket{2}\}$11 bits, and CGLMP parameter $\{\ket{0},\ket{1},\ket{2}\}$12 (Li et al., 28 Jun 2026). The experiment does not explicitly target non-Clifford qutrit T-states or magic states, but it does provide the local single-qutrit control, phase tuning, tomography, and networked qutrit transmission required to prepare, transmit, and characterize states of the form

$\{\ket{0},\ket{1},\ket{2}\}$13

This suggests a near-term route from abstract qutrit T-state theory to distributed superconducting implementations (Li et al., 28 Jun 2026).

Taken together, these developments show that “qutrit T-state” is best understood as a family resemblance term rather than a single universally fixed object. In the narrowest sense it is the magic state $\{\ket{0},\ket{1},\ket{2}\}$14 for a third-level non-Clifford qutrit phase gate. In the broader magic-state literature it often points instead to the strange state as the maximally magical and maximally symmetric single-qutrit resource, with the equatorial state as the cleanest direct injection resource. In encoded and hardware settings, the same term now also marks the search for robust, characterizable, and eventually distributable non-Clifford qutrit resources.