Diagonal Transversal Clifford Gates

• Diagonal transversal Clifford gates are logical operations that are diagonal in the computational basis and implemented transversally on encoded qubits or qudits.
• They are characterized through phase gadget representations and polynomial-phase descriptions, with canonical examples including Z, S, and CZ gates.
• Frameworks like CSS codes and stabilizer symmetries elucidate their fault-tolerance limits and guide extensions beyond traditional Pauli stabilizer constraints.

Searching arXiv for recent and foundational papers on diagonal transversal Clifford gates. Diagonal transversal Clifford gates are logical operations that are simultaneously diagonal in the computational basis and implemented transversally on encoded qubits or qudits, typically by tensor products of single-site phase gates and, in generalized LDPC settings, by uniformly bounded-locality diagonal operations acting across code blocks. Canonical examples are $Z$, $S$, and $CZ$; within the diagonal part $\mathcal D_\ell$ of the Clifford hierarchy they occupy the Clifford sector $\mathcal D_2$, while gates such as $T$, $CCZ$, and higher controlled phases lie above it. The topic sits at the intersection of stabilizer-code symmetry, phase-polynomial descriptions of the Clifford hierarchy, CSS code design, and fault-tolerance no-go theorems. It also has a terminological subtlety: some recent multi-block classification work uses “diagonal” to mean repeated action of the same Clifford on each physical qubit position, rather than diagonal-in-basis action (Cui et al., 2016, Kissinger et al., 2024, Dasu et al., 14 Jul 2025).

1. Terminology and conceptual scope

The phrase “diagonal transversal Clifford gate” is used in two related but distinct ways. In the most common sense, it refers to a logical gate whose matrix is diagonal in the computational basis and whose physical implementation is transversal. In the multi-block stabilizer-symmetry literature, “diagonal” can instead refer to the repetition of the same $\ell$-qubit Clifford on each physical qubit column across $\ell$ code blocks, whether or not that Clifford is diagonal in the computational basis (Dasu et al., 14 Jul 2025).

Usage	Meaning	Typical examples
Computational-basis diagonal	Basis-state phases, usually products of $Z(\theta)$-type gates	$S$0, $S$1, $S$2
Repeated blockwise action	Same Clifford applied on each physical position across code blocks	Groups $S$3

The notion of transversality also varies. In the simplest CSS setting it means a tensor product of single-qubit unitaries on the physical qubits. In folded-surface-code constructions it is generalized to depth-1 gates on disjoint blocks of size at most $S$4. In recent qLDPC work it is used in a generalized LDPC sense: multi-qubit gates act only across code blocks and each physical qubit is acted on a constant number of times (Kissinger et al., 2024, Moussa, 2016, Li et al., 2 Apr 2026).

This distinction matters because the computational-basis-diagonal literature is organized by phase structure and hierarchy level, whereas the repeated-blockwise literature is organized by endomorphism algebras and symplectic matrix groups. The two viewpoints are compatible, but they answer different classification questions.

2. Placement in the Clifford hierarchy

For prime-dimensional qudits, diagonal elements of the Clifford hierarchy admit a complete polynomial-phase description: a diagonal unitary is determined by phases of the form $S$5, where $S$6 is a polynomial over $S$7, and its hierarchy level is fixed by the denominator precision $S$8 together with the degree, or multivariate weight, of the highest relevant monomial (Cui et al., 2016). For qubits this reproduces the familiar pattern $S$9, $CZ$0, and $CZ$1.

A useful operational description comes from phase gadgets. In the ZX-calculus formulation, phase gadgets span all diagonal unitaries, and gadgets with angles that are multiples of $CZ$2 span the diagonal part $CZ$3 of the $CZ$4-th hierarchy level. In particular, diagonal Clifford gates $CZ$5 use phases that are multiples of $CZ$6, whereas $CZ$7 uses multiples of $CZ$8 (Kissinger et al., 2024).

For qubits, an important structured subclass is given by quadratic-form diagonal gates

$CZ$9

with $\mathcal D_\ell$0 symmetric over $\mathcal D_\ell$1. These gates form a large subgroup of diagonal $\mathcal D_\ell$2-th-level gates, include all $\mathcal D_\ell$3-local diagonal gates, and recover the standard diagonal Cliffords at $\mathcal D_\ell$4. In that language, $\mathcal D_\ell$5 and $\mathcal D_\ell$6 are diagonal Clifford examples, while $\mathcal D_\ell$7 and $\mathcal D_\ell$8 appear at the next level (Rengaswamy et al., 2019).

This hierarchy-theoretic viewpoint fixes the boundary of the subject. Diagonal transversal Clifford gates are the level-$\mathcal D_\ell$9 sector of a larger family of transversal diagonal gates, and much of the modern literature studies them alongside non-Clifford relatives because the same algebraic machinery often treats $\mathcal D_2$0, $\mathcal D_2$1, $\mathcal D_2$2, $\mathcal D_2$3, and $\mathcal D_2$4 uniformly.

3. Stabilizer and CSS criteria for diagonal transversality

For qubit stabilizer codes, transversal diagonal unitaries are severely constrained. A foundational classification shows that if a nontrivial stabilizer code admits a transversal diagonal logical operation, then each physical one-qubit factor must have diagonal entries of the form $\mathcal D_2$5. Consequently, every diagonal logical gate obtainable in this way lies somewhere in the Clifford hierarchy; diagonal Clifford gates are precisely the cases where the induced logical phases have denominator $\mathcal D_2$6 in lowest terms (Anderson et al., 2014).

For CSS codes, recent work gives an explicit module-theoretic description. If $\mathcal D_2$7 and $\mathcal D_2$8 with $\mathcal D_2$9, one defines

$T$0

together with the submodules $T$1 of vectors inducing transversal logical operators and $T$2 of vectors acting as the logical identity. The characterizing equations are written explicitly in terms of coordinatewise products of basis vectors of $T$3 and $T$4, and the framework recovers and extends CSS-$T$5, triorthogonal, and divisible-code conditions (Camps-Moreno et al., 29 Jan 2026).

A complementary channel-theoretic formulation uses generator coefficients $T$6, indexed by $T$7-syndromes and logical $T$8-cosets. In that language a diagonal physical gate preserves the code space exactly when

$T$9

and the induced logical operator is obtained directly from the coefficients $CCZ$0. This formulation makes syndrome dependence, logical correction rules, and the role of $CCZ$1-stabilizer signs explicit (Hu et al., 2021).

For higher-level diagonal gates, the same CSS machinery specializes to strong combinatorial constraints. In the $CCZ$2-gate case, the QFD formalism yields Heisenberg-picture conditions that reduce on CSS codes to triorthogonality and mod-$CCZ$3 weight congruences. This is why triorthogonal matrices occupy a central position in the study of diagonal transversal gates: they characterize the most general CSS codes for which physical transversal $CCZ$4 realizes logical transversal $CCZ$5, while divisible-code conditions govern when a given transversal diagonal acts as the logical identity (Rengaswamy et al., 2019).

4. Graphical and algebraic classification frameworks

The ZX-calculus gives a graphical characterization of transversal diagonal gates on CSS codes. A central result states that a product of $CCZ$6-phase gadgets with connectivity matrix $CCZ$7 is Clifford if and only if $CCZ$8 is semi-triorthogonal, and it is the identity if and only if $CCZ$9 is triorthogonal. The same paper proves that a CSS code with $\ell$0-logical and $\ell$1-stabilizer matrices $\ell$2 and $\ell$3 admits a transversal implementation of a diagonal $\ell$4 gate $\ell$5 exactly when there exists a matrix $\ell$6 whose rows have Hamming weight $\ell$7 such that

$\ell$8

is triorthogonal; it further notes that the same argument specializes to $\ell$9, giving a characterization of all transversal diagonal Clifford unitaries (Kissinger et al., 2024).

A different algebraic viewpoint classifies repeated-blockwise diagonal transversal Clifford groups through endomorphism algebras. For a qubit stabilizer code $\ell$0, the group

$\ell$1

is shown to belong, up to local-diagonal Clifford equivalence, to exactly one of six families: $\ell$2 Here “diagonal” refers to repeating the same $\ell$3-qubit Clifford on every physical column, not to computational-basis diagonality; nevertheless, the result is part of the modern taxonomy of transversal Clifford symmetries (Dasu et al., 14 Jul 2025).

The quadratic-form diagonal formalism supplies a third unifying language. Symmetric matrices over $\ell$4 parametrize a subgroup of diagonal $\ell$5-th-level gates, with group law given by matrix addition and conjugation on Paulis expressed through an integer-symplectic matrix

$\ell$6

At $\ell$7, this reduces to the familiar binary symplectic description of diagonal Clifford gates; for larger $\ell$8, it recursively generates higher-level diagonal actions (Rengaswamy et al., 2019).

5. Code families and concrete realizations

Several code families realize nontrivial diagonal transversal Clifford gates explicitly. For the Steane $\ell$9 CSS code, strongly transversal $Z(\theta)$0 implements the logical phase gate $Z(\theta)$1, making it a canonical example of a diagonal transversal Clifford gate on a finite-distance code (Anderson et al., 2014).

Folded surface codes provide a topological construction with transversal Clifford structure in two spatial dimensions. Their logical $Z(\theta)$2 is implemented by applying $Z(\theta)$3 and $Z(\theta)$4 on fold qudits and $Z(\theta)$5 and $Z(\theta)$6 on paired top-bottom qudits, so the logical diagonal Clifford is realized by a depth-1 local circuit built entirely from diagonal Clifford gates (Moussa, 2016).

The generator-coefficient framework exhibits smaller CSS examples. On the $Z(\theta)$7 code, a suitable QFD operator with $Z(\theta)$8 yields $Z(\theta)$9, and this preserves the code space while inducing a logical diagonal Clifford $S$00. The same framework also shows how changing the signs of $S$01-stabilizers can alter syndrome statistics and the induced logical diagonal action (Hu et al., 2021).

Recent qLDPC and qLTC work places diagonal Clifford transversality in an asymptotic setting. A general algebraic-topological “cupcap” construction yields transversal logical multi-controlled-$S$02 gates on almost-good qLDPC and qLTC families; the non-Clifford emphasis is on $S$03, but the same framework includes the Clifford cases. In particular, for $S$04 it gives logical $S$05, and the paper states that earlier work obtains qLDPC codes with optimal parameters $S$06 supporting a nontrivial transversal logical $S$07 gate (Li et al., 2 Apr 2026).

Taken together, these examples show that diagonal transversal Clifford gates appear in several distinct regimes: small stabilizer and CSS codes, topological constructions with folded geometry, and asymptotically good or almost-good LDPC families. What changes across these regimes is not the gate algebra but the mechanism certifying nontrivial logical action.

6. Limitations, obstructions, and current directions

Diagonal transversal Clifford gates are valuable precisely because they occupy a narrow region between fault-tolerant usefulness and strong no-go theorems. The standard obstruction is Eastin–Knill: no code family of fixed local dimension can possess a universal transversal gate set. Recent qLDPC work explicitly emphasizes that its diagonal constructions do not circumvent Eastin–Knill and remain compatible with Bravyi–König-type constraints (Li et al., 2 Apr 2026).

Concrete code families can be much more restrictive. For hypergraph product codes in the vertical sector, a Bravyi–König-style argument shows that any transversal gate must act logically as a Clifford. As a result, any nontrivial diagonal transversal gate on these codes is necessarily a diagonal Clifford; transversal $S$08, $S$09, and higher diagonal non-Cliffords are excluded (Burton et al., 2020).

At the same time, newer work indicates that the apparent boundary is tied to the Pauli-stabilizer framework rather than to diagonality itself. Non-Abelian surface-code constructions realize purely $S$10D transversal phase gates at arbitrary hierarchy levels, including $S$11, by stacking symmetry-protected topological phases on quantum doubles $S$12. This suggests that the obstruction to high-level diagonal transversality is not universal once one leaves Pauli stabilizer codes (Warman et al., 15 Dec 2025).

Nonadditive codes point in the same direction. A recent search over small nonadditive codes found a $S$13 code with a transversal $S$14 gate, $S$15 codes with transversal $S$16 and $S$17, and an $S$18 code with transversal $S$19, using an SS-LP construction specialized to diagonal gates. This suggests a richer landscape of diagonal transversal groups than is visible in additive stabilizer codes alone (Zhang et al., 29 Apr 2025).

The contemporary picture is therefore stratified. In stabilizer and CSS codes, diagonal transversal Clifford gates are well understood through dyadic phase restrictions, orthogonality conditions, and code-symmetry classifications. In qLDPC and topological codes they can coexist with strong asymptotic parameters, but only under carefully structured constructions. Outside the additive Pauli setting, the broader diagonal-transversal landscape appears substantially larger, and current research increasingly treats the Clifford cases not as isolated curiosities but as the lowest nontrivial layer of a general theory of transversal diagonal gates.