SqrtGate: Theory, Synthesis, and Applications

• SqrtGate is a gate whose repeated application yields fundamental operations like Pauli or Clifford, and it is defined both in quantum circuits and categorical models.
• It plays a central role in circuit synthesis by enabling efficient cancellation and commutation rules, thereby reducing resource costs in fault-tolerant quantum designs.
• SqrtGate is also applied in quantum arithmetic circuits and MoE neural architectures, ensuring optimized operations such as RMS invariance and minimal qubit usage.

A SqrtGate is any gate whose repeated application yields a fundamental quantum or classical gate—in particular, it is often used to denote gates that “square” to a Pauli or Clifford operation (e.g., $\sqrt{X}$, $\sqrt{Z}$, and related generalizations), or, in algorithmic contexts, a quantum circuit implementing an arithmetic square root. The term also refers to square root operations and their categorical or circuit-theoretic axiomatizations in a variety of quantum computing frameworks. Notably, “SqrtGate” has further acquired meaning as a precise mechanism in recent optimizations for both near-term fault-tolerant quantum circuits and large-scale neural architectures. This article surveys the theoretical definitions, formal properties, circuit synthesis, categorical presentations, and algorithmic realizations of SqrtGate across quantum information science and quantum-inspired machine learning.

1. Formal Definitions and Representations

The archetypal SqrtGate in quantum circuit theory is the $\sqrt{X}$ gate, commonly denoted $V$. Its defining characteristic is $V^2 = X$, where $X$ is the Pauli-X operator (bit-flip):

$$V = \frac12 \begin{pmatrix} 1 + i & 1 - i \ 1 - i & 1 + i \end{pmatrix}, \qquad V^2 = X$$

Similarly, the Controlled-$V$ (C$V$) acts on two qubits and is defined by:

$$\mathrm{C}\text{-}V = \begin{pmatrix} 1 & 0 & 0 & 0 \ 0 & 1 & 0 & 0 \ 0 & 0 & \tfrac{1+i}{2} & \tfrac{1-i}{2} \ 0 & 0 & \tfrac{1-i}{2} & \tfrac{1+i}{2} \end{pmatrix}$$

(0710.4694)

Categorically, SqrtGates are abstracted as morphisms satisfying specific square-root equations. In the rig groupoid model of quantum circuits, one introduces a primitive $\sqrt{Z}$0 such that $\sqrt{Z}$1, where $\sqrt{Z}$2 is the symmetry (swap map) on the biproduct. This $\sqrt{Z}$3 becomes the categorical $\sqrt{Z}$4, and together with an appropriate “phase” morphism $\sqrt{Z}$5, suffices for equational completeness of Clifford-type circuits (Carette et al., 2023).

In algorithmic settings, notably in arithmetic circuit synthesis, the SqrtGate may refer to a quantum circuit that computes integer square roots via reversible logic, with resource and depth optimizations tailored to fault-tolerant Clifford+T architectures (Muñoz-Coreas et al., 2017).

2. Algebraic and Categorical Foundations

The importance of SqrtGate types is grounded in their algebraic role as root operations in circuit decompositions and categorical models:

• Clifford circuits and completeness: In the categorical model by Carette et al., the addition of $\sqrt{Z}$6 (with $\sqrt{Z}$7) and an eighth-root-of-unity $\sqrt{Z}$8 yields an equationally complete presentation for Clifford and Clifford+T gate sets on up to two qubits. The crucial non-degeneracy condition (Euler axiom) is:

$\sqrt{Z}$9

which enforces the correct Euler decomposition of the Hadamard gate in the categorical language (Carette et al., 2023).

• Group-theoretical synthesis: In multiple-valued logic synthesis for quantum circuits, gates such as $\sqrt{X}$0 and controlled-$\sqrt{X}$1 are interpreted as generators of permutation groups acting on four-symbol encodings (binary plus “half-rotated” states). The entire circuit synthesis problem is transformed into factoring group permutations by these generators, simplifying search for minimal quantum-cost implementations (0710.4694).

3. Circuit Synthesis and Optimization

SqrtGates play a central role in synthesis and optimization strategies for reversible circuits:

• Cost model and exact synthesis: The approach of Yang et al. assigns unit cost to two-qubit gates (CNOT, C$\sqrt{X}$2), with single-qubit SqrtGates (V, $\sqrt{X}$3) considered “free.” The synthesis procedure is enumerative: all possible products of generators up to a fixed cost are examined, and those realisable at lower cost are filtered out, yielding minimal two-qubit gate decompositions for arbitrary reversible functions. Example results include canonical decompositions of the Toffoli and Peres gates using minimal C$\sqrt{X}$4 and CNOT counts (0710.4694).
• Cancellation and commutation rules: The relation $\sqrt{X}$5 and similar rules for $\sqrt{X}$6, and controlled gates permit aggressive cancellation and commutation during synthesis, critical for reducing circuit depth and resource usage in larger designs (0710.4694).

4. Categorical Quantum Programming and Equational Theories

The abstract treatment of SqrtGate as a primitive in universal quantum programming languages enables diagrammatic and algebraic reasoning for quantum circuits:

• Extension of classical reversible languages: By adjoining two primitives (an eighth-root scalar and a square root of swap) to the syntax of the classical $\sqrt{X}$7 language, one obtains an extension ($\sqrt{X}$8) with a sound and complete equational theory for all Clifford and Clifford+T operations. The square-root gate $\sqrt{X}$9 directly defines $V$0 in any model, and the resulting category supports direct diagrammatic composition without explicit matrices (Carette et al., 2023).
• Universal circuit identities: The equational theory induced by these primitives recovers standard identities underlying circuit optimization, including Selinger’s relations for the Clifford+T family, via the rig-groupoid and distributivity structure (Carette et al., 2023).

5. Quantum Arithmetic: SqrtGate as an Algorithmic Circuit

In computational quantum arithmetic, SqrtGate refers specifically to the optimized construction of a square-root function as a quantum circuit:

• Non-restoring square-root algorithm: The SqrtGate$V$1 is a reversible circuit mapping $V$2, where $V$3 is the $V$4-bit input, $V$5, $V$6, and the ancillary register is garbage-free. This is achieved via a sequence of three subcircuits implementing initialization, iterated operations, and final restoration steps, using Clifford+T arithmetic with optimized T-count and qubit usage (Muñoz-Coreas et al., 2017).
• Resource scaling: The construction achieves T-count $V$7, T-depth $V$8, and qubit count $V$9, strictly improving asymptotic complexity relative to prior proposals (which required $V^2 = X$0 qubits or garbage removal via Bennett’s uncomputation) (Muñoz-Coreas et al., 2017).
• Practical deployment: SqrtGate circuits of this kind are readily integrated into mid-scale modular quantum computers, particularly for scientific computing workloads or cryptographic primitives requiring arithmetic square roots, since fault-tolerant architectures (surface-code, color-code) support their “garbage-free” realization (Muñoz-Coreas et al., 2017).

Design	T-count	Qubits
Sultana et al.	$V^2 = X$1	$V^2 = X$2
Bhaskar et al.	$V^2 = X$3	$V^2 = X$4
Ananthalakshmi I	$V^2 = X$5	$V^2 = X$6
Proposed (Muñoz-Coreas et al., 2017)	$V^2 = X$7	$V^2 = X$8

6. Physical Realizations and Protected Non-Clifford SqrtGates

Advances in protected gate synthesis have produced physical SqrtGates that achieve non-Clifford rotations with exponential robustness:

• Exponentially robust $V^2 = X$9-gate protocol: Driven-dissipative circuits with ancillary Josephson junctions and engineered quartic flux potentials enable the realization of a protected $X$0 gate at the physical level. The protocol combines quartic “drive” segments and free-evolution intervals to obtain the unitary $X$1 on the logical code space (GKP-like states) (O'Brien et al., 25 Jul 2025).
• Topological protection and error analysis: Fourfold degenerate code wells and strong four-photon dissipation provide topological protection, with infidelity suppressed exponentially in device imperfection parameters and bath temperature: $X$2. Typical gate timescales are $X$3–$X$4s for GHz-class resonators, and realistic error budgets yield fidelities $X$5 (O'Brien et al., 25 Jul 2025).
• Comparison to surface-code state injection: Unlike magic-state injection, which is resource-intensive and limited by noisy T-state distillation, the protected $X$6 protocol delivers non-Clifford gates natively on hardware, with fidelity and hardware demands favorable to near-term implementations (O'Brien et al., 25 Jul 2025).

7. SqrtGate Mechanisms in Machine Learning: MoE Granularity Scaling

In large-scale Mixture-of-Experts (MoE) neural models, the SqrtGate concept appears as a solution to scaling-induced RMS shrinkage:

• Hypersphere optimization and RMS invariance: Under Frobenius-norm sphere constraint (MuonH-style), normal Top-$X$7 softmax gating in MoE architectures causes output RMS to scale as $X$8. The SqrtGate mechanism modifies the routed output to be the weighted sum of $X$9, ensuring the output RMS remains invariant with respect to $$V = \frac12 \begin{pmatrix} 1 + i & 1 - i \ 1 - i & 1 + i \end{pmatrix}, \qquad V^2 = X$$0. This cancellation is exact in the regime of equal-per-expert RMS and weak inter-expert correlations (Ren et al., 30 Mar 2026).
• Empirical effects: The SqrtGate MoE mechanism maintains constant routed-branch RMS across $$V = \frac12 \begin{pmatrix} 1 + i & 1 - i \ 1 - i & 1 + i \end{pmatrix}, \qquad V^2 = X$$1 and ensures bounded router $$V = \frac12 \begin{pmatrix} 1 + i & 1 - i \ 1 - i & 1 + i \end{pmatrix}, \qquad V^2 = X$$2-values, eliminating pathological logit spikes and improving validation losses. Its compatibility with hypersphere-constrained auxiliary load-balancing objectives enables larger tuning parameters without instability (Ren et al., 30 Mar 2026).
• Practical impact: SqrtGate introduces no extra learnable parameters and minimal computational overhead (one square root per routed weight). Its invariance property simplifies hyperparameter transfer across MoE granularities and improves device utilization for training LLMs (Ren et al., 30 Mar 2026).

• (0710.4694) Yang, Grinshpon, Perkowski, "Exact Synthesis of 3-Qubit Quantum Circuits from Non-Binary Quantum Gates Using Multiple-Valued Logic and Group Theory"
• (Muñoz-Coreas et al., 2017) Muñoz-Coreas, Thapliyal, "T-count and Qubit Optimized Quantum Circuit Design of the Non-Restoring Square Root Algorithm"
• (Carette et al., 2023) Carette, Lemay, Staton, "With a Few Square Roots, Quantum Computing is as Easy as Π"
• (O'Brien et al., 25 Jul 2025) Nathan et al., "Exponentially robust non-Clifford gate in a driven-dissipative circuit"
• (Ren et al., 30 Mar 2026) Tran, Barham, Kakade, et al., "Rethinking LLM Scaling under Transferable Hypersphere Optimization"