The challenge of quantum computing is to combine error resilience with universal computation. There are many finite sets of gates that are universal, and a standard choice is to augment the set of Clifford gates by a non-Clifford unitary such as the T gate. Given a CSS code, we introduce a method of synthesizing all possible diagonal physical gates that preserve the codespace and induce a target logical gate. We denote an 〚n, k = k1− k2, d〛 CSS code $\mathcal{C}$ by $\operatorname{CSS} \left( {X,{\mathcal{C}_2};Z,\mathcal{C}_1^ \bot } \right)$, where the [n, k2] binary code ${\mathcal{C}_2}$ determines the X-stabilizers in $\mathcal{C}$, and the [n, n−k1] binary code $\mathcal{C}_1^ \bot $ determines the Z-stabilizers in $\mathcal{C}$. The diagonal entries of a diagonal physical gate are indexed by binary vectors in $\mathbb{F}_2^n$. We show that a diagonal physical gate preserves the CSS codespace if and only if entries from the same coset of ${\mathcal{C}_2}$ in ${\mathcal{C}_1}$ (same X-logical) are identical. We also show that the target logical operator only specifies ${2^{{k_1}}}$ out of 2ndiagonal entries of the diagonal physical gate. The remaining degrees of freedom can be used to optimize implementation of the physical gate within a particular quantum computing infrastructure. This encompasses optimization with respect to locality of the physical gate, a criterion that is essential to fault tolerance. When the target logical operator is the identity, the physical gates that preserve the CSS code represent noise operators to which the codespace is oblivious. We illustrate our method by providing several examples of code-gate pairs for which the target logical gate is a non-Clifford unitary. The framework is extended to stabilizer codes in https://arxiv.org/abs/2109.13481.

Physical platforms such as trapped ions suffer from coherent noise that does not follow a simple stochastic model. Stochastic errors in quantum systems occur randomly but coherent errors are more damaging since they can accumulate in a particular direction. We consider coherent noise acting transversally, giving rise to an effective error which is a$Z$-rotation on each qubit by some angle$\theta $. Rather than address coherent noise through active error correction, we investigate passive mitigation through decoherence free subspaces. In the language of stabilizer codes, we require the noise to preserve the code space, and to act trivially (as the logical identity operator) on the protected information. Thus, we develop necessary and sufficient conditions for all transversal$Z$-rotations to preserve the code space of a stabilizer code. These conditions require the weight-$2~Z$-stabilizers to cover all the qubits that are in the support of the$X$-component of some stabilizer. Furthermore, the weight-$2~Z$-stabilizers generate a direct product of single-parity-check codes with even block length. By adjusting the sizes of these components, we are able to construct a large family of QECC codes oblivious to coherent noise, one that includes the$[[4L^{2}, 1, 2L]]$Shor codes. The Shor codes are examples of constant excitation codes, where logical qubits are encoded as a code state that is a sum of physical states indexed by binary vectors with the same weight. Constant excitation codes are oblivious to coherent noise since a transversal$Z$-rotation acts as a global phase. We prove that a CSS code is oblivious to coherent noise if and only if it is a constant excitation code, and that if the code is error-detecting, then the (constant) weights in different cosets of the$X$-stabilizers are identical.

Physical platforms such as trapped ions suffer from coherent noise that does not follow a simple stochastic model. We view coherent errors as rotations about a particular axis, and observe that since they can accumulate coherently over time, they can be more damaging. It is natural to consider coherent noise acting transversally giving rise to an effective error, which is a$Z$-rotation on each qubit by some angle$\theta$. Rather than addressing coherent noise through active error correction, we instead investigate passive mitigation through decoherence free subspaces. In the language of stabilizer codes, we require the noise to preserve the code space, and to act trivially (as the logical identity operator) on the protected information. Thus, we develop necessary and sufficient conditions for all transversal$Z$-rotations to preserve the code space of a stabilizer code. These conditions require the existence of a large number of weight 2$Z$-stabilizers, and together, these weight 2$Z$-stabilizers generate a direct product of single-parity-check codes. By adjusting the size of these components, we are able to construct a large family of CSS codes, oblivious to coherent noise, that includes the$[[4L^{2}, 1,2L]]$Shor codes. Given$m$even and given any$[[n, k, d]]$CSS code, we can construct an$[[mn, k, d^{\prime}\geq d]]$CSS code that is oblivious to coherent noise. This result is generalized to stabilizer codes in [Hu, Liang, Rengaswamy, and Calderbank 2020]. The MacWilliams Identities play a central role in the technical analysis, and classical coding theorists may be interested in connections to classical codes with all weights divisible by some integer$d$.

Efficient simulation of large-scale mammalian brain models provides a crucial computational means for understanding complex brain functions and neuronal dynamics. However, such tasks are hindered by significant computational complexities. In this work, we attempt to address the significant computational challenge in simulating large-scale neural networks based on biophysically plausible Hodgkin-Huxley (HH) neuron models. Unlike simpler phenomenological spiking models, the use of HH models allows one to directly associate the observed network dynamics with the underlying biological and physiological causes, but at a significantly higher computational cost. We exploit recent commodity massively parallel graphics processors (GPUs) to alleviate the significant computational cost in HH model based neural network simulation. We develop look-up table based HH model evaluation and efficient parallel implementation strategies geared towards higher arithmetic intensity and minimum thread divergence. Furthermore, we adopt and develop advanced multi-level numerical integration techniques well suited for intricate dynamical and stability characteristics of HH models. On a commodity GPU card with 240 streaming processors, for a neural network with one million neurons and 200 million synaptic connections, the presented GPU neural network simulator is about 600X faster than a basic serial CPU based simulator, 28X faster than the CPU implementation of the proposed techniques, and only two to three times slower than the GPU based simulation using simpler phenomenological spiking models.