发表机构
QpiAI India Private Ltd.（QpiAI印度私人有限公司）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
该研究结合周期性DFT与有源空间量子计算，通过AVAS和自然轨道截断将Ru单原子位点N2加氢反应映射为16量子比特问题，并用ADAPT-VQE求解，验证了量子模拟可行性与精度。
AI 中文摘要
在温和条件下选择性活化二氮（N$_2$）是困难的：N$\equiv$N是化学中最强的键之一，且大多数能够断裂该键的多相催化剂需要高温高压。原子级分散的过渡金属位点为研究涉及的强关联中间体提供了一条计算上易处理的途径。我们将周期性DFT计算与有限、非周期性表面碎片上的关联有源空间计算相结合，展示了在Ru(0001)上孤立Ru$_1$位点处加氢步骤RuH$_2$(N$_2$)* $\rightarrow$ RuH(NNH)*的工作流程。从周期性弛豫结构中提取第一壳层碎片，围绕Ru-N/N-H键重排使用活性原子价空间（AVAS）方法选择有源空间，并通过自然轨道截断降低其成本，同时保留相关的关联自由度。约化的哈密顿量通过Jordan-Wigner变换映射到量子比特，并使用自适应导数组装的赝Trotter变分量子本征求解器（ADAPT-VQE）求解。使用强收缩NEVPT2和DSRG-MRPT2方法考察了有源空间之外的动态关联。AVAS给出22量子比特的有源空间，自然轨道截断将其压缩至16量子比特，与未截断的CASCI态能量误差在0.21 kcal/mol以内（反应能误差为0.2 kcal/mol）。在该约化空间中，态矢量ADAPT-VQE在固定池梯度停止准则下对两个反应态均收敛。NEVPT2对有限Ru碎片产生异常大且态不平衡的修正，而DSRG-MRPT2在其默认流参数下保持吸热反应能，尽管其大小强烈依赖于该参数。
英文摘要
Selective activation of dinitrogen (N$_2$) under mild conditions is difficult: N$\equiv$N is one of the strongest bonds in chemistry, and most heterogeneous catalysts capable of breaking it require high temperature and pressure. Atomically dispersed transition-metal sites offer a computationally tractable route to studying the strongly correlated intermediates involved. We connect periodic DFT calculations with correlated active-space calculations on a finite, non-periodic surface fragment, demonstrating the workflow for the hydrogenation step RuH$_2$(N$_2$)* $\rightarrow$ RuH(NNH)* at an isolated Ru$_1$ site on Ru(0001). A first-shell fragment is extracted from the periodically relaxed structure, an active space is selected with Active Atomic Valence Space (AVAS) around the Ru-N/N-H bond reorganization, and natural-orbital truncation reduces its cost while retaining the relevant correlated degrees of freedom. The reduced Hamiltonian is mapped to qubits with the Jordan-Wigner transformation and solved with the adaptive derivative-assembled pseudo-Trotter variational quantum eigensolver (ADAPT-VQE). Dynamic correlation beyond the active space is examined with strongly contracted NEVPT2 and DSRG-MRPT2. AVAS gives a 22-qubit active space, which natural-orbital truncation compresses to 16 qubits, reproducing the untruncated CASCI state energies to within 0.21 kcal/mol (reaction-energy error 0.2 kcal/mol). Statevector ADAPT-VQE in this reduced space converges for both reaction states under a fixed pool-gradient stopping criterion. NEVPT2 yields anomalously large, state-imbalanced corrections for the finite Ru fragment, while DSRG-MRPT2 retains an endothermic reaction energy at its default flow parameter, although its magnitude depends strongly on that parameter.
