Zhao Qing, Zhang Xing, Martirez John Mark P, Carter Emily A
Department of Mechanical and Aerospace Engineering, Princeton University, Princeton, New Jersey 08544-5263, United States.
J Chem Theory Comput. 2020 Nov 10;16(11):7078-7088. doi: 10.1021/acs.jctc.0c00341. Epub 2020 Oct 20.
Embedded (emb-) correlated wavefunction (CW) theory enables accurate assessments of both ground- and excited-state reaction mechanisms involved in heterogeneous catalysis. Embedded multireference second-order perturbation theory (emb-MRPT2) based on reference wavefunctions generated via embedded complete active space self-consistent field (emb-CASSCF) theory is currently state-of-the-art. However, the factorial scaling of CASSCF limits the size of active space and the complexity of systems that can be studied. Here, we assess the efficacy of an alternative CW method, adaptive sampling configuration interaction (ASCI)-which enables large active spaces to be used-for studying surface reactions. We couple ASCI with density functional embedding theory (DFET) and benchmark its performance for two reactions: H desorption from and CH dissociation on the Cu(111) surface. Unlike embedded complete active space second-order perturbation theory (emb-CASPT2) that accurately reproduces a measured H desorption barrier, embedded ASCI, using a very large active space (though one that still comprises a small portion of the full set of orbitals) fails to do so. Adding an extra correlation term from embedded Møller-Plesset second-order perturbation theory (emb-MP2) improves the desorption barrier and endothermicity predictions. Thus, the inaccuracy of embedded ASCI comes from the missing dynamic correlation from the many other electrons and orbitals not included in the active space. For CH dissociation, again embedded ASCI overestimates the dissociation barrier compared to emb-CASPT2 predictions. Adding dynamic correlation from emb-MP2 helps correct the barrier. However, this composite approach suffers from double counting of correlation within embedded ASCI followed by emb-MP2 calculations. We therefore conclude that the state-of-the-art emb-MRPT2 based on reference wavefunctions generated via emb-CASSCF remains the method of choice for studying surface reactions. emb-ASCI is useful when large active spaces beyond the limit of emb-CASSCF are essential, such as to study complex surface reactions with significant multiconfigurational character (static correlation) but weak dynamic correlation.
嵌入（emb-）相关波函数（CW）理论能够准确评估多相催化中涉及的基态和激发态反应机理。基于通过嵌入完全活性空间自洽场（emb-CASSCF）理论生成的参考波函数的嵌入多参考二阶微扰理论（emb-MRPT2）是目前的最先进方法。然而，CASSCF的阶乘缩放限制了活性空间的大小以及可研究系统的复杂性。在这里，我们评估了一种替代的CW方法——自适应采样组态相互作用（ASCI）——用于研究表面反应的有效性，该方法能够使用大活性空间。我们将ASCI与密度泛函嵌入理论（DFET）相结合，并对其在两个反应中的性能进行基准测试：H从Cu(111)表面脱附以及CH在Cu(111)表面解离。与能准确再现测量的H脱附势垒的嵌入完全活性空间二阶微扰理论（emb-CASPT2）不同，使用非常大的活性空间（尽管它仍然只包含全套轨道的一小部分）的嵌入ASCI却无法做到这一点。添加来自嵌入的莫勒-普列斯特定则二阶微扰理论（emb-MP2）的额外相关项可改善脱附势垒和吸热性预测。因此，嵌入ASCI的不准确之处在于活性空间中未包含的许多其他电子和轨道所缺失的动态相关性。对于CH解离，与emb-CASPT2预测相比，嵌入ASCI同样高估了解离势垒。添加来自emb-MP2的动态相关性有助于校正势垒。然而，这种复合方法在嵌入ASCI随后进行emb-MP2计算时会出现相关项的双重计算问题。因此，我们得出结论，基于通过emb-CASSCF生成的参考波函数的最先进的emb-MRPT2仍然是研究表面反应的首选方法。当超出emb-CASSCF极限的大活性空间至关重要时，例如研究具有显著多组态特征（静态相关性）但动态相关性较弱的复杂表面反应时，emb-ASCI是有用的。
