When I was working on quantum transport at Tianjin University, I often used the non-equilibrium Green's function (NEGF) method to calculate electron transmission and current through a device. Given a Hamiltonian describing electronic energies and couplings, along with electrodes and an applied bias, the program produces a result. But if we want to know how much the current changes when a parameter is adjusted, we usually have to run the calculation again. That gets slow when there are many parameters.
diff --git a/_pages/research-deeptb-negf-zh.md b/_pages/research-deeptb-negf-zh.md
new file mode 100644
index 0000000..d5e0457
--- /dev/null
+++ b/_pages/research-deeptb-negf-zh.md
@@ -0,0 +1,28 @@
+---
+layout: research
+permalink: /zh/research/deeptb-negf/
+title: "DeePTB-NEGF：从学习到的哈密顿量计算量子输运"
+title_en: "DeePTB-NEGF: transport from learned Hamiltonians"
+description: "A DeePTB and DPNEGF workflow for quantum transport in graphene, h-BN and MoS2, compared with first-principles DFT-NEGF calculations."
+description_zh: "DeePTB 和 DPNEGF 结合，计算石墨烯、氮化硼和二硫化钼的量子输运，并与 DFT-NEGF 对照。"
+paper_year: 2025
+paper_url: "https://arxiv.org/abs/2512.11291"
+code_url: "https://github.com/deepmodeling/dpnegf"
+research_locale: zh
+translation_url: "/research/deeptb-negf/"
+translation_of: "/research/deeptb-negf/"
+alternate_en_url: "/research/deeptb-negf/"
+alternate_zh_url: "/zh/research/deeptb-negf/"
+---
+
+
A quantum transport calculation usually begins with a Hamiltonian. Constructing it directly from DFT can be reliable, but it becomes expensive when the system is large or when we want to compare many different structures. DeePTB can predict a tight-binding Hamiltonian from an atomic structure, which raises a natural question: can we use that Hamiltonian directly in a quantum transport solver?
diff --git a/_pages/research-ferh-magnetism-zh.md b/_pages/research-ferh-magnetism-zh.md
new file mode 100644
index 0000000..5484e62
--- /dev/null
+++ b/_pages/research-ferh-magnetism-zh.md
@@ -0,0 +1,25 @@
+---
+layout: research
+permalink: /zh/research/ferh-magnetism/
+title: "FeRh 中交换关联泛函对磁性的影响"
+title_en: "Exchange-correlation functionals and magnetism in FeRh"
+description: "A note on why density-functional predictions of magnetism in FeRh change with lattice constant and exchange-correlation functional."
+description_zh: "关于 FeRh 合金的磁性预测为什么依赖晶格常数和 DFT 交换关联泛函。"
+paper_year: 2025
+paper_url: "https://doi.org/10.1016/j.commatsci.2024.113561"
+research_locale: zh
+translation_url: "/research/ferh-magnetism/"
+translation_of: "/research/ferh-magnetism/"
+alternate_en_url: "/research/ferh-magnetism/"
+alternate_zh_url: "/zh/research/ferh-magnetism/"
+---
+
+
FeRh is an alloy of iron and rhodium whose magnetism is unusually sensitive to the lattice constant. A small change in the distance between atoms can change the relative energies of different magnetic states. That makes it a difficult system for DFT: how much of a prediction comes from the material, and how much comes from the approximation we chose?
diff --git a/_pages/research-logdet-entropy-zh.md b/_pages/research-logdet-entropy-zh.md
new file mode 100644
index 0000000..1b1d25f
--- /dev/null
+++ b/_pages/research-logdet-entropy-zh.md
@@ -0,0 +1,32 @@
+---
+layout: research
+permalink: /zh/research/logdet-entropy/
+title: "LogDet：用矩阵估计神经网络里的信息量"
+title_en: "LogDet: estimating information in neural networks"
+description: "The LogDet entropy estimator measures high-dimensional information using matrix operations and was applied to the analysis of neural networks."
+description_zh: "LogDet 熵估计器利用矩阵运算处理高维数据，用于分析神经网络特征中的信息变化。"
+paper_year: 2025
+paper_url: "https://doi.org/10.1016/j.neucom.2025.131520"
+paper_title: "Understanding neural networks with logarithm determinant entropy estimator"
+paper_doi: "10.1016/j.neucom.2025.131520"
+paper_authors:
+ - "Zhanghao Zhouyin"
+ - "Ding Liu"
+research_locale: zh
+translation_url: "/research/logdet-entropy/"
+translation_of: "/research/logdet-entropy/"
+alternate_en_url: "/research/logdet-entropy/"
+alternate_zh_url: "/zh/research/logdet-entropy/"
+---
+
+
As an undergraduate, I spent some time studying neural-network theory. I was interested in what happens to information inside a network: what survives as features pass through layers, and what gets discarded. Information theory gives us entropy and mutual information, but estimating them for high-dimensional neural-network features is difficult in practice.
diff --git a/_pages/research-moire-phonon-zh.md b/_pages/research-moire-phonon-zh.md
new file mode 100644
index 0000000..3f26b7e
--- /dev/null
+++ b/_pages/research-moire-phonon-zh.md
@@ -0,0 +1,34 @@
+---
+layout: research
+permalink: /zh/research/moire-phonon/
+title: "魔角石墨烯里的声子凝聚"
+title_en: "Moiré phonon condensation in twisted bilayer graphene"
+description: "A note on moiré phonon condensation, soft modes, and structural reconstruction in magic-angle twisted bilayer graphene."
+description_zh: "关于魔角双层石墨烯结构重构、软声子模式和莫尔声子凝聚的一些计算与想法。"
+paper_year: 2026
+paper_url: "https://arxiv.org/abs/2607.06711"
+paper_title: "Moiré Phonon Condensation in Magic-Angle Twisted Bilayer Graphene"
+paper_doi: "10.48550/arXiv.2607.06711"
+paper_authors:
+ - "Zhanghao Zhouyin"
+ - "Jyun-jie Jiang"
+ - "Xianghua Kong"
+ - "Hong Guo"
+research_locale: zh
+translation_url: "/research/moire-phonon/"
+translation_of: "/research/moire-phonon/"
+alternate_en_url: "/research/moire-phonon/"
+alternate_zh_url: "/zh/research/moire-phonon/"
+---
+
+
Put two sheets of graphene on top of each other and rotate one slightly. A moiré pattern appears, much larger than the original atomic lattice. As the twist angle gets smaller, the moiré period grows and the atoms rearrange. In a recent project, I wanted to understand what causes this reconstruction near the magic angle.
diff --git a/_pages/research-nqs-impurity-zh.md b/_pages/research-nqs-impurity-zh.md
new file mode 100644
index 0000000..c7bbcb9
--- /dev/null
+++ b/_pages/research-nqs-impurity-zh.md
@@ -0,0 +1,35 @@
+---
+layout: research
+permalink: /zh/research/nqs-impurity/
+title: "用神经网络求解量子嵌入里的杂质问题"
+title_en: "Neural quantum states for quantum embedding"
+description: "A research note on a neural-quantum-state impurity solver in ghost Gutzwiller quantum embedding and its sampling bottleneck."
+description_zh: "用神经量子态求解 ghost Gutzwiller 量子嵌入中的杂质模型，以及高精度采样带来的计算开销。"
+paper_year: 2026
+paper_url: "https://doi.org/10.1103/rkd8-q6yl"
+paper_title: "Neural-quantum-states impurity solver for quantum embedding problems"
+paper_doi: "10.1103/rkd8-q6yl"
+paper_authors:
+ - "Zhanghao Zhouyin"
+ - "Tsung-Han Lee"
+ - "Ao Chen"
+ - "Nicola Lanatà"
+ - "Hong Guo"
+research_locale: zh
+translation_url: "/research/nqs-impurity/"
+translation_of: "/research/nqs-impurity/"
+alternate_en_url: "/research/nqs-impurity/"
+alternate_zh_url: "/zh/research/nqs-impurity/"
+---
+
+
When electrons interact strongly, calculating the electronic state of a material becomes difficult. We could try solving for many electrons at once, but the number of possible quantum states grows rapidly with system size. Quantum embedding takes a different approach. It treats a smaller, difficult part as an impurity problem and repeatedly exchanges information between that problem and an effective environment.
diff --git a/_pages/research-slem-zh.md b/_pages/research-slem-zh.md
new file mode 100644
index 0000000..8e5352b
--- /dev/null
+++ b/_pages/research-slem-zh.md
@@ -0,0 +1,29 @@
+---
+layout: research
+permalink: /zh/research/slem/
+title: "SLEM：从原子结构预测量子算符"
+title_en: "SLEM: predicting quantum operators from atomic structures"
+description: "SLEM uses local, rotation-equivariant representations to predict quantum-operator matrices from atomic structures."
+description_zh: "SLEM 利用局域性和旋转等变性，从原子结构预测哈密顿量、重叠矩阵等量子算符。"
+paper_year: 2025
+paper_url: "https://arxiv.org/abs/2407.06053"
+research_locale: zh
+translation_url: "/research/slem/"
+translation_of: "/research/slem/"
+alternate_en_url: "/research/slem/"
+alternate_zh_url: "/zh/research/slem/"
+---
+
+
A band diagram is only part of what comes out of an electronic-structure calculation. The Hamiltonian matrix records orbital energies and couplings, the overlap matrix describes how the orbitals overlap, and the density matrix relates to where electrons are distributed. These are the quantities we often need for later property calculations.
diff --git a/_pages/research-vbcsr-zh.md b/_pages/research-vbcsr-zh.md
new file mode 100644
index 0000000..3367f1b
--- /dev/null
+++ b/_pages/research-vbcsr-zh.md
@@ -0,0 +1,50 @@
+---
+layout: research
+permalink: /zh/research/vbcsr/
+title: "VBCSR：给大规模稀疏矩阵写的一套工具"
+title_en: "VBCSR: distributed sparse matrices for scientific computing"
+description: "VBCSR is a distributed variable-block sparse matrix library for electronic-structure and scientific computing workloads, with MPI and thresholded matrix products."
+description_zh: "VBCSR 是一套支持 MPI 的变块稀疏矩阵工具，面向电子结构计算和其他科学计算任务。"
+project_kind: "software"
+code_url: "https://github.com/DeepElectron/vbcsr"
+paper_url: "https://arxiv.org/abs/2609.04397"
+paper_year: 2026
+research_locale: zh
+translation_url: "/research/vbcsr/"
+translation_of: "/research/vbcsr/"
+alternate_en_url: "/research/vbcsr/"
+alternate_zh_url: "/zh/research/vbcsr/"
+---
+
+
In electronic-structure calculations, I often have to work with very large matrices. Atomic orbitals mainly couple to nearby orbitals, so many matrix entries are zero. Storing and multiplying all those zeros would be wasteful. Sparse matrix formats avoid that, but distributing the calculation across many machines introduces other problems.
diff --git a/_pages/research-zh.html b/_pages/research-zh.html
new file mode 100644
index 0000000..89cb691
--- /dev/null
+++ b/_pages/research-zh.html
@@ -0,0 +1,112 @@
+---
+layout: raw
+permalink: /zh/research/
+title: "研究"
+description: "周寅张皓的研究项目与科研软件：机器学习电子结构、量子材料、量子输运和科学计算 Agent。"
+description_zh: "周寅张皓的量子材料、电子结构、量子输运和科学计算软件研究。"
+research_locale: zh
+translation_url: "/research/"
+alternate_en_url: "/research/"
+alternate_zh_url: "/zh/research/"
+---
+
+
+
+
+
+研究 | 周寅张皓
+
+{% include lamp-head.html %}
+
+
+
+
+
+ {% include lamp-nav.html active="research" %}
+
+
A neural network can read only a material’s local structure and predict its quantum operators directly — reaching near first-principles accuracy while staying local enough to scale toward whole devices.
Learning local equivariant representations for quantum operators (SLEM)
Neural quantum states can act as the impurity solver inside quantum embedding, matching exact diagonalization on correlated models — where the real bottleneck is how precisely you sample observables, not training the network.
Neural-Quantum-States Impurity Solver for Quantum Embedding Problems
Trained only on first-principles eigenvalues, a deep-learning tight-binding model predicts the electronic structure of structures it has never seen — making million-atom, finite-temperature electronic simulations feasible.
Deep learning tight-binding approach for large-scale electronic simulations (DeePTB)
Making the non-equilibrium Green’s function method differentiable end to end turns a quantum transport simulator into something you can optimize through — providing gradients for sensitivity analysis and inverse design.
Automatic differentiable nonequilibrium Green's function formalism (AD-NEGF)
Infrastructure for representing and computing with very large sparse operators in electronic-structure and quantum-transport work — the quiet plumbing that lets bigger simulations run on smaller machines.