1.Method
1. 方法
1.1 Model of MgO
1.1 MgO 的模型
Fig.S1. Models of pure MgO in (a) top view and (b) side view. (c) electronic differential density when CO2 adsorbed in Pure MgO with adsorption site O. (d) Band Structure when CO2 adsorbed in Pure MgO with adsorption site O.
无花果。S1.（a） 俯视图和 （b） 侧视图中的纯 MgO 模型。（c） 当 CO2 吸附床位于具有吸附位点 O 的纯 MgO 中时的电子差分密度 。（d） 当 CO2 吸附在具有吸附位点 O 的纯 MgO 中时的带结构。
As shown in Fig. S1a and S1b, the pure MgO is illustrated, where silver atoms represent Mg and blue atoms represent O. Since MgO is a hexahedral crystal, it has twelve potential adsorption sites. To identify the most favorable adsorption configuration, we recorded the adsorption energies and quantified electrical properties of various adsorption configurations after relaxation from different initial states. We referenced previous studies on Al/Ca/Fe doped adsorption structures [1] to confirm that the computational results align with existing trends, ensuring the reliability of the findings.
如图 1 所示 。 S1a 和 S1b， 说明了纯 MgO， 其中银原子代表 Mg，蓝色原子代表 O。 由于 MgO 是六面体晶体，因此它有 12 个潜在的吸附位点。为了确定最有利的吸附构型，我们记录了从不同初始状态弛豫后各种吸附构型的吸附能和量化的电学性质。我们参考了以前关于 Al/Ca/Fe 掺杂吸附结构的研究 [1]，以确认 计算结果与现有趋势一致，从而确保 研究结果的可靠性。
Subsequently, we computed the band structure and differential charge density before and after CO2 adsorption (as shown in Fig. S1c and S1d). Based on these results, machine learning methods are used to evaluate the changes in various quantifiable properties before and after adsorption. Consistent with previous experimental and theoretical studies, doping has been shown to be an effective method to modulate the CO2 adsorption capacity of MgO, significantly altering its crystal structure's electrical properties, such as valence electrons, orbital electron counts [2], HOMO/LUMO values [3], and other related attributes that greatly influence adsorption energy. Additionally, we explored various doping methods involving four atoms, but most structures experienced significant deformation as a result. After CO2 adsorption, in addition to studying the electrical properties, we also investigated the geometry of doped MgO, including bond lengths and distances between the dopant atoms and the remaining oxygen and magnesium atoms. The dopant atoms caused certain surface sites of MgO to bend, resulting in stronger CO2 adsorption at these top sites. Previous studies have also indicated that the bond angle of CO2 is closely related to the adsorption energy [4]. By building a dataset, machine learning can be better leveraged to assist in analyzing the effects of metal atom doping on CO2 adsorption properties on the surface of MgO.
随后，我们计算了 CO 2 吸附前后的能带结构和差分电荷密度 （如图 2 所示 ）。 S1c 和 S1d）。基于这些结果，使用机器学习方法来 评估吸附前后各种可量化特性的变化 。与以前的实验和理论研究一致，掺杂已被证明是调节 MgO 对 CO2 吸附能力的有效方法 ，显着改变其晶体结构的电学性质，如价电子、轨道电子数 [2]、HOMO/LUMO 值 [3]] 以及其他对吸附能有很大影响的相关属性。此外，我们还探索了涉及四个原子的各种掺杂方法，但大多数结构因此发生了明显的变形。 在 CO2 吸附后，除了研究电学性质外，我们还研究了掺杂 MgO 的几何形状 ，包括掺杂剂原子与剩余的氧和镁原子之间的键长和距离。掺杂原子导致 MgO 的某些表面位点弯曲，导致这些顶部位点的 CO2 吸附更强 。以前的研究还表明，CO2 的键角与吸附能密切相关 [4]。 通过构建数据集， 可以更好地利用机器学习来帮助分析金属原子掺杂对 MgO 表面 CO2 吸附特性的影响 。
All calculations in this study were based on density functional theory (DFT) and implemented using the CASTEP module within Material Studio. The exchange-correlation functional chosen was the Perdew-Burke-Ernzerhof (PBE) with generalized gradient approximation (GGA). Ultrasoft pseudopotentials were used to model the interactions between atomic nuclei and electrons. However, using the GGA-PBE method alone often underestimates binding energies due to its inaccurate description of van der Waals forces in layered materials [5]. Previous reports [6] have validated that the inclusion of the TS (GGA-PBE+TS) DFT-D method can correct this computational error.
本研究中的所有计算均基于密度泛函理论 （DFT），并使用 Material Studio 中的 CASTEP 模块实现。选择的交换相关函数是具有广义梯度近似 （GGA） 的 Perdew-Burke-Ernzerhof （PBE）。超软赝势用于模拟原子核和电子之间的相互作用。然而，单独使用 GGA-PBE 方法往往会低估结合能，因为它对层状材料中的范德华力的描述不准确 [5]。以前的报告[6]已经验证了 TS （GGA-PBE+TS） DFT-D 方法可以纠正这种计算错误。
In this project, all parameters used in the computational simulation software were tested. As shown in Fig. S2, preliminary convergence tests indicated that the energy stabilized at a cutoff energy of 500 eV, so a cutoff energy of 500 eV was selected for this study.
在这个项目中，对计算仿真软件中使用的所有参数进行了测试。如图 1 所示 。 S2 的初步收敛测试表明，能量稳定在 500 eV 的截止能量，因此本研究选择了 500 eV 的截止能量。
Fig.S2. Cut-off energy test.
无花果。S2.截止能量测试 。
In this study, the crystal model for MgO in each doping structure was set as follows: a = b = c = 4.2112 Å, and α = β = γ = 90°. Based on preliminary calculations and considering periodic boundary conditions and the size of CO2 molecules, a 5×5×3 slab model with periodicity along the x, y, and z axes was used, with a 20 Å vacuum layer added to minimize interactions between the slabs. After constructing the models, calculations were performed using the CASTEP module in Materials Studio. Most parameters in the calculations were set to their default values, with adjustments made to the following parameters: cutoff energy of 500 eV, k-point mesh of 3 × 3 × 1, SCF convergence threshold of < 10^-6 eV/atom, energy convergence threshold of < 10^-5 eV/atom, maximum displacement of < 0.001 Å, maximum force of < 0.03 eV/Å, and maximum stress of < 0.05 GPa.
在这项研究中，每种掺杂结构中 MgO 的晶体模型 设置如下：a = b = c = 4.2112 Å，α = β = γ = 90°。基于初步计算并考虑周期性边界条件和 CO2 分子的大小，使用了沿 x、y 和 z 轴具有周期性的 5×5×3 板模型，并添加了 20 Å 真空层以最大限度地减少板之间的相互作用。构建模型后，使用 Materials Studio 中的 CASTEP 模块进行计算。计算中的大多数参数都设置为默认值，并对以下参数进行了调整：截止能量为 500 eV，k 点网格为 3 × 3 × 1，SCF 收敛阈值为 < 10^-6 eV/原子，能量收敛阈值为 < 10^-5 eV/原子，最大位移为 < 0.001 Å，最大力为 < 0.03 eV/Å， 最大应力 < 0.05 GPa。
Since GGA-based DFT calculations cannot accurately describe van der Waals (vdW) interactions, this study uses the TS method for correction. Additionally, to identify the most stable adsorption configuration of the target gas, we carefully considered the initial positions and orientations of CO2 molecules on the MgO material.
由于基于 GGA 的 DFT 计算无法准确描述范德华 （vdW） 相互作用，因此本研究使用 TS 方法进行校正。此外，为了确定目标气体最稳定的吸附构型，我们仔细考虑了 CO2 分子在 MgO 材料上的初始位置和方向。
The adsorption energy (Ea) of CO2 on the MgO material is calculated using the following equation:
CO2 在 MgO 材料上的吸附能 （Ea） 使用以下公式计算：
Ea =ECO2+substrate – (Esubstrate+ ECO2)
Ea =ECO2+底物 –（E 底物 + ECO2）
ECO2+substrate, Esubstrate and ECO2 represent the total energy of the CO2 adsorption structure in its stable state, the energy of the substrate material, and the energy of the free CO2 molecule, respectively [7].
ECO2+底物、E 底物和 ECO2 分别代表 CO2 吸附结构在稳定状态下的总能量、底物材料的能量和游离 CO2 分子的能量 [7]。
1.2 Dataset Building
1.2 数据集构建
In the machine learning section, for the ECA (Element-Concentration-Adsorption sites) structure in this study, we categorized the electronic properties in the dataset into four types, totaling 68 properties:
在机器学习部分， 对于本研究中的 ECA（元素-浓度-吸附位点） 结构，我们将数据集中的电子属性分为四种类型， 总共 68 种属性：
1.Properties of the Dopant Atom in the Free State (such as atomic mass, atomic radius, formation enthalpy of oxides, electronegativity, first ionization energy, and number of valence electrons);
1. 自由态掺杂原子的性质 （如原子质量、原子半径、氧化物的形成焓、电负性、第一电离能和价电子数）;
2.Properties of the Dopant Substrate Before Adsorption (for example, the average net charge of oxygen atoms surrounding the dopant atom, the number of s-orbital electrons of the dopant atom, the average bond length between magnesium atoms surrounding the dopant atom, and the average bond order of oxygen atoms surrounding the dopant atom);
2. 吸附前掺杂剂衬底的性能 （例如，掺杂剂原子周围氧原子的平均净电荷、掺杂剂原子的 s 轨道电子数、掺杂剂原子周围镁原子之间的平均键长、掺杂剂原子周围氧原子的平均键序）;
3.Properties of the Dopant Substrate After Adsorption (including the properties mentioned above after adsorption and additional properties such as the C of CO2 in the structure after adsorption);
3. 吸附后掺杂剂基材的性能 （包括上述吸附后的性能和 吸附后结构中 CO2 的 C 等附加性能）;
4.System Before and After Adsorption Difference (i.e., comparisons of similar properties among the DAF, BADS, and AADS categories, such as the difference in s-orbital electrons between DAF and BADS states, or the differences between BADS and AADS states).
4. 系统吸附前后差异 （即 DAF、BADS 和 AADS 类别之间相似性质的比较，例如 DAF 和 BADS 状态之间 s 轨道电子的差异，或 BADS 和 AADS 状态之间的差异）。
It is worth mentioning that electronic differential density (EDD) is a spatially extensive property [8], making it difficult to quantify. The following Fig. S3 illustrates the data related to differential charge density. For example, Fig.S3a shows the case of EN-C1 structure. From the Fig. S3a, it can be observed that the EDD at the center of the N atom is positive which color is blue.
值得一提的是，电子差分密度 （EDD） 是一种空间广泛的特性 [8]，因此难以量化。下面的图 S3 说明了与差分电荷密度相关的数据。例如， 图 .S3a 显示了 E N-C1 结构的情况 。从图 S3a 中可以观察到 N 原子中心的 EDD 为正，其中颜色为蓝色。
Fig.S3. Quantitative value method of EDD special data of (a) two layers. and (b) CO2 adsorption site far away from doped atom.
无花果。S3.（a） 两层 s. 和 （b） 远离掺杂原子的 CO2 吸附位点的 EDD 特殊数据的定量值法 。
Considering that the gain or loss of electrons is interrelated, when outer-layer electrons are lost, the relative number of inner-layer electrons appears to increase. Conversely, if outer-layer electrons are gained, the inner-layer electrons relatively decrease. This is a relative trend: an increase in outer-layer electrons corresponds to a relative decrease in inner-layer electrons, as inner-layer electrons do not participate in the reaction. Therefore, when recording the EDD, it is important to focus on changes in the outer layer, so the red region in the Fig. S3a should be our data recorded.
考虑到电子的增益或损失是相互关联的，当外层电子丢失时，内层电子的相对数量似乎会增加。相反，如果获得外层电子，则内层电子相对减少。这是一个相对趋势：外层电子的增加对应于内层电子的相对减少，因为内层电子不参与反应。因此，在记录 EDD 时，关注外层的变化很重要，因此图 S3a 中的红色区域应该是我们的数据记录。
As shown in Fig. S3b, when there is a presence of both positive and negative charge density between the dopant atom and the CO2 molecule, we focus on collecting data in the region along the line connecting the atoms. According to previous studies, electronic charge tends to concentrate in the bonding region, which is characteristic of most covalent bonds, although it may not apply to all [9].
如图 1 所示 。S3b，当掺杂原子和 CO 2 分子之间同时存在正电荷密度和负电荷密度时 ，我们专注于沿连接原子的线收集区域的数据。根据以前的研究，电子电荷往往集中在键合区域，这是大多数共价键的特征，尽管它可能不适用于所有共价键 [9]。
2.Results and Discussion
2.结果和讨论
2.1 TDOS difference
2.1 TDOS 的差异
Fig. S4 shows the total Density of States (TDOS) for Pure MgO and PDOS of O and Mg.
无花果。S4 显示了纯 MgO 的总态密度 （TDOS） 以及 O 和 Mg 的 PDOS。
Fig.S4. TDOS of Pure MgO.
无花果。S4. 纯 MgO 的 TDOS。
As shown in the Fig. S4, the bottom of conduction band of the MgO substrate is composed of the O`s 2p orbitals.
如图 S4 所示 ，MgO 衬底的导带底部由 O 的 2p 轨道组成。
Fig. S5. PDOS of ELi-C1 and E Li/Na/K -C2.
无花果。 S5.E Li-C 1 和 ELi/Na/K -C 2 的 PDOS。
As shown in Fig. S5, for Li and Na, the main change occurs in the ELi-C2/3 structures, where the 2s orbitals of Li and Na do exhibit some increase, but this change is relatively small compared to the 2p orbitals of O and can be considered negligible. The primary difference between the two elements lies in the 2p orbitals. It is evident that the 2p orbital of K is positioned lower and shifted more to the left, indicating that K is more stable in the substrate compared to Na. Consequently, the adsorption strength of the K doped structure is weaker relative to that of the Na doped structure.
如图 1 所示 。S5，对于 Li 和 Na，主要变化发生在 ELi-C 2/3 结构 s，其中 Li 和 Na 的 2s 轨道确实表现出一些增加，但与 O 的 2p 轨道相比，这种变化相对较小 ，可以认为可以忽略不计。两种元素之间的主要区别在于 2p 轨道。很明显，K 的 2p 轨道位于较低位置并更多地向左移动，这表明与 Na 相比，K 在衬底中更稳定。因此，K 掺杂结构的吸附强度相对于 Na 掺杂结构的吸附强度较弱。
Fig. S6. TDOS of E Li/Na/Al/P -C2.
无花果。 S6.E Li/Na/Al/P -C 的 TDOS 2”
As shown in Fig. S6, for P and Al, comparing with pure MgO, we can clearly observe that their TDOS peaks shift to lower potential positions, indicating that the active electron states become more stable. Similarly, N and Si also exhibit this behavior. The adsorption energy levels of structures doped with P, N, Al, and Si are significantly superior compared to other structures. The decrease in electronic energy levels, i.e., the leftward shift of the DOS, is a contributing factor to the change in adsorption energy.
如图 1 所示 。S6，对于 P 和 Al，与纯 MgO 相比 ，我们可以清楚地观察到它们的 TDOS 峰向较低的电位位置移动，表明活性电子态变得更加稳定。同样，N 和 Si 也表现出这种行为。与其他结构相比，掺杂 P、N、Al 和 Si 的结构的吸附能级明显更高。电子能级的降低，即 DOS 的左移，是导致吸附能变化的一个因素。
Among them, the 2p orbitals of N contribute significantly to the top of valence band, with N’s 2p contribution reaching 2/3. For Si, as the doping concentration increases, the top of valence band is consistently composed of 25% from Si's 3s and 3p orbitals and 75% from Mg's 2s and 2p orbitals.
其中，N 的 2p 轨道对价带顶部有显著贡献，N 的 2p 贡献达到 2/3。对于 Si，随着掺杂浓度的增加，价带顶部始终由 Si 的 3s 和 3p 轨道的 25% 和 Mg 的 2s 和 2p 轨道的 75% 组成。
After a comprehensive analysis, as shown in Fig. S7, we find that, among all structures, Li, Na, K, Ca, and Be Doped structures invariably have their Fermi levels close to the bottom of the conduction band.
经过综合分析，如图 2 所示 。S7 中，我们发现，在所有结构中，Li、Na、K、Ca 和 Be 掺杂结构的费米能级总是接近导带的底部。
From Fig. 13 in the manuscript, it is evident that only Al doped structure has its Fermi level close to the top of the valence band; Fe and Co doped structure has their valence bands crossing the Fermi level; Si, Ge, N, P, S, Cl and Ga doped structure has their Fermi levels situated in the middle of the bandgap, with various small peaks present. This result indicates that there are differences both between different groups and among different periods within the same group.
从图 13 在手稿中 ，很明显只有 Al 掺杂结构的费米能级接近价带的顶部;Fe 和 Co 掺杂结构的价带穿过费米能级;Si、Ge、N、P、S、Cl 和 Ga 掺杂结构的费米能级位于带隙的中间，存在各种小峰。此结果表明，不同组之间以及同一组内的不同期间之间存在差异。
However, from the TDOS plots, Li, Na, K, Ca and Be doped structure can be considered similar in terms of DOS behavior, with the only notable difference being that the conduction band peak for Be is reduced by 9% compared to the others.
然而，从 TDOS 图中， 可以认为 Li、Na、K、Ca 和 Be 掺杂结构在 DOS 行为方面相似，唯一显着的区别是 Be 的导带峰值与其他结构相比减少了 9%。
Fig. S7. TDOS of E Li/Na/K/Ca/Be -C1.
无花果。 S7.E Li/Na/K/Ca/Be -C 的 TDOS 1.
Regarding the adsorption sites of Li, Na, K, Ca, and Be doped structure, it is observed that the adsorption energy increases significantly when adsorbed at the oxygen site. However, the overall TDOS does not show significant changes, except for a notable decrease in Be. Nonetheless, from the comprehensive adsorption energy table, it is evident that Be exhibits significantly higher adsorption energy levels compared to Li, Na, K, and Ca doped structure.
关于 Li、Na、K、Ca 和 Be 掺杂结构的吸附位点 ，观察到吸附在氧位点时吸附能显着增加。但是，除了 Be 的显著下降外，整体 TDOS 没有显示出显着变化。尽管如此，从综合吸附能表中可以明显看出，与 Li、Na、K 和 Ca 掺杂结构相比，Be 表现出明显更高的吸附能级 。
Fig. S8. TDOS of E Li/Na/K/Ca/Be -C1-AO.
无花果。 S8. E Li/Na/K/Ca/Be -C 1-AO 的 TDOS。
We compared additional adsorption sites, as shown in Fig. S9. Due to limitations on the length of a document, only the adsorption sites of Li and Be doped structure at Doped atom/Mg/O are displayed. However, overall, it can be observed from Fig. S7 and S8 that the performance of Li, Na, K, and Ca doped structure is very similar.
我们比较了其他吸附位点，如图 1 所示 。S9 中。由于文档长度的限制，仅显示 Li 和 Be 掺杂结构在 Doped atom/Mg/O 处的吸附位点。然而，总的来说，可以从图 1 中观察到。S7 和 S8 认为 Li、Na、K 和 Ca 掺杂结构的性能非常相似。
Fig. S9. TDOS of E Li/ Be -C1-AEC/Mg/O.
无花果。 S9 中。E Li/ Be -C1-A EC/Mg/O 的 TDOS。
By comparing the TDOS plots of Be and Li doped structure at various adsorption sites, it is evident that their differences are quite similar at equivalent sites. Therefore, we need to use the PDOS for a more detailed comparison.
通过比较 Be 和 Li 掺杂结构在不同吸附位点的 TDOS 图，可以明显看出它们在等效位点的差异非常相似。因此，我们需要使用 PDOS 进行更详细的比较。
2.2 PDOS difference
2.2 PDOS 的区别
Fig. S10 TDOS and PDOS of E Li/ Be/Na -C1-AO and E Li -C1-AEC.
无花果。 E Li/Be/Na -C 1-AO 和 ELi-C 1-A EC 的 S10 TDOS 和 PDOS。
As shown in Fig. S10, the PDOS plots for Be, Li, and Na doped structure at the oxygen site reveal that the conduction band peak in the TDOS is predominantly composed of O's 2p orbitals. This indicates that the O atom is the primary reason for the reduction and shift of the conduction band peaks in the TDOS of various structures. From the manuscript section 3.1, ANC-O-BA and ABL-O-d were found to have moderate correlations with the group containing Li, Na and Be. We can infer that the left shift of the PDOS peak represents the core of electron transfer, while ANC-O-BA and ABL-O-d reflect the external manifestations.
如图 1 所示 。S10，氧位点 Be、Li 和 Na 掺杂结构的 PDOS 图表明，TDOS 中的导带峰主要由 O 的 2p 轨道组成。这表明 O 原子是各种结构的 TDOS 中导带峰减少和偏移的主要原因。从手稿第 3.1 节中 ，发现 ANC-O-BA 和 ABL-O-d 与包含 Li、Na 和 Be 的基团具有中等相关性 。我们可以推断，PDOS 峰的左移代表电子转移的核心，而 ANC-O-BA 和 ABL-O-d 反映了外部表现。
Additionally, it is observed that, regardless of the dopant, when the adsorption energy increases, the PDOS peak of C (CO2) significantly diminishes. This seems to be related to the strong correlation we found between the NC-C attribute and adsorption energy in our previous machine learning analysis in the manuscript section 2 fig.4. To confirm this finding, Fig. S11 shows a comparison of the PDOS peak of C (CO2) for Li, Na, and Be doped structures at various adsorption sites, as the sensitivity of the adsorption energy to the NC-C attribute was found to be almost identical for alkali and alkaline earth metals in the earlier machine learning analysis which in Table S3 in this Supporting Information below.
此外， 据观察，无论使用哪种掺杂剂，当吸附能增加时，C （CO2） 的 PDOS 峰显着降低。这似乎与我们在之前的机器学习分析中发现的 NC-C 属性和吸附能之间的强相关性有关 ， 请参见手稿第 2 节图 4。为了证实这一发现， 图 1S11 显示了 Li、Na 和 Be 掺杂结构在不同吸附位点的 C （CO2） 的 PDOS 峰的比较 ，因为在早期的机器学习分析中发现碱和碱土金属的吸附能对 NC-C 属性的敏感性几乎相同 ，如下面的支持信息表 S3 所示 。
![]()
Fig.S11 PDOS of C (CO2) for E Li/Na/Be-C1-AEC/Mg/O.
无花果。E Li/Na/Be-C 1-AEC/Mg/O 的 C （CO2） 的 S11 PDOS。
It can be observed that, for the EBe -C1-AO structure, the PDOS peak of C (CO2) exhibits a deeper reduction compared to other similar structures, and the corresponding adsorption energy is significantly higher. In comparison, peak of C in Li doped structure is slightly lower than Na doped structure, and the corresponding adsorption energy of Li doped structure is still slightly higher than that of Na doped structure.
可以观察到，对于 EBe -C1-A O 结构， 与其他类似结构相比，C （CO2） 的 PDOS 峰表现出更深的还原，相应的吸附能明显更高。相比之下， Li 掺杂结构中的 C 峰略低于 Na 掺杂结构 ， 而 Li 掺杂结构的相应吸附能仍略高于 Na 掺杂结构 。
2.3 PDOS of Mg and C Difference
2.3 Mg 和 C 的 PDOS 差异
in the manuscript section 3.1. Transition metals have high correlations with both NC-C and NE-s-dA. For Transition metal doped structures EFe/Pt/Ir-C3-AO and EFe-C3-AEC, we examined the PDOS of C (CO2). Since transition metals also show high correlation with NC-C, the PDOS peak of C (CO2) for these structures are shown in Fig. S12. Surprisingly, the EFe-C3-AEC `s PDOS peak of C (CO2) are identical, and the EPt/Ir-C3-AO graphs are similar as well. This suggests that the adsorption relationships involving CO2 may correspond to two different patterns.
在手稿第 3 节中。1.过渡金属与 NC-C 和 NE-s-dA 都具有高度相关性 。 对于过渡金属掺杂结构 E Fe/Pt/Ir-C 3-A O 和 EFe-C 3-A EC， 我们研究了 C （CO2） 的 PDOS。由于过渡金属也显示出与 NC-C 的高度相关性， 因此这些结构的 C （CO2） 的 PDOS 峰如图 2 所示。S12.令人惊讶的是，E Fe-C 3-A EC 的 PDOS 峰 C （CO2） 相同，EPt/Ir-C 3-A O 图也相似。这表明涉及 CO2 的吸附关系可能对应于两种不同的模式。
Fig.S12 PDOS of C (CO2) for EFe/Pt/Ir -C3-AO and EFe-C3-AEC.
无花果。E Fe/Pt/Ir-C 3-A O 和 EFe-C 3-A EC 的 C （CO2） 的 S12 PDOS。
For Al and Ga, these semi-transition metals doped Substrate show high correlations with the attributes NC-C, ABL-Mg-BA, and NE-d-sA. Consequently, we investigated the PDOS of both the dopant atoms in each adsorption structure and Mg in each structure before adsorption. Fig. S13 below shows the PDOS of Mg in each structure prior to adsorption.
对于 Al 和 Ga，这些半过渡金属掺杂的 Substrate 与 NC-C、ABL-Mg-BA 和 NE-d-sA 属性表现出高度相关性 。因此，我们研究了吸附前每个吸附结构中掺杂原子的 PDOS 和每个结构中 Mg 的 PDOS。 无花果。 下面的 S13 显示了吸附前每种结构中 Mg 的 PDOS。
Fig.S13 PDOS of Mg for EGa -C1 and EAl-C1/2/3 structures.
图 S13 EGa -C1 和 EAl-C1/2/3 结构的 Mg PDOS。
For the classification of the adsorption energy of EGa -C1 and EAl-C1/2/3 structures, the adsorption energy levels for these structures are related to the extent to which the Fermi level crosses the valence band and the height of the peaks shown in the Fig. S13. Furthermore, since the NE-d-sA attribute also shows a high correlation with adsorption energy, we compared the Al and Ga atoms in each adsorption structure. Similar to the analysis of transition metal atoms, Al and Ga are major contributors of d orbitals in the entire structure. However, surprisingly, d orbitals were found to be involved in almost none of the structures. Upon examining the dataset, it is discovered that this issue was due to a machine learning artifact, which incorrectly identified this attribute as having high relevance.
对于 EGa -C1 和 EAl-C 1/2/3 结构 s 的吸附能分类 ， 这些结构 s 的吸附能级与费米能级穿过价带的程度和图 S13 中所示的峰的高度有关。 此外，由于 NE-d-sA 属性也显示出与吸附能的高度相关性，我们比较了每种吸附结构中的 Al 和 Ga 原子 。与过渡金属原子的分析类似，Al 和 Ga 是整个结构中 d 轨道的主要 c 贡献者 。然而，令人惊讶的是，发现 d 轨道几乎没有参与任何结构。在检查数据集时， 发现此问题是由于机器学习工件造成的，该工件错误地将此属性标识为具有高相关性。
3. DOS plot
3. DOS 图
3.1 DOS plot of structures without CO2 adsorption
3.1 无 CO2 吸附的结构的 DOS 图
It should be noted that this chapter designs a large number of pictures, so they are all in the form of combined pictures. The upper left corner of each single picture is the Label declares structure and object corresponding to the DOS diagram. In order to make each image more recognizable, the label naming method is "doped [amount] [element]-[adsorption site]-[PDOS object] ", when the plot is structures without CO2 adsorbed, it will be named like "doped [amount] [element]-[PDOS object] "
需要注意的是，本章设计了大量的图片，所以都是以组合图片的形式。 每张图片的左上角是 Label declares structure 和 object 对应的 DOS 图。 为了使每张图像更具辨识度， 标签命名方式为“掺杂 [量] [元素]-[ 吸附位点]-[PDOS 对象] ”，当小区是没有 CO2 吸附的结构时， 会命名为 “掺杂 [量] [元素]-[PDOS 对象] ”
Fig. S14-1-1. PDOS of all Mg in E-C1 structures without CO2 adsorbed.
无花果。 S14-1-1 中。E-C 1 结构中所有 Mg 的 PDOS， 未 吸附 CO2。
Fig. S14-1-2. PDOS of all O in E-C1 structures without CO2 adsorbed
无花果。 S14-1-2 中。E-C 1 结构中所有 O 的 PDOS， 无 CO2 吸附
Fig. S14-1-3. PDOS of Doped Atom in E-C1 structures without CO2 adsorbed
无花果。 S14-1-3 中。E-C 1 结构中掺杂原子的 PDOS， 无 CO2 吸附
Fig. S14-1-4. TDOS of E-C1 structures without CO2 adsorbed
无花果。 S14-1-4 中。E-C1 结构的 TDOS， 无 CO2 吸附
Fig. S14-2-1. PDOS of all Mg in E-C2 structures without CO2 adsorbed
无花果。 S14-2-1 中。E-C 2 结构中所有 Mg 的 PDOS， 无 CO2 吸附
Fig. S14-2-2. PDOS of all O in E-C2 structures without CO2 adsorbed
无花果。 S14-2-2 中。E-C 2 结构中所有 O 的 PDOS， 无 CO2 吸附
Fig. S14-2-3. PDOS of Doped Atom in E-C2 structures without CO2 adsorbed
无花果。 S14-2-3 中。E-C 2 结构中掺杂原子的 PDOS， 无 CO2 吸附
Fig. S14-2-4. TDOS of E-C2 structures without CO2 adsorbed
无花果。 S14-2-4 中。E-C2 结构的 TDOS， 无 CO2 吸附
Fig. S14-3-1. PDOS of all Mg in E-C3 structures without CO2 adsorbed
无花果。 S14-3-1 中。E-C 3 结构中所有 Mg 的 PDOS， 无 CO2 吸附
Fig. S14-3-2. PDOS of all O in E-C3 structures without CO2 adsorbed
无花果。 S14-3-2 中。E-C 3 结构中所有 O 的 PDOS， 无 CO2 吸附
Fig. S14-3-3. PDOS of Doped Atom in E-C3 structures without CO2 adsorbed
无花果。 S14-3-3 的。E-C 3 结构中掺杂原子的 PDOS， 无 CO2 吸附
Fig. S14-3-4. TDOS of E-C3 structures without CO2 adsorbed
无花果。 S14-3-4 中。E-C3 结构的 TDOS， 无 CO2 吸附
问题反馈