Introduction

Homochirality, a signature of life, refers to the dominance of one enantiomer of a pair across the entire biosphere1,2. In chemistry, an enantiomer containing an asymmetric carbon atom cannot be superimposed on its own mirror image. When comparing their analogous physical and chemical properties, it is crucial to note that only one of the enantiomers exhibits bioactivity, while the other may be inactive or even toxic3. Therefore, the creation of chiral compounds to achieve chiral recognition and separation is of paramount importance in the fields of chemistry, agriculture, food industry, and particularly in pharmaceutics.

Metals hold great promise as materials for chiral sensing, achieved through the breaking of bulk metal crystals symmetry4,5,6,7, chiral molecular imprinting8,9,10,11, and functionalization with chiral molecules12,13. The utilization of high-Miller-index surface (h ≠ k ≠ l ≠ h and h, k, l ≠ 0) of face-centered cubic metals and chiral-encoded mesoporous metals, leveraging strong substrate-molecule interactions, imparts a high selectivity. However, these defective surfaces often suffer from atomic roughness6,14,15 due to entropy gain6, high surface energy16, and low coordination of the kink atoms15. For example, metal films can lose their chirality within 30 min under operating conditions17,18. A significant breakthrough was reported, in which single-crystal copper foils with high-index facets were synthesized through seeded growth16. Nevertheless, the controlled preparation of chiral high-Miller-index facets remains challenging due to the thermodynamic and kinetic instability of kink steps. To address the thermal stability of high-Miller-index surfaces, an intermetallic compound of PdGa(111) has emerged for enantioselective catalysis and adsorption, where chirality is induced by a helix structure beneath the flat surface19,20. Combining experimental and theoretical approaches, it has been demonstrated that the natural termination of bulk chirality through dispersion interactions, significantly enhances chiral recognition capabilities. These findings underscore the potential of natural chiral termination as a platform for chiral recognition and separation.

Chiral nanostructures made of ceramics, including metal oxides, nitrides, carbides, and silicates, have also found applications in asymmetric catalysis, chiral recognition, and separation21,22. The first observation of homochirality resulted from the asymmetric adsorption of amino acids on quartz, yielding an enantiomeric excess (e.e.) ranging from 1.0% to 1.8%23. More significant enantioselective adsorption was observed for aspartic acids on calcite crystal faces, with a maximum e.e. value of 10%24. In addition to natural minerals, a controllable method has been developed for synthesizing chiral CuO films25. Recently, multicomponent ceramics, including high- and medium-entropy ceramics, have been proposed to enhance material functionality through diverse compositions, versatile structures, and moderate band gaps26,27,28. Medium-entropy ceramic (MEC) typically exhibits a mixing configurational entropy of 1–1.5 kB per cation, where kB represents the Boltzmann constant29. Among various MECs, (CrMoTa)Si2 with a C40 hexagonal crystal structure has been synthesized28, meeting the criteria for asymmetry and high thermal stability.

In this study, we propose the use of MECs for investigating chiral recognition capability through selective adsorption of serine, a prototype chiral molecule that has been widely studied. The flat, disordered (CrMoTa)Si2(0001) surface, lacking step and kink atoms, exhibits higher thermal stability compared to high-Miller-index Cu facets, such as Cu(531)/(643)/(874). This enhanced stability is due to the higher coordination number, which stabilizes the surface configuration by preventing atomic ejection from the surfaces. To evaluate the enantioselectivity of chiral molecules on MECs, we extend the universal isotherm adsorption model30, which captures adsorption uptake and energy distribution on porous heterogeneous surfaces, to multicomponent ceramics. Through high-throughput calculations of serine adsorbed on (CrMoTa)Si2(0001) (Supplementary Fig. 1), we observe two peaks in the adsorption energy distribution corresponding to the μ3’ and μ4 configurations. By considering the competitive adsorption of enantiomers on a given site, we plot a revised adsorption energy distribution of L- and D-enantiomers co-adsorbed on (CrMoTa)Si2(0001), achieving a promising 42% e.e. value. This enantioselectivity is attributed to the significant difference in charge transfer between L-serine and the substrate, as compared to that of D-serine.

The MEC films were synthesized using non-reactive magnetron sputtering technology. A Cr transition layer of 7.45 nm was pre-sputtered to enhance the binding strength of the interface, and a 92.8 nm thick MEC film was sputtered to maintain the bulk chirality. Grazing incidence X-ray diffraction (GIXRD) revealed a phase transition from multiphase to single-phase with a C40 hexagonal crystal structure. Energy dispersive X-ray spectroscopy (EDS) verified a uniform distribution of Cr/Mo/Ta/Si. The chiral nature of the MECs was confirmed by the combination of transmission electron microscopy (TEM) and circular dichroism (CD) spectra. Furthermore, the enantioselectivity was verified using quartz crystal microbalance (QCM) systems when the MEC film was deposited on the electrode. A linear relationship is observed between the adsorption amount and the analyte concentration with a high α value of 1.58 (indicating the ratio of adsorption capacity for L- to that for D-serine). These findings highlight the potential of MECs as a chiral stationary phase for chiral recognition and separation.

Results

Structure and preparation of medium-entropy ceramics

To enhance the stability of chiral stationary phases and minimize the atomic roughness of the surface atoms, particularly for kink atoms, materials with a natural termination of bulk chirality are highly promising due to their high thermal and chemical stability, as well as their high density of chiral binding sites19. In this study, we have selected a MEC, (CrMoTa)Si2, as the chiral stationary phase. The C40 hexagonal lattice structure of (CrMoTa)Si2 is depicted in Fig. 1a. Similar to the helix structure beneath PdGa(111)31,32,33, the natural layer arrangement in (CrMoTa)Si2 along the (0001) direction follows an ABC stacking pattern, with the lower layer rotated by 120° compared to the upper layer (Supplementary Fig. 2). This helix structure results in three adsorption geometries for serine (Fig. 1b). The lattice constants of (CrMoTa)Si2 are 6.461/4.588 for c/a, which align well with previous findings (Table 1 and Supplementary Fig. 3). The difference between the calculated and measured (empirical) lattice parameters is 0.65% (0.31%) for a and 0.95% (0.54%) for c. The well-matched lattice parameters obtained from calculations, measurements, and empirical methods indicate that the generated MECs adhere to the ideal rule of mixing with a solid solution structure. Furthermore, the average formation energy of − 1.24 eV and the entropy formation ability of 151 (eV/atom)−1 support the preference for the formation of a single-phase (CrMoTa)Si2. For more detailed information, please refer to the Supplementary Information and Supplementary Fig. 4.

Fig. 1: Structure of CrSi2-type ceramics and schematic diagram of enantiomeric adsorption.
figure 1

a CrSi2 with a C40 hexagonal crystal structure and the slab of MECs, (CrMoTa)Si2(0001), with a random arrangement of metal elements. On the slab of MECs, there is no symmetry due to the random arrangement of metal elements. b Possible adsorption orientations of L-serine on CrSi2(0001). There are three possible orientations (0°, 60°, and 120°) due to the rotation diad (C2) of CrSi2(0001). c Schematic diagram of the enantioselective adsorption on CrSi2(0001). Negligible enantioselectivity is foreseen resulting from the rotation diad (C2) along the (0001) direction. d Schematic diagram of enantioselective adsorption on (CrMoTa)Si2(0001). The rotation diad (C2) constituted by metallic atoms is broken by introducing the multiple elements in the helical site. The asymmetric surface configuration confines the enantiomers to locate at the distinct sites, enhancing the asymmetric effect from the helix structure.

Full size image
Table 1 Lattice constant and axial ratio (c/a) of pure and medium-entropy ceramics
Full size table

Next, we examine the structural evolution capability of MEC films. The local configuration evolution of the surface typically encompasses three steps: (i) atomic ejection from the least stable sites, (ii) atomic diffusion on the terrace, and (iii) atomic aggregation to form sub-nanometer clusters (Supplementary Fig. 5a)34. Among these three steps, the first step requires energy to break the bonds between the ejected atom and the substrate. The adatom formation energy serves as a descriptor to assess the ease of atomic ejection in this initial step (calculation details are provided in the Methods section)35. Taking Cu surfaces as an example, the adatom formation energy is strongly dependent on the coordination number (CN) of the ejected atom (Supplementary Fig. 5b). High-index chiral surfaces exhibit comparable and low adatom formation energies, suggesting a reduced level of thermal stability when compared to low-index surfaces. Nevertheless, in ceramics, the adatom formation energy compasses the breakdown of chemical bonds between metallic and nonmetallic atoms, making it not solely dependent on the CN (Supplementary Fig. 5b). The adatom formation energy of MEC films is 1.13 eV, which exceeds that of high-index metal surfaces by 0.23 eV (Supplementary Fig. 5b and Supplementary Tables 1−3). These findings imply that initiating structural evolution in MECs is more difficult compared to high-index metal surfaces. Another critical factor determining the ability of structural evolution is atomic diffusion. The energy barrier of transition states during the atomic diffusion process was evaluated using the climbing image nudged elastic band (CI-NEB) method. The reaction pathway is shown in Supplementary Fig. 6. A significant amount of energy is required to overcome the breaking of chemical bonds in ceramics, especially when dragging the ejected atom out of the surface. The energy barrier of transition states on the MEC film is 1.34 eV, which is twice as high as that on Cu(874) (Supplementary Fig. 5c). This indicates that, according to the Arrhenius equation, the probability of atomic diffusion on the MEC film is two orders of magnitude lower than on high-index metal surfaces. Atomic aggregation is an exothermic process due to the formation of bonds between adatoms, and it does not hinder structural evolution. Taking into account atomic ejection, diffusion, and aggregation, it is evident that MEC films exhibit higher stability compared to high-index metal surfaces.

We synthesized the (CrMoTa)Si2 film using the non-reactive magnetron sputtering method. EDS analysis confirms the uniform distribution of elements in the MEC film, as depicted in Fig. 2a, c–f and detailed in Supplementary Table 4. The high-angle annular dark-filed transmission electron microscopy (HAADF-TEM) image reveals lattice fringe spacings of 0.231 and 0.227 nm, corresponding to the (\(5\bar{4}\bar{1}0\)) and (\(11\bar{2}0\)) planes, respectively (Fig. 2b). The HAADF-TEM and aberration-corrected high-angle annular dark-filed scanning transmission electron microscopy (AC-HAADF-STEM) images at various resolutions and orientations indicate lattice constants of c and a to be 0.675 and 0.458 nm, respectively, confirming the nature of MECs (Supplementary Figs. 7, 8). The structural properties of the target and film were characterized by X-ray diffraction (XRD) and GIXRD, revealing a distinct phase transition from multiphase to single-phase with a C40 hexagonal lattice structure (Fig. 2g). The calculated PDF card of MECs is shown in Supplementary Table 5. The similarity of diffraction angles (2θ) in different incident angles suggests the phase of MEC films at different depths is uniform. The film thickness, determined by atomic force microscopy (AFM), is found to be 92.8 nm for the MEC film, with an additional 7.45 nm for the transition layer (Fig. 2h–j). The surface roughness under different sputtering conditions indicates a flat and smooth coating on the QCM electrode (Supplementary Table 6). A distinct ultraviolet-visible (UV-vis) absorption peak is observed around 300 nm through UV-vis diffuse-reflectance spectroscopy (Supplementary Fig. 9). Based on the UV-vis absorption wavelength, CD spectra were conducted in range of 190–400 nm, showing a positive Cotton effect for four matters, meaning an R configuration (Fig. 2k). The CD spectra of CrSi2, MoSi2, TaSi2, and the MECs exhibit distinct absorption features, confirming the synthesis of chiral MECs analogous to C40 MoSi2 and TaSi2. Surface observations obtained through these characterization techniques confirm the single-phase nature and uniform element distribution in the MEC film. These results demonstrate the robustness of the magnetron sputtering method for fabricating MEC films. This fabrication approach provides an alternative method for producing medium/high entropy silicide films, complementing techniques such as spark plasma sintering27,36 and electron beam evaporation28.

Fig. 2: Preparation and characterization of MEC films.
figure 2

a TEM image of the MEC films, cut by a focused ion beam at 30 kV. b HAADF-STEM image of MECs, with the inset showing the corresponding diffraction spot. c–f Elemental mapping of Cr, Mo, Ta, and Si in the MEC films by EDS, respectively. g XRD pattern of the composite silicide target and GIXRD patterns at different grazing incidence angles for MECs deposited via non-reactive magnetron sputtering. h, i AFM image and topography profile of the MEC films and the Cr transition layer. j Thickness of MEC films and Cr transition layer. k CD spectra of CrSi2, MoSi2, TaSi2, and MECs. Source data are provided as a Source Data file.

Full size image

Enantioselective adsorption on medium-entropy ceramics

The adsorption energy of enantiomers can be considered a crucial descriptor for characterizing their adsorption behaviors and tendencies. Serine was chosen as the model analyte because it is one of the smallest polar amino acids and exhibits strong molecule-substrate interactions due to its side chain. The possibility of different orientations is an intrinsic property of the crystalline structure, which is related to surface energies16. Among the eight possible crystalline orientations, the (0001) surface is the most stable and is chosen as the representative surface to investigate enantioselectivity (Supplementary Table 7). The possible adsorption geometries and corresponding adsorption energies of serine on CrSi2/MoSi2/TaSi2(0001) are employed to investigate the adsorption behaviors on CrSi2-type ceramics (Supplementary Fig. 10 and Supplementary Tables 8, 9). There are three stable adsorption sites (Site0°, Site60°, and Site120°) due to the rotation diad (C2) of CrSi2(0001), where L-serine prefers to locate on Site0° and D-serine prefers Site60°. The adsorption energies of serine follow the order: CrSi2 < MoSi2 < TaSi2, while the adsorption energy difference between L- and D-serine remains small (< 0.03 eV) on these ceramic surfaces. The small adsorption energy difference originates from the symmetry (Fig. 1c), while there is significant adsorption selectivity at a given site (such as Site0° and Site60°). Introducing the medium-entropy concept could be an efficient way to enhance the adsorption selectivity by breaking the symmetry and confining the adsorption geometries (Fig. 1d).

The chiral recognition of the MECs was investigated through high-throughput calculations based on DFT. Inspired by the surrogate model37, the MEC surfaces with intricate local chemical structures are decomposed into extensive small structures. To ensure representative sampling of the MECs surfaces and avoid randomness, a large number of nearly equiprobable and sufficient samples are required to fully analyze adsorption behavior (Fig. 3a, b). In this study, we selected a (2 × 2) unit cell of (CrMoTa)Si2(0001) with a coverage of 0.25 ML, consisting of 8 Si atoms and 4 metallic atoms in the first layer. A total of 81 surfaces can be generated, covering all possible atomic configurations in the first layer, considering the three possibilities on each of the four metallic sites (34). For each surface configuration, the random seed method was employed to generate 10 distinct 4-layer-cell structures, resulting in a total of 810 configurations.

Fig. 3: Enantioselective adsorption of serine enantiomers on the (CrMoTa)Si2(0001) surface.
figure 3

a, b Schematic diagrams illustrating the high and low energy adsorption configuration (μ4 and μ3’) for L- and D-serine. The subscript denotes the number of bonds between enantiomers and MECs. c, d Adsorption energy distribution of different arrangements of surface atoms. The inset in (d) illustrates the defined arrangement of surface atoms, where the red and blue triangles indicate the location of adsorbed molecules for L- and D-serine, respectively. e, f Total adsorption energy distribution function X and the corresponding adsorption uptake θt as a function of adsorption energy. Two distinct peaks are observed, corresponding to the μ4 and μ3’ adsorption configurations. Source data are provided as a Source Data file.

Full size image

High-throughput calculations were conducted to determine the stable adsorption geometries and energies of serine on the 810 configurations of MEC films. To examine these adsorption geometries, a μ4 adsorption configuration with high adsorption energy is visualized in Fig. 3a, b, while the major energy peak corresponds to the μ3’ adsorption configuration. Analysis of the adsorption geometries reveals that the high energy site (μ4) is characterized by the presence of two or more Ta atoms in the first layer. In comparison to the continuous adsorption energies of intermediates on high-entropy alloys, the adsorption energies of L-serine on MECs exhibit a discontinuous energy distribution, as shown in Fig. 3c, d. The two distinct adsorption energy peaks correspond to two adsorption geometries (μ4 and μ3’). The statistical results of adsorption heights show that the bond length of the deprotonated hydroxyl group with MECs(0001) is 1.83 Å (1.83 Å) for D- (L-) serine, respectively, while the bond length of the amidogen group with MECs(0001) is 2.23 Å (2.34 Å) for D- (L-) serine (Supplementary Fig. 11a). In addition, the Bader charge analysis indicates that the N and O gain an average of 1.174 e (1.188 e) and 1.233 e (1.228 e) for D- (L-) serine adsorbed on MECs(0001) (Supplementary Fig. 11b). All these findings suggest the occurrence of chemisorption of serine on the MEC surfaces.

To model the uptake of serine on MECs, an adsorption isothermal model based on the Langmuir model was employed to establish the relationship between the coverage of monolayer adsorbate and temperature (adsorption energy)30. The adsorption energies shown in Fig. 3c, d are fitted using the Kennel smooth approach to obtain the adsorption energy distribution function of serine enantiomers, as depicted in Fig. 3e, f. Following a type-I isotherm behavior, only high energy sites (μ4) are available for serine at low concentrations, maximizing the Gibbs free energy drop. The cumulative uptake of serine as a function of adsorption energy is plotted in the right axes of Fig. 3e, f. A gradual increase in coverage is observed due to the high threshold energy and low probability among the surface sites. As the concentration of serine increases, the low energy sites (μ3’) with higher availability become occupied, resulting in a rapid increase in surface coverage.

To elucidate the enantioselectivity of MECs, it is necessary to establish a connection between the difference in adsorption energy and enantioselectivity. Drawing upon concepts from statistical perturbation theory38,39, the enantioselectivity between the adsorbed L- and D-enantiomers on the surface can be described by ref. 40

$$\alpha=\frac{{N}_{{{\rm{L}}}}}{{N}_{{{\rm{D}}}}}=\exp \left(-\frac{{\sum }_{i=1}^{n}{g}_{i}\left({E}_{i}^{{{\rm{L}}}}-{E}_{i}^{{{\rm{D}}}}\right)}{{{{\rm{k}}}}_{{{\rm{B}}}}T}\right)$$
(1)

where gi represents the degeneracies of state i, and \({E}_{i}^{{{\rm{L}}}({{\rm{D}}})}\) denotes the adsorption energy of the L(D)-enantiomer for state i. On ideal stepped metal surfaces, enantioselectivity is solely dependent on the difference in adsorption energy between enantiomers at the most stable locations. However, due to the random nature of MECs, versatile discrete energy states are generated. By considering multiple adsorption configurations, the enantioselectivity α value of serine is determined to be 1.55.

Another crucial characteristic of chiral surfaces is their ability to selectively adsorb one enantiomer over the other after the deposition of a racemic mixture. By introducing the concept of competitive adsorption, where different adsorbates compete for adsorption sites to maximize the adsorption energies (Fig. 4a), we weigh the adsorption energies of enantiomers on each site to thoroughly assess the propensity of the given chiral surface. This approach helps mitigate potential errors arising from outliers in adsorption energy data. An enantiomeric adsorption energy distribution, based on competitive adsorption, is presented in Fig. 4b, while the corresponding adsorption uptake, utilizing the condensation approximation (CA)30, is shown in Fig. 4c. The CA indicates that all sites with adsorption energy higher than a predetermined value are occupied, while all sites below this value remain unoccupied. At low coverage, high adsorption energy sites are preferentially occupied, and the D-enantiomers exhibit a stronger preference for adsorption on MECs. As the coverage increases, the low adsorption energy sites become populated, leading to a higher possibility of L-enantiomer adsorption. Ultimately, at a coverage of 0.25 ML, the adsorption uptake of L- and D-enantiomers is 71% and 29%, respectively.

Fig. 4: Competitive adsorption of serine enantiomers on the (CrMoTa)Si2(0001) surface.
figure 4

a Schematic diagram of competitive adsorption. For multicomponent adsorption, different adsorbates compete for the adsorption sites characterized by maximum adsorption energies. Here, we weigh the adsorption energies of L- and D-serine on given sites to evaluate the adsorption preference of those sites. b, c Revised adsorption energy distribution considering the competitive adsorption of serine enantiomers (b) and the corresponding adsorption uptake (c). The adsorption uptake is obtained by \({\theta }_{{{\rm{t}}}}={\int }_{-\infty }^{{\varepsilon }_{{{\rm{e}}}}}\{\theta (\varepsilon )X(\varepsilon )\}{{\rm{d}}}\varepsilon\). Source data are provided as a Source Data file.

Full size image

To understand this phenomenon, we computed the deformation energy (Edef) of the substrate, the molecules, and the systems across 1620 samples (Fig. 5a). A negative correlation between the deformation energy difference and the adsorption energy difference is observed in our systems (Supplementary Fig. 12). This suggests that the origin of the enantioselective adsorption is presumably due to the larger structural deformation upon D-serine adsorption compared to L-serine. A similar mechanism has also been observed for serine on Cu(531), cysteine on Au(532), and a series of α-amino acids on stepped metal surfaces5,41,42.

Fig. 5: Deformation energy and electronic structure analysis of enantioselective adsorption.
figure 5

a Statistical results of deformation energy of serine enantiomers on MECs. The left column presents the deformation energies of the substrate, molecules, and systems for D-serine adsorbed on MEC films, while the right column presents these energies for L-serine. b Schematic diagram of charge distribution before and after the adsorption of the molecule on MECs. c–f Spatial distribution of charge transfer for L- and D-serine enantiomers on CrSi2 (c, d) and MECs (e, f). The isosurface value is set to 0.005 and 0.05 e Å−3 for serine on CrSi2 and MECs, respectively. g–i Projected density of states (PDOS) of (g) L- and (h) D-serine on CrSi2 and MECs, and (i) ΔPDOS, which represents the difference in PDOS between L-enantiomers and D-enantiomers. Source data are provided as a Source Data file.

Full size image

Furthermore, we delve into the electronic properties from the perspective of charge transfer to gain a deeper understanding of enantioselectivity. The presence of abundant unpaired electrons, contributed by Ta with a valence electron configuration of 5d36s2, imparts metallic properties to the MECs. Upon serine adsorption, these unpaired electrons are injected into the unoccupied molecular orbitals, including the lowest unoccupied molecular orbital (LUMO), LUMO + 1, and higher levels of molecular orbitals (Fig. 5b). Analysis of the spatial distribution of charge transfer reveals that it predominantly occurs at the amidogen (−NH2), deprotonated β-hydroxyl (−O−), and carboxylate (−COO−) groups (Fig. 5c–f). The charge transfer between L- and D-serine on CrSi2 is nearly equal, while a distinct charge difference is observed on MECs.

To describe the number of states and the differences at specific energy levels upon enantiomeric adsorption, we utilize the projected density of states (PDOS) and projected density of states difference (ΔPDOS). This approach is particularly suitable for systems involving multi-orbital charge transfer. In most cases, the PDOS on MECs appears more broadened compared to those on CrSi2 (Fig. 5g, h), indicating higher adsorption energy and increased charge transfer on MECs. ΔPDOS represents the difference in PDOS between L- and D-enantiomers at the same energy level. Similar to the molecular-orbital density of states difference5, ΔPDOS directly characterizes enantioselectivity. On CrSi2, ΔPDOS exhibits only small differences, suggesting low enantioselectivity (Fig. 5i). On MECs, ΔPDOS displays more pronounced differences compared to that of CrSi2, implying that incorporating multiple elements to break symmetry upon adsorption can enhance enantioselectivity.

Besides the most stable crystalline orientation, the second most stable crystalline orientation, (\(11\bar{2}0\)), is chosen to investigate enantioselectivity. Unlike MECs(0001), MECs(\(11\bar{2}0\)) exhibits local asymmetry. High-throughput calculations were performed to study the enantioselective adsorption of serine on 270 MECs(\(11\bar{2}0\)). The obtained adsorption energy distribution function follows an approximate Gaussian distribution (Supplementary Fig. 13). There is a significant difference in the adsorption energy distribution peak between L- and D-serine on MECs(\(11\bar{2}0\)). The average adsorption energy for L-serine (D-serine) is − 1.94 eV (− 1.62 eV). Considering competitive adsorption, we find that 68.9% of sites prefer to adsorb L-serine, while 31.1% prefer D-serine on MECs(\(11\bar{2}0\)).

To verify the university of enantioselective adsorption on MECs, we investigated two additional amino acids, cysteine, and alanine, on 810 random MECs(0001). One is polar, and the other is non-polar. Similar to serine, cysteine enantiomers exhibit two possible adsorption geometries (μ3’ and μ4). However, the average adsorption energies of cysteine (− 2.89 eV for L-cysteine and − 2.86 eV for D-cysteine) are higher than those of serine (− 2.18 eV for L-serine and − 2.16 eV for D-serine) on MECs(0001) (Supplementary Fig. 14a). This is attributed to the stronger interaction between the deprotonated sulfhydryl group and MECs compared to the hydroxyl group, as observed with serine and cysteine on chiral metal surfaces (e.g., Cu(531), Au(324), etc.). Considering competitive adsorption, 56.9% of the surfaces preferentially adsorb L-cysteine, while 43.1% preferentially adsorb D-cysteine (Supplementary Fig. 14b). In contrast to the adsorption behaviors of serine and cysteine, alanine, which only has two functional groups (amidogen and carboxyl group), results in two possible adsorption configurations (μ2 and μ3). Due to steric constraints, most alanine adsorbs on MECs(0001) in the μ2 configuration. The average adsorption energy for L-alanine (D-alanine) is − 1.90 eV (− 1.91 eV) (Supplementary Fig. 14c). Considering competitive adsorption, 49.7% of the sites prefer L-alanine, while 50.3% prefer D-alanine (Supplementary Fig. 14d). This negligible difference originates from the μ2 adsorption configuration, which is ineffective for enantioselective adsorption. Consequently, we conclude that MECs(0001) can effectively recognize polar amino acids, as their side chains can interact with MECs, but are unable to recognize non-polar amino acids.

The adsorption states of serine on MECs were characterized using X-ray photoelectron spectroscopy (XPS). The MECs were immersed in both L- and D-serine solutions at a concentration of 0.1 M for 24 h. The C 1 s spectra exhibit three peaks at 288.4 (288.3), 286.2 (286.2), and 284.8 (284.8) eV, corresponding to the deprotonated carboxylic acid carbon, α- and β-carbon atoms, and decomposition species of L-serine (D-serine), respectively (Fig. 6a, d). The main peak in the N 1 s spectra is centered around 400 eV, which originates from the bond between the neutral NH2 group and the MECs (Fig. 6b, e). A noticeable difference in peak intensity is evident in the O 1 s spectra (Fig. 6c, f). The narrow O 1 s peak at 531.4 eV can be assigned to the two oxygen atoms of the deprotonated carboxylate group, while the peak at 532.7 eV corresponds to β-OH groups, which are not bonded to the surface. Therefore, the XPS spectra confirm that both L- and D-serine are chemically bound to the MECs in the experiment. This observation aligns with previous findings of serine chemically adsorbed on Cu(531) surfaces43.

Fig. 6: Enantioselective adsorption of serine on MEC films.
figure 6

a–f XPS spectra of C 1 s, N 1 s, and O 1 s for (a–c) L- and (d–f) D-serine adsorbed on MECs. g Real-time adsorption behaviors of serine enantiomers based on the frequency shift measured by QCM. The concentration of serine solution is 0.3 M. h The frequency shift changes of serine enantiomers on the MEC surface according to the concentration changes. The error bars represent the standard deviation of three independent samples. The red dots and lines represent the L-serine, while the blue ones correspond to D-serine. Source data are provided as a Source Data file.

Full size image

The enantioselective adsorption of serine on MECs was investigated using a modified electrode in the QCM system. The adsorption uptake was determined by monitoring the frequency shift of quartz oscillation under an applied voltage, in accordance with the Sauerbrey equation. Our QCM system exhibits a high sensitivity of 0.62 ng cm−2 Hz−1, providing sufficient accuracy for detecting amino acid adsorption44. In our experiments, real-time adsorption behaviors were recoded until the equilibrium adsorption capacity was reached, indicated by a frequency shift variation of less than 1 Hz min−1 (Supplementary Fig. 15). As shown in Fig. 6g, the frequency shift for L-enantiomers (139 Hz after adding 100 μL solution) is significantly larger than that for D-enantiomers (92 Hz after adding 100 μL solution), highlighting the distinct adsorption capabilities of the enantiomers on MECs. To ensure the reliability of the enantioselective adsorption, the process was repeated three times (Fig. 6h), and the atomic distributions of these three electrodes were characterized by scanning electron microscope-energy dispersive X-ray spectroscopy (Supplementary Fig. 16). The results reveal a linear relationship between adsorption amount and analyte concentration, consistent with the Henry adsorption isotherm. At a concentration of 60 mM, the enantioselectivity (α value) of serine was 1.58 and e.e. was 22.34%. A comparison of enantiomeric adsorption shows that L-serine exhibits a higher preference for adsorption on the MEC film, aligning with the predictions derived from DFT calculations. In addition, the frequency shift increases quasi-linearly with concentrations ranging from 0.05 to 0.20 M (Supplementary Fig. 17), accompanied by a decrease in the e.e. value. This phenomenon is attributed to the formation of a multilayer adsorbate, whose preference is influenced by the adjacent layer near the chiral surfaces rather than the chiral surfaces themselves. A nearly equal frequency shift is observed on clean and achiral Au electrode surfaces, indicating that the enantioselective adsorption on MECs is due to the nature of chirality (Supplementary Fig. 18).

To further evaluate the adsorption behaviors of serine on MECs, both the Langmuir and Freundlich isotherms were employed to fit the adsorption amount to the concentration of the solution (definitions of both models are provided in the Methods section). Both adsorption models demonstrate a good fit with the enantioselective adsorption of serine on the MEC surface (Supplementary Fig. 19). In the Langmuir model, the intercept (equivalent to 1/ΔFmax) approaches 0, indicating that the adsorption behavior occurs in a rapid and linear increase stage. In the Freundlich model, the n values for L- and D-serine are determined to be 0.82 and 0.95, respectively, suggesting nearly linear adsorption behavior of serine on the MEC surface. The Henry isotherm was also used to estimate the adsorption behaviors of serine enantiomers on three electrodes (Fig. 6h). The pre-factors of L- and D-serine in the Henry isotherm were calculated to be 2279 and 1489 Hz M-1, respectively. These analyses provide further evidence of the enantioselective adsorption of serine on the MECs surface.

Discussion

In conclusion, we have presented an effective approach utilizing MEC surfaces for chiral recognition. Our method stands out from previous approaches as it necessitates the elimination of stepped and kink atoms on the chiral surfaces, thereby preventing kink coalescence and the formation of kinks at the intersections of different long step edges, ensuring the stability of the chiral structure. Through extensive high-throughput calculations, we have thoroughly investigated the thermal stability of MECs and the adsorption behaviors of serine on MECs. By extending the isotherm adsorption model to multicomponent silicides, we observed two distinct energy distribution peaks arising from μ4 configurations with high adsorption energy and μ3’ configurations with low adsorption energy. Taking into account the coadsorption of L- and D-serine, we estimated a significant adsorption uptake difference (42% e.e.) for serine on MECs. The MECs have been synthesized using non-reactive magnetron sputtering, and the structure has been confirmed by GIXRD, TEM, and CD spectra. The enantioselectivity of serine on the MECs surface has been further confirmed through QCM, yielding an α value of 1.58. The practical application of MECs holds great potential for enantiomeric catalysis and asymmetry synthesis.

Methods

Computational methods

DFT calculations were performed using the Vienna ab initio simulation package (VASP) within a high-throughput frame. The generalized gradient approximation (GGA) of the Perdew-Burke-Ernzerhof (PBE) functional was utilized as the exchange-correlation functional45,46. The semiempirical DFT-D3 scheme was employed to account for dispersion interactions, ensuring the generation of appropriate adsorption structures and energies47. An energy cutoff of 400 eV was used for structural optimization and single-point energy calculations. The Brillouin zone was sampled with Monkhorst-Pack sampling of (5 × 5 × 9) and (5 × 5 × 1) k-points for the bulk and slab of (CrMoTa)Si2, respectively48. The convergence criterion for the energy difference was set to 10−4 eV for structural optimizations and 10−5 eV for single-point energy calculations. The structural optimization was iterated until the force is less than 0.02 eV Å−1.

The thermal stability was estimated by the adatom formation energy and the energy barrier of the transition barrier, which involves atomic ejection and migration. The procedure to evaluate the adatom formation energy (Eform) involves the utilization of two individual and separated slab models. The corresponding equation is as follows35:

$${E}_{{{\rm{form}}}}={E}_{{{\rm{s}}}+{{\rm{vac}}}}+{E}_{{{\rm{d}}},{{\rm{adatom}}}}-{E}_{{{\rm{s}}}}-{E}_{{{\rm{d}}}}$$
(2)

where Es+vac /Es and Ed,adatom/Ed represent the source surface energy with/without the atomic ejection and the drain surface energy with/without adatom, respectively. The energy barrier of the transition state was calculated using the CI-NEB method49,50. In CI-NEB calculations, initial pathways for the transition state were generated with five intermediates between the reactants and products. The intermediates were then optimized using transition state tools within the VASP code51.

Experimental procedures

Materials

The electrodes with Au coating for quartz crystal microbalance (QCM) measurement were purchased from Initium Co., Ltd (Japan). L- and D-serine (> 99% purity) were purchased from Aladdin Reagent Co., Ltd. A multiple-phase (CrMoTa)Si2 target with 50.8 (diameter) × 2 (thickness) mm2 was purchased from Hefei Kejing Material Technology Co., Ltd. The water used in the experiments was purified to a resistivity of 18.2 MΩ cm−1 using a Purelab Prima system. Sulfuric acid solution (95%), hydrogen peroxide solution (30%), and absolute ethyl alcohol (99.7%) were purchased from Sinopharm Chemical Reagent Co., Ltd. Sodium dodecyl sulfate (99.5%) was purchased from Shanghai Macklin Biochemical Co., Ltd.

Magnetron sputtering

The (CrMoTa)Si2 films were deposited using a non-reactive magnetron sputtering system with a high vacuum (base pressure < 6.6 × 10−5 Pa). The system was equipped with 2 direct current (DC) magnetrons for metals and 1 radio frequency (RF) magnetron for semiconductors and ceramics. The distance between the target and the substrate was set to 80 mm. A power-metallurgically produced 50.8 (diameter) × 2 (thickness) mm2 composite silicide target, consisting of 33.3 % (molar ratio) CrSi2, MoSi2, and TaSi2, was used. After reaching a base pressure of < 1 × 10−4 Pa, the substrates were etched with Ar-ion using DC for the metallic transition layer and RF for the MEC film at an Ar gas pressure of 0.4 Pa. The sputtering power was set to 40 W for Cr and 60, 70, and 80 W for the MEC film. The sputtering time was set to 10 s for Cr and 10 or 20 min for MECs.

Characterization

Grazing incidence X-ray diffraction (GIXRD, Bruker D8, Cu-Kα radiation) was used to identify the crystal phases of the materials at 40 kV and 40 mA. The MECs surface and the Cr transition layer on the QCM electrode were observed using an AFM (ICON, Bruker) in tapping mode. The morphology and elemental distribution of the samples were characterized using a scanning electron microscope (SEM, TESCAN mira4) combined with energy disperse spectroscopy (EDS, Ultim Max 65, Oxford spectrum).

The QCM sensing system

The QCM (Affinix QN Pro; Initium, Tokyo, Japan) was utilized for the sensing section, which featured an electrode modified by MECs. The base frequency of the AT-cut quartz crystal was 27 MHz, with a sensitivity of 0.62 ng cm−2 Hz−1. The mass of the adsorbate was determined from the frequency shift using the Sauerbrey equation:

$$\Delta m=-\frac{A\sqrt{{\mu }_{{{\rm{q}}}}{\rho }_{{{\rm{q}}}}}}{2{F}_{0}^{2}}\Delta F$$
(3)

where Δm, A, μq, ρq, F0, and ΔF are the mass of adsorbate (in g), the surface area of electrode (in cm−2), the shear modulus of quartz (2.95 × 1011 dyn cm−2), the density of quartz (2.65 g cm−3), the fundamental frequency of quartz crystal (2.7 × 107 Hz), and the measured frequency shift (in Hz), respectively. The mass of serine monolayer was estimated using Eq. (4):

$$\Delta m={\rho }_{V}A=\frac{{m}_{{{\rm{unitcell}}}}}{{A}_{{{\rm{unitcell}}}}}\times 4.909\times {10}^{-2}{{\rm{ng}}}=4.699{{\rm{ng}}}$$
(4)

where munitcell and Aunitcell represent the mass of 1 ML serine and the area of the unit cell, respectively. A is the area of the QCM electrode, which is 4.909 × 10−2 cm2 in the QCM system used. The frequency shift of the monolayer was obtained from the mass of the adsorbate using Eq. (3):

$$\Delta {F}_{{{\rm{m}}}}=-\frac{2{F}_{0}^{2}}{A\sqrt{{\mu }_{{{\rm{q}}}}{\rho }_{{{\rm{q}}}}}}\times \Delta m=154.39{{\rm{Hz}}}$$
(5)

In our experiment, the adsorption was conducted in an aqueous environment. First, 400 microliters of water were injected into the electrode cell, and the system was allowed to reach adsorption equilibrium. This step is crucial for eliminating the influence of the solvent. Then, 20 μL of serine solution was injected into the electrode cell. The frequency shift was recorded until adsorption equilibrium was achieved. This operation was repeated to record the real-time adsorption behavior at different concentrations. The multiple additions of the solution aimed to fully occupy the adsorption sites through multiple adsorption equilibria.

Enantioselectivity of serine on MECs

The enantioselectivity, given in terms of alpha value (α) and enantiomeric excess (e.e.), can be estimated using the following formulas:

$$\alpha=\frac{\frac{{m}_{{{\rm{ads}}}}\left(L\right)}{A}\cdot \Delta t}{\frac{{m}_{{{\rm{ads}}}}\left(D\right)}{A}\cdot \Delta t}=\frac{\Delta {F}_{{{\rm{ads}}}}\left(L\right)}{\Delta {F}_{{{\rm{ads}}}}\left(D\right)}$$
(6)
$${{\rm{e}}}.{{\rm{e}}}.=\frac{{m}_{{{\rm{ads}}}}\left(L\right)-{m}_{{{\rm{ads}}}}\left(D\right)}{{m}_{{{\rm{ads}}}}\left(L\right)+{m}_{{{\rm{ads}}}}\left(D\right)}=\frac{{\Delta F}_{{{\rm{ads}}}}\left(L\right)-{\Delta F}_{{{\rm{ads}}}}\left(D\right)}{{\Delta F}_{{{\rm{ads}}}}\left(L\right)+{\Delta F}_{{{\rm{ads}}}}\left(D\right)}$$
(7)

where mads(L) and mads(D) are the adsorption amounts of L- and D-enantiomers, respectively. A is the surface area of the electrode. Δt is the adsorption time.

Adsorption isotherms

The Langmuir, Freundlich, and Henry adsorption isotherms were used to characterize the adsorption of serine on the surfaces of MECs. The frequency shift as a function of concentration was obtained by the QCM system. For the Langmuir adsorption isotherm, the adsorption equation can be written as the following formula:

$${\Delta F}_{{{\rm{e}}}}=\frac{{\Delta F}_{\max }{K}_{{{\rm{L}}}}{c}_{{{\rm{e}}}}}{1+{K}_{{{\rm{L}}}}{c}_{{{\rm{e}}}}}$$
(8)

where ΔFe, ΔFmax, KL, and ce represent the amount of adsorbed analyte at equilibrium, the theoretical maximum adsorption capacity, the Langmuir equilibrium constant, and the concentration of analyte at equilibrium, respectively. For the Freundlich adsorption isotherm, the adsorption equation can be written as the following formula:

$${\Delta F}_{{{\rm{e}}}}={K}_{{{\rm{F}}}}{c}_{{{\rm{e}}}}^{\frac{1}{n}}$$
(9)

where KF is the Freundlich equilibrium constant, and n is a constant related to the adsorption intensity. For the Henry isotherm, the adsorption equation can be written as the following formula:

$${\Delta F}_{{{\rm{e}}}}={K}_{{{\rm{H}}}}{c}_{{{\rm{e}}}}$$
(10)

where KH is the Henry equilibrium constant.