Separation of enantiomers by their enantiospecific interaction with achiral magnetic substrates

See allHide authors and affiliations

Science  22 Jun 2018:
Vol. 360, Issue 6395, pp. 1331-1334
DOI: 10.1126/science.aar4265

Taking enantiomers for a spin

There are two common ways to distinguish mirror-image molecules, or enantiomers. The first relies on their distinct interactions with circularly polarized light, the second on their interactions with a pure enantiomer of some other molecule. Now Banerjee-Ghosh et al. report a conceptually different approach to chiral resolution. Experiments showed that, depending on the direction of magnetization, chiral oligopeptides, oligonucleotides, and amino acids have enantiospecific differences in initial adsorption rates on ferromagnetic surfaces. This effect is attributed to enantiospecific induced spin polarization.

Science, this issue p. 1331


It is commonly assumed that recognition and discrimination of chirality, both in nature and in artificial systems, depend solely on spatial effects. However, recent studies have suggested that charge redistribution in chiral molecules manifests an enantiospecific preference in electron spin orientation. We therefore reasoned that the induced spin polarization may affect enantiorecognition through exchange interactions. Here we show experimentally that the interaction of chiral molecules with a perpendicularly magnetized substrate is enantiospecific. Thus, one enantiomer adsorbs preferentially when the magnetic dipole is pointing up, whereas the other adsorbs faster for the opposite alignment of the magnetization. The interaction is not controlled by the magnetic field per se, but rather by the electron spin orientations, and opens prospects for a distinct approach to enantiomeric separations.

The relation between magnetism and chirality was the basis for an argument between two giants of science, Pasteur and Kelvin (1, 2), over a century ago. Pasteur, after discovering that the chemistry of life showed a preference for molecules with a particular handedness (3, 4), tried to carry out experiments aimed at finding asymmetric physical forces that could explain the biological homochirality (5). He attempted to induce handedness by stirring the reactants in a centrifuge and applying a magnetic field, but both attempts failed. In 2001, Ribo and co-workers reported that stirring solutions of achiral porphyrins could induce chiral-symmetry breaking in the resulting mesophases (6). Chiral supramolecular organization was also induced by vortex flow (7). However, further attempts to induce chirality by magnetic field failed (8). This lack of success is consistent with Kelvin’s conclusion that “the magnetic rotation alone has neither left-handed nor right-handed quality” (9). Moreover, de Gennes has demonstrated that even the superposition of a magnetic field and an electric field, originally suggested by Curie (10), does not induce asymmetry in reactions (11). He claimed, however, that if the final state is out of equilibrium, asymmetry remains possible. Here we present another twist in this long-standing debate by demonstrating an enantioselective interaction of chiral molecules with a substrate magnetized perpendicular to its surface. The discrimination is mediated by a spin-specific interaction but not by the magnetic field per se.

Recent studies have shown that, when electrons move through chiral molecules, their transport is spin dependent, with the preferred spin orientation determined by the handedness of the molecule and the direction of motion (12, 13). The effect, which by now is well established (1418), is known as chirality-induced spin selectivity (CISS). As a corollary, charge redistribution in chiral molecules is also accompanied by enantiospecific spin polarization (19). These results led us to consider the possible interaction between chiral molecules with perpendicularly magnetized surfaces. This interaction should be spin sensitive by virtue of short-range magnetic-exchange interactions. Therefore, it could result in enantiospecificity through the above-mentioned relation between spin and chirality via the CISS phenomenon. If this were the case, it would facilitate the separation of a racemic mixture into its two enantiomeric components simply by allowing the mixture to interact with the magnetic substrate.

As a first examination, an l– or d–polyalanine (PAL)–based oligomer, a thiolated α-helix oligopeptide, SH-CAAAAKAAAAKAAAAKAAAAKAAAAKAAAAKAAAAK (C, A, and K represent cysteine, alanine, and lysine, respectively), was exposed to a ferromagnetic (FM) cobalt film covered with 5 nm of gold for 2 s (see supplementary materials). This oligomer was chosen because of its stable helical structure and the existence of vast information on its adsorption on gold (12). The substrate configuration was chosen on the basis of prior studies, which have established that a thin layer of a noble metal like gold or platinum (up to about 10 nm), deposited on a ferromagnet, does not diminish magnetic or spin transport properties but protects the ferromagnet from oxidation (2022). SiO2 nanoparticles (NPs) were attached to the adsorbed PAL to act as a marker in the scanning electron microscope (SEM) images for the monolayer adsorption density. Figure 1A, (i) and (ii), shows SEM images of the l-enantiomer adsorbed on the FM substrate, with the latter magnetized up or down, respectively. The concentration of adsorbed particles in Fig. 1A, (i), is 4 ± 0.4 × 1010 NPs/cm2, whereas in Fig. 1A, (ii), it is 6 ± 1 × 109 NPs/cm2. Results of the complementary experiment with d-PAL are shown in Fig. 1A, (iii) and (iv), with concentrations of 1 ± 0.2 × 1010 and 4 ± 0.5 × 1010 NPs/cm2 found for the up- and down-magnetized substrate. As a control experiment, l- and d-enantiomers were also adsorbed on a nonmagnetic, pure gold substrate, with an external magnetic field applied either normally or antinormally to the substrate. In this case, no enantioselectivity was observed [see a full comparison in Fig. 1A, (v)]. In addition, the adsorption efficiency in this case was found to be between that of the FM substrate with the favored and unfavored spin alignment. Moreover, when the substrate was magnetized in-plane, no selective adsorption was observed. The experiments described here were conducted on 10 separate sets of samples, and the error analysis presented in the figures is based on particle-number counting at different parts of each sample. A complete discussion of the reproducibility of this experiment, as well as of the additional ones reported below, is provided in the supplementary materials.

Fig. 1 Enantiospecific adsorption of a peptide.

(A) Adsorption of the PAL oligopeptide [shown in inset of (v)] on FM samples (silicon with a 1.8-nm Co film and a 5-nm Au film), magnetized with the magnetic dipole pointing up (H+) or down (H) relative to the substrate surface. SiO2 nanoparticles were attached to the adsorbed oligopeptides. Panels (i) and (ii) l-PAL and (iii) and (iv) d-PAL were adsorbed for 2 s on a substrate magnetized up or down. Panel (v) summarizes the nanoparticle adsorption densities shown in (i) to (iv), compared with the adsorption density on Au with an applied external magnetic field (red bars). Double-headed arrows represent error bars. The errors are the standard deviation among 10 measurements conducted on each of the 10 samples, hence a total of 100 measurements. (B) Panel (i) shows the enantiospecific CD spectra of PAL, obtained from exposure of a racemic PAL mixture exhibiting no CD to a substrate with magnetization pointing down (red) or up (blue). CD spectra were obtained postadsorption. After the specific adsorption of one enantiomer, the resulting CD spectra clearly indicate the presence of the opposite enantiomer. The linewidth reflects the uncertainty of the results. Panel (ii) shows the CD spectra of the pure enantiomers, provided for comparison. mdeg, millidegrees.

The above results clearly demonstrate enantioselectivity in the adsorption process. We note that, in one magnetization direction, the l-PAL adsorption rate is 8 ± 2 times faster than that of d-PAL, whereas in the other magnetization direction, the d-PAL adsorption rate is 4 ± 1 times faster than that of l-PAL. However, the d-PAL purification level is lower than that of l-PAL, likely explaining the asymmetry in adsorption rate ratios (see supplementary materials). Furthermore, repetition of this experiment with a longer adsorption time (2 min) resulted in a reduction in the enantioselectivity of adsorption (see figs. S1 and S2), that is, the process exhibits substantial kinetic differences.

The kinetic effects in the selective adsorption are related to the adsorption-desorption kinetics typical to the formation of self-assembled monolayers of thiolated molecules (23). Specifically, it is well known that, when self-assembled monolayers are formed, the process is initially controlled by the rate of adsorption, the kinetics. At long time scales, the thermodynamics defines the adsorption, and, for two species that have the same energy for binding to a surface, the quantity of adsorbed molecules will be defined by their concentration. In the specific case described here, at short adsorption time, selectivity is observed because of the different rates of adsorption for the two enantiomers; however, at longer times, because the binding energies are identical, the two enantiomers will be equally adsorbed, and hence the enantioselectivity is reduced. See more on this subject below.

We next explored the separation of a racemic PAL mixture into its two enantiomeric components upon exposure to a magnetic substrate. The original mixture exhibited no circular dichroism (CD). As shown in Fig. 1B, (i), upon sequential exposure to 100 1 cm–by–1 cm substrates with up- or down-pointing magnetization, a clear CD signature corresponding to d- or l-PAL, respectively, was obtained from the residual solution. Furthermore, this signature is similar to that obtained for the pure enantiomers, as shown in Fig. 1B, (ii).

To implement a larger-scale enantioseparation apparatus, we coated a flow column on one side with a thin film consisting of 50 Al/5 Ti/0.3 Co/(0.7 Ni/0.3 Co)6/5 Au (all numbers are in nm), magnetized externally either normally or antinormally to the surface. This particular film was chosen because it is a relatively soft magnet and could easily be magnetized with the external magnetic field (see supplementary materials). Figure 2 shows the CD spectra of a racemic [Cya-(Ala-Aib)5-CO2H] solution at the inlet and outlet of the apparatus for the two magnetization directions, where Cya is cystamine, Ala is alanine, and Aib is aminoisobutyric acid. Clearly, for a given magnetic direction, one enantiomer is adsorbed on the Au. The opposite enantiomer therefore accumulates in excess at the outlet and dominates the CD spectrum. The results here indicate once again that the l-enantiomer favors a surface that is magnetized up (that is, H+), whereas the d-enantiomer favors the down magnetization (that is, H). Quantitative details are provided in the supplementary materials.

Fig. 2 Flow-based separation of enantiomers.

CD spectra of a racemic d- and l-[Cya-(Ala-Aib)5-CO2H] oligopeptide mixture at the inlet (red) and outlet of a magnetic column (i), with an external (ii) magnetic field pointing up (black) or down (blue). Enantioseparation is clearly obtained.

To monitor the kinetics of enantioselective adsorption directly, we monitored the fluorescence of double-stranded DNA (dsDNA) oligonucleotides (20 base pairs in length), to which a fluorescent dye was attached. The Cy-3 (cyanine) dye was tagged at the 3′ position (cytosine) of the dsDNA (see supplementary materials). In this case, the molecules were adsorbed on a Ni/Au surface, and the fluorescence was measured for different adsorption times and for different Ni magnetization directions (Fig. 3A). We verified that the same results were obtained with the Ni/Au as with the Co/Au substrates. Figure 3B shows the adsorption as a function of time and demonstrates that the adsorption rate is considerably different for up and down magnetization of the substrate.

Fig. 3 Selective adsorption of dsDNA and cysteine.

(A) Fluorescence spectrum measured from adsorbed dsDNA on a Si wafer coated with 8 Ti/100 Ni/8 Au (all numbers are in nm), with the magnet pointing up or down. a.u., arbitrary units. (B) Time dependence of the adsorption process, obtained when the magnetic dipole of the substrate is pointing up (blue) or down (red). (C) Adsorption specificity, defined as Embedded Image, where AD and AU are HPLC-measured quantities of adsorbed cysteine on the magnetic substrate with magnetic moment pointing down or up, respectively, as a function of time. The size of the points on the graph represents the error in the adsorption specificity.

To establish that the effect extended beyond oligopeptides and dsDNA, we probed the adsorption kinetics for a single amino acid, cysteine. Briefly, solutions of l- or d-cysteine were exposed to the same magnetic substrates used in prior experiments, with the magnetization pointing up or down. The change in concentration was measured by applying high-performance liquid chromatography (HPLC). We define the adsorption specificity as Embedded Image, where AD and AU denote the quantity of adsorbed molecules measured for the two magnetization directions, namely, down or up, respectively. The results, presented in Fig. 3C, represent data averaged over four sets of experiments (see supplementary materials). They indicate that, for short adsorption times, the adsorption is enantioselective exactly as before. For long adsorption times, the selectivity is lost, a result which is, again, consistent with that obtained for PAL and which indicates different adsorption rates on the magnetized substrate. Clearly the kinetics vary for different molecules and different concentrations, as indicated by the relatively slow process observed in dsDNA (Fig. 3B).

The above results demonstrate the generic nature of the effect. Unambiguous enantioselectivity, based on different adsorption rates, on a perpendicularly magnetized FM substrate is obtained throughout for a variety of chiral molecules and magnetic substrates. We now explain in more detail how this can be rationalized in terms of CISS-induced spin polarization in the chiral molecule, facilitating a selective magnetic-exchange interaction with the FM substrate. Under electric dipole polarization, induced by the substrate, an excess of electrons and holes on the negative and positive poles of the molecule, respectively, is obtained. This charge polarization is accompanied by spin polarization, as previously shown both experimentally and theoretically (19, 24). The specific spin orientation at each pole depends on the chirality of the molecule (see Fig. 4A). The spin polarization results from the dynamic process of the electron redistribution in the chiral potential presented by the molecule.

Fig. 4 Theoretical underpinnings.

(A) A schematic cartoon of the polarization. The electrical polarization of the molecule is accompanied by spin polarization. The spin alignment at each electric pole depends on the specific enantiomer. The horizontal arrows indicate the rotation, and the vertical arrows indicate the spin direction. (B) Therefore, for a specific enantiomer, the interaction between the magnetized surface and the molecule (circled in blue and red) follows either a low-spin (i) or a high-spin (ii) potential, depending on the direction of magnetization of the substrate. (C) Calculated interaction energies as a function of distance between a hydrogen atom and a surface represented by a 2 atom–by–2 atom–by–2 atom cube of Ni atoms. All spins of the Ni atoms are aligned parallel to each other, and the H atom spin is aligned either parallel (red curve) or antiparallel (blue curve) to the spin of the Ni atoms. At longer distances, the two different configurations merge in terms of total energy, and that energy is considered as the zero of the energy axis.

If there is an exchange interaction between the molecular spin and the spin of the FM substrate, then the substrate-molecule interaction should be stabilized if the two spins are antiparallel (low-spin configuration) and destabilized if they are parallel (high-spin configuration), as shown in Fig. 4B. Indeed, a prior study found that adsorption of chiral molecules on a nonmagnetized FM substrate tends to magnetize it in opposite directions for opposite enantiomers (25).

The remaining question, then, is whether this spin interaction is large enough to promote enantioselectivity. To explore this, we assume that spin polarization has already been obtained through the dynamic CISS effect and study spin interaction with the FM substrate. The simplest model possible for net spin polarization is given by a H atom with a spin-polarized electron. We then represent the FM layer by a cube of Ni atoms with two atoms in each dimension, whose H-facing surface is oriented along the (111) direction of bulk Ni, with spins aligned parallel to each other owing to the FM interaction. The model we present is not a model of chirality; rather, it is a model for spin polarization. In fact, theory for the CISS effect is an ongoing topic of investigation [see, for example, reference (24)]. However, the model we present does show that the spin energetics can indeed drive the effect.

We performed density functional theory (DFT) calculations exploring the interaction energies as a function of distance, using the Perdew-Burke-Ernzerhof functional (26) (computational details are given in the supplementary materials). In one case, the H-atom spin was considered to be along the same direction of all the Ni atoms (high-spin, FM configuration) and in the other in the opposite direction (low-spin, anti-FM configuration). Clearly, this is but a crude model for the real system, whose accuracy is further limited by known limitations on charge (27) and spin (28) dissociation with conventional DFT. Nevertheless, the results, given in Fig. 4C, show that for all the distances up to about 0.35 nm, the total energy of the system for the high-spin configuration is higher than that of the low-spin configuration. The energy difference at the minimum of the potential is about 0.6 eV, which is well above thermal energy at room temperature. The H atom represents an extreme case of charge polarization (a full electron). Previous studies indicated that the polarization involves a spin of about a full electron (19, 29). However, even if only 1/10 of an electron charge is polarized, the difference between the two spin states would still easily allow for an observable effect.

To show that a similar effect is obtained with a more realistic molecular model, we considered a small chiral molecule, CH3C(OH)H(NH), whose radical character results in a net spin polarization in the gas phase. Interaction energy curves, similar to those shown in Fig. 4C, are given in fig. S3. Here, too, the energy difference is sizeable (~0.4 eV at the minimum of the potential) and decaying with increasing molecule-substrate separation, even though the spin on the atoms close to the surface is less than that of a full electron.

The curves of Fig. 4C show two binding-potential landscapes. If the molecule-surface interaction time is long enough, compared to the time it takes to cross from the upper potential to the lower one by spin orbit coupling or by hyperfine interaction, then the molecular spin orientation may flip, and the molecule will interact with the surface on the lower potential energy surface [see also (30) for a discussion in terms of “true” and “false” chirality]. However, at shorter times, each spin polarization will experience a different interaction potential and, therefore, different adsorption kinetics, consistent with the experimental results of Figs. 1 and 3B. It should be realized that, although the spin polarization controls the kinetics of the process, because of the spin flipping, on a long time scale, thermodynamics will prevail, namely, all the molecules will be adsorbed on the low–potential energy surface independent of the chirality. This process is related to the kinetics of monolayer formation in which molecules adsorb and desorb. After a long enough time, the transient spin effect diminishes, and the formation is controlled by concentration. It is important to appreciate that the “crossing time” between the potentials depends on molecular properties (spin orbit and hyperfine couplings) as well as on the number of collisions with the surface per second and, hence, will vary with concentration. Upon reaching thermodynamic equilibrium, molecules adsorb and desorb. Hence, molecules of the “wrong chirality” replace adsorbed molecules, and the selectivity diminishes. As can be realized from the data presented, the time scale to reach thermodynamics depends on the molecules and the experimental conditions, such as surface area versus concentration in the solution This is why the kinetics can actually be obtained only experimentally. We emphasize once again that the effect reported here is not associated with the direct interaction of chiral molecules with magnetic fields (3134) but rather with the magnetic-exchange interaction of the spin-polarized molecules with the spin-polarized substrate [see Figs. 1A, (v), and 4B]. The direct interaction energy of an electron spin with a magnetic field is typically of the order of micro–electron volts, whereas the interaction energies obtained in Fig. 4C are at least five orders of magnitude greater.

Enantioselectivity is ubiquitous in nature, and many of the molecules in plants and living organisms have specific enantiomeric properties (35). Chiral recognition and enantiomeric selectivity are commonly assumed, both in nature and in artificial systems, to be related to a spatial effect, with the recognition process typically described by a “lock-and-key”–type model (36). Accordingly, chromatography-based enantioseparation requires the chiral substrate to be adjusted so that it interacts optimally with a specific enantiomer (37). Indeed, enantioseparation is an extremely important process in the pharmaceutical and chemical industries. Chromatography and electromigration techniques have long been the methods of choice in this field (3746). However, despite intensive efforts, obtaining enantiomerically pure synthetic materials remains a challenge, as the cost of separation is relatively high and an extensive effort is required.

The enantioselective interaction of chiral molecules with a magnetic substrate, presented in this article, provides a potentially generic chromatographic method for enantioseparation, which does not require a specific separating column. Because the observed effect depends on the electrical polarizability of the system (that is accompanied by spin polarization) and because this polarization depends on the global structure of the chiral molecule, the method described here may also allow the separation of chiral molecules from a mixture of molecules, either chiral or achiral. In addition, this technique could potentially be applied for separating chiral molecules on the basis of their secondary structure and/or for separating two secondary structures of the same chiral molecule. Clearly, the efficiency of the method depends on the electric and spin polarization properties of the molecules and may vary considerably for different systems. We hope that this work will motivate further exploration of the detailed mechanisms and ultimate efficiency of this approach.

Supplementary Materials

Materials and Methods

Figs. S1 to S5

References (4751)

References and Notes

Acknowledgments: We thank A. Brandis for performing the HPLC measurements. Funding: Y.P. and R.N. acknowledge the support of the John Templeton Foundation. R.N. acknowledges the support of the Israel Science Foundation and the European Research Council under the European Union’s Seventh Framework Program (FP7/2007–2013)–ERC grant agreement n° 338720 CISS. Y.P. acknowledges the support of the Israel Science Foundation and the Ministry of Science Israel. O.B.D. acknowledges support from the Israeli Ministry of Science, Technology, and Space, grant number 0399174. L.T.B. acknowledges institutional support from the Institute of Physics Polish Academy of Sciences. Author contributions: K.B.-G. performed the experiments on the DNA and with the HPLC; O.B.D. performed the adsorption experiments on the oligopeptides; F.T. performed the experiments applying the CD; E.C. prepared the setup for the adsorption experiments; S.Y. prepared and characterized the oligopeptides; A.C., S.-H.Y., and S.S.P.P. designed, prepared, and characterized the Ni substrates; S.S. performed and analyzed the calculations; L.K. participated in planning and analyzing the calculations; L.T.B. designed, prepared, and characterized the Co surfaces; R.N. and Y.P. conceived the experiments; and L.K., Y.P., and R.N. wrote the manuscript. Competing interests: E.C., S.Y., R.N., and Y.P. filed a patent on magnetic separation of chiral compounds, U.S. application no. 62/636,903. Data and materials availability: All the data are presented in the manuscript and supplementary materials.

Stay Connected to Science

Navigate This Article