In experiments, 2D quasi-hexagonal phase (qHP) and quasi-tetragonal phase (qTP) C60 have been successfully synthesized [30–32]. The qHP C60 demonstrates better stability [29, 35] than qTP C60 and can exist in both monolayer and few-layer forms. In qHP C60, fullerenes adopt two possible orientations[29–31], which can be distinguished from the highlighted pentagons on the top of a C60, as shown in Fig. 1 (a).
The monolayer qHP C60 has a centrosymmetric structure. In stable few-layer forms, a fullerene tends to align with the center of the triangle consisting of three nearest-neighboring fullerenes from the adjacent layer, as observed in experiments [30, 31]. There are two distinct bilayer configurations (labeled as AB and AB’) of qHP C60, which are both centrosymmetric, as shown schematically in Fig. 1 (b) and (c). In both cases, the bottom layer is shifted relative to the top layer by approximately \(\:b/3\), where \(\:b\) is the lattice constant along the \(\:b\)-direction.
Next, to examine whether trialyer qHP C60 can be noncentrosymmetric, we enumerate all possible trilayer qHP C60’s and identify six different stacking structures. The naming convention and schematic plots of all these trilayer structures are presented in Supplemental Fig. S1 and S2. Among them, two stacking configurations (AB’A’, AB’C’) possess both out-of-plane and in-plane polarizations, two stacking configurations have only in-plane polarizations, and the other two are centrosymmetric. The stacking configurations, point group, and electric-polarization directions are summarized in Table 1. In the following discussion, we focus on the trilayer phases with out-of-plane polarizations.
Figure 1. (a) (left) The structural model of monolayer qHP C60. The blue and yellow pentagon rings are used to highlight C60 clusters in two different orientations. (right) A simplified illustration of the qHP C60 monolayer, where the blue and yellow spheres represent two different orientations of the C60. (b) Bilayer qHP C60 in the AB and (c) AB’ stacking configuration, with the inversion center labeled by the red star. (d) Trilayer C60 with AB’A’ stacking configuration; (e) The isosurface plot (with isosurface value \(\:\pm\:6.5\times\:{10}^{-5}\:e/\text{B}\text{o}\text{h}{\text{r}}^{3}\)) of the differential charge density of the AB’A’ structure, where yellow and green isosurfaces indicate electron accumulation and depletion due to non-centrosymmetric layer stacking; (f) The relative energies of 6 stacking configurations and the out-of-plane polarization of the polar structure.
Table 1
The symmetry properties and polarization directions of 6 different trilayer stacking configurations.
Stacking configurations | Point Group | Out-of-plane (OP) polarization | In-plane (IP) polarization |
AB’A’, AB’C’ | m | ✓ | ✓ |
ABA, AB’A | mm2 | × | ✓ |
ABC, AB’C | 2/m | × | × |
Figure 1 (d) schematically depicts the non-centrosymmetric AB’A’ stacking configuration with both IP and OP polarization. To better illustrate the origin of its polarization, we calculated the charge density differences \(\:{\Delta\:}\rho\:\left(\varvec{r}\right)\) between qHP C60 layers in the AB’A’ stacking configuration and free-standing C60 monolayers. As shown in Fig. 1 (e), due to the weak interlayer interaction, the redistributed charge density mostly occurs at the boundary of the Van der Waals gap between adjacent layers and evidently reveals the asymmetric shape in the AB’A’ structure. In addition, the planar-averaged distribution \(\:{\Delta\:}\rho\:\left(z\right)=\int\:{\Delta\:}\rho\:\left(\varvec{r}\right)dxdy\) along the z-axis quantitatively demonstrates symmetry broken, giving rise to the out-of-plane polarization (see Supplemental Fig. S3 for more details).
As shown in Fig. 1 (f), the energy differences between different stacking configurations are on the order of 0.1 meV/atom, which is beyond the chemical accuracy of DFT calculations. Accordingly, these stacking configurations are nearly degenerate ground-state structures. In the Berry phase formulation of macroscopic polarization, the calculated vertical polarizations of polar structures AB’A’ and AB’C’ are 0.22 and 0.25 pC/m, respectively. This is on the same order of magnitude as the out-of-plane polarization of tetralayer graphene [24], InSe [8], and many others [20]. The ABA and AB’A structures exhibit in-plane polarization of 0.17 pC/m and 0.18 pC/m, respectively, but no out-of-plane polarization. DFT calculations reveal trilayer qHP C60 with different stacking configurations have a band gap of around 0.8 eV (as shown in Supplemental Fig S4). Compared to metal or semi-metallic ferroelectrics [24, 27, 36], the finite band gap and OP polarizations of trilayer qHP C60 make it easier to measure the electric polarization in experiments and more suitable for diverse applications.
SHG response is a well-established technique for identifying layer stacking of 2D materials with variations of centrosymmetric structures [33, 37]. Figure 2 (a) shows the calculated frequency-dependent second-order nonlinear susceptibility tensor component \(\:{\chi\:}_{yyy}^{\left(2\right)}\) of AB’A’ and AB’C’ structures. The highest peak of \(\:{\chi\:}_{yyy}^{\left(2\right)}\) for the AB’C’ structure corresponds to a photon energy of 1.213 eV (equivalent to a wavelength of 1022 nm), while the highest peak for the AB’A’ structure is located at 1.329 eV (corresponding to 933 nm). The dependencies of other second-order nonlinear susceptibility tensor components are presented in Supplemental Material Fig. S5. Under an incident light with a wavelength of 1022 nm and an incident angle of \(\:{\theta\:}\) = 0°, the SHG responses of two trilayer structures are plotted as a functions of polarization angle \(\:\varphi\:\) in Fig. 2 (b) and (c) [38, 39]. The maximum parallel components \(\:{\varvec{I}}_{\parallel\:}\left(\varphi\:\right)\) of both AB’A’ and AB’C’ configurations appear at \(\:\varphi\:=90^\circ\:\), but with a substantial difference in their magnitude. Additionally, the vertical component \(\:{\varvec{I}}_{\perp\:}\left(\varphi\:\right)\) of AB’A’ peaks at \(\:\varphi\:=\:0^\circ\:,\) with its maximum value comparable to the parallel component. In contrast, the vertical component intensity of AB’C’ is only one-twentieth of its parallel component. Clearly, SHG responses can serve as a useful tool for distinguishing trilayer qHP C60 stacking configurations.
Similar to other sliding ferroelectrics [30–32], all stacking configurations are related by the relatively translational displacements of C60 layers, i.e., interlayer sliding. For example, the displacement of the middle layer by \(\:(\varvec{a}+\varvec{b})/2\) directly transforms AB’A’ into ABA’. One should note that ABA’ is equivalent to the A’B’A structure and has an out-of-plane polarization opposite to that of AB’A’. Consequently, the polarization reversing process is achieved under the process depicted in Fig. 3 (b). The calculated energy barrier required to achieve this transition is about 7 meV/atom, which is on the same order-of-magnitude as the sliding energy barrier of tetra-layer graphene (about 4 meV/unit cell) [24]. Moreover, the symmetry of the AB’A’ structure suggests that displacement of the middle layer by \(\:(-\varvec{a}+\varvec{b})/2\), \(\:(\varvec{a}-\varvec{b})/2\), \(\:-(\varvec{a}+\varvec{b})/2\) results in the same structure as that of \(\:(\varvec{a}+\varvec{b})/2\). This can be illustrated by the energy landscape as a function of different translational movements of the middle layer (see Supplemental Fig. S6).
Notably, the energy barrier can be reduced when the polarization reversing process goes through intermediate states with stable stacking configurations. For example, Fig. 3 (b) also schematically illustrates the transition process AB’A’⇢ABA⇢ABA’ (= A’B’A). One may note that the intermediate state ABA is non-polar and equivalent to A’B’A’. This process may be understood by first sliding the top layer C60 by \(\:(\varvec{a}+\varvec{b})/2\), followed by sliding the bottom layer C60 by \(\:(\varvec{a}+\varvec{b})/2\). Similarly, another transition path AB’A’⇢;AB’A⇢ABA’ (= A’B’A) is accomplished by first shifting the bottom layer and then the top layer. The energy barriers for these two transition paths are around 3 meV/atom.
While the AB’A’ structure resembles the Bernal stacking order of graphene, the AB’C’ structure resembles the rhombohedral stacking configuration [25]. Consequently, the polarization reversal process of AB’C’ is distinct from that of AB’A’ and can be achieved by sliding the bottom layer and middle layer by \(\:\varvec{a}/2+\varvec{b}/3\) and \(\:\varvec{b}/3\), as shown in Fig. 3 (d). This transition path leads to the final structure ACB’, which is equivalent to C’B’A. Even though this transition path involves a simultaneous movement of two layers, the energy barrier is only around 2 meV/atom, which is lower than many reported 2D ferroelectrics [8, 40–42]. Similar to the AB’A’ structure, the polarization reversal path of AB’C’ can be achieved by passing through a meta-stable intermediate state, such as ACA. This can be accomplished by shifting the top layer by \(\:-\varvec{a}/2-\varvec{b}/2\), followed by sliding the bottom layer by \(\:\varvec{a}/2+\varvec{b}/2\) to get ACB’ (= C’B’A, which has a lower energy barrier of ~ 1 meV/atom.
In addition to the reversal of polarization, switching between different polarization states can also be effectively achieved by interlayer sliding [43]. We calculated the transition energy barriers between four stacking configurations with different out-of-plane and in-plane polarizations, as shown in Fig. 4. Overall, the energy barriers required for transitioning between different polarization states are close to those required to reverse the direction of polarization. For instance, the energy barrier for transitions AB’C’⇢;AB’A’, AB’A’⇢AC’A’, and AC’A’⇢AC’B’ are 0.8, 1.7, and 0.8 meV/atom, respectively.
Notably, the AB’A’ and AC’A’ configurations are physically equivalent, since they are related by a rotation around c-axis by \(\:{180}^{\circ\:}\). Similarly, AB’C’ and AC’B’ structures are physically equivalent. As a result, AB’A’ and AC’A’ structures (or AB’C’ and AC’B’ structures) have the same OP polarization but opposite IP polarizations. In other words, the out-of-plane and in-plane polarizations can be reversed independently in trilayer qHP C60 systems. This property is different from monolayer In2Se3 [9] and many others [20], whose OP polarization direction is coupled with the IP polarization direction. Markedly, the decoupling of IP and OP polarization can be a key feature that allows one to construct Van der Waals structures with ferroelectrically switchable chirality.
It is noteworthy that intrinsic chirality in 2D Van der Waals materials is uncommon and have been pursued [44]. Chirality in low-dimension materials can be artificially introduced by twisting [45, 46], kirigami [47, 48], adsorption of chiral molecules [49, 50], and other methods [51, 52], which are challenging to manipulate and do not allow for nonvolatile switching of chirality through electric fields. To demonstrate the concept of ferroelectrically switchable chirality, we consider two slabs of AB’A’ trilayers that are rotated by \(\:\theta\:={90}^{\circ\:}\) with each other, as shown in Fig. 5. In practice, the rotation angle \(\:\theta\:\) can take any value ranging between \(\:0\) to \(\:{180}^{\circ\:}\). As shown in Fig. 5, the top A’B’A’ trilayer has an OP polarization \(\:{\mathbf{P}}_{OP1}\) pointing along the \(\:z\)-direction and an IP polarization \(\:{\mathbf{P}}_{IP1}\) pointing along the \(\:x\)-direction, while the bottom A’B’A’ trilayer has an OP polarization \(\:{\mathbf{P}}_{OP2}\) pointing along the \(\:z\)-direction and an IP polarization \(\:{\mathbf{P}}_{IP2}\) pointing along the \(\:y\)-direction. Such hexa-layer configuration has a total polarization given by \(\:({\mathbf{P}}_{IP1},{\mathbf{P}}_{IP2},{\mathbf{P}}_{OP1}+{\mathbf{P}}_{OP2})\), that forms a basis with right-handed chirality. The chirality can be readily switched by applying an out-of-plane electric field that reverse the OP polarization while keeping the IP polarization unchanged. This leads to a new configuration with polarization \(\:({\mathbf{P}}_{IP1},{\mathbf{P}}_{IP2},{-\mathbf{P}}_{OP1}-{\mathbf{P}}_{OP2})\), which has left-handed chirality. As we noted, the key for this ferroelectric chirality switching is to decouple the OP and IP polarizations, i.e., switching the OP polarization independently.