Compass, one of the most ancient artificial tools recorded both in Chinese and western historical materials12, has played an important role in the process of human development and exploration. In a normal compass, a magnetized pointer could orient itself roughly north-south showing the pole direction of geomagnetic field, which may be the earliest principle used to measure the distribution of electromagnetic field. Over decades, it is still challenging to achieve the detection of smaller-scale interactions between atoms and molecules under the guidance of such classical principles. The van der Waals (vdW) interactions, between the neutral and polarizable atoms, molecules or particles, are ubiquitous in various physical, chemical and biological processes across different scales18,19. However, the direct measurements of vdW force usually involved various sophisticated nanotechniques20-22 difficult to be extended to more complex multi-atoms systems. For example, the vdW potentials in nanoscale channels determine various physical and chemical behaviors of confined molecules, such as sorption, transport, catalysis and phase transition7-11, but they are still a great challenge to investigate at the single-molecule level with current methods. Analogous to the principle of traditional compass, we can introduce a rotating molecular pointer with the defined direction marks into the channels. And, the orientations of rotating pointers will reveal the spatial vdW potentials in the surrounding circumstances, if they can be precisely identified.
Directly observing the molecular configurations helps us better understand the interactions at the atomic scale1-6. And, in the compass-analogy strategy for detecting the vdW potentials, the single-molecule imaging of compass construction is also highly demanded to identify the exact pointer orientations. However, it still requires very advanced imaging techniques to ‘see’ the real single molecules, especially the small organic molecules, due to the its high instability and sensitivity to electrons. Recently, an emergent imaging mode, the integrated differential phase contrast (iDPC) scanning transmission electron microscopy (STEM)13-15, brings us new possibility to observe small organic molecules. The iDPC-STEM has been proved to be an efficient tool for imaging the beam-sensitive materials with ultra-high resolution and signal-to-noise ratio, and the contrasts of light elements are enhanced in the iDPC-STEM images16,17. These advantages allow the iDPC-STEM to be considered as an efficient tool to image and investigate the single-molecule behaviors in the real-space.
In this work, we establish a nanoscale vdW compass with single para-xylene (PX) molecule as a pointer confined in the straight channels of quasi-2D MFI-type frameworks23,24. According to the orientations of the rotating PX pointers, we can identify the vdW interactions between the PXs and surrounding Si/O atoms. And, using the Cs-corrected iDPC-STEM, we image not only the atomic structures of the MFI-type channels but also the single PX molecule in each channel with ultimate resolution. We find the imaged orientation of PX pointer is closely related to the channel geometry and the corresponding vdW potential field in each channel. Then, we observe the rotation of PX pointer in response to the channel deformations both in the spatial and temporal dimensions. These results provide a compass-inspired strategy to detect the vdW interactions under the confinement of nanochannels using a rotating pointer, which will help us investigate various single-molecule behaviors in the sorption, transport, catalysis and phase transition.
The MFI-type framework owns a cross-linked channel system consisting of straight and sinusoidal channels, which can be imaged from the  and  projections respectively (Fig. S1 and S2). Inspired by the reconstruction of traditional compass in Fig. 1a, we used the single PX molecule as a rotating pointer and the Si10/O10 ring of MFI straight channel as direction marks to construct our vdW compass in Fig. 1b and c. The elliptical Si10/O10 ring is numbered to indicate 20 discrete directions about 18° apart. And, such straight channel (with the size of ~5.6×5.3 Å) can just contain only one C6-ring (with the kinetic diameter of ~5.8 Å) as a size-matched pointer. Based on the previous adsorption studies in MFI-type zeolites25-31, monocyclic aromatics can be stably confined at the intersections of the straight and sinusoidal channels, and their long molecular axes are nearly parallel to the straight channel (b-axis of MFI framework). Then, the aromatic pointer will show an energetically preferred orientation due to the vdW interaction with surrounding Si/O atoms, and rotate with the change of channel geometry to detect the different vdW potential fields inside the channels as it does in a traditional compass.
The quasi-2D MFI-type zeolites we used in this work are derived from the short-b-axis ZSM-5 (a typical MFI-type zeolite) nanocrystals16 (Fig. S3). In Fig. 1d, the annular dark field (ADF) STEM image from the  projection shows the shape of short-b-axis ZSM-5 with a smooth (010) surface. After etched by the sodium carbonate (Na2CO3), the ZSM-5 nanocrystals own a hollow shape with very thin (quasi-2D) areas as shown in Fig. 1e and Fig. S4. The corresponding intensity profile in Fig. 1f indicates the different contrasts due to a wide range of thicknesses. The pristine ZSM-5 crystals (thick areas) are uniformly 30-50 nm thick, while the etched parts (thin area) are estimated to show a thickness less than 5 nm. From the lateral view (the  projection), the ADF-STEM image and profile analysis in Fig. 1g and h indicate the thickness of the thinnest areas in the etched ZSM-5 can be reduced to 2~3 nm, which is nearly the thickness of single unit cell in the b-axis direction. And, in the magnified image (Fig. 1i), we obtain the lateral view of the quasi-2D ZSM-5 lattice, where each straight channel has complete intersection with only one sinusoidal channel as a stable adsorption site of PX molecule. These results confirm that the quasi-2D ZSM-5 area for the single-molecule imaging is successfully built by this etching method. Then, at this quasi-2D area, the iDPC-STEM helps us resolve the atomic structure of MFI-type framework from the  projection. As shown in Fig. 1j, the projected positions of Si atoms and the arrangement of empty straight channels can be clearly displayed, which are highly consistent with the inset structural model.
Then, PX molecules were adsorbed into the straight channels of ZSM-5 directly from the PX liquid phase as the pointers of vdW compass. And, we imaged an etched ZSM-5 crystal adsorbing PXs, the shape and thickness of which are shown in the ADF-STEM image and corresponding intensity profile in Fig. 2a and b. Fig. 2c and d are the iDPC-STEM image and the intensity profile of the thick area (marked by the blue frame), while Fig. 2e and f are those of the thin area (marked by the cyan frame). After the PX filling, we can clearly observe the oriented spindle-shaped spots in these channels compared to the empty ZSM-5 framework in Fig. 1j. And, their orientations (more obvious in the thick area) can be directly identified in the iDPC-STEM images, which are required for detecting the interactions in the vdW compass. In the profiles in Fig. 2d and f, we normalized the intensities from 0 to 1 to quantitatively compare these profiles in different images. And, the black dash line is the normalized profile of the empty channels, and the red line is the normalized profile of the PX-filling channels. The adsorbed PX molecules can be expressed by the small peaks in the valleys of empty channels.
Moreover, the intensity profiles of the thick and thin areas are a little different. In the thick area, the PX spots in channels represent the arrays of PX columns, not the single molecules. Thus, the number of PX molecules in each channel is naturally varied when the PX adsorption isn’t saturated. As we observed in Fig. 2d, the intensities of PX peaks, which are related to the numbers of PX molecules in channels, are varied as marked by a wide red band. However, in the intensity profile of the thin area in Fig. 2f, the intensities of PX peaks keep constant in these channels, since there will be only one intersection site for the PX adsorption within 2 nm thickness. Furthermore, we calculated the projected electrostatic potentials as the simulated image from the  projection (Fig. 2g). The simulation is conducted using the structural model with one PX molecule in single unit cell and based on the same parameters with the iDPC-STEM imaging. And, in Fig. 2f, the blue dash line is the normalized intensity profile of the simulated image. We find that the intensities of the PX peaks in the iDPC-STEM and simulated images are highly consistent. These results also confirmed, in another way, that the thickness of the etched ZSM-5 crystal was reduced to about 2 nm and we have imaged the single PX molecule in each channel.
To further study the relation between the PX orientations and host-guest interactions, we detailedly checked the magnified iDPC-STEM images of individual channels with different PX orientations. Fig. 3a shows the iDPC-STEM image of an empty ZSM-5 channel. After the PX filling, there is a spindle-shaped spot indicating a vertical PX molecule which appears in the dark channel as shown in Fig. 3b. The red arrow shows the orientation of PX molecule, which points to the O-2 direction as defined in Fig. 1c. In this image, we can identify the exact positions of ten Si atoms around the channel as marked by the blue dots. More accurately, we used the intensity profiles to measure the projected distances of opposite Si atoms in Fig. S5 and Table S1. Then, we fixed the positions of Si atoms (the distances of opposite Si atoms) and optimized the structure of single MFI unit cell. And, we calculated the energetically preferred configuration in this unit cell based on the density functional theory (DFT) method. Fig. 3c gives the channel geometry and PX configuration with the most optimal energy corresponding to Fig. 3b. The PX pointer in this vdW compass points out the vdW potential field from by the surrounding atoms along the O-2/O-7 direction.
Moreover, in Fig. 3d-f, the iDPC-STEM images show other three PX orientations in different MFI channels. We did the same image processing and structural optimization based on the channels in these images, and then, calculated the theoretical configurations of inside PX pointers as shown in Fig. S6. It is obvious that the channel geometry is totally different in each channel. And, these calculated results are all highly consistent with the imaging results, where the PX pointers point to the Si-1, Si-2 and O-3 in Fig. 3d-f respectively. Meanwhile, we also calculated the energies of PX molecules with different orientations in the channel models obtained from these four images. In each channel model, we rotated the PX molecules so that it pointed to the atoms O-1, Si-1, O-2, Si-2, O-3 and Si-3 respectively. The results are summarized in Fig. 3g, and the imaged orientation indeed shows the lowest energy as we expected. The energy difference between the adjacent PX orientations is about 0.04-0.12 eV. Based on the Boltzmann distribution laws, the most stable configuration occurs with about 5-100 times probability of the adjacent ones in the dwell time of probe scanning (16 μs), which, thus, can be finally reflected in the images. These calculations reveal that the imaged orientation of PX pointer is closely related to the channel geometry and the corresponding vdW potential field inside the channels. That is, we firstly noticed the aperiodicity in the ZSM-5 framework by the real-space iDPC-STEM imaging with the atomic resolution, which leads to the varied host-guest vdW interaction in each channel. Such different vdW interactions, especially their directions, can be directly detected by the imaged (most probable) orientations of every PX pointers analogous to the principle of traditional compass.
The above analysis of single iDPC-STEM image reveals the spatial distribution of vdW potential field in each channel on the 2D  projection. Next, we continuously took several images of the same region to study the varied orientations of PX pointer in temporal dimension. Fortunately, we clearly captured the single-molecule orientation changing in a same channel from these continuous images (Fig. 4a-f and Fig. S7). Due to the time-consuming probe scanning and image integration processes of iDPC-STEM, the time interval for these images is about 1-2 minutes. The changes in this time scale do not result from the thermal vibrations of the framework and molecules, but the shift of equilibrium configuration due to the changing channel geometry. As shown in Fig. S8, the profile analysis reveals that this channel was clearly deformed when continuously imaged. Based on the mechanism of electron beam damagexx, the framework deformation may be induced by the locally accumulated charging and heating under a long-time irradiation. Then, in Fig. 4g, the red data points indicate the change of PX orientation, while the blue data points provide the distances between the opposite atoms Si-1 and Si-6 in these images, which are the most volatile data in Fig. S8. We find that the trends of these two sets of data points are highly consistent. That is, the PX orientation changes in response to the channel deformation and points out the resulting change of vdW potential field in this channel. These results accord with the conclusion about the relation between the channel geometry and pointer orientation we derived from Fig. 3. And, the imaged rotation of PX pointer further prove the availability of using compass-analogy strategy to real-time detect the time-variant channel geometry and corresponding vdW potential field inside.
In summary, the unique structure of a single-molecule vdW compass, consisting of a MFI-type straight channel and a PX pointer inside, can be directly resolved by the iDPC-STEM. Especially, the configuration of single PX molecule can be identified in the iDPC-STEM images to detect the host-guest vdW interactions in the MFI-type frameworks. That is, we directly ‘see’ the orientations of PX pointers in these channels, making the strategy based on the single-molecule vdW compass possible to be applied to the detection of molecular interactions in a confined space. For example, the PX pointer provides both the spatial distribution and the real-time change of the vdW potential field in each channel, proving the availability of this strategy in the PX/ZSM-5 system. Moreover, achieving the single-molecule imaging under various microscopes is always a milestone in the nanotechnology and molecular science. The iDPC-STEM help us real-space image single organic molecules in porous materials for the first time. And other organic molecules, such as homologues of aromatics, can also be confined and imaged in MFI and other size-matching porous frameworks. Thus, this work may provide a general method to investigate the host-guest interactions in a series of organic/inorganic systems by direct imaging, which have attracted wide research interests in the sorption and catalysis, however, difficult to be studied at the molecular level. Such progress on the imaging technique will promote the further study on the physical and chemical properties of guest molecules, and bring new insights into diverse single-molecule behaviors.