The composition and structure of the ubiquitous hydrocarbon contamination on van der Waals materials

The behavior of single layer van der Waals (vdW) materials is profoundly influenced by the immediate atomic environment at their surface, a prime example being the myriad of emergent properties in artificial heterostructures. Equally significant are adsorbates deposited onto their surface from ambient. While vdW interfaces are well understood, our knowledge regarding atmospheric contamination is severely limited. Here we show that the common ambient contamination on the surface of: graphene, graphite, hBN and MoS2 is composed of a self-organized molecular layer, which forms during a few days of ambient exposure. Using low-temperature STM measurements we image the atomic structure of this adlayer and in combination with infrared spectroscopy identify the contaminant molecules as normal alkanes with lengths of 20-26 carbon atoms. Through its ability to self-organize, the alkane layer displaces the manifold other airborne contaminant species, capping the surface of vdW materials and possibly dominating their interaction with the environment.

1. Please specify the temperature of the IR measurement, because it is related to the phase of the specimen (smectic or crystalline). Figure 2, the splitting of the scissoring mode at 1472 cm-1 is not attributed to the bilayer formation in ref. 40. On the contrary, it seems that the first layer in contact with the Ag(111) surface shows the split peaks. It may be attributed to the molecule-metal substrate interactions, which may not be the case for the HOPG surface. According to another paper of the same group (J. Phys. Chem. B 2000, 104, 7363-7369), split scissoring mode is attributed to the crystalline phase, not a double layer. 3. The STM contrast of normal alkane layer shown in Figure 3 seems to be well explained by the simulation. However, I wonder that the discrepancy between the distances obtained by the STM observation and DFT calculation in Figure 3 can be serious. The experimentally obtained intermolecular distances 0.42 nm and 0.57 nm are fairly larger than the simulated values of 0.378 nm, 0.485 nm, which is not a minor difference in my opinion. This difference makes the positional relation between the molecule and the substrate atoms different, and therefore, the real moleculesubstrate interactions may not be reproduced by the simulation. Moreover, the larger intermolecular distance may allow a more dramatic orientation changes from the edge-on, beyond a slight tilt shown in Figure 3e. Hence, a more detailed simulation or discussion is required concerning this matter, such as the possibility of the further orientation changes. Even if the superlattice is not reproduced for the molecular arrangement with all the molecules with the flaton orientation as shown in Fig. S11, it does not necessarily exclude the mixture of the edge-on, tilted-on, and flat-on orientation of the molecules.

As for the IR spectra shown in
Reviewer #3: Remarks to the Author: In the current paper, the authors have investigated the chemical composition of the airborne hydrocarbon contaminants on 2D materials using grazing angle IR, AFM, STM, ellipsometry and DFT calculation. They conclude that the contaminants are saturated alkane w/ 20-26 carbons that self-organize to form ordered "stripe" on the surface of 2D materials. The characterization and analysis are carefully conducted. The conclusion is supported by the experimental and DFT results. This is a high-quality work on a very important topic. While it has been recognized in the past decade that airborne hydrocarbon contamination significantly impacts the properties of 2D materials, it is still unclear what those contaminant are. Moreover, there has been very limited efforts on this topic. This reviewer suggest that the paper be accepted for publication after the following minor issues are addressed: 1. In grazing angle IR experiments, there should be another control: A hydrocarbon (e.g., SAMs) with functional groups (e.g., COO). This is to make sure the IR has the resolution to detect small amounts of functional groups.

#1 (line 78-80):
The authors write that the stripes are observable in [..] adhesion, deformation, dissipation. This seems to be AFM wording. Perhaps the authors could briefly mention that they mean the signal recorded in the given channel? Similar, a PF topography map could be termed differently?
We thank the reviewer for pointing out that the AFM jargon may be inadequate for a wide readership. The PF topography map can safely be referred to as just topography. We have modified the main text with a brief description of the quantities being measured by the various AFM modes.
Revised main text: "The stripes are observable in the PF imaging modes of: surface adhesion, deformation, dissipation and Young's modulus (for a description of these modes see Methods)." Revised Methods section: "In the PeakForce QNM mode, in each image pixel a complete force curve is measured and the surface adhesion, deformation, dissipation and Young's modulus measurement channels are extracted. These measurement channels are proportional to the following quantities: the maximum adhesion force between tip and sample, the maximum deformation of the sample surface at peak force, the energy dissipated during a force curve and a fit to the elastic response of the sample." #2 (line 112) The authors write that the molecules in adjacent stripes tend to be shifted by a half-molecule width. Which shift is meant? Perpendicular or parallel to the long axis of the molecule? If parallel is meant, then this is compared with a quantity in perpendicular direction?
We thank the reviewer for pointing out the need to explain this aspect in more detail. In the text, the perpendicular direction is meant. We have revised the text to make this clearer. New text in the revised manuscript: "The molecules in adjacent stripes tend to be shifted by a half-molecule width, perpendicular to the molecule axis. The crystal structure of the contaminant layer varies between centered rectangular (cell angle: 90°) and centered oblique (cell angle: 80-85°) regions (see FigS12)." Furthermore, we added a new figure to the Supplementary Information, FigS12 regarding the crystal structure of the airborne contaminant and the dotriacontane (C32) monolayers in order to make a clearer statement. For more details on the crystal structure please read below.
Regarding the homogenous C32 monolayer, it is much easier to pinpoint the exact crystal structure, whereas in the heterogeneous airborne layer we find slightly different arrangements in different regions. In their bulk form the normal alkanes form two kinds of crystal structures depending on their molecular parity: the even-numbered alkanes exhibit triclinic (cell angle 75-85°), while the oddnumbered alkanes exhibit primitive (or close to the melting point, face-centered) orthorhombic (cell angle: 90°) structure [1]. The molecular parity dependence is more pronounced for short n-alkane (<16 carbons) monolayers, as diffraction studies reveal an odd/even alternation such that the even-length n-alkanes form herringbone assemblies, while odd-numbered ones form rectangular lamella-molecular backbone structures. Our particular interest is of mid-length alkane monolayers where the odd/even alternations in morphology are also observed (for monolayers of C16-C21 molecules): the structures are of the centered rectangular (odd) or oblique (even) type [1]. The dependence on molecular-parity is still observable in longer alkanes; however, there is only a 4-5° difference from 90° in the case of even alkanes, which is often disregarded. However, there are reports in the literature for longer, even alkanes (C24 or C32) where the authors stated centered rectangular arrangements [2][3]. As can be seen in FigS12 a-b, in the airborne contaminant layer, thanks to the heterogenous length of the C20-C26 molecules, we could observe both centered rectangular and centered oblique regions. On the other hand, in the homogenous C32 monolayer we found only the centered oblique structure, where the cell angle was ~84°.

#3 (line 217) It is mentioned only in the method's section how the STM calculation is done. It would perhaps be good to mention that briefly in the main text as well?
We agree with the Referee, therefore we added further sentences related to STM simulation methods in the revised version of the manuscript. We added the following text to the manuscript: "we modelled the arrangement of an alkane by using ab initio density functional theory (DFT) calculations implemented in Vienna Ab initio Simulation Package (VASP) 43 . Based on the calculated electronic structures of the different arrangement of the alkanes we also simulated their STM images within the Tersoff-Hamann approximation 44 (see Methods)."

#4 (line 254) Same as above. A few words more than just ''ab-initio calculation'' might be helpful.
We have extended the revised version of the manuscript with a brief description of the calculations. See the new text above.

#5 (line 299) The minor quantitative differences appear rather major to me. I do not understand the explanation why these distances should differ that much. Why is there a choice in nearest-neighbor molecule distance? Shouldn't this distance be more or less known?
We thank the referee for pointing this out. The exact inter-molecule distance is uncertain, as we point out in more detail below. Due to the computationally expensive nature of the problem of reproducing our STM images, we only strive to model the qualitative features of our STM measurements, namely: -the appearance of brighter and darker molecules and the presence of a quasi-periodic superlattice thereof, -having two inter-molecule distances in the measurement (see Fig 3) Our DFT calculations reproduce both of these effects. Our findings are further supported by additional calculations with a larger alkanealkane distance. We have removed the word "minor" from the main text, describing the quantitative differences.
Details of the inter-molecule distance: Despite the great number of observations and the fundamental role of the n-alkane chains in the literature of the self-organized monolayers, the exact value of the molecule-molecule distance is still elusive both experimentally and theoretically. It is related to the fact that the delicate balance between the molecule-molecule, molecule-substrate and molecule-environment forces governs the exact structure of the molecule monolayer. Furthermore, the structure can depend on the temperature and the number of alkane layers as well.
From an experimental point of view, the molecule-molecule distances in 3D, bulk alkane crystals can be exactly known from numerous studies [1]. However, these values cannot be used as such, because the crystal structure of the adsorbed 1-4 layers of alkanes is distinct from the bulk counterpart [2]. The main advantage of the STM investigation is that the moleculemolecule distance of the adsorbed layers on the surface can be measured directly. However, the great majority of the STM examinations of the structure of alkanes on graphite are conducted on monolayers deposited from solution (from phenyloctane or n-decane) [3][4]. The comparison with these experiments is not straightforward in our case, because of the so called "solventeffect" [5]. The liquids above the deposited solute molecules imply different forces; the formed monolayer may exhibit slightly different packing in different solvents, which means that it is a distinct system to our substrate-vacuum interface [5][6]. It was shown that even in the low temperature STM images, the distribution of the molecule distances can reach the 0.2 nm value on a graphite surface [7]. In our own STM studies, the intermolecular distance is not directly visible, sinceas we point out in our manuscriptthe largest contribution to the STM image stems from the hydrogen atoms, the apparent height of which is determined by the possible rotation of the alkane chain and it's position relative to the graphite below. Thus, a direct measurement of the inter-molecule distance is difficult.
From the point of view of theoretical calculations, similar differences appear in the moleculemolecule distance due to the different calculation methods. For example, classical molecule dynamics calculations (MD) predict 3.8 Å molecule-molecule distance for the edge-on configuration on graphite ( Figure 2 in [8]), while DFT calculations define 3.6 Å (Yang et al. [9]). These differences probably arise from the different vdW treatment of the molecule-molecule and molecule-substrate interactions in the MD and DFT simulations. The cited DFT calculations also pointed out that the slope of the interaction energies is rather flat around the energetically most favorable molecule-molecule distance position, which means that there is only a slight energy difference between the structures with different molecular distances. In particular, for the edge-on (flat-on) orientation similar energies appear between 3.5-4 Å (4-4.5 Å) molecule-molecule distances. This wide region of the calculated values can explain the differences in the measured molecule-molecule distances highlighting the importance of the external conditions in the measurements. As we explained above, defining an exact molecule-molecule distance in general is not straightforward, neither experimentally nor theoretically. In our edge-on geometry we used 3.8 Å, because we found it the energetically best interchain distance in our reference DFT calculations, in good agreement with previous literature data [8]. In this case, the cell with the graphite substrate has 34.08 Å length, containing 9 alkanes (527 atoms together). According to the cited DFT paper [9], we can further increase the molecule-molecule distance in our calculation to approach the experimentally observed value. We have performed new DFT calculations with 4.0 Å molecule-molecule distance in a 68.2 Å supercell consist of 17 alkanes (1159 atoms). We observed qualitatively the same STM image features of the moiré pattern as in the case of the 3.8 Å distance, namely the appearance of brighter and darker molecules and the two inter-molecule distances (see Fig S11) in good agreement with the measurement. Therefore, our new results demonstrate the applicability of our previous conclusions for different molecule-molecule distance as well.

#6 (line 302) The flat-on layer does not describe the experimental data. So? A concluding sentence in the sense of …this is why we suggest the layer to be formed from edge-on molecules? Why is that? Wouldn't one expect the opposite?
The orientation of the alkane molecules (edge-on, flat-on or mixed) and their exact distance is a very peculiar and hard to solve problem. From the early observations of the alkane monolayers on graphite there are arguments on both sides [1][2], still the majority of the (room-temperature) experiments found equidistant alkane-alkane distance and flat-on orientation. However, a temperature driven phase transition is reported by Diama et al. [3] and Endo et al. [4], where edge-on orientation together with mixed phases appear at low-temperature conditions. In our paper, we restricted ourselves to the examination of the two distinct, simple cases, the solely flat-on and the solely edge-on arrangements, which can be handled within the DFT methodology. As we have shown, the features of the observed moiré patterns in our samples are better reproduced by the simulated STM images of the edge-on orientation rather than the flat-on orientation at low temperature, in agreement with previous experiments [4]. However, we cannot exclude the possibility of mixed phases in the samples, as these geometries are beyond our DFT computational limit, thus no simulated STM images are available for the comparison. We have extended the manuscript to point out this limitation of our calculation and mention the possibility of mixed phase in our samples. Addition to the main text is as follows: "However, from our DFT calculations we cannot rule out the presence of mixed phases in the molecule orientation, where the edge-on and flat-on configurations may be intermixed."

This paper describes that the identity of the common ambient contamination on the surface of vdW materials such as graphene, graphite, hBN and MoS2 is composed of a self-organized molecular layer of normal alkanes with lengths of 20-26 carbon atoms, by
using the low-temperature STM measurements and infrared spectroscopy. The relation between the contaminant layer and the anisotropic friction measured by AFM is also described. The main conclusion of the paper is well supported by the experimental results. Therefore, I would recommend this paper to be published after a minor revision concerning the following points.

Please specify the temperature of the IR measurement, because it is related to the phase of the specimen (smectic or crystalline).
This detail is clearly missing from the manuscript, we thank the reviewer for drawing our attention to this. The measurements were done at room temperature (~300 K) and the temperature is indeed a very important factor to understand the IR spectra and the structure of the samples. Vapor-deposited, homogenous monolayers of C24 or C32 are in smectic phase at room temperature. The crystalline to smectic transition occurs between 215-230 K, while the smectic to liquid transition occurs around 340 K for the C24 monolayers [1][2][3]. The respective values for the C32 monolayer are just around 10 K above those temperatures. We cannot find an exact value for the C32 monolayer's crystalline to smectic transition, while the smectic to liquid transition occurs at 350 K [1][2]. On the other hand, a slightly higher than monolayer coverage (1.15 monolayer), a denser package shifts these temperatures, resulting in a denser layer of C24 that is still in a crystalline phase at room temperature, up to 310-320 K [1]. We suggest that the same applies for the denser C32 layer as well. We have extended the Methods section to clarify the measurement temperature.  Figure 2, the splitting of the scissoring mode at 1472 cm-1 is not attributed to the bilayer formation in ref. 40. On the contrary, it seems that the first layer in contact with the Ag(111) surface shows the split peaks. It may be attributed to the molecule-metal substrate interactions, which may not be the case for the HOPG surface. According to another paper of the same group (J. Phys. Chem. B 2000, 104, 7363-7369), split scissoring mode is attributed to the crystalline phase, not a double layer.

As for the IR spectra shown in
We thank the reviewer for pointing out this clear mistake in our manuscript. Actually, we wanted to refer to J. Phys. Chem. B 2000, 104, 7363-7369 [1]. In fact, the mistakenly referred paper (Ref. 40 in the original manuscript, J. Phys. Chem. B 2000, 104, 7370-7376) says only a little about the scissoring mode, it is discussing the effect of molecule-metal interaction mainly on the higher frequency stretching modes. Furthermore, the splitting of the scissoring mode closes as the layer thickness increases. We apologize for the mistake and thank the reviewer for the suggestion.
From the correct paper [1], to which the reviewer drew our attention, we understood that the splitting of the scissoring mode of the deposited alkane layer is attributed to an emerging crystalline phase. However, in the experiments presented by Yamamoto et al. [1] the peak splitting occurs only in thicker layers (>8.8 Å) of C44, which they described as a sign of the presence of more than two (1.: flat-on, 2.: gauche and 3+: crystalline) layers of C44. In our understanding, the splitting of the scissoring mode means a crystalline phase and a thicker layer.
In the same paper Yamamoto et al. [1] describes "the excluded volume effect" near the completion of the first monolayer, where in the denser layer the molecules tilt-out from the flaton geometry and their molecular plane is perpendicular to the substrate, meaning a partly edgeon geometry. It has to be noted, that in this denser monolayer, no splitting of the scissoring mode was reported. It could be possible that our vapor-deposited C32 layer is not an exact monolayer, but that it contains some areas with multilayers, consisting of crystalline phases. However, the fact that we could make stable STM imaging on the samples, which could be hardly possible in thick layer, imply that the regions consisting of thicker layers cover a small part of the surface.
We modified our interpretation on the C32 layer's scissoring mode splitting in the light of the above-discussed two questions and added the correct reference. New text added to the manuscript: "… the peak of the scissoring mode shows a splitting, which could be due to the presence of thicker layers of dotriacontane, with partial coverage, which could be in a crystalline phase 38 ."  Figure 3 seems to be well explained by the simulation. However, I wonder that the discrepancy between the distances obtained by the STM observation and DFT calculation in Figure 3 can be serious. The experimentally obtained intermolecular distances 0.42 nm and 0.57 nm are fairly larger than the simulated values of 0.378 nm, 0.485 nm, which is not a minor difference in my opinion. This difference makes the positional relation between the molecule and the substrate atoms different, and therefore, the real molecule-substrate interactions may not be reproduced by the simulation. Moreover, the larger intermolecular distance may allow a more dramatic orientation changes from the edge-on, beyond a slight tilt shown in Figure 3e. Hence, a more detailed simulation or discussion is required concerning this matter, such as the possibility of the further orientation changes. Even if the superlattice is not reproduced for the molecular arrangement with all the molecules with the flat-on orientation as shown in Fig. S11, it does not necessarily exclude the mixture of the edgeon, tilted-on, and flat-on orientation of the molecules.

The STM contrast of normal alkane layer shown in
We agree with the Referee that the difference of the intermolecular distance between the experiments and the calculations also makes the positional relation between the molecule and the substrate atoms different in the two cases. Since the molecules deposited from air are clearly not homogenous (see length distribution in Fig. 3c of the main text) we can't fully capture their behavior within DFT calculations, because of the need for periodic supercells. As we discuss in the main text, the following effects, observed in the STM are also reproduced in our DFT calculations: 1. Having two inter-molecule distances in the measurement.
2. The appearance of brighter and darker molecules in the STM images, due to the differing molecular environment. These brighter molecules form a quasiperiodic superlattice. Both of these effects are captured by the DFT calculations of edge-on alkane molecules as a result of the differing molecular environments of the alkane chains on the graphite support. To further strengthen our observations, we have done calculations on much larger supercells with a larger inter-molecular distance (4.0 Å). These new calculations also reproduce the above two observations, as detailed below. We added the following text to the manuscript: "We have also checked that increasing the alkanealkane distance to 4 Å in our calculation does not change the qualitative agreement with our STM measurement. This enlarged distance also results in a modulation of the apparent height of the molecules, corresponding to the measured quasiperiodic superlattice and the measured two inter-molecule distances of the alkanes (see Supplementary Figure 11b)." Detailed discussion of the intermolecular distance: In our calculations, we applied 3.8 Å distance for the edge-on orientation, as we have found this distance as the energetically most stable configuration in our reference DFT calculations, similarly to ref [1]. It is also known from previous DFT calculations [2] that the total energy of the structure depends weakly on the molecular distance between alkanes. Namely, for the edgeon orientation geometries have roughly the same energy on the graphite substrate in the range of 3.5-4 Å molecular distance. This allows us to increase the intermolecular distance in our calculations, since these geometries are close to equilibrium and can be also relevant in experiments. With the help of our new calculations, we have investigated the effect of the intermolecular distance on the Moiré patterns formed on the graphite substrate as the Referee suggested. It is worth noting that the increased molecular distance also significantly increases the commensurate supercell of the alkane-graphite system, making these calculations computationally challenging. More specifically, the Moiré wavelength has a divergence at 4.26 Å, increasing the needed supercell impossibly large to simulate even at 4.1 Å. In accordance with this, in our new DFT calculations we were able to reach only a maximal molecular distance of 4.0 Å before exceeding our computational limits. The supercell length then became 68.2 Å with 17 alkane chains. This geometry results in just over one thousand atoms (1159) in the simulation. The results of the two different molecule distances (3.8 Å and 4.0 Å) allows us to compare the main features of the observed Moiré periodicities and to make qualitative predictions. The first important result of the new DFT calculation is the confirmation of the rotation of the alkane molecules, which is responsible for the geometrical effects in the observed Moiré patterns. In the case of the 4 Å molecular distance, we did not find evidence of more drastic rotation of the alkanes as was suggested by the Referee compared to the case of the previous 3.8 Å geometry. It is worth noting that our new calculation revealed a non-uniform rotation of the alkane molecules, namely that the molecules situated at a symmetric position over the graphite surface have much smaller rotation angle (see Figure below and FigS11b). Our results highlight the complex behavior of the rotation effect and show that the larger intermolecular distance does not result in increased rotation of the molecules. As a next step, we have produced the simulated STM image for this geometry (see Figure below and FigS11b). The main important observation is that the simulated STM image is in qualitative agreement with our previous calculation. Namely, the fine details of the moiré pattern for increased intermolecular distances (4.0 Å) also show that additional periodicities appear below the supercell period in a similar way as for shorter molecular distances (3.8 Å). We observed the variation of the apparent height of the molecules along the stripe in the newly executed edge-on simulations, where the combined geometrical and electronic effects results in 4-5 brighter molecules together with 2-3 molecules with lower contrast. This variability in the simulated STM images, where different periodicities appear below the supercell period, is in good agreement both with our previous calculations and the experimentally observed STM images. It is worth noting that in contrast to the edge-on orientation, the flat-on geometry does not show such complex variability of the observed periodicities in our STM image simulations (see Fig S11c). This qualitative difference supports the rather edge-on nature of our findings. We also quantified the observed periodicities by using the same FFT analysis. We found four peaks in the new geometry calculation at 0.4 nm, 0.52 nm, 1.7 nm and 3.4 nm values (see Figure). If we compare these values to our previous results (0.378 nm, 0.485 nm and 1.13 nm), where the molecule distance was 3.8 Å, we can recognize increased periodicities. This is not surprising for the first two values (0.378 nm, 0.485 nm), which correspond to geometrical effects (H atom positions). In those cases, the applied 4 Å molecular distance simply increases these periodicities, namely the molecular distance of the alkanes and the distance of the furthest hydrogens, respectively. The remaining two peaks (1.7 nm, 3.4 nm) also reflect electronic effects. This is in agreement with our previous findings, in which we revealed that the peaks with even larger values in the FFT spectrum originated from both geometrical and electronic structure effects. Our new calculation further confirms this observation, where periodicities with 1.7 nm and 3.4 nm values are obtained. These values are smaller than the supercell period and could be comparable with the experimentally observed 1D quasiperiodic patterns (of around 1.9 nm period). As we can see, the comparison of the two DFT calculations show a tendency in the observed periodicities, where the periodicities in the alkane-graphite system are shifting towards the larger values by increasing the molecular distance. Therefore, we can conclude that the distance between the alkane molecules changes the observed periodicities in the system, as the Referee mentioned. However, the main origins of these periodicities are the same as we have demonstrated by our additional DFT calculation. Therefore, we assume that further changes of the molecule distances (approaching the experimental values) will have the same qualitative behavior and will reproduce the observed periodicities of the measured STM topography images (0.42 nm, 0.57 nm and 1.95 nm). Finally, we agree with the Referee that our calculations on pristine edge-on and flat-on geometries cannot exclude the possibility of mixed phases in our samples. Although the pure edge-on geometries show nice agreement with the experiments, mixed phases can also appear in samples. Unfortunately, we are not able to compare theoretical calculations of the mixed phase with the experiments, because these mixed phases are beyond our DFT computational limit. We have extended the manuscript to point out this limitation of our calculation and mention the possibility of mixed phase. Addition to the main text is as follows: "However, from our DFT calculations we cannot rule out the presence of mixed phases in the molecule orientation, where the edge-on and flat-on configurations may be intermixed."