Structure and thermodynamics of supported lipid membranes on hydrophobic van der Waals surfaces

Understanding the adsorption and physical characteristics of supported lipid membranes is crucial for their effective use as model cell membranes. Their morphological and thermodynamic properties at the nanoscale have traditionally been studied on hydrophilic substrates, such as mica and silicon oxide, which have proved to facilitate the reconstruction of biomembranes. However, in more recent years, with the advent of the van der Waals crystals technology, two-dimensional crystals such as graphene have been proposed as potential substrates in biosensing devices. Membranes formed on these crystals are expected to behave differently owing to their intrinsic hydrophobicity, however thus far knowledge of their morphological and thermodynamic properties is lacking. Here we present a comprehensive nanoscale analysis of the adsorption of phosphatidylcholine lipid monolayers on two of the most commonly used van der Waals crystals, graphite and hexagonal boron nitride. Both morphological and thermodynamic properties of the lipid membranes were investigated using temperature-controlled atomic force microscopy. Our experiments show that the lipids adsorb onto the crystals, forming monolayers with their orientation dependent upon their concentration. Furthermore, we found that the hydrophobicity of van der Waals crystals determines a strong increase in the transition temperature of the lipid monolayer compared to that observed on hydrophilic substrates. These results are important for understanding the properties of lipid membranes at solid surfaces and extending their use to novel drug delivery and biosensing devices made of van der Waals crystals.

Supported lipid bilayers (SLB) are a fundamental and widely-used experimental pla orm in biophysics. 1,2By integra ng other biomolecular complexes, they allow for reconstruc on of model cell membranes and inves ga on of their structural and physical-chemical proper es in vitro.For example, they have been used to study transmembrane proteins, 3,4 and to visualise me-dependent processes such as protein-lipid and drug-lipid interac ons 5,6 , as well as molecular recogni on. 7Atomic force microscopy (AFM) based techniques have been extensively used to study SLBs for their ability to access morphological and physical proper es of the lipid membranes down to atomic scale resolu on and with precise control over the system's environment. 8,9High-resolu on AFM has also been deployed to resolve the morphological organisa on of lipids together with their water solva on structures. 10,11[14] SLBs spontaneously form through self-assembly on various flat substrates using either vesicle fusion methods 15,16 or the Langmuir-Blodge technique. 16,17[20] The lipid molecules in the bilayer form two adjacent leaflets where the head groups expose towards the water and the substrate -known as the distal and proximal leaflets, respec vely -whilst their hydrophobic tails are buried inside.2][23][24][25][26] In this case, lipids immersed in water form stable monolayers with the tails adjacent to the solid interface and the head groups poin ng away from the hydrophobic surface towards the water.It has been predicted that subsequent lipid bilayers might then form on top of the first interfacial monolayer. 23Being electrically conduc ve, 27 graphite/graphene substrates are an ideal pla orm to carry out bio-electrochemical experiments on lipid membranes 28 , o en used as a subs tu on for gold substrates which have been tradi onally employed for biosensing applica ons. 29,30Hence, with the recent advances in nanoscience and two-dimensional (2D) van der Waals (vdW) technology that allow the development of novel 2D nanosensors, it has become crucial to prepare, in a controlled and reproducible manner, lipid membranes on graphene and other vdW crystals.
A key feature of lipid membranes is their ability to exist in different thermodynamic phases, which crucially depends upon their chemical composi on and environmental condi ons.Addi onally they also undergo reversible phase transi ons, as characterised by their phase transi on temperatures. 31,32In par cular, the mel ng temperature, Tm, indicates the main phase transi on of lipids from a solid-ordered (So) phase, where the lipids are regularly packed with their tails extended, to a liquid-disordered (Ld) phase, where the lipids' tails compress and the lipids are more free to diffuse laterally, leading to a less ordered membrane.This results in an overall 'shrinking' of the lipid membrane thickness.Depending upon the model system the phase transi on can be studied by numerous techniques.][38][39][40][41][42][43][44] Importantly, SLBs have demonstrated different behaviour compared with lipid vesicles due to the presence of the solid support. 37,42Only recently has a clear understanding of the thermodynamic behaviour of the two lipid leaflets cons tu ng the bilayer been achieved. 35Whilst the phase transi on of vesicles occurs at a single temperature, two transi ons have been observed for SLBs on hydrophilic surfaces such as mica and silicon.The first occurs at a temperature similar to that found for vesicles and corresponds to the phase transi on of the distal leaflet.The second occurs at a higher temperature and is associated to the mel ng of the proximal leaflet -si ng adjacent to the solid surface.[46][47] Interes ngly, it has been suggested that by modula ng the environmental and prepara on condi ons of SLBs, one may couple or decouple the phase transi on of the two leaflets.38 However, to the best of our knowledge, no studies have been reported on the effect of hydrophobic substrates on the thermodynamics of supported lipids membranes.
Here, we u lised temperature-controlled amplitude modula on AFM to inves gate the morphological and thermodynamic proper es of two commonly used phospha dylcholine (PC) lipids (1,2-dimyristoyl-sn-glycero-3-phosphocholine, DMPC, and 1,2-dilauroyl-sn-glycero-3-phosphocholine, DLPC) deposited, via vesicle-fusion methods, on two hydrophobic vdW crystals: highly ordered pyroly c graphite (HOPG) and hexagonal boron nitride (h-BN).First, the adsorp on and structural arrangement of lipids on HOPG was inves gated.Secondly, we studied the effect of the temperature on fully formed lipid monolayers and subsequently determined the transi on temperature for both DMPC and DLPC monolayers on HOPG.We studied the nanoscale morphological structure of lipid monolayers below and above the transi on temperature.We contrasted these results with those obtained on hydrophilic supports by repea ng the experiments on mica and silicon oxide substrates, ul mately finding important differences in the transi on temperature.To understand the impact of graphite conduc vity on these findings, we analysed the case of h-BN, a similar vdW crystal to HOPG as it shares the same hydrophobic character and hexagonal la ce structure but importantly differs in its electrical proper es with it being electrically insula ng. 48We found a large increase of the transi on temperature of PC lipids on HOPG with respect to those obtained on lipids membranes on hydrophilic substrates and on h-BN, indica ng the important role of the substrate metallicity on the structure and phase of lipid membranes.

Lipid sample prepara on
Mul -lamellar liposomes were a ained following methods reported by Atwood et al., 15 in order to prepare samples via vesicle fusion for analysis using AFM.PC lipids, DMPC and DLPC, were bought from Avan Lipids and stored at -20 °C.Chloroform (anhydrous, ≥99%, Sigma-Aldrich) was added to both to make stock solu ons of 10 mg/ml, which were stored in amber vials to reduce oxida on.Under a nitrogen flow, the chloroform was evaporated whilst rota ng the vial to form an even film.The vial was le overnight to ensure the absence of any chloroform.The lipid film was re-hydrated using deionised (DI) water, of resis vity 18.2 MΩ (Millipore), to form concentrated lipid suspensions and was followed by 10 minutes of sonica on to remove any le over film from the vial.No further extrusion was carried out, however before producing the diluted suspensions used for deposi on, the concentrated suspension was sonicated for 10 more minutes.The liposome suspensions were stored away from light, at approximately 4-8 °C.
For experiments on hydrophilic surfaces, mica and p-doped silicon (with its na ve oxide), 0.2 mg/ml concentra on suspensions of DMPC-water were deposited on the surface and le to incubate at room temperature for 10 minutes before rinsing with DI water, ensuring a water droplet remained on the surface.Prior to deposi on, mica surfaces were freshly cleaved, and silicon chips were treated with a piranha solu on (9:1 sulphuric acid:hydrogen peroxide, Sigma-Aldrich) at 80 °C for two minutes before rinsing with DI water.Silicon chips were then sonicated with a solu on (5% in DI water) of Decon-90 (Decon Laboratories Ltd, UK).For hydrophobic surfaces, DMPC and DLPC suspensions of concentra ons ranging from 1 µg/ml to 0.1 mg/ml were used.Around 200 µl of the suspension was dropped onto freshly cleaved HOPG/hBN and was le to incubate for 30 minutes at 40 °C before rinsing with water, as with the hydrophilic surfaces.For both types of surfaces, the samples were immediately taken to the AFM for measurements ensuring that the sample remained hydrated.

AFM measurements
AFM measurements were carried out in water solu ons using a commercial AFM (Cypher ES, Asylum Research, Oxford Instruments) in amplitude modula on with photothermal excita on of the can lever.Gold-coated can levers (HQ NSC19/Cr-Au, µMasch) were used, with a typical spring constant of  1 N/m and resonance frequency of  35 kHz in liquid.The temperature-controlled sample stage allowed us to inves gate the temperature dependence of the supported lipid membranes between 15 °C and 70 °C, with a hea ng rate of 0.1 °C/s.The sample was le for 2-5 minutes to thermalise before reapproaching the AFM p to the surface and resuming imaging.Care was taken to re-align the scan at each temperature to ensure the same areas were imaged consecu vely.Furthermore, mul ple scans at each temperature were performed to ensure that the sample had thermalised.Post processing and analysis of the AFM images were done using WSxM and Gwyddion so ware.0][51][52] Can levers were calibrated by taking forcedistance curves towards the sample and the sta c and first mode op cal lever sensi vity, σ0 and σ1 (σ1 = 1.09σ0), were extracted from the slope of the deflec on (V)-piezo extension (nm) curve.

Growth and structure of DMPC lipid membranes
We started by inves ga ng the forma on of DMPC lipid membranes on HOPG.Ini ally, low concentra ons (1 µg/ml) of DMPC/DI suspensions were deposited on freshly cleaved HOPG surfaces, and me-lapse AFM images were taken to visualise the growth of the lipid membrane.For such low concentra ons, we found the presence of small patches at the beginning of the imaging that, under stable condi ons, con nuously grew up un l a satura on point.Figure 1a shows three consecu ve images of a representa ve patch taken at mes t = 0, 6 and 14 minutes, showing the structural evolu on of the patch.The area covered by the patch increases with me due to the con nued adsorp on of lipid molecules from the lipid suspension.Corresponding topography profiles (Figure 1b) reveal that the height of the lipid patch does not change with me and is  0.6 nm.This value is much lower than expected for a DMPC monolayer ( 2.2 nm) 29,30 , indica ng that at such low concentra ons the lipid molecules lie essen ally parallel to the HOPG surface.We found this configura on only for dilute vesicle suspensions which yielded small monolayer patches on the surface.Such a lted, flat-lying, conforma on has been previously reported for DMPC deposited on hydrophobic gold 29,30 and is also typically found for alkanes molecules on HOPG. 53As for the case of alkane molecules, this orienta on of DMPC molecules can be explained by the high-affinity of the lipid alkyl chains for the HOPG surface. 54xt, we increased the incuba on concentra on of DMPC vesicles.Using approximately two orders of magnitude higher concentra on ( 0.1 mg/ml), a full-supported lipid membrane on HOPG was formed, with few defects to reveal the bare HOPG substrate below.These defects allowed us to measure the thickness of lipid membrane with respect to the HOPG. Figure 2a shows a representa ve topography image of the fully formed DMPC monolayer on HOPG.The cross sec on taken across a defect shows that the thickness of the formed layer is 1.9 ± 0.1 nm, which was furthermore confirmed by performing force-distance curves upon the layer, yielding 2.2 ± 0.2 nm (see Figure S5 in ESI).The obtained thickness matches the value expected for the DMPC monolayer, indica ng that for membranes formed from higher concentra on suspensions, the DMPC molecules are oriented perpendicularly to the HOPG with the tails adjacent to the substrate.Again, this is consistent with results previously obtained on hydrophobic gold. 29,30Importantly, these measurements were performed at a temperature < 20 °C to ensure that the lipid layer was in its expected So phase.As a control of our experimental method, we carefully repeated the DMPC deposi on on hydrophilic surfaces of muscovite mica and silicon, which have been previously used as supports of lipid bilayers. 38In both cases, the AFM images (Fig. 2c and Fig. 2d) clearly indicate the forma on of a bilayer of thickness  5 nm in good agreement with expecta on and previous reports for DMPC bilayers (average height of 5.3 ± 0.5 nm and 5.0 ± 0.2 nm for mica and silicon respec vely). 55e thickness of the formed lipid membrane was not the only morphological difference between the case of HOPG and that of the hydrophilic surfaces.AFM images taken at higher resolu on [50][51][52] (ESI Fig. S4) revealed the existence of stripe-like domains that extended across the whole DMPC lipid monolayer surface, unlike the smooth surface seen for hydrophilic supports.
To inves gate this, we obtained images at various temperatures ranging from 21 °C up to 60 °C.The stripe-like domains, which displayed an average periodicity of  8 nm, were stable and did not change significantly with temperature.This rules out that these features are a result of the so-called ripple phase seen in SLBs, associated with alterna ng domains at temperatures near the main phase transi on of the DMPC bilayer. 39,41,56Instead, they are consistent with previous reports of stripe-like domains for lipids deposited on hydrophobic gold and HOPG supports, 24,29,30,57 sugges ng that on hydrophobic materials the lipid monolayer organises into hemimicellar structures.
To be er understand the impact of the surface hydrophobicity on the forma on of PC lipid membranes, we inves gated the behaviour of DMPC lipids on h-BN crystals, offering an alterna ve substrate to HOPG.h-BN is a similar van der Waals crystal to HOPG, sharing features such as moderate hydrophobicity and atomic flatness.Furthermore, h-BN is structurally a-like graphite, displaying a honeycomb atomic la ce, however possessing a boron−nitride pair instead of the double bonded carbon atoms, as found for HOPG.This results in h-BN being electrically insula ng, as opposed to HOPG which is a semi-metal, and thus the two surfaces display very different conduc ve proper es.As for HOPG, DMPC lipid membranes were formed on h-BN via vesicle fusion via the same procedure.
We obtained similar monolayer forma on, as for the case of HOPG (see Fig. 2b).Importantly, we note the existence of many more defects in the lipid monolayer on h-BN that revealed the bare h-BN surface, in contrast to the few present on HOPG.As before, from the defects we could directly measure the layer thickness (Fig. 2b).AFM cross sec ons yielded a membrane thickness of  2.2 nm, in agreement with values found on HOPG and again indica ng the forma on of a lipid monolayer.

Temperature-dependent behaviour of DMPC monolayers
Next we analysed the temperature-dependent behaviour of DMPC membranes on hydrophobic and hydrophilic surfaces using the established van't Hoff analysis.As previously men oned, a change in the lipid phase can be recognised by the different height of lipids domains in the So and Ld phases. 44The main phase transi on temperature, Tm, can then be obtained by fi ng the frac on of So/Ld domains (as a func on of temperature) with a sigmoidal func on, i.e. using the van't Hoff equa on. 32Wri ng s and l as the frac onal occupancy of the So and Ld domains, respec vely, we can define the equilibrium constant,  = , of the phase transi on in terms of the temperature as where ΔHvH and R are the van't Hoff enthalpy and the gas constant, respec vely.If we express s as a func on of T, we then obtain which is a sigmoidal func on that can be used to describe the behaviour of the frac onal occupancy of the So lipid domains. 38,41First, we verified this procedure on DMPC bilayers on hydrophilic surfaces, analysing the changes in the AFM images taken at increasing temperature.DMPC lipids have an expected transi on temperature of approximately 23 °C, as determined from liposome suspensions by DSC experiments. 58igure 3c shows the frac onal occupancy of the lipid bilayers in the So phase as a func on of temperature as obtained on mica (green symbols) and silicon (blue symbols) substrates (see also Fig. S1 and Fig. S2 in ESI).41,59 Specifically, we found the Tm of the distal and proximal layer to be 22.2 ± 0.1 °C and 33.8 ± 0.1 °C on mica, and 19.5 ± 0.1 °C and 30.7 ± 0.1 °C on silicon, respec vely.Notably, for both substrates, the transi on temperature of the distal layer is consistent with those found in DSC experiments ( 23 °C), 58 as expected.
Next, we looked at the temperature-dependent proper es of the DMPC monolayer on HOPG.As for the hydrophilic surfaces, we performed AFM topography images at increasing temperatures.
In contrast with the previous experiments on hydrophilic substrates, we did not observe the appearance of domains with changing temperature but rather an overall change in the thickness of the monolayer as measured via the defects.From this, we were able to determine the transi on temperature (see ESI S2 sec on for further details).Given that only a DMPC monolayer is present on the HOPG substrate, a temperature dependence similar to the proximal layer of DMPC bilayers on hydrophilic substrates was expected, i.e. a phase transi on temperature occurring at higher temperatures than found in DSC measurements.Indeed, we found that the transi on did not occur up to approximately 40 °C.Figure 3a shows topography images and height profiles at temperatures 40 °C, 50 °C and 60 °C.At 40 °C and 50 °C, the average thickness of the monolayer was found to be 1.9 ± 0.1 nm and 1.8 ± 0.1 nm, respec vely, remaining prac cally constant to the ini al thickness measured at 20 °C.Upon further increasing the temperature to 60 °C, the membrane thickness decreased to 1.4 ± 0.1 nm, which is close to the value reported for a DMPC monolayer in the Ld state. 60his implies the phase change occurs at around 50 °C, much higher than expected.Figure 3c shows the observed height of the lipid monolayer as a func on of the temperature (red symbols) and the fi ng with a sigmoidal func on (red solid line), which yields Tm = 52.6 ± 0.1 °C.This is approximately 20 °C higher than the value we found for the proximal leaflet on mica and silicon and 30 °C higher than that found for the distal leaflet.
As for the morphology, we compared the behaviour of the DMPC monolayer on HOPG with that on h-BN by repea ng the temperature-dependent experiments (Figure 3b).Again, from the forma on of domains with higher/lower height, we could recover the frac onal occupancy of the lipid So phase (as shown for the SLBs on hydrophilic substrates).Surprisingly, the transi on temperature on h-BN is clearly smaller than the one obtained for HOPG (Fig. 3c, pink symbols).By fi ng the data to Eq. 2, we found Tm = 36.7 ± 0.1 °C, which is close to the value obtained for the Tm of DMPC proximal leaflet on hydrophilic materials.

Temperature-dependent behaviour of DLPC monolayers
To verify whether the anomalous transi on temperatures observed on hydrophobic surfaces for DMPC also occurs for other PC lipids, we proceeded to repeat the previous experiments with DLPC.DLPC was chosen as its expected Tm from So to Ld phase is at approximately -1 °C. 58This is much lower than the Tm for DMPC and therefore helps to avoid experimental difficul es at high temperatures, in par cular the evapora on of the water solu on during the experiment.Following the same protocol as for DMPC, full DLPC membranes were formed on HOPG and h-BN.The thickness of these layers, determined via AFM topographic images and sta c force curves (see Fig. S5, S6, S7 in ESI), was found to be  1.1 nm, on both HOPG and h-BN indica ng the forma on of a lipid monolayer.As for the case of DMPC, we found that DLPC monolayers presented stripe-like nano-domains on both HOPG and hBN with a periodicity of  5 nm (see ESI Fig S4 and S7).AFM images of the DLPC monolayer on HOPG were recorded as a func on of the temperature between 15 °C and 45 °C. Figure 4a shows that at 15 °C, the lipid monolayer was characterised by the high density of domains approximately 0.3 nm higher than the adjacent lipids (also shown is a terrace of the graphite substrate).Such domains correspond to the So phase and were found over the whole surface.The frac on of the membrane consis ng of higher domains changes with temperature, as it decreases at higher temperatures.At 45 °C they almost disappeared, indica ng that the monolayer transi oned to its Ld phase.The temperature was cycled, cooling and hea ng the sample mul ple mes whilst taking AFM images, showing the reversibility of the process (see the full temperature cycle in Fig. S8 in ESI), as previously shown for lipid bilayers on hydrophilic substrates. 37,43Moreover, analysis for both hea ng and cooling yielded very similar results.By plo ng the frac onal occupancy of the So domains with respect to the total area against the temperature and the fi ng the data to Eq. 2 (Fig. 4c, red), we obtained the transi on temperature of DLPC to be Tm = 30.9°C, around 30 °C higher than expected from DSC. 58 This confirmed our previous results obtained on DMPC and suggests that the transi on temperature of PC lipids on HOPG substrates is shi ed upwards by approximately 20-30 °C.It is also interes ng to note that, as for DMPC, similar ripple structures were found on both the Ld and So phase (see ESI, Figure S4), again sugges ng these stripe-like domains are not the ones commonly associated with the phase transi on of lipids but rather a morphological feature characterising lipids on these substrates.
Next, we studied the temperature-dependent behaviour of DLPC on h-BN.Assuming a similar behaviour as on HOPG, the temperature was decreased down star ng from 40° C.However, no significant structural changes were observed down to 20 °C.At 10 °C, domains of a higher thickness in the So phase were observed approximately occupying half of the imaged area.Due to experimental limita ons in our setup, temperaturedependent AFM measurements further below 8 °C were not possible.Despite this, we could proceed with extrac ng the transi on temperature.Figure 4c shows the frac on of the So phase with respect to the total area as a func on of the temperature and the fi ng to Eq. 2, yielding Tm = 8.2 °C.This value is higher than the expected transi on from DSC, however, it is more than 20 °C lower than that observed on HOPG.
A summary of all the transi on temperatures of DLPC and DMPC obtained in this work is shown in Table 1, together with previously reported values obtained from DSC. 61 Table 1 Summary of the thermodynamic proper es of supported DMPC and DLPC lipid monolayers on HOPG and h-BN.Control data obtained on lipids bilayers on muscovite mica and silicon surfaces are also reported. 58iscussion and conclusions In this work, we used temperature-controlled AFM to study the effect of two hydrophobic vdW crystals, HOPG and h-BN, on the morphological and thermodynamic proper es of supported PC lipid membranes.We found that, on these substrates, lipid molecules organised themselves into monolayers, whilst on the hydrophilic substrates a lipid bilayer formed.The self-assembly of lipid molecules into a monolayer is likely to be driven by the strong interac on that hydrophobic vdW materials have with lipid alkyl tails.When using low lipid concentra ons, the lipids seemingly adsorbed with their long axis parallel to the surface.However, upon increasing the lipid concentra on, we obtained a rearrangement of the lipid layers, with the lipid molecules stacked perpendicular to the materials surface.In the la er case, the probed lipid thickness matched the predicted thickness for a single monolayer, confirming previous structural measurements. 55Moreover, at these higher concentra ons, temperature-independent stripe-like domains were present.We disassociate this from the commonly observed 'ripple-phase' of SLBs which is known to be due to the temperature-dependent compe ng phases of the lipid membrane. 39Rather, we argue that it is linked to the presence of the hydrophobic substrate, as already suggested in previous literature. 24,28,29 then characterised the phase transi on of the lipid monolayers as a func on of temperature.A general increase of the transi on temperature with respect that found by DSC experiments was expected due to the interac on between the substrate and the lipid membrane.This has previously been reported for the proximal layer of the lipid bilayers on hydrophilic surfaces.Indeed, in our control experiments on mica and silicon supports we observed a decoupled phase transi on of the distal and proximal leaflets, with the proximal leaflet showing higher transi on temperatures.However, on HOPG we found that the phase transi on shi ed substan ally by over 30 °C more than expected for both DMPC and DLPC.Previous reports have shown that the mel ng temperature of monolayers of linear alkyl molecules formed on graphite, which is qualita vely similar to the main phase transi on of lipids from their So to Ld phase, 62 is expected to shi to higher temperatures with respect to the bulk. 63,64Although this seems analogous to our experimental observa on, to our knowledge, this effect has never been reported for lipid membranes.
Our experiments on h-BN, which is both hydrophobic and atomically flat similarly to HOPG, shows that the transi on temperature of the DMPC lipid membrane shi s again to higher temperatures.However, by only a few degrees more than the proximal leaflet of DMPC SLBs.Addi onally, repea ng the experiments on h-BN with the DLPC lipid, we found a transi on temperature more than 20 °C less than that found for HOPG, confirming that PC lipids on h-BN have quite a different thermodynamic behaviour with respect to HOPG.To ra onalise our observa ons, we note that in our experiments we see a reduc on of defects in the monolayers on hydrophobic surfaces in comparison to SLBs on hydrophilic surfaces, sugges ng a more regular lipid packing on hydrophobic surfaces.This is shown in Figure 5. Addi onally, it is apparent that the frequency of defects in lipid monolayers on HOPG in comparison to h-BN is reduced.For the hydrophobic vdW surfaces we observed fewer defects, and we did not record cracks in the layer.This is par cularly relevant, since the main phase transi on of lipid membranes is known to start from defects present in the membrane (so called 'cracks' and holes), which arise due to packing irregulari es of the lipid molecules. 46,47This is concurrent with our results, presented in ESI Fig. S1,2, where the transi on can be seen to originate from the defects in the bilayer.Furthermore, higher density packing of lipids in the membranes has demonstrated an increase in the transi on temperature of the membrane, as for the case of the proximal leaflet on hydrophilic substrates. 38,59This can be further illustrated by the lower transi on temperature of the SLB on silicon in comparison to mica, rela ng an increased frequency of membrane defects of the DMPC SLB on the silicon surface, as shown in Fig. 5. Hence, we argue that on hydrophobic supports (i) a highly packed lipid layer and (ii) the absence of defects in the monolayer lead to an increase in Tm.In addi on, we observed that an even more highly ordered monolayer formed on HOPG than h-BN shi s the Tm to even higher temperatures.Further inves ga on is needed to understand the origin of this effect.However, as h-BN and HOPG surfaces share similar honey-comb atomic la ce and only differ in their conduc vity and surface charge, we speculate that this effect may originate from long-range forces, such as vdW forces, rather than steric constraints between the crystal la ce and the lipid alkyl chains.
In conclusion, we have studied the impact of hydrophobic vdW substrates on both the physical and thermodynamic proper es of reconstructed lipid membranes.Our findings improve our understanding of lipid membranes' proper es at solid surfaces and will be useful in various applica ons of lipid membranes, in par cular the development of novel bioelectric devices as well as experimental biosensing setups using vdW crystals as electrodes or supports.

Figure 1 .
Figure 1.(a) AFM topography images of the adsorp on of DMPC molecules on a HOPG substrate.Images were taken at consecu ve mes (0, 6, and 14 min) in a water a er deposi ng a droplet of DMPC/DI suspension at low concentra on ( 1 µg/ml).(b) Corresponding height profiles taken across the do ed line in the topography images.They show the height of the lipid patch to be around 0.6 nm and constant with me, indica ng a lted flat-lying orienta on of the lipid molecules.

Figure 2 .
Figure 2. From top to bo om: DMPC lipid membranes formed on (a) HOPG, (b) hBN, (c) muscovite mica, and (d) on silicon at temperature < 20 °C and corresponding cross sec ons.A DMPC monolayer with thickness  2 nm was formed on HOPG and hBN, whilst a bilayer was obtained on mica and silicon with thickness  5 nm.The height of the monolayer on the hydrophobic surfaces indicates that the lipid molecules are oriented perpendicularly to the surface with their tails towards the substrate.

Figure 3 .
Figure 3. (a) Series of AFM images in water of a DMPC monolayer on HOPG (concentra on 0.1mg/ml) at temperatures 40°C, 50°C, and 60°C, respec vely, and corresponding height profiles taken along a defect in the monolayer indicated by the dashed lines.(b) Same as (a), but on h-BN at different indicated temperatures.(c) The experimental frac onal occupancy of the lipid solid phase of DMPC monolayers on HOPG (red triangles) and h-BN (pink stars).For comparison, control data taken on DMPC bilayers on hydrophilic surfaces of mica (green squares) and silicon (blue circles) are also reported, showing both the distal and proximal leaflet transi on.Solid lines are fi ngs to Eq. 2.

Figure 4 .
Figure 4. (a) AFM images in water of DLPC monolayer on HOPG (concentra on 0.1mg/ml) at various indicated temperatures.Domains with a larger thickness appear as the lipid monolayer changes from Ld to So phase.As with DMPC, the transi on of DLPC on HOPG is much higher than that reported with DSC (~-1°C).(b) Same as (a) but on h-BN.(c) Frac onal occupancy of the solid phase as a func on of the temperature on HOPG (red triangles) and h-BN (pink stars).Solid lines are fi ngs of Eq. 2.

Figure 5 .
Figure 5. DMPC lipid layers on hydrophilic (mica and silicon) and hydrophobic surfaces (HOPG and hBN) in their So phase, showing presence of defects in the membranes.Examples of holes and cracks in the membrane are indicated by white circles and white arrows respec vely (note that cracks are only present in the SLBs, which are a result of the packing irregulari es of lipid molecules in SLBs).For hydrophilic surfaces, we no ced the presence of many holes and cracks.For the hydrophobic vdW surfaces we observed fewer defects, and we did not record cracks in the layer.

Figure S2 .
Figure S2.A complete DMPC supported lipid bilayer with small defects formed on clean silicon wafers using 0.2 mg/ml DMPC/DI suspension.(a) The temperature-dependent phase transition of the bilayer was tracked using AFM in liquid environment.Increasing the temperature led to a decoupled leaflet transition, with the top leaflet (lipid layer facing the aqueous solution) transitioning from the gel ordered to fluid disordered phase, as was seen on mica.(b) To analyse the transition temperature of each leaflet separately, the fraction of the gel phase vs temperature was plotted and fitted using van't Hoff analysis.This revealed a transition temperature of the upper/lower leaflet as 22 °C (295 K) and 32 °C (307 K),respectively.The transition temperature of the first leaflet is very close to the one extracted on mica and those found via DSC.The second transition is higher than the expected transition temperature due to the interaction between the substrate and the lipids.

S3
Figure S4.Morphology of zwitterionic lipid membranes formed on HOPG.The lipid monolayer shows stripe-like features, indicative of a hemimicellar conformation (a).Similar features were recorded for both DMPC (b) and DLPC (c).Importantly they were stable and did not change significantly with temperature.

Figure S5 .
Figure S5.Force distance curves (FDC) obtained on lipid membranes formed on HOPG for both DMPC and DLPC.The breakthrough event into the lipid monolayer allows to determine the thickness of the lipid membrane.

Figure S6 .
Figure S6.DLPC supported lipid membrane on HOPG.Topography map showing a defect in the lipid membrane from which a cross section was taken to determine the thickness of the membrane, which is consistent with the presence of a single monolayer adsorbed at the surface of HOPG.

Figure S7 .
Figure S7.(a) Topography map showing a fully formed DLPC supported lipid membrane on at T = 7 °C (gel phase).A defect in the membrane was used to determine a membrane thickness of ~1.3 nm.The thickness agrees with that found on HOPG.(b) Topography map showing that like on HOPG, stripe-like structures in the membrane form.The periodicity of the ripples was measured to be around 7.5 nm.