Structure of Normal-Alkanes Adsorbed on Hexagonal-Boron Nitride

Jan 7, 2014 - Diamond Light Source, Diamond House, Harwell Science and Innovation Campus, Chilton, Didcot OX11 0DE, United Kingdom. J. Phys. Chem...
1 downloads 0 Views 1MB Size
Article pubs.acs.org/JPCC

Structure of Normal-Alkanes Adsorbed on Hexagonal-Boron Nitride Thomas Arnold,* Matthew Forster, Achilles Athanasiou Fragkoulis, and Julia E. Parker Diamond Light Source, Diamond House, Harwell Science and Innovation Campus, Chilton, Didcot OX11 0DE, United Kingdom S Supporting Information *

ABSTRACT: We report the structures of submonolayer films of the homologous series of normal-alkanes hexane (C6H14) to hexadecane (C16H34) adsorbed on the surface of hexagonal-boron nitride (h-BN), as determined by X-ray powder diffraction. These n-alkanes are demonstrated to form solid monolayers and exhibit structures that are similar to those seen on graphite but with some important differences. Each n-alkane studied shows a fully commensurate structure, even at submonolayer coverage. Hexane and octane adopt structures in which the molecules are arranged in a “herringbone” pattern (2 × m√3abn for CnHn+2, where abn is the unit cell parameter of the underlying h-BN surface, n = 6 or 8 and m = (2n − 2)/2). Heptane, nonane, and undecane−hexadecane adopt structures in which the molecules within adjacent lamellae are arranged with their principal axis parallel to each other (√3 × mabn for CnHn+2, where n = 7, 9, 11−16 and m = n + 2). Decane is a transitional case with evidence of both structure types, depending on coverage and temperature. This transitional case occurs for shorter molecules on h-BN than on graphite, indicating a difference in the relative balance of intermolecular and molecule−surface interactions that has significant implications for the phase behavior of adsorbates on these two surfaces.



INTRODUCTION The structure and properties of physically adsorbed molecules has been of interest for many years1−4 and is central to our understanding of many important phenomena such as film growth, wetting, and two-dimensional melting. In particular, the normal-alkanes represent a particularly interesting subset of adsorbates since they are of great relevance industrially and provide a homologous series of subtly varying species that can be used to systematically investigate the balance between intermolecular and molecule−substrate interactions. In this study we obtained structures of submonolayer films of n-alkanes adsorbed on the surface of hexagonal-boron nitride (h-BN) powders by X-ray powder diffraction. We are interested in how the chemical composition and relative size of a substrate lattice can influence the structures and properties of molecular films adsorbed on them. As such we made a direct comparison of these data with similar measurements of the same n-alkanes adsorbed on graphite.5−8 Diffraction measurements of the short-chain n-alkanes adsorbed on graphite show an odd−even effect in the monolayer structure, which has also recently been observed by STM.9 Specifically these studies show that short “even” n-alkanes (i.e., CnHn+2 where n is even) adopt a “herringbone” structure, while longer even alkanes and all the “odd” n-alkanes (i.e., CnHn+2 where n is odd) show a structure in which the molecules within adjacent lamellae are arranged with their principal axis approximately parallel to each other. Many STM images of longer n-alkanes10 do not show herringbone structures for the n-alkanes and the transition between the herringbone and the parallel structures for even nalkanes appears to occur for chain lengths of 12 or 14 carbon © 2014 American Chemical Society

atoms. In fact, the precise location of the transition may depend on the exact experimental conditions. For example, neutron diffraction data showed the structure for dodecane on graphite was dependent on the surface coverage, while both herringbone and parallel structures have been observed for tetradecane depending on the technique used.7,11 Interestingly, a small phase transition is seen in calorimetry measurements from monolayers of dodecane immediately prior to melting.7,12 Although this transition has not been properly characterized it may be related to a transition between a herringbone and a parallel structure. The reason for the transition between structural symmetries of n-alkane monolayers on graphite may be due to the frustration between intermolecular forces and the surface potential. There is a small mismatch between the carbon bond lengths and angles in the fully saturated n-alkanes and in the delocalized graphite basal plane which means that, particularly for longer n-alkanes, it is not possible to sustain perfect registry with the underlying substrate. This manifests itself in some STM images13 as a Moiré pattern. The effect is less significant for shorter molecules, and so they seem to be able to remain commensurate while simultaneously satisfying the demands of intermolecular packing. There is considerably less data on physically adsorbed nalkanes on substrates other than graphite. A summary of such structures can be found in recent reviews,1,2 but in general, both Received: June 26, 2013 Revised: January 6, 2014 Published: January 7, 2014 2418

dx.doi.org/10.1021/jp4063059 | J. Phys. Chem. C 2014, 118, 2418−2428

The Journal of Physical Chemistry C

Article

in resolution. We also performed some characterization measurements on beamlines I07 and I22 also at Diamond Light Source. Characterization of h-BN Powders. All n-alkanes used in this study were purchased from Sigma-Aldrich (≥99.8%) and used without further purification. h-BN was obtained from Momentive Performance Materials Inc. Materials can be obtained in different grades with increasing surface area. The grade can have a significant influence on adsorption isotherm data in that the lower surface area powders showed better defined layering steps in volumetric adsorption isotherms.16 Our DSC results, however, showed much less sensitivity to the grade of powder,17 although the measured monolayer melting enthalpy was found to scale with surface area. In this study we used some lower grade powders with high surface area, specifically “HCP” and “HCPH”, since our initial attempts to use the lower surface area HCM grade were unsuccessful. A comparison of these powders and more detailed characterization is presented in the Supporting Information. In order to increase the number density of scatterers in the beam we compressed the powder into 3 mm diameter pellets using a standard pellet press (0.5 tons). All diffraction data of h-BN samples in this study have a significant small-angle-scattering contribution to the observed scattering. This is comparable to that seen for recompressed graphite powders (“Papyex” or “Grafoil”)30 and depends on factors such as the size of the crystallite particles, the porosity of the powder, and the extent of partial filling of these pores with hydrocarbon. Importantly, however, in order to obtain a clear diffraction pattern of the adsorbed monolayer, for which some Bragg peaks can occur at relatively low Q, it is preferable to subtract the signal from the underlying substrate, particularly the small-angle contribution. Unfortunately this process is complicated by small variations from sample to sample, so that a simple subtraction of a background pattern is never perfect. Even a very small variation in the scattering is relatively very large by comparison to the extremely small monolayer diffraction peaks of interest. This is particularly true for the small-angle region and around the bulk BN Bragg peaks but is also manifest in an uneven background scattering between the diffraction features. Since our powder diffraction data does not generally extend below Q ≈ 0.2 Å−1, we cannot perform a quantitative analysis of the small-angle scattering for each sample. Instead, as in previous studies6,7,14 we subtracted from each of the patterns an arbitrary function with Q−4- and Q−2-dependent terms in order to flatten the experimental diffraction pattern for subsequent analysis. It is important to note that the arbitrary nature of this subtraction does introduce some uncertainty into the measured intensity of the low-Q Bragg peaks. As such the intensity and peak shape of the peaks in this region are less reliable than at higher Q. Determination of Parameters Used in 2D Diffraction Data Analysis. The classic models used to calculate 2D peak profiles31−33 contain parameters that define the characteristic “sawtooth” peak shape. These parameters are sample dependent and reasonably well established for graphite but less so for h-BN powders. Therefore we examined the bulk diffraction patterns to independently determine some of the most important parameters. Specifically, we made estimates of the characteristic length (L) of the crystallite size and the preferential orientation of the crystallites in the powder (defined by two parameters, δ, the mosaic spread of crystallites

the herringbone and the parallel structures are observed with the parallel structure commonly preferred as the chain-length increases. Of particular interest is the case of magnesium oxide, which despite the considerable differences in symmetry and chemical composition shows remarkably similar commensurate structures for monolayers of butane14 to those seen for alkanes on graphite.6−9,15 Such a similarity might suggest the domination of intermolecular adsorbate interactions in these systems, yet the fact that for both substrates the structures are commensurate means that the molecule−substrate interaction plays an important role. In contrast to magnesium oxide, h-BN is a substrate that is very similar to graphite and as such provides a more direct comparison that should allow an understanding of how small changes in the substrate can influence the properties of the adsorbed monolayer. High-quality h-BN powders are suitable for both thermodynamic and diffraction measurements, and as such there have been a few studies of simple molecular adsorbates.16−25 In most of these studies, similar structures and phase behavior are seen on both substrates, although the results suggest a weaker corrugation of the surface potential for h-BN. For example, the tricritcal temperature of N2 adsorbed on BN (65.8 K) is some 20 K below the equivalent temperature on graphite (85.4 K).23 To our knowledge, there have been no studies of more complex molecules adsorbed on h-BN, until recently when we performed an extensive study using differential scanning calorimetry to investigate the n-alkanes adsorbed on h-BN powders.17 This study found that the n-alkanes (hexane and longer) show a very small peak about 5−8% higher than the bulk melting point of the n-alkane concerned. By comparison with similar behavior seen on graphite26−28 this peak was assigned to the melting transition of a stabilized solid monolayer adsorbed on the surface. As with N2 molecules mentioned above, stabilization on h-BN is less than for the equivalent monolayers on graphite, which melt about 10−14% above the bulk. In addition, this study also suggested that there may be a difference in the miscibility of coadsorbed n-alkanes. This difference, derived from a regular solution theory analysis of binary mixture phase diagrams, shows that for any given mixture of adsorbed n-alkanes the miscibility on h-BN is marginally higher than on graphite. In order to try to understand this behavior, we have now performed a full structural determination of the pure n-alkanes (hexane− hexadecane) at submonolayer coverage. This has been achieved using X-ray scattering alone since the neutron scattering and STM techniques discussed above have substantial experimental problems for studies on h-BN. Specifically, boron-10 is a very good neutron absorber, which means that isotopically enriched 11 BN must be synthesized (which is not straightforward) in order to use neutron scattering techniques. Similarly, techniques such as STM generally require an electrically conducting substrate, but h-BN is an insulator.



EXPERIMENTAL SECTION The bulk of the measurements presented here were measured on beamline I11 at Diamond Light Source29 at 12 keV (1.034 Å) using the Mythen detector. The position of this detector and the wavelength were calibrated against a silicon standard. Samples were held in 3.5 mm capillaries and cooled using a standard cryo-cooler to ∼100 K. The large size of the capillaries maximizes the number of surface scatterers in the relatively large beam (1 mm (v) × 2 mm (h)) at the cost of a reduction 2419

dx.doi.org/10.1021/jp4063059 | J. Phys. Chem. C 2014, 118, 2418−2428

The Journal of Physical Chemistry C

Article

Table 1. Comparison of the Parameters (L and H0/H1) Used To Obtain the Best Fit for Guassian, Lorentz, LorentzSquared Bragg-Rod Profiles (δ was fixed at 71.5°)a

and H0/H1, the ratio of random to oriented crystallite planes). We then considered how these parameters influence which form of the integrated Bragg rod intensity gives the best fit to our experimental data. We have not made corrections for the Debye−Waller factor since the patterns were all collected at relatively low temperature. Nor have we included a polarization correction, since the scattering plane of the measurements was perpendicular to the fully polarized synchrotron radiation source. We can make an estimate of L by examining the width of the bulk BN diffraction peaks. This can be done with the Scherrer equation34

characteristic length, L(110)/Å determined from bulk measurements on h-BN #7 determined for graphite sample (Papyex) Gaussian Lorentz Lorentz squared

δ/deg

H0/(H0 + H1)

H0/ H1

343

71.5

0.46

0.85

200−600

∼30

0.35

0.54

220 ± 50 700 ± 50 400 ± 50

71.5 71.5 71.5

0−0.5 0.5−1 0−0.5

0 1 0

a

L=

These are compared with the values obtained independently for the HCPH grade h-BN (see text) and those reported for Papyex in the literature.30,33

0.94λ B cos θ

where L is the dimension of the average crystal, λ is the wavelength, θ is the Bragg angle, and B is the full-width halfmaximum (fwhm) of the peak concerned (in radians), corrected for instrument resolution. The constant 0.94 shows a small variation in the literature, which is relatively insignificant in this context.30 We considered an analysis of the individual peak widths in the measured powder diffraction patterns from each of the h-BN samples. In this case the peak widths were fitted individually using a fundamental parameters approach35,36 in order to estimate the particle size in the direction of each reflection. Using fundamental parameters the features of the synchrotron beamline are used to give the instrumental contribution to peak widths and convolutions for the sample crystal size and strain are then included in order to extract the sample contribution to the peak broadening. As a highresolution instrument the instrumental contributions to the peak widths of Beamline I11 are negligibly small, so the peak width in the diffraction patterns can be mainly attributed to sample effects. Since adsorption occurs predominantly on the basal plane, we can use the (110) peak to estimate the size of the crystallites in this dimension, L(110). We determined this for each batch of powder used (see Supporting Information Table S1), the average of which is ∼350 Å. Preparation of exfoliated graphite substrates “Papyex” and “Grafoil” involves a recompression stage that is known to result in partial alignment of the graphite crystallites with the macroscopic surface of the material. This preferential orientation has been used to enhance the scattering in the 2D plane of the adsorbed monolayers.32 Although there is no similar step for the as-purchased h-BN powders, we compressed the powder into pellets for use in our measurements. This could result in preferential orientation of the crystallites, which we directly examined (using Beamline I07 at Diamond). The intensity of the (002) h-BN Bragg peak was monitored as a function of the azimuthal angle for a single pellet of h-BN. After making appropriate geometrical corrections for the absorption, the measured intensity is proportional to the number of crystallites scattering at a given azimuthal angle. From this we can, therefore, extract a distribution of preferentially oriented crystallites. The result is expressed as a function of the two parameters mentioned earlier, δ and H0/H1.32,33 The full-width half-maximum of the measured distribution is the mosaic spread, δ, and the proportion of the scattering that is isotropic can be directly related to H0/H1. Table 1 lists the values obtained by this method and compares them with values reported for Papyex.30,33 From this we can see that although there is some preferential orientation in the h-BN pellets it is relatively much weaker than seen for the recompressed

exfoliated graphite samples. In effect the h-BN powder is almost completely isotropic. For our data we considered three different forms for the integrated intensity of Bragg rods: Gaussian, Lorentz, and Lorentz squared.33 For each case it is possible to obtain a fit to the experimental peak shape by varying the parameters L and H0/H1 (though we kept δ fixed at 71.5°). To illustrate this the (1,1) and (3,1) peaks of undecane are shown in Figure 1,

Figure 1. Comparison of the calculated powder diffraction peak shape for Gaussian (red, L = 220 and H0/H1 = 0), Lorentzian (green, L = 700 and H0/H1 = 1), and Lorentzian-squared (blue, L = 400 and H0/ H1 = 0) Bragg-rod profiles. For each calculation δ = 71.5°, and parameters L and H0/H1 have been roughly optimized. Experimental data shows the (1,1) and (3,1) peaks for a monolayer of undecane (C11H24) on h-BN.

together with the best fits using each of Gaussian, Lorentz, and Lorentz squared for the Bragg rod intensity profile. Best-fit parameters are compared with the independently determined values in Table 1. Qualitatively, the best peak shape seems to be for the Lorentz profile, but in this case the L parameter is a relatively poor match to this parameter independently determined. The alternative Lorentz squared is probably the next best fit and does show a relatively good match for L and H0/H1. We therefore used this profile for all fits presented below and set δ = 71.5°, L = 400, and H0/H1 = 0.85. However, it is important to note that this analysis is not conclusive, especially if we consider the uncertainty introduced by the significant corrections made to the experimental diffraction patterns. 2420

dx.doi.org/10.1021/jp4063059 | J. Phys. Chem. C 2014, 118, 2418−2428

The Journal of Physical Chemistry C

Article

Sample Cleaning and Dosing. There are several approaches to cleaning h-BN powders discussed in the literature.16,37 Here we have chosen to simply wash the pellets with methanol for 6−12 h (to remove soluble borates from the sample) prior to heat treatment (at 450 °C) under vacuum (1 × 10−8 mbar) for approximately 24 h. Once cleaned the powders were stored in sealed bottles but handled in air during sample preparation. We verified the efficacy of this treatment by examining diffraction patterns from powders that have and have not been washed in methanol (see Figure 2). In other respects the

the uneven background levels seen in some of our diffraction data. As in previous studies on graphite, after cleaning the samples have been dosed with the relevant n-alkane and annealed at 50−80 °C. Several grams of pellets were used in order to add a measurable amount of n-alkane to give the desired surface coverage, which is assumed to be evenly distributed over the whole sample. Coverage (θ) was calculated based on the specific surface area of the powder (see Table S1, Supporting Information) and the known area per molecule of the commensurate n-alkanes on graphite.6,7 However, the error in our calculation of the surface coverage is likely to be relatively high due to evaporation losses and uneven spreading of the monolayer over the surface of a relatively large amount of powder. For all our samples we estimate the coverage to be between 0.6 and 0.9 monolayers, but due to the relatively large error in this estimate we have not generally quoted precise coverage values for the data presented. However, we can be confident that our estimates are not wildly inaccurate by noting that our assignment is corroborated by the melting points of the films, as we will see below.



RESULTS To calculate the 2D diffraction patterns shown here we used a python program written by Christopher Richardson, Adam Brewer, and Stuart M. Clarke.40 The trial structures calculated are based on symmetry and close-packing considerations,41,42 with reference to the n-alkane structures determined on graphite.5−9 Atomic positions within a molecule are fixed and to a first approximation based on the molecular bond lengths and angles found in bulk n-alkane crystal structures.43−48 As stated earlier, we fixed the coherence length, preferential orientation parameters, and peak shape using independently obtained values. This approach deliberately minimizes the number of parameters that can be modified when fitting the experimental patterns. This is done because there are very few diffraction peaks and the background subtractions mean that their intensities may not be reliable, particularly at low Q. As a result it is debatable whether it is meaningful to improve the fits by introducing more fit parameters. Instead, we infer that the proposed structures are correct by noting a consistent variation along the extensive homologous series of alkanes and by comparing the patterns with alkanes on graphite, where there is additional evidence from neutron scattering to support those structures. Since the weak van der Waals forces governing this behavior are similar on both substrates and because they show very similar DSC results,17 we have chosen to use the structures of n-alkanes on graphite as a starting point for our new models of n-alkanes on h-BN. We can readily calculate the ideal dimensions of a perfectly commensurate structure analogous to the observed structure on graphite. These fully commensurate cell parameters are shown in Tables 2 and 3. Below we show that in each case, using these unit cell parameters as a starting point, we have been able to get a good fit to the new X-ray data for n-alkanes on h-BN. In each case we made an estimate of the accuracy of the fit in terms of each of the parameters used for the calculated pattern. These rough estimates are based on varying the parameters until the calculated patterns can no longer be said to be a good fit to the experimental data. Although the Q range shown is limited to 0.2 < Q < 1.6 Å−1, which is below the first major Bragg peak from the h-BN substrate, we also measured data to much higher Q. However, this data does not show any obvious Bragg peaks that can be

Figure 2. Comparison of the scattering of h-BN that has (top) and has not (bottom) been washed with methanol prior to the vacuumcleaning treatment described in the text. For clarity, the small-angle scattering contribution has been arbitrarily subtracted. Feature at 1.05 Å−1 has a characteristic Warren peak shape, indicating the presence of some 2D crystal which fits well to a monolayer of boric acid (see text, black line shows this calculation). After washing with methanol this peak is substantially removed from the scattering. Broad feature at 1.5 Å−1 is not removed in this process.

powders were treated in the same way: compressed into pellets prior to washing and subsequently heat treated as described above. Notably the unwashed powder has a small Bragg peak at Q = 1.03 Å−1, while this peak is no longer obvious for the washed powder. This peak has the characteristic sawtooth peak shape discussed above. The diffraction pattern from such a 2D monolayer of boric acid can be calculated based on the relevant bulk cell parameters (a = b = 7.04, γ = 120°)38,39 and is also shown in Figure 2. The calculation shows a single prominent (0,1) Bragg peak and is in remarkable agreement with the observed data. This suggests a well-defined monolayer of boric acid-like material is adsorbed on the surface on the h-BN powder, which is mostly removed by the methanol washing treatment. In addition to the well-defined 2D peak, there is a broad feature at around Q = 1.5 Å−1. The origin of this is not known, though it does seem to be sample dependent. It may be related to the presence of small amounts of amorphous BN and other BN isomorphs (such as r-BN, w-BN, or c-BN) in the supplied powder. However, this feature does not vary with temperature and is present in the absence of adsorbate. Therefore, in order to account for imperfect subtractions, we arbitrarily removed a Gaussian fit to this feature in addition to the overall background subtraction. This has unfortunately resulted in some inconsistency between samples and is probably the main reason for 2421

dx.doi.org/10.1021/jp4063059 | J. Phys. Chem. C 2014, 118, 2418−2428

The Journal of Physical Chemistry C

Article

Table 2. Cell Parameters for n-Alkanes with the Parallel Structure, as Determined on h-BN and Graphite,6,7 with a Comparison to the Theoretical Perfect Commensurate Layersa fully commensurate (b × a)/Å substrate C7 (heptane)

C9 (nonane)

C11 (undecane)

C12 (dodecane)

C13 (tridecane)

C14 (tetradecane)

C15 (pentadecane)

C16 (hexadecane)

h-BN graphite(low cov) graphite(hi cov) h-BN h-BN graphite(low cov) graphite(hi cov) h-BN graphite(low cov) graphite(hi cov) h-BN graphite(low cov) graphite(hi cov) h-BN graphite(low cov) graphite(hi cov) h-BN graphite(low cov) graphite(hi cov) h-BN graphite(low cov) graphite(hi cov) h-BN graphite11

√3 × 9

4.34 × 22.54 4.26 × 22.14

√3 × 11

4.34 × 27.54 4.26 × 27.06

√3 × 13

4.34 × 32.55 4.26 × 31.98

√3 × 14

4.34 × 35.06 4.26 × 34.44

√3 × 15

4.34 × 37.56 4.26 × 36.90

√3 × 16

4.34 × 40.06 4.26 × 39.36

√3 × 17

4.34 × 42.57 4.26 × 41.82

√3 × 18

4.34 × 45.07 4.26 × 44.28

measured (b × a)/Å

area per molecule/Å2

4.316(±0.01) × 22.2(±0.2) 4.75 × 21.8 4.26 × 22.0 4.316(±0.015) × 27.2(±0.3) 4.340(±0.015) × 27.5(±0.3) 4.8 × 27.0 4.24 × 27.0 4.340(±0.015) × 32.6(±0.4) 4.8 × 32.0 4.3 × 32.0 4.316(±0.015) × 34.3(±0.3) 4.74 × 35.2b 4.5 × 34.6 4.316(±0.015) × 37.0(±0.5) 4.8 × 37.4 disordered × 38.0 4.316(±0.015) × 39.9(±0.5) c 4.45 × 39.0 4.316(±0.015) × 42.3(±0.6) 4.8 × 42.2 d10 4.305(±0.015) × 45(±0.7) 4.5 × >38d

47.91 52.17 47.29 58.70 59.68 63.45 57.24 70.74 76.8 68.8 74.02 83.42 77.85 79.85 80.41 ? 86.10 ? 86.78 91.28 101.28 ? 96.03 ?

Using a = 2.46 Å for graphite and a = 2.504 Å for BN, i.e., the literature value49,50 rather than the measured value of 2.5053 Å. For each structure γ = 90°, there are 2 molecules per unit cell and the principal molecular axis of the n-alkanes is parallel to the unit cell axis (±5°). bOnly observed shortly before melting. cOnly composite pattern observed. dAlthough the STM images report the molecular length, the equivalent of a unit cell parameter was not specifically reported. a

Table 3. Cell Parameters for n-Alkanes with Herringbone Symmetry, as Determined on h-BN and Graphite,7 with a Comparison to the Theoretical Perfect Commensurate Layersa fully commensurate (b × a)/Å substrate C6 (hexane)

C8 (octane)

C10 (decane) (HB)

C10 (decane) (||)

a

h-BN graphite(low cov) graphite(hi cov) h-BN h-BN graphite(low cov) graphite(hi cov) h-BN graphite(low cov) graphite(hi cov) h-BN h-BN graphite

2 × 4√3

5.01 × 17.35 4.92 × 17.04

2 × 5√3

5.01 × 21.69 4.92 × 21.30

2 × 6√3

√3 × 12

5.01 × 26.02 4.92 × 25.57 30.05 × 4.34 29.52 × 4.26

measured/Å

HB angle/deg

area per molecule/Å2

5.01(±0.02) × 17.15 (±0.1) 5.4 × 16.9 4.9 × 16.9 5.01(±0.03) × 21.4(±0.15) 5.02(±0.03) × 21.6(±0.15) 5.37 × 21.2 4.9 × 21.2 5.00(±0.03) × 25.6(±0.3) 5.45 × 25.5 5.4 × 25.5 4.325(±0.05) × 29.5(±0.5) 4.35(±0.01) × 30.2(±0.2) not observed

20 ± 5 25 25 25 ± 5 28 ± 3 27 25 26 ± 3 29 29 0(±5) 0(±5)

42.96 45.63 41.41 53.61 54.22 56.92 51.94 64.00 69.49 65.03 63.72 65.69

For each structure γ = 90° and there are 2 molecules per unit cell. As in Table 2, a = 2.46 for graphite and a = 2.504 for h-BN.

lengths and angles, but once determined these have been fixed for all subsequent fits. Unfortunately the insensitivity of X-rays to the position of hydrogen atoms means that we have very poor sensitivity to the rotation of n-alkane molecules about their principal axis. As such we cannot absolutely determine the symmetry of the structures since this strictly depends on the precise orientation of the molecules within the unit cells. Loosely, the symmetries of the structures used here are based on the Pgg “herringbone”

resolved above the artifacts generated by imperfect background subtraction. As a result we are determining the structures with very few diffraction peaks, sometimes as few as three. We therefore limited the number variables used in the models; specifically, these are the unit cell parameters and the “herringbone angle” (HB), which is the angle made between the unit cell and the principal axis of the n-alkane molecule. We also considered the best values of the intramolecular bond 2422

dx.doi.org/10.1021/jp4063059 | J. Phys. Chem. C 2014, 118, 2418−2428

The Journal of Physical Chemistry C

Article

and P2 “parallel” structures observed on graphite.6,7 However, to ensure that we have chosen the best fit to our experimental data, we surveyed a range of other possible structures. These structures included consideration of some simple oblique unit cells and substantially varying the herringbone angles. Within the constraints described above, in each case we could not find any molecular arrangements that reproduced the observed data better than the structures reported here. Our final analysis is to monitor the temperature variation of the diffraction data, since this can give information about thermal expansion and melting of the adsorbed film. The disappearance of Bragg peaks can be used as a guide to the loss of long-range order within the adsorbed film and is a good first approximation for the melting of the monolayer.51 We therefore monitored the peak height and width as a function of temperature for each of these n-alkanes. From this data we are able to confirm that for each n-alkane melting occurs at approximately 75−90% of the bulk melting temperature for that n-alkane. As stated earlier, this is a good indication that the coverage is indeed in the submonolayer regime. We will now discuss the structures and temperature variation of the n-alkanes adsorbed on h-BN in detail. Most of the observed structures have the parallel arrangement of molecules, and so we begin with these systems. We will then present data for the herringbone structures of hexane, octane, and the special case of decane. Examples of these two structure types are illustrated in Figure 3.

Figure 4. Experimental data and corresponding fits for CnHn+2, where n = 7, 9, 11, 12, 13, 14, 15, and 16 as indicated. These patterns were obtained between 100 and 180 K, and fit parameters are listed in Table 2.

Figure 3. Examples of the two structure types: the herringbone structure of hexane (left) and the parallel structure of heptane (right). In each case the unit cell is shown as a red box. Molecules are shown with their C−C zigzag in the plane of the substrate, but in fact our Xray data is not sensitive to this orientation. We also approximately illustrated the assumed commensurate relation to the underlying h-BN surface.

actually very similar to the h-BN distance of 2.504 Å. Assuming that the bond lengths remain at their bulk values of 1.53 Å, this would give a C−C−C bond angle of ∼110°, which is slightly smaller than seen in the bulk crystals (112−114°).43−48 Given that the perfect tetrahedral angle is 109.5°, it is not unreasonable to expect that interaction with a surface potential could perturb these bond angles. This gives a spacing along an n-alkane backbone of 2.507 Å, and thus, we see that the mismatch between the n-alkane chains, and the substrate spacing is much less on h-BN than on graphite. With reference to the experimental data, we examined how our calculated patterns vary on changing this bond angle. Figure 5 compares the calculated diffraction patterns for hexadecane molecules with C−C−C bond angles of 108°, 110°, 112°, and 114°, together with the corresponding experimental data. The main difference between these calculated patterns is the relative intensities of the (2,0), (4,0), and (1,1) peaks. Thus, we see that as the bond angle is reduced from 114° to 108° the intensity of the (2,0) and (4,0) peaks grows relative to that of (1,1). It is clear that our experimental data favors smaller angles over larger ones, perhaps with the best fit at 108°. However, given the small-angle background correction discussed earlier, we cannot reliably use this variation to precisely determine the internal bond angles of the molecule. We therefore fixed the angle to 110°, which both gives a reasonable fit to the data and is corroborated by other STM experiments on graphite.10,11 For

Heptane, Nonane, and Undecane−Hexadecane. Figure 4 shows the diffraction patterns and corresponding fits for the n-alkanes, CnHn+2, for n = 7, 9, 11, 12−16. This includes all of the odd n-alkanes studied and the even n-alkanes longer than decane. In each case the structures have 2 molecules per unit cell and are, within error, fully commensurate. Parameters for each calculation are listed in Table 2 and compared with the fully commensurate parameters and the equivalent data for nalkanes on graphite. Note that some of the experimental patterns show artifacts between the obvious 2D diffraction peaks that are generated by imperfect subtractions of the background and the arbitrary small-angle scattering and Gaussian features discussed above. These features may seem quite prominent when compared to the size of the diffraction peaks. However, because there is a very small amount of scattering material in the beam, the measured peaks are up to 10 times smaller than equivalent data on graphite substrates for example. As such background imperfections can seem correspondingly larger. STM images of n-alkanes on graphite show a separation between alternate carbon atoms of around 2.51 Å.10,11 This is notably longer than the graphite repeat distance of 2.46 Å but is 2423

dx.doi.org/10.1021/jp4063059 | J. Phys. Chem. C 2014, 118, 2418−2428

The Journal of Physical Chemistry C

Article

contributions of the decrease in coherence length with the increase of the Debye−Waller factor, particularly when there are uncertainties over the temperature of the sample and the background subtraction. Thus, although quantitative analysis of peak widths will always be limited, it is still possible to use the peak heights as a reliable indicator of the loss of long-range order within the 2D crystal, because these are independent of the precise functional form of the diffraction peaks. To illustrate the temperature variation of our diffraction data, we will discuss the cases of dodecane and heptane in detail (see Figures 7 and 8). First, we discuss the example of dodecane which is typical of most of our data obtained for the n-alkanes. Figure 7b shows a series of diffraction patterns for dodecane as the temperature was increased at a rate of approximately 5 K per minute, and the temperature reported is the average temperature during scans of approximately 1 min. There is some thermal lag between the recorded temperature and the sample, which we estimate to be about 5−10 K, based on a comparison of the phase transition temperatures in heating and cooling scans. Since these measurements were recorded upon heating, the sample temperature is expected to be no more than 10 K lower than that quoted. The dodecane monolayer has undoubtedly melted by 225 K, which is significantly below the bulk melting point (263.5 K52) and a good confirmation that the coverage is in the submonolayer regime. However, the loss of long-range order that is implied by the disappearance of Bragg peaks seems to occur at different rates for peaks at 0.37 and 1.47 Å−1. These peaks are indexed to (2,0) and (1,1), respectively. These observations are illustrated in Figure 8, which plots the peak heights and positions as a function of temperature. Figure 8a shows that for dodecane these two peaks clearly reduce in size at different rates. Notably, the (1,1) peak is the first to broaden at around 180−198 K, and in doing so it shifts to lower Q, indicating a thermal expansion of the unit cell (shown in Figure 8b). Meanwhile, the (2,0) peak persists to significantly higher temperature, finally disappearing between 216 and 225 K. This peak also shows a shift to lower Q over this temperature range. Data for the n-alkanes undecane−hexadecane is broadly consistent with the trends outlined for dodecane. The shortest n-alkanes, however, do not show such a gradual melting transition. Figures 7(a) and 8(a) also show comparable plots for heptane. In this case the melting transition occurs much more quickly, and we cannot detect any significant changes in the cell parameters nor broadening of the diffraction peaks. Melting occurs between 138 and 148 K, which is again well below the bulk melting temperature (182.6 K52). Such sharp transitions are only seen for the shortest n-alkanes studied (hexane and heptane). From octane and longer the melting begins to occur over a broader temperature range, while the discrepancy between a and b cell parameters becomes more obvious. In general, since this broadening of peaks seemingly occurs at different rates it is difficult to precisely determine a melting temperature. However, in each case the melting transition occurs between 75% and 95% of the bulk melting temperature. This variation may be related to differences in the true coverage of the individual samples concerned. Decane. The structure of decane shows a transition between herringbone and parallel structures that depends significantly on the surface coverage and temperature. For decane on graphite, similar behavior has not been reported for any coverage or temperature, with only the equivalent

Figure 5. Comparison of various calculations of the diffraction pattern for a hexadecane monolayer with the experimental data. Atomic positions used in these calculations are determined based on molecular C−C−C bond angles of 114° (green), 112° (pink), 110° (red), and 108° (blue). In each case bond lengths were fixed and all other fit parameters were not varied.

the sake of consistency, all fits presented here use this bond angle. Hexane and Octane. Both hexane and octane show the herringbone structure seen on graphite with cell parameters that are, within error, fully commensurate. The data and corresponding fits for these two n-alkanes are shown in Figure 6, and cell parameters are shown in Table 3. The herringbone

Figure 6. Experimental data and corresponding fits for CnHn+2, where n = 6, 8, and 10. Decane example shows two fits that correspond to herringbone (red) and parallel (blue) structures. These patterns were obtained between 100 and 150 K, and fit parameters are listed in Table 3.

angle can be defined reasonably well by matching the relative intensities of the (1,1) and (2,1) peaks. We found that the best fit for hexane occurs for a HB angle of 20 ± 5°. For octane we were able to measure different samples. Cell parameters for these two patterns vary slightly but remain close to fully commensurate in both cases. As with hexane, by comparing the relative intensity of the (2,1) and (3,1) peaks, we can estimate the herringbone angle to be 28 ± 5°. Temperature Variation. Before we consider the special case of decane, we pause to examine the variation of the structures so far reported with temperature up to melting of the 2D crystals. In this regime, the Debye−Waller factor is significant and a substantial complication to analysis of the diffraction peaks. It is not straightforward to separate the 2424

dx.doi.org/10.1021/jp4063059 | J. Phys. Chem. C 2014, 118, 2418−2428

The Journal of Physical Chemistry C

Article

Figure 7. Temperature variation during a heating scan of the principal diffraction peaks in a monolayer of (a) heptane, (b) dodecane, and (c) decane (θ ≈ 0.9). Intensities of these peaks are plotted in Figures 8 and 9.

We measured decane nominally in two coverage regimes, at θ ≈ 0.6−0.8 and 0.9. Although the accuracy of these values is unknown, this assignment is supported by the melting points of the layers. The higher coverage pattern fits well to a commensurate parallel structure, while the lower coverage data shows a coexistence between a similar parallel structure (with slightly different cell parameters) and a commensurate herringbone structure. To understand this more complex behavior we examined the temperature variation of the diffraction patterns in some detail. The θ ≈ 0.9 data shows behavior that is similar to the other n-alkanes already discussed above (see Figures 7(c) and 9). Thus, the (2,0) and (1,1) peaks are seen to broaden at different temperatures, while there is a small expansion of the unit cell as the temperature is increased. In this case the expansion is relatively small, ∼1−2% (see Figure 8(c)). The melting point occurs between 200 and 219 K, which is around 80−90% of the bulk melting point (243.3 K52). Importantly the structure clearly remains with the parallel arrangement of molecules right up to the melting transition. However, the nominally lower coverage pattern at θ ≈ 0.6− 0.8 shows some complex phase behavior. In this case, melting occurs somewhere between 185 and 195 K, which is approximately 75−80% of the bulk melting point and significantly lower than for the higher coverage sample. This difference in melting point supports our estimates of the difference in coverage, with the higher coverage sample closer to monolayer completion but not above. For the lower coverage sample we see some significant temperature-dependent structural changes prior to melting, as seen in Figure 9, which shows both heating and cooling scans. As above the heating scans were measured at a rate of about 5 K/min, while the cooling scan was uncontrolled but occurred at 4 K/min. We used the observed discrepancy between the temperatures of the phase transitions for the heating and cooling scans to estimate the accuracy of the recorded temperature. We can therefore be confident that the true sample temperature was no more than 15 K above the quoted temperature. First, we consider cooling shown in Figure 9(a). The first structural peaks appear at Q ≈ 0.42 and ∼1.4 Å−1. These peaks index well with a parallel structure with a unit cell slightly larger than that quoted above (at 178 K a = 30.1 ± 0.1 Å and b = 4.58 ± 0.03 Å). Both of these prominent peaks then move gradually to higher Q, indicating a thermal contraction of the monolayer (