Sublimation of Li@C60

Experiments that probe the fundamental properties of endohedral fullerenes often require the preparation of molecular beams or thin films of the neutral molecules. It is challenging to cleanly sublime this class of molecules without producing some thermal degradation. We report combined gas phase and scanning tunnelling microscopy studies that probe the thermal decay of commercial [Li+C60]PF6- in a quartz ampoule and provide treatment conditions that will allow the sublimation of intact, neutral Li@C60 accompanied by a well-characterised component of neutral C60. The decay of the material at appropriate temperatures can be modelled with the assumption of a second order decay process in the oven yielding Arrhenius parameters that can predict the ratio of Li@C60 to C60 in the sublimed material.


Introduction
Endohedral metallo-fullerenes, where at least one metal atom is encapsulated within the fullerene cage, have been the subject of many studies probing their formation, properties and potential applications [1,2]. The most commonly studied metallofullerenes are formed by growing the carbon cage around the metal atoms. In many cases the most energetically favourable cage does not have the same atomic structure as the equivalent empty fullerene cage and it has proved particularly challenging to isolate M@C 60 using the conventional arc discharge or laser vapourisation methods [1]. An alternative way to produce C 60 metallofullerenes is to implant the metal ion into the already-formed carbon cage. This method was used in early experiments to produce Li@C 60 and has the advantage that the cage structure is the same as the well-known, icosahedral empty C 60 cage [3]. Experiments on the isolated material showed a higher reactivity compared to C 60 with a tendency for the purified material to oligomerise [4]. The production of Li@C 60 has recently been upscaled and commercialised by using a plasma implantation technique [5,6]. It was shown that the purified material could be stabilised by forming a hexafluorophosphate salt [6]. The availability of macroscopic amounts of the purified material in the form of the salt has allowed many characterisation studies to be carried out [7]. The Li@C 60 molecule is particularly attractive to study from the fundamental Contribution to the Topical Issue "Atomic Cluster Collisions (2019)", edited by Alexey Verkhovtsev, Pablo de Vera, Nigel J. Mason, Andrey V. Solov'yov.
Supplementary material in the form of one pdf file available from the Journal web page at https://doi.org/10.1140/epjd/e2020-10146-0. a e-mail: eleanor.campbell@ed.ac.uk point of view due to the high symmetry of the encapsulating carbon cage and the off-centre position of the encapsulated metal with a relatively strong charge transfer component between the metal and the cage. However, it is challenging to produce isolated molecules for detailed study either in the gas phase or deposited in the form of thin films. We have recently combined gas phase studies using mass spectrometry and photoelectron spectroscopy with scanning tunnelling microscopy to obtain insight into the socalled Super-Atom Molecular Orbital states of the molecule [8] and showed that it is possible to demonstrate multistate switching in a single Li@C 60 molecule with 14 distinct switching states [9]. These experiments required the preparation of a clean beam of Li@C 60 molecules (containing some proportion of empty C 60 ) that was prepared by heating the commercial [Li + @C 60 ]PF − 6 material in a quartz ampoule at elevated temperatures. In this paper we present results of combined gas phase and scanning tunnelling microscopy experiments that characterise the thermal decay of the material within the oven. The behaviour can be modelled empirically with the assumption of a second order decay process occurring in the oven, providing Arrhenius parameters that can predict the ratio of Li@C 60 /C 60 in the evaporated molecular beam.

Methods
An effusive molecular beam of Li@C 60 was produced by heating [Li + @C 60 ]PF − 6 (Idea International Inc., > 80% purity) in a small quartz capillary inserted into a heated molybdenum cylinder within an UHV vacuum chamber (background pressure 10 −8 mbar). It is important to ensure that hot fullerene materials do not make contact with heated metals to avoid extensive destruction of the cage and amorphization of the material in the oven, this is particularly important for metallo-endohedral fullerenes. Laser-desorption FTICR mass spectrometry of the starting material in both the positive and negative mode showed only minor traces of C 60 with the non-fullerene content being predominantly the stabilising anions and solvent residues. As we have shown previously [8,9] neutral Li@C 60 is evaporated from the oven along with impurities and some C 60 . The material is typically heated to a temperature of ca. 590 K for at least 20 h to remove solvent residue and other impurities before gas phase experiments such as photoelectron spectroscopy [8] are carried out. The neutral species are ionised with an ultrashort pulse laser in the extraction region of a simple linear Wiley-MacLaren time-of-flight mass spectrometer. The positively charged ions are detected at an MCP detector situated 42 cm from the interaction region. The time-offlight signal was recorded with a 500 MHz oscilloscope. All spectra reported in this paper were collected for 4 × 10 5 laser shots. Unless otherwise stated, the ions were produced with the third harmonic of a 120 fs Ti:Sapphire laser (267 nm). This wavelength efficiently ionises both C 60 and Li@C 60 with 2 photons (ionisation energies of 7.6 eV [10] and 6.5 eV [11], respectively) and reduces the influence of fragmentation. The relative photoionisation efficiencies are known to be dependent on the wavelength. The laser power, unless otherwise stated was within the range 3-3.5 mW for all reported measurements. This power produced only very small amounts of fullerene fragments.
The scanning tunnelling microscopy (STM) images were obtained in an ultrahigh vacuum surface analysis system with a base pressure below 10 −10 mbar, consisting of a preparation chamber allowing for standard sample preparation and molecular deposition, and a microscope chamber housing a CreaTec low-temperature STM. STM acquisition was performed at liquid helium temperature with electrochemically etched W tips and applying the bias voltage to the sample (tip remaining grounded). Imaging was achieved using the constant current mode. For the results reported here, the fullerene molecules were deposited on Au (1 1 1). The fullerene material was heated in a molecular evaporator consisting of a small quartz tube surrounded by a heating element. The evaporator was degassed at ca. 570 K for several hours prior to experiment, similar to the treatment used for the mass spectrometry experiments. The molecules were then dosed for 4 min. at ca. 665 K at a partial pressure of 5 × 10 −8 mbar and then annealed on the surface at 570 K for 30 s to form large hexagonally close-packed islands by mass diffusion before being quenched down to liquid He temperature for STM imaging.
3 Results and discussion 3.1 Mass spectrometry Figure 1 shows the development of the positive ion mass spectra produced by ionising the neutral evaporated material from the quartz oven with 267 nm, ultrashort laser pulses. The evaporation was carried out at an oven temperature of 590 K. There are strong signals that can be attributed to tetrabutyl ammonium phosphorus hexafluoride along with other impurities including H 2 O + . As the material continues to be heated at 590 K the intensities of the impurities and peaks due to TBAPF 6 decrease. Figure 1c shows the spectrum obtained after cooling and reheating, with a total of 15 h heating at 590 K. Here the impurity peaks have almost disappeared.
In order to check whether the empty C 60 is coming from the oven or is produced via fragmentation due to the laser excitation, we show the detected integrated intensity ratio Li@C + 60 /C + 60 (summing over all fullerene parent and fragment species) as a function of laser power in Figure 2 for two similar oven temperatures (600 K and 608 K) measured on different days, after the initial "de-gassing" process had been completed by heating for ca. 20 h at 590 K. The ratio is constant for both measurement series showing that the production of empty fullerenes is not a consequence of the ionisation process. The lower value for ratio in the mass spectra. Circles: oven temperature of 600 K. Squares: oven temperature of 608 K after the same material was heated to 623 K, cooled and reheated 6 days later. The ratio does not change significantly with laser power over the measured range but does change between measurements.
the ratio obtained at 608 K is due to the time the material has been heated in the oven between measurement series, as will be discussed later. The ratio of Li@C + 60 to C + 60 that is observed in the mass spectra changes with time and oven temperature, as well as with laser wavelength. The laser wavelength effect can be attributed to different photoionisation probabilities for Li@C 60 and C 60 . As discussed above in Section 2, we choose 267 nm to minimise this effect (only 2 photons required to ionise each molecule) and also to reduce the likelihood of fragmentation due to multiple photon absorption. However, we do not know what the relative detection efficiency is. The influence of heating and exposure to air after heating can be seen in Figure 3. Figure 3a shows the Li@C + 60 component to be stronger than the C 60 component on first heating to a temperature of 688 K. The material was then heated to 767 K where there is a significant change in the ratio of the two intensities with Li@C + 60 now significantly smaller than C + 60 (see further discussion in Sect. 3.4). The ratio continues to decrease as the material is cooled back down to 644 K. The same material was then cooled down to room temperature and removed from the vacuum chamber for a period of 12 days before being re-examined. Figure 3d shows the resulting mass spectrum on re-heating the material to 668 K, there is a further significant drop in the intensity of Li@C + 60 and the achievable intensity for similar laser ionisation conditions is noticeably less (however, note that the absolute intensity values in Fig. 3 are comparable for (a)-(c) but not for (d) due to slightly different apparatus settings).
There is clearly a change in the heated material, particularly when exposed to atmosphere after heating, with a component of insoluble material within the oven in addition to the changed mass spectra. Raman spectroscopy confirms significant changes at the surface of the material left in the oven after heating, with evidence for amorphisation of the material (S.I. Fig. S1). However, MALDI mass spectrometry of the soluble component of the remaining material (sonicated in dichlorobenzene for 20 min prior to filtering) shows identical spectra to the starting material (S.I. Fig. S2), indicating that the material changes are predominantly at the surface of the material in the oven. This is also known to be the case for empty C 60 where studies have shown that the vapour pressure can decrease significantly on successive heating cycles accompanied by the appearance of an insoluble amorphous residue, particularly in the presence of any organic solvent residue in the starting material [12,13].
In addition to the decreasing Li@C + 60 intensity with time at elevated temperatures, there is an accompanying increase in the intensity of Li + in the mass spectra. If we assume that the isolated Li + signal is a consequence of the destruction of Li@C 60 within the oven (there is no non-endohedral Li in the starting material) we can use the temperature dependence to extract an effective activation energy for the appearance of Li and thus the destruction of the endohedral species, assuming that the Li efficiently escapes from the oven. An Arrhenius-type plot of the natural logarithm of the Li + intensity measured in the mass spectra multiplied by temperature (proportional to the partial pressure of Li in the oven) [14] as a function of inverse temperature is shown in Figure 4. This yields an activation energy for the cage destruction within the oven of 1.27±0.05 eV (errors are the 95% confidence limits from the straight line fit). This value is remarkably close to the activation energy determined previously for Li loss from the dimerised endohedral species in solution (1.1 ± 0.2 eV) [15]. It is much lower that the value of 5.4±0.2 eV obtained for release of Li from the isolated gas phase cationic endohedral molecules that is similar to the threshold energy for Li + implantation in C 60 [16] and the bias voltage at which Li loss from the C 60 cage is observed in STM experiments [9]. The determined activation energy for the material within the oven is thus not a consequence of a simple unimolecular decay reflecting the loss of Li from an intact C 60 cage but is a consequence of more complex reactions with impurity species in the oven. However, the "effective" activation energy that is determined in this way can give a useful estimation/prediction of the thermal stability of the endohedral material in the oven.

Enthalpy of sublimation
The enthalpy of sublimation can be determined from the intensity of the molecular species in the mass spectra. We follow the mass spectrometry procedure used previously to determine the enthalpy of sublimation of C 60 , C 70 [13,17] and Er 3 N@C 80 [18]. The ion intensity measured in the mass spectrometer is proportional to the vapour pressure divided by the temperature. A plot of ln(Intensity × Temperature) as a function of inverse temperature should yield a straight line with the gradient determined by the activation energy for vapourisation, −E a /R where R is the gas constant. The E a value is equivalent to the internal energy of sublimation, ∆ sub U at the average temperature T av for which the measurement was carried out (since sublimation occurs into vacuum). This is converted to the enthalpy of sublimation at T av , ∆ sub H Tav by addition of RT av [19]. To convert ∆ sub H Tav to ∆ sub H 298 we consider the difference  in heat capacity between the gas and solid, ∆ g s C P , at constant pressure, ∆ sub H 298 = ∆ sub H Tav − ∆ g s C P (T av − 298), with ∆ g s C P = −37.4 J K −1 mol −1 [20]. Figure 5 shows a plot of the flux of Li@C 60 and C 60 from the endohedral material for two heating cycles as well as the results from a pure C 60 sample from the first heating cycle for comparison. The results from the different measurement series have been shifted vertically on the plot for clarity. The extracted values for the enthalpies of sublimation are summarised in Table 1. The value obtained for the pure C 60 material (99.95% purity, SES research), ∆ sub H 298 = 156 ± 18 kJ mol −1 , is consistent with previous reported measurements for a first heating cycle [18]. Note that, as discussed in the literature [12,13,18], the value obtained for the enthalpy of sublimation depends strongly on how the material was thermally treated prior to measurement. In this case it was heated at 368 K for 3 days to remove solvent residue. As is usual for fullerene materials, the second and subsequent heating cycles produce a significantly lower vapour pressure than the first cycle. The deviation of the data points at high temperature from the straight-line fit is indicative of changes occurring within the oven that reduce the vapour pressure. This can also be seen from the results from the Li@C 60 material where the extracted ∆ sub H 298 increases from 162 ± 17 kJ mol −1 on the first heating cycle to 203 ± 11 kJ mol −1 on the second cycle. The enthalpy of sublimation determined for C 60 that is emitted from the endohedral fullerene material in the low temperature range where the values have been determined is identical to that for Li@C 60 within the 95% confidence limits of the fits. It is also the case that the enthalpy of sublimation for pure C 60 on the first heating cycle is the same as for Li@C 60 and C 60 from the endohedral material within the experimental errors. The Table 1. Values of the enthalpy of sublimation (in units of kJ mol −1 ) extracted from the data shown in Figure 5. The error limits represent the 95% confidence levels of the regression analysis.

Sample
TAv  6. STM constant current images of fullerene islands deposited from [Li + C60]PF − 6 exposed to thermal pre-treatment at 570 K in a quartz evaporator prior to deposition. The images were taken at a bias voltage of +2.5 V that allows the Li@C60 molecules (diffuse circular image) to be distinguished from the empty C60 molecules (dumbbells and trefoils). (a) Pre-heating at 570 K for 6.5 ± 0.5 h, 42 ± 3% Li@C60. (b) Pre-heating at 570 K for 105 ± 5 h, 13 ± 3% Li@C60. Insert illustrates the three distinct structures: red circle: Li@C60; black circles: C60 (dumbbell and trefoil).
results show a similar trend to those obtained from studies of the sublimation of Er 3 N@C 80 [18] where the sublimation enthalpy for an average temperature of ca. 870 K was 165 kJ mol −1 on the first heating cycle increasing to 237 kJ mol −1 for subsequent heating cycles at an average temperature of 955 K.

Scanning tunnelling microscopy images
In our previous studies of scanning tunnelling spectroscopy of Li@C 60 deposited on Au(1 1 1) [8,9] we showed that it was possible to distinguish Li@C 60 from C 60 by imaging the molecules at a bias voltage of +2.5 V. The thermal treatment of the endohedral fullerene material prior to deposition is similar to that used for the gas phase experiments (to ensure that deposition of non-fullerene material from the oven is minimised) and this leads to a mixture of Li@C 60 and C 60 being deposited on the gold surface. The fullerene molecules form hexagonally close-packed islands. At negative bias voltages (imaging the filled states) the molecules are indistinguishable, however at the positive bias voltage of +2.5 V, the S-SAMO orbital of Li@C 60 is accessed and the endohedral molecules appear to "light up" showing a diffuse circular image with no apparent structure [9]. It is therefore possible to directly count the endohedral and empty fullerenes that are deposited on the substrate for a given thermal treatment in the oven. Figure 6 shows two example images taken with a +2.5 V bias on fullerene islands that were deposited after different thermal pre-treatment. The material deposited in Figure 6a was de-gassed at a temperature of 570 K for 6.5 ± 5 h while the material deposited in Figure 6b was treated at the same temperature for 105 ± 5 h. The diffuse circular structures are Li@C 60 while the smaller "dumbbell" and darker "trefoil" structures are due to empty C 60 that has different orientations on the surface [9]. It is clear from these images that there are significantly smaller amounts of Li@C 60 deposited in Figure 6b. Statistical analysis of the deposited islands for the two different pre-deposition thermal treatments yields 42 ± 3% Li@C 60 for the shorter thermal treatment and 13 ± 3% Li@C 60 for the longer thermal treatment. The STM analysis gives additional information about the decay of the endohedral material. Although we have very few data points we can estimate the decay rate within the oven. There are many different models to describe the kinetics of solid state reactions depending on the most probable mechanism for a particular case [21]. Our data do not fit well to a simple first order reaction but can be reasonably fitted by e.g. the assumption of a second order reaction or a more complex 3D diffusion model. Here, we do not have sufficient data points to determine the reaction mechanism but are mainly concerned with comparing the extracted rate constant at the de-gassing temperature used for the STM experiments with the decay behaviour observed from the gas phase mass spectrometry results in order to The detection probability, that influences the absolute value of the ratio detected in the gas phase experiments was found to be six times larger for Li@C60 than for C60.
determine an empirical estimation of the extent of decay of the endohedral fullerene material as a function of thermal treatment. Using a second order decay model, the rate constant determined from the STM measurements is k = (1.7 ± 0.2) × 10 −5 s −1 at 570 K, where the error limits are given by the standard error of the regression fit. Note that the value of the rate constant determined from the data is dependent on the model assumed. Taking the activation energy for the decay from the gas phase measurements (1.27 ± 0.05 eV) allows us to then estimate the range of possible pre-exponential factors for the decay (A ≈ 9 × 10 5 -8.7 × 10 6 s −1 ). Figure 7a shows the measured Li@C + 60 /C + 60 intensity ratios as a function of temperature for the measurement series plotted in Figure 5 as "Li@C 60 1st" and "Li@C 60 2nd". The typical measurement time at each temperature was ca. 10 min, however there are temperatures at which the material was held longer (e.g. at ca. 635 K while other checks were being carried out). The arrows show the order of the measurements. It is clearly seen that the endohedral component of the fullerenes that are emitted from the oven decreases with the temperature (and measurement time). The low ratio measured for the highest temperature in the series continues to decay as the oven is cooled back down again. The ratios obtained for the two series of measurements up until a temperature of ca. 600 K are similar, indicating that no significant change occurred to the amount of endohedral component in the oven during the first measurement series, although some degradation was caused prior to the measurement due to the long de-gassing time (ca. 20 h) at 590 K giving a starting value of the Li@C + 60 /C + 60 ratio of ca. 2.3. The second series extended to higher temperatures and shows a strong decrease in the ratio from 2.3 at the start (corresponding to 70% Li@C 60 ) to 0.3 at the end of the measurements (23% Li@C 60 ). Figure 7b shows the same data from "Li@C 60 2nd" along with calculations (dashed and full lines) of the expected ratio. The full line provides the best fit to the data with values of A = 3.8 × 10 6 s −1 and E a = 1.285 eV, while constraining the fit to remain within the appropriate range to describe the STM experiments, k (570 K) = 1.74 × 10 −5 s −1 . The dashed lines indicate the range for these values to still be compatible with the trend observed in the experimental measurements, with A = (3.8 ± 0.2) × 10 6 s −1 , E a = 1.285 ± 0.005 eV and k(570 K) = (1.74 ± 0.26) × 10 −5 s −1 , indicating the sensitivity of the fit to the three parameters. Note also that the optimum fit values are dependent on the model assumed to extract k. A similar analysis using the 3D diffusion model to fit the STM data (giving a smaller value for k) provided a poorer fit to the decay data shown in Figure 7, predicting a more rapid drop in the detected ratio at higher temperatures. The only other fit parameter is the combination of the photoionisation and detection probability of Li@C 60 compared to C 60 that changes the absolute value of the ratio measured in the gas phase experiments but does not influence the dependence on temperature/time. The fit in Figure 7b uses a detection ratio of Li@C 60 /C 60 = 6 at the wavelength of 267 nm. The calculations take account of all prior heating stages, including the 20 h of de-gassing at 590 K and the heating that took place during the first series of measurements. Very good agreement is obtained with the experimental data up to a temperature of 700 K. The simple model, however, over-estimates the decay that continues to happen at higher temperatures. It is striking that such a simple model can provide such good agreement with all the experimental observations and allow predictions of the decay of the endohedral species within a temperature range that is relevant for preparation of gas phase targets or deposition of thin films.

Conclusion
We have shown that it is possible to sublime neutral Li@C 60 and C 60 from commercially available [Li + @C 60 ]PF − 6 . In order to obtain a "clean" molecular beam it is necessary to heat the material at a low temperature for many hours to remove components related to the stabilising anion and solvent residues ("de-gassing"). This treatment leads to thermal decay of the material in the oven, reducing the Li@C 60 component and leading to the formation of an insoluble component. We have characterised this decay by carrying out mass spectrometry studies using gentle 2-photon ionisation with 267 nm, 100 fs laser pulses, to monitor the relative intensity of Li@C 60 and C 60 in the beam of material and used the presence of photoionised "free" Li to determine the activation energy for the decay process. Complementary STM experiments allowed us to directly determine the numbers of Li@C 60 and C 60 molecules deposited on a substrate for the de-gassing temperature used in the STM experiments (570 K). The percentage of Li@C 60 observed was then used to extract an effective rate constant for the decomposition reaction at 570 K. By combining the gas phase and STM experiments it was possible to fit the decrease in the observed Li@C 60 /C 60 ratio in the mass spectra as a function of temperature and time of heating. More detailed studies are required to obtain further insight into the mechanism of the reaction(s) happening within the oven but the data provided in this paper provides a useful practical means of estimating the remaining Li@C 60 that is available in the gas phase or deposited on a substrate as a function of the thermal treatment that the [Li + @C 60 ]PF − 6 material has been exposed to.