In Situ Surface-Sensitive Investigation of Multiple Carbon Phases on Fe(110) in the Fischer–Tropsch Synthesis

Carbide formation on iron-based catalysts is an integral and, arguably, the most important part of the Fischer–Tropsch synthesis process, converting CO and H2 into synthetic fuels and numerous valuable chemicals. Here, we report an in situ surface-sensitive study of the effect of pressure, temperature, time, and gas feed composition on the growth dynamics of two distinct iron–carbon phases with the octahedral and trigonal prismatic coordination of carbon sites on an Fe(110) single crystal acting as a model catalyst. Using a combination of state-of-the-art X-ray photoelectron spectroscopy at an unprecedentedly high pressure, high-energy surface X-ray diffraction, mass spectrometry, and theoretical calculations, we reveal the details of iron surface carburization and product formation under semirealistic conditions. We provide a detailed insight into the state of the catalyst’s surface in relation to the reaction.


XPS experimental setup.
In this work we employ the POLARIS instrument at the P22 beamline of the PETRA III synchrotron at DESY. [1][2][3] It is based on a Scienta-Omicron R4000-HiPP-2 electron analyser and is designed to be capable of acquiring photoelectron spectra at pressures from several hundred mbar to 1 bar and beyond. The spectra were recorded using photons at 4.6 keV energy that were impinging on the surface at an incident angle of 0.4 degrees with respect to the surface plane. This is below the angle of total external reflection for iron at that energy. High energy increases the photoelectron mean free path while low incident angle enhances the surface sensitivity. 1 The pass-energy of the electron analyser was 100 eV. The incoming beam was focused down to ca. 1015 m 2 with the photon flux of about 10 13 photons per second distributed over the entire footprint, which at this angle was about 140015 m yielding about 4.810 14 photons/s per 1 mm 2 . The sample sits in a custom-made holder manufactured from 1.4762 -AISI 446 steel and heated with a BN heater, which is in direct contact with the backside of the crystal. Due to the specific design of this experimental setup, no contamination from the heater or chamber can reach the sample surface when gas flow is applied.

SXRD experimental setup.
The diffraction experiments were performed using synchrotron radiation with an energy of 83 keV at the P21 beamline of PETRA III, Deutsches Elektronen-Synchrotron (DESY) in Hamburg, Germany. The details of the High-Energy Surface X-Ray Diffraction methodology, data recording and treatment can be found elsewhere. 4,5 The sample was placed in a specially designed chamber 6 mounted on a Huber® diffractometer stage, which allows for a precise surface alignment to the incident beam. In order to maximize the surface to bulk signal ratio, the angle of incidence was set to 0.035º, which is slightly above the critical angle of total external reflection for iron when irradiated by 83 keV X-rays. The chamber combines an ultra-high vacuum (UHV) part for sample preparation and a 50 ml gas flow reactor in which the sample can be exposed to a gas mixture. During the experiment the sample was continuously exposed to a photon flux on the order of 1.3·10 10 photons/sec with an incident beam size of 4x10 µm 2 . The beam induced effects were tested and found to not influence the experiment. To record the diffraction patterns a 430x430 mm 2 Varex Imaging flat panel X-ray detector was used. It has a resolution of 2880x2880 pixels with the physical size of 150x150 µm 2 . In order to avoid the saturation by highly intense diffraction maxima, the corresponding parts of the detector were covered with Tungsten pieces blocking the electromagnetic radiation. This protection is seen in experimental patterns as black circles.

Mass-spectrometry.
To measure the reaction products in the XPS experiment, a Hiden mass spectrometer (HAL/3F RC 301 PIC system) attached to the differential pumping of the spectrometer was used. It allows for sensing the gas flowing directly from the reaction volume to the pump through the aperture of the analyser's nozzle. Thus, XPS and MS data are from an as close as possible probing volume. In the SXRD experiments, the partial pressures of the reactants and the reaction products were followed with a RGA200 Stanford Research Systems residual gas analyzer accessing the reaction volume via a leak.
Recoil effect in XPS data treatment. Using the approximation described by Takata et al. 7 it is possible to estimate the magnitude of the shift and the broadening caused by the recoil effect for, respectively, C 1s, O 1s and Fe 2p spectral lines measured at 4.6 keV photon energy, i.e. the values relevant to the present work. The energy shift can be calculated using a simple formula , where is the mass of the atom, = ( / ) • is the electron mass, and is the photon energy. The broadening of the spectral lines, then, can be approximately estimated as , where is the Bolzman constant ( ) and is temperature in .

S2. Sample preparation
In all experiments the surface of a 4N5 purity Fe(110) single-crystal purchased from Surface Preparation Lab (SPL) was prepared by multiple cycles of 30 min Ar + ion sputtering at 1.5 kV and annealing at up to 700C for 5 minutes alternating with hydrogen treatment at 10-100 mbar pressure and 400C temperature (occasionally the H 2 flow was exchanged with CO 2 flow for mild surface oxidation).
For XPS experiments, survey scans were acquired every time before and after a set of measurements in the experiment to ensure the absence of surface contamination. The main focus was set on checking for such common contaminants as silicon and sulfur. The carbon C 1s region was also required to be featureless, although, small amounts of adventitious carbon were tolerated if present since it is almost impossible to get rid of it outside of Ultra-High Vacuum (UHV) conditions. Besides, pure hydrogen flow ushering every measurement effectively removed adventitious carbon. A mild oxidation of the surface in the form of iron oxide was also not considered as a contamination since it is notoriously difficult to keep iron metallic outside UHV conditions and it inevitably oxidizes upon introduction to the chamber of a CO containing reaction gas mixture and reduces later in the carburization process.
In Figure S2-1, one can see two representative spectra taken at 100 mbar pressure of pure hydrogen at 152C prior to the measurement and at 550 mbar pressure of the 1CO : 2H 2 reaction gas mixture at 318C after the corresponding experiment. The spectrum in the top panel taken in hydrogen, shows the absence of contaminations on the surface before the set of measurements. The spectrum at the bottom features the residual carbon signal in C 1s region and the signal from the gas phase CO in both C 1s and O 1s regions (see Figure S2-2 for more details). Note that a high background of photoelectrons inelastically scattered in the gas phase as well as additional electron energy loss features (H 2 excitation 12.8 eV, CO excitation 8.5 eV) are present in the spectra on the higher binding energy side of each peak due to the high gas pressure.  For the diffraction experiment, the sample was prepared in the same manner as for the XPS studies. Before every set of measurements, a survey pattern was taken and was required to represent a flat metallic surface. One such pattern is shown in Figure S2-3, panel d. The black circular shapes represent the tungsten pieces covering the detector from oversaturation in the places where the diffraction maxima from the crystal should occur. The vertical streaks of enhanced intensity represent the crystal truncation rods originating from the Fe(110) surface. The insignificant traces of polycrystalline iron phase indicated by the weak rings in the diffraction patterns were assigned to the edges of the sample that could not be efficiently sputtered and were tolerated. The absence of a significant amount of disordered contaminants like Sulphur was implied based on the previous X-ray photoelectron spectroscopy studies of the same sample where a similar treatment did not cause any significant segregation of impurities on the surface. To describe the surface structure, a monoclinic basis set of vectors a 1 , a 2 with the angle = 54.74º between them lying in the surface plane and a 3 perpendicular to them was used. In terms of the bulk iron lattice constant a 0 = 2.866 Å, the lengths of these vectors are |a 1 | = a 0 = 2.866, |a 2 | = a 0 = 2.482, and |a 3 | = a 0 = 4.053 Å. The Reciprocal Lattice Units (RLU). Diffraction data was collected by recording the diffraction patterns corresponding to different azimuthal angles during the sample rotation by 125º under the photon beam. The angular step was chosen to be 0.2º resulting in 625 diffraction images in one data set. To resolve the surface changes in time, separate measurements at a constant azimuthal rotational angle corresponding to the hl-plane in reciprocal space were performed as well recording the detector images every 5 seconds.

S3. X-ray beam induced effects
In order to ensure the absence of X-ray beam induced effects, we performed the following experiment. Two separate measurements were done under identical conditions (200 mbar pressure, 233C temperature, 1CO:4H 2 gas mixture at 2.1 L n /min total flow, incidence angle 0.4) with the same initial clean state of the surface. In the first measurement, the surface was exposed to the X-ray beam and the surface evolution was recorded for 20 minutes. In the second measurement, the X-ray flux was set to zero by closing the beamline shutter and the surface stayed under the same conditions as in the previous measurement for 20 minutes. Then the beam was switched on and the state of the surface was recorded. Figure S3-1 shows the results of these measurements. As one can see from the left column of Figure S3-1, the first sweep of the measurement records iron oxide on the surface as always when the transition between UHV and high-pressure mode is done. At the same time no significant carbon signal is visible. Very quickly, the oxide is reduced and the C 1s signal reveals the appearance and growth of carbon species between 283 and 284 eV binding energy. Later, also signal between 284 and 286 eV binding energy gets more pronounced. The right column shows the state of the surface after it had been exposed to the same conditions but without the photon beam for 20 minutes. The logics of the experiment tells that if no beam-induced effects are involved, the last (topmost) sweep of each electron level in the left column should be similar to the first (bottommost) sweep of each corresponding level in the right column, as they both are meant to capture the surface after 20 minutes of exposure to the same reaction conditions. Naturally, these sweeps should be significantly different if beam-induced effects are involved. In Figure  S3-2, we take a closer look at these two sweeps for C 1s electron level. Figure S3-2. Comparison of C 1s XP spectra recorded using 4.6 keV photons for Fe(110) single crystal surface exposed to 1CO : 4H 2 gas mixture at 2.1 L n /min total flow at 200 mbar pressure and 233C temperature for 20 minutes with and without photon beam.
From Figure S3-2 it becomes clear that even though the XP spectra don't look entirely equal, the same carbon species are present on the surface in a comparable amount. This means that the beam-induced effects are either non-existent or don't influence the principles of surface evolution under reaction conditions. A slight difference between the spectra could have been caused by a slight difference in the initial state of the surface, such as a presence of small amount of adventitious carbon, or a fluctuation in the gas flow -the experimental uncertainties that cannot be entirely avoided in such a complex experiment. It should be noted that the spectrum measured after 20 minutes of no beam exposure is noisier than the spectrum with the beam on. The reason for that is a small sample to beam misalignment that accumulates over time due to the thermal drift and is corrected by maximizing the signal that is being detected. Naturally, when the beam is disabled for 20 minutes, no such correction is being performed resulting in lower signal after the beginning of the exposure. The corresponding O 1s and Fe 2p 3/2 spectra are identical and don't reveal any beam induced traces. The same approach was taken in the diffraction experiment with the same result that did not show any significant difference between the diffraction patterns after keeping the sample for 20 minutes under the reaction conditions with the X-ray beam on and off.

S4. Sample heating and surface temperature measurements
In the XPS experiment, the sample rests in a custom-made holder manufactured from 1.4762 -AISI 446 steel and heated with a BN heater, which is in the direct contact with the back side of the crystal. The temperature is measured by an N-type thermocouple placed between the heater and the sample's back side. In the high-pressure regime, high flow of room-temperature gas supplied to the surface cools the surface down by a certain amount, which should be accounted for. To do this, we have performed a number of calibration tests, for which purpose a dummy iron sample with two thermocouples spot-welded to both front and back sides was subjected to a series of temperature measurements at various gas flow and pressure conditions. The results of the tests are briefly summarized in Figure S4  From the figure it can be seen that the main contribution to the surface temperature offset is related to the absolute temperature of the sample's back side, while the dependence on the pressure and the gas flow is rather small. Specifically, at flows below 3 L n /min and pressures below 700 mbar (the upper limits in the currently discussed experiment), the temperature change due to these factors are negligible. Figure S4-2 shows the values of surface temperature offset measured at 100, 200, 300 and 400ºC (=sample backside temperature) and the non-linear interpolation of the measured points with 2 nd and 3 rd order polynomic functions. Both non-linear approximations give very similar values in the region of interest of the experiment, namely [100ºC; 400ºC]. Subtracting the corresponding values of the surface temperature offset from the thermocouple measurements allows for the determination of the actual temperature on the sample surface in the high-pressure regime.

S5. Fermi level correction
The binding energy scale in all spectra was referenced to the Fermi level measured after every change of experimental conditions and stayed constant throughout the entire experiment. An example of recorded Fermi level and its fitting is shown in Figure S5-1.

S6. XPS experimental data examples
Here we provide more examples of the experimental data for the reader's reference. Two representative sets of XPS measurements of a Fe(110) surface at 85 mbar and 550 mbar pressure and 1:1, 1:2, 1:4, and 1:10 CO to H 2 gas mixtures with 2-2.5 L n /min total flow have been chosen for display in Figures S6-1 and S6-2. The data are normalized by the number of sweeps in every scan, by the dwell time, and by the background level on the low binding energy side of C 1s region (for scans of the same set taken simultaneously at the same conditions). To simplify the visual comparison of spectra, the intensity was also corrected by the corresponding values of the photoionization cross-section and the total scattering loss of photoelectrons in gas. In Figure S6-3, the C 1s level of the same measurements is shown in a different representation for more details and easier direct comparison. Representative mass-spectrometry data corresponding to the measurements at 550 mbar for 1CO : 1H 2 and 1CO : 10H 2 gas mixtures are also presented in Figures S6-4 -S6-7. Note that the MS signal m/z = 18, which corresponds to water, is greatly affected by interactions with the chamber walls of the differential pumping stage, which results in an accumulation in the system over time, and this signal can thus not be directly correlated to the activity at the indicated temperature.