https://doi.org/10.1016/j.elecom.2022.107375
“A micro-electrochemical cell is sealed with 2–5 layers of graphene to monitor the changing oxidation state of Cu nanoparticles (NPs) with X-ray photoelectron spectroscopy (XPS) in a mildly alkaline aqueous solution under electrochemical control. The main role of graphene is to ensure an abrupt change between the liquid and vacuum environments, where the latter is required for conducting XPS experiments. Decent transparency to the generated photoelectrons with a kinetic energy of few hundred eV is another requirement that graphene fulfils for performing such experiments. Graphene also acts as an electrically conducting support material for Cu NPs, ensuring that a bias can be applied to them. The proof-of-concept measurements presented in this work show that relatively lower flux X-ray sources, such as those with Al-Kα emission that are commonly used in laboratories, are sufficient for probing the solid–liquid interfaces with this approach.”
”
X-ray photoelectron spectroscopy (XPS) is among the most powerful techniques for probing solid materials, as it is element specific, sensitive to chemical changes in the material, and surface sensitive. Probing solid–liquid interfaces with XPS has been a long-standing goal because of its surface sensitivity [1], which relies on the limited inelastic mean-free path (IMFP) of electrons, but this also has limited XPS to high vacuum (HV) conditions. Over the last two decades, ambient pressure (AP)XPS has been developed, extending the operating pressure into the mbar pressure regime (Fig. 1a) [2], but a standard APXPS setup is still not sufficient for working with liquids. There are currently two approaches for probing the electrified solid–liquid interfaces conveniently with APXPS: the dip-pull method, which requires hard X-rays and synchrotron radiation [3], and the graphene-capping approach [1]. Whilst the former technique has been successfully applied to a couple of electrochemical reactions [3], the second approach has been successfully used in electroplating [4] and in characterising IrO2 at anodic potentials [5]. Here, we demonstrate that the graphene-capping approach can be used both at positive and negative polarisations for multiple cycles under constant potential conditions between the working and counter electrodes, while monitoring the changing chemical state with XPS. More importantly, our approach requires neither hard X-rays nor synchrotron radiation, making ‘electrochemical-XPS’ applicable to the lab-source soft X-rays that are widely available in APXPS instruments.
Fig. 1. (a) Schematic of APXPS setups. The sample can be in the presence of ∼ 1 mbar of gas, in which the photoelectrons have an IMFP of a few mm. To avoid massive losses in XPS intensity, the sample is kept at a distance below 1 mm from a nozzle. (b) SEM image from one of the perforated regions collected with a transmission detector. Dark spots are due to Cu NPs. Scale bar is 100 nm. (c) Examples of Raman spectra (with 532 nm excitation) of two layers of graphene stacked on top of each other with Cu NPs behind it. (d) Schematic of our micro-electrochemical cell that is placed in the sample position shown in (a). A droplet of aqueous NaOH solution is stored in the cell (light blue). The WE is graphene (dark grey) that is supported on Au/SiNx (gold/purple) with a perforated array of holes. Photoelectrons can penetrate to the vacuum through these holes that are covered with graphene. The backside of graphene is decorated with Cu NPs (orange). A Pt wire is used as the CE. Electrical connections to both electrodes are from outside of the vacuum chamber using electrical feedthroughs. (e) Photograph of the micro-electrochemical cell under the nozzle. The circular part at the centre of the cell is the SiNx chip with the rectangular membrane just below the nozzle. Photons only hit this 200-nm thick rectangular window part. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
The current APXPS systems utilise electron optics to guide the generated photoelectrons through several orifices as part of a multistage differentially pumped lens system to extend the measurement pressure range (Fig. 1a) [6], [7], [8], [9]. However, this approach is insufficient for liquids. By capping the liquid reactor with an ultra-thin membrane, an abrupt change between the reactor cell and measurement chamber can be obtained [10], [11]. The membrane must be strong enough to sustain the large pressure difference, thin enough to allow sufficient photoelectrons to pass through, and electrically conductive to avoid any major charging-related issues. Graphene fulfils all of these requirements [11], and indeed, graphene layers have been used to maintain liquid environments under vacuum conditions for both electron microscopy [12], [13], [14], [15], [16] and XPS measurements [4], [10]. A similar approach was previously used to probe electrode–electrolyte interfaces with X-ray absorption spectroscopy (XAS) and photoemission electron microscopy (PEEM) [17], [18]. Building on this, we developed graphene-capped micro-electrochemical cells for Kelvin probe force microscopy (KPFM) studies on electrochemical gating of graphene [19], and the changing electronic interaction between graphene and Cu nanoparticles (NPs) due to the chemical state changes of Cu under electrochemical control [20]. However, KPFM does not directly measure the changing chemical state; therefore, the chemical state was rather inferred from the doping behaviour of graphene [20].
”
”
2. Experimental
We used the same preparation method for capping the micro-electrochemical cell as in our previous studies [19], [20]. In brief, graphene layers grown on a Cu foil were transferred onto Au-coated SiNx grids after etching away the Cu foil and diluting the Na₂S₂O₈ etchant solution with MilliQ H2O. SiNx grids have a 200 nm thick ‘window’ region (0.45 × 0.45 mm2) with an array of perforations (Φ ∼ 500 nm) at their centre [19]. The Au coating ensures electrical continuity over a large area to minimise charging issues during XPS experiments. Localised tears can form in single-layer graphene during this transfer process which are detrimental to our experiments because graphene cannot perfectly seal the cell inside a vacuum chamber. This was not a problem in our previous experiments in air, because there is enough H2O vapour pressure in air to avoid evaporation of H2O through the graphene tears. To cope with this issue, we either repeated the transfer procedure twice and prepared unstacked double-layer graphene or sealed the cell initially with multilayers (3–5 layers) of graphene. Both approaches come at the expense of XPS intensity because photoelectrons have to travel through multilayers of graphene [21], but the sacrifice in signal is necessary to avoid failure of the membranes in vacuum. The transfer of graphene is done without the use of a support polymer [19], in order to avoid polymer residues that can potentially cause additional side reactions and can decrease the XPS intensity further [21].
Cu was evaporated on the back side of a graphene-covered Au/SiNx sample with a nominal thickness of 3 nm, resulting in the formation of Cu NPs. The samples were then annealed to ∼ 230 °C in HV for 2–3 h to desorb H2O and hydrocarbons. A scanning electron microscopy (SEM) image of one of the graphene-covered perforations that is decorated with Cu NPs is shown in Fig. 1b. Raman spectra of two layers of graphene from three randomly chosen spots are shown in Fig. 1c, verifying the high quality of our graphene membrane after polymer-free transfer, exposure to Cu vapour, and post-annealing. The spectra contain the typical D, G, and the 2D peaks, the former being associated with the defect density [22]. The intensity of the D peak, as well as the relative intensity of the G and 2D peaks, appears different at different regions of the sample, due to its heterogeneity. The total area of the perforated regions in our samples is ∼ 22 % of the total area under X-ray illumination. Assuming that Cu coverage is roughly half, and more than half of the signal is attenuated due to multiple graphene layers, the Cu signal is expected to be roughly 20 times smaller than the signal from a solid Cu sample.
Electrochemical experiments were performed using a custom micro-reactor in a two-electrode configuration (Fig. 1d) [19]. A Φ 1 mm Pt wire served as the counter electrode (CE) and graphene/Cu served as the working electrode (WE). The electrolyte was enclosed in a polyether ether ketone housing and sealed with the SiNx grid (Φ 3 mm) glued to an alumina washer (Φ 4 mm) that was sandwiched between an O-ring and a stainless steel lid, which has a Φ 2 mm aperture. We used a mildly alkaline aqueous NaOH solution (10-4 M) as electrolyte. The solution was prepared using MilliQ H2O and was bubbled with Ar prior to the experiments.
APXPS experiments were performed in a lab-based spectrometer at the Center for Functional Nanomaterials (CFN) at Brookhaven National Laboratory (BNL) [23]. The SiNx grid was placed ∼ 0.6 mm from an aperture (Φ 0.3 mm) that leads to a differential pumping system, which contains a series of electrostatic lenses to focus the photoelectrons into a SPECS PHOIBOS 150 NAP hemispherical analyser. The excitation source is a monochromatic Al-Kα (Ehν = 1486.6 eV) anode that is focused to a ∼ 0.3 mm spot. A pass energy of 20 eV was used. After adjusting the distance between the cell and the nozzle by maximising the intensity of the Au 4f signal on an off-membrane region, the xy-position of the electrochemical cell was aligned by first minimising the intensity of the Au 4f signal and then maximising the Cu 2p signal, to make sure that the Φ 0.3 mm spot is on the window region (Fig. 1e).
Several precautions were taken. When the cell was first introduced into the loadlock chamber, the chamber was pumped down slowly to 0.1 mbar to avoid rapid bulging and subsequent failure of the SiNx window. The SiNx window indeed bulges slightly despite the slow chamber evacuation, but this does not cause any problem as long as the cell retains the aqueous solution. Another necessary precaution is sustaining a 0.1 mbar pressure of H2O vapour in the APXPS analysis chamber prior to sample introduction and during the data collection. This background pressure slows any potential leaking from the tears in graphene. Finally, the resistance between the two electrodes was measured both in air prior to the experiments and in the APXPS chamber during the experiments. Maintaining a rather constant resistance throughout the experiments ensures contact between the electrodes through the electrolyte solution, i.e., H2O leaking is not significant.
Electrochemical control of the cell during XPS data acquisition was maintained by electrical feedthroughs to the pins that are connected to the electrodes (Fig. 1d). A potential difference between two electrodes was applied with a power supply that is grounded to the same ground as the electron analyser. Then either the WE or the CE was kept at ground while applying a positive or negative potential to the other. If the WE was at a non-zero potential, this causes a shift in the XPS binding energies which we present here after shifting them back.
3. Results and discussion
Fig. 2 shows the Cu 2p3/2, C 1 s, and O 1 s core-levels of the XPS spectra obtained from a sample that had been stored 2 days in air after annealing prior to electrochemical XPS experiments. The initial peak positions are not exact and appear slightly shifted; for instance at VWE – VCE = 0 V conditions, the C 1 s peak appears at 285.2 eV (bottom spectrum in Fig. 2b) and the Au 4f7/2 peak appears at 84.3 eV (not shown) in the binding energy scale. These are higher than the typical values of 284.5 eV and 84.0 eV obtained for graphite and Au reference samples. Moreover, the exact positions of the graphene and Au peaks are slightly different for each experiment. The reason behind this is the imperfect ohmic contact between graphene and Au, which leads to slight charging of the sample during measurements, thereby shifting the XPS peaks to higher positions. Defect-free graphene is an excellent electrical conductor along the plane thanks to its delocalised π-electrons, but the conductivity along the surface normal is poor. Furthermore, the high defect density of CVD-grown graphene reduces its conductivity significantly. We estimate the resistance of our graphene (without the Au layer) to be in the 0.4–1.5 kΩ range across ∼ 2 mm, with the exact value depending on the graphene quality and number of layers. Similarly, Cu peaks also appear at shifted positions because their only electrical contact is through graphene. For instance, an annealed test sample with graphene/Cu2O produced a peak at the expected position of 933 eV (calibrated position with a Cu foil reference sample), but another sample prepared in exactly the same way produced the same peak due to Cu2O at 934 eV. Therefore, it is better to compare the relative positions of the XPS peaks between each other than to compare their exact positions.
Fig. 2. XPS spectra of the (a) Cu 2p3/2, (b) C 1 s, and (c) O 1 s regions acquired under electrochemical control. The binding energy scale is not calibrated (see text). The cell potential is adjusted by applying a potential to the graphene/Cu WE while keeping the Pt CE grounded. Peak positions are shifted by the amount of the applied potential to the working electrode. Spectra shown in red and blue underline different polarisations. Grey dashed lines are used as eye guides for different peaks discussed in the main text. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
In Fig. 2a, initially the Cu 2p3/2 feature appears as a very broad peak at 935.5 eV when VWE − VCE is 0 V and 0.3 V. Here, the oxidation state of Cu is not well-defined due to storage in air, hence the broad peak. At VWE − VCE = 0.6 V, which is an oxidising condition, the Cu 2p3/2 peak is split into two peaks appearing at ∼ 934.2 eV (β peak) and ∼ 935.7 eV (α peak), and the peak remains split once the cell potential is lowered to 0.3 V. There is a negligible difference in XPS binding energy between the Cu 2p3/2 peaks of metallic Cu and Cu2O [24], [25]. The CuO chemical state produces a peak at 1–1.2 eV higher than the binding energy of Cu and Cu2O [24], [26]. Cu(OH)2, another possible chemical species in our system, produces a peak at over 2 eV higher binding energy than the peak of Cu and Cu2O [26]. Using these reference data, we can conclude that the β peak in Cu 2p3/2 spectra is due to either Cu, Cu2O, or their mixture. The α peak that appears ∼ 1.5 eV higher than the β peak is then a mixture of CuO and Cu(OH)2 chemical states. Cu2+ oxidation state obtained in cyclic voltammetry measurements in the literature was indeed interpreted as a mixture of these two chemical states [27], [28]. Upon changing the cell potential first to 0 V and then changing the polarisation to reducing conditions, the intensity of the α peak at higher binding energy diminishes compared to the other peak. This change indicates a reduction of the Cu2+ state back to Cu+ and/or Cu0.
The spectra shown in Fig. 2a are remarkable because the changing oxidation state of the surface of the Cu NPs under electrochemical control was detected directly with XPS at a solid–liquid interface. At positive VWE − VCE, the NP surface is composed of the Cu2+ species CuO, Cu(OH)2 (α peak, Fig. 2a), and a mixture of Cu2O and Cu0 (β peak, Fig. 2a). At negative VWE − VCE, the NP surface consists mostly of a mixture of Cu2O and Cu0. We can explain these results as the oxidation and reduction of the shell of the NPs respectively at positive and negative cell potentials, whereas the core of NPs remains reduced. Incomplete oxidation of the core of the Cu NPs can be related to the low molarity of NaOH used in this study.
“