Resolving X-Ray Photoelectron Spectra of Ionic Liquids with Difference Spectroscopy

X-ray photoelectron spectroscopy (XPS) is a powerful element-specific technique to determine the composition and chemical state of all elements in an involatile sample. However, for elements such as carbon, the wide variety of chemical states produce complex spectra that are difficult to interpret, consequently concealing important information due to the uncertainty in signal identity. Here we report a process whereby chemical modification of carbon structures with electron withdrawing groups can reveal this information, providing accurate, highly refined fitting models far more complex than previously possible. This method is demonstrated with functionalised ionic liquids bearing chlorine or trifluoromethane groups that shift electron density from targeted locations. By comparing the C 1s spectra of non-functional ionic liquids to their functional analogues, a series of difference spectra can be produced to identify exact binding energies of carbon photoemissions, which can be used to improve the C 1s peak fitting of both samples. Importantly, ionic liquids possess ideal chemical and physical properties, which enhance this methodology to enable significant progress in XPS peak fitting and data interpretation. surface modification of low-density polyethylene (LDPE) thin films. 15 The C 1s XP spectrum of untreated LDPE was subtracted from the C 1s spectra of a range of chemically oxidised LDPE samples. The difference spectra were used to identify changes in chemical state and the area of the photoemissions were compared to oxygen at.% to assess peak fitting variables (e.g. lineshapes and FWHMs). Importantly, these overlooked experiments demonstrate that incremental structural modifications can produce consistent XP photoemissions that differ by discreet and quantifiable changes. This work develops upon this principle to provide the first targeted chemical functionalisation of samples for identification of XP photoemissions. respectively (relative to their [C n Py-2-Cl][NTf 2 ] C 1s photoemissions). studies have measured an average of 10% signal loss from sp 2 -hybridised carbon atoms due to the shake-up/off phenomenon. The measured shake-up/off signals for the [C n Py-2-Cl][NTf 2 ] ILs are 12.9% (n = 4) and 9.9% (n = 8), giving an average signal loss of 11.4%. The positive and negative signals therefore originate from a single carbon atom that has experienced shake-up/off losses, i.e. the sp 2 hybridised C 2 carbon atom.


Introduction
X-ray photoelectron spectroscopy (XPS) gives information regarding the chemical states of elements in a sample. [1][2][3] Different oxidation states and electronic environments can be identified and quantified for all elements (except for hydrogen and helium), 4 making XPS a powerful technique for surface analysis of inorganic and organic samples. Prominent areas that utilise XPS include solid-state materials, polymers, nanoscience, and more recently, ionic liquids (ILs). [5][6][7][8] When an element occupies multiple chemical states, such as the carbon 1s photoemission of organic structures, signals appear complex and unresolved. This phenomenon is inherent to XPS as the FWHM of core photoemissions are large compared to the binging energy (B.E.) ranges of most common chemical states. For this reason, peak fitting models are used to interpret the characteristic structures of convoluted photoemissions, imparting an uncertainty to the information obtained from complex spectra. 9,10 Many of the fitting models described to date assign peaks by conjecture or examining vast numbers of samples relative to each other to incrementally improve peak fitting paramteres. [11][12][13][14][15][16][17] Importantly, there is no absolute peak fitting method and expert analysts have warned that interpretation of XP spectra is likely to lead to mistakes. 10 Therefore, a reliable methodology is essential to extract valuable information from complex photoemission spectra.
As an ultra-high vacuum (UHV) based technique, most liquids would rapidly evaporate under XPS experimental conditions. However, the extremely low volatility of ionic liquids (ILs) have enabled the investigation of liquid phase processes (e.g. solvent-solvent and solvent-solute interactions) by XPS. [18][19][20][21][22] Investigations of IL surfaces have also produced a wealth of information regarding the liquid-gas interface, nanostructure, and surface enrichment of solutes, primarily due to the element specific nature of XPS. [23][24][25][26] Importantly, XPS studies of ILs are complimented by strong photoelectron fluxes that give rise to intense, narrow signals emitted by flat IL surfaces. Furthermore, ILs are electrically conducting, which prevents significant differential charging, and have an apparent high beam stability due to the dynamic liquid surface. 27 IL XP spectra are consequently exceptionally high quality and are often superior to the XP spectra of solid organic powders.
IL chemical structures produce complex C 1s photoemission spectra because of the diverse chemical states of carbon, which occupy both electron-rich aliphatic environments and electron-poor ionic environments. The complexity of IL C 1s spectra are dictated by the cation (i.e. covalent bonding, charge delocalisation), anion (i.e. charge transfer, presence of carbon), alkyl chain lengths, and presence of functional groups. 13,[28][29][30][31] Hence, accurate and reliable C 1s fitting models are needed to unlock all of the information present in an XP spectrum. However, due to the lack of standard procedures, different C 1s fitting models have been developed for even the most basic IL chemical structures. For example, the C 1s region of imidazolium ILs have been fitted with 2-3 component models, which broadly account for polar and non-polar regions, 32 or more complex fittings with 3+ components. 21,33 While the latter can potentially provide more information, simpler models are a conservative approach aimed at minimising over interpretation of XP spectra.
Post-analysis data interpretation is problematic for any complex system with multiple chemical states and often a limiting factor for the successful application of XPS as an analytical method. Although there are numerous methodologies and tools for determining the goodness of fit for XP spectra, most serve as error analyses to identify poor peak fittings and do not indicate correct peak assignments. [34][35][36] Peak fitting by intuition can also produce misleading results as XP spectra are a combination of initial and final state effects. XPS has significant final state effect contributions and previous publications have warned against interpreting B.E. shifts < 0.5 eV in IL XP spectra in terms of atomic charges. 32,37 Despite this, correlations between core-electron B.E.s and physicochemical properties (e.g. Kamlettaft parameters) 21,22,31 support the drive for accurate peak fittings, regardless of the physical interpretation of the data.
Difference spectra can be generated by subtracting one core-level XP spectrum from the same corelevel spectrum of another sample. This method can provide useful information when the samples are structurally related. A relevant example by Cremer et al. used the difference between two homologous imidazolium ILs to definitively identify the C 1s photoemission of a C 2 -methyl group. 22 15 The C 1s XP spectrum of untreated LDPE was subtracted from the C 1s spectra of a range of chemically oxidised LDPE samples. The difference spectra were used to identify changes in chemical state and the area of the photoemissions were compared to oxygen at.% to assess peak fitting variables (e.g. lineshapes and FWHMs). Importantly, these overlooked experiments demonstrate that incremental structural modifications can produce consistent XP photoemissions that differ by discreet and quantifiable changes. This work develops upon this principle to provide the first targeted chemical functionalisation of samples for identification of XP photoemissions.
ILs are considered neoteric designer solvents as their physicochemical properties can be tuned by chemical modification to improve their functions, i.e. they are task-specific (TSILs). 38 There are   numerous examples of TSIL XPS investigations, most focused towards characterising the impact of   functional groups on IL physical properties. 20,28,31 However, some studies have sought to utilise the designer aspect of ILs to expand XPS as an analytical tool. Prominent examples include monitoring gasliquid (e.g. CO 2 capture by amines) 39 43 Here, we exploit the tunable nature of ILs to effect electronic changes in the IL cation to facilitate XPS peak fittings. The ILs presented in this work are therefore task-specific for XP spectroscopic measurements; further support that their unique properties are perfectly complimentary to XPS.

Data Acquisition
All XP spectra were recorded using a Kratos Axis Ultra Spectrometer equipped with a monochromated Al Kα source (1486.6 eV), hybrid (magnetic/electrostatic) optics, concentric hemispherical analyser (CHA) and a multi-channel plate and delay line detector (DLD). The incident angle of the X-rays was 30° and the collection angle was 0 °, relative to the surface normal. The entrance aperture was 300 x 700 μm 2 and pass energies were set to either 80 eV for wide scans or 20 eV for high resolution scans.
The Ag 3d5/2 photoemission had an intensity of 7. photoemissions and the experimental error was determined by the manufacturer to be ± 0.1 eV.
Liquid samples were placed on to a stainless steel sample bar as single drops and degassed overnight in a sample transfer chamber (≈ 10 -7 mbar) before being moved to the analysis chamber (≤ 1 x 10 -8 mbar). Liquid samples were not charge neutralised as they are electrically conducting and therefore do not experience significant differential charging. 27

Data Analysis
All data sets were converted to VAMAS (.vms) format and imported to CasaXPS for quantification and peak fittings. A detailed description of the C 1s peak fitting procedures used in this work and previously published work is given in the ESI. The C 1s photoemission peaks are numbered by the carbon atom

Physical Chemistry Chemical Physics Accepted Manuscript
numbers, which are ordered by IUPAC priority rules (see Figure 2). All XP spectra presented in this study were charge corrected using existing procedures, whereby the C aliphatic (abbr. C ali ) components for long alkyl-chain ILs (≥ C 8 ) are set to 285.0 eV. The B.E. shifts are then used to correct the N 1s photoemissions and the resulting values are used to reference all XP spectra of the same cationic type.
All XP spectra were normalised by adjusting the areas of the non-functionalised IL F 1s photoemissions to equal the areas of the analogous functionalised IL F 1s photoemissions. For C 1s difference spectra, the normalisation was checked by comparing the -CF 3 C 1s photoemission from the [NTf 2 ]anions, and any deviations were omitted by adjusting the areas of the non-functionalised IL C 1s photoemissions until the signals were aligned (N.B. further normalisation was often required but the adjustments were relatively minor). In the absence of this reference (i.e. [BF 4 ] -ILs), spectra were normalised to their F 1s photoemissions and the C ali signals were used to fine-tune the normalisation in the same manner as described above.
The difference spectra reported here are plotted on a common X-axis and the data points are aligned to allow data subtraction. This work uses the functionalised IL B.E. axis, therefore non-functionalised XP spectra are referenced to the functionalised IL XP spectra. This is achieved by setting the -CF 3 signal maximum of the non-functionalised IL to the same B.E. as the functionalised ILs, or in the case of [BF 4 ]salts, the C ali component. All photoemissions are plotted on the functionalised IL x-and y-axes with the normalised non-functionalised IL photoemissions overlaid. Most y-axes therefore display arbitrary units and the difference spectra are generated on the functionalised IL scale. All area quantifications are relative to normalised spectra. A list of difference spectra and structural representation of the subtraction process are given in the ESI (Figure S2-6). The non-functionalised XP spectra reported here have been previously published; 22,27,30 the fully analysed spectra are shown in the ESI (Figure S25-29) for comparative purposes. In addition, the experimental elemental compositions and nominal stoichiometries for the ILs presented in this work are displayed in the ESI (Table S1).

Pyridinium ILs
Survey and high resolution XP scans of [C n Py-2-Cl][NTf 2 ] are shown in the ESI (Figure S11-14); both ILs produce high quality XP spectra with no evidence of impurities or beam damage. Furthermore, all photoemission signal B.E.s (see Table 1    Again, the IL produces high quality XP spectra with no evidence of impurities in the survey scan, and no signs of beam damage from the high resolution scans. The B.E.s are summarised in Table   1 and the C 1s photoemission is displayed in Figure 6   The survey and high resolution XP scans of [C n C 1 Im-4-Cl][NTf 2 ] are shown in the ESI (Figure S21-24); the ILs produce high quality XP spectra with no impurities or signs of beam damage. The B.E.s for Table   2. The C 1s photoemissions are also displayed in Figure 7a-b, along with the respective [C n C 1 Im][NTf 2 ] C 1s photoemissions and the resulting difference spectra (difference 8 and 9). As before, difference 8

Difference Spectra
The difference spectra reported herein were generated by subtracting the C signal loss previously measured. 27 The measured values support the single carbon assignments and provide further evidence that shake-up/off losses have been accurately calculated for both pyridinium and imidazolium IL C 1s photoemissions.

Significance of the data
For the non-functional pyridinium ILs, B.E.s provided by the difference spectra support previous peak assignments and strengthen the C 1s peak fitting model previously presented by our group (Figure   8a). 30 The symmetry of the pyridinium cation and the shape of their C 1s photoemission spectra make this assignment far easier than other cationic systems. In comparison, the EWG-functionalised pyridinium C 1s signals are not spread across a large B.E. range, and assignment of the coincident signals between 286-288 eV is far harder because of the lack of features (i.e. single peak maxima).
Fortunately, the difference spectra for both function and non-functional pyridinium ILs provide accurate B.E. signals that can be used to build precise C 1s peak fitting models. Importantly, the information obtained from the analytical procedure can benefit both C 1s spectra under consideration.
The fully fitted C 1s spectra for all pyridinium salts are presented in the supporting information.
For the non-functional imidazolium ILs, the negative C 4 peak position (286.5 eV) indicates the position of the signal before chlorine functionalisation. The C 1s B.E. shows that previous peak assignments are incorrect, and a new peak fitting model is required (see Figure 8b). In this model, the C 4 and C 5 photoemissions appear at lower B.E.s than the C 6 and C 7 carbon photoemission signals, suggesting that most of the positive charge is spread about the NCN portion of the imidazolium ring and around the N-C carbons of the pendant alkyl chains. The back of the imidazolium ring (i.e. the C 4 and C 5 carbons) therefore has a higher electron density than previously thought (through XPS interpretations). Further investigations are currently underway to investigate whether this new observation is in fact related to electron density or is a result of final state effects. Regardless of the interpretation, the new data obtained from the difference spectra has led to a refinement in the C 1s fitting model. Furthermore, the accurate B.E.s of the chlorine functionalised imidazolium salts have been presented and fully peak fitted XP spectra are presented in the supporting information. Figure 8 The C 1s high resolution XP spectra with peak fittings and assignments (a) [C 8

Conclusions
The difference spectra reported herein demonstrate that targeted functionalisation of ILs with EWGs such as chlorine and trifluoromethane can produce B.E.s shifts large enough to clearly show initial and final peak positions. The exact C 2 , C 3 , and C 4 C 1s photoemission B.E.s of the functional and nonfunctional pyridinium ILs presented in this work have now been experimentally determined, along with the C 4 C 1s photoemissions of the functional and non-functional imidazolium ILs. The obtained B.E. values have confirmed previous pyridinium C 1s peak fitting models and enabled refinement of the imidazolium peak fitting models. The accurate C 1s peak fittings of the functional pyridinium and imidazolium salts have been presented; without difference spectroscopy peak assignments would likely be incorrect due to the lack of defining features in each spectrum. Furthermore, quantification of the difference spectra supports previous shake-up/off losses of sp 2 hybridised carbons by directly measuring the photoemission signals, as opposed to indirect measurement of the shake-up signal which is subject to data analysis error (e.g. choice of background).
Overall, this work demonstrates that difference spectroscopy can significantly enhance XPS analysis, providing more reliable information than previously though possible. High quality XP spectra are required to produce high quality difference spectra. For this reason, and in combination with robust charge correction procedures, this work demonstrates that ILs are ideal small molecules for XPS difference spectroscopy. The success of this technique as a post data acquisition analytical method is unlike to be matched by other systems such as polymers.