Transparent Conductive Two-Dimensional Titanium Carbide Epitaxial Thin Films

Since the discovery of graphene, the quest for two-dimensional (2D) materials has intensified greatly. Recently, a new family of 2D transition metal carbides and carbonitrides (MXenes) was discovered that is both conducting and hydrophilic, an uncommon combination. To date MXenes have been produced as powders, flakes, and colloidal solutions. Herein, we report on the fabrication of ∼1 × 1 cm2 Ti3C2 films by selective etching of Al, from sputter-deposited epitaxial Ti3AlC2 films, in aqueous HF or NH4HF2. Films that were about 19 nm thick, etched with NH4HF2, transmit ∼90% of the light in the visible-to-infrared range and exhibit metallic conductivity down to ∼100 K. Below 100 K, the films’ resistivity increases with decreasing temperature and they exhibit negative magnetoresistance—both observations consistent with a weak localization phenomenon characteristic of many 2D defective solids. This advance opens the door for the use of MXenes in electronic, photonic, and sensing applications.


I. X-ray reflectometry and thickness determination
Films' thickness has been determined from XRR for Ti 3 AlC 2 films, before and after etching, deposited for 5 and 10 min, examples of the XRR data and their fittings shown in Figures S1, and S2. For Ti 3 AlC 2 film deposited for 30 mins, thicknesses before and after etching were obtained by direct measurement in TEM (Figures 2a-c, and Figure S9c). Figure S1 shows the measured X-Ray reflectometry for Ti 3 AlC 2 (black curve), and bestfitted simulation from TiC incubation layer/Ti 3 AlC 2 (red curve): for a film deposited for 5 min ( Figure S1a) giving a thickness of 15.2 ± 0.5 nm and a film deposited for 10 min ( Figure S1b) giving a thickness of 27.7 ± 0.8 nm. Figure S1c shows the relationship between the thickness of Ti 3 AlC 2 , Ti 3 C 2 T x and Ti 3 C 2 T x -IC and deposition time of Ti 3 AlC 2 . Films deposited for 20 min: their thickness before and after etching was obtained from interpolation ( Figure S3c). Figure S1. Measured X-Ray reflectometry for (a) Ti 3 AlC 2 sputtered for 5 min and, (b) Ti 3 AlC 2 sputtered for 10 min, Ti 3 AlC 2 (black curve), and best fit simulation for TiC incubation layer/ Ti 3 AlC 2 (red curve) and (c) Thickness vs. deposition time of Ti 3 AlC 2 for Ti 3 AlC 2 , Ti 3 C 2 T x and Ti 3 C 2 T x -IC.

II. XPS analysis of Ti 3 AlC 2 , Ti 3 C 2 T x , and Ti 3 C 2 T x -IC
Analysis of the high-resolution XPS spectra was performed through peak fitting using symmetric Gaussian-Lorentzian curves resting on a Shirley background. Fig. S2 presents the Ti 2p, C 1s, and Al 2p regions for the Ti 3 AlC 2 thin film together with the obtained Shirley background and Gaussian-Lorentzian curves for each region. In the Ti 2p region ( Figure S2a), which contains both the 2p 1/2 and the 2p 3/2 spin-orbit split components, the XPS spectrum could be best fit with four pairs of Gaussian-Lorentzian curves, where each pair is the 2p 1/2 and the 2p 3/2 component that we assign to Ti-Al, Ti-C, Ti(II) oxide, and Ti(III) oxide, respectively. [1][2][3][4] The C 1s region ( Figure S2b) could be fit with four Gaussian-Lorentzian curves. The low binding energy feature is a sharp asymmetric peak that is assigned to the Ti-C bond. 1,2 The asymmetry is due to extrinsic energy losses caused by delocalized states. This asymmetry, in turn, required the Ti-C XPS peak to be fit with two symmetric Gaussian-Lorentzian curves. In addition to the Ti-C peak there are two peaks assigned to surface hydrocarbon (-CH 2 -& -CH 3 ) and carboxylate (-COO) 5 contamination common for samples exposed to laboratory air. 6 In the Al 2p region ( Figure S2c) the low binding energy feature contains the 2p 1/2 and the 2p 3/2 spin-orbit split components assigned to Ti-Al. 3 The high binding energy feature is assigned to an aluminum oxide components, 2 fitted with a symmetric Gaussian-Lorentzian curve, due to surface oxidation which is common for MAX phase materials. 2 The results obtained from the peak fitting of the Ti 3 AlC 2 thin film are summarized in Table S1.

Figures S3a-d and Figures
S3e-h represent the Ti 2p, C 1s, O 1s, and F 1s regions for the Ti 3 C 2 T x thin film and the Ti 3 C 2 T x -IC thin film, respectively. For both MXene thin films, the XPS spectrum of the Ti 2p region (Figures S3a, and S3e) could be best fitted with five pairs of Gaussian-Lorentzian curves, where each pair is the 2p 1/2 and the 2p 3/2 component that we assign to Ti-C, Ti(II) oxide, Ti(III) oxide, Ti(IV) oxide, and Ti-F, respectively. [3][4][5][7][8][9] The C 1s region for both MXene thin films (Figures S3b, and S3f) shows features assigned to Ti-C, 1, 2 hydrocarbons (-CH 2 -& CH 3 -), and carboxylates (-COO). In this case intense contributions from graphite (C-C) and alcohol (C-O) formation are also present. 5 The O 1s region for both MXene thin films (Figures S3c, and S3g) could be fit by two components assigned to titanium oxide: one component for stoichiometric TiO 2 and one component for sub-stoichiometric TiO x . 9 The O 1s region also suggests Ti-OH formation and H 2 Figure S3. Deconvolution of high resolution XPS spectra for elements in Ti 3 C 2 T x , and Ti 3 C 2 T x -IC. The F 1s region for both MXene thin films (Figs. S3d and S3h) shows a dominating contribution from fluorinated titanium, 8 Ti-F, but also a small component assigned to aluminum fluoride, 10 Al-F, which is corroborated by the appearance of a weak feature at 73.6 eV in the Al 2p spectra shown in Figure 2b. The results obtained from the peak fitting of the Ti 3 C 2 T x thin film and the Ti 3 C 2 T x -IC thin film are summarized in Tables S2  and S3.   Table S2. Ti 3 C 2 T x XPS peak fitting result. Parameters obtained from the peak fitting of the Ti 3 C 2 T x thin film XPS spectra using symmetric Gaussian-Lorentzian curves.

BE [eV] a fwhm [eV] a Fraction Assigned to Reference
Ti 2p   The peak for the aluminum fluoride component shifts to a higher binding energy compared to that for Ti 3 C 2 T x , while its FWHM maximum increases, which may be due to the formation of (NH 4 ) 3 AlF 6 rather than AlF 3 as indicted by XRD in Figure S6.
The high-resolution XPS spectra of the Ti 2p and C 1s regions for the Ti 3 AlC 2 , Ti 3 C 2 T x , and Ti 3 C 2 T x -IC thin films show that the removal of Al causes a shift of the Ti-C contribution in the Ti 2p and C 1s XPS spectra toward higher binding energies, indicative of a loss of charge, and a concomitant charge redistribution within the material. Note that this decrease of charge is not due to the removal of Al atoms, because the latter needs to leave behind the charge they gained when forming Ti 3 AlC 2 . The charge redistribution is instead a consequence of the higher electronegativities of the OH, O and F surface functional groups, as mentioned in the main text.
In addition there is a slight shift to higher binding energies for the Ti 2p spectrum of the Ti 3 C 2 T x -IC thin film compared to the corresponding spectrum of Ti 3 C 2 T x . Since the matching C 1s spectra do not show any changes in the binding energies, the observed binding energy shift could be due to the intercalated nitrogen species (Figure 2e) interacting with the surface of the Ti 3 C 2 T x -IC thin film.
The peak fitting of the XPS high-resolution spectra suggests the chemical compositions to be Ti 3 C 2.2 O 2 F 0.6 and Ti 3 C 2.3 O 1.2 F 0.7 N 0.2 for Ti 3 C 2 T x and Ti 3 C 2 -IC, respectively. Fitted peaks that correspond to surface oxidation and contamination, i.e. TiO 2 , H 2 O, hydrocarbons, alcohols, carboxylates, and aluminum fluoride, are not included in these elemental quantifications. The obtained quantification data indicates that the etching of Ti 3 AlC 2 , using HF or NH 4 HF 2 , provides near-stoichiometric Ti 3 C 2 2D-stuctures, surface terminated with a mixture of fluoride-and hydroxyl groups. Additionally, etching in NH 4 HF 2 results in the intercalation of the nitrogen species, NH 4 + and NH 3 . 11,12

III. XRD analysis of the byproducts from etching Ti 3 AlC 2 with NH 4 HF 2
One half of a gram of Ti 3 AlC 2 powder (the method of preparation is described elsewhere 13 ) were soaked in 5ml of 1M NH 4 HF 2 solution at room temperature. The mixture was left untouched until the solvent evaporated. XRD diffraction of the dry powders ( Figure S4) indicates the existence of two byproduct compounds (NH 4 ) 3 AlF 6 (PDF# 22-1036) and AlF 3 .3H 2 O (PDF# 46-1459) comparing the intensity of the maximum peaks for both gives a ratio of about 7:1 respectively. It follows that the major byproduct from etching Ti 3 AlC 2 with NH 4 HF 2 is (NH 4 ) 3 AlF 6 . Figure S4: X-Ray diffraction pattern of Ti 3 C 2 T x intercalated MXene (Ti 3 C 2 T x -IC) after etching Ti 3 AlC 2 powder with 1M NH 4 HF 2 , and allowing the mixture to sit until the solvent dried. In contrast to our previous work, here the resulting salt was not washed away with water. The XRD pattern also includes peaks associated with un-reacted Ti 3 AlC 2 and TiC present as an impurity in the as-received powders.

IV. Intercalation and de-intercalation of Ti 3 C 2 T x
In order to verify the assumption that intercalation is the pertinent mechanism, Ti 3 C 2 powder (the method of preparation is described elsewhere 13 ) were immersed in 1 M of NH 4 F or in 5 M NH 4 OH for 24 h at room temperature while stirring. XRD patterns before and after treatment, shown in Fig. S5a, confirm that the increase in c, from 19.8 Å to 25 Å, in both solutions is similar to that observed when NH 4 HF 2 was used as etchant. It is thus reasonable to assume that, intercalant compound should be the common species between NH 4 HF 2 , NH 4 F and NH 4 OH, viz. NH 4 + . To investigate the reversibility of the intercalation process, a 43 nm nominal thickness Ti 3 C 2 T x -IC film was heated in a vacuum at 250 º C for 90 min. The XRD patterns for the film before and after de-intercalation are compared in Figure S5b. After the vacuum treatment, the (0002) peak shifts to an angle that corresponds to a c lattice parameter of 21 Å. It is thus possible to de-intercalate ammonia from Ti 3 C 2 T x -IC. This value is also quite similar to, the lattice parameter obtained from SAED pattern shown in Figure 3e, suggesting that the TEM sample was partially de-intercalated during the sample preparation.

V. Optimization of the etching process
The resistivities of all films increased with increase in etching times. Table S4 lists the resistivities of Ti 3 C 2 T x films produced by HF etching for samples, of the same nominal thicknesses, etched for different times. For example, the resistivities of Ti 3 AlC 2 films, 60 nm nominal thick, etched for 160 mins and 360 mins were 1.8 µΩm and 3.4 µΩm, respectively.
The etching time required to fully transform the Ti 3 AlC 2 films to MXenes was determined by repeatedly measuring the XRD diffraction patterns, of a given film, after successive etching steps until the XRD peaks belonging to Ti 3 AlC 2 disappeared ( Figure  S6). This procedure was adopted in order to compare the properties of all etched films at the point all the Al layers were etched out. Said otherwise, the etching times varied from film to film. The dependence is not linear, however. For example, the etching times for Ti 3 AlC 2 films of nominal thicknesses 15 and 28 nm were 10 and 15 mins, respectively. Similarly, at 150 min and 160 min, the time needed to fully etch 15 nm and 28 nm thick Ti 3 AlC 2 films in NH 4 HF 2 , were quite comparable (See Table 1).
A perusal of the results shown in Tables 1 and S4 suggest that the final resistivities one obtains is a complicated function of film thickness, etching times and the nature of the etchant. On the one hand, thin films are more resistive than their thicker counterparts. It is this variability -reflected in Table 1 -that is most probably responsible for some of the anomalies observed. For example, the etching times for Ti 3 C 2 T x films of nominal thicknesses of 15, 28 and 43 nm were 10, 15 and 60 min, respectively.   Figure S7 presents an EDX map showing the distribution of C, Ti, F, and O atoms over the corresponding high-resolution STEM image of Ti 3 C 2 T x produced by HF etching. The fluorine signal is concentrated primarily in the spaces between the Ti-C layers, which suggests that F atoms are attached to the surfaces of the Ti-C layers.  Figure S8a shows the morphology of a typical Ti 3 AlC 2 surface, where the hexagonal striations reflect to the symmetry of the basal planes together with some three-sided grains that are protruding out of plane due to their tilted orientations. Such nonbasal grains are preferentially etched, leading to premature over-etching and thus higher resistivities. Figure S8b shows the surface of Ti 3 AlC 2 film after etching (Ti 3 C 2 T x ). Pinholes have appeared near the tilted grains in the Ti 3 C 2 T x , presumably a result of the preferential etching of the tilted grains with respect to the grains parallel to the substrate surface. Figures S9a,b show TEM images of the tilted grains of Ti 3 AlC 2 and Ti 3 C 2 T x , respectively. In Figure S9a, Ti 3 AlC 2 grain has grown over Ti 2 AlC. The nonbasal grain nucleated on the substrate and overgrew the basal grains, since growth along basal planes is faster than normal to it. For the Ti 3 C 2 T x grain ( Figure S9b), defects appear in the interface between both the tilted and horizontal grains.

VII. Morphologies of Ti 3 AlC 2 and Ti 3 C 2 T x films
Low magnification cross-sectional TEM images of two Ti 3 C 2 T x films are shown in Figure S9c. The two samples are facing each other so that the surface of Ti 3 C 2 T x is inwards. The globules appearing between the two films are carbon particles coming from the glue being used to hold the samples together on the TEM grid. The Ti 3 C 2 T x films shown are uniform with some tilted grains ( Figure S9a).

IX. Electrical transport measurements and analysis
To understand what causes the low-temperature insulating behavior, there are numerous possible models to consider. We consider the following four models: i) thermally activated, in which case ρ ~ exp(E A /kT); ii) 3D variable range hopping for which ρ ~ exp(T 0 /T) 1/4 , iii) 2D variable range hopping for which ρ ~ exp(T 0 /T) 1/3 , and, iv) a weak localization model for which ρ ~ ln(T). [18][19][20][21] The low-temperature (< 75 K) resistivity data for Ti 3 C 2 T x films of 28 nm nominal thickness were fitted to all of the aforementioned models. Results of these fits are shown in Figures S10 and S11. The poor fit for the first three models to the experimental obtained ( Figure S11) suggest that they can be discarded. In contradistinction, the fit for the weak-localization model is nearly perfect for all samples (insets in Figure S10). The low-temperature behavior of the resistivity is thus consistent with the weak localization model, a phenomenon typically observed in 2D metallic films. [19][20][21] The negative magnetoresistance observed in the same temperature range is also consistent with the weak localization model 22 . This evidences a truly 2D behavior of the electronic transport properties of Ti 3 C 2 T x , in that the charge carriers are confined and weakly localized within individual Ti 3 C 2 T x layers. Figure S11. Temperature dependencies of electrical resistivities of a: (a) 60 nm thick Ti 3 C 2 T x film, (b) 28 nm thick Ti 3 C 2 T x -IC film. Insets in both figures show the fits of resistivities, in the 2 to 74 K temperature range, to the weak localization model, viz. ρ ~ ln(T). Figure S12. Fitting of resistivity of 28 nm nominal thickness Ti 3 C 2 T x films on temperature in the 2 to 74 K temperature range assuming: (a) a thermally activated process; (b) a 3D variable range hopping model; 18 (c) 2D variable range hopping model. 18