Electronic and optical characterization of 2D Ti₂C and Nb₂C (MXene) thin films

Two-dimensional (2D) solids have been shown to be promising candidates for electronic and optical applications [1, 2]. For instance, graphene possesses ballistic conductivity and high electron mobility reaching 2  ×  10⁵ cm² V⁻¹ s⁻¹ at room temperature (RT) and transmits 97.7% of the light in the near-infrared (IR) to ultraviolet (UV) range [3]. This combination allows graphene to excel in optoelectronic applications such as transparent conductive electrodes which can be used for touch screens [4, 5]. Other 2D materials have different electronic properties such as the semiconducting MoS₂ which can be used in applications such as transistors, sensors, and memory devices [6, 7].

Recently, a new family of 2D materials based on transition metal carbides and nitrides, labelled 'MXene' was discovered [8, 9]. MXenes are mostly produced by etching the A element from the MAX phases. The latter are a family of hexagonal layered ternary transition metal carbides and/or nitrides with a general formula of M_n+1AX_n, where M stands for an early transition metal, A stands for mainly group 13 and 14 elements, X stands for carbon and/or nitrogen and n  =  1, 2, or 3 [10].

Various acidic solutions containing fluoride ions (HF, NH₄HF₂, LiF  +  HCl, HF  +  LiCl, or NaF  +  HCl) are used to selectively etch the A layers (usually Al, but more recently Ga and Si), to convert MAX phases into MXenes [9, 11–17]. The A layers are replaced with oxygen, hydroxyl and/or fluoride surface terminations, T, such that the proper MXene designation is M_n+1X_nT_z [18]. Using the appropriate solutions, MXenes can be delaminated as single flakes in colloidal suspensions [19–21]. MXenes were first synthesized in powder or colloidal suspension form. In 2014, the first MXene thin films, Ti₃C₂T_z, were produced by etching—using HF or NH₄HF₂ [12]—physical vapor deposited (PVD) films of the MAX phase Ti₃AlC₂ on sapphire substrates. Later, other methods were used to produce Ti₃C₂T_z thin films, such as spin coating, spray coating and electrohydrodynamic atomization [22–25].

MXenes have the potential to be used for various applications, such as battery electrodes, supercapacitors, fuel cells, transparent conducting electrodes, photocatalytic applications, water treatment, electromagnetic shielding, gas sensors, and biosensors among many others [26–28]. For most, if not all, of these applications it is crucial to understand electron transport, a topic that is still very much a work in progress since there are many variables to consider, including the nature of the terminations, the multiple chemistries and defects at various levels. To put the difficulty in perspective we note that electron transport situation for the relatively much simpler MAX phases is still far from being understood since the transport depends on the shape of the Fermi surfaces, among other variables [29].

In general, transport in Ti₃C₂T_z is metal-like, with a resistivity, ρ, that decreases linearly with decreasing temperatures. This response is not only true for thin films, but also for single flakes [12, 30].

Before proceeding much further it is instructive to review the variable range hopping, VRH, mechanism since it has been shown to be applicable to several MXenes, including Nb₂CT_z explored herein. In the VRH model the DC conductivity, σ, is assumed to vary as

$$\begin{align} \newcommand{\e}{{\rm e}} \displaystyle \sigma ={{\sigma }_{0}}{\rm exp}-{{\left( \frac{{{T}_{0}}}{T} \right)}^{x}}\nonumber \end{align} \tag{ 1 }$$

where T is the absolute temperature, σ₀ is a conductivity pre-factor and T₀ a characteristic temperature with an exponent x that depends on dimensionality, d, of the transport, and a term p  characterizing the variation of the density of states at the Fermi level D(E_F) with energy [31, 32]. p  and d are related by

$$\begin{align} \newcommand{\e}{{\rm e}} \displaystyle x=\left(\,p+1 \right)/\left( d+p+1 \right).\nonumber \end{align} \tag{ 2 }$$

If, as assumed here, the D(E_F) is not a strong function of energy, then p   =  0 and x reduces to 1/4 for d  =  3, and 1/3 for d  =  2 [33]. In the Efros–Shklovskii VHR model, where prominent charging or Coulomb effects are operative [32], x is equal to 1/2 and is independent of d [34].

This brief review of VRH is important for understanding electron transport in MXenes other than Ti-based. For example, in Mo-containing MXene thin films ρ increases with decreasing temperatures and was thus termed 'semiconductor-like' [16, 35, 36]. Recently we measured σ of Mo₂CT_z, Mo_1.33CT_z, Mo₂TiC₂T_z and Mo₂Ti₂C₃T_z thin films—made by filtering colloidal suspensions—in the 10–300 K temperature range and concluded that, with the exception of the heavily defective Mo_1.33CT_z composition, VRH, between individual 2D sheets is the operative conduction mechanism, especially at the lower temperatures [37]. Most importantly, for this work, ρ of the highly defective Mo_1.33CT_z composition, was also concluded to be by VRH, however, in that case, it was surmised that the VRH within individual flakes was rate limiting. This is important because this is the conclusion reached herein concerning transport in our Nb₂CT_z films.

The motivaton for this work is two fold. One is scientific in that we are trying to systematically understand electronic transport in MXene films [12, 16, 36, 37]. From a practical point of view, we have shown that spincast transparent conductive Ti₂CT_z films on glass were as promising as Ti₃C₂T_z or undoped chemically vapor-deposited graphene films [46]. Conversely, Nb₂CT_z has attracted some interest recently since it has shown promise in several applications such as Li-ion battery electrodes [21, 38], a photocatalyst for hydrogen evolution [39], photothermal tumor eradication [40], and support for platinum catalysts [41], among others. To date, and as far as we are aware, there have been no experimental studies on the electronic and optical properties of Nb₂CT_z. The only systematic study on a Nb-containing MXene was recently reported by Halim et al who measured the temperature dependencies of the resistivity of thin films of heavily defective Nb_1.33CT_z films [42]. Here again it was concluded that the VRH model was applicable.

In this report, physically vapor-deposited epitaxial Ti₂AlC and Nb₂AlC thin films grown on sapphire substrates were converted to their respective MXenes, viz Nb₂CT_z and Ti₂CT_z, by etching them using a mixture of LiF and HCl. We subsequently measured the electronic and optical properties and showed them to depend on the transition metal.

2.1. Deposition of Ti₂AlC and Nb₂AlC

2.2. Synthesis of Ti₂CT_z and Nb₂CT_z

2.3. Structural and chemical characterization

2.4. Optical and electrical characterization

Figure 1 shows XRD diffractograms of as-deposited and etched Ti-based (figure 1(a)) and Nb-based (figure 1(b)) films. Consistent with previous work, the as-deposited films are epitaxial since they (bottom curves in figure 1) exhibit only 0 0 l (l  =  2, 4,6, ...) peaks [43, 44]. The corresponding c lattice parameters, LPs, are 13.61 Å and 13.85 Å for Ti₂AlC and Nb₂AlC, respectively, again in agreement with previous work [43, 45]. After etching (top curves in figure 1), the 0 0 2 peaks are shifted to lower angles and all other 0 0 l peaks are so reduced in intensity that they are barely visible (figure S6 for Ti₂CT_z–y Li and figure S7 for Nb₂CT_z–y Li).

Zoom In Zoom Out Reset image size

The etching therefore increases the interlayer spacing, δ, (=c/2) from 6.805 Å to 11.35 Å for the Ti-based films and from 6.925 Å to 11.65 Å for the Nb-based films. In general, the large expansion in δ is a characteristic of the MAX to MXene transformation [9, 38]. More specifically, it is typical for LiF etched MXene. The expansion is due to the intercalation of water and Li between the 2D carbide layers [11, 13, 22, 46]. It follows that the proper designation of our films should be: Ti₂CT_z–y Li and Nb₂CT_z–y Li, with T_z represents the various termination and -y Li denotes intercalated Li. Evidence for the presence of Li is given below.

Figure 2 shows HAADF-STEM images of the Ti- and Nb-based films of 40 and 20 nm thick, respectively. Figures 2(a) and S8 are a typical overview image of Ti₂CT_z–y Li on top of the Ti₂AlC–TiC mixed incubation layer grown on the Al₂O₃. Figure 2(b) shows atomically resolved Ti₂CT_z–y Li layers. Figures 2(c) and S8 are typical overview images of the Nb₂CT_z–y Li films and figure 2(d) shows the atomically resolved image of the Nb₂CT_z–y Li films. At 11.4  ±  0.2 and 11.8  ±  0.3 Å, the δ values for the Ti and Nb-based films, respectively, estimated from the TEM micrographs, are in good agreement with the values obtained from XRD. The overall chemical composition of the films, determined by XPS measurements, were found to be Ti₂CO_0.4(OH)_0.5 F_0.6 · 0.1H₂O–0.2Li and Nb₂CO_0.5(OH)_0.8F_0.4 · 0.3H₂O–0.3Li. Detailed analysis of the XPS data is given in the supplementary information (stacks.iop.org/JPhysCM/31/165301/mmedia). As typical of most MXenes, both contain mixed terminations of comparable ratios.

Zoom In Zoom Out Reset image size

As in our previous work on Ti₂CT_z and Nb₂CT_z cold pressed from multilayers, the sum of the terminations, z, are  ≈1.5  ±  0.2 and 1.7  ±  0.2, respectively [18]. When all termination sites are occupied, the total of the O, OH and F terminations should be 2 [47]. The fact that here z here is  <2, implies that not all M atoms are terminated. The exact reason for this state of affairs is unclear at this time. The most likely explanation is that some of the terminations desorb in the high vacuum of the XPS. These comment notwithstanding, thermodynamically, it is almost impossible for these terminations to be absent when the films are in air and the assumption made hereafter is that all M atoms are terminated during characterization in air.

Figures 3(a) and (c) plot the temperature dependencies of ρ for Ti- and Nb-based films, respectively. Figure 3(c) also plots the results of previous MXene work in which T₀ was determined to be  >10⁴ (see below) [37, 42]. The former exhibits metallic behavior, with a linear increase in ρ with temperature in the range from ~80 to 300 K temperature range. The RT resistivity, ρ_RT, is ~4.4 μΩ · m. Below 50 K, ρ increases weakly with decreasing temperature, which can possibly be attributed to weak localization in this 2D metallic material.

Zoom In Zoom Out Reset image size

This is indirectly evidenced by the fit to a weak-localization model, WL, shown in figure 3(b). The metallic behavior at high temperatures and WL at low temperatures for Ti₂CT_z–y Li is reminiscent of that of Ti₃C₂T_z thin films, made from sputter-deposited epitaxial films [12].

In sharp contrast, ρ of Nb-based film increases more than two orders of magnitude in the 5 K to 300 K temperature range (figure 3(c)). In this case, at 414 μΩ · m, ρ_RT is roughly two orders of magnitude higher than ρ_RT ~ of the Ti-based films. At 50 K the difference is more than four orders of magnitude. Note, no results were obtained at T  <  35 K because the resistances of the epitaxial films were simply too high given their extreme thinness (≈20 nm).

As done previously [16, 37], the transport mechanism was evaluated by fitting the transport data to several transport models including simple thermal activation (figure S3(a)) a power law model (figure S3(b)) and various VRH (equation (1)) models. A comparison of the various fits indicated that the 3D VRH mechanism showed the best fit (figure 3(d)) over the 40 to 300 K temperature range. As discussed in more detail in our previous work, in general it is difficult to distinguish between the various VRH mechanisms, viz between, the 3D, Efros–Shklovskii, or 2D models, the fits for which are shown in figures 3(b), S3(c) and S3(d), respectively. In this case, what can be ruled out are the Efros–Shklovskii VRH and the thermal activation models. The fit for the 3D model is ever so slightly better than the 2D, which is why it is the one shown in figure 3(d). However, based on these results we simply cannot differentiate between power-law models, 2D or 3D VRH ones (figures 3(d), S3(b) and (d)).

As noted above, figure 3(c), in addition to plotting the results obtained here also plots the results for Mo_1.33CT_z and Nb_1.33CT_z thin filtered films. The similarities between the four is somewhat surprising given that the results for Mo_1.33CT_z [37, 48] and Nb_1.33CT_z (figure S4) were obtained on thin filtered films, whereas those obtained here were obtained on etched epitaxial thin films.

Table 1 summarizes the fitting parameters, T₀ and σ₀ (see equation (1)) for the Nb-films tested herein, together with those of other films we tested for which T₀ is  >10⁴ K [37]. The parameters were determined for x  =  1/4 and 1/3 that correspond to the 3D and 2D VRH models, respectively.

In our previous work, we argued that when T₀  <  10⁴ K, the most likely rate limiting step is flake-to-flake transport [37]. We also made the case, that when T₀ was  >10⁴ K, the more likely scenario is one where transport within individual flakes is rate limiting.

In the VRH model, this conclusion was reached by first appreciating that in Mott's VRH model, T₀, is related to the density of states at the Fermi level, D(E_f), by

$$\begin{align} \newcommand{\e}{{\rm e}} \displaystyle {{T}_{0}}=\frac{a}{{{k}_{{\rm B}}}D\left( {{E}_{{\rm F}}} \right){{\gamma }^{-3}}}\nonumber \end{align} \tag{ 3 }$$

where 1/γ is the decay length or reciprocal tunneling exponent of the localized states, a is a constant initially estimated by Mott to be 16, which later, more accurate, numerical calculations showed to be 64 in case D(E_F) does not vary over a range  ≈k_BT [49]. In case of an exponential DOS, a was found to exceed 300 [49]. In our previous work, we assumed a range of γ and estimated the D(E_F) using the experimental values of T₀. By doing so we showed that since in many of our films, T₀ was  <10⁴ K, the jump distances were consistent with interflake conductivity.

Here we approach the problem slightly differently; we assume a  =  310, D(E_f) to be 1.5  ×  10²² eV⁻¹ cm⁻³—i.e. that of a decent metal—and making use of the experimentally determined values of T₀ (table 1) we calculate γ to obtain a sense for the magnitude of the jump distances involved. It is important to note that the choice of a  =  64 or 310, does not affect much the calculated values of γ. For example for Nb₂CT_z–y Li: assuming 3D VRH, for a  =  64, γ  =  0.31 nm while for a  =  310, γ  =  0.52 nm. For simplicity's sake, in the remaining of this discussion we will assume the operative mechanism to be 3D VRH. Column 4 in table 1 lists the values of γ obtained. The value obtained, ≈0.52 nm, is reasonable and is consistent with a system in which the conductivity is rate limited by hopping within individual flakes, rather than between flakes as is the case when T₀ is low. For example, if T₀ is 10 000, γ is  ≈3 nm, which is double the typical interlayer distance in MXene. Similarly, for Mo_1.33CT_x, the values obtained for γ, 0.34 and 0.18 nm, are consistent with a system where the conductivity is dominated by hopping within individual flakes. As for Nb_1.33CT_z, the value for γ is 1.12 nm which is close to the interlayer distance in MXene and the size of the cluster vacancies in the flakes, thus in this case the conductivity could be dominated by either interflake or intraflake hopping. Lastly, it is important to note that given the high resistivities measured it is reasonable to assume that the charge carriers percolate between flakes acting in series with the hops within flakes. Without this assumption, no current would be measured. This assumption is thus consistent with the significantly higher ρ measured for the Nb-films compared to their Ti-based counterparts.

Figures 4(a) and (b) plot the magnetoresistances, MRs, of the Ti- and Nb-based films as a function of applied magnetic field at 10 K, respectively. The negative MR for Ti₂CT_z–y Li (figure 4(a)) is reminiscent of that of Ti₃C₂T_z thin films etched in HF or NH₄HF₂ [12], and is consistent with WL. In contrast, the MR of Nb₂CT_z–y Li (figure 1(b)) is positive. In that respect it is similar to that for Mo₂CT_z, Mo₂TiC₂T_z and Mo₂Ti₂C₃T_z (MXenes) [16, 35], but somewhat surprisingly, opposite in sign to Nb_1.33CT_z and Mo_1.33CT_z [37, 42]. This reversal in sign in MR is clearly important but at this stage is not understood. More theoretical work is indicated, although the problem has proven to be anything but trivial.

Zoom In Zoom Out Reset image size

Figures 5(a) and (b) display the dielectric functions—determined from analysis of spectroscopic ellipsometry data in the 0.73 eV to 3.34 eV spectral range—for the Ti and Nb-based films, respectively. Straight lines represent results from model-dielectric-functions (MDF) based analysis employing several parametrized contributions (dashed lines). Further details on the data analysis are given in the supplementary information.

Zoom In Zoom Out Reset image size

For Ti₂CT_z–y Li, an absorption feature is found in the visible and a strong decrease of ${{\boldsymbol{\varepsilon }}_{{\bf 1}}}$ and an increase of ${{\boldsymbol{\varepsilon }}_{{\bf 2}}}$ is observed towards the infrared, IR, spectral range. This behavior is qualitatively similar to the corresponding MAX phase Ti₂AlC [50] and Ti₃C₂T_z–y Li MXene [22]. The absorption coefficient $\alpha$ ranges between $1.9\times {{10}^{5}}\ {\rm c}{{{\rm m}}^{-1}}$ and $3.0\times {{10}^{5}}\ {\rm c}{{{\rm m}}^{-1}}$, and the value at 550 nm, ${{\alpha }^{550}}=2.22\times {{10}^{5}}\ {\rm c}{{{\rm m}}^{-1}}$, is in excellent agreement with theoretical values [51] and experimental observations for Ti₂CT_z–y Li [46] and Ti₃C₂T_z–y Li [22]. The behavior towards the IR is expected for metals and is modeled by a Drude MDF [52]. The resistivity parameter was determined as 5.89 μΩ · m, which is comparable with the RT value obtained from the electrical measurements, viz 4.4 μΩ · m. Furthermore, the model includes a Lorentz oscillator and a UV-pole for UV absorptions to adequately model the experimental data.

In comparison, the Nb-based film exhibits significantly different behavior. Here both ${{\boldsymbol{\varepsilon }}_{{\bf 1}}}$ and ${{\boldsymbol{\varepsilon }}_{{\bf 2}}}$ increase towards the IR spectral range indicating an absorption in the near IR. In contrast to Ti-based film this behavior differs significantly from that of the corresponding MAX phase Nb₂AlC [50]. The absorption coefficient $\alpha$ ranges between 2.5  ×  10⁵ cm⁻¹ and 5.7  ×  10⁵ cm⁻¹ in the investigated spectral range, with a value of ${{\alpha }^{550}}=5.7\times {{10}^{5}}\ {\rm c}{{{\rm m}}^{-1}}$ at 550 nm. The near IR absorption is modeled by a Lorentz MDF. While the Lorentz MDF describes ${{\boldsymbol{\varepsilon }}_{{\bf 2}}}$ sufficiently well over the whole measured range, an offset mainly in ${{\boldsymbol{\varepsilon }}_{{\bf 1}}}$ remains. This offset decreases with increasing photon energy and can possibly originate from different mechanisms, for example further IR absorptions or free charge carriers that behave unlike a metal, e.g. carriers with a high mobility and low concentration. In addition, carrier localization and hopping may also play a role. For consistency we use Drude MDF to model the offset. The data analysis provides a resistivity value of 0.1 μΩ ⋅ m, which differs significantly from values from electrical measurements. However the uncertainty of the resistivity value is high due to depolarization in the sample. In addition, carrier localization and hopping may also play a role. The exact nature of the free charge carrier properties in Nb₂CT_z–y Li cannot be determined from the experimental data acquired here and more detailed work is needed.

Since both MXenes contain mixed terminations of comparable ratios, and, as importantly, comparable interlayer distances, our results show that the electrical and optical properties are strong functions of the transition metal (Ti versus Nb). For that reason, we suggest that the difference in the electronic and optical properties arises from the electronic configuration of the M elements, i.e. Ti is in group 4 and Nb is in group 5 and their bonding with the C atoms and surface terminations. This addition of an extra valence electron is proposed to cause the difference in electronic transport properties from metallic behavior to VRH transport. The 2D nature of the properties of Ti₂CT_z–y Li is evidenced by the low-temperature resistivity and MR. The low-temperature behavior of Nb₂CT_z–y Li is more complex, but consistent with VRH models.

The MXenes Ti₂CT_z–y Li and Nb₂CT_z–y Li were produced by etching Ti₂AlC and Nb₂AlC epitaxial thin films deposited by PVD on sapphire substrates and their structure, surface terminations, and electrical and optical properties were characterized. XPS of the films revealed their chemistry to be, respectively,

$$\begin{align} \newcommand{\e}{{\rm e}} \displaystyle \begin{array}{@{}cccccccccccccccccccc@{}} & {\rm T}{{{\rm i}}_{{\rm 2}}}{\rm C}{{{\rm O}}_{0.{\rm 4}}}{{\left( {\rm OH} \right)}_{0.{\rm 5}}}{{{\rm F}}_{0.{\rm 6}}}\cdot 0.{\rm 1}{{{\rm H}}_{{\rm 2}}}{{{\rm O}}_{{\rm ads}}}{{\rm {-}}}0.{\rm 2Li}\quad {\rm and} \nonumber \\ & {\rm N}{{{\rm b}}_{{\rm 2}}}{\rm C}{{{\rm O}}_{0.{\rm 5}}}{{\left( {\rm OH} \right)}_{0.{\rm 8}}}{{{\rm F}}_{0.{\rm 4}}}\cdot 0.{\rm 3}{{{\rm H}}_{{\rm 2}}}{\rm O{{-}}}0.{\rm 3Li}. \nonumber \\ \end{array}\nonumber \end{align}$$

The Ti-based films exhibit typical metallic behavior with a room temperature ρ ~ 4.4 μΩ · m. The optical properties determined from spectroscopic ellipsometry measurements confirm the metallic behavior, where ${{\boldsymbol{\varepsilon }}_{{\bf 1}}}$ decreases and ${{\boldsymbol{\varepsilon }}_{{\bf 2}}}$ increases. The optically determined resistivity parameter is close to the one determined from the electrical measurements.

In contrast, ρ of the Nb-based films increase with decreasing temperature. With a room temperature ρ of 414.1 μΩ · m, this film is two orders of magnitude more resistant than its Ti-based counterpart. At 50 K the difference is  ≈4 orders of magnitude. Analysis of the ρ versus T, results were found to be consistent with a 3D VRH mechanism, where hopping within individual flakes was rate limiting. At this stage, however, neither the 2D model nor a power law model can be eliminated.

This implies that the flakes are quite defective, the nature of which is unclear at this stage. The higher ρ values for the Nb-based films, are ascribed to a percolation between flakes. The optical response of the Nb-based film is also different than its Ti-based counterpart, since both ${{\boldsymbol{\varepsilon }}_{{\bf 1}}}$ and ${{\boldsymbol{\varepsilon }}_{{\bf 2}}}$ increase towards the IR spectral range indicating an absorption in the near IR.

The present work therefore clearly demonstrates how the electronic and optical properties of MXenes can be tailored, by the choice of the transition metal.

We thank Dr Steve May for access to the Physical Properties Measurement System and Professor T Ouisse for many fruitful discussions. The authors acknowledge the Swedish Research Council for funding under grant nos. 642-2013-8020, 621-2014-4890, 2016-04412 and 2016-00889. The Knut and Alice Wallenberg (KAW) Foundation is acknowledged for support of the electron microscopy laboratory in Linköping, and Fellowship grants. The authors also acknowledge Swedish Foundation for Strategic Research (SSF) through project funding (EM16-0004), Future Research Leader program FL12-0181, and the Research Infrastructure Fellow programs no. RIF 14-0074 and RIF14-055. The authors finally acknowledge support from the Swedish Government Strategic Research Area in Materials Science on Functional Materials at Linköping University (Faculty Grant SFO-Mat-LiU No 2009 00971).

There are no conflicts to declare.