Control of MXenes’ electronic properties through termination and intercalation — TEI-tagged view

James L. HartpersName, Kanit HantanasirisakulpersName, Andrew C. LangpersName, Babak AnasoripersName, David PintopersName, Yevheniy PivakpersName, J. Tijn van OmmepersName, Steven J. MaypersName, Yury GogotsipersName, Mitra L. TaheripersName (2019date). “Control of MXenes’ electronic properties through termination and intercalation.” Nature Communicationstitle·j.

Object of study —

Research question —

Contribution —

Full text — tagged

The reference list is not part of the text we hold for this paper. Our copy was extracted from the publisher's page and the bibliography did not survive that extraction, so the article below ends before its references. Nothing has been removed by us.

Abstract. MXenesterm·dom are an emerging family of highly-conductive 2D materialsterm·dom which have demonstrated state-of-the-art performance in electromagnetic interference shielding, chemical sensing, and energy storage. To further improve performance, there is a need to increase MXenesterm·dom’ electronic conductivityterm. Tailoring the MXeneterm·dom surface chemistryterm could achieve this goal, as density functional theory predicts that surface terminations strongly influence MXenesterm·dom' Fermi level density of states and thereby MXenesterm·dom’ electronic conductivityterm. Here, we directly correlate MXeneterm·dom surface de-functionalizationterm with increased electronic conductivityterm through in situ vacuum annealingterm, electrical biasingterm, and spectroscopic analysis within the transmission electron microscope. Furthermore, we show that intercalation can induce transitions between metallic and semiconductor-like transport (transitions from a positive to negative temperature-dependence of resistance) through inter-flake effects. These findings lay the groundwork for intercalation- and termination-engineered MXenesterm·dom, which promise improved electronic conductivityterm and could lead to the realization of semiconducting, magnetic, and topologically insulating MXenesterm·dom. Two-dimensional transition metal carbides and nitrides (MXenesterm·dom) have emerged as highly conductive and stable materials, of promise for electronic applications. Here, the authors use in situ electric biasing and transmission electron microscopy to investigate the effect of surface termination and intercalation on electronic properties. Introduction Discovered in 2011, MXenesterm·dom are a rapidly growing family of 2D transition metal carbides, nitrides, and carbonitrides with the general formula M n +1 X n T x ( n = 1, 2, or 3; M = transition metal, e.g., Ti, V, Nb, Mo; X = C and/or N; T = surface termination, e.g., –OH, –F, =O) 1 – 5 . MXenesterm·dom are formed by selective etching parent ternary carbide MAX compounds to remove the A-group element, e.g., Ti 3 AlC 2 (layered MAX) → Ti 3 C 2 T x (2D MXeneterm·dom) 6 , 7 . For both device applications and fundamental studies, MXeneterm·dom samples are generally thin films comprised of many MXeneterm·dom flakes, though some studies have focused on single-layer MXeneterm·dom 8 , 9 . In contrast to most other 2D materialsterm·dom, MXenesterm·dom offer an attractive combination of high electronic conductivityterm, hydrophilicity, and chemical stability 1 – 5 , 10 – 13 . With these properties, MXenesterm·dom show exceptional promise in areas including electromagnetic interference shielding 14 , 15 , wireless communication 16 , chemical sensing 17 – 20 , energy storage 21 – 23 , optoelectronics 24 – 27 , triboelectrics 28 – 30 , catalysis 31 – 33 , and conformal/wearable electronics 34 . Performance in these applications is directly related to electronic conductivityterm, and as such, there is motivation to further increase the MXeneterm·dom metallic conductivity. In parallel, there is excitement surrounding the potential realization of semiconducting MXenesterm·dom, which are predicted to be excellent materials for spintronics and thermoelectrics 35 – 37 . To meet these demands on MXenesterm·dom’ electronic properties, researchers have mostly focused on the development of new M n +1 X n chemistries. So far, over 30 MXenesterm·dom have been synthesized, but the first discovered MXeneterm·dom—Ti 3 C 2 T x —remains the most conductive 13 . Recently, certain Mo- and V-based MXenesterm·dom have garnered interest due to their so-called semiconductor-like behavior 38 – 40 , i.e., a negative temperature-dependence of resistance (d R /d T ). However, the cause of negative d R /d Tterm in these MXenesterm·dom remains unclear, and their underlying electronic structure is debated 38 , 41 . A potentially more useful approach to control MXenesterm·dom’ conductivity is to manipulate their surface chemistryterm. Surface terminations, which are introduced during MXeneterm·dom synthesis 6 , have been predicted to control metal-to-insulator transitions 37 , 38 and to affect functional properties such as magnetism 1 , 42 , Li-ion capacity 43 , catalytic performance 31 , 44 , band alignment 45 , mechanical properties 46 , and predictions of superconductivity 47 . While promising, these predicted effects lack direct experimental confirmation 48 , 49 . An additional mechanism which can affect MXeneterm·dom conductivity is intercalation. Intercalantsterm are not thought to alter the intra-flake (intrinsic) MXeneterm·dom properties, but for multi-layer samples, intercalation can increase device resistance by over an order of magnitude 27 , 40 , 41 , 50 – 52 . This effect is generally attributed to intercalantsterm increasing the inter-flake spacing and thereby the inter-flake resistanceterm. For any multi-layer MXeneterm·dom sample, intercalation, termination, and M n +1 X n chemistry all contribute to the measured electronic conductivityterm, and this convolution of effects greatly complicates experimental interpretation. As a result, our understanding and ability to control MXenesterm·dom’ electronic properties are lacking. To address this challenge, we perform in situ vacuum annealingterm (up to 775 °C) and electrical biasingterm of MXenesterm·dom within the transmission electron microscope (TEM)term. We observe de-intercalationterm and surface de-functionalizationterm with in situ electron energy loss spectroscopy (EELS)term and ex situ thermogravimetric analysis with mass spectroscopy (TGA-MS)term. Importantly, we utilize low-dose direct detection (DD) EELSterm 53 to avoid electron beam-induced specimen damageterm 54 (Supplementary Figure 1 ). With this approach, we correlate the desorption of –OH, –F, and =O termination species with increased MXeneterm·dom conductivity. Additionally, we report transitions from ensemble semiconductor-like (negative d R /d Tterm ) to metallic behaviorterm in Ti 3 CNT x and Mo 2 TiC 2 T x after the de-intercalationterm of water and organic molecules. This work furthers our fundamental understanding of conduction through MXeneterm·dom films and opens the door to intercalation- and termination-engineered MXenesterm·dom. Results Sample synthesis and experimental approach We investigated three MXenesterm·dom with diverse macroscopic electronic transport behavior: Ti 3 C 2 T x , Ti 3 CNT x , and Mo 2 TiC 2 T x (Table 1 ). Ti 3 C 2 T x (its structure is shown schematically in the Fig. 1a top inset) is the most studied MXeneterm·dom and is known to be metallic and highly conductive 7 . Density functional theory (DFT) studies have consistently predicted that surface functionalizationterm reduces the Ti 3 C 2 T x density of states (DOS) at the Fermi level ( E F )term, suggesting a decrease in the charge carrier densityterm and thereby a decrease in the conductivity 43 , 55 , 56 . Understanding of the next two MXenesterm·dom—Ti 3 CNT x and Mo 2 TiC 2 T x —is limited. The structure of Ti 3 CNT x is similar to Ti 3 C 2 T x but with a mixture of C and N on the X-sites (Fig. 1b top inset). DFT predicts Ti 3 CNT x to be metallic for all terminations 55 , 57 , 58 , but to date, this MXeneterm·dom has only shown semiconductor-like transport (unpublished results). Mo 2 TiC 2 T x is an ordered, double transition metal MXeneterm·dom analogous to Ti 3 C 2 T x but with the outer Ti layers replaced by Mo layers 59 (Fig. 1c top inset). Terminations have been predicted to induce metallic, semiconducting 38 , and topologically insulating 60 states in this MXeneterm·dom. Experimentally, Mo 2 TiC 2 T x shows semiconductor-like behavior in its as-prepared state 38 , but it is unclear if this behavior is due to intrinsic (M n +1 X n T x stoichiometry) or extrinsic (intercalation, inter-flake hopping) effects. Table 1 Summary of in situ heating and biasing results MXeneterm·dom chemistry Termination Intercalation d R /d T Resistance (Ω) Ti 3 C 2 T x (OH) 0.4 F 0.4 O 0.5 → F 0.2 O 0.5 H 2 O, Li + → Li + M → M 41 → 10 Ti 3 CNT x (OH) 0.9 F 0.5 O 0.7 → F 0.2 O 0.7 H 2 O, Li + → Li + S → M 159 → 27 Ti 3 CNT x (TBA + ) (OH) 0.6 F 0.1 O 1.2 → O 1.2 H 2 O, TBA + → none S → S 3330 → 290 Mo 2 TiC 2 T x (OH) 0.5 F 0.01 O 1.5 → O 0.8 H 2 O, TBA + → none S → M 2500 → 387 In each cell, the arrow symbol represents changes induced during in situ vacuum annealingterm up to temperatures of 775, 700, 750, and 775 °C for Ti 3 C 2 T x , Ti 3 CNT x , Ti 3 CNT x (TBA + ), and Mo 2 TiC 2 T x , respectively. In the d R /d T column, M indicates metallic conduction (positive d R /d T ) while S indicates semiconductor-like conduction (negative d R /d Tterm ). Initial termination concentrations were determined through ex situ X-ray photoelectron spectroscopy (XPS) measurements (Supplementary Tables 3 - 6 and Supplementary Figure 12 ). The loss of intercalantsterm and –OH terminations was determined from ex situ TGA-MS, and the reduction in –F and =O terminating species was measured with in situ EELS Fig. 1 Evolution of MXeneterm·dom electronic properties with in situ vacuum annealingterm. Resistance versus temperature measurements are shown for Ti 3 C 2 T x ( a ), Ti 3 CNT x ( b ), and Mo 2 TiC 2 T x ( c ). The measurements were conducted during various vacuum annealing steps performed in the TEM. Each vacuum annealing step is represented with a different color and symbol. For each annealing step, both the heating and cooling curves are shown. In all cases, the resistance decreased during annealing, hence, the cooling curve is always beneath the heating curve. The atomic structures of the various MXenesterm·dom are shown as insets. The initial state schematics (top schematics) show Ti 3 C 2 T x and Ti 3 CNT x with intercalated water molecules on their surfaces, and Mo 2 TiC 2 T x is shown with water and TBA + molecules. With annealing, the MXeneterm·dom sample resistance is affected by the loss of adsorbed species, intercalantsterm, and terminating species. Some of these processes are shown schematically. The resistance data in this figure are also shown in Supplementary Figures 13-15 , where the resistance is plotted as a function of annealing time We produced the Ti 3 C 2 T x and Ti 3 CNT x samples through etching of their 3D parent MAX phases, i.e., Ti 3 AlC 2 and Ti 3 AlCN, in a mixture of LiF and HCl. This process results in –OH, –F, and =O terminations and H 2 O and Li + intercalation 6 , 23 . Mo 2 TiC 2 T x was instead produced through etching of Mo 2 TiAlC 2 in HF and delaminating via tetrabutylammonium hydroxide (TBAOH) intercalation 59 . This method reduces the concentration of –F termination and results in tetrabutylammonium (TBA + ) and H 2 O intercalation 40 , 61 . TBA + is a large organic ion which can significantly increase the inter-flake spacing and electrical resistance 6 , 40 , 52 . To directly compare H 2 O and TBA + intercalation, we additionally studied Ti 3 CNT x prepared with HF etching and TBAOH delamination. For clarity, this sample is referred to as Ti 3 CNT x (TBA + ). After synthesis, MXeneterm·dom films were spray-casted 24 onto MEMS (microelectromechanical systems)-based nanochips 62 designed for heating and biasing within the TEM column (Supplementary Figure 2 ). The MXeneterm·dom films were at least several flakes thick (Supplementary Figures 1 and 2 ), and the electrode spacing (~20 μm) was considerably larger than the MXeneterm·dom flake diameters (~100 nm up to 2–3 μm). As such, sample resistance measurements were dependent upon both intra- and inter-flake contributions. The cross-sectional sample areas were not well-defined, so we report the relative change in sample resistance and not the absolute resistivity. TEM imaging and electron diffraction confirmed the MXeneterm·dom structure both before and after in situ annealing at ≥700 °C (Supplementary Figure 2 ). The temperature-dependent resistance of each sample was initially measured in ambient atmosphere from room temperature (RT) up to 75 °C in order to understand the as-prepared sample properties. In agreement with previous results, Ti 3 C 2 T x displayed metallic behaviorterm while Ti 3 CNT x and Mo 2 TiC 2 T x displayed semiconductor-like (negative d R /d Tterm ) behavior 27 , 38 (Table 1 and Fig. 1 ). After measurement of the as-prepared MXeneterm·dom electronic properties, the samples were inserted into the TEM and vacuum annealed at temperatures up to 775 °C. Before providing a detailed analysis of our in situ heating and biasing experimentsterm·mth, we first summarize our two main findings. First, we observed transitions from semiconductor-like to metallic behaviorterm in Ti 3 CNT x and Mo 2 TiC 2 T x after the annealing-induced loss of intercalated species. These transitions reveal the intra-flake metallicity of Ti 3 CNT x and Mo 2 TiC 2 T x and demonstrate that intercalantsterm can cause negative d R /d Tterm in multi-layer MXenesterm·dom. Second, high temperature annealing and the partial loss of surface terminations (Fig. 1 , bottom insets) increased the conductivity of all three MXenesterm·dom. This finding is in agreement with past predictions that non-terminated MXenesterm·dom exhibit metallic behaviorterm with high carrier concentrations 43 , 55 , 56 , 59 . In the following, we describe the in situ heating and biasing data in greater detail; results are organized based on the different mechanisms which influence MXeneterm·dom electronic properties. Adsorbed species Prior to thermal annealing, insertion of Ti 3 C 2 T x and Ti 3 CNT x into the TEM vacuum (~10 −5 Pa) caused an immediate reduction in resistance (Fig. 1a, b and Supplementary Figure 3 ). For both of these samples, the resistance decreased roughly 20% after 150 s of insertion into the TEM. This change in resistance is attributed to the loss of adsorbed atmospheric species, e.g., H 2 O and O 2 . These species are known to cause doping in 2D materialsterm·dom such as graphene 63 , and similar effects have previously been reported in Ti 3 C 2 T x 8 , 25 . We note that both of these samples were only intercalated with H 2 O and Li + . For Mo 2 TiC 2 T x intercalated with TBA + , the sample resistance did not significantly change upon exposure to the TEM vacuum (Fig. 1c and Supplementary Figure 3 ). This observation suggests an increased effect of the large TBA + molecule relative to H 2 O, specifically, that TBA + intercalation limits the film resistance and masks the effect of atmospheric species desorption on the electronic resistance. The effect of adsorbed species on the resistance of Ti 3 CNT x (TBA + ) could not be determined, since the sample resistance prior to in situ vacuum annealingterm was too high to be accurately measured. Intercalation For all studied MXenesterm·dom, de-intercalationterm significantly increased electronic conductivityterm. We first consider the role of H 2 O, which began to de-intercalate at lower temperatures than TBA + . For all samples, mass spectroscopy (MS) showed a large peak in the H 2 O (mass to charge ratio ( m / e ) = 18) ion current centered at ~150 °C, indicating H 2 O de-intercalationterm. This behavior is shown for Ti 3 CNT x in Fig. 2a , and full TGA-MS results for all samples are presented in Supplementary Figure 4 . In situ EELS data is consistent with the de-intercalationterm of H 2 O during low temperature annealing. Figure 2b shows the O K -edge of Ti 3 CNT x after various in situ annealing steps; three distinct peaks are observed. For =O or –OH bonded with surface Ti, these three peaks are expected, with peak 1 arising due to hybridization between the termination moiety and the Ti 3 d orbitals 64 . Conversely, the O K -edge of water does not contain peak 1 65 . With annealing, peaks 2 and 3 decrease with respect to peak 1, signifying the loss of intercalated H 2 O. The loss of =O or –OH terminations cannot explain the observed changes in fine structure, as surface de-functionalizationterm would produce a more uniform decrease in the O K- edge intensity. Fig. 2 Influence of intercalated species on the MXeneterm·dom resistance and d R /d T . a TGA-MS of Ti 3 CNT x showing the loss of H 2 O intercalantsterm and –OH termination species. b In situ EELS of the normalized O K -edge of Ti 3 CNT x indicating the loss of H 2 O intercalantsterm with annealing. c Ex situ XRD of the Ti 3 CNT x (002) peak for an as-prepared thick sample, i.e., vacuum filtered and free-standing, as well as thick samples annealed at 150 °C (in vacuum) and 400 °C (in Ar). The peak shift to higher 2 θ values represents a decrease in the inter-layer spacing. The dashed vertical line represents the approximate position of the Ti 3 C 2 T x (002) peak after annealing at 150 °C, taken from ref. 6 . d Temperature-dependent PPMS resistance measurements of a thick as-prepared (intercalated) Ti 3 CNT x sample and the de-intercalated sample annealed to 700 °C within the TEM. R RT is the room temperature resistance. e TGA-MS of Mo 2 TiC 2 T x showing the loss of H 2 O intercalantsterm and the decomposition of TBA + intercalantsterm. f Temperature-dependent PPMS resistance measurements of a thick as-prepared (intercalated) Mo 2 TiC 2 T x sample and the de-intercalated sample annealed to 775 °C within the TEM. The as-prepared data is taken from ref. 38 . g Schematic of intercalantsterm’ influence on conduction through multi-flake Mo 2 TiC 2 T x By annealing samples at 200 °C, we isolate the influence of H 2 O de-intercalationterm on the Ti 3 C 2 T x and Ti 3 CNT x electronic properties from any effects of surface de-functionalizationterm. To prove this point, we consider the de-functionalization of –OH, the least stable termination species and thus the first to desorb 12 , 66 . The MS –OH ( m / e = 17) ion current in Fig. 2a shows two peaks for Ti 3 CNT x . The first peak perfectly mirrors the H 2 O ion current and is thus attributed to de-protonated H 2 O (Supplementary Figure 4 ); the second peak at ~375 °C is attributed to –OH termination loss. Our conclusion that –OH terminations are stable at <200 °C is supported by nuclear magnetic resonance spectroscopy and neutron scattering studies performed previously on Ti 3 C 2 T x 67 , 68 , assuming that the termination interaction with surface Ti atoms is similar in both carbide and carbonitride MXenesterm·dom. Moreover, it was shown that hydroxyl groups do not desorb from titania powder until 350–500 °C 69 . For 200 °C annealed Ti 3 C 2 T x , H 2 O de-intercalationterm caused an 18% reduction in sample resistance, in qualitative agreement with past results (Fig. 1a ) 50 , 70 . Subtle increases in the dR / dT of Ti 3 C 2 T x with 125 and 200 °C annealing suggest the decrease of an insulating inter-flake resistive term, as opposed to a change in the intra-flake metallic conductivityterm (Supplementary Note 1, Supplementary Figure 5 , and Supplementary Table 1 ). These changes in d R /d T with low temperature annealing are in agreement with a decrease in resistance driven by H 2 O de-intercalationterm. In comparison to Ti 3 C 2 T x , Ti 3 CNT x displayed an anomalously large response to H 2 O de-intercalationterm. After annealing at 200 °C, there was a 36% decrease in resistance and a transition from negative to positive d R /d T (Fig. 1b ). Ex situ XRD shows a contraction of the Ti 3 CNT x c -lattice parameter (signifying a decrease in the inter-layer spacing) after 150 °C annealing (Fig. 2c ). This measurement further confirms the partial de-intercalationterm of water and supports the argument that de-intercalationterm improves conductivity through a reduced inter-flake resistanceterm. The measurement of metallic conductivity in Ti 3 CNT x confirms numerous theoretical predictions of its metallic nature 55 , 57 , 58 . To better understand the electronic properties of the de-intercalated Ti 3 CNT x , we studied the sample’s low-temperature conduction ex situ. To do so, the sample was further annealed to 700 °C (ensuring maximum de-intercalationterm), removed from the TEM, and then inserted into the physical property measurement system (PPMS). Figure 2d shows the normalized resistance of as-prepared (intercalated) and 700 °C annealed (de-intercalated) Ti 3 CNT x . The intercalated sample shows semiconductor-like behavior across the entire temperature range of the PPMS measurement (from RT to −263 °C), while the de-intercalated sample displays metallic behaviorterm down to roughly −150 °C. At this temperature, the de-intercalated Ti 3 CNT x sample shows a transition to negative d R /d Tterm , which is similar to the reported behavior of multi-layer Ti 3 C 2 T x samples 27 , 38 . It is an open question whether this low temperature transition is due to intra-flake or inter-flake effects. Both the as-prepared and annealed samples displayed negative magnetoresistance (MR) at −263 °C (10 K) (Supplementary Figure 6 ). We note that annealing Ti 3 CNT x above 300 °C causes partial surface de-functionalizationterm, which we discuss in the next section. However, high temperature annealing and surface de-functionalizationterm did not significantly affect d R /d T (Fig. 1b and Supplementary Table 1 ), so we ascribe the changes in normalized resistance shown Fig. 2d to de-intercalationterm. While H 2 O intercalation was known to increase MXeneterm·dom resistance values, the findings shown in Figs. 1a, b and 2a–d demonstrate that H 2 O intercalation can additionally act to decrease the value of d R /d T . Moreover, given a sufficient concentration of intercalantsterm in Ti 3 CNT x , the inter-flake resistanceterm can mask the MXeneterm·dom’s intra-flake metallic conductivityterm. We stress that the resulting negative dR /d T of intercalated Ti 3 CNT x is not due to a bandgap opening, rather it is the consequence of the temperature-dependence of the inter-flake electron hopping process. It is noteworthy that H 2 O intercalated Ti 3 C 2 T x displayed metallic conduction while H 2 O intercalated Ti 3 CNT x displayed semiconductor-like conduction, despite both MXenesterm·dom being intrinsically metallic. To understand this difference, we show the position of the Ti 3 C 2 T x (002) peak after annealing at 150 °C (vertical dashed line in Fig. 2c ). The c -lattice spacing of 150 °C-annealed Ti 3 CNT x (23.0 Å) is significantly larger than that of 150 °C-annealed Ti 3 C 2 T x (20.6 Å). The lattice spacing of Ti 3 CNT x only contracts to ~20 Å after annealing at 400 °C. This data suggests an increased lattice response of Ti 3 CNT x to intercalated H 2 O, which in turn increases the inter-flake resistanceterm and drives the transition to negative d R /d Tterm . Measurements of Mo 2 TiC 2 T x , having both H 2 O and TBA + intercalantsterm, showed an increased effect of TBA + relative to H 2 O. Annealing Mo 2 TiC 2 T x at 200 °C led to the de-intercalationterm of H 2 O and a 24% reduction in resistance, but there was no change in the sign of d R /d T (Figs. 1 c and 2e ). Subsequent annealing up to 500 °C led to a 69% decrease in resistance and a transition from negative to positive d R /d T (Fig. 1c ). This transition to metallic behaviorterm is due to the decomposition and loss of TBA + , as seen in the MS signal of C 2 H 6 ( m / e = 30) (Fig. 2e ). The measured temperature of TBA + de-intercalationterm/decomposition was ~350 °C, which is similar to the decomposition temperature of TBA + intercalated in graphite 71 . Previous work has shown that annealing Mo 2 TiC 2 T x at 530 °C reduces the c -lattice parameter from 37.7 to 24.5 Å 52 , which supports our claim that the decomposition of TBA + causes a decrease in the inter-flake resistanceterm and consequently a transition to metallic behaviorterm. To further study the effect of TBA + decomposition, the low-temperature resistance of Mo 2 TiC 2 T x was measured. We annealed the sample up to 775 °C in an attempt to eliminate residual TBA + , and then we removed the sample from the TEM and inserted it into the PPMS. The normalized resistance of the as-prepared (intercalated) and the annealed (de-intercalated) Mo 2 TiC 2 T x are shown in Fig. 2f . The de-intercalated Mo 2 TiC 2 T x sample shows metallic behaviorterm down to the lowest measured temperature of −263 °C (10 K), and both the as-prepared and de-intercalated samples show positive MR (Supplementary Figure 6 ). These results demonstrate the intrinsically metallic nature of Mo 2 TiC 2 T x , in agreement with the previously reported metallic conductivity in this MXeneterm·dom after annealing at ~530 °C 52 . We note that annealing Mo 2 TiC 2 T x at 775 °C leads to the partial loss of termination groups; however, annealing at these high temperatures did not significantly alter the value of d R /d T (Fig. 1c and Supplementary Table 1 ). Thus, the difference in normalized resistance shown in Fig. 2f is attributed to de-intercalationterm. The observation of intercalation-induced negative d R /d Tterm in two different MXenesterm·dom with two different intercalantterm species provides strong evidence that this is a general phenomenon. For sufficient levels of intercalation, the thermally activated inter-flake hopping processterm becomes the rate-limiting step for conduction through multi-layer samples. This extrinsic effect causes MXenesterm·dom to display negative d R /d Tterm regardless of their intrinsic electronic properties, as shown schematically in Fig. 2g . Next, we consider Ti 3 CNT x intercalated with TBA + . With annealing, Ti 3 CNT x (TBA + ) showed a larger total reduction in resistance than Ti 3 CNT x intercalated with only H 2 O and Li + (Table 1 and Supplementary Figure 7 ), consistent with the increased effect of TBA + intercalantsterm. However, even after annealing at 750 °C, Ti 3 CNT x (TBA + ) continued to display negative d R /d Tterm , albeit with a reduced temperature-dependency (Supplementary Table 1 ). Given the previous demonstration of metallic intra-flake conduction in Ti 3 CNT x , we speculate that incomplete removal of TBA + is responsible for the persistence of the semiconductor-like behavior. After annealing, the temperature-dependent resistance closely followed R ∝ T ×exp( W / kT ), which maypersName be related to the inter-flake hopping mechanism (Supplementary Figure 7 ). It is presently unclear why annealing led to metallic conduction in Mo 2 TiC 2 T x (TBA + ) but not in Ti 3 CNT x (TBA + ). This difference could be due to the surface Mo atoms of Mo 2 TiC 2 T x affecting the release of TBA + and/or the inter-flake hopping process. Termination After de-intercalationterm, vacuum annealing at higher temperatures led to surface de-functionalizationterm and improved electronic conductivityterm. Initially, each sample had –OH, –F, and =O terminations in varying concentrations (Table 1 ). Based on DFT calculations, –OH should be the first species to desorb 12 , 66 , and for certain samples, TGA-MS data indicated the loss of –OH ( m / e = 17) at ~375 °C (Fig. 2a ). However, for the majority of TGA-MS measurements, the release of de-protonated H 2 O completely masked the signal of –OH loss (Supplementary Figure 4 ). De-functionalization of –OH was also difficult to detect with in situ EELS (Fig. 2b ), which could be due to the electron beam transforming –OH groups into more stable =O terminations prior to EELS acquisition 72 . Conversely, the partial loss of –F and =O species were clearly identified. As such, we focus on the effects of –F and =O desorption; however, we stress that the release of these species indicates prior loss of less stable –OH groups 12 , 66 . Annealing Ti 3 C 2 T x from 300 to 775 °C led to a significant loss of –F termination species and an associated increase in sample conductivity. In situ EELS measurements show a decrease in the F K -edge intensity after annealing at 500, 700, and 775 °C (Fig. 3a ), demonstrating the partial de-functionalization of –F. Consistent with the EELS measurements, ex situ TGA-MS shows the release of –F termination ( m / e = 19) beginning around 400 °C (Fig. 3b ). The broad peak in the m / e = 19 ion channel at ~150 °C closely mirrors the H 2 O channel, indicating that this peak arises from a mechanism related to H 2 O de-intercalationterm, e.g., release of protonated H 2 O. Despite the large loss of –F, there was no evidence of =O release, as demonstrated by the constant value of peak 1 of the O K -edge EELS spectra (Fig. 3c ). Fig. 3 Correlation of MXeneterm·dom surface termination and conductivity. a In situ EELS of the Ti 3 C 2 T x F K -edge showing the loss of –F with annealing. b TGA-MS of Ti 3 C 2 T x showing de-intercalationterm of H 2 O completing by ~400 °C and the onset of –F de-functionalization at ~400 °C. c In situ EELS of the Ti 3 C 2 T x O K -edge with annealing. The relative decrease in peaks 2 and 3 relative to peak 1 indicates the loss of intercalated H 2 O and the retention of =O. d Comparison of the post-anneal RT resistance of Ti 3 C 2 T x (black circles) and –F concentration (green triangles) given in units of x based on the chemical formula Ti 3 C 2 O 0.5 F x . e , f show the time-resolved resistance and F K -edge intensity, respectively, of Ti 3 CNT x during in situ annealing. The solid (dashed) arrows represent heating (cooling). For the time axis, t = 0 s corresponds to the beginning of annealing at RT. The stars mark 500 °C, the maximum temperature of prior annealing. The green line in f is a Gaussian smooth of the EELS data (black circles). The dark orange line in f is the calculated RT resistance, determined using the measured value of d R /d T upon heating. Because d R /d T changes with annealing, this calculation is only valid up to ~700 °C. To highlight the correlation between the –F concentration and the calculated RT resistance, both quantities were normalized by their initial value. g In situ EELS spectra of the O K -edge of Mo 2 TiC 2 T x . h Comparison of the post-anneal RT resistance (black circles) with the concentration of =O (red squares), given in units of x based on the chemical formula Mo 2 TiC 2 O x . For a , c , and g , EELS spectra were acquired at RT after annealing, and the edge intensities were normalized to the Ti L -edge intensity. In d and h , the change in elemental concentration was determined with in situ EELS. The temperature offset between elemental concentration data and resistance measurements in d and h are due to thermal gradients within the nanochip (Supplementary Figure 10 ). The range of these thermal gradients are represented by the horizontal error bars in the RT resistance data The observed loss of –F beginning at ~400 °C and the retention of =O terminations is consistent with ref. 73 , where Persson et al. report in situ annealing of Ti 3 C 2 T x with scanning TEM (STEM) and XPS analysis. However, our results differ from the recent in situ STEM annealing of Ti 3 C 2 T x reported by Sang et al. 54 . In their study, =O terminations were lost and large voids were formed after 500 °C annealing and electron irradiation. We attribute these contradictory results to beam induced effects. Sang et al. utilized a focused STEM probe (~10 9 e − s −1 Å −2 ) while we used a spread TEM beam (~10 e − s −1 Å −2 ). Moreover, our use of low-dose DD EELS 53 allows short acquisition times, further decreasing the potential for beam-induced sample degradation (Supplementary Figure 1 ). In Fig. 3d , we show the correlation between changes in the Ti 3 C 2 T x –F concentration and the RT resistance as a function of annealing temperature. The decrease in sample resistance produced during the 500 °C annealing step likely has a significant contribution from the loss of –OH groups. Additionally, there maypersName be a small contribution from the final stages of H 2 O de-intercalationterm. However, after the 500 °C annealing step, ex situ TGA-MS (Fig. 3b ) and in situ EELS (Fig. 3c ) show no evidence of further H 2 O loss. Additionally, XRD studies have shown that there is no change in the Ti 3 C 2 T x c -lattice spacing for annealing beyond 500 °C 74 . With no evidence of intercalation loss, no change in the inter-flake spacing, and no observation of a conducting secondary phase (Supplementary Figure 2 ) 75 , we conclude that the improvement in Ti 3 C 2 T x conductivity during annealing steps at 700 and 775 °C is entirely due to the loss of –F terminations. Similar to Ti 3 C 2 T x , in situ annealing of Ti 3 CNT x at 500 and 700 °C led to a reduction in –F species and improved conductivity (Fig. 1b and Supplementary Figure 8 ). In addition to the Ti 3 CNT x sample discussed previously in the text and shown in Figs. 1 and 2 , another Ti 3 CNT x sample was measured with time-resolved in situ EELS. The time-resolved measurements of Ti 3 CNT x show the simultaneous release of –F termination species alongside an increase in electronic conductivityterm (Fig. 3e, f and Supplementary Figure 9 ). This data further demonstrates that the loss of termination species influences the MXeneterm·dom electronic properties. For Mo 2 TiC 2 T x , surface de-functionalizationterm occurred at lower temperatures than Ti 3 C 2 T x or Ti 3 CNT x , indicating a weaker interaction between termination species and surface Mo atoms compared to Ti atoms. In situ EELS shows the complete loss of –F from Mo 2 TiC 2 T x after annealing at 500 °C (Supplementary Figure 8 ), and annealing at 700 and 775 °C produced a large reduction in the =O concentration (Fig. 3g ). As =O terminations were partially removed from the surface, the Mo 2 TiC 2 T x resistance decreased a total of 32% (Fig. 3h ). These findings suggest that the release of =O from Mo 2 TiC 2 O x increases the MXeneterm·dom intra-flake conductivity 38 , 59 . However, we cannot rule out the possibility that inter-flake effects contribute to the measured changes in resistance for annealing steps at 700 and 775 °C. As opposed to H 2 O intercalantsterm which cleanly de-intercalate, the TBA + intercalantsterm decompose (Fig. 2e and Supplementary Figure 4 ). Even though the decomposition of TBA + leads to the onset of metallic behaviorterm in Mo 2 TiC 2 T x (Fig. 1c ), it is possible that some residue remains even after annealing at >500 °C. This is likely the case for Ti 3 CNT x (TBA + ), since this MXeneterm·dom does not show metallic behaviorterm even after annealing at 750 °C. To the best of our knowledge, these results constitute the first direct experimental correlation of MXeneterm·dom surface chemistryterm and electronic conductivityterm. Previous DFT studies have predicted that termination of Ti 3 C 2 T x and Ti 3 CNT x with –OH, –F and/or =O significantly alters the electronic states near E Fterm 7 , 43 , 55 , 56 , 66 . For non-terminated Ti 3 C 2 and Ti 3 CN, DFT predicts a local maximum in the DOS at E F , but with complete surface functionalizationterm, the DOS( E F ) is greatly reduced. Consequently, surface functionalizationterm maypersName alter the electronic resistance through a decrease the carrier concentration, nterm . To test the validity of this proposed mechanism, we analyzed the concurrent changes in the Ti 3 C 2 T x and Ti 3 CNT x resistance and d R /d T with in situ annealing. We assume that the metallic intra-flake conductivity of these MXenesterm·dom can be described by the Drude equation, and that over the temperature range of our in situ TEM experimentsterm·mth, electron-phonon scattering is approximately linear in temperature (Fig. 1 ) 76 , 77 . With these assumptions, the Drude equation predicts that an increase in n with annealing will produce a proportional decrease in the resistance and d R /d T , i.e., Δ R ∝Δ dR / dT ∝( n 1 / n 2 −1), where n 1 and n 2 are the carrier concentrations before and after annealing, respectively (Supplementary Note 1). To visualize this behavior, we define ηterm as the ratio of the proportional change in the RT d R /d T to the proportional change in the RT resistanceterm for a given annealing step. A value of ηterm = 1 is predicted for a change in resistance driven solely by a change in the intra-flake carrier concentrationterm. For annealing Ti 3 C 2 T x and Ti 3 CNT x at high temperatures, ηterm ~ 1, supporting the claim that surface de-functionalizationterm increases the MXeneterm·dom conductivity through an increase in n (Fig. 4 ). Fig. 4 Analysis of concurrent resistance and d R /d T changes with annealing. For a given annealing step, ηterm is the ratio of the proportional change in the RT d R /d T to the proportional change in the RT resistanceterm (Supplementary Note 1). A positive value of ηterm (indicating a decrease in both the d R /d T and the resistance) is consistent with a change in the intra-flake resistance, and a negative value of ηterm (indicating an increase in d R /d T and a decrease in resistance) is consistent with a change to the inter-flake resistanceterm. For a change in resistance solely due to a change in the intra-flake carrier concentrationterm, the Drude formula predicts that ηterm = 1. The colored lines are a guide to the eye. Error bars represent the measurement standard errorterm·stat accounting for the linear fit to the d R /d T data and assuming a base uncertainty of 1.2% in the resistance measurements. For the inset equation, n 1 and n 2 refer to the intra-flake carrier concentrationsterm before and after an annealing step, respectively. See Supplementary Figure 5 for d R /d T and ηterm analysis for all studied MXenesterm·dom In contrast to the ηterm ~ 1 behavior for high temperature annealing, ηterm is negative for annealing Ti 3 C 2 T x and Ti 3 CNT x at low temperatures ( ηterm < 0 reflects a decrease in resistance and an increase in d R /d T ). As we describe in Supplementary Note 1, a negative value of ηterm is inconsistent with a change to the intra-flake metallic conductivityterm driven by, e.g., a change in n , effective electron mass, or defect density. However, a negative value of ηterm can be explained by a decrease in the insulating inter-flake resistanceterm, assuming that the inter-flake and intra-flake resistances act in series. Thus, the observed transition from negative to positive ηterm is indicative of a transition from control over the inter-flake resistanceterm (due to low temperature annealing and de-intercalationterm) to control over the intra-flake resistance (due to high temperature annealing and surface de-functionalizationterm). For Mo 2 TiC 2 T x and Ti 3 CNT x (TBA + ), ηterm < 0 for all annealing temperatures, suggesting that residue from the TBA + decomposition continues to affect the MXeneterm·dom inter-flake resistanceterm even after annealing at >500 °C (Supplementary Figure 5 ). Discussion In this study, we vacuum annealed multi-layer MXeneterm·dom samples within the TEM and measured 4, 6, >10, and 6 times increases in the conductivity of Ti 3 C 2 T x , Ti 3 CNT x , Ti 3 CNT x (TBA + ), and Mo 2 TiC 2 T x , respectively. With annealing, we studied de-intercalationterm and surface de-functionalizationterm with both in situ and ex situ spectroscopic techniques. By correlating this chemical analysis with in situ resistance and d R /d T measurements, we were able to delineate the effects of intercalation and surface termination on MXeneterm·dom electronic properties. Considering the role of intercalantsterm first, we found that both H 2 O and TBA + intercalantsterm increase sample resistance and can induce negative d R /d Tterm . Vacuum annealing caused de-intercalationterm, significant decreases in sample resistance, and the onset of metallic conduction (except for Ti 3 CNT x (TBA + )). The effects of TBA + intercalation on electronic properties were far greater than that of H 2 O, and even after annealing at >750 °C, lingering effects of TBA + intercalation persisted. These findings are relevant for the optimization of MXeneterm·dom devices where large metallic conductivity is required and the increased inter-flake spacing associated with intercalation does not affect performance, e.g., wireless communication and wearable electronics. The ability to control d R /d T through intercalation could find use in multifunctional sensors or in the development of MXeneterm·dom films with arbitrary d R /d T values. Additionally, we note the striking similarities between the intercalated Ti 3 CNT x and Mo 2 TiC 2 T x samples measured here with previous reports of semiconductor-like transport in Mo- and V-based MXenesterm·dom 38 – 40 . Confirmation of intrinsic semiconducting MXeneterm·dom behavior will require temperature-dependent resistance measurementsterm of single-flake MXene devicesterm·mth, which is beyond the scope of this study. Regarding the effects of termination, vacuum annealing was shown to cause partial surface termination removal and increases in the MXeneterm·dom electronic conductivityterm. Oxygen terminations were more stable than –F terminations, and Mo 2 TiC 2 T x experienced a greater degree of surface de-functionalizationterm than the Ti-based MXenesterm·dom. These findings provide an avenue to further improve performance in MXeneterm·dom applications such as electromagnetic interference shielding and optoelectronics. For other applications, e.g., triboelectrics and catalysis, improved conductivity is desired, but the functionalized MXeneterm·dom surfaces offer chemical benefits. Hence, further analysis is needed to understand how partial de-functionalization affects performance in these areas. From a broader perspective, our findings provide a first step towards realizing non-functionalized MXenesterm·dom and termination-engineered MXenesterm·dom, which are predicted to exhibit magnetism 56 , fully spin-polarized transport 42 , semiconducting behavior 36 , 38 , and nontrivial topological order 60 , 78 . Methods Syntheses of MXenesterm·dom Ti 3 C 2 T x MXeneterm·dom was synthesized by selective etching of Ti 3 AlC 2 MAX phase (Materials Research Center, Ukraine) in a mixture of LiF and HCl 6 . Specifically, 0.5 g of Ti 3 AlC 2 powder was allowed to react with a premixed etchant (0.8 g of LiF and 10 mL of 9 M HCl) for 24 h at room temperature. Then the acid mixture was washed with 150 mL of deionized water for 3–5 cycles until the pH of the supernatant reached the value of 5. After that, the mixture was hand-shaken for 10 min followed by centrifugation at 3500 rpm for 10 min to remove unreacted MAX particles and reaction by-products. The dark supernatant was centrifuged at 3500 rpm for another 1 h to yield delaminated Ti 3 C 2 T x solution. Ti 3 CNT x MXeneterm·dom was synthesized from Ti 3 AlCN MAX 11 phase by two different routes. For the first route, the synthesis protocol is similar to the Ti 3 C 2 T x protocol mentioned earlier except that the reaction mixture was stirred at 500 rpm at 40 °C for 18 h. For the second route, etching is performed in HF acid and delamination is achieved with molecular intercalation. Initially, 2 g Ti 3 AlCN was etched in 20 mL of ~30% HF acid for 24 h at room temperature. The reaction mixture was washed with 150 mL of deionized water for 3–5 cycles until the pH of the supernatant reached the value of 5. The resulting mixture was filtered through a filter paper (3501 Coated PP, Celgard, USA) to collect multilayer Ti 3 CNT x powder. To delaminate the Ti 3 CNT x powder, 1 g of dry powder was stirred in a mixture of 9 mL of water and 1 mL of tetrabutylammonium hydroxide (TBAOH) solution (48 wt%, Sigma Aldrich) for 24 h. The mixture was hand-shaken for 10 min and washed with 3–5 cycles of deionized water. After neutral pH was reached, the mixture was centrifuged at 3500 rpm for 10 min to remove unreacted MAX particles and reaction by-products. The brownish supernatant was centrifuged at 3500 rpm for another hour to yield delaminated Ti 3 CNT x solution. Mo 2 TiC 2 T x was synthesized from Mo 2 TiAlC 2 MAX phase 59 . To produce the MAX phase, Mo, Ti, Al, and graphite powders (Alfa Aesar, Ward Hill, MA) were mixed in a ratio 2:1:1.1:2 and ball milled for 18 h in a plastic jar with zirconia balls. The powder mixture was transferred in an alumina crucible and held at 1600 °C for 4 h (5 °C min −1 , Ar). The resulting MAX block was drilled and sieved (400 mesh, particle size <38 µm). Specifically, 2 g of Mo 2 TiAlC 2 powder was added to 40 mL of 48–51% aqueous HF solution and stirred at 50 °C for 48 h. The mixture was washed and collected the same way as the HF-prepared Ti 3 CNT x MXeneterm·dom. To delaminate Mo 2 TiC 2 T x powder, 1 g of the powder was stirred in 10 mL of 48% TBAOH aqueous solution for 12 h. The mixture was washed with 3 cycles of deionized water (3500 rpm for 15 min, each cycle). After a pH of 7–8 was reached, the supernatant was decanted the same way as the Ti 3 CNT x (TBA + ). The sediment was dispersed in 30 mL of deionized water and sonicated for 30 min in a water-ice sonication bath under argon bubbling. The final solution was centrifugate at 3500 rpm for 30 min. Finally, the supernatant is collected. We note that the exact changes in conductivity, intercalation, and surface termination reported here are likely related to MXeneterm·dom film thickness, flake size, and general flake quality, which are all a function of the synthesis process. As such, the measurement of different MXeneterm·dom films—produced with differing synthesis procedures—will likely produce a differing degree of conductivity changes with annealing. Electron microscopy and spectroscopy A JEOL 2100F microscope with a Schottky electron emitter was used for these experimentsterm·mth. The microscope is equipped with the high-resolution pole piece with a C S of 1 mm. EELS measurements were performed in TEM (diffraction) mode in order to sample a large area of the MXeneterm·dom film and to reduce irradiation-induced sample degradation. The EELS collection angle was set to 11 mrad. A direct detection (DD) EELS system was used as recently reported in ref. 53 . The energy dispersion was set to 0.125 eV per channel, and the zero-loss peak (approximate energy resolution) was measured to be ~1.0 eV. For EELS analysis, Gatan DigitalMicrograph was used. In situ heating and biasing The DENSsolutions Lighting D9+ heating and biasing sample holder was used for in situ TEM/EELS experimentsterm·mth with the 8 contact (A1-type) heating and biasing nanochips 62 . MXenesterm·dom were deposited onto the nanochips through spray casting 24 (Supplementary Figure 2 ). A mask was used to confine the MXeneterm·dom deposition to an area centered on the biasing electrodes (Supplementary Figure 2 ). Resistance measurements were performed with a 4-probe geometry using a Keithley 2400 SMU. Due to the high conductivity of MXeneterm·dom and the large sample cross-sectional area (relative to in situ TEM standards), application of voltages above 0.1 V led to high current densities. The resulting Joule heating strongly affected heating measurements and damaged the heating and biasing chips. As such, the resistance measurements were obtained with applied voltages of ≤5 mV. The maximum annealing temperatures were 775, 700, 750, and 775 °C, for Ti 3 C 2 T x , Ti 3 CNT x , Ti 3 CNT x (TBA + ), and Mo 2 TiC 2 T x , respectively. The Ti 3 CNT x samples were not heated to 775 °C since previous experimentsterm·mth showed a larger extent of TiO 2 formation at 775 °C. Heating and cooling rates were 1 °C s −1 . For annealing steps performed at 75, 125, and 200 °C, the maximum temperature was held for 10 s, and for annealing steps performed at 300, 500, 700, and 775 °C, the maximum temperature was held for 5 min. We note that the average temperature within the samples deviated from the input temperature due to temperature gradients within the films (Supplementary Figure 10 ). A different heating procedure was used for the Ti 3 CNT x (TBA + ) sample. Initially, free TBA + molecules prevented good electrical contact, and the sample was annealed in situ at 200 °C to achieve adequate contact between the MXeneterm·dom film and Pt electrodes. Then the sample was heated to 750 °C and held at this temperature for 1 h. This modified procedure was aimed at ensuring maximum de-intercalationterm/decomposition of the TBA + molecules without significant TiO 2 formation. We note that for each M n +1 X n chemistry, two samples were tested with in situ TEM heating and biasing, and each sample showed qualitatively similar behavior (Supplementary Table 2 ). Thermogravimetric-mass spectrometry analysis Simultaneous thermogravimetric-mass spectrometry analysis was performed on a Discovery SDT 650 model connected to Discovery mass spectrometer (TA Instruments, DE). Vacuum-filtered films of MXenesterm·dom were cut to small pieces and packed in a 90 μL alumina pan. Before heating, the analysis chamber was flushed with He gas at 100 mL min −1 for 1 h to reduce residual air. The samples were heated to 1500 °C at a constant heating rate of 10 °C min −1 in He atmosphere (100 mL min −1 ). Low temperature electronic transport Electronic transport properties of MXenesterm·dom after heat treatments were measured in a Quantum Design EverCool II Physical Property Measurement System (PPMS). After the in situ annealing experimentsterm·mth, the heating and biasing nanochip was removed from the TEM and wired to the PPMS sample puck using conductive Ag paint (Supplementary Figure 11 ). Temperature-dependent resistance was recorded from 25 °C (300 K) down to −263 °C (10 K) in a low pressure helium environment (∼20 Torr). Magnetoresistance data was recorded at 10 K by sweeping the magnetic field perpendicular to the chip from −70 kOe to 70 kOe. We note that for both the Ti 3 CNT x and Mo 2 TiC 2 T x samples, the room temperature resistance and d R /d T values (but not the sign of d R /d T ) differed between the final measurement within the TEM and subsequent measurements within the PPMS. This difference is likely due to the differing experimental set-up and/or sample degradation during exposure to atmosphere after removal from the TEM. X-ray photoelectron spectroscopy X-ray photoelectron spectroscopy analysis was performed in a spectrometer (Physical Electronics, VersaProbe 5000, Chanhassen, MN) using a 100 μm monochromatic Al K α X-ray beam. Photoelectrons were collected by a takeoff angle of 45° between the sample surface and the hemispherical electron energy analyzer. Charge neutralization was applied using a dual beam charge neutralizer irradiating low-energy electrons and ion beam. Vacuum-filtered films were mounted on double-sided tape and were electrically grounded using a copper wire. Prior to data acquisition, an Ar beam operating at 2 kV and 2 µA was used to sputter the sample surface for 2 min. Quantification and deconvolution of the core-level spectra was performed using a software package (CasaXPS Version 2.3.16 RP 1.6). Background contributions to the measured intensities were subtracted using a Shirley function prior to quantification and deconvolution. X-ray diffraction The X-ray diffraction data were acquired from vacuum-filtered films of the same solution used for the in situ TEM study. The data were acquired by a diffractometer (Rigaku Smart Lab, USA) with Cu K α radiation at a step size of 0.03° with 0.6 s dwelling time. Supplementary information Supplementary Information Peer Review File