Introduction

Transition metal carbides have been used in unique applications where oxides lack potential due to their high melting points (such as ~3900 °C for HfC)1,2, thermal conductivity (such as 63 W·m−1·K−1 for WC)3, and mechanical behavior (elastic modulus up to 549 GPa for TiC)4. In current studies, carbon vacancies5, rapid heating6, or noble metal decoration7 give tools for modification of the inherent material behavior of transition metal carbide systems8,9,10,11,12,13,14,15,16,17. Although some methods, such as flash or long-term sintering at low (~750 °C) temperatures provide some phase control for ideal performance6,12, there remains opportunity for advances to precisely control the phase of transition metal carbide systems for ideal phase-property relationships18.

The introduction of MXenes in 2011 launched transition metal carbides into the 2D realm19, which has served to add a diverse and tunable family of few-atom-thick (~1-nm-thick) and solution-processable transition metal carbides to materials science20,21. MXenes’ chemical diversity is apparent by their broad chemical formula Mn+1XnTx, where M stands for n + 1 layers of one or more 3d−5d and group 3–6 transition metals, X stands for n layers of carbon and/or nitrogen interleaving the M layers, and Tx stands for surface-M bonded surface groups, such as −O, −(OH), −Cl, and −F22,23. These surface groups are derived from MXenes’ scalable top-down synthesis methods from their respective MAX precursor24,25,26,27,28,29, which results in processable dispersions of single-to-few layer sheets of MXene (with ~ −40 mV in water)20,30,31,32.

Early studies on MXenes and their hybrid composites investigated their use in energy applications33,34, such as alkali ion batteries (Li, Na, K)35,36, and have persisted as a major area for application of these materials (nearly ~1500 publications to date)37. In part, this is due to the layered nature of MXenes, their electrochemically active surface, and their high electrical conductivity (up to 24,000 S∙cm−1 for Ti3C2Tx)38. When MXenes were first investigated for these applications, it became clear that alkali cations could effectively shuttle between the interlayers and could bind with MXenes’ surface39. Broadly, on the surfaces of MXenes (and some other 2D materials)40, there exist both surface terminations, which come from the etchant solution29, and atomic vacancies, which can be increased by overetching the surface41. Although previous studies considered that alkali cations can interact with the surface terminations42,43,44,45, to our knowledge, no study to date has comparatively evaluated the preferential site (surface groups, transition metal defects) for the interaction of alkali cations.

Beyond energy applications of MXenes46, recent studies have looked to using MXenes inherent M-X chemistry to enable their use in extreme environment applications as precursors for ultra-high temperature ceramics47,48. In previous studies, MXenes’ high-temperature behavior under an inert atmosphere can be described in four separate regimes. First, MXenes lose adsorbed surface species, such as adsorbed water, between room temperature to ~300 °C49,50,51. Second, MXenes lose their surface groups in order of the terminations’ bond strength between 300 °C to ~700 °C50,51. Third, Ti3C2Tx MXenes have shown that they undergo the start of phase transition through initiation of M and X paired diffusion to the surface to begin homoepitaxial nano-lamellar TiCy carbide growth along the (111) TiCy plane between 500–800 °C47,52. Past 800 °C, this diffusion results in nano lamellar TiCy carbide structures, which have been shown stable up to 1500 °C47,51. Broadly, it has been shown that defects in MXenes are primary sites for phase transition, as these are the energetically favorable sites for atomic migration of M and C out of the vacancy sites onto the basal plane during transition52. In addition, alkali cations have been shown to stabilize V2CTx MXenes toward hydrolysis53, while in bulk carbides, alkali cations have been hypothesized to serve as grain growth control and phase stability improvement agents54,55. Therefore, based on the gap in understanding of alkali cation occupation in defective sites, we believed that there was an opportunity for defect engineering of MXenes’ basal plane if alkali cations preferred to occupy defective sites in the basal plane.

To answer these questions, in this study, we first demonstrate that alkali cations prefer to occupy the transition metal atomic vacancy sites in Ti3C2Tx MXene’s basal plane. Next, we show that the occupation of alkali cations in the vacancy defective sites on the basal plane serves as an effective atomic migration control of MXenes’ phase transition at high temperatures to further stabilize MXene. After establishing this defect engineering with Ti3C2Tx, we next illustrate this behavior with a more compositionally complex ordered double transition metal Mo2TiC2Tx MXene by 1) establishing the locality of γ-Mo2C crystal formation around defective areas in phase-transformed Mo2TiC2Tx and 2) suppression of this γ-Mo2C growth around defective sites using alkali cations. To do so, we use a combination of in situ x-ray diffraction (XRD) and scanning transmission electron microscopy (STEM) techniques with ex-situ atomic-layer-resolution secondary ion mass spectrometry (SIMS), thermogravimetric analysis, and x-ray photoelectron spectroscopy methods paired with density functional theory (DFT) simulations. Overall, this study establishes the use of alkali metal cations as defect engineering techniques to control the phase stability of 2D MXenes in ambient and high temperatures, which can enable the further use of MXenes as elevated temperature stable energetic or extreme environment materials.

Results

Occupation and high-temperature behavior of alkali cations on Ti3C2Tx

In order to evaluate how alkali cations potentially interact with defects on MXenes, we first focused on controlling the defects on the surface of Ti3C2Tx MXene. The as-synthesized Ti3C2Tx (referred to as as-is Ti3C2Tx) was synthesized using the modified MXene synthesis method, which uses 5 wt% HF (HF-HCl acidic mixture) to etch an optimized Ti3AlC2 phase56. To induce defects on the Ti3C2Tx MXene, we increased the HF concentration to 9.1 and 12.5 wt% HF (HF–HCl acidic mixture), which has been shown to induce more surface titanium vacancies (VTi) in the resultant Ti3C2Tx MXene (Fig. 1a)41. Further, we confirmed a reduction in film conductivity from ~19,500 S/cm for as-is Ti3C2Tx made with 5 wt% HF in the etchant solution to ~2200 S/cm for Ti3C2Tx made with an etchant with 12.5 wt% HF, which is indicative of the detrimental effects of increased HF concentration on MXene’s flake properties.

Fig. 1: Layer-by-layer SIMS analysis of alkali cation decorated overetched Ti3C2Tx.
figure 1

a Mechanism of defect formation through using increased HF concentration induces titanium vacancies (VTi) defects on the surface of Ti3C2Tx with later alkali cation occupation. b–d SIMS of Ti3C2Tx MXenes synthesized with 5, 9.1, and 12.5 wt% HF, respectively. e–g SIMS of Na-decorated Ti3C2Tx MXenes synthesized with 5, 9.1, and 12.5 wt% HF, respectively.

Full size image

After observing a reduction in the electrical conductivity of Ti3C2Tx films due to an increase of HF concentration during MXene synthesis, we used dynamic atomic-resolution layer-by-layer SIMS to measure Ti vacancy concentrations at the basal planes and evaluate where alkali cations occupied on MXenes. Using SIMS on variable HF synthesized Ti3C2Tx MXenes, we measured the defect concentration on Ti3C2Tx’s surface with increasing concentrations of HF by directly comparing the Ti content between the central (assumed defect-free) to that of the surface Ti layers. SIMS measurements show a surface VTi content of 0.92 ± 0.04 at% for the as-is Ti3C2Tx (5 wt% HF used in the etchant), which was increased to surface VTi content of 8.69 ± 0.37 at% and 16.35 ± 0.68 at% for the Ti3C2Tx made with increased concentrations of HF at 9.1 wt% (Figs. 1c) and 12.5 wt% (Fig. 1d), respectively, in the HF-HCl etchant (Supplementary Fig. 1). This is in agreement with previous STEM studies which measured increased VTi contents on Ti3C2Tx by increasing the HF concertation in the etchant41. In addition, we also detected some oxygen (~3 at%) in the carbon sites in Ti3C2Tx.

To first investigate where alkali cations go (on the surface or in the VTi sites), we focused on determining where Li+ occupied, as Li comes from LiCl used to chemically delaminate flakes Ti3C2Tx. Using SIMS, we observed Li attachment only to the surface for the 5 wt% HF with <1 at% VTi, confirmed by Li signal at the surface termination layers (Supplementary Fig. 2a). In contrast, when the content VTi in the surface Ti layers (Supplementary Figs. 2b, c) was increased using HF overetching, we detected Li signal only overlapped with the surface Ti layers, which shows increasing vacancies results in Li occupation at the vacancy sites. Importantly, this is the first study that has shown that 1) alkali cations occupy defective sites in MXenes, and 2) controlling the surface VTi defects changes their occupancy site, which potentially has major implications on the design of MXenes for applications such as Li-ion batteries. However, with this knowledge, we can now evaluate that previous high-temperature studies of Ti3C2Tx MXenes likely already included Li occupation at defect sites45, so we then focused on determining where added larger alkali cations (such as Na) go on MXenes (Figs. 1e–g).

After the cation exchange process (see “Methods” section), in as-is Ti3C2Tx synthesized with 5 wt% HF (Fig. 1e), Na is found on the surface of the MXene sheets, similar to Li (Supplementary Fig. 2). Similar to Li, with an increase in HF concentration from 9.1 and 12.5 wt% HF and resulting increased in VTi concentration, we again observe that Na cations occupy the surface VTi sites. To understand the trend in preference of alkali cations to occupy the surface vs. defective sites, by comparing 5 wt% and ≥9.1 wt% HF, we can see that alkali cations only occupy defective sites once the molar content of surface Ti vacancies exceeds the content of Na (Supplementary Fig. 3). Broadly, this demonstrates that alkali cations can bind to MXenes surface, but likely energetically prefer partial occupation of the defective sites over sites atop the surface groups. Further, comparing between 9.1 and 12.5 wt% (Figs. 1e and f), we note an increased replacement of Na in the surface VTi sites from Li. Particularly, Li is nearly entirely removed by Na for the 12.5 wt% HF synthesized Ti3C2Tx (calculated from SIMS in Supplementary Fig. 4). Overall, these results show that an increase in VTi causes an increase in alkali cation occupation in the surface Ti layer.

After observing the cation occupation in defective sites using SIMS, we next compared high-temperature phase behavior as-is (5 wt% HF) and overetched (12.5 wt% HF) Ti3C2Tx using in situ XRD. Our results (Fig. 2a) indicated that the (00 L) peaks of as-is Ti3C2Tx MXene were retained even at ~900 °C with the partial formation of a rock-salt ordered carbon vacancy superstructure Ti2C and disordered carbon vacancy TiCy visible in the peak shoulders at ~17.8 and ~36 ° 2θ47. The higher temperature stability of Ti3C2Tx MXene compared to the previously reported results could be attributed to the VTi and oxygen substitution in the X-lattice site (Fig. 1) due to the use of an optimized Ti3AlC2 precursor57, which is in contrast to the MXene made from the non-optimized Ti3AlC2 used in the previous study (up to 30 at% O in the C sites compared to 3 at% in this study)47,57. However, when we compare this phase behavior to that of the overetched Ti3C2Tx MXene (Fig. 2b), we note for the overetched Ti3C2Tx 1) we have an increased (~6 times at 800 °C) peak intensity of δ-TiCy 2) a decreased shoulder intensity of Ti2C, and 3) a contraction in the a-LP of the resulting δ-TiCy phase, which changes from 4.41 Å to 4.29 Å for as-is and overetched MXenes, respectively. Notably, we observed that TiO2 was not present after annealing in all samples, which indicates that our phases present up to 900 °C are indeed carbide structures and not oxides. Based on previous reports of the thermogravimetric analysis (TGA) high-temperature behavior of MXenes51, all three of differences of overetched Ti3C2Tx MXene to the as-is Ti3C2Tx MXene is could be attributed to loss of C in the MXene structure either due to overetching, which likely destabilizes the MXene structure and allows for reaction of interior C with O to form CO2. Overall, this shows that defects both affect MXenes’ stability and the phase transformed phases.

Fig. 2: Phase control through defect stabilization in Ti3C2Tx MXene using alkali cations.
figure 2

a In situ XRD annealing of as-is Ti3C2Tx (00 L peaks shown with the blue ⬨symbol) synthesized using an optimized MAX method and etched with HF-HCl with 5 wt% HF etchant, demonstrating the formation of ordered carbon vacancy Ti2C (shown by the purple Δ symbol) and disordered carbon vacancy TiCy (demonstrated by the blue-green δ symbol). The XRD in situ hot stage window cover peaks (graphite) are marked with gray boxes. b In situ XRD annealing of overetched Ti3C2Tx (with 12.5 wt% HF etchant) which defects were induced by raising the HF concentration. c In situ XRD annealing of Na+-decorated overetched Ti3C2Tx, where a clear reduction in phase transformed Ti2C or TiCy is observed compared to as-is or overetched Ti3C2Tx. d–f Focus on the (002) and (004) peak regions of Ti3C2Tx MXene, which is used to calculate the peak intensity ratios in panel g which demonstrates the change in (002)/(004) peak intensity ratio across in situ annealing temperatures.

Full size image

After understanding how defects negatively impact the high-temperature stability of Ti3C2Tx, we investigated the effects of alkali cation addition on the stabilization of Ti3C2Tx. In the XRD patterns of the Na+-decorated overetched Ti3C2Tx film (Fig. 2c), the δ-TiCy and Ti2C peaks were difficult to detect, which may suggest a slowed phase transition of this Na+-decorated MXene. To gain better insight into this phase transition, we further investigated the effect of titanium surface vacancies (due to atomic migration) using the peak ratios of the (002)/(004) peaks of Ti3C2Tx. We first simulated the XRD patterns of Ti3C2Tx with increasing surface vacancies, using VESTA, where the (002)/(004) peak ratio increases with increasing surface Ti vacancies (Supplementary Fig. 5).

To evaluate the potential loss of surface Ti atoms due to surface diffusion during epitaxial growth, we calculated the (002)/(004) peak ratio changes in our in situ XRD results for each sample by increasing the temperatures for the as-is, overetched, and Na+-decorated overetched Ti3C2Tx films. We used 400 °C as our start temperature, as TGA studies have indicated 400 °C is when all intercalation species are removed from the spacing between the MXene sheets51. Using this analysis, we find that the overetched sample has the highest increase in (002)/(004) ratio, while the as-is sample has a slight decrease in (002)/(004) ratio. This matches our expectations for the surface diffusion of Ti atoms from surface Ti sites for atomic migration during phase transition, as vacancy sites are the initiation sites for this surface diffusion52. In contrast, when the overetched sample is decorated with Na+ (the Na+-decorated overetched Ti3C2Tx film), (002)/(004) peak intensity changes are minimized (Fig. 2g). This suggests that the Na+-decorated overetched Ti3C2Tx film had a reduced loss of surface Ti atoms to atomic migration, which agrees with the indeterminable δ-TiCy or Ti2C peaks after annealing at 900 °C. Overall, this analysis indicates that alkali cations may control the phase behavior of MXenes at high temperatures. Based on this finding, we next aimed to evaluate alkali cation effects on the more compositionally complex Mo2TiC2Tx MXene that forms multiple phases after phase transition.

Phase transition of Mo2TiC2Tx

The full analysis of the phase transition of Mo2TiC2Tx is shown in the Supporting Information (Supplementary Figs. S6–13 with text discussion). To briefly describe it here, we heat treated Mo2TiC2Tx (Fig. 3a) under inert atmosphere up to ~ 1000 °C using in situ XRD and STEM techniques paired with electron energy loss spectroscopy (EELS) measurements, and we identified that Mo2TiC2Tx MXenes transition to an NaCl-type carbide δ-(Ti, Mo)Cy, an orthorhombic carbide γ-Mo2C, and a body-centered cubic β-(Ti, Mo)ss intermetallic phases (Figs. 3a, b). Particularly, we observe that the γ-Mo2C and δ-(Ti, Mo)Cy structures initiate out of the 500 °C-formed vacancy clusters in Mo2TiC2Tx, as shown in Fig. 3c. By focusing on a vacancy cluster at 500 °C and ramping and holding at 600 °C (Figs. 3d–f), we observe evidence for this hypothesis, as the vacancy clusters grow to form larger clusters and a few visually identifiable phases. Similarly, we can visualize both adatoms and the defective area growth around the pore in the Mo2TiC2Tx similar to previous STEM observations41,52. To identify these local phases around the vacancy clusters, we used fast Fourier transform (FFT) analysis and mapping (FFT pattern matching in Supplementary Fig. 11), as shown at 700 °C in Fig. 3g. Using FFT mapping and EELS mapping analysis (Supplementary Fig. 12), it is evident that γ-Mo2C crystals are localized around the vacancy clusters while δ-(Ti, Mo)Cy forms in the original basal plane of Mo2TiC2Tx, as shown in Fig. 3g. Therefore, we believe these vacancies in the Mo2TiC2Tx basal plane act as preferential sites for atomic migration and crystal nucleation for the γ-Mo2C crystals to grow out of the basal plane.

Fig. 3: Flake-scale phase behavior of Mo2TiC2Tx.
figure 3

a, b Cross-sectional SEM of pre- and post-annealed Mo2TiC2Tx with various formed phases. c Phase change of Mo2TiC2Tx to its decomposed phases initiates at defect cluster sites. d–f Vacancy clustering with increasing temperature and time. g Localization of Mo2C around vacancy clusters with (Ti, Mo)Cy in the flake structure, with corresponding FFTs for the marked phases. h Mo2C grains grow with increasing temperature with corresponding FFTs for marked phases. i Shown STEM image of the various structures formed at 900 °C with relevant EELS mapping data showing distribution of Ti, Mo, and C among the marked phases.

Full size image

Figure 3h illustrates that, even after increasing the temperature to 900 °C, the δ-(Ti, Mo)Cy maintains the original flake-like structure while γ-Mo2C grows to form large crystals out of the basal plane of the δ-(Ti, Mo)Cy structure. By increasing temperatures (Fig. 3i and Supplementary Fig. 13), γ-Mo2C crystals grow dramatically on the original δ-(Ti, Mo)Cy structure, which agrees with the rapid grain growth of γ-Mo2C crystals at such temperatures from previous in situ TEM works58,59. Broadly, as this type of crystal growth of MXene-derived carbides may be potentially helpful for energy storage or conversion60,61, we dialed in on controlling these defects to see how this affected the phase transition of Mo2TiC2Tx MXene.

To examine our hypothesis that defects were critical to the formation of γ-Mo2C crystals, we reduced the MXene etching time to half (2 days instead of 4 days) to reduce the number of vacancies and measured the phase transformation temperatures via in situ XRD41. While the in situ annealing XRD data did not reveal changes in the phase stability of the Mo2TiC2Tx MXene to higher temperatures, unlike Ti3C2Tx, the data did show a reduction of the γ-Mo2C as compared to the δ-(Ti, Mo)Cy phase (Supplementary Fig. 14). Therefore, combining this in situ XRD data with our in situ STEM data showing defects as the preferential formation site of γ-Mo2C, we next look to investigate if the vacancy occupation using alkali cations could also control (hinder) the formation of these γ-Mo2C crystals as it had controlled atomic migration of Ti and C out of vacancy sites in Ti3C2Tx.

Phase control of Mo2TiC2Tx with alkali metal cations

In order to control the phase behavior of Mo2TiC2Tx MXene, we dissolved NaCl and KCl in Mo2TiC2Tx’s aqueous solution to decorate Mo2TiC2Tx with alkali metal (Na+, K+) cations similar to the previous section of Ti3C2Tx. Based on the established behavior of 1) alkali cations on surface migration in Ti3C2Tx and 2) the localization of γ−Mo2C phases around defective sites in Mo2TiC2Tx, we here present how alkali cations occupied the Mo2TiC2Tx flakes at elevated temperatures and if these cations affected the presence of the γ−Mo2C phase (Fig. 4a).

Fig. 4: Cation occupation and phase behavior of cation decorated Mo2TiC2Tx.
figure 4

a Mo2TiC2Tx is decorated using alkali cation chloride salts in an aqueous solution which then is filtered to form a film to test for its high-temperature phase behavior. b–d SIMS of 500 °C annealed as-is and decorated (Na, K, respectively) Mo2TiC2Tx MXene films to illustrate atomic composition layer-by-layer through the MXene sheets. e–g In situ XRD of Na+ and K+ decorated Mo2TiC2Tx shows increasing phase stability with cation decoration, where the blue diamonds represent Mo2TiC2Tx while the blue-green δ, pink γ, and purple β symbols represent (Ti, Mo)Cy, Mo2C, and (Ti, Mo)ss, respectively. The XRD in situ hot stage window cover peaks (graphite) are marked with gray boxes.

Full size image

To do so, we used SIMS to confirm the presence of alkali cations in-plane with the exterior molybdenum vacancy (VMo) sites in Mo2TiC2Tx annealed at 500 °C. First, we confirmed the ordered out-of-plane MXene Mo2TiC2Tx structure with O-Mo-C-Ti-C-Mo-O atomic ordering after annealing at 500 °C (Fig. 4b), which is similar to room temperature layer-by-layer SIMS analysis done on Mo2TiAlC2 MAX57. Further, we can observe that this Mo2TiC2Tx is indeed also an oxycarbide (O occupation in the X site between 27–29 at%). This is due to the presence of oxygen in the MAX precursor (Mo2TiAlC2) similar to the previous reports of MXene from non-optimized oxycarbide MAX phases57. Regardless, by directly comparing this Mo composition with the Mo2TiAlC2 MAX (Supplementary Fig. 15), using SIMS we calculated a 4.96 ± 0.18 at% VMo concentration in the surface Mo site likely induced by overetching our MXene flakes. In addition to the SIMS measurements, the cation presence in the Mo2TiC2Tx films was confirmed using EDS methods (Supplementary Fig. 16). Further, after adding Na/K to this Mo2TiC2Tx, our SIMS data indicated both Na (Fig. 4c) and K (Fig. 4d) occupy the same atomic plane as Mo in the Mo2TiC2Tx structure even after annealing at 500 °C, which shows that Na/K occupy the surface VMo sites in Mo2TiC2Tx similar to the surface VTi sites in Ti3C2Tx. It is important to note that the intensity ratio of the Mo vs. Na/K peaks is not directly reflective of the atomic percentage contents of Na and K, as SIMS is more sensitive to lighter elements. To further understand this behavior, we ran density functional theory (DFT) predictions, which show alkali cations energetically prefer the vacancy sites over the surface terminations on Mo2TiC2Tx (Supplementary Figs. 17a–c).

After confirming the alkali cation occupation in the VMo sites in Mo2TiC2Tx, we next used in situ XRD to observe the high-temperature phase behavior of Mo2TiC2T. These XRD results reveal that MXene’s (00 L) peaks are present up to 800 °C and 900 °C when decorated with Na+ and K+ (Figs. 4e, f), respectively, which contrasts with the behavior of regular Mo2TiC2Tx films which showed phase transition at 500 °C (Supporting Information). We also conducted XRD post annealing (after 1100 °C annealing) and our ex situ XRD measurements at room temperature show that the γ-Mo2C and β-(Ti, Mo)ss peaks are lower in intensity in the cation-decorated ones as compared to regular Mo2TiC2Tx (Fig. 3d). Further, EDS detected Na/K in the annealed Mo2TiC2Tx film (reduction by 57.1 at%/72.2 at%, respectively) (Supplementary Fig. 18) and XPS confirmed both the presence and binding states of Na/K even after annealing at 1100 °C (Supplementary Fig. 19). Based on this data and the phase transition behavior of Mo2TiC2Tx, we hypothesized that the cations must interact with the defect sites (vacancies) to limit the formation of γ-Mo2C.

To evaluate the effect of alkali cations vacancy occupation at high-temperatures using in situ STEM, we calculated the local a-LP (strain, ε) around the vacancy clusters at 500 °C in the Na- and K-decorated Mo2TiC2Tx flakes, which were 3.376 ± 0.151 Å (0.224 ± 0.014 ε) and 3.244 ± 0.179 Å (0.095 ± 0.006 ε), respectively, an increase over the 3.166 ± 0.180 Å a-LP (0.035 ± 0.003 ε) for non-decorated (bare) Mo2TiC2Tx (Supplementary Fig. 17d). Further, when we compared the decorated and non-decorated single-flake Mo2TiC2Tx at 600 °C, Mo2TiC2Tx single-flakes showed a reduced concentration of basal plane vacancy clusters (9.2 ± 2.1 % for Na, 8.3 ± 3.4 % for K) as compared to the non-decorated (as-is) Mo2TiC2Tx single flake (11.3 ± 4.2 %) (Supplementary Figs. 17e–g, Supplementary Fig. 20). Using in situ STEM, we also observed a reduction in γ-Mo2C grain size for Na/K decorated single-flake Mo2TiC2Tx compared to as-is Mo2TiC2Tx (Supplementary Fig. 21). Interestingly, FFT mapping determined that Mo2TiC2Tx MXene is still present on K+-decorated Mo2TiC2Tx even up to 800 °C (Supplementary Fig. 17f), which was not the case on the non-decorated (as-is) Mo2TiC2Tx structure (Supporting Information). Overall, these observations using in situ STEM in the cation-decorated Mo2TiC2Tx suggest that the cations are locally suppressing phase transition from MXene to 3D crystalline carbides and growth of γ-Mo2C around the vacancy clusters, which shows alkali cations interact with these vacancy clusters even at high-temperatures.

Further, using thermogravimetric analysis (TGA) and gas chromatography-mass spectrometry (GC-MS) methods during high-temperature annealing (Supplementary Fig. 22), we observed a reduction in both the rate (δW/δT) (up to 45% reduction for K-decorated compared to as-is) and total mass loss (up to 7 times reduction in the mass loss for K-decorated compared to as-is) around the carbon loss region (GC-MS presence of CO/CO2 confirmation in Supplementary Fig. 23 after annealing at 500 °C)51 in the Mo2TiC2Tx MXene films (Supplementary Figs. 18 & 23). Additionally, we observe that Tx = O loss occurs at temperatures ~700 °C, which matches previous work50 that O surface groups are typically lost due to reduction with C from MXenes. When analyzing the O content between as-is Mo2TiC2Tx and Na-/K-decorated Mo2TiC2Tx, we further observe a reduction of O loss as determined using EDS of 18.4 at%/39.2 at%, respectively, which may be due to prevention of loss of O in the C sublattice (as shown in SIMS) or the Tx = O groups. It is important to note that, as shown in the raw EDS (Supplementary Figs. 18 & 24), O is still present even up to 1100 °C, which matches the O content in the C sublattice as shown using SIMS (Fig. 4b). With regards to the cation-decorated samples at this O loss region as CO/CO2 species, we believe that this may suggest that alkali cation occupation in the VMo sites affect the migration of C at the defective areas onto the surface of the MXene during phase transition52, which can contribute to the reduction in rate and total loss of C through reduction of O surface groups as determined using TGA. However, other methods, such as neutron diffraction, can further help to understand the C/O occupancies in the carbon sublattice and Tx sites before and after annealing in cation-decorated MXene structures.

Regardless, in order to identify any electronic structure changes by increased Mo content in the δ-(Ti, Mo)Cy lattice on the electronic structure of Mo and C, we used XPS analysis (full deconvolutions in Supplementary Figs. 25–28, Supplementary Table 1). In Na- and K-decorated Mo2TiC2Tx, there is an increased contribution of Mo2+ to the pattern as compared to Mo2C (Supplementary Fig. 25b). This feature is accompanied by a similar increase in the Ti3+-C contribution in the Ti 2p pattern in the post-annealed Na- and K-decorated Mo2TiC2Tx samples, as shown in Supplementary Fig. 26b. Paired with the in situ XRD, STEM, and EELS data, this points to an increased Mo in the M6C octahedra of the cubic δ-(Ti, Mo)Cy lattice62, which would pronounce the effect of valence electrons in the antibonding states in Mo on the M-C bond63. As a result, a shift in the C 1 s pattern is visible in Supplementary Fig. 27b, where the M-C contribution shifts to a lower binding energy for the annealed Na- and K-decorated Mo2TiC2Tx as compared to control Mo2TiC2Tx, suggesting a reduced binding between Mo-C as Mo occupies a non-preferable crystal structure in the cubic δ-(Ti, Mo)Cy63,64. Similarly, using EELS, we determined the Mo:Ti ratio in the δ-(Ti, Mo)Cy at 0.65, 1.62, and 0.85 of regular (as-is), Na- and K-decorated Mo2TiC2Tx MXene, respectively (raw EELS shown in Supplementary Fig. 29). Therefore, the in situ XRD, STEM, and EELS data paired with this XPS Mo 3d5/2 and C 1 s peak shifting indicates that a higher content of Mo is present in the M-X octahedra of cubic δ-(Ti, Mo)Cy lattice rather than migrating out to form Mo2C62.

After establishing the effect of Na+ and K+ on the phase formation and transformation at defective sites on Mo2TiC2Tx MXenes, to further expand our work, we also examined the high-temperature stability of the thicker M4C3 Mo2Ti2C3Tx MXene films and the effect of alkali cation stabilization. We observed that alkali cations had similar effects on the stability of Mo2Ti2C3Tx MXene films (Supplementary Fig. 30), which likely suggests that alkali cations also occupy surface VMo sites in Mo2Ti2C3Tx MXene as well. However, further studies are necessary to fully understand the generalizability of this behavior to all MXene systems. Overall, we believe future studies monitoring the alkali cation occupation during either controlled concentration in SIMS or electrochemical cycling in TEM may be necessary to provide further evidence on the cation locality in MXene flakes during cycling in applications such as alkali cation batteries. After the findings in this study, we strongly believe that future studies must investigate the effect of bonding and charge transfer from alkali cations to the MXene to understand why alkali cations prefer this vacancy occupation.

Discussion

As increasing attention is paid to phase control of nano-sized transition metal carbides to meet energy and extreme environment applications, the development of atomic-level control of phase stability of these phases at the nanoscale is critical. As a result of these observations in three MXene systems (Ti3C2Tx, Mo2TiC2Tx, and Mo2Ti2C3Tx), we believe that the alkali metal cations, and potentially other metal cations, serve to affect the ambient and high-temperature behavior of in two main features. First, we observe that alkali cations occupy vacancies of the defective surface transition metal sites, which is essential for the complete understanding of MXenes behavior in alkali batteries and provides evidence for other reports on the increased stability of cation-decorated MXenes. Second, we see that the alkali metal cations stabilize the MXene crystal structure to higher temperatures by occupying vacant surface transition metal sites, which are the initiation site of phase transition for MXenes, as evidenced by the higher temperature at which the (00 L) peaks were visible in all MXene systems. Lastly, we observe that alkali cations further affect the phases formed during high-temperature annealing by reducing the loss of C during phase transition of the MXenes, as evidenced by Mo2TiC2Tx phase transition and TGA data as well as the Ti3C2Tx stabilization in Na+ overetched Ti3C2Tx as compared to the overetched Ti3C2Tx. At high temperatures, we demonstrate the stabilization potential metal cations have to 1) stabilize the defects in MXenes in high-temperature environments, 2) control the defect-based phase transition of MXenes, and 3) expand into phase control and stabilization in other 2D materials or nanocrystalline ceramic systems. Overall, we believe that future studies using cations as control mechanisms of 2D carbide and nitride phase stability and transition will further enable the development of this MXene 2D material family for advanced application as a nano-sized block for nanoceramics.

Methods

MAX phase synthesis

In order to synthesize the eventual MXenes, the precursor MAX phases (Optimized Ti3AlC2, Mo2TiAlC2, and Mo2Ti2AlC3) were synthesized in a similar manner to previous publications reporting phase-pure protocols56,65,66. In short, to synthesize this MAX phase, we mixed molar ratios of elemental powder mixtures using elemental powder precursors of each needed transition metal (−325 mesh, Alfa Aesar), Al (−325 mesh, Alfa Aesar), and calcinated coke powder as our C source. For the optimized Ti3AlC2, we mixed non-stoichiometric mixtures of TiC (−325 mesh, Alfa Aesar) in TiC:Al:C molar ratios of 2:2.2:1.25. We mixed each MAX phase in 80 g batch sizes by placing the aforementioned molar mixture by placing the powder in a 250 mL Nalgene high-density polyethylene bottle (HDPE) with yttria-stabilized zirconia balls at a ball-to-powder mass ratio of 2:1. After preparing the powder and ball mixture in the bottle, the HDPE bottle was then mixed using a rolling jar ball mill (Shimpo) for 18 h. After mixing, the powders were then removed from the HDPE bottle and added to a 100 mL rectangular alumina crucible and fired in a Carbolite Gero 1700 at temperature (1400 °C and 1600 °C for Ti3AlC2 and Mo2TiAlC2/Mo2Ti2AlC3, respectively) for 4 h using a 3.5 °C/min ramp rate under Ar flow (99.999%). After firing, the sintered MAX phases were allowed to cool in the furnace at a de-ramp rate of −10 °C/min until at room temperature. The MAX phase block was then removed from the furnace and drilled and sieved to a fine (<71 µm size) powder in preparation for etching. For the optimized Ti3AlC2, we treated this MAX phase in 9 M HCl in an ice bath overnight, washed it to neutral, and filtered the powder before continuing for MXene synthesis.

MXene synthesis

To synthesize the MXenes from the MAX phase precursor, we used the appropriate protocol for each MXene. Overall, we slowly added the MAX powder (~1 g/2 min) to an appropriate HDPE Nalgene bottle containing the acidic solution stirring at 500 RPM with a 1.5 inch stir Teflon-coated stir bar at room temperature on a Corning hot/stir plate. For 1 g Ti3C2Tx, we etched in a solution of 3 mL 50 wt% HF, 9 mL de-ionized (DI) water, and 18 mL 12 M HCl at 35 °C for 24 h. For 1 g of Mo2TiC2Tx and Mo2Ti2C3Tx, we etched in 10 mL of 50 wt% HF at 55 °C for 96 h. For the 2 day etched Mo2TiC2Tx, we reduced the etching time while keeping all other conditions the same. After the etching procedure, the solution was then centrifuged at 3234 × g for 5 min until the natant was clear. The clear natant was then discarded, deionized (DI) water was added to the centrifuge tube, the newly formed exfoliated MXene sediment was redistributed, and then subsequently centrifuged at the aforementioned speed and time. This cycle was repeated until the pH reached >6, which typically occurred after 150–200 mL of DI water was added per gram of the original MAX added.

After neutralization, the sediment was filtered on a ~ 1 µm pore size filter paper (Whatman). To yield single-flake MXenes from the exfoliated MXene, we used typical delamination conditions for each MXene stirring at 500 RPM. For 1 g of Ti3C2Tx, we delaminated in a lithium chloride (LiCl, Fisher Scientific) water solution by adding 1 g of LiCl to 50 mL of DI water at 60 °C for 1 h. For 1 g of Mo2TiC2Tx and Mo2Ti2C3Tx, we delaminated in 10 mL of a 5 wt% TMAOH solution at 55 °C for 4 h. After each delamination time, the solution was then centrifuged at 21900 × g for 5–10 min until the natant was clear, which was then discarded. After discarding the clear natant, the newly formed MXene clay at the bottom of the tube was agitated with a cleaned glass rod, which was then cleaned of MXene clay using a spray DI water bottle. After the clay was removed from the bottom of the centrifuge tube, DI water was added to the centrifuge tube and the redispersed Mo2TiC2Tx clay was then subsequently centrifuged at the aforementioned speed and time. This cycle was repeated until the pH reached <8, which typically occurred after 200–250 mL of DI water was added per gram of the original MAX added. After reaching <8 pH or after 4–5 washes for LiCl, we again redispersed the MXene clay, added DI water, and centrifuged the dispersed MXene clay at 2380 × g for 30 min to yield single-to-few layer MXene in the natant.

After 2380 × g–30 min, we removed the MXene single-to-few layer natant from the centrifuge tube using a pipette for storage. For the in-situ scanning transmission electron microscopy (STEM) annealing measurements, the Mo2TiC2Tx single-to-few layer natant was added to a fresh centrifuge tube and centrifuged at 21900 × g for 10 min until the natant was cleared to form a clay. Before the STEM measurements were conducted, the clay was redispersed into DI water to form a dilute colloidal solution.

To conduct the in-situ x-ray diffraction (XRD) annealing experiments, we filtered ~5–10 mL of the MXene single-to-few layer natant from the stock solution to form free-standing MXene films on a 20 mm glass vacuum filter using pre-wet 0.3 µm cellulose membrane filter paper, which we allowed to filter overnight. After removing the film from the filter paper, we then cut each MXene film into roughly 5 x ~ 20 mm rectangular ribbons using a cleaned razor blade on a cleaned glass slide, which yielded typically 3 ribbons of MXene film samples from one film.

MXene cation decoration

To decorate the MXene with cations, ~30 mg of was added to a DI water solution in a cleaned glass beaker and stirred at 500 RPM for 5 min to ensure dispersion. The amount of DI water added to the beaker was to attain an alkali cation salt concentration of 25 mg/mL to stay well below the saturation point for each salt in water. After 5 min, a 500x molar ratio of alkali cation salt to MXene was added to the solution, and the solution was stirred for 15 min to ensure dissolution and dispersion of the salts. After 15 min, the solution was washed with a total of 100 mL of DI water using centrifugation at 21900 × g for 1 min to wash away excess salt. After centrifugation, a cleaned glass rod was used to redisperse the cation-decorated MXene from the centrifuge tube, then probe sonicated using a Fisherbrand ultrasonic probe with a microtip at a amplitude of 40 for 5 min with a 5 s on/off rate in an ice bath to redisperse the MXene. After probe sonication, the solutions were used for STEM preparation, while the cation-decorated MXene was filtered using a Celgard filter paper overnight in a 20 mm glass filter to form a free-standing film.

In-situ XRD measurements

An Anton-Paar DHS1100 domed hot stage paired with a Bruker D8 Discover diffractometer was used for the in-situ XRD measurements in this study. To prepare for the in-situ annealing experiments, the aforementioned MXene film ribbons were stacked and affixed to a AlN stage using stainless steel clips at the ribbon edges. The sample height was calibrated for the in-situ runs using two angled laser beams, where the sample was considered calibrated for z-offset once the two laser beams were concentric. This laser focusing method was confirmed using a corundum standard, of which our results in this study are consistent with previous reports on the corundum standard before our in-situ experiments47. After laser focusing, a protective graphite dome was affixed to the DHS 1100 stage to protect the sample during in-situ annealing from environmental (non-gas related) effects. To ensure our sample was an inert environment, high purity (99.998%) Ar was flowed bottom-up at a regulator pressure output of 50 kPa from the AlN stage to prevent sample oxidation.

For our in-situ annealing results, we first took the XRD diffraction pattern of our MXene stage at room temperature for 15 min, which was then followed by a ramp from room temperature to 300 °C at a ramp rate of 60 °C/min before the next 15 min scan. Each diffractogram in this study (in-situ or ex-situ) was taken using a Vantec 2D area detector paired with a Cu Kα radiation (monochromatic wavelength at λ = 1.5406 Å) focused from 5–75° 2θ using a step size of 5° 2θ, which captured a full diffractogram area of 0–90° 2θ. After the 300 °C scan, each subsequent scan was taken at increments of 100 °C (300 °C, 400 °C, 500 °C, …) with a ramp rate of 60 °C/min until the max temperature of 1100 °C was reached. After the 1100 °C 15 min scan was completed, the stage was allowed to cool to room temperature under forced air convection (cool rate of an averaged ~20 °C/min) over the graphite dome until the stage reached 200 °C, at which Newtonian cooling was allowed to cool the stage to room temperature. After the stage reached room temperature, the graphite dome was removed and the annealed MXene was then removed from the AlN stage for ex-situ testing. For ex-situ XRD testing, the same diffractogram scan was used, but the pre- and post-annealed films were placed onto ~1 × 1 cm double sided amorphous carbon tape on a cleaned glass slide.

Scanning transmission electron microscopy (STEM)

The previously mentioned Mo2TiC2Tx colloidal solution was then dropped cast onto a plasma-cleaned commercial microelectromechanical (MEMS) based in situ heating platform from Protochips, Inc. with a SiN support membrane. Arrays of holes were created using a Ga+ ion focused ion beam to ensure the presence of freestanding Mo2TiC2Tx flakes over a vacuum. The STEM imaging, EELS characterization and in situ heating experiments were conducted using a Nion UltraSTEM operating at 100 kV with a beam current of 40 pA and equipped with a spherical aberration (Cs) corrector to achieve 1 Å spatial resolution. A convergence angle of 31 mrad was used, with a HAADF detector inner and outer collection angles of 86 mrad and 200 mrad. Additional in situ STEM heating experiments and EELS acquisitions taken using a JEOL NEOARM operated at 80 kV.

For image analysis, Nion Swift software was used in conjunction with a custom python script67 for fast-Fourier transform mapping for phase analysis. In all shown STEM images, with the exception of Supplementary Fig. 10, we re-colorized the plots for consistent phase identification. For the vacant area calculations shown in Supplementary Fig. 17g, a series of 0.5 nm-thick line scans were taken across the entire area of the Mo2TiC2Tx single-flakes for control, Na+, and K+ decorated samples. To determine the vacant area, the line scan intensity values for the center of the vacancy clusters were taken for each Mo2TiC2Tx flake and used as a cutoff value. This cutoff value was then used to determine whether an individual data point along the line scan was of a vacant area. For each Mo2TiC2Tx flake, the counted vacant area points were then divided by the total number of line scan data points to determine the vacant area on the flake. For each sample, four to five flakes per sample category (control, Na+, K+), which comprised over 300 total line scans per sample category, were taken to calculate the vacant area.

To calculate strain around vacancy sites, a roughly 3 × 3-nm square was placed around each nanopore vacancy site for each Mo2TiC2Tx flake at 500 °C to calculate the FFT. Using the FFT, the a-lattice parameter (a-LP) of the local area of the Mo2TiC2Tx flakes was calculated (a). For each image, typically 4–5 vacancy areas were captured. To calculate strain, the “un-strained” a-LP value (a0) was taken by taking an FFT map of the full flake area (including defects, typically ~15-nm × ~ 15-nm). Using this analysis, three-to-four flakes of Mo2TiC2Tx of the various sample categories (control, Na+, K+) were used to calculate the averages. The equation used to calculate strain was as follows:

$${\varepsilon }_{a-{LP}}=\frac{a-{a}_{0}}{a}$$
(1)

where a represents the a lattice constant of Mo2TiC2Tx around the vacancy site, a0 represents the overall a lattice constant in the non-defective areas of the Mo2TiC2Tx flake, and \({\varepsilon }_{a-{LP}}\) represents the calculated strain in the a lattice constant around the vacancy sites.

Secondary ion mass spectrometry

SIMS experiments were performed using a CAMECA IMS SC Ultra instrument. Cs+ primary ions with ultra-low impact energy (100 eV) and high angle of incident (75°) were used. To achieve atomic depth resolution several modifications of the measurement procedures were introduced, such ion polishing, extraction parameters optimization, super-cycling, and advanced beam positioning57.

Scanning electron microscopy

In this study, a JEOL JSM-7800F field-emission scanning electron microscope was used to capture the morphological features of the pre-and post-annealed MXene films. For image clarity, all samples were sputter coated with Au using a Denton Desk V Turbo to enhance the clarity of the SEM images. To image the samples, the SEM settings were set to a 15 kV acceleration voltage, a probe current of 8, and an aperture of 4. For compositional information, energy dispersive x-ray spectroscopy (EDS) was used using a EDAX octane super detector with its EDAX TEAM software for analysis. The EDS scans were conducted by zooming in on the film surface to 1,000,000x followed by 30 s of EDS exposure in the point scan.

X-ray photoelectron spectroscopy

For x-ray photoelectron spectroscopy (XPS) measurements, a Thermo K-Alpha XPS system with a spot size of 400 µm and a resolution of 0.1 eV was used after using Ar+ milling to take off the surface oxidation features of the Mo2TiC2Tx film. The spectra were processed using its propriety Thermo Avantage software.

Thermogravimetric analysis and gas chromatography

The weight loss of the samples was collected on a thermogravimetric analyzer (TGA, TA Instruments TGA55). The samples were heated from room temperature to 1000 oC at 10 oC/min. To understand the gas evolved during this heating, we also performed gas chromatography-mass spectrometry (GC-MS, Shimadzu GCMS-QP2020 NX) analysis which is connected directly to the gas exhaust from the TGA. The gases evolved were confirmed based on the retention times for standard samples using a thermal conductivity detector (TCD). Data was compared with a standard gas mixture from Scotty Analyzed Gases.

Density functional theory

To perform density functional predictions of the adsorption energies we utilized plane wave projector augmented wave density functional theory as implemented in the VASP code68,69,70,71, version 6.3.2. We simulated the MXene surface utilizing a supercell with 80 atoms of a single unterminated Mo2TiC2. Adsorption energies were obtained for a single Li, Na and K atom on the bare surface and with a single Mo atom removed to simulate an isolated single surface vacancy. The structures were fully geometrically relaxed in all cases, with the structures and eventual energies obtained using both the PBE72 density functional and the self-consistent van der Waals functional OptB8873. Calculations used a well-converged 500 eV plane wave cutoff, a 3 × 3 × 1 gamma-point centered k-point grid, and VASP’s “accurate” settings.