Photoredox phase engineering of transition metal dichalcogenides

Transition metal dichalcogenides (TMDs) exhibit a wide range of electronic properties based on their crystallographic structure^1,2. Of particular interest is the phase transition from the trigonal 2H to the octahedral 1T (including 1T and distorted 1T) phase, which changes the electronic behaviour from a semiconducting to a quasi-metallic^1,2,3 state. This phase transition has been exploited to enhance field effect transistor device performances by lowering the contact resistance^3,4,5, to increase the hydrogen evolution reactivity^6,7,8,9 and to unlock a wide range of chemical exfoliations and functionalization down to single-layer TMDs^10,11,12,13. More recently, it has also been used to advance flexible rectifying antennas¹⁴ and lithium–sulfur battery technology¹⁵.

Chemical lithiation using the organolithiation agent n-butyllithium (n-BuLi), which was first proposed in 1975 (ref. ¹⁶), remains the most popular route to achieve the 2H to 1T/1T′ phase transition. Here, the semiconducting 2H-TMD is immersed in n-BuLi, which yields the 1T/1T′ phase after several days. Mechanistically, n-BuLi first donates an electron to the 2H-TMD, which enables Li-ion intercalation into the van der Waals (vdW) gap to complete the 2H to 1T/1T′ phase transition^1,16. This charge-transfer-limited reaction mechanism is further supported by electrochemical and electron microscopy studies, which suggest a sequential process of lithium adsorption through initial charge transfer, followed by the phase transition and irreversible phase conversion at lower potentials^{17,18,19,20,21,22,23,24,25}. However, we still lack a fundamental real-time nanoscale picture of the 2H to 1T/1T′ phase transition. This hampers the use of phase engineering of TMDs in crucial aspects: (1) the current reaction kinetics are extremely slow, requiring, for example, tens of hours for complete conversion of 2H-MoS₂ to the 1T/1T′ phase; and (2) the use of pyrophoric n-BuLi poses a substantial safety risk for any scale-up of these materials.

Here, using in situ optical reflectance interferometric contrast microscopy (RICM), we uncover mechanistic details of the 2H to 1T/1T′ phase transition with diffraction-limited resolution (about 200 nm). We discover a photoredox-activated reaction pathway that overcomes the otherwise slow charge-transfer-limited reaction occurring under ‘dark’ conditions, allowing us to markedly speed up the phase transition and demonstrate a new photoredox phase patterning methodology. Finally, with the obtained insights, we can eliminate the use of pyrophoric n-BuLi in favour of non-pyrophoric polycyclic aromatic hydrocarbon lithiation systems.

Figure 1a shows the experimental setup (left) used to monitor in situ the chemically induced phase transition from the semiconducting 2H to the quasi-metallic 1T(1T/1T′) phase (right). Mechanically exfoliated 2H-TMD flakes were placed in an air-tight, optically accessible cell that could be filled with n-BuLi using a syringe pump. The reaction was spatio-temporally resolved with an RICM using LED illumination at 730 nm and 455 nm.

a, Optical setup for this study and the schematic for light-driven 2H to 1T phase transition of TMD. Illumination with 455 nm accelerates the phase transition by up to two orders of magnitude. The bottom side of the flake is observed through a glass substrate, whereas the top side is in contact with the solution. b, Image of mono- (1L) and bilayer (2L) 2H-MoS₂ and 1T-Li_xMoS₂ under different illumination wavelengths. The 2H phase (left images) changed to 1T(1T/1T′) phase (right) in 698 min under illumination with 730 nm (red, top row), and 21 min with 455 nm (blue, bottom row). c,d, Photoluminescence and Raman spectra of 2L 2H-MoS₂ (as exfoliated) and 1T-MoS₂ produced using three different methods: 48 h n-BuLi immersion in the dark (grey), 24 h of 730 nm (below band gap) illumination (red) and 25 min of 455 nm (above band gap) illumination (blue). The higher optical resolution achieved at 455 nm shows more fine structure in the flake (see Supplementary Fig. 10 for a comparison of the same flake under 455 nm and 730 nm). Scale bar, 5 μm (b).

Typical RICM images for mono- and bilayer MoS₂ before and after the phase transition under both illumination conditions are shown in Fig. 1b. We note that the higher diffraction-limited resolution at 455 nm shows a more detailed morphology compared with that at 730 nm (for more information, see Supplementary Information section 1). Independent of the illumination, at the start of the reaction the bilayer (2L) appeared brightly reflective, whereas the monolayer (1L) was substantially darker with intensity values similar to that of the glass substrate. The brightness reduces during the reaction and the images after 698 min under 730 nm illumination show a drastic reduction in the intensity, leaving behind a smooth surface at the edges of the flake and a more rugged central region characterized by irregular brighter and darker spots. We propose that the darkening is associated with the 2H to 1T phase transition. Flake darkening was also observed under 455 nm, but it took only 21 min, corresponding to a 33-fold acceleration of the reaction speed under 455 nm illumination.

We confirmed that the darkened 2L regions correspond to the 1T-MoS₂ phase by steady-state photoluminescence, Raman and reflectance microscopy measurements. Before chemical lithiation, bilayer 2H-MoS₂ exhibited photoluminescence at around 660 nm (direct bandgap) and around 800 nm (indirect bandgap) alongside pronounced Raman peaks at approximately 385 cm^−1 and 407 cm^−1 (E_2g and A_1g modes, respectively; Fig. 1c,d, black line). After the 2L fully darkened, we observed a loss of all photoluminescence emission, as well as red shifts of the E_2g Raman peaks and the emergence of J-peaks in both illumination conditions, indicating successful 1T formation (Fig. 1c,d; 730 nm (red) and 455 nm (blue))²⁶. We verified that the same photoluminescence and Raman signatures are observed when exposing 2H-MoS₂ to n-BuLi for 48 h under dark conditions (Fig. 1c,d, dark grey) and that the monolayer region is also successfully converted to 1T (refs. ^1,26,27) (for more details, see Supplementary Information sections 2 and 3). Moreover, reflectance measurements show the loss of the A- and B-exciton absorption peak in line with the formation of a 1T phase^28,29 (for more sample characterization details, see Supplementary Information section 4).

These results confirm that the darkening of the flake observed in RICM corresponds to the phase transition from 2H-MoS₂ to its 1T phase and can be sensitively and optically monitored in situ. Furthermore, using below-bandgap illumination at 730 nm enables us to capture the phase transition in MoS₂ as it would occur under nominally dark conditions (Extended Data Fig. 1). By contrast, the above-bandgap excitation at 455 nm coincides with the C-exciton absorption band of 2H-MoS₂, suggesting a photo-activated mechanism that markedly accelerates the phase transition³⁰.

We study the mechanistic origin of the phase transition under 730 nm illumination by analysing the RICM images as a function of time after n-BuLi addition (Fig. 2a). After 74 min, the 2L edges become visibly darker, and by 147 min, a progression of a phase front at the 2L/1L terrace can be seen. A similar phase front progression at the 2L/glass edge was reached after 220 min. Once the 1T phase nucleated, the 1T/2H phase boundary moved into the flake with a constant velocity of 0.1 nm s^–1, remaining sharp throughout. Interestingly, after 366 min, brighter and darker regions formed whenever two-phase boundaries of different orientations converged. We attribute these regions to microscopic wrinkles arising from a lattice mismatch^20,31. Similar phase frustration effects were previously observed for phase transitions occurring in lithium cobalt oxide³². After the wrinkles formed for the first time, the surface morphology of the 1T phase was no longer smooth but instead stayed rugged with brighter and darker features emerging throughout the remaining phase transition (698 min, see also Supplementary Video 1; for more details, see Supplementary Information section 1).

a, Snapshots of the phase transition visualized at 730 nm. b, Histogram analysis of the 1L (left) and 2L (right) regions. c, Selected reaction dynamics from two different points of the 2L indicated in a. Both spots are located at a similar distance (1 μm) from the edge. d, Snapshots of the phase transition visualized at 455 nm. e, Histogram analysis of the 1L (left) and 2L (right) regions. f, Power dependence of the reaction dynamics under 455 nm laser illumination for three different flakes. Each curve was obtained 1 μm away from the 2L/glass edge of the studied sample. Histograms are normalized to the mean of the intensity distribution 1 min after n-BuLi was added. Scale bar, 5 μm (a,d).

To quantify the reaction dynamics at 730 nm, we performed a histogram analysis of the 1 L and 2 L regions. As shown in Fig. 2b (left), the 1L exhibited a single-peaked distribution in the histogram map. This feature quickly reduced in intensity to 0.9 over 50 min with no changes to its width, indicating a homogeneous surface change during the initial reaction with n-BuLi. As the 1L cannot intercalate Li ions, this initial decay suggests a facile electron donation to the surface. We note that the intensity at the glass/n-BuLi interface changes up to 15% over 12 h in this experiment, which accounts for the long-term increase in reflection intensity (Supplementary Information section 1).

Similarly, at the beginning of the reaction, the 2L (Fig. 2b, right) exhibited a single-peaked feature in the histogram map that rapidly decayed to a value of 0.9 over 50 min—the same timescale observed for the 1L. To confirm that this reduction corresponds to a surface change of the 2L, we carried out the photoluminescence measurements that showed that the vdW interaction between each layer is weakened within around 50 min (for more details, see Supplementary Information section 7). As the width of the initial feature does not change during this time, we attribute this process to a homogeneous surface electron donation, as observed for the 1L. After 50 min, the histogram peak at 0.9 reduces in occurrences and is converted into a lower intensity feature at an intensity of 0.7, which becomes prominent after about 300 min. This feature signifies the transition of the 2H phase (bright) to the 1T phase (dark) at the edges of the 2L (compare with Fig. 2a). After around 400 min, a second peak emerges at a lower intensity (approximately 0.6), accompanied by a broadening of both lower intensity features, coinciding with the formation of a wrinkled 1T region at the centre of the 2L.

To further illustrate the intrinsic variability of the phase transition as observed at 730 nm, Fig. 2c shows the intensity profile over time at the 2L/1L terrace and 2L/glass edge (2L_Terrace and 2L_Edge, respectively). The 2L_Terrace completed the phase transition in just 140 min, whereas the equidistant 2L_Edge required 340 min. This discrepancy probably originates from different activation barriers along these two interfaces, highlighting how our approach can be used to map out intercalation energy barriers in these materials measurements (for more details, see Supplementary Information section 6).

Taken together, the first step of the phase transition in MoS₂ is rapid surface electron donation from n-BuLi to the whole flake. In 1L-MoS₂, this is sufficient to form the 1T phase. In 2L-MoS₂, Li-ion intercalation causes the 1T phase to form at the edges of the flake with a sharp phase boundary that moves with a constant velocity into the centre of the flake, suggesting a charge-transfer limited reaction mechanism. Finally, once the two-phase boundaries of different orientations converge, a wrinkled 1T phase is formed.

We repeated the same analysis for the 455 nm illumination. As shown in Fig. 2d, after 5 min, we already observed the 2L edges darkening and forming a sharp phase boundary between the bright 2H and dark 1T phases. After 13 min, however, this phase boundary was more diffuse and the centre of the flake appears noticeably darker. After 17 min, no clear phase boundaries could be identified, and the flake appeared to darken homogeneously with no signs of wrinkle formation (Supplementary Video 2).

Analysis of the histogram maps in Fig. 2e (left) for the 1L showed an identical, albeit accelerated surface reaction compared with the 730 nm illumination. Again, we observed an initial homogeneous darkening of the whole 1L to an intensity of 0.9 in just 5 min, followed by no further changes. The 2L (Fig. 2e, right) mimicked this intensity reduction in the first 5 min after the addition of n-BuLi, confirming that the 455 nm illumination accelerates the surface electron donation step of the reaction. After about 5 min, the initial distribution reduces in occurrences and shifts in intensity to form a new distribution at 0.4 within 10 min, in a markedly different behaviour than what we observed at 730 nm.

Mechanistically, these observations suggest a photoredox-activated mechanism. Initially, photo-excitation into the C-exciton of MoS₂ at 455 nm must lower the activation barrier for electron donation, to explain the accelerated surface darkening³³. We propose that C-exciton excitation causes efficient surface-hole generation, thereby remarkably increasing the electron transfer rate (as seen by the change time of the initial darkening around 100 min at 730 nm compared with about 5 min at 455 nm) in line with previous transient absorption reports³⁴. The associated opening of the vdW gap then enables rapid Li-ion intercalation from the edges of the flake. Owing to the continued photo-induced hole formation, this process is no longer strictly charge-transfer limited, causing the phase boundary to broaden and to accelerate the overall reaction speed.

To further support this mechanistic picture, we carried out power-dependent measurements on 2L-MoS₂, varying the illumination fluence at 455 nm between 20 mW cm^−2, 110 mW cm^−2 and 510 mW cm^−2. Figure 2f shows the intensity curves for regions located 1 µm away from the 2L/glass edge. We find that the photoredox-activated phase transition depends on power. Lower fluences (20 mW cm^−2; Fig. 2f, light blue) approach the shape of the intensity curves observed under 730 nm illumination (Supplementary Video 3), indicative of a charge-transfer-limited reaction composed of an initial surface change followed by a period of constant intensity before the phase boundary traverses over the region. By contrast, higher fluences (510 mW cm^−2; Fig. 2f, purple), lack clear features of a phase boundary and simply reduce in intensity to a constant value in as little as 8 min (about 90-fold acceleration), demonstrating that the process is deviating from a strict charge-transfer-limited reaction profile.

A particularly promising application of the discovered photoredox-driven reaction pathway lies in the possibility of inscribing arbitrary phase patterns with diffraction-limited edge resolution into few-layer TMDs. Figure 3a shows the experimental setup used to create a well-defined T-shaped pattern of the 1T phase into 1L and 2L 2H-MoS₂. Using 455 nm illumination at a power density of 510 mW cm^−2, we were able to inscribe the desired phase pattern into 1L-MoS₂ in 30 s and into 2 L MoS₂ in 13 min (Fig. 3a, top right) and achieved a channel width of 3.3 µm with a deconvolved 1T/2H edge sharpness of σ = 161 nm (Fig. 3a, bottom right). Photoluminescence and Raman imaging confirmed the successful phase transition (Extended Data Fig. 2 and Supplementary Fig. 9). It is worth noting that these power densities are about six orders of magnitude lower than conventional laser-writing approaches (for more details, see Supplementary Information section 8).

a, Schematic of the photoredox phase patterning of a 1T phase pattern on 1L- and 2L-MoS₂ (left). The image of the patterned flake (right) and the average across the red rectangle show the profile of the patterned 1T phase. b, Response comparison of non-patterned (2H) and photoredox-phase-patterned (1T/2H/1T) 1L-MoS₂ photodetector. The bottom panel shows the ohmic-like transport behaviour by photoredox phase patterning. Scale bar, 5 μm (a).

To demonstrate the electronic quality of the photoredox-patterned phase, we used our approach to inscribe a 1T/2H/1T phase junction into 1L 2H-MoS₂. Using a power density of 1 W cm^−2 enabled us to manufacture this junction in 10 s, without affecting the material. We then deposited gold electrodes onto the 1T phase area to fabricate a simple phase-engineered photodetector. As shown in Fig. 3b, the phase-engineered device exhibited enhanced photocurrent density to devices made with 2H-MoS₂, showing responsivity of 85 A W^−1 compared with 10 A W^−1 (ref. ³⁵) (for more details, see Supplementary Information section 9). Using Fowler–Nordheim analysis, we verified that this enhanced photoresponsivity arises by switching from a Schottky contact behaviour to an ohmic-like contact behaviour (Fig. 3b, bottom), which enhances the transport properties^3,36 (Supplementary Information section 10). Furthermore, it demonstrates that the diffraction-limited edge resolution does not notably impede electron mobility at the phase edges³⁵.

Next, we explored this photoredox reaction across different TMDs, such as WS₂, MoSe₂ and WSe₂. Here we will focus on the comparison of MoS₂ and WS₂, but a full discussion of all samples is provided in Supplementary Information sections 2, 12–14. Figure 4a compares the spatio-temporal changes of MoS₂ (top, blue) and WS₂ (bottom, orange) in the low-power regime on the addition of n-BuLi at 455 nm illumination. As described previously, the 2L-MoS₂ shows surface darkening together with phase boundary propagation, whereas the 1L-MoS₂ reduces only in intensity. Conversely, the 1L and 2L WS₂ proceed by a homogeneous surface darkening mechanism with no visible phase boundaries at any stage during the reaction (Supplementary Video 4).

a, Chemically induced phase transition of MoS₂ and WS₂ observed in RICM. b, Reaction dynamics at selected points indicated in a for MoS₂ and WS₂ during chemical lithiation. No data were recorded for WS₂ between about 220 min and 250 min. c, Electrochemical discharge/charge voltage profiles of MoS₂ and WS₂ versus Li/Li⁺. d, Energy landscape proposed for the phase transition and conversion reaction of MoS₂ and WS₂. The blue line shows the light-activated pathways. Arrows at the conversion barriers indicate the thermodynamic preference of the system. Scale bar, 5 μm (a).

To further analyse this difference, Fig. 4b shows the intensity curves for comparable 1L and 2L regions of the respective TMD. Although the 1L- and 2L-MoS₂ exhibit a simultaneous, rapid surface darkening followed by a decrease in intensity in 2L-MoS₂ because of the phase boundary motion, this behaviour is absent in WS₂. The 1L-WS₂ shows a rapid intensity decay within about 100 min, which is not matched by any dynamic feature observed in the 2L-WS₂. Instead, the 2L-WS₂ decays initially to an intermediate intensity not observed in the 1L-WS₂ and then continues to decay slowly over the next 300 min.

These results highlight that WS₂, in contrast to MoS₂, proceeds exclusively by a surface-driven reaction mechanism that seems different between 1L- and 2L-WS₂. We carried our Raman and photoluminescence measurements before and after the reaction of WS₂ with n-BuLi (Supplementary Information sections 11 and 12). 2H-WS₂ showed the expected photoluminescence and Raman peaks. However, after exposure to 455 nm light in the presence of n-BuLi, the Raman spectrum indicated a 2H phase with a high signal-to-noise ratio, and no photoluminescence could be observed. On the basis of these results, we conclude that the chemical lithiation of WS₂ under 455 nm photo-excitation does not produce the desired 1T phase, but instead leads to phase conversion to tungsten and lithium sulfides³⁷. We further attempted to resolve the phase transition in WS₂ under below-band-gap illumination at 730 nm; however, over the time frame of 48 h, no changes to the flakes could be observed, suggesting less favourable energetics for the chemical lithiation by n-BuLi, as compared with MoS₂.

To rationalize these findings, we fabricated electrochemical devices based on (multilayer) 2H phase TMD powders and measured the galvanostatic discharge–charge response of MoS₂ and WS₂ as shown in Fig. 4c (for more details, see Supplementary Information section 13). MoS₂ (blue) exhibited two plateaus during the galvanostatic discharge process at 1.1 V, corresponding to the reversible phase transition from 2H to 1T, and another at 0.57 V, assigned to conversion by irreversible decomposition^17,18. As the redox potential of n-BuLi is 1 V (refs. ^1,38), the charge-transfer-limited phase transition in MoS₂ proceeds by a small driving force of about 0.1 V. The slow reaction speed is, therefore, because of a reasonably large activation energy for electron donation. Illumination with 455 nm light increases the (surface) hole concentration and thereby reduces the activation barrier for electron donation, resulting in an accelerated reaction rate. Importantly, photo-excitation of MoS₂ does not provide enough driving force to activate the conversion reaction at 0.57 V.

This picture changes for WS₂, for which the phase transition and conversion potential are close together and located between 0.75 V and 0.8 V (Fig. 4c, orange). Chemical lithiation is, therefore, thermodynamically unfavourable, resulting in an extremely slow reaction with low yield after more than a week of n-BuLi soaking at room temperature (for more details, see Supplementary Information section 11). As for MoS₂, illumination at 455 nm reduces the activation barrier for electron transfer. However, owing to the close proximity of the phase transition and conversion, continuous photo-excitation cannot selectively stop at the 1T phase for WS₂. Instead, the conversion reaction is also photo-activated and occurs shortly after the 1T phase has formed. Examination of the MoSe₂ and WSe₂ corroborates this picture, in which MoSe₂ could be successfully transitioned to the 1T phase, and WSe₂ showed conversion behaviour (Supplementary Information sections 12–14).

Taken together, we propose the energy level scheme presented in Fig. 4d to explain our findings. For MoX₂ (X = S, Se), the thermodynamic landscape puts the reversible lithiated 1T phase into a local minimum with respect to its 2H phase and decomposition products^39,40,41,42 (for more details on reversibility, see Supplementary Information section 15). Although the activation barrier for the 2H–1 T phase transition is large, photo-excitation at 455 nm effectively reduces this barrier, leading to fast electron donation and Li-ion intercalation. In WX₂, the initial electron donation reaction is thermodynamically unfavourable and can be initiated only under 455 nm illumination. However, as the energy difference between 1T and decomposition products is very small, continuous illumination pushes the system too far, preventing controlled 1T formation, and ultimately resulting in conversion.

Finally, we sought to improve the sustainability of TMD phase engineering by removing the toxic and pyrophoric n-BuLi from the reaction. By considering the requirements of the reaction on the electrochemical potential, the chemical redox potential and illumination conditions, we developed several safe organo-lithiation agents based on polycyclic aromatic hydrocarbons (PAHs) and lithium metal in tetrahydrofuran solvent. These chemicals furnish a solution of pi-radical anions counterbalanced by Li cations with sufficient driving force for electron donation to TMDs, offering a route towards replacing n-BuLi.

Figure 5a presents the Raman spectra of (multilayer) MoS₂ treated with different PAH-Li agents. Anthracene-Li (AnLi, green, 0.91–1.03 V) and pyrene-Li (purple, 0.73–0.93 V), fall into the acceptable redox potential range between 0.57 V and 1.1 V and, therefore, successfully convert 2H-MoS₂ to its 1T phase, as evidenced by their close spectral match to n-BuLi (blue). By contrast, perylene-Li (grey, 1.19–1.35 V) did not react with the 2H phase because of insufficient redox potential, whereas naphthalene-Li (orange, 0.26–0.46 V) resulted in undesirable decomposition because of a high driving force⁴³.

a, Raman spectra of PAH-Li treated MoS₂ with the chemical structure of PAHs. Blue-shaded region denotes redox-matched chemicals. Anthracene-Li (green) and pyrene-Li (purple) show the 1T Raman signature identical to the n-BuLi-treated sample (blue). b, Optical image of phase transition of thick MoS₂ by anthracene-Li treatment for 600 s in the dark. c, Optical image of phase transition of thick MoS₂ by anthracene-Li treatment for 10 s with 445 nm illumination. Scale bar, 10 μm (b,c).

To explore the effect of changing the redox reagent on the photoredox-driven phase transition dynamics, we investigated exfoliated, thick MoS₂ flakes on glass and added AnLi. Figure 5b shows an ex situ optical image of a thick flake under ‘dark’ conditions after 10 min of reaction time, showing a partial phase transformation. Noticeably, the extracted phase front speed of about 6 nm s^−1 with AnLi in a thick flake was remarkedly faster than with n-BuLi in a bilayer (0.08 nm s^−1), showcasing the effect of the higher driving force provided by AnLi. On above-gap illumination (Fig. 5c; 10 s illumination at 130 mW cm^−2), the reaction is yet again substantially accelerated with an estimated phase front speed of up to 5 μm s^−1, about six orders of magnitude faster than under dark conditions for a bilayer. Taken together, our results show that even thick TMD flakes can be readily transformed by combining the effect of redox potential matching with above-gap illumination. Moreover, the AnLi system also worked for MoSe₂ and WS₂ by accelerating the lithiation reaction (Supplementary Information sections 16 and 17).

In summary, our study provides mechanistic insights into the 2H to 1T phase transition in few-layer TMDs by real-time visualization of the chemical lithiation reaction. We develop a photo-activated reaction pathway that accelerates the phase transition reaction, allowing for rapid phase-pattering of MoS₂. Finally, we use these insights to replace pyrophoric n-BuLi from the reaction with safe organolithiation agents (PAH-Li). These reagents make the process safer and greener and also reduce reaction times by six orders of magnitude compared with conventional n-BuLi treatment.