Judicious Selection of Precursors with Suitable Chemical Valence State for Controlled Growth of Transition Metal Chalcogenides

Transition metal chalcogenides (TMCs) have attracted wide attentions as a class of promising material for both fundamental investigations and electronic applications due to their atomic thin thickness, dangling bond‐free surface, and excellent electronic properties. Specifically, TMCs show outstanding properties such as good thermal conductivity, robust mechanical properties, and extraordinary electronical characteristics, bestowing them utility in both fundamental research and applications. Recently, the development of post‐Moore electronics based on TMCs calls for their large‐size and single‐crystal growth. However, researchers about synthesis usually focus on controlling several growth parameters (such as growth temperature, flow rate, and time). Herein, it is reported that the chemical valence states of transition metal precursors play an important role in controlling the lateral size and crystal quality for TMCs. The study discusses the valence states‐dependent growth mechanism for WS2 and MoS2 from four factors: evaporation temperature, skipping of reaction steps, atomic binding energy of the precursors, and formation energy. In addition, the as‐grown WS2 and MoS2 nanoflakes exhibit good photoelectric response properties. For EuS, the growth results are obviously different by using EuBr3 and EuBr2 as precursors. The studies provide a unique perspective and also new knowledge to controllably grow large‐size and good crystal quality TMCs.


Introduction
3] TMCs nanosheets with a thickness ≈100 Å were reported way back in 1966 by Frindt, pioneering the exfoliation of MoS 2 by adhesive-tape peeling at that time. [4]However, this kind of materials were only regarded as a class of solid lubricant and did not arouse extensive interests.Until 2004 when the star material "graphene" occurred, the 2D material realm saw explosive development, also inducing renaissance of TMCs materials. [5]The interests in TMCs have grown exponentially across various science and engineering disciplines due to their unique atomic structure manifested in a 2D lattice. [6]The chemical bonds between M and X are predominantly covalent in nature, while alternating layers are held together by comparably weaker vdW interactions (MoS 2 : ≈28 meV Å −2 ). [7][20][21][22] Their extraordinary electronic and transport properties (single layer MoS 2 with a mobility of ≈200 cm −2 V −1 s −1 ) [23] are desired for applications in electronic and optoelectronic devices, sensors and electrochemical devices.Moreover, the bandgap of some TMCs changes from an indirect to a direct band gap when thinned down from several layers to a single layer, which offers a choice for ultrathin highperformance optoelectronics. [24] Moreover, the unique valley properties in the energy band structures and the dipolar interaction in plane have triggered research activities from fundamental science (e.g., topological materials) to far-reaching cutting-edge device applications such as quantum computers. [28]he growth of the material lays the foundation for the further applications of the 2D TMCs.Preparation of materials determines the future.The development of post-Moore integrated circuits based on 2D materials calls for their large-size and singlecrystal growth. [29]Chemical vapor deposition (CVD) is considered as an ideal method for growing large-size 2D materials due to its low cost and easiness for handling. [30]However, researchers about synthesis usually focus on controlling several growth parameters (such as growth temperature, flow rate, and time) and typically tuning the size of TMCs can be laborious.The selection of the precursors that nonnegligibly influence the synthesis of 2D TMCs calls for judiciousness, but the concerning knowledge is very lacking.In this article, we report that the transition metal precursors with different chemical valence states have impact on the results of the growth, which shows importance of the selection of the precursors and provides further unique knowledge and guidance on the large-size growth of 2D TMCs.

Results and Discussion
The TMCs samples were grown by home-built atmospheric chemical vapor deposition (APCVD), which is described in the Methods section in details. [31][32][33] It is essential to precisely control the crystal size of WS 2 nanosheets during the growth process. [34]ifferent from tuning conventional growth conditions such as temperature, growth time and so on, [35][36][37] we find that the chemical valence states of precursors play a crucial role in tuning the grain size and crystal quality for TMCs.As a demonstration, first, we focus on the controllable synthesis of WS 2 nanosheets.By varying the valence states of precursors while keeping other conditions (for example, growth temperature) constant, we find that the grain size of WS 2 nanosheets (Figure 1a-i) show a regular evolution with the valence states of precursors.In detail, when the growth temperatures are all at 920 °C, there are no samples on the sapphire substrate if WO 3 or WO 2.9 is used as the precursor (Figure 1a,b).On the contrary, there are some small triangular WS 2 nanoflakes on the substrate if WO 2.72 is used as the precursor (Figure 1c).As the growth temperature increases to 940 °C, all three precursors can generate WS 2 and obviously the grain size of WS 2 nanoflakes increases as the composition of oxygen in the precursor decreases (Figure 1d-f).When the growth temperature increases to 970 °C, the grain size of WS 2 nanoflakes grown by all three precursors further increases.The size increases with the oxygen ingredient decreases (marked by black dashed arrow in Figure 1g-i).Some WS 2 nanoflake grown by WO 2.72 even exceeds 300 μm (Figure S1, Supporting Information).Atomic force microscope (AFM) results (Figure 1j; Figure S2, Supporting Information) confirm that almost all WS 2 nanoflakes grown by three precursors are monolayer.
In order to clearly show the variation of grain size with valence states, we draw the statistical grain size distributions of the WS 2 nanosheets synthesized by WO 3 , WO 2.9 , and WO 2.72 at 940 °C, respectively (Figure 1k).It can be obviously shown that the grain size of WS 2 nanoflakes increases as the composition of oxygen in the precursor decreases, whose mechanism will be discussed in detail in the following part.To characterize the crystal quality of WS 2 nanoflakes, we studied the photoluminescence (PL) and Raman mapping of WS 2 grown by all three precursors.Figure 1l,m,n show the PL intensity mapping for WS 2 nanosheets grown by WO 3 , WO 2.9 , and WO 2.72 , respectively, indicating the improved homogeneity of the crystal quality as the composition of oxygen in the precursor decreases.Figure 1o shows the statistical PL full width at half maximum (FWHM) distributions of thousands of pixel data of the WS 2 nanosheets, which is extracted from their PL FWHM mapping (Figures S3-S5, Supporting Information), respectively.Figure S6 (Supporting Information) shows the corresponding histogram.[40] The mean values of PL FWHMs and their distribution of WS 2 grown by WO 2.9 (19.9 nm) and WO 2.72 (19.8 nm) are narrower than that grown by WO 3 (22.4nm) (Figure 1o, statistical data in Table S1, Supporting Information).In addition, Raman integrated intensity mapping (330-365 cm −1 ) and PL peak position mapping also indicate the improved homogeneity of crystal quality as the composition of oxygen in the precursor decreases (Figures S3-S5, Supporting Information).
To further explore how these precursors influence the grain size and crystal quality of the as-grown WS 2 nanoflakes, we characterize all three WO x (x = 3, 2.9, and 2.72) precursors.Figure 2a,b show the OM images of WO 3 (yellow powder in appearance, known as "yellow tungsten oxide") and WO 2.9 (blue powder in appearance, known as "blue tungsten oxide"), respectively, suggesting their particle-shape with a lateral size of dozens of micrometers.However, scanning electron microscope (SEM) image shows that WO 2.72 (purple powder in appearance, known as "purple tungsten oxide") are uniquely nanowires with a length of hundreds of nanometers (Figure 2c).The morphology imparts larger surface area and enables reactions with sulfur vapor to a greater extent.[43] The increasing FWHMs of XRD patterns for all three WO x precursors with the decreasing composition of oxygen may attribute to the increasing oxygen vacancy, which can also be evidenced by X-ray photoelectron spectroscopy (XPS) in the following discussions.To investigate the composition and valence state of WO x precursors, we perform XPS characterization for them.Figure 2d-f show the W 4f corelevel XPS spectra of WO x precursors.The XPS spectrum of the W 4f consists of two spin-orbit doublets (W 4f 7/2 in 35.8 eV and W 4f 5/2 in 37.9 eV) corresponding to W 6+ in WO 3 (Figure 2d). [44]As a comparison, besides W 6+ , an additional small additional fraction of W 5+ (W 4f 7/2 in 34.6 eV and W 4f 5/2 in 36.6 eV) can be fitted in WO 2.9 (Figure 2e).Additional peaks (≈41.5 eV) in the larger binding energy direction in Figure 2d,e are attributed to W 5p 3/2 .In WO 2.72 (Figure 2f), new W 4f doublets (W 4f 7/2 in 33.3 eV and W 4f 5/2 in 35.7 eV) appear, corresponding to W 4+ .To clearly show the valence state of WO x precursor, we list the fraction of different valence states in Table S2 (Supporting Information) according to peak area of each fitted peak.WO 3 contains 100% W 6+ and WO 2.9 contains 9.3% W 5+ and 90.7% W 6+ .WO 2.72 contains 45.6% W 4+ , 24.7% W 5+ , and 29.7% W 6+ . According to pervious related reports, [47][48][49] the conversion from WO x to WS 2 is a gradual reduction reaction of losing oxygen, which can be expressed as the following two equations: Thus, the smaller x in WO x may favor to skip the additional reduction reaction of losing oxygen (Equation 1), in favor of growing high crystal quality WS 2 .Figure 2g-i show the XPS spectra of O 1s orbit for WO x precursors.[52] According to Figure 2g-i and Table S3 (Supporting Information), we find that as the composition of oxygen in the precursor decreases, oxygen defects increases (22.9% for WO 3 , 30.4% for WO 2.9 , and 41.7% for WO 2.72 ).
The WO x precursors are further characterized by Raman spectroscopy, which is sensitive on analyzing their structural phase.As shown in Figure 2j, compared with WO 3 , all Raman vibration peaks of WO 2.9 and WO 2.72 are obviously broader.Smeared and broader Raman vibration peaks can be attributed to the related c) SEM image of WO 2.72 powder.The scale bar is 250 nm.d,f) W 4f core-level XPS spectra of d) WO 3 , e) WO 2.9 , and f) WO 2.72 powder, respectively.O 1s orbit XPS spectra of g) WO 3 , h) WO 2.9 , and i) WO 2.72 powder, respectively.j) Raman spectra of WO 3 (black line), WO 2.9 (red line) and WO 2.72 (blue line) powder, respectively.k) TG curves of WO 3 (black line), WO 2.9 (red line) and WO 2.72 (blue line) powder, respectively.l) Calculated atomic binding energy of WO 3 (black line), WO 2.9 (red line) and WO 2.72 (blue line), respectively.The insets are their crystal structure model.bonds induced by oxygen deficiency. [42]Together, some vibration peaks exhibit a small red shift (Raman softening) as the composition of oxygen in the precursor decreases, indicating the oxidation states of W from high valence state to low valence state, in which case fewer chemical bonding coordinated to the W atom leads to smaller stiffness of the bonds and then vibrations at lower energies. [42]Moreover, in order to clearly clarify the effect of evaporation temperature (T e ) for all three WO x precursors on the growth of WS 2 , we perform thermogravimetric (TG) analysis on them to quantify the weight of WO x powder.After gradually heating up these precursors to high temperature in inert gas atmosphere, the remaining weights were accurately recorded.After normalizing the initial weight, we plot their TG curves as shown in Figure 2k.We find that as the composition of oxygen in the precursor decreases, the T e is lower (from 1271 °C for WO 3 to 802 °C for WO 2.72 ).The thin nanowire morphology of WO 2.72 can par-tially lead to its lower evaporation temperature. [53]The difficulty for the evaporation of the tungsten oxide precursors decreases in the order of WO 3 , WO 2.9 , and WO 2.72 .The typical appropriate temperature zone of CVD in quartz tube furnace is 500-1000 °C.The evaporation temperature for the precursor of tungsten series is relatively high and the vapor pressure of tungsten oxide is pretty low (9.67 × 10 −7 atm. at 1314 K for WO 3 , not applicable for WO 2.9 and WO 2.72 ). [54]Easier evaporated tungsten precursors contribute to the size-increasing tendency.Note the statistical densities of nucleation sites of these precursors are in Figure S8 (Supporting Information), increasing in order but not acutely.
Beside experimental characterizations, we further calculate their atomic binding energy as shown in Figure 2l.The calculation of atomic binding energy is described in the Methods.The much lower atomic binding energy for WO 2.72 (−5.2 eV) and slight lower atomic binding energy for WO 2.9 (−5.7 eV) implies that formation of W─O bond releases lower energy and simultaneously breaking of W─O bond requires lower energy, comparatively to counterpart WO 3 (−5.8eV), favoring the conversion from WO 2.72 and WO 2.9 to WS 2 .In addition, we compare formation energy of W─O─S complex for two conditions (Figure S9, Supporting Information): i) direct filling the WO 3 lattice with interstitial sulfur (≈3.95 eV); ii) replacing oxygen atoms with sulfur (≈3.58 eV).It suggests that oxygen vacancies are beneficial for assisting the sulfur atoms to enter into the tungsten oxide lattice to bind with tungsten.In conclusion, we can summarize the following factors for mechanisms: i) for WO 2.72 , nanoscale morphology can effectively lower T e , assisting the evaporation of the precursors and benefiting for growing large-size WS 2 ; ii) the increasing proportion of low valence state with the decreasing composition of oxygen in the precursor skips the additional reduction reaction, in favor of growing high crystal quality WS 2 ; iii) the lower atomic binding energy of WO 2.72 also favors the conversion from WO x to WS 2 ; iv) the existence of oxygen vacancies in tungsten oxide lower the formation energy from the complex of sulfur and tungsten oxide, making it easier for the sulfur atoms to enter into the tungsten oxide lattice to bind with tungsten.Based on the above analysis, Interestingly, we can grow WS 2 continuous membrane by WO 2.72 when we raise the growth temperature to 980 °C and elongate the time (Figure S10, Supporting Information).
As another typical transition metal system, we synthesize MoS 2 nanoflakes by MoO 3 and MoO 2 (both of them are white powder in appearance) to demonstrate that the valence states of precursors play a crucial role in tuning the growth for TMCs.By keeping other conditions (for example, growth temperature) constant, we find that the grain size of MoS 2 nanosheets (Figure 3a,b) is dependent on the valence states of precursors.As shown in Figure 3a,b, MoS 2 nanoflakes grown by MoO 2 yield an obvi-ously larger lateral size compared with those grown by MoO 3 .In order to clearly show the variation of grain size with valence states, we draw the statistical grain size distributions of the MoS 2 nanosheets synthesized by MoO 3 and MoO 2 , respectively (Figure 3c).It can be obviously shown that the grain size of MoS 2 nanoflakes increases as the composition of oxygen in the precursor decreases.Some nanoflakes grown by MoO 2 can even reach 120 μm (as shown by black dashed arrow in Figure 3b).AFM images and height curves (Figure S11, Supporting Information) show that MoS 2 nanosheets grown by both MoO 3 and MoO 2 are almost monolayers.PL and Raman mapping are used to characterize the crystal quality of MoS 2 nanoflakes grown by these two precursors.Figure 3d,e show the PL intensity mapping for MoS 2 nanosheets grown by MoO 3 and MoO 2 , respectively.From the PL intensity mapping, we can find that the homogeneity of the crystal quality of MoS 2 nanoflakes grown by MoO 3 (Figure 3d) are worse compared with those grown by MoO 2 (Figure 3e).By extracting numerous data of PL FWHMs from their corresponding PL mappings (Figures S12 and S13, Supporting Information), respectively, we can plot the statistical PL FWHM distributions of the MoS 2 nanosheets, as shown in Figure 3f and Figure S14 (Supporting Information).The average PL FWHM of MoS 2 grown by MoO 2 (37.8 nm) is obviously narrower than that grown by MoO 3 (54.0nm) (Table S4, Supporting Information).In addition, Raman integrated intensity mapping (395-410 cm −1 ) and PL peak position mapping also suggest that the homogeneity of the crystal quality of MoS 2 nanoflake grown by MoO 2 are improved compared with that grown by MoO 3 (Figures S12 and S13, Supporting Information).
To further explore how MoO 3 and MoO 2 tune the grain size and crystal quality of the as-grown MoS 2 nanoflakes, we characterize MoO 3 and MoO 2 precursors.images of MoO 3 and MoO 2 powder, respectively, suggesting their particle-shape with a lateral size of dozens of micrometers and the particle size of MoO 3 is slightly (≈20-30 μm) larger than that of MoO 2 . The larger FWHM of XRD patterns for MoO 2 precursor may attribute to the increased oxygen vacancy, which can be evidenced by XPS in the following discussions.To investigate the composition and valence state of MoO x (x = 3 and 2) precursor, we perform XPS characterization for them.Figure 4c,d show the Mo 3d orbit XPS spectra of MoO x precursors.The XPS spectrum of the Mo 3d consists of two spin-orbit doublets (Mo 3d 5/2 in 232.5 eV and Mo 3d 3/2 in 235.6 eV) corresponding to Mo 6+ in MoO 3 (Figure 4c). [57]However, a majority fraction of Mo 4+ (Mo 3d 5/2 in 230.0 eV and Mo 3d 3/2 in 233.2 eV) can be fitted in MoO 2 (Figure 4d).Besides Mo 4+ , there is a small fraction of Mo 5+ (Mo 3d 5/2 in 231.6 eV and Mo 3d 3/2 in 235.5 eV) in MoO 2 , which is consistent with previous reports. [58]The Mo 5+ can be attributed to the oxidation process when MoO 2 is stored in ambient condition. [59]o clearly show the valence state of MoO x precursors, we list the fraction of different valence states in Table S5 (Supporting Information) according to peak area of each fitted peak.MoO 3 contains 100% Mo 6+ and MoO 2 contains 57.8% Mo 4+ and 42.2% Mo 5+ .
According to previous report, [60] during a stepwise sulfurization of MoO 3 to the final product of MoS 2 , intermediate products of MoO 2 and MoOS 2 can form, which can be described by three intermediate reactions: Thus, the lower valence state in MoO 2 makes the conversion skip the additional reduction reaction (Equation 3), in favor of growing high crystal quality MoS 2 .Figure 4e,f show the XPS spectra of O 1s orbit for MoO x precursors.In all MoO x precursors, there are two peaks (≈530 and ≈531 eV) in XPS spectra of O 1s orbit, corresponding to lattice oxygen and oxygen defects, respectively. [50]According to Figure 4e,f and Table S6 (Supporting Information), we find that the oxygen defects in MoO 2 (44.8%) yield higher proportion compared with MoO 3 (27.3%).
The MoO x precursors are further characterized by Raman spectroscopy to analyze their structural phase.As shown in Figure 4g, all Raman vibration peaks of MoO 2 are obviously broader compared with MoO 3 , which can be attributed to the related bonds induced by oxygen deficiency. [42]Together, some vibration peaks in MoO 2 exhibit a small red shift and softened compared with those in MoO 3 , indicating the oxidation states of Mo from high valence state to low valence state, in which smaller stiffness of the bonding by the reduced number of bonds and then vibrations at lower energies. [42]In order to clearly clarify the effect of evaporation temperature T e for MoO x precursors on the growth of MoS 2 , we perform TG analysis on them to quantify the residual weight of MoO x powder as the temperature increases, as shown in Figure 4h.Interestingly, we find that the T e of MoO 2 (865 °C) is higher compared with that of MoO 3 (803 °C).However, the typical appropriate temperature zone of CVD in quartz tube furnace is 500-1000 °C.The evaporation temperatures of MoO 2 and MoO 3 fall just right in this appropriate temperature zone.Furthermore, as we know, the vapor pressure of MoO 3 (4045 × 10 −7 atm. at 958 K) is very high compared with that of MoO 2 (883.3 × 10 −7 atm. at 959 K), [61] leading to numerous and dense nucleation sites which makes the size of the grown grains small (Figure S16, Supporting Information). [62]The judicious selection of the slightly higher evaporation temperature and lower vapor pressure of MoO 2 , on the contrary, make it more suitable for fewer nucleation sites and larger flake size (Figure S16, Supporting Information). [63]We further calculate their atomic binding energy as shown in Figure 4i.The lower absolute value of negative atomic binding energy for MoO 2 implies that breaking of Mo─O bond in MoO 2 requires lower energy.In addition, we compare formation energy of inserting sulfur atoms to MoO 3 lattice for two conditions: i) direct inserting interstitial sulfur atoms (≈3.04 eV); ii) filling oxygen vacancies with sulfur (≈2.40 eV), as shown in Figure S17 (Supporting Information).It suggests that oxygen vacancies are beneficial for sulfur atoms entering the lattice of molybdenum oxide to bind with molybdenum.In conclusion, we can summarize the following considerations: i) Taking the factors into account that the temperature for growing TMCs of Mo series adequately falls in the typical CVD temperature zone and the vapor pressure is relatively quite high for the molybdenum oxides, the higher T e in MoO 2 can reduce the number of nucleation sites than that by MoO 3 , benefiting for growing large size MoS 2 ; ii) the lower valence state in MoO 2 makes the reaction skip the additional reduction reaction, in favor of growing high crystal quality MoS 2 ; iii) the lower atomic binding energy in MoO 3 also favors the conversion from MoO x to MoS 2. iv) the oxygen vacancies in molybdenum oxide lower the formation energy to more negative for the sulfur atom to enter the lattice of molybdenum oxide to bind with molybdenum.Interestingly, based on the above analysis, it should be noted that we can grow MoS 2 continuous membrane with lateral size above 1 cm by MoO 2 at 950 °C (Figure S18, Supporting Information).
WS 2 and MoS 2 with properties such as direct band gap transition in low dimensional structures, strong light-matter interaction and good carrier mobility, combined with the feasible and low-cost growth, have triggered extensive interests for this material in the field of optoelectronics. [64]To further characterize the crystal quality of our as-grown WS 2 and MoS 2 nanoflakes, we study their photo-electronical properties.Figure 5a,b show the dark and photocurrent sweep when incident light power intensity increases from zero to 81.1 mW cm −2 for WS 2 and MoS 2 nanoflakes, respectively.Based on the above current sweep, we can obtain responsivity (R) and detectivity (D * ) of WS 2 (Figure 5c) and MoS 2 (Figure 5d) under various light power intensity with R = I ph /P in and D * = A 1/2 R/(2qI d ) 1/2 , respectively, where I ph , P in , A, q and I d are photocurrent, illumination power, the effective area of the photodetector, element charge and dark current. [65]he maximum R and D * of WS 2 are 12.9 mA W −1 and 7.0 × 10 10 Jones under 650 nm light illumination with voltage of 0.5 V.The maximum R and D * of MoS 2 are 20.8 mA W −1 and 5.9 × 10 11 Jones.Further, we can obtain their 1/f noise by S I (f) =  H I 2 /(fN), where  H , I, f, and N are Hooge parameter, current, frequency and the number of carriers engaging in the conduction process (Figure 5e,f). [65]The 1/f noise of WS 2 and MoS 2 under 0.6 V bias are ≈10 −23 A 2 Hz −1 and 10 −22 A 2 Hz −1 magnitude, respectively.68][69][70][71][72] 2D non-layered EuS, is an emerging TMCs material, which presents distinctive magnetic and optical properties and exhibits great potential in spintronic devices. [73]However, the challenge in their controllable synthesis limits their experimental investigations of monocrystalline 2D EuS nanoflakes.We also demonstrate that the valence state of precursors (EuBr 3 and EuBr 2 ) can play an important role in controlling the growth of EuS nanoflakes.Figure 6a,b show the OM images of EuS nanoflakes grown by EuBr 2 and EuBr 3 , respectively, under the same other growth conditions.It indicates that EuS nanoflakes grown by EuBr 2 yield a higher coverage and larger lateral size, compared with those grown by EuBr 3 (Figure 6c). Figure 6d,e show Raman and PL spectra for EuS nanoflakes grown by EuBr 2 , respectively, which is consistent with a previous report. [73]In Raman spectrum, a fundamental vibrational mode at ≈237 cm −1 and an overtone sequence at ≈494 cm −1 can be observed, which only contains longitudinal optical phonon modes.In PL spectrum, a peak is located at 743 nm, corresponding to a band gap of ≈1.67 eV.According to the database, the melting point of EuBr 2 and EuBr 3 are 683 and 702 °C respectively. [74,75]The lower melting point for EuBr 2 may help to grow larger size EuS.We calculate the atomic binding energy for EuBr 3 and EuBr 2 , as shown in Figure 6f.The lower atomic binding energy for EuBr 2 implies that breaking of Eu─Br bond in EuBr 2 requires lower energy, consistent with the data of melting points.It is in favor of growing larger coverage and larger lateral size EuS.This case also proves that the judicious selection of the valence state of the precursors is beneficial for the improvement in the 2D material growth.

Conclusion
We demonstrated that the chemical valence states of precursors play a crucial and unique role in tuning the grain size and crystal quality for TMCs.For WS 2 , the lateral size and crystal quality are improved as the composition of oxygen in the precursor decreases.The MoS 2 grown by MoO 2 yields lager lateral size and better crystal quality compared with that grown by MoO 3 .We discuss the judicious selection of valence states of precursors for WS 2 and MoS 2 based on mainly four factors: evaporation temperature (T e ), skipping of reaction steps, atomic binding energy of the precursors, and formation energy.For EuS, EuBr 2 favors to grow larger lateral size and higher quality EuS compared with EuBr 3 .The as-grown WS 2 and MoS 2 exhibit good photoelectric device properties, which make them excellent alternatives for photodetectors, showing the validity of the approach of judicious selection of the valence state in precursors.

Experimental Section
CVD Growth of WS 2 , MoS 2 , and EuS: These three TMCs materials were grown by home-built APCVD system (Figure S19, Supporting Information).Typically, S powder (0.2 g) was placed in the upstream lowtemperature zone.The transition metal precursors (WO x , MoO x , or EuBr x ) and NaCl were put on a substrate (SiO 2 /Si, sapphire or mica, Shanghai Onway Technology Co. Ltd.) in the heating center of the downstream hightemperature zone.Before the growth, the quartz tube was purged with Ar gas flow for 10 min.The upstream zone was heated to 300 °C in 35 min with 180 sccm Ar.For WS 2 nanoflakes, the downstream zone was heated to 920, 940, and 970 °C as needed.For WS 2 continuous membrane, the downstream zone was heated to 980 °C and WO 2.72 was used as precursor.For MoS 2 nanoflakes, the downstream zone was heated to 800 °C.For MoS 2 continuous membrane, the downstream zone was heated to 950 °C and MoO 2 was used as precursor.For EuS, the downstream zone was heated to 1000 °C and the growth time was 5 min.The S powder was pushed into the upstream zone by a magnet when the temperature of downstream zone reaches the set value.For WS 2 and MoS 2 , after reaction for 10 min, the furnace was cooled naturally down to room temperature.
Material Characterizations: AFM experiments were performed in tapping mode under ambient conditions (Bruker, Dimension XR).Raman and PL mapping were characterized by Horiba Lab RAM Odyssey.XRD patterns were measured by Bruker APEXII.TG curves were collected by NETZSCH-STA449F3.Devices were fabricated by DWL system (Heidelberg Instruments DWL-66) to create electrode patterns and subsequently evaporate Cr/Au electrodes.Photocurrent test was conducted by probe station (METATEST Corp. E2) equipped with Keithley 2636B.The 1/f noise was recorded by a noise measurement system (PDA NC300L, 100 kHz bandwidth).
Density Functional Theory (DFT) Calculation: The geometrical optimizations were conducted by DFT with Perdew-Burke-Ernzerhof generalized gradient approximation (PBE-GGA) functional with projector augmented wave (PAW) potentials, realized in Vienna Ab-initio Simulation Package (VASP).The kinetic cutoff energy was selected as 450 eV for planewave basis set.The valence electron configurations for Mo (4p 6 5s 1 4d 5 ), W (5p 6 6s 2 5d 4 ), Eu (4f 7 6s 2 ), and O (2s 2 2p 4 ) were employed.The first Brillouin zone is characterized by a Γ-point-centered Monkhorst-Pack k-mesh with a grid configuration of 6 × 6 × 4. The tolerance for energy convergence was set at 1.0 × 10 −4 eV, for both structural optimizations and self-consistent-field (SCF) iteration.The criteria for force components convergence was set at −0.02 eV Å with symmetrization of the charge density used.For the atomic binding energy, calculation is taken as the followed example: W+ n 2 O 2 → WO n .Atomic binding energy of WO n is defined as . The rest ones can be done in the same manner.For the formation energy of a sulfur atom inserted into the interstitial positions or the oxygen vacancy positions, the calculation is done as the following: WO 3 +S → WO 3 S. Formation energy can be defined as E(WO 3 S)− E(WO 3 )− E(S).The rest ones can be done in the same manner.

Figure 1 .
Figure 1.Growth and characterization of WS 2 .OM images of WS 2 nanosheets grown by a,d,g) WO 3 , b,e,h) WO 2.9 , and c,f,i) WO 2.72 at a,b,c) 920 °C, d,e,f) 940 °C, and g,h,i) 970 °C, respectively.The scale bars for (a-i) are 60 μm.j) AFM images and height curves of WS 2 grown by WO 3 .k) The statistical grain size distributions of the WS 2 nanosheets synthesized by WO 3 , WO 2.9 , and WO 2.72 at 940 °C, respectively.l-n) PL intensity mapping for WS 2 nanosheets grown by l) WO 3 , m) WO 2.9 , and n) WO 2.72 , respectively.o) The statistical PL FWHM mean value (the middle line in the box) and standard deviation (error bars) of the WS 2 nanosheets synthesized by WO 3 (black box), WO 2.9 (red box), and WO 2.72 (blue box), respectively.The bottom and top horizontal lines in the box stand for the 25 and 75 percentiles.

Figure 2 .
Figure 2. Characterization of WO x precursors.OM images of a) WO 3 and b) WO 2.9 powder, respectively.The scale bars in (a) and (b) are 60 μm.c)SEM image of WO 2.72 powder.The scale bar is 250 nm.d,f) W 4f core-level XPS spectra of d) WO 3 , e) WO 2.9 , and f) WO 2.72 powder, respectively.O 1s orbit XPS spectra of g) WO 3 , h) WO 2.9 , and i) WO 2.72 powder, respectively.j) Raman spectra of WO 3 (black line), WO 2.9 (red line) and WO 2.72 (blue line) powder, respectively.k) TG curves of WO 3 (black line), WO 2.9 (red line) and WO 2.72 (blue line) powder, respectively.l) Calculated atomic binding energy of WO 3 (black line), WO 2.9 (red line) and WO 2.72 (blue line), respectively.The insets are their crystal structure model.

Figure 3 .
Figure 3. Growth and characterization of MoS 2 .OM images of MoS 2 nanosheets grown by a) MoO 3 and b) MoO 2 under the same condition, respectively.The scale bar in (a,b) is 60 and 100 μm, respectively.c) The statistical grain size distributions of the MoS 2 nanosheets synthesized by MoO 3 (black) and MoO 2 (red), respectively.PL intensity mapping for MoS 2 nanosheets grown by d) MoO 3 and e) MoO 2 , respectively.f) The statistical mean value (the middle horizontal line in the box) and standard deviation (error bars) of PL FWHM distributions of the MoS 2 nanosheets synthesized by MoO 3 (black) and MoO 2 (red), respectively.The bottom and top horizontal lines in the box stand for the 25 and 75 percentiles.
Figure 4a,b show the OM

Figure 4 .
Figure 4. Characterization of MoO x precursors.OM images of a) MoO 3 and b) MoO 2 powder, respectively.The scale bars in (a,b) are 100 μm.Mo 3d orbit XPS spectra of c) MoO 3 and d) MoO 2 , respectively.O 1s orbit XPS spectra of e) MoO 3 and f) MoO 2 powder, respectively.g) Raman spectra of MoO 3 (black line) and MoO 2 (red line) powder, respectively.h) TG curves of MoO 3 (black line) and MoO 2 (red line) powder, respectively.i) Calculated atomic binding energy of MoO 3 (black line) and MoO 2 (red line), respectively.The insets are their corresponding crystal structure model.

Figure 5 .
Figure 5. Photocurrent characterizations for WS 2 and MoS 2 nanoflakes.a) Dark and photocurrent, b) responsivity and detectivity, and c) 1/f noise for a typical WS 2 nanoflake.d) Dark and photocurrent, e) responsivity and detectivity, and f) 1/f noise for a typical MoS 2 nanoflake.The insets in (a,d) are OM images of WS 2 and MoS 2 .The scale bars are 20 μm.

Figure 6 .
Figure 6.Growth and characterization of EuS nanoflakes.OM images of EuS nanoflakes grown by a) EuBr 2 and b) EuBr 3 under the same condition, respectively.The scale bars in (a,b) are 10 μm.c) The statistical grain size distributions of the EuS nanosheets synthesized by EuBr 3 (black) and EuBr 2 (red), respectively.d) Raman and e) PL spectra for EuS nanoflakes grown by EuBr 3 , respectively.f) Calculated atomic binding energy of EuBr 3 (black line) and EuBr 2 (red line), respectively.The insets are their corresponding crystal structure model.