Topochemical Synthesis of Two‐Dimensional Transition‐Metal Phosphides Using Phosphorene Templates

Abstract Transition‐metal phosphides (TMPs) have emerged as a fascinating class of narrow‐gap semiconductors and electrocatalysts. However, they are intrinsic nonlayered materials that cannot be delaminated into two‐dimensional (2D) sheets. Here, we demonstrate a general bottom‐up topochemical strategy to synthesize a series of 2D TMPs (e.g. Co2P, Ni12P5, and CoxFe2−xP) by using phosphorene sheets as the phosphorus precursors and 2D templates. Notably, 2D Co2P is a p‐type semiconductor, with a hole mobility of 20.8 cm2 V−1 s−1 at 300 K in field‐effect transistors. It also behaves as a promising electrocatalyst for the oxygen evolution reaction (OER), thanks to the charge‐transport modulation and improved surface exposure. In particular, iron‐doped Co2P (i.e. Co1.5Fe0.5P) delivers a low overpotential of only 278 mV at a current density of 10 mA cm−2 that outperforms the commercial Ir/C benchmark (304 mV).


Introduction
Tr ansition-metal phosphides (TMPs) display many interesting chemical and physical features,including superconductivity, [1] thermoelectric properties, [2] luminescence, [3] and magnetism, [4] derived from the unique role of the phosphorus atoms.T he electronegativity of phosphorus is generally higher than those of metals,t herefore,m ost TMPs are insulators or semiconductors,b ecause their electron delocalization is strongly restricted. [5] Recently,T MPs were rediscovered as "star materials" in catalysis and energy harnessing.F or example,c obalt-and iron-based phosphides demonstrate high catalytic activity and long-term stability for electrocatalytic water splitting in both acidic and basic solutions and are superior to many traditional catalysts based on expensive noble metals. [6] From af undamental point of view,t he electrocatalytic behavior of TMPs is dominated by their interfacial chemistry as well as electronic structures, which govern, respectively,t he formation of active surface sites and charge transport across the interfaces during the electrochemical reactions. [7] Although the chemical compositions and crystal structures of TMPs are well known, experimental access to their electronic properties remains in an ascent stage,b ecause TMPs are inherently nonlayered materials that cannot be delaminated into thin layers by topdown exfoliation methods. [8] Consequently,bottom-up synthesis has become the mainstream approach to prepare multidimensional TMPs.Recent progress in this direction highlights the key impact of phosphorus (P) sources.O rganophosphorus precursors,s uch as tri-n-octylphosphine (TOP) [9] and other alkyl-and arylphosphines, [5] are versatile Ps ources in wet-chemical synthetic procedures.T hese result in ag reat number of monodispersed TMP nanoparticles at relatively low temperatures (220-385 8 8C) through anucleation and growth mechanism. [10] Alternatively,i norganic Ps ources,c onsisting of phosphates (e.g. (NH 4 ) 2 HPO 4 ), [11] hypophosphines (e.g.N aH 2 PO 2 ), [12] tris(trimethylsilyl)phosphine (i.e.P(SiMe 3 ) 3 ), [13] and red phosphorus, [14] are widely used in solid-solid or gas-solid reactions. Nonetheless,these precursors require high-temperature thermal annealing (e.g.650 8 8C) to decompose or to transform into reactive intermediate species (e.g.P H 3 ,w hite phosphorus gas), [15] thus providing poor control over the shapes of the TMPs,w hich range from nanoparticles, [16] bulk pieces, [11a] films, [11b,13] and nanorods [12b] to nanowire arrays. [12a] Although salt templating is able to guide the growth of TMP nanoplatelets, [17] subsequent hydrogen (H 2 )annealing results in reformation of nanoscale clusters.T od ate,i tr emains ag reat challenge to synthesize 2D TMPs,e specially,f rom conventional molecular precursors.
Herein, we demonstrate anovel topochemical strategy to prepare 2D TMPs in solution by using 2D phosphorene sheets as the Ps ource and sacrificing templates.T he intrinsic high chemical reactivity of phosphorene offsets the need for aggressive conditions [18] and facilitates fast phosphidation of dispersed metal ions.The chemical structures of TMPs can be tailored by varying the feed species and ratios of metal salts, thereby resulting in 2D Co 2 P, Ni 12 P 5 ,a nd Co x Fe 2Àx Pn anosheets with mean lengths of 2.5 mm 2 and thicknesses of 3.6 nm. Importantly,t he 2D TMPs provide the feasibility to explore their charge-transport properties as field-effect transistors. Notably,t he fabricated 2D Co 2 Ps hows p-type semiconducting properties with ah ole mobility of 20.8 cm 2 V À1 s À1 at 300 K. When serving as electrocatalysts,C o 1.5 Fe 0.5 Ps heets exhibit alow overpotential of 278 mV at acurrent density of 10 mA cm À2 and good stability (10 h) for the oxygen evolution reaction (OER), superior to the catalytic performance of commercial Ir/C.Our method opens up anew avenue for the design and construction of nonlayered 2D materials.

Results and Discussion
Given that most of the TMPs are covalent compounds,in which the electron density easily shifts from the metal center (M) to the phosphorus atoms (P), the synthesis of TMPs does not require specific oxidation states of metal ions or phosphorus precursors. [5] However,t he formation of metalmetalloid (M-P) bonds needs highly active precursors or harsh reaction conditions. [19] In this respect, black phosphorus (BP) represents an ideal Psource to realize the 2D structures of TMPs,because of its high reactivity arising from abundant lone pairs of electrons. [18] Defect-free phosphorene was prepared by electrochemical exfoliation of layered black phosphorus crystals by our previously reported method ( Figure 1a). [20] Compared with other scalable exfoliation methods such as tip sonication [21] and shear mixing, [22] this mild intercalation strategy is able to maintain the structural integrity of phosphorene flakes.T he metal sources consisted of several soluble coordination compounds,s panning from ammonium molybdate ((NH 4 ) 2 MoO 4 ), vanadyl acetylacetonate (VO(acac) 2 ), ferrocene (Fe(C 5 H 5 ) 2 ), iron(III) acetyl-acetonate (Fe(acac) 3 ), cobalt(II) acetylacetonate (Co(acac) 2 ) and nickel(II) acetylacetonate (Ni(acac) 2 ;s ee Scheme S1 in the Supporting Information). As olvothermal reaction was carried out between the mixed dispersion of exfoliated BP sheets,transition-metal salt, and anhydrous N,N-dimethylformamide (DMF) in an autoclave.W efound that the Mo VI ,V IV , and Fe II salts failed to react with phosphorene,mostly because of competitive side reactions with the solvent or lattice mismatch with the target TMPs. [17] TheFe III salt mainly led to amorphous 2D Fe 2 Psheets ( Figure S1), while the Co II and Ni II salts gave rise to 2D Co 2 Pa nd Ni 12 P 5 crystals,r espectively ( Figure 1b). Note that, 2D Ni 2 Pc annot be produced by this method as it is thermodynamically unstable. [23] Dilute colloidal dispersions of the metal phosphide sheets (0.05 mg mL À1 ) show ac lear Ty ndall effect ( Figure S2) and are stable for at least four weeks without sedimentation. In addition to the metal sources,the reaction temperature was also crucial. For example,attemperatures higher than 200 8 8C, the majority of the samples were metal oxide nanoparticles,w hich was ascribed to the strong reducing ability of DMF at high temperature. [24] However,a t1 50 8 8C, TMPs were not formed even after 24 h. Therefore,amoderate temperature of 180 8 8C was selected as being optimal in terms of the resulting structures and reaction kinetics.
Themorphology of the 2D sheets was examined by atomic force microscopy (AFM). As depicted in Figure 1c,exfoliated BP flakes present mean diameters of (2.8 AE 1.5) mma nd thicknesses of (3.2 AE 1.1) nm. This polydispersed size/thickness distribution is associated with af ragmentation-exfoliation mechanism, [25] well-known for top-down exfoliation methods.B ased on energy-dispersive X-ray spectroscopy (EDX;F igures S3 and S4), the cobalt and nickel phosphides comprise atomic ratios of 2.1:1.0 (Co/P) and 2.3:1.0 (Ni/P), almost identical to their theoretical stoichiometry (i.e.2.0:1.0 and 2.4:1.0, respectively). AFM images of both Co 2 Pa nd Ni 12 P 5 (Figure 1d,e) display clean surfaces and well-defined edges with height profiles of 2-3 nm. Thetransformation from phosphorene to metal phosphides did not cause apparent variation in the dimensions and thickness.F or example,Co 2 P and Ni 12 P 5 both had an average lateral size of 2.5 mma nd thickness of 3.6 nm, close to the parameters of the parent phosphorene flakes,based on astatistical calculation of more than 80 flakes (Figure 1f,g). By contrast, the crystal structures of the resulting TMP sheets were distinctly different. The starting phosphorene shows two strong characteristic (040) and (060) peaks at 34.28 8 and 52.48 8,r espectively,i nt he X-ray diffraction (XRD) spectrum ( Figure 1h). However, neither peak was detectable after phosphidation reactions,t hus indicating the broken ordering in the out-of-plane direction. Consequently,the XRD patterns of the products confirm the formation of orthorhombic Co 2 P( JCPDS No.3 2-0306) and tetragonal Ni 12 P 5 (JCPDS No.7 4-1381). Co 2 Pc ontains atetrahedral and apyramidal Co center,surrounded by four and five Pa toms,r espectively,w hile the Pc enters in Ni 12 P 5 share 10 or 8N ia toms ( Figure S5). Although the final products do not contain any phosphorene,t he total yield is less than 100 %, because of the possible decomposition of phosphorene during the solvothermal process. c-e) AFM images of phosphorene, Co 2 P, and Ni 12 P 5 ,r espectively. f, g) Statisticala nalysis of Co 2 Pand Ni 12 P 5 flakes in terms of their dimensionsa nd thickness. h) XRD patterns of thin films prepared by filtration of the dispersed 2D sheets.
Co 2 Pw as selected as am odel material to track the structural evolution from phosphorene to 2D TMPs,a nd samples at different stages of the reactions were examined by high-resolution transmission electron microscopy (HR-TEM; Figure 2). Initially,phosphorene has the typical orthorhombic structure.T he lattice fingers of 0.43 nm and 0.33 nm are in accordance with the respective [001] and [100] directions of the intact crystal (Figure 2a,d). [26] Each phosphorus atom connects to three adjacent phosphorus atoms through sp 3 hybridization, thereby leading to two free electrons. [18] As shown in Figure 2b,e,lattice defects started to show up on the surface within 20 min and some phosphorus atoms were missing, while the intrinsic (111) facet of phosphorene remained on the main skeleton. Although al ow density of defects did not cause ac ollapse of the 2D morphology,t his intermediate geometry showed ac lear distortion upon electron beam irradiation during the TEM measurement. As the solvothermal reaction proceeded, an ew lattice orientation appeared on the phosphorene interfaces after 60 min, which corresponds to the incubation of Co x P(0< x < 2; Figure 2c,f). Thes pacing fingers of 0.35 nm and 0.25 nm were identified as the (001) and (210) facets of Co 2 P. Besides the Co 2 Pl attice,t he (111) facet of phosphorene was still pronounced. However,itvanished in the final Co 2 Pproduct. This observation suggests that the synthesis of Co 2 Pf ollows ac onversion-growth mechanism, in which cobalt atoms gradually react with phosphorus atoms and are incorporated into the 2D matrix.
To verify the underlying mechanisms,v arious combinations of precursors were tested in the same solvothermal environment. As an example,adispersion of phosphorene in DMF became clear after 5h reaction. Thes tructural degradation of phosphorene indeed occurred, even though it was maintained in an oxygen-free anhydrous system. In contrast, the solutions of metal salts in DMF resulted in amorphous solid or metal nitride ( Figure S6). Thus,the interplay between the phosphorene,m etal salts,a nd the solvent is essential to grow 2D TMPs.D espite DMF being ap olar aprotic solvent, under suitable circumstances the DMF molecule loses ah ydrogen atom and becomes aD MF radical, for example, by nucleophilic attack from as trong Lewis base (i.e. potassium methylate). [27] Similarly,p hosphorene can donate its unpaired electrons to DMF molecules and remove hydrogen atoms from their methyl groups.T herefore,i ti sr ational to propose the following growth mechanisms (Figure 2g): At elevated temperature (e.g.1 80 8 8C), the surrounding DMF molecules react at the edges and basal planes of phosphorene (denoted as BP), thereby causing phosphorus vacancies in the superlattice (i.e.B P nVac ). Consequently,P atoms are detached in the form of PH 3 .S imilar pathways towards ad efective BP structure have been observed in the presence of water molecules at room temperature [Eq. (1)]. [28] ThePatoms neighboring the vacancies continue to react with DMF molecules according to Equation (2). Afterwards,the growth of the 2D TMPs involves two steps: [29] First, reduction of the metal ions to the metal (M) at hydrogenated Pv acancies [Eq. (3)] followed by phosphidation of the metal clusters at the phosphorene interfaces [Eq. (4) 2 n Co þ BP-H 3nÀ2 ! n Co 2 P þ 3n À 2 2 H 2 ð4Þ Despite the exact details of the phase transformation remaining elusive,t he repetition of reactions (1)-(4) and subsequent structural rearrangement are responsible for transforming the entire phosphorene flakes into 2D metal phosphides.
Without metal ions in this system, the increasing density of Pvacancies (BP nVa c and BP-H 3n )was not able to support the 2D shape of phosphorene,w hich eventually broke down to soluble phosphorus-containingcompounds.Itisworth noting that, as side reactions,n anoparticles could grow in the dispersion because of the accumulation of PH 3 .F or example, with an extension of the reaction duration to 10 h, noticeable nanoparticles appeared on the surfaces of the TMP flakes ( Figure S7). These experimental results agree well with our proposed mechanisms.
Scanning electron microscopy (SEM) images (Figure 3a, Figure S8a) confirm the polydispersed lateral dimensions of the Co 2 Pa nd Ni 12 P 5 flakes,w hich generally range from 1t o 5 mm, consistent with their AFM images.T he representative TEM images (Figure 3b,F igure S8b) display thin sheets with irregular shapes.T he corresponding selected area electron diffraction (SAED) patterns indicate good crystallinity for both samples.T he surface compositions and electron states were uncovered by X-ray photoelectron spectroscopy (XPS; Figure 3c,d). ForC o 2 P, the characteristic peaks of Co-P corresponding to Co 2p 3/2 and Co 2p 1/2 appear at 777.9 and 792.8 eV,r espectively. [30] Tw os atellite peaks at 785.6 and 802.5 eV and the peaks at 781.2 (2p 3/2 )a nd 797.1 eV (2p 1/2 ) arise from Co À Ob onding in oxidized Co species. [31] Ni 12 P 5 exhibits similar electron states.However,the peaks of Ni 2p 3/2 and Ni 2p 1/2 appear at higher binding energies of 852.8 and 870.1 eV,r espectively ( Figure S8d). [32] Theh igh-resolution P2ps pectra (Figure 3d,F igure S8e) of Co 2 Pa nd Ni 12 P 5 are very similar. Thedoublets at 129.3 and 130.2 eV are assigned to PÀCo or PÀNi, while the broad peak at 133.3 eV originates from the oxidized Pspecies. [30] Theslight surface oxidation is acommon phenomenon of TMP nanomaterials as aresult of air contact, and is consistent with many other reports. [6,33] The2 DT MP sheets constitute ideal platforms for the experimental studies on their electronic properties in fieldeffect transistor (FET) devices.B ottom-gate and gold bottom-contact FET devices were fabricated by spin coating of dilute metal phosphide dispersions onto commercial electrode-patterned substrates with 300 nm SiO 2 dielectric layers (Figure 3e,f), followed by thermal annealing in vacuum to remove the residual solvents and surface oxide layers.A ll measurements on the charge-transport behavior were recorded at room temperature (300 K) in vacuum (ca. 1 10 À5 mbar). As shown in Figure 3g,athin Co 2 Pl ayer (6.7 nm thick) bridges the source and drain electrodes with ac hannel width of 2.5 mm. Thet ransfer curve (Figure 3h) proves that the synthesized 2D Co 2 Ps heet is ap -type semiconductor with ah ole mobility of 20.8 cm 2 V À1 s À1 , calculated from the linear slope of the source-drain current (I ds )p lot. Them obility in 2D Co 2 Pi so nt he same order of magnitude as few-layer metal chalcogenides,varying from 10 to 50 cm 2 V À1 s À1 at 300 K. [34] The I-V characteristics show good linearity in the regime of the source-drain voltage (V ds ) from À3Vto 0V (Figure 3i). The I on /I off ratio based on draincurrent modulation is 1. 8 10 3 .O ur observation agrees with early experimental studies that cobalt phosphides have narrow band gaps of 0.67-0.88 eV in mixed phases, [35] although some density of states (DOS) calculations claim that the Co 2 Ps tructure is metallic without ab and gap. [36] In comparison, the 2D Ni 12 P 5 sheet demonstrates similar features but has al ower hole mobility of 8.7 cm 2 V À1 s À1 and an I on /I off ratio of 1.6 10 2 ( Figure S9). Ultrathin 2D TMPs are promising semiconductor electrocatalysts despite their intrinsic low carrier concentration. [2] At the semiconductor-electrolyte interfaces,t he pronounced "self-gating" phenomenon [37] switches the surface conductance between "on" and "off" states,w hich correspond to highly conductive and insulating phases,r espectively.T he charge-transfer process occurs primarily at the active regions,w hile carrier concentration accumulates to an extremely high level (over 10 14 ecm À2 )a ti nert regions. [37] Thec apability for surface charge modulation strongly supports their fast charge transport kinetics in the electrocatalytic reactions.
As ap roof of concept, 2D Co 2 P, Ni 12 P 5 ,a nd Fe 2 Pw ere prepared to illustrate their potential applications in the OER. Atypical three-electrode system in aN 2 -saturated 1.0 m KOH aqueous electrolyte was applied, using an Ag/AgCl electrode and aP tw ire as the reference and counter electrodes, respectively.A ll potentials were referred to the reversible hydrogen electrode (RHE), and the ohmic potential drop from the electrolyte resistance were subtracted. As depicted in Figure 4a,phosphorene and Fe 2 Pnanosheets have sluggish activity,whereas Co 2 Pand Ni 12 P 5 exhibit smaller onset OER  . Electrocatalytic performances of 2D TMPs towards the OER. a) OER polarization curves of the Co 2 P, Ni 12 P 5 ,F e 2 P, Co 1.5 Fe 0.5 P, phosphorene, and Ir/C electrocatalysts. b) Corresponding Tafel plots of Co 2 P, Ni 12 P 5 ,C o 1.5 Fe 0.5 P, and Ir/C. c) Polarization curves of the 2D Co 1.5 Fe 0.5 Pbefore and after 5000 CV cycles. d) Long-term OER stability test of the Co 1.5 Fe 0.5 Pand Ir/C at acurrent density of 10 mA cm À2 . overpotentials of 280 mV and 285 mV,r espectively.M etal doping or alloying is an efficient method to reduce the kinetic energy barriers,a nd thus to improve the electrochemical performance of TMPs. [38] As expected, iron-doped bimetallic phosphide Co x Fe 2Àx P( 0 < x < 2) shows enhanced electrocatalytic activity (Figures S10 and S11). Remarkably,w hen x = 1.5, Co 1.5 Fe 0.5 Pg enerates oxygen at an overpotential of only about 180 mV,w hich is substantially lower than that of commercial Ir/C (ca. 260 mV; Figure S12). At ac urrent density of 10 mA cm À2 ,C o 1.5 Fe 0.5 Pd elivers al ow overpotential of 278 mV,s uperior to those of Co 2 P( 335 mV), Ni 12 P 5 (375 mV), and Ir/C catalysts (304 mV). Thec atalytic performance of Co 1.5 Fe 0.5 Pa lso outperforms many state-of-the-art OER electrocatalysts,s uch as CoMnP nanoparticles, [39] NiC-oP/C nanoboxes, [40] Mn-Co oxyphosphide multishelled particles, [41] CoFe-layered double hydroxides, [42] meso/micro-FeCoN x -CN, [43] and double perovskite LaFe x Ni 1Àx O 3 nanorods [44] that have overpotentials ranging from 290 to 450 mV (Table S1). Moreover,C o 1.5 Fe 0. 5 Pu ndergoes ar apid chargetransfer process,i mplied by its electrochemical impedance ( Figure S13). TheT afel slope of Co 1.5 Fe 0.5 P(57 mV decade À1 ) is lower than that of Co 2 P( 71 mV decade À1 ), Ni 12 P 5 (76 mV decade À1 ), and Ir/C (78 mV decade À1 ), thus suggesting it has fast OER kinetics (Figure 4b). Theo verpotential of Co 1.5 Fe 0.5 Ps lightly increases by only 1mVa fter 5000 cyclic voltammetry (CV) scans performed between 1.0 and 1.6 Vat as can rate of 50 mV s À1 (Figure 4c), thus indicating its excellent electrocatalytic durability.According to along-term electrocatalytic OER process at aconstant current density of 10 mA cm À2 ,C o 1.5 Fe 0.5 Ps hows excellent stability that retains asteady OER activity over aperiod of 10 h. By contrast, the Ir/C catalyst gradually loses its activity during the measurements (Figure 4d). Theremarkable electrocatalytic performances of 2D Co 1.5 Fe 0.5 Pa re correlated with three major factors:1 )Intrinsic high catalytic activities of Co and Fe as well as the highly exposed surfaces permit rapid oxidation and transformation into the corresponding active metal oxide/ oxyhydroxides; [45] 2) the "self-gating" effect largely improves the concentration of the surface charge during the OER; [37] and 3) the confined 2D sheet structure ensures fast charge transport at the interfaces.

Conclusion
In summary,wehave developed aversatile topochemical strategy to prepare various solution-processable 2D metal phosphides.T ransition-metal precursors,including cobalt and nickel salts,a re able to transform phosphorene nanosheet templates into the corresponding 2D TMPs with an average lateral size of 2.5 mma nd thickness of 3.6 nm. Field-effect transistors disclose the underlying p-type semiconducting properties of as-prepared 2D TMPs,i nw hich, 2D Co 2 P demonstrates ah ole mobility of 20.8 cm 2 V À1 s À1 at room temperature.T he unique electronic structure and highly exposed active surface sites enable high electrocatalytic activity of 2D TMPs for the OER. Remarkably,i ron-doped Co 2 P(i.e.Co 1.5 Fe 0.5 P) exhibits alow overpotential of 278 mV at acurrent density of 10 mA cm À2 and superior stability over ap eriod of 10 hw ithout apparent performance decay.T his topochemical method opens up new horizons for the design and synthesis of nonlayered 2D materials,w hich do not fit traditional exfoliation approaches.