Synthetic Mechanism Studies of Iron Selenides: An Emerging Class of Materials for Electrocatalysis

: Solution-processed iron selenide nanocrystals (NCs) have recently attracted considerable attention in electrocatalysis water splitting. Nevertheless, a primary challenge in current iron-based NCs chemical synthesis is controlling phase purities between each chalcogen (monochalcogenide, dichalcogenides, and oxides), which requires a comprehensive understanding of the reaction mechanisms at the early stages of nucleation. Herein, we investigate the fundamental steps in transforming molecular organoiron and organoselenium precursors to iron selenides NCs with the view of de-veloping universal synthesis protocols for phase pure metal selenium and metal oxides NCs. The main intermediate species and volatile by-products are identiﬁed by high-resolution electron microscopy and Nuclear Magnetic Resonance (NMR) spectroscopy ( 1 H, 13 C, and 31 P). Experimental evidence suggests that the phase determining factor is the coordinating reactivity difference between oleﬁns (1-octadecene, oleylamine), tributylphosphine and trioctylphosphine associated with their corresponding Se bond cleavage. This work proposes organoselenium interconversion reaction mechanisms during iron selenides synthesis, offering a universal synthetic strategy for other electrocatalytically or photocatalytically active layered metal selenides materials.

In the present work, we investigate the fundamental steps in transforming molecular organoiron and organoselenium precursors to iron selenides NCs during the organometallic reaction to develop universal synthetic protocols for preparing phase pure FeSe and FeSe 2 NCs. Four common organic reaction media have been chosen, and the main intermediate species and volatile by-products were identified by Transmission Electron Microscopy (TEM), High Resolution Transmission Electron Microscopy (HRTEM) and NMR spectroscopy ( 1 H, 13 C, and 31 P). Phosphine Lewis bases such as tributylphosphine (TBP) or trioctylphosphine (TOP) were chosen due to their strong coordination to the chalcogenide component. In addition, long alkane chain olefin compounds such as 1-octadecene (1-ODE) and oleylamine (OLA) were also selected since both reagents are the most popular phosphine-free solvents for metal chalcogenide nanocrystal synthesis. Experimental evidence suggests that the phase determining factor is the coordinating reactivity difference between olefins (1-ODE and OLA), TBP, and TOP associated with their selective Fe-Se bond cleavage.

Chemical Reaction Mechanisms in the 'Hot-Injection' Approach
A double cold trap apparatus was employed to extract the volatile reaction intermediates. Se precursor was prepared by dissolving Se in TBP. The iron precursor was formed by iron (II) acetate (Fe(AC) 2 ) reacting with stearic acid (SA) in 1-ODE. As shown in Figure S1 (Supplementary Information), the reaction product generated from the first reaction of iron acetate with SA and 1-ODE is monitored. The volatiles generated from the reaction were trapped in the double 'cold trap', which was equipped with a Dewar with liquid nitrogen. The formation of the iron precursor was examined by taking aliquots during the heating stage prior to the injection of Se precursors. Figure 1a,b shows 1 H and 13 C NMR analysis of the product (Sample I) collected by the 'double trap', which was generated after loading iron acetate, SA and 1-ODE into the reaction vessel and heating at 60 • C and 130 • C for half an hour, respectively. In Figure  1a, the formation of acetic acid is revealed by the strong intermolecular hydrogen bonding in carboxylic acid dimers, which leads to very low-field broad signal δ = 9.74 (sample I) and 11.00 (commercial, acetic acid, ≥99.7%, Sigma-Aldrich) in the range of 9-15 ppm. The up-field proton signal is similar between Sample I and commercial acetic acid, such as the singlet at around 2 ppm is also characteristic (CH 3 ) of acetic acid. Further evidence is provided by the 13 C NMR spectra, as shown in Figure 1b, in which identical carbon shifts are observed for Sample I and acetic acid standard sample.  Figure 1c shows the 13 C NMR analysis of iron-containing species collected from the iron precursor. Compared to the 13 C NMR spectra of standard SA (analytical grade, Fisher Scientific) and acetic acid, the reaction solution shows negligible acetic acid content. On the other hand, iron stearate (FeSt 2 ) can be resolved with identical chemical shifts to stearic acid except for the absence of carbonyl's chemical shift ( 13 C, δ = 180 ppm) ascribed to the screening effect upon bonding with iron.
However, after injection of Se precursor (Se in TBP), no color change occurs even when the heating time is extended to 30 min. Furthermore, as shown in Figure 1d, there are no characteristic or prominent lattice reflections from the XRD analysis of the crude solution, especially at large angles. From the result achieved so far, the only conclusion is the formation of amorphous metal or organic iron complex rather than highly crystalline nanoparticles.
Based on the NMR and XRD analysis, a plausible reaction pathway is proposed in Scheme 1. Fe(AC) 2 reacts with SA in the presence of 1-ODE via a ligand exchange reaction, generating FeSt 2 as iron precursors and volatile acetic acid. Due to the strong bonding between Se and phosphorus, the Se-TBP bond is not readily cleaved under these conditions after the injection of Se precursors. Consequently, no crystalline products are formed in the end. The second investigation is carried out by adjusting Se precursors' reactivity from replacing TBP by OLA. Since the coordination between OLA and Se is weaker than that of phosphines [45,55], we anticipate a more labile Se-OLA bond cleavage, thus accelerating the reaction of Se with FeSt 2 and resulting in the formation of the desired iron selenide NCs. Figure 2a shows the XRD of as-prepared nanocrystals obtained in the presence of OLA. In contrast to the TBP approach, highly crystalline orthorhombic FeSe 2 is achieved, with each XRD feature assigned according to the standard PDF card index (PDF = 21-0432 and 74-0247).   Figure 2d resolved the reflection from the (110) plane, with a lattice spacing of 3.8 ± 0.2 Å, which is close to the value for single crystal orthorhombic iron diselenide.
Based on the XRD, TEM and HRTEM analysis results, a plausible reaction pathway of the second approach is proposed, as shown in Scheme 2. The mechanism for the FeSt 2 formation is the same as Scheme 1. However, the reaction between selenium and OLA will generate organoselenium complexes, as reported in our previous works [45,59] These less reactive organoselenium complexes react with FeSt 2 at elevated temperature and facilitate the formation of FeSe 2 NCs. Scheme 2. Plausible reaction pathway for the second approach for the synthesis of FeSe 2 , employing FeSt 2 as an iron precursor and OLA for solubilization of Se.
The third investigation involves the replacement of 1-ODE by OLA in the iron precursor. In that case, both iron and Se precursor were prepared in the presence of OLA. After injection of the Se precursor, strongly magnetic particles are formed, as revealed by the strong adsorbing on the magnetic stirring bars. TEM analysis revealed spherical shape nanoparticles. Figure 3b shows the XRD analysis of the final product from the third approach, with no evidence of iron selenides. High purity Fe 3 O 4 nanocrystals are found as confirmed by the well-aligned XRD reflections compared with standard face-centered cubic Fe 3 O 4 PDF card (PDF = 75-0033). The crystal domain size is calculated to be 10.9 ± 4.3 nm from the Scherrer equation. HRTEM analysis resolved the reflection from the (220) plane in Figure 3c with a lattice spacing of 3.0 ± 0.2 Å confirmed the cubic structure. Furthermore, the (111) plane ( Figure 3d) with a lattice spacing of 5.0 ± 0.2 Å also resolved, which are close to the bulk Fe 3 O 4 structure value.  Based on XRD, TEM, and HRTEM analysis of the data, a plausible reaction pathway for the third approach is proposed in Scheme 3. The critical step in iron precursor formation is the anion exchange process driven by the depletion of acetic acid and the construction of FeSt 2 . As shown in Scheme 3, when 1-ODE is replaced by OLA, the formation of FeSt 2 is prohibited by a condensation reaction between OLA and SA instead of the anion exchange reaction between SA and acetic acid. Therefore, when elevating the temperature after the injection of Se precursors, iron acetate will first undergo a thermolysis reaction generating Fe 3 O 4 NCs rather than reacting with Se species.
Furthermore, as discussed earlier in Schemes 2 and 3, the formation of FeSt 2 and Se-C organoselenium species is crucial to achieving high purity FeSe 2 without the formation of Fe 3 O 4 . Therefore, an excess amount of SA, vacuum drying and suitable selenium precursors are the critical factors in the FeSe 2 NCs synthesis through the 'hot-injection' approach. It should be noted that the organic metallic chemical synthesis of iron nanoparticles, iron metal or iron-nonmetal alloy is also a rather complex reaction process [59]; however, similar approaches to our current work could be carried out in the future to elucidate the underlying synthetic mechanisms.

Selenium Extraction for the Synthesis of FeSe Nanoparticles
The synthesis of FeSe was carried out through the extraction of selenium from FeSe 2 by TBP, analogous to the previous report [74]. However, no FeSe nanocrystals could be found in the final solution after the chemical extraction process ( Figure S2). The XRD analysis in Figure S2 of the crude solution shows no prominent reflections from FeSe NCs, especially at large angles. A plausible chemical pathway can be proposed, as shown in Scheme S1. It can be concluded that after reacting with TBP, all the FeSe 2 changed to amorphous metal or organic iron complex rather than highly crystalline NCs.
An alternative approach was carried out by replacing TBP with TOP. During the reaction between FeSe 2 and TOP, 1 mL of the reaction solution was syringed out under Ar protection, and J. Young NMR tube was employed for 31 P NMR analysis (denoted as Sample II) since TOP is air sensitive. Aliquots were also taken out for monitoring crystal structure evolution by XRD analysis. Figure 4 shows the typical 31 P NMR analysis result. Taking TOP standard as a reference, a new singlet with a chemical shift of δ = 36.28 ppm is appeared after the phase transformation. The signal is indexed to P = Se correlation which is consistent with the literature value [75].
The as-prepared FeSe nanoparticles show flake-like features, as shown in Figure 5a of the TEM analysis. The HRTEM image in Figure 5b resolves the lattice fringes of the (001) plane, with a lattice spacing of 5.9 ± 0.2 Å, consistent with the bulk FeSe tetragonal crystal value (PDF = 85-0735). Figure 5c shows the XRD analysis of the aliquots during the extraction reaction. Well-defined reflections can be revealed in Figure 5c, which displays series of reflection evolutions during phase transformation from orthorhombic FeSe 2 to a tetragonal FeSe crystal structure.  Based on the analysis of reaction intermediates extracted from TEM and XRD, a plausible reaction mechanism was proposed, as shown in Scheme 4. FeSe 2 reacted with TOP and diffuse half the equivalents of Se generating FeSe and TOPSe. It should be noted that the delicate detail step reaction mechanism remains unclear; however, a conclusive claim can be made that the amount of TOP added into the solution and the reaction duration are crucial factors. For instance, we found an excess amount of TOP or extending the reaction time will result in the formation of amorphous materials instead of FeSe NCs.

Optical Properties Analysis
Reflectance is the fraction of light reflected from a surface as a function of wavelength. When properly measured, reflectance provides information about the optical properties of materials, since the light that is not reflected is either absorbed due to its chemical composition, or is scattered and/or transmitted, depending on mechanical structure.
Generally, the surfaces of substances reflect light differently depending on their texture. Smooth surfaces such as mirrors reflect the beam in exactly the same direction as the incident beam. On the other hand, rough surfaces substance exhibit diffuse reflection, whereby the rays of the incident beam are scattered in all directions. Normally, a typical surface is in neither smooth or completely rough, and therefore has both specular and diffuse components. Therefore, using an integrating sphere which collects all of the beams reflected by the material surface, one can estimate the optical properties of the substance. To extract the band gap from reflectance measurements, the functions proposed by Tau [76] and Kubelka-Munk were employed (see derivative process below).
where: h: Plank's constant, ν: frequency of vibration, α: absorption coefficient, E gap : band gap, A: proportional constant. The value of n is denoted according to the nature of the materials: for direct allowed transition, n = 1 2 ; for direct forbidden transition, n = 3/2; for indirect allowed transition, n = 2; for indirect forbidden transition, n = 3. For both FeSe and FeSe 2 , direct allowed transitions have previously been reported. Therefore, in this case, the value of n in Equation (1) is 2. Figure 6a,e shows the multiple total reflection measurements on different areas of the glass slides covered by FeSe 2 and FeSe, respectively. Since it is difficult to ensure the homogeneous coverage of the slides by drop-casting, the position of the slide was directed toward the light source together with multiple measurements to minimize errors. By transforming the reflectance experimental data (Figure 6a,e) following Equations (1)-(3), the optical band gap of FeSe 2 (Figure 6b-d) and FeSe (Figure 6f-h) were predicted from the intercepts of the tangent line of the curves to the abscissa.
High absorption and low reflectance can be observed from Figure 6a, indicating that the bandgap of FeSe 2 is less than 1.2 eV. Based on the Tau and Kubelka-Munk equations, [76] the optical band gap of FeSe 2 is estimated to be 0.67 eV. For FeSe NCs, the reflectance measurement shows a change in the reflectance at a wavelength of around 700 nm for all three measurements. By performing the same transformation, the optical band gap of FeSe is estimated to be 1.73 eV. It should be noted that in Figure 6a,e, the reflectance increases swiftly when the incident light has a wavelength below 400 nm. This could be attributed to a scattering effect. In the Rayleigh approximation as depicted in Equation (4): where I 0 is the light intensity before the interaction with the particle, R is the distance between the particle and the observer, θ is the scattering angle, n is the refractive index of the particle, and d is the diameter of the particle. For instance, taking some common semiconductors as examples, and setting the distance between the particle and observer at 0.2 m, the size of the particle as 100 nm, I 0 equal to unity, and taking and the refractive index of CdSe, [77] CdTe, [77] CdS, [77] ZnSe, [77] ZnO, [78] TiO 2 , [79] PbSe [77] from literature values. The evolution of I as a function of λ and refractive index n is shown in Figure 7. From the plot, an increase in scattering is observed as the wavelength of incident light decreases. Further explanation should be given about the index of refraction, since the index of refraction of materials is also a function of wavelength. The values taken from the literature are measured under different wavelength conditions. However, in the current study, it is assumed that the refraction index of the selected materials remains constant during the light scattering process, and focusing only on the variation of light scattering at the surface of the semiconductor as a function of incident light wavelength. The term scattering refers to a general physical process, such as when the ray form of light, sound or even moving particles are forced to change their trajectory by one or more localized non-uniformities in the medium which they are in. Reflection can be simply understood as being a special type of scattering process, whereby, the line normal to the surface equally divides the angle between incident light and reflected light, in accordance with the so-called law of reflection. Therefore, as noted before, a light scattering will induce the enhancement of reflection within the integrating sphere reflection measurement.

'Hot-Injection' Approach
The synthesis of FeSe 2 was carried out under standard oxygen/water-free conditions using a Schlenk line. Two types of selenium precursors are prepared: selenium in oleylamine (OLA) and selenium in tributylphosphine (TBP). Se in oleylamine: 0.1342 g (1.7 mmol) Se (99.99%, Sigma-Aldrich, Dorset, UK ) and 10 mL oleylamine (technical grade, 70%, Sigma-Aldrich, Dorset, UK) were loaded into a two-neck flask under Ar and vacuum drying at 60 • C and 130 • C for 30 min, respectively. Then, Se was allowed to dissolve completely by heating at 200 • C under Ar overnight. Se in TBP: a Se stock solution (1 M) was prepared by dissolving 0.79 g of selenium in 10 mL of TBP, yielding a colorless solution.
The Se precursor (transparent colorless for Se in TBP or transparent yellowish solution for Se in OLA) was injected into the iron precursor (prepared in 1-ODE or OLA) under Ar at 130 • C.
After injection, the temperature increased to 200 • C under vigorous stirring. The reaction temperature was held at 200 • C for 20 min before cooling to room temperature. Toluene (7 mL) was injected into the final solution, and the solution stored under Ar. The purification of as-prepared FeSe 2 involves precipitation by adding isopropyl alcohol and centrifugation at 8000 rpm for 10 min. To remove the excess of organic stabilizer, centrifugation was carried out three times.

TBP and TOP Extraction Se from Iron Diselenide
The procedure is similar to the method reported recently8] Approximate 30 mg of purified FeSe 2 was loaded into a two-neck flask under Ar. 10 mL TOP or TBP was added, and the mixture was dried in a vacuum at 60 • C for 30 min. The temperature increased to 180 • C and held at this temperature for 50 min with vigorous stirring. The mixture was allowed to be cooling down to room temperature. The NCs were purified by precipitation with hexane and methanol mixture and centrifugation (3 times). The final purified FeSe NCs were re-dispersed in toluene and stored under Ar.

NMR, XRD and TEM Methodologies and Instrumentation
Reactions requiring anhydrous conditions were performed under an atmosphere of either anhydrous nitrogen or argon. All glassware was flame-dried prior to use, and glass syringes and needles were placed in an oven (150 • C) for at least 2 h and allowed to cool in desiccators under an atmosphere of anhydrous nitrogen. 1H NMR and 13 C NMR spectra were measured in a 400 MHz Varian INOVA 400 instrument (Varian, Crawley, UK). Chemical shifts are quoted in parts per million (ppm) and referenced to SiMe 4 (1H NMR 0 ppm) and CDCl 3 ( 13 C NMR 77.16 ppm). The 13 C NMR sequence parameters were set up as follows: 45 • pulse of 7.70 µs, τ = 1.0 s, and acquisition time of 1.28 s. The 31 P NMR spectra were measured in a Jeol ECP (Eclipse) 300 instrument (JEOL, Herts, UK). The crystal structure of the products was investigated by powder XRD (BRUKER D8, Cu Kα radiation λ = 1.54 Å, Durham, UK), JEOL 2011 200 kV Hi Resolution TEM (HRTEM, JEOL, Herts, UK) fitted with an EDX Oxford Instruments ISIS 300 system (Oxford Instruments plc, Abingdon, UK).

Optical Properties Analysis
Typically, FeSe 2 or FeSe nanocrystals which dispersed in toluene were drop-casting onto a quartz slide. Then, the reflectance of the slides is analyzed at a 90-degree incident bean angle position on a Perkin Elmer Lambda 35 spectrometer (PerkinElmer, Beaconsfield, UK).

Conclusions
In summary, FeSe 2 , FeSe and Fe 3 O 4 NCs were successfully produced through a wellcontrolled synthetic strategy. The underlying reaction mechanism was, for the first time, monitored by in situ NMR analysis of intermediates generated during the synthesis of FeSe 2 from Fe(AC) 2 . Plausible reaction mechanisms are proposed based on the detected intermediate species. Upon understanding the chemical reaction process, iron oxide NCs were synthesized by thermolysis reaction. It was found that Fe 3 O 4 is the most stable iron oxide phase during the thermolysis of iron acetate, rather than Fe 2 O 3 or FeO. By 'chemical etching' of FeSe 2 , tetragonal crystal structure FeSe nanoparticles were synthesized successfully. As determined by 31 P NMR, TOP was found to play a role in the lattice diffusion of Se from FeSe 2 . The formation of FeSe was found subsequently with the cleavage of the Fe-Se bond and the formation of the Se-P bond. A plausible reaction mechanism of this phase transformation process was also proposed. The optical properties of as-prepared FeSe and FeSe 2 were investigated by integrating sphere total reflection measurements. It was found that the optical band gap of FeSe 2 is around 0.67 eV, and FeSe is 1.73 eV. Given the complex nature of the phase diagram of iron chalcogenides, the fundamental understanding of phase pure synthesis of FeSe and FeSe 2 NCs are of paramount importance for their photocatalysis and electrocatalysis applications. This work provides comprehensive studies on the organoselenium interconversion reaction mechanisms during iron selenides synthesis, which offer a universal chemical synthetic pathway for preparing phase pure electrocatalytically or photocatalytically active layered metal selenides materials.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/ 10.3390/catal11060681/s1, Figure S1: A 'double cold trap' set up for the collection of the volatile reaction intermediate species, Scheme S1: Plausible reaction mechanism for the synthesis of FeSe from FeSe 2 and TBP, Figure S2: The powder XRD pattern of the final product generating from the preparation of FeSe, which employs TBP as an extraction reagent.