Electrodeposition of an iron thin film with compact and smooth morphology using an ethereal electrolyte

Abstract Electrodeposition of iron (Fe) from an ethereal solution was investigated. The bath consisted of ferrous chloride (FeCl2), diglyme (G2), and aluminum chloride (AlCl3), in which iron species were estimated to be [Fe(G2)2]2+ complex cations. The effect of hydrogen gas evolution on the morphology of iron deposits was determined by comparing common aqueous electrolytes. An Fe thin film was fabricated using the FeCl2–G2–AlCl3 bath without the influence of hydrogen gas evolution, and the nucleation of Fe was explained by an instantaneous nucleation mechanism. As a result, the surface morphology of the Fe thin film was compact and smooth compared with the cases of aqueous and other nonaqueous electrolytes.


Introduction
Iron (Fe) is one of the least expensive and most abundant elements. Metallic Fe shows ferromagnetism, high electrical conductivity, and other advantageous mechanical properties. Therefore, Fe thin film has been well-studied and has many proposed applications in electronics and spintronics, such as magnetic memory and electromagnetic shielding [1e3]. The giant magnetoresistance effect was discovered originally in Fe/Cr multilayers [4]. Ultrahighvacuum techniques are usually used to fabricate Fe thin films [5,6]. An electrodeposition method for thin films can be an alternative to ultrahigh-vacuum techniques, because the costs involved are relatively low.
The electrodeposition of an Fe thin film from aqueous solutions has been studied extensively [7e14]. Because the standard potential of the Fe 2þ /Fe 0 deposition isotherm is more negative than that of the hydrogen evolution reaction d À0.44 V vs. a standard hydrogen electrode (SHE) d the electrodeposition of Fe is inevitably accompanied by side reactions [10]. The H 2 evolution results in a local pH increase near the cathode, forming iron hydroxide or iron hydroxy-chloride, and thereby the deposits are contaminated [11]. In addition, the hydrogen bubbles attached to the deposits inhibit the nucleation and growth of Fe deposits, resulting in a noncompact and rough surface morphology [11e13]. To prevent the formation of iron hydroxide, the pH of Fe electroplating baths should be lower than 3.5 [14]. Meanwhile, a pH buffer is necessary to stabilize the pH values during electrodeposition [15,16].
Nonaqueous baths without active protons are advantageous over conventional aqueous baths, in that H 2 gas evolution can be prevented. There are several studies on Fe electrodeposition from nonaqueous baths, including deep eutectic solvents and ionic liquids [17e21]. The Fe nanoparticles have been obtained from choline chloride (ChCl)eurea with FeCl 3 [17] and an ionic liquid AlCl 3 e1methyl-3-butylimidazolium chloride with electrochemicallydissolved Fe 2þ [18], respectively. In ChCleureaeFeCl 2 [19] and ChCleethylene glycoleFeCl 2 [20] baths, coarse Fe deposits grew on glassy carbon electrode through a progressive nucleation mechanism. Air-and water-stable ionic liquids composed of N-butyl-Nmethylpyrrolidinium (BMP) cation and amide anions with FeCl 2 have also been used to electrodeposit Fe [21].
However, a compact Fe thin film with a smooth surface deposited from nonaqueous electrolytes has not been established. Fe thin films that are not compact are not suitable for many applications. In this work, a nonaqueous electrolyte for fabricating a compact and smooth Fe thin film is investigated. A relatively safe and costeffective organic solvent, i.e., a high-boiling-point ether, diglyme (G2), is used [22]. The dissolved chemical species and the electrodeposition mechanism are discussed.

Reagents and bath preparation
G2 and ferrous chloride (FeCl 2 ) were purchased from Kanto Chemical Co. and FUJIFILM Wako Pure Chemical Co., respectively. High-purity aluminum chloride (AlCl 3 ) was supplied by Nippon Light Metal Co. The water content in the G2 solvent was less than 25 ppm after drying using molecular sieves. All baths were prepared in an Ar-filled glovebox with H 2 O and O 2 contents of less than 5 ppm. The mixing ratio of AlCl 3 eG2 solution was 1:5 by mole. Different concentrations for FeCl 2 in AlCl 3 eG2 solution were prepared.
For the aqueous electrolytes, the FeCl 2 concentration varied from 0.025 mol dm À3 to 0.5 mol dm À3 . Based on previous research [13], some additives, including sodium chloride (NaCl, 0.7 mol dm À3 ), boric acid (H 3 BO 3 , 0.4 mol dm À3 ), saccharin (0.0075 mol dm À3 ), and L-ascorbic acid (0.05 mol dm À3 ), were added into 0.2 mol dm À3 FeCl 2 aqueous solution. Prior to the experiments, the aqueous electrolytes were purged with N 2 for 2 h to decrease the O 2 content. The pH of the electrolytes was adjusted to 3.0 using HCl or NaOH solutions.

Bath characterization
The electrolyte was sealed with a septum in the glovebox, which made it possible to perform viscosity measurements in an Ar atmosphere. The viscosity of the electrolytes was determined using an EMS Viscometer (EMS-1000, Kyoto Electronics Manufacturing  Co.). The ionic conductivities of the electrolytes were measured using electrochemical impedance spectroscopy in a temperature chamber (Espec Co., SU-222) with a self-made two-stainless-steelelectrode cell, and the cell constant was calibrated with a 0.1 mol dm À3 KCl aqueous solution. 27 Al nuclear magnetic resonance (NMR) spectra were obtained (200 scans; acquisition time, 1.5 s) by an NMR spectrometer at 600 MHz (JNM-ECA 600) referenced to DMSO-d 6 (99.9 at% D, Sigma-Aldrich). UVeVis spectroscopy (Hitachi U-3500) was performed at a scan rate of 30 nm s À1 in 1-mm quartz cuvettes over the wavelength range of 200e800 nm. Raman spectroscopy was conducted using an InnoRam 785 (B&W Tek) equipped with a 785-nm semiconductor laser light source.

Electrochemical measurements and deposition
Cyclic voltammetry and potentiostatic electrodeposition were carried out with an electrochemical working station (Bio-Logic Science Instruments SAS, VSP-300). In the FeCl 2 eG2eAlCl 3 baths, the three-electrode system consisted of a working electrode (Cu sheet, 2.0 Â 0.5 cm), counter electrode (Fe sheet, 2.0 Â 2.0 cm), and quasi-reference electrode (QRE, Fe sheet, 2.0 Â 0.5 cm). Electrodeposition was conducted at different constant potentials. In the aqueous baths, Cu, Pt, and Hg/Hg 2 SO 4 (ALS Co., Ltd.) were used as the working electrode, counter electrode, and the reference electrode (0.657 V vs. SHE), respectively. The cathodic electrodeposition in FeCl 2 aqueous electrolytes was conducted at a constant current density of 5 mA cm À2 , pH 3.0, and room temperature with a stirring speed of 500 rpm, which is the common condition for the Fe electrodeposition from aqueous solutions [8,13].

Characterization of deposits
The surface morphology of the deposit was characterized by a scanning electron microscope (SEM, Keyence VE-7800), and an Xray diffraction (Rigaku RINT2200, Cu Ka) experiment was performed at a scan rate of 0.3 min À1 . A transmission electron microscope (TEM, JEM-2100F) equipped with an energy dispersive Xray spectrometer (EDX, JED-2300T) was used for high-resolution transmission electron microscopy (HRTEM) observations, corresponding selected area electron diffraction (SAED), and chemical composition analysis. The samples for TEM and HRTEM were prepared by a focused ion beam (FIB, JEOL JFIB-2300) system.

Solubility of FeCl 2
The solubility of the FeCl 2 in pure G2 and AlCl 3 eG2 solution with a molar ratio of AlCl 3 :G2 ¼ 1:5 was investigated at 25 C. Fig. 1(a) shows that the brown FeCl 2 powder in pure G2 remained undissolved. This insolubility of FeCl 2 in pure G2 can be explained by the strong Coulomb interaction between Fe 2þ and Cl À , which is not affected by ionedipole interactions between Fe 2þ and G2. By contrast, a pale green solution, as shown in Fig. 1(b), was obtained by adding small amounts of FeCl 2 into the AlCl 3 eG2 solution. Prior to the tests, the AlCl 3 eG2 solution was purified to be colorless through pre-electrolysis. As shown in Fig. 1(c), a higher molar ratio  of FeCl 2 led to a saturation with a pale-yellow color. Such limited solubility of FeCl 2 in Lewis acidic solutions has been reported for Lewis acidic chloroaluminate ionic liquids [23,24].
In Fig. 2(c) and (d), baseline correction was carried out using AlCl 3 eG2 solution. The absorbance of each sample increased linearly with the concentration of FeCl 2 . The relationship between the absorbance and the concentration was used to estimate the saturated concentration. The maximum solubility of FeCl 2 in AlCl 3 eG2 solution with a molar ratio of 1:5 was approximately 0.15 mol dm À3 . 27 Al NMR spectroscopy was performed to check the influence of the FeCl 2 addition on Al complexes in the G2 electrolytes. As shown in Fig. 3, each spectrum for AlCl 3 eG2 and FeCl 2 eAlCl 3 eG2 solutions has a sharp peak at high frequencies corresponding to an AlCl 4 À anion and a broad peak with lower intensity resulting from an [AlCl 2 (G2) 2 ] þ cation [30,31] . Therefore, the Fe 2þ eG2 complex forms with the help of AlCl 3 , which is consistent with the UVeVis results.

27 Al NMR and Raman spectroscopy
FeCl 2 dissolved into Lewis basic chloride baths could exist as a  anion [20,32]. However, the Raman spectra of FeCl 2 eAlCl 3 eG2 solution shown in Fig. 4 does not correlate well with that for FeCl 4 2À with a characteristic peak at 265 cm À1 [32]. In addition, the spectrum of AlCl 4 À formed in AlCl 3 eG2 and FeCl 2 eAlCl 3 eG2 solutions presents the same feature. This indicates a weak correlation between AlCl 4 À and Fe(II), which is also reported in the Lewis neutral haloaluminate ionic liquids [23,24]. Therefore, the reactions between FeCl 2 and AlCl 3 eG2 solution can be represented by the following equations: In Eq. (1), AlCl 3 reacts with G2 to form a hexacoordinate [AlCl 2 (G2) 2 ] þ complex cation and AlCl 4 À anion. As a result of the UVeVis, 27 Al NMR, and Raman spectrum analysis, Eq. (2)

Cyclic voltammetry and potentiostatic electrodeposition
In general, Al can be employed as a QRE in AlCl 3 -containing electrolytes for subsequent measurement of cyclic voltammetry and potentiostatic electrodeposition [33e37]. However, when Al was used as QRE in the FeCl 2 eAlCl 3 eG2 electrolyte, the potential difference between two Al electrodes fluctuated considerably ( Fig. 5(a)). This is because a displacement reaction occurred between Al metal and Fe 2þ to give Al 3þ and Fe metal, and the potential-determining reaction cannot be unique. Here, the feasibility of Fe as a 'reference electrode of the first kind' in FeCl 2 eAlCl 3 eG2 electrolyte was examined, by referring a previous report of Sn electrode in an ionic liquid electrolyte [38]. First, the potential difference between two Fe electrodes was measured, proving that the Fe electrode possesses a long-term stability within 1 mV (Fig. 5(a)). A further experiment was carried out using a couple of Fe electrodes immersed in FeCl 2 eAlCl 3 eG2 electrolyte. As shown in Fig. 5(b), after applying potential pulses of þ100 and À100 mV vs. Fe electrode, the potential difference returned to the original value within a few seconds, substantiating that the immersion potential of Fe electrodes is stable against external electrical disturbance as is always required for a reference electrode. Therefore, Fe can be employed as a QRE in this system. A half reaction should be written as [Fe(G2)] 2þ þ 2e # Fe þ 2G2. Fig. 6(a) displays cyclic voltammograms (CVs) with varying switching potentials at points B, C and D for the FeCl 2 eAlCl 3 eG2 electrolyte with 0.025 mol dm À3 FeCl 2 . The CVs differ from those of the AlCl 3 eG2 electrolyte, as shown in Fig. 6(b), indicating that Fe electrodeposition has occurred. As for the FeCl 2 eAlCl 3 eG2 electrolyte, the scan starts from open-circuit voltage toward the negative direction, and the redox couple HI/IJ refers to Cu oxidation [36]. The reduction to Fe starts from point A, as described below. The ironediglyme complex [Fe(G2) 2 ] 2þ indicated by Eq. (2) should be an active species for the reduction to Fe. The following reduction wave BC is also attributed to Fe electrodeposition. This assignment is supported by the fact that the reduction onset AB and a plateau BC correspond to an almost identical oxidation process. The reduction wave BC reaches a plateau at 3.2 mA cm À2 , because the system is in a diffusion-limited condition.
A constant potential À0.5 V vs. Fe QRE was applied for cathodic electrodeposition. Fig. 7(a) shows that the Cu substrate was covered by a deposit that was bright. The brightness originates from a voidfree and smooth surface, as displayed by the SEM image in Fig. 7(b). TEM observations and electron diffraction measurements d see Figs. 7(c)e(e) d show that the deposit was composed of pure Fe with a body-centered cubic (bcc) structure, and the size of the crystallite, marked by magenta circles, was approximately 10 nm. This is consistent with the results of CV, as shown in Fig. 6(a). An XRD profile with broadened peaks is shown in Fig. 7(f), and it confirms that the deposition film, peeled from the copper substrate, was composed of crystalline a-Fe. Additionally, the average crystallite size was estimated to be 10 nm by Scherrer's equation [39], which is consistent with the TEM result in Fig. 7(e). The smaller Fe crystallite size contributes to a smooth surface. The observed lattice distances shown in Fig. 7(e) are consistent with the unit cell size of a-Fe (0.287 nm), suggesting that the nanocrystals exhibit high purity. The absence of Al and C impurities is supported by the EDX mappings shown in Fig. 8, suggesting that Al 3þ eglyme complexes are not reduced at À0.5 V vs. Fe QRE. A potentiostatic electrodeposition at À0.6 V vs. Fe QRE successfully gave a bright Fe thin film, as in the case at À0.5 V. It is emphasized that such a bright and compact Fe thin film is extremely difficult to obtain from aqueous baths [13]. In previous studies on nonaqueous baths, only noncompact deposits of Fe nanoparticles, rather than film morphology, have been obtained in choline chloride (ChCl)eurea deep eutectic solvent with dissolved Fe(III) [17] and AlCl 3 e1-methyl-3butylimidazolium chloride ionic liquid with dissolved Fe(II) [18].
Without FeCl 2 , metallic Al can be produced from the AlCl 3 eG2 bath d see Fig. 6(b). In the CVs shown in Fig. 6(a) from point C to point E, there is a steep increase of current density. The standard redox potential of Al/Al 3þ is À1.67 V vs. SHE, which can be converted to approximately À1.2 V vs. Fe QRE d see Fig. 6(b). As the transition stage at D is around À1.2 V, the CVs shown in Fig. 6(a) gave faradaic current densities similar to that of AlCl 3 eG2 electrolyte ( Fig. 6(b)). It is speculated that the electrodeposition of FeeAl alloys and pure Al and may occur below approximately À1.2 V vs. Fe QRE. New oxidation peaks at points F and G appear below 0 V vs. Fe QRE. Since the oxidation wave at point F is absent for the voltammogram switched at point D (dashed pink curve in Fig. 6(a)), points F and G is considered to be dissolution of electrodeposited pure Al and FeeAl alloys, respectively.
A set of potentiostatic electrodeposition at more-negative potentials was conducted. In Figs. 9(a)e(c), the compact deposits obtained at a potential of À1.0 V vs. Fe QRE are also pure Fe with a metallic luster. At potentials of À1.5 V (Figs. 9(d)e(f)), the TEM-EDX results ( Fig. 9(f)) revealed that the deposits contained Fe, O and Al. To obtain average information of the whole deposits, XRD measurements were conducted at the same condition as the case of À0.5 V. In Fig. S1 (ESIy), Fe and Fe oxides were observed, consistent with the SAED results shown in Fig. 9(f). At more negative potentials of À2.0 V (Figs. 9(g)e(i)), the SAED and TEM-EDX results shown in Fig. 9(i) revealed that the deposits had a much lower crystallinity and also contained Fe, O and Al. Therefore, only metallic Fe with low intensity was observed in Fig. S1 (ESIy) and the XRD peaks of Fe oxides and the possible Al/Al oxide phases should be hidden in the background. Deposits obtained at À1.5 V and À2.0 V had many cracks, and these larger specific surface areas resulted in rapid oxidation in air to give Fe oxides as a main phase.
There were no diffraction peaks of metallic Al or Al oxides for the whole deposits (Fig. S1, ESIy). In general, contaminants with their contents less than a few percentage are not detectable by Xray diffraction. This further confirmed that only a very small amount of Al existed in the deposits, supported by the TEM-EDX results (Figs. 9(f) and (i)). However, it seems contradictory to what is predicted by the CVs shown in Fig. 6(a), where reduction of Al 3þ was observed at À1.5 V and À2.0 V in the short-term electrolysis. In the case of long-term electrodeposition, however, Fe 2þ reduction became dominant even though the FeCl 2 concentration was two orders of magnitude lower than AlCl 3 and the applied potentials were fairly negative. We speculate that the Cu surface still appeared during a short-term electrolysis like CVs and Al 3þ reduction on Cu may take place, while for a long-term electrolysis Cu was totally covered with Fe deposits and the Al 3þ reduction became hard to occur. Therefore, the reduction rate of Fe 2þ should be higher than that of Al 3þ for the long-term electrolysis. Besides, according to Nernst equation, Fe 2þ /Fe still had a more positive reduction potential than Al 3þ /Al in FeCl 2 eAlCl 3 eG2 electrolyte with FeCl 2 of 0.025 mol dm À3 . Therefore, the driving force for Fe 2þ reduction became larger at more negative potentials and the bulk concentration of FeCl 2 (0.025 mol dm À3 ) was not so dilute.
The results should also be considered from the viewpoint of the iron impurity effect on Al electrodeposition. Iron is the major impurity in AlCl 3 [40]. Recently, AlCl 3 eorganic baths have attracted attention for plating and batteries [41]. There have been reported other AlCl 3 -containing organic baths [33,42,43], where cationic complexes [AlCl 2 (ligand) n ] þ similar to [AlCl 2 (G2) 2 ] þ . Based on the results of FeCl 2 eAlCl 3 eG2 electrolyte, it can be deduced that the rate of Al 3þ reduction could also be much lower if FeCl 2 contaminates the baths and reacts with the cationic complexes [AlCl 2 (ligand) n ] þ to form [Fe(ligand) n ] 2þ . Therefore, even a very small amount of iron impurity in the electrolyte prevents the Al electrodeposition and influences the properties of batteries.  Fig. 10(a) shows the typical currentetime transient for the electrodeposition of Fe at a potential of À0.5 V vs. Fe QRE. The current density increased initially with time and reached a maximum i max for the nucleation and growth of Fe particles. Then, the currentetime transient was followed by a decrease resulting from a diffusional overlap. The ScharifkereHills (SeH) models were used to analyze the currentetime transient, to distinguish between the instantaneous nucleation mechanism and the progressive one [44]. In instantaneous nucleation, all the sites for nucleation are activated at the same time, and the nuclei grow at the same rate. In progressive nucleation, the nuclei are still produced along with their growth, and the ages of the nuclei are different. The processes of instantaneous and progressive nucleation, as in Fig. 10(b), can be expressed as Eqs.

Currentetime transient of Fe electrodeposition for the G2 electrolyte
The experimental currentetime transient can be replotted as in Fig. 10(b) and compared with the SeH models. As a result, the potentiostatic electrodeposition at À0.5 V can be understood by the model of the instantaneous nucleation mechanism. Hence, the reason why the Fe deposits present a compact and smooth surface is not only because there is no interference from H 2 gas evolution and the additives. By contrast, previously-reported electrodeposition from nonaqueous solutions are based on progressive nucleation, failed to obtain a compact and smooth films [19,20].
The SeH models can provide an estimate of the nucleation number density of active sites N 0 . For the three-dimensional instantaneous nucleation process, the value of N 0 can be estimated from the values of i max and t max in by the following equations [45]: where n is the number of electrons transferred for each Fe atom deposited, F is Faraday constant, C is concentration of Fe 2þ , M and r are molar weight and density of Fe. There have been several literatures where N 0 is estimated using Eq. (5) [46,47]. However, the previous studies showed that the considerable discrepancies exist in the values of N 0 between that estimated from Eq. (5) and that estimated by counting particles on SEM images deposited for a short time, e.g. t max . The reported values of N 0 are as large as 10 13 e10 16 cm À2 at t max , a several orders of magnitude larger than those estimated from Eq. (5). Moreover, for the previous studies, aggregation of particles occurred before t max [46,47]. This strongly suggests that the preposition of the SeH model, i.e. the assumption of three-dimensional diffusion controlled growth of the nuclei, is not satisfied for the previous studies.
In our case, however, the discrepancy is rather small and no aggregation was observed at t max , evidencing the utility of SeH models in our case. For the potentiostatic electrodeposition at À0.5 V, the calculated value of N 0 using Eq. (5) is 1.13 Â 10 6 cm À2 . Fig. 11(a) is the SEM image of the Fe deposits obtained by potentiostatic electrodeposition at À0.5 V for t max ¼ 4 s. The value of N 0 , calculated by counting the particles in Fig. 11(a), was about 1.5 Â 10 7 cm À2 . Although this is about ten times higher than N 0 obtained using Eq. (5), it is more important that the individual Fe nuclei are isolated from one another and distributed on the Cu substrate at t ¼ t max ¼ 4 s (Fig. 11(a)). Even at t ¼ 30 s, the nuclei became larger with the similar particle densities (Fig. 11(b)), proving that the preposition of the SeH model, i.e. the assumption of three-dimensional diffusion controlled growth of the nuclei, is satisfied. As a result, the deviation is ignorable between the fitting and the experimental curves in Fig. 10(b), in stark contrast to the previous studies [46,47].  Figs. S2(a) and (e), the electrolyte contains 0.025 mol dm À3 FeCl 2 d the same as that in the FeCl 2 eG2eAlCl 3 electrolyte. There was a clear increase in current density when the potentials were swept to À0.87 and À1.07 V. Because the standard potential of Fe 2þ /Fe 0 (À0.44 V vs. SHE) is more negative than that of the H 2 evolution reaction, the initial increase of current density in region I d see Figs. S2(e)e(h) d results from the H 2 evolution. Based on the Nernst equation, the estimated reduction potential of H þ /H 2 in these aqueous electrolytes with pH ¼ 3 is À0.09 V, which is consistent with the potential initiating the increase of the current density in the CVs d see  Figs. S3(a)e(f) (ESIy) show that the Fe deposits had a metallic luster and were scattered with voids, where the reduced H 2 adhered to Fe deposits during electrodeposition. With additives, however, the samples showed a black appearance, as shown in Fig. S3(h), and granular microstructure d see Fig. S3(g). Therefore, H 2 gas evolution and the additives in aqueous electrolytes interfere with the growth of deposits, and subsequently lead to a noncompact and rough surface morphology. Instead of aqueous electrolytes, Fe can be deposited uniformly from nonaqueous FeCl 2 eAlCl 3 eG2 electrolyte without H 2 gas evolution, resulting in a compact and smooth Fe thin film.

Conclusions
A nonaqueous electrolyte, FeCl 2 eAlCl 3 eG2, made the electrodeposition of Fe thin film successful. Electrodeposition at a constant potential of À0.5 V vs. Fe QRE gave a compact structure and smooth surface, not only because the Fe nucleates in an instantaneous nucleation mechanism, but also because the influence of H 2 gas evolution was excluded. The obtained Fe deposits were composed of nanocrystals, the grain size of which was much smaller than those in previous reports. This burnished Fe film has potential applications in high-precision optical devices [48], as well as electromagnetic devices. Related studies are planned for the future. FeCl 2 cannot be dissolved into G2 without the help of AlCl 3 , and the electrochemically active species should be the [Fe(G2) 2 ] 2þ cation. Moreover, this study has shown the possibility that the properties of Al-ion batteries and Al plating baths could be significantly degraded by a very small amount of iron "impurity".
CRediT authorship contribution statement