One million cyclable blue/colourless electrochromic device using K2Zn3[Fe(CN)6]2 nanoparticles synthesized with a micromixer

Well-defined K2Zn3[Fe(CN)6]2-nanoparticles (NPs) synthesized with a micromixer showed robust redox reaction cyclability. Crystal structure analysis revealed that the robustness results from the maintenance of the original rhombohedral crystal structure in the oxidation state. The blue/colourless electrochromic device with the K2Zn3[Fe(CN)6]2-NPs and Prussian blue NPs showed recyclability over 1 million redox reactions.

Electrochromic devices (ECDs), otherwise known as colourswitching devices, produced by electrochemical redox reactions, are attracting attention owing to their several applications in various systems such as in electric paper display devices [1][2][3] and smart window systems for the external control of light and heat ow. [4][5][6][7] Several types of materials have been reported for use in ECDs, 8 e.g. Prussian blue analogues, [9][10][11][12] tungsten oxide, [13][14][15] and organic materials, including p-conjugated organic polymers. 16 When these materials are used as working electrodes in ECDs, the application of electrochemically active material for a counter electrode is also essential. Particularly, if the ECDs are used for smart windows or transparent displays, the counter electrode material must satisfy either of the following requirements: it must be transparent for visible light in both the redox states or show complementary reaction with the working electrode. For example, potassium zinc hexacyanoferrate (K 2 Zn 3 [Fe(CN) 6 ] 2 , KZnHCF), a Prussian blue analogue, was found to have high transmittance in the visible region at both the oxidation (Zn II 3 [Fe III (CN) 6 ] 2 ) and reduction states (K 2 Zn II 3 [Fe II (CN) 6 ] 2 ). 17 In addition, the KZnHCF nanoparticles (NPs) show a relatively high cyclic stability in ECD application. 18 In this case, KZnHCFs were prepared using classication to pick up small NPs, implying that the yield was rather low. In this study, to increase the yield to prepare the well-dened KZnHCF NPs, we developed a new synthesis method of KZnHCFs using a micromixer. Using the KZnHCF-NPs synthesized with the micromixer, high cyclic stability was achieved with over 1 million runs using the KZnHCF-NPs as the counter electrode in ECDs in combination with the Prussian blue NPs as the working electrode.
KZnHCF-NPs were synthesized with the micromixer and by a batch method by mixing the same volume of 4 mmol L À1 of K 4 [Fe(CN) 6 ] aqueous solution and 6 mmol L À1 of ZnCl 2 aqueous solution and adjusting the pH to 5 using HCl and NaOH (hereinaer referred to as the standard concentration). The reaction is expressed by the following equation: The temperatures of the solutions before and aer utilising the micromixer were adjusted from 8 to 10 C using a 10 m precooling tube consisting of an internal diameter (i.d.) of 0.5 mm stainless steel (SUS304) and a 100 m reaction tube consisting of an i.d. of 3 mm silicon tube, respectively (here-inaer referred to as the standard condition), as shown in Fig. 1(a). To evaluate the effectiveness of the micromixer, the dependence of the particle size on the ow rate was investigated. Subsequently, KZnHCF was synthesized by mixing standard concentrations of ZnCl 2 and K 4 [Fe(CN) 6 ] aqueous solutions with the total ow rates set to 10, 40, 100, and 140 mL min À1 , separately. The linear velocities in the micromixer were 9.4, 38, 94, and 132 m s À1 at 10, 40, 100, and 140 mL min À1 , respectively. The batch sample was synthesized by mixing 50 mL of each reagent, whose solutions were the same as the standard solutions in a beaker stirred and cooled in ice water. The dependence of the reagent concentration on the NP size was evaluated in the standard concentration and at 132 m s À1 with different concentrations, where the concentrations of ZnCl 2 were 0.60, 2.0, 4.0, 6.0, 60, and 600 mmol L À1 , and the concentrations of K 4 [Fe(CN) 6 ] were 0.40, 1.32, 2.67, 4.0, 40, and 400 mmol L À1 .
The particle morphology was evaluated using a scanning electron microscope (SEM, S-4800, Hitachi High-Technologies Corp.). The particle size was evaluated based on the longest diagonal distance in the SEM gure. The average distribution of the particle size was evaluated with considering more than 100 particles per sample.
The crystal structure of KZnHCFs was evaluated by powder Xray diffraction (PXRD, D2 phaser, Bruker Corp.). A KZnHCF thin lm was prepared on a stainless steel (SUS-304) substrate using the spin-coating method. The crystal structure of the sample was evaluated by Rietveld renement using DIFFRAC.TOPAS 5 (Bruker Corp.). The experimental details of the lm preparation are provided in ESI. † The crystal structure of the oxidised state of KZnHCF was analysed aer the electrochemical oxidation, as shown in ESI. † Three terminal electrochemical evaluations of the KCuHCF electrode were made as follows: the KZnHCF thin lms on an indium tin oxide (ITO) glass substrate were also fabricated by spin coating under the conditions of 1000 rpm for 10 s on the 2.5 cm Â 2.5 cm ITO glass substrate. The area of the working electrode was set to 1.0 cm 2 . A propylene carbonate (PC) solution comprising 0.1 mol L À1 (PC) of potassium bis(tri-uoromethanesulfonyl) imide, saturated calomel electrode (SCE), and platinum wire, were used as the electrolyte, the reference electrode, and the counter electrode, respectively. The electrochemical behaviour was analysed using an electrochemical analyser (ALS6115D BAS inc.). The cyclic voltammogram (CV) was obtained at a scanning speed of 5 mV s À1 and a potential range of 0.4-1.4 V (vs. SCE). The ultraviolet-visible (UV-vis) spectra of the samples were obtained using a UV-vis spectrometer (USB-4000, ocean optics corp.) aer the spincoating process on the ITO glass substrate.
The ECDs were fabricated by sandwiching the electrolyte between the Prussian blue lm on the ITO glass and the KZnHCF lm on the ITO glass. The detailed method for the fabrication and preparation of the ECDs are described in ESI. † For the electrolyte, the propylene carbonate solution containing 0.1 mol L À1 of potassium bis(triuoromethane sulfonyl)imide and 360 g L À1 of polymethyl methacrylate was used. The waterdispersible ink of Prussian blue NPs (Fe III [Fe II (CN) 6 ] 3/4 ) was purchased from Kanto-chemical co., inc. The electrochemical reactions at each electrode are detailed below.
The cyclic stability test of the ECDs was performed for each ECD consisting of KZnHCFs under the following condition. The voltage was set at À0.3 V for 4 s and at À1.2 V for 2 s cyclically with the use of an electro power supply (GS200, Yokogawa Denki) for 1 million cycles. To evaluate the performance change of the ECDs in the cyclic stability test aer 110 000 600 000, and 1 000 000 cycles, the performances of the ECD samples were evaluated by chronocoulometry processes and UV-vis spectroscopy. In chronocoulometry processes 1 and 2, the voltage was set from À0.3 V to À1.2 V (CC1) and from À1.2 V to À0.3 V (CC2), respectively, and the potential was maintained for 30 s. Five samples of each KZnHCF were taken and their averages were evaluated.
First, we investigated the KZnHCF NP samples synthesized with different linear velocities, v L , by SEM. The averages and standard deviations of the particle size were 124 AE 38, 99 AE 34, 88 AE 29, and 80 AE 24 nm for v L ¼ 9.4, 37, 94, and 132 m s À1 , respectively. In the case of batch synthesis, the particle size was 192 AE 50 nm. The averages and dispersions of the particle size decreased as v L increased, as shown in Fig. 1(d). The particle size distributions in each linear velocity are shown in ESI. † In general, the micromixer synthesis with a relatively high v L provides relatively small and homogeneous NPs due to intense and rapid mixing in the micromixer. 19 The mixed ow in the micromixer, except for the case of v L ¼ 9.4 m s À1 , was considered to be turbulent because the Reynolds number was greater than 3000. Particularly, 86% of the particles synthesized at the highest linear velocity (v L ¼ 132 m s À1 ) are smaller than 100 nm.
We discovered that the reagent concentrations have a strong impact on the morphology and particle size of KZnHCF. The concentration of ZnCl 2 , c Zn , was set to be 1.5 times higher than that of K 4 [Fe(CN) 6 ], c Fe . The dependence of the particle size on c Zn is shown in Fig. 2. The particle size had minimal values, 2.0 < c Zn < 6.0 mmol L À1 . In the case of c Zn < 0.6 mmol L À1 , some KZnHCF NPs had various polyhedral shapes. Such shapes may not be suitable for ECDs made by the coating method, because the polyhedron corn prevents particle interactions, implying weak electron migration. Conversely, in the case of c Zn > 6.0 mmol L À1 , very large particles were obtained, which are also unsuitable for ECDs due to the slow ion migration in the particles. As a result, we conclude that 2.0 < c Zn < 6.0 mmol L À1 would be suitable for the ECD. In particular, for c Zn ¼ 6.0 mmol L À1 , the synthesis yield of KZnHCF was beyond 90%. Conversely, there was no dependence on the tube length in the reactor.
We also evaluated the inuence of temperature on the particle size. The relatively high temperature provided relatively large particle sizes of KZnHCF (See Fig. S1 †). The cooling condition with ice water is better for synthesizing the KZnHCF NPs.
For the detailed studies, the KZnHCFs synthesized using the two methods were compared. The batch method (KZnHCF-B) and the micromixer method (KZnHCF-M) with RF ¼ 132 m s À1 and c Zn ¼ 6 mmol L À1 . The as-synthesized KZnHCF-B has low dispersibility in water and could not be well coated on the ITO glass substrate with aggregations. To make KZnHCF-B dispersible, we conducted surface treatment with a K 4 [Fe(CN) 6 ] solution in accordance with previous studies. 20 Conversely, KZnHCF-M was sufficiently dispersible for the fabrication of the thin lm by spin-coating. The obtained KZnHCFs were modied with these K 4 Table S1. † The number of the surface sites of KZnHCFs is much lower than that of [Fe(CN) 6 ] in the crystal structure; the surface treatment caused a slight difference in the chemical composition. The PXRD patterns are consistent with the previously reported structure of Na 2 Zn 3 [-Fe(CN) 6 ] 2 , and the space group is R 3c, 23 as shown in Fig. S4(b). † The electrochemical properties of the thin lms of KZnHCF-BS and KZnHCF-MS on the ITO glass substrates were investigated with the three-terminal electrode method. Their CV curves showed nearly similar proles with two pairs of redox peaks, as shown in Fig. 3(a). The main peak pairs correspond to the redox reaction of [Fe II/III (CN) 6 ] constructing a porous network in KZnHCF. The second peak could be caused by the redox reaction of the remaining K 4/3 [Fe II/III (CN) 6 ] surface. The chronocoulometry curves of KZnHCF-BS and KZnHCF-MS were obtained as shown in Fig. 3(b). The injected charge density of the KZnHCF-BS thin lm was 5.3 mC cm À2 and that of KZnHCF-MS was 6.5 mC cm À2 .
The UV-vis spectra of the KZnHCF-BS and KZnHCF-MS electrodes were also measured. The transmittance of the KZnHCF-MS electrode on the ITO glass is higher than that of the KZnHCF-BS electrode on the ITO glass as shown in Fig. 3(c). The low transmittance of KZnHCF-BS could be caused by the scattering of the relatively large particles. The high transmittance of the KZnHCF-MS electrode indicates less surface roughness even in comparison with the bare ITO substrate.
The crystal structures of the electrochemically oxidised states of KZnHCF-MS were evaluated by PXRD and Rietveld renement. The XRD pattern of the oxidised and reduced states shown in Fig. S4(a) † indicates the same space groups, R 3c, as    6 ] 3 ) which has large absorbance at the peak of 700 nm due to the metal to metal charge transfer, as shown in Fig. 4(c) 6 ] 3 ), which is known as Prussian white owing to its low absorbance in the visible region.
Concerning the colour-switching speed, we evaluated the reaction time for colouration, t C , and for bleaching, t B . These indicators are dened as the time required for 80% of transmittance change from the initial state to the completely switched states of CC1 and CC2. 18 For colouration, we evaluated t C ¼ 1.1, 1.0, and 1.3 s for the ECDs using KZnHCF-MS, KZnHCF-M, and KZnHCF-BS, respectively, while t B ¼ 1.3, 3.0, and 3.8 s for each ECD as shown in ESI. † The result shows that both the particle sizes and the surface modication affect the colour-switching speed. The high surface area of the small particles and low aggregation of the surface modication would increase the redox activity on the surface.
The t B ratio of KZnHCF-MS to KZnHCF-BS, and 0.34. The particle size ratio of KZnHCF-MS to KZnHCF-BS were 0.34 and 0.42, respectively. The relation between the response speed and the particle size is reasonable because the specic surface area is inverse proportional to the particle size.
The cycle stabilities of the ECDs were evaluated for specics with sequential step-like voltage change. The result shown in Fig. 4 (a) and (b) indicates that KZnHCF-MS and KZnHCF-M have longer cyclic stability than KZnHCF-BS, retaining 90% of the effective charge and 120% of the transmittance ratio to the initial values, respectively, in the oxidation reaction until one million cycles. Aer one million cycles, KZnHCF-MS shows similar transmittance in both redox states to that in the precyclic test, as shown in Fig. 4(c).
Finally, the thermal inuence on the ECD device was evaluated. An ECD of KZnHCF-MS was fabricated in same method of cycle test. The ECD in colour phase was kept in applying V ¼ À1.3 V at 80 C with evaluation of UV-vis. The absorbance at 680 nm was deceased from 0.85 to 0.52 for 21 hours as shown in ESI. † The result shows temperature has large inuence on the memory effect of the ECD.
In summary, we developed the well-dened K 2 Zn 3 [-Fe(CN) 6 ] 2 $7H 2 O (KZnHCF)-NPs using a micromixer. The KZnHCF-NPs are transparent in both redox states and have high cyclic stability in the electrochemical redox reactions even aer 1 million cycles. The non-change of the main framework of Zn-NC-Fe might be attributed to the cyclic stability of KZnHCF. Thus, it is a suitable candidate for application as a counter electrode in transparent-type ECDs.

Conflicts of interest
There are no conicts to declare.