Mechano-catalytic overall water-splitting into hydrogen and oxygen on some metal oxides

Applied Energy 67 (2000) 159±179

www.elsevier.com/locate/apenergy

Mechano-catalytic overall water-splitting into hydrogen and oxygen on some metal oxides Kazunari Domen *,1, Shigeru Ikeda, Tsuyoshi Takata, Akira Tanaka 2, Michikazu Hara, Junko N. Kondo Research Laboratory of Resources Utilization, Tokyo Institute of Technology, 4259 Nagatsuta, Midori-ku, Yokohama 226-8503, Japan

Abstract This is a novel way to generate hydrogen from pure water by converting mechanical energy to chemical energy. The phenomenological aspects of the mechano-catalytic overall watersplitting are reviewed, and the basic feature of the catalytic process is clearly established. The estimation of the conversion from mechanical to chemical energy is also shown. Experimental results, carried out to reveal the mechanism of the mechano-catalytic reaction are described. # 2000 Elsevier Science Ltd. All rights reserved.

1. Introduction The authors have been working, for more than two decades, to develop photocatalytic systems to accomplish overall water-splitting based on metal oxides [1±21]. During such experiments under visible light irradiation, it was accidentally found that some Cu(I)-containing oxides evolved H2 and O2 simultaneously from distilled water [22]. So, it was ®rst believed that the reaction was driven by the energy of visible-light photons. When further studies were conducted on the system, however, some unusual aspects judging from the conventional reaction mechanism of semiconductor-based photocatalytic reactions were found. For example, H2 and O2 evolutions continued for a long period even after the irradiation was ®nished. We at ®rst tried to interpret such a phenomenon as follows: photoexcited electrons and * Corresponding author. Tel.: +81-45±924-5238; fax: +81-45-924-5282. E-mail address: [email protected] (K. Domen). 1 CREST, JST (Japan Science and Technology). 2 Nikon Corporation, 1-10-1 Asamizodai, Sagamihara 228-0828, Japan. 0306-2619/00/$ - see front matter # 2000 Elsevier Science Ltd. All rights reserved. PII: S0306-2619(00)00012-X

160

K. Domen et al. / Applied Energy 67 (2000) 159±179

holes reacted with the components of the oxide: in other words, the excited states were localized to form chemical species such as Cu(0) and Oÿ species. Then, the chemical species gradually reacted with water to form H2 and O2. If this mechanism was correct, we might be able to obtain some H2 and O2 after the irradiation was stopped. But the amounts should be much smaller than that of the catalyst employed, because no signi®cant changes of the catalysts before and after the reaction were observed by several spectroscopic methods. To our surprise, however, in carefully controlled reactions the H2 and O2 evolution continued in the dark and the total molar amount of evolved gases far exceeded that of the catalyst. On the other hand, when magnetic stirring of the catalyst was stopped, no evolution of H2 and O2 occurred. Thus, we realized that the H2 and O2 evolution is not due to a photocatalytic reaction, but to some other mechanism that had been missed [23]. From various kinds of experimental results that will be described below, we named this somewhat curious reaction as ``mechano-catalytic'' overall water splitting. In this reaction, mechanical energy is converted into chemical energy to form H2 from water. In this paper, phenomenological aspects of ``mechano-catalytic'' overall water splitting will be reviewed. The reaction mechanism is still not fully understood, but several possibilities will be discussed based on the experimental results [23±27]. 2. Experimental 2.1. Materials Powders of simple oxides were purchased from either Kanto Chemical Co. or Wako Purechemical Co. and used as received. The NiO powder, for example, a high purity reagent (99.9%), had the BET surface area of 1 m2 gÿ1 and the particle size of about 0.5 mm. As the reproducibility of the Cu2O activity was not good when commercial samples were used, Cu2O was prepared according to the procedure as follows: Cu2O powder was prepared by the reduction of Fehling's solution. This was prepared by boiling a mixture of 1.0 M aqueous CuSO4 solution (50 cm3), 1.3 M aqueous potassium sodium tartrate (KNaC4H4O6) solution (50 cm3) and 18.8 M aqueous NaOH solution (50 cm3). Cu2O was precipitated by adding 0.5 M aqueous D-glucose solution (25 cm3) to boiling Fehling's solution with vigorous stirring under a Ar ¯ow. Cu2O, as a red precipitate, was washed with distilled water (200 cm3) for 5±7 times followed by decantation and drying in vacuo. The particle size and surface area of Cu2O were estimated to be 4±6 mm and 2.5 m2 gÿ1 by means of a scanning electron microscope (SEM ) and BET measurements, respectively. When Na®oncoated Cu2O was used, it was prepared as follows: 0.5±8.0 wt% of Na®on was deposited onto the Cu2O by evaporating a solvent from a mixture of 1 g of Cu2O, Na®on and 15 cm3 of 2-propanol in a rotary evaporator at room temperature. Na®on (5 wt% solution in a mixture of lower aliphatic alcohols and water) was obtained from the Aldrich Chemical Company, Inc.

K. Domen et al. / Applied Energy 67 (2000) 159±179

161

The mixed oxides such as CuMO2 (M=Fe, Ga, Al, Cr, La, Y) were prepared by a solid-state reaction. The stoichiometric amount of M2O3 and Cu2O (purchased) was mixed in a mortar and the mixture was calcined at 1050 C in N2 for 36 h. In the case of the preparation of CuAlO2, calcination was carried out in air at the same temperature. The crystal structure was con®rmed by X-ray di€raction (XRD). From the Cu 2p3/2 X-ray photoelectron spectra (XPS), Cu existed dominantly as Cu1+ as expected from the formula. The particles' dimensions were estimated to be 110 mm in diameter from the scanning electron microscopy (SEM). 2.2. Reaction system The mechano-catalytic reaction was carried out in a ¯at-bottomed vessel made of Pyrex glass. Typically, 0.1 g of metal oxide powder was suspended by magnetic stirring (F-205, Tokyo Garasu Kikai) in 200 cm3 of distilled water. The stirring rod was sealed by PTFE (polytetra¯uoroethylene, Te¯on1). A closed gas-circulation and evacuation system made of Pyrex glass was connected to the reaction vessel and the evolved gases were collected in it for a gas chromatographic analysis without any contamination by air (Fig. 1). The gas phase was evacuated to remove N2 and O2 prior to the reactions and to let only water vapor remain. When performing the dark reaction, the reaction vessel was completely covered with aluminum foil and when performing the reaction under photo-irradiation, a Xe lamp of 300 W placed at the side of the reaction vessel was used.

Fig. 1. A schematic view of the apparatus. The reaction was carried out in a ¯at-bottomed reaction cell made of Pyrex glass with 0.1 g of oxide powder and 200 cm3 of distilled water, which were magnetically stirred [F-205, Tokyo Garasu Kikai, stirring rate 1500 rpm, triangular prism-type stirrring rod (8334 mm) sealed by PTFE (Te¯on)] in most of the experiments. A closed gas-circulation and evacuation system made of Pyrex glass was connected to the reaction cell and the evolved gases were accumulated in it (470 cm3) for a gas chromatographic analysis.

162

K. Domen et al. / Applied Energy 67 (2000) 159±179

3. Results As will be shown below, the mechano-catalytic overall reaction is driven solely by mechanical energy and no photo-irradiation e€ect is observed for all the mechanocatalysts except Cu2O. In the case of Cu2O, the band-gap irradiation a€ects the rates of H2 and O2 evolutions to some extent, which previously led us to a misunderstanding of the reaction mechanism. It is now obvious that, with Cu2O, no photon energy is converted into chemical energy even under band-gap irradiation. Therefore, we will focus our attention on the mechano-catalytic reactions without any irradiation. 3.1. Mechano-catalytic activities of simple oxides The 'mechano-catalytic' activities of simple oxides for overall water splitting are summarized in Table 1 [25]. All reactions were carried out for 24 h and the rates shown in Table 1 are averages. NiO, Co3O4, Cu2O and Fe3O4 are seen to exhibit profound activities. It should be noted that CuO, FeO, Fe2O3, and CoO did not evolve both H2 and O2, although small amounts of H2 were detected in the latter three oxides. It was also con®rmed that RuO2 and IrO2 had much lower yet de®nite activities for the stoichiometric evolution of H2 and O2. Cr2O3 also evolved both H2 and O2, but the amount of O2 was much less than the stoichiometric one. The reason why in some cases only H2 or O2 evolved is not clear, but the redox reaction of the Table 1 Activities of various oxides upon the water-splitting (mmol hÿ1) a,b Oxide

H2

O2

Oxide

H2

O2

Oxide

H2

O2

Cr2O3 MnO Mn3O4 MnO2 FeO Fe2O3 Fe3O4 CoO Co3O4 NiO CuO Cu2O ZnO Sc2O3 MgO Y2O3

1.0 0 0 0 0.5 0.02 1.68 0.3 44.2 46.0 0 5.7 0 0 0 0

0.001 0 0.01 0 0 0 0.97 0 22.5 22.7 0 3.7 0 0 0 0

TiO2 ZrO2 V2O5 Nb2O5 Ta2O5 MoO3 WO3 RuO2 Rh2O3 IrO2 PdO Ag2O CdO Al2O3 Ga2O3 In2O3

0 0 0.06 0 0 0 0 0.1 0 00.22 0 0 0.05 0 0.04 0

0 0 0 0 0 0 0 0.05 0 0.07 0.2 0 0 0 0 0

SiO2 SnO2 SnO PbO Pb3O4 PbO2 Bi2O3 La2O3 CeO2 Pr5O11 Nd2O3 Sm2O3 Dy2O3 Ho2O3 Er2O3 Tm2O3

0 0 0.006 0 0 0 0 0.1 0 0 0 0 0 0 0 0

0 0 0 0 0.1 0.3 0.007 0 0 0 0 0 0 0 0 0

a Each reaction was continued for 24 h and the total amounts of evolved H2 and O2 were divided by 24 to give the stated values. b All the reactions were carried out using a ¯at-bottomed vessel with 0.1 g of oxide powders and 200 cm3 of distilled water, which was magnetically stirred. A triangular prism-type stirring rod (8334 mm), sealed by PTFE, was used (see Fig. 1). Catalyst 0.1 g, H2O 200 cm3, stirring rate: 1500 rpm.

K. Domen et al. / Applied Energy 67 (2000) 159±179

163

metal component may be responsible for such behavior. Whether this kind of oxide works as a mechano-catalyst after a long-term reaction has not been examined. On the other hand, TiO2, ZnO and WO3, which are widely used as photocatalytic materials, were completely inert. 3.2. Stoichiometric H2 and O2 evolutions on NiO Fig. 2 shows the typical time courses of H2 and O2 evolutions on NiO [23]. The reaction system was evacuated at 10±15 h intervals. Stoichiometric evolutions of H2 and O2 were observed, while the rates of H2 and O2 evolution decreased with the accumulation of the evolved gas, probably due to the e€ect of the gas phase pressure. A similar decrease of the activity was observed when Ar, H2 or O2 was introduced into the gas phase of the reaction system. There was no noticeable dependence on the kind of gas so suggesting that the possibility of the reverse reaction (H2+1/2O2!H2O) was excluded. Actually, no reverse reaction was observed when the NiO powder was stirred in the presence of H2 (20 kPa) and O2 (20 kPa) in the gas phase. Almost the same time courses were reproduced in the subsequent runs after evacuation of the gas phase, although the rates of H2 and O2 evolutions slightly decreased during successive runs. The total amounts of evolved H2 and O2 reached 1700 and 840 mmol, respectively, after 5 runs, while the amount of the used NiO was 1300 mmol (0.1 g). Thus, the amounts of evolved H2 and O2 exceeded that of the used NiO, so implying that the reaction proceeded catalytically. Similar time courses were also observed when Co3O4, Cu2O, or Fe3O4 were used (results not shown). Inductively coupled plasma (ICP) and pH measurements of the

Fig. 2. Time courses of H2 (*) and O2 (*) evolution from NiO suspended in a distilled water system. The gas phase was evacuated at 10 h after the start-up (run 1) and at 15 h intervals in subsequent runs. The molar amounts of evolved H2 and O2 after 5 runs reached 1700 and 840 mmol, respectively. Catalyst (NiO): 0.1 g, H2O: 200 cm3, stirring rate: 1500rpm.

164

K. Domen et al. / Applied Energy 67 (2000) 159±179

aqueous solution showed that there was no appreciable dissolution of the oxides or change of the pH in suspension during the reaction. 3.3. Mechano-catalytic overall water splitting on CuMO2 [22] The observations shown in Table 1 suggests that the reaction proceeds on the oxides that contain some special elements in speci®c oxidation states. This agrees with our ®nding that Cu(I) containing delafossites, CuMO2 (M=Al, Ga, Fe), were also active during the reaction. Ternary Cu(I)-containing oxides of the general formula Cu14M3+O2 are crystallized in the delafossite structure. The schematic structure is shown in Fig. 3. The MO6 octahedra share edges to form a triangular plane. The monovalent Cu ions are linearly coordinated with two oxygen ions on the surface of the MO6 layer. Fig. 4 shows the time courses of the H2 and O2 evolutions on CuFeO2. During the ®rst 20 h, the reaction (run 1) was carried out under visible light irradiation (>420 nm) and then the reaction (run 2) was continued in the dark for the subsequent 20 h after evacuation. H2 and O2 evolved in a stoichiometric ratio in both runs 1 and 2, and there was no appreciable di€erence in activities between runs 1 and 2. This con®rms that there is no photo-irradiation e€ect on the mechanocatalytic overall water-splitting except with Cu2O. On the other hand, H2 and O2 evolutions were not observed in run 3 under visible light irradiation without stirring. The results indicate

Fig. 3. Schematic structure of the delafossite oxide, CuMO2. The structure belongs to the trigonal system with cell parameters ah=3.035 AÊ and c=17.066 (i.e. CuFeO2) in the hexagonal description.

K. Domen et al. / Applied Energy 67 (2000) 159±179

165

that the reaction occurs by stirring (mechanical e€ect), and the light irradiation (photon energy) has no e€ect on the reaction. To examine the speci®city of the component, the activities of overall water splitting on several Cu(I)-containing delafossite oxides in dark conditions were examined and the results are summarized in Table 2. Stoichiometric H2 and O2 evolution was observed on CuFeO2, CuGaO2, and CuAlO2 and other delafossite oxides did not exhibit the activity. As shown in Table 2, Cu2O is an active material for the reaction, while Fe2O3, Ga2O3 and Al2O3 did not show the activity. CuO is also an inert material. Cu(I) is, therefore, essential for the reaction on these delafossite oxides. As shown in Table 2, however, M(III) cations in the octahedral plane also in¯uence the activity. This

Fig. 4. Time histories of H2 and O2 evolutions on CuFeO2 under visible light (>420 nm) irradiation and in the dark. The closed gas circulation system was evacuated about every 20 h. The rates of the rotation of the stirring rod were 1500 rpm for runs 1 and 2 and 0 rpm for run 3. Catalyst: 0.21 g; H2O: 200 cm3; light source: Xe lamp (300 W).

Table 2 Activity for overall water-splitting on some delafossite oxides CuM(III)O2 (delafossite)

CuAlO2 CuFeO2 CuCrO2 CuGaO2 CuLaO2 CuYO2 CuEuO2

Rate of evolved gases (mmol hÿ1) H2

O2

10.95 3.08 0.07 13.36 0 0.01 0.02

5.10 1.56 0 6.89 0 0 0

166

K. Domen et al. / Applied Energy 67 (2000) 159±179

result provides a clue to the understanding of the mechanism of mechano-catalytic overall water-splitting. Fig. 5 shows typical time courses for the H2 and O2 evolutions on the CuAlO2 catalyst (0.1 g=820 mmol was used). The reaction was continued for about 500 h with evacuation at about 250 h after the start of the ®rst run. The observed time course resembled that for NiO (see Fig. 2), and it was again con®rmed that the molar amounts of the evolved H2 and O2 after the reaction for 500 h de®nitely exceeded the amount of the used CuAlO2. 3.4. Mechanical energy responsible for the reaction Fig. 6 shows the relationship between the revolution rate (rpm) of the stirring rod and the rates of H2 and O2 evolution on NiO. H2 and O2 evolutions were not observed without stirring. The stirring rod was rotated without any supply of such energy as light or heat, but the rates of H2 and O2 evolutions increased monotonically with the revolution rate. It again con®rms that the reaction proceeded as a result of the mechanical energy imparted by the stirring rod. The role of the magnetic ®eld was examined during an experiment in which an electric motor was used to rotate a stirring rod made of alumina (10 mm of radius): 0.01 g of NiO powder was attached to the ¯at bottom of the stirring rod that was pressed to the bottom of the reaction cell. As shown in Fig. 7, simultaneous H2 and

Fig. 5. Time courses of H2 (*) and O2 (*) evolutions for the CuAlO2 suspension. The gas phase was evacuated for 250 h after start-up. The molar amounts of evolved H2 and O2 (about 5000 and 2500 mol, respectively) exceeded the amount of CuAlO2 used (820 mmol). Catalyst (CuAlO2): 0.1 g; H2O: 200 cm3.

K. Domen et al. / Applied Energy 67 (2000) 159±179

167

Fig. 6. The rates of H2 (*) and O2 (*) evolutions for the NiO-suspended system as functions of the revolution rate (rpm) of the stirring rod. The rates of H2 and O2 evolutions monotonically increased with increasing the revolution rate of the stirring rod. Catalyst (NiO): 0.1 g; H2O: 200 cm3.

Fig. 7. Time courses of H2 and O2 evolutions in the reaction performed under the free-of-magnetic-®eld condition. Schematic view of the motor-drive type reaction cell is shown in the inset. 0.01 g of NiO powder was stuck to the ¯at bottom of the stirring rod made of alumina. Simultaneous H2 and O2 evolutions were observed. H2O: 200 cm3; stirring rate: 1500 rpm.

168

K. Domen et al. / Applied Energy 67 (2000) 159±179

O2 evolutions were observed again so proving that a magnetic ®eld is not responsible for the mechano-catalytic reaction. Similar results were obtained when other active oxides were used. Mechanical energy injected into the reaction system via stirring may be classi®ed into (1) the energy consumed for stirring the aqueous suspension and (2) the energy deposited by scraping at the interface between the rotating stirring rod and the bottom wall of the reaction vessel. In order to determine which mechanical energy leads to the mechano-catalytic reaction, the relationship between the reaction and the stirring manner was examined using di€erent shaped stirring rods. The shapes of the tested two stirring rods and the time courses of the H2 and O2 evolutions are depicted in Fig. 8. In the ®rst experiment, 0.1 g of NiO powder was suspended in distilled water by rotating the stirring rod (a) that was kept from contact with the bottom wall of the reaction vessel and there were no evolutions of H2 and O2 (Fig. 8a). In the second experiment, ca. 0.01 g of NiO was stuck by a double-sided adhesive tape onto the bottom face of the stirring rod (b) with a ¯at bottom and the rod was rotated on the bottom of the reaction vessel in distilled water without any suspended NiO powder (Fig. 8b). By using the suspension system similar to that shown in Fig. 1, a stoichiometric evolutions of H2 and O2 were observed. These results proved that the reaction was driven by the mechanical energy consumed at the interface between the rotating stirring rod and the bottom of the reaction vessel. Moreover, it is indicated that the collision among the oxide particles or at the stirrer surface is not responsible for the reaction. Additionally, the results precluded the possibility that the Te¯on

Fig. 8. Changes by stirring manner of the rates of H2 (*) and O2 (*) evolutions. (a) Stirring NiO (0.1 g)suspended in distilled water with a ¯oating type stirring rod, the stirring rod was rotated without contact with the bottom of the reaction vessel. There was no evolution of H2 and O2. (b) Both-sided adhesive tape was stuck on the ¯at bottom of the stirring rod and ca. 0.01 g of NiO powder was stuck onto the other side of the tape. Stirring the NiO (ca. 0.01 g)-glued-¯at-bottomed rod in suspension-free water. Similar to the suspension system shown in Fig. 2. Simultaneous H2 and O2 evolutions were observed. H2O: 200 cm3; stirring rate: 1500 rpm.

K. Domen et al. / Applied Energy 67 (2000) 159±179

169

®lm that was coated over the stirring rod took part in the reaction, because in the second experiment no Te¯on but adhesive tape contacted with NiO powder was used. As a whole, the simultaneous H2 and O2 evolutions is the ``mechano-catalytic'' overall water splitting in which mechanical energy supplied by the ``rubbing'' of these oxide powders on the bottom of the reaction vessel is converted to chemical energy with the help of these oxide powders activity as catalytic materials. 3.5. Eciency of mechanical-to-chemical energy conversion Estimation of the eciency of mechanical-to-chemical energy conversion was carried out as follows. We denote by Ec, Et, Ei, and Z the chemical energy accumulated for 1 h by water decomposition, the total mechanical energy provided into the system for 1 h for stirring, the mechanical energy consumed at the interface for 1 h between the rotating rod and the bottom of the reaction vessel, and conversion eciency from Ei to Ec, respectively, and express Ec, Et, and Ei in kJ hÿ1. Ec was obtained from the measured rate of water decomposition. Estimations of Et and Ei were made as follows. The stirring rod, mounted at the end of a driving shaft, was rotated by an electric motor to stir the suspension (catalyst: 0.5 g, H2O:200 cm3) in the reaction vessel (see Fig. 9). The stirring rod was pressed against the bottom wall of the reaction vessel with the force of 1.9 N, which was the same as that exerted by

Fig. 9. Schematic views of the apparatus used for estimation of the eciency of mechanical-to-chemical energy conversion: (A) for Et (under rubbing) and (B) for Et, (without rubbing) (see text).

170

K. Domen et al. / Applied Energy 67 (2000) 159±179

the magnetic stirrer. Et was then derived by measuring the torque and the rate of revolution according to the following equation: ÿ
Et kJ hÿ1 ˆ 2  …revolution rate …rpm†=60…s††  torque …Nm†  3600…s† Ei, which is the mechanical energy responsible for the reaction, was obtained by subtracting from Et the energy, Et0 , derived by releasing the pressing force of 1.9 N [i. e. Ei=Et (under rubbing)- Et0 (without rubbing)]. The energy for such free stirring of the suspension was determined by measuring the torque when the stirring rod was rotated in the suspension without touching the bottom of the reaction vessel. The conversion eciency, Z, was then obtained by the following equation. Z…%† ˆ Ec =Ei  100 According to the de®nition and the method described above, Ec and Ei0 were estimated as follows when we used with data of NiO as described in Table 1. Ec ˆ 46 mmol hÿ1 ˆ 0:011 kJ hÿ1 Ei ˆ 0:83 kJ hÿ1 Z ˆ 0:011=0:83  100 ˆ 1:3…%† When a larger stirring rod with a triangular prism-shape (13354 mm) was used with the NiO catalyst, a higher eciency was obtained as follows: Ec ˆ 200 mmol hÿ1 ˆ 0:047 kJ hÿ1 Ei ˆ 1:10 kJ hÿ1 Z ˆ 0:047=1:10  100 ˆ 4:3…%†

3.6. E€ect of aqueous solution So far, the H2 and O2 evolutions from the distilled water have been described. The e€ect of the reaction solution was next examined in order to obtain information concerning the reaction mechanism. It is well known that, in the case of photocatalytic decomposition of water, O2 evolution is completely suppressed in an aqueous methanol solution, because the intermediate species of water oxidation, i.e. OH.radicals and/or Oÿ ions, preferentially attack the methanol molecules which are oxidized eventually into CO2. Fig. 10 shows time courses for H2 and O2 evolutions on NiO from an aqueous methanol solution (10 vol%). In contrast to photocatalytic reactions, both H2 and O2 were evolved in the present system although the amount of O2 was a little less than expected from the stoichiometric ratio. The result suggests that the formation of O2 molecules by mechano-catalysis does not proceed through these intermediate species formed in the bulk of the aqueous solution.

K. Domen et al. / Applied Energy 67 (2000) 159±179

171

Fig. 10. Time courses for H2 (*) and O2 (*) evolutions from the NiO-suspended aqueous methanol solution. Almost stoichiometric H2 and O2 evolutions were observed. Catalyst (NiO): 0.1 g, 10% methanol aq.: 200 cm3; stirring rate: 1500rpm.

When pure methanol was used, H2 accompanied by CO and a small amount of CH4 evolutions were observed. The rate of H2 evolution was almost comparable with that for pure water. No O2 evolution was observed in pure methanol solution. To examine the contribution of electrolytes in the aqueous solution, the reactions in the presence of some electrolytes were carried out. Fig. 11 shows the typical dependence of the mechano-catalytic activity of NiO on the concentration of NaCl in a reaction mixture. The rates of H2 and O2 evolutions monotonically decreased with increasing concentration of NaCl. Similar results were observed in the presence of other electrolytes such as Na2SO4, NaNO3, KCl etc. as shown in Table 3, and also on Co3O4, Cu2O and Fe3O4. These results indicate that the addition of any electrolyte to the aqueous solution retards the rate of mechano-catalytic overall water splitting. 3.7. Materials of the bottom of the reaction vessel From the experimental results described above, it has become clear that the rubbing of the mechano-catalyst's powder on the bottom of the reaction vessel is essential. All the data presented so far were obtained in the reaction vessel of Pyrex glass. The reaction was also examined by employing other materials at the bottom of the reaction vessel as shown in Table 4. In the case of a quartz glass plate, stoichiometric evolutions of H2 and O2 were observed, but the rate of the reaction was about a half of that using a Pyrex glass

172

K. Domen et al. / Applied Energy 67 (2000) 159±179

plate. When Al2O3 (sapphire) and CaF2 plates were used, both H2 and O2 evolutions were again con®rmed although the rates decreased by more than an order of magnitude. When PTFE or acrylate board was used, a small amount of H2 but no O2 were detected within 20 h.

Fig. 11. Dependence on the concentration of NaCl of the rates of H2 (*) and O2 (*) evolutions NiOwater system. With increasing concentrations of NaCl, the rates of H2 and O2 evolution monotonically decreased. Catalyst (NiO): 0.1g, H2O: 200 cm3; stirring rate: 1500 rpm. Table 3 Dependence of the activity for overall water-splitting upon the kinds of employed electrolytes in H2Oa Rate of evolved gases (mmol hÿ1)b

Conductivity

Electrolyte

Concd (mol dmÿ3)

H2

O2

pH

(25 C, 18 C) ( ÿ1 cmÿ1)

None NaCl

± 0.001 0.05 0.01 0.001 0.0005 0.001 0.001 0.001 0.001 0.001 0.001 0.0005

42.0 26.1 6.6 4.1 21.8 21.2 13.7 18.1 16.4 5.2 3.2 7.6 7.0

21.0 13.2 3.2 1.7 13.2 11.9 5.9 8.6 10.0 2.4 1.2 1.6 3.8

6.8 6.8 6.8 6.8 6.8 6.2 6.0 6.8 6.8 11.0 11.0 3.0 2.9

<110ÿ5 7.210ÿ3 3.510ÿ2 6.910ÿ2 1.010ÿ2 7.210ÿ3* 1.410ÿ2* 1.110ÿ2 1.410ÿ2 8.610ÿ3* 1.310ÿ2* 1.510ÿ2 2.010ÿ2

NaNO3 Na2SO4 KCl KNO3 NaOH KOH HCl H2SO4 a b

NiO 0.1 g, H2O 200 cm3; stirring rate 1500 rpm. Initial activity.

K. Domen et al. / Applied Energy 67 (2000) 159±179

173

3.8. Formation of metallic elements during mechano-catalytic overall water splitting Fig. 12 shows the XRD patterns of Cu2O and Na®on/Cu2O before and after the reaction occuring for 150 h. As will be shown below, Na®on-coating enhances the activity of mechano-catalytic overall water splitting of Cu2O although the mechanism of the enhancement is not yet clear. The XRD pattern of Na®on/Cu2O before the reaction was the same as that of the original Cu2O (Fig. 12A). After the reaction, Table 4 Dependence of the activity for overall water-splitting upon the quality of the materials at the bottom of the rection vessel Plate

Rate of evolved gases (mmol hÿ1) H2

Quartz glass 19 Pyrex glass 40 Sapphire 0.9 CaF2 0.8 PTFE 1.0 Acryl 0.8 Pyrex frosted glass 6

O2 10 20 0.5 0.3 ± ± 3

Fig. 12. XRD patterns of Cu2O before and after the reaction for 160h. A; B: Cu2O after the reaction.

174

K. Domen et al. / Applied Energy 67 (2000) 159±179

a small peak due to metallic Cu (Cu(111) di€raction) appeared as shown in Fig. 12B. The peak was observed even after the reaction for 10 h, and remained unchanged after 20 h. The XRD pattern of the Na®on/Cu2O after the reaction (not shown) exhibited a larger peak of metallic Cu than that in Fig. 12B. When Cu2O and Na®on/Cu2O were kept in distilled water for 500 h without stirring, no H2 and O2 evolved, and no peak of metallic Cu was observed for Cu2O and Na®on/Cu2O. Considering the fact that the peak of metallic Cu in Na®on/Cu2O was larger than that in Cu2O, and that Na®on/Cu2O has a higher activity for water splitting, the formation of metallic Cu may be related with the mechanism of the mechano-catalytic overall water splitting. In NiO or Co3O4 after the mechano-catalytic reaction for 50 h, neither metallic Ni or Co was observed by XPS and XRD. As the redox potentials of Ni and Co are more positive than that of Cu, metallic Ni and Co may be reoxidized by the water right after the formation. In order to examine whether NiO and Co3O4 are reduced into metals during the reaction or not, the mechano-catalytic overall water splitting on NiO and Co3O4 were examined for the KI±I2 solution. The latter is often utilized for the detection of metals in metal oxides. I2 in the KI±I2 solution, oxidizes a transition metal denoted as M such as Ni and Co as follows: M ‡ I2 ! MI2 Even in the KI±I2 solution, both H2 and O2 evolutions were observed, although the ratio was not stoichiometric. The deviation from stoichiometry was attributed to the reduction of I2 which was estimated by UV-visible absorption spectroscopy and it coincided with the shortage of evolved H2. Further, the amounts of dissolved Ni2+ and Co2+ agreed with those expected from the decrease of I2. It is noted that, without I2, no dissolutions of Ni or Co were detected and even under a H2 atmosphere, metallic Ni or Co was formed. (The full description of this experiment will be reported elsewhere.) From these experiments, it was therefore indicated that during the mechano-catalytic overall water splitting on NiO and Co3O4, metallic Ni and Co respectively were formed. Thus, the formation of metallic elements in mechano-catalysts during the reaction seems to be a common phenomenon at least for the three oxides examined so far. This, however, does not necessarily mean that metals themselves directly participate in the reaction, because the metals might be formed as by-products. In order to investigate the role of the metals, the reactions of water with metallic Cu, Ni and Co were examined. For example, 0.11 g of CuO powder (1.4 mmol) was reduced at 423 K for 24 h by H2 in the reaction vessel. The particles' mean size and surface area of metallic Cu were estimated to be 6 mm and 1.0 m2 gÿ1 repectively by SEM and BET measurements. A volume of 120 cm3 of distilled water, that was free from air, was introduced into the reaction vessel, and then the solution was stirred magnetically. Only H2 was evolved during the early stages of the reaction, but O2 evolution was also observed after the reaction for 30 h. The evolution rate of H2 increased with the beginning of O2 evolution, and the ratio of evolution rates of H2 to O2 (H2/O2) approached 2:1 with reaction time. The XP spectra and XRD patterns showed that the surface of metallic Cu is oxidized into

K. Domen et al. / Applied Energy 67 (2000) 159±179

175

Cu2O during the reaction. These results indicate that metallic Cu was oxidized into Cu2O by reducing water into H2 during the early stages of the reaction, where only H2 was evolved. When the surface of Cu particles was covered with Cu2O, O2 began to evolve with H2. In the cases of Ni and Co, we obtained similar results. Hence, it was concluded that metallic Cu, Ni and Co were oxidized into Cu2O, NiO and Co3O4, respectively, by reducing water into H2 and when these metals were covered with metal oxides, O2 evolution began to occur with the H2. 3.9. Change of particle sizes of catalysts Because mechano-catalytic overall water splitting proceeds by rubbing of the catalysts, the grinding of catalyst particles may be responsible for the reaction as in the cases of mechano-chemical or tribo-chemical reactions. The particle sizes and surface areas of Cu2O, Na®on/Cu2O, NiO and Co3O4 before and after the reaction for 24 h were examined. The particle sizes and surface areas were estimated by secondary electron microscope (SEM) and BET measurements, respectively. Each particle was ground to smaller ones during the reaction for 24 h, while the particles were not further ground during the subsequent reaction for 100 h. These results imply that the grinding of the metal oxide particles are completed at an early stage of the reaction and are not directly related to the reaction because the rates of H2 and O2 evolution on the metal oxides were stable for more than 100 h. 3.10. Origin of the evolved O2 This was examined by using H216O/H218O mixture. The reaction was carried out with 0.01 g of CuAlO2 and a mixture of H216O and H218O (H216O/H218O=4.7), which was stirred magnetically. After the stoichiometric H2 and O2 evolution was con®rmed by gas chromatography, the evolved O2 species were analyzed by mass spectral analysis. Fig. 13 shows a time course of the isotope composition in the evolved O2 on CuAlO2 from the H216O/H18 2 O mixture. At the beginning stage of the reaction for 1 h, the ratio, 16O2/16O18O/18O2, in the evolved O2 was estimated to be 22.0/9.5/ 1.0. Therefore, the atomic ratio, 16O/18O, in the total amount of evolved O2 is 4.7/ 1.0, which exactly coincided with that in the water. The 16O/18O ratio did not change in subsequent measurements up to 76 h. The same results were also obtained for NiO, Co3O4, Cu2O, and Fe3O4. 3.11. On the mechanism of mechano-catalytic overall water-splitting The standard Gibbs free-energy change,
G , of overall water splitting is very large; for liquid water at 298 K, it is 237 kJmolÿ1, and for water vapor at 298, 1000 and 2000 K, it is 229, 193 and 135 kJ molÿ1, respectively. This indicates that the equilibrium pressures of H2 with H2O(g) of 1 bar are 310ÿ27 and 510ÿ4 bar at 298 and 2000 K, respectively. In the present experiment, more than 0.1 bar of H2 was accumulated under 0.04 bar of H2O(g) Ð i.e. the vapor pressure of H2O at 298 K (see Fig. 4). Although the increase of the temperature at the interface between a

176

K. Domen et al. / Applied Energy 67 (2000) 159±179

Fig. 13. Time courses of isotope composition in evolved O2 on CuFeO2 from H216O/H218O mixture. Catalyst: 0.01 g, H2O (H216O/H218O=4.7): 4 cm3; stirring rate: 1500rpm.

rotating stirring rod and the bottom of the reaction vessel is expected to some extent, a temperature above 2000 K is extremely unlikely. It is, therefore, obviously concluded that the H2 and O2 evolutions are not due to a thermally-driven reaction. Mechano-catalytic overall water-splitting must proceed by some other process than the heat of friction. In mechano-chemistry, some reactions with
G>0 (e.g. Au +3/4CO2!1/2Au2O3+3/4C,
G =377 kJ.molÿ1) are known to proceed using mechanical energy. The activation processes of these reactions are generally attributed to the grinding or frictional wear of materials: the reactions are classi®ed into stoichiometric and irreversible ones. As mentioned above, in the cases of NiO, Cu2O, Co3O4 and Fe3O4, the particle sizes of those oxides became smaller at the start of the reaction but they remained almost unchanged thereafter, while the reaction proceeded steadily for a long time. These results indicate that the grinding or frictional wear is not really essential for the mechano-catalytic overall watersplitting. ``Mechano-catalytic'' overall water-splitting is one of the catalytic reactions, which are di€erent from the mechanochemical, and/or tribochemical reactions, and the molar amounts of evolved H2 and O2 far exceed that of the catalyst. Reversible mechanical e€ects without any wear of the materials have to be considered in interpreting the reaction mechanism of the mechano-catalytic overall water-splitting. One possible explanation is the so-called ``tribo-electricity'' phenomenon, which is not fully understood even now. It is known that electrostatic charge separation can take place easily in vacuum or under a gas-phase condition when two materials are frictionally in contact. As a result, the oxide powder may be negatively charged and the bottom of the reaction vessel may be positively charged. These tribo-electrically-generated charges may be able to cause the redox reactions

K. Domen et al. / Applied Energy 67 (2000) 159±179

177

resulting in overall water-splitting. In our preliminary experiments, photoemission was actually observed from the interface where the stirring rod is rotating. This may suggest such electrostatic charge separation can take place even in an aqueous solution. As mentioned in the preceding section, the rates of H2 and O2 evolutions decreased when some electrolytes were added to the distilled water. This seems to indicate that the charge separation at the interface becomes more dicult in an aqueous solution containing electrolytes. Although, at present, tribo-electricity is the most probable candidate for the ®rst step of the mechano-catalytic overall water splitting, there are several experimental results that seem to be in con¯ict with to this model. First, no evolutions of H2 and O2 were observed when we carried out the reaction in water vapor. The gas-phase reaction seems to be more facile for the tribo-electric charge separation. Secondly, if tribo-electricity is responsible for the mechano-catalysis, the eciency of conversion from mechanical-to-chemical energy seems to be very high. As shown above, the estimated value was as high as 4.3%. As we do not know how the maximum eciency of tribo-electricity is estimated, this value might be reasonable. Also what is the ultimate eciency of this reaction? Thirdly, only a limited number of oxides can eciently decompose water into H2 and O2 and most oxides are inert. If charge separation is essential for the reaction, various kinds of materials, especially insulating ones such as MgO, Al2O3, and ZrO2 are expected to be active. This suggests that there is another essential role of the oxide powders for the accomplishment of the overall water-splitting by the mechano-catalysis.The fact that only a few oxides, with some speci®c oxidation states, are active for the reaction suggests that these materials have some kind of catalytic activity for overall water splitting which is driven by tribo-electricity. As shown in Table 1, Fe3O4, Co3O4, NiO, Cu2O, RuO2 and IrO2 evolved H2 and O2 in a stoichiometric ratio. Some of them (Co3O4, RuO2 and IrO2) are known to be good catalysts for O2 evolution in water oxidation, but Cu2O and Fe3O4 seem to be less e€ective. Co3O4, NiO, RuO2 and IrO2 are expected to work as good catalysts for water reduction to form H2, but again Fe3O4 and Cu2O are not. Therefore, a simple explanation from the viewpoint of catalytic activities for water oxidation and reduction is not appropriate. Furthermore, O2 evolution from a methanol aqueous solution indicates that at least a water-oxidation process does not proceed through a reaction mechanism similar to the photo-catalytic oxidation of water. One of the possible reaction mechanisms for H2 and O2 evolution is based on the redox reaction of the oxide materials themselves. In our experiments, the formations of metallic Cu, Ni and Co were indicated during the mechano-catalytic reaction on Cu2O, NiO and Co3O4. When the metallic Cu, Ni and Co were used, under the condition of the mechanocatalytic reaction, H2 evolution was also con®rmed. Therefore, if the oxides are reduced to the metallic states, it is possible to reduce water into H2 although the reduction mechanism of the oxides is unclear. Other observations worth noting when considering the reaction mechanism are the results obtained in the experiments using H218O as shown in Fig. 13. The ratio of the isotopes of 18O /16O in evolved O2 was coincident with that for water. One of the interpretations of this result is that no

178

K. Domen et al. / Applied Energy 67 (2000) 159±179

oxygen in the oxide is incorporated in the catalytic cycle of water splitting. Another possibility is that, although the redox reaction of the oxide is responsible for the water splitting, the amount of oxygen of the oxide, which is incorporated in the reaction, is very small so that the ratio of the isotopes appears to be the same as that of water. If some of the oxygen in the oxide and that in water exchange very quickly under the mechano-catalytic condition, and only such oxygen atoms are incorporated in the redox reaction, then we again obtain the same isotope ratio. At present, we do not have any plausible model to explain all the observed results satisfactorily. The detailed mechanistic properties of the reaction are still under investigation. In this review article, we tried to describe the phenomenological aspects as much as possible. At present, our interests in this new and somewhat curious reaction To answer two questions, namely what is the reaction mechanism and what will be the ultimate eciency of mechano-catalytic overall water-splitting? Acknowledgements This work was supported by CREST of JST (Japan Science and the Technology) and Research Institute of Innovative Technology for the Earth (RITE). References [1] [2] [3] [4] [5] [6] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20]

Domen K, Naito S, Soma M, Onishi T, Tamaru K. J Chem Soc Chem Commun, 543 1980. Domen K, Naito S, Onishi T, Tamaru K, Soma M. J Phys Chem 1982;86:3657. Domen K, Kudo A, Onishi T. J Phys Chem 1986;90:292. Domen K, Kudo A, Shinozaki A, Tanaka A, Maruya K, Onishi T. J Chem Soc, Chem Commun 356 (1986). Domen K, Kudo A, Shibata M, Tanaka A, Maruya K, Onishi T. J Chem Soc, Chem Commun 1706 (1986). Kudo A, Domen K, Maruya K, Onishi T. Chem Phys Lett 1987;133:517. Kudo A, Tanaka A, Domen K, Maruya K, Aika K, Onishi T. J Catal 1988;111:67. Kudo A, Tanaka A, Domen K, Onishi T. J Catal, 111, 296 (1988). Kudo A, Sayama K, Tanaka A, Asakura K, Domen K, Maruya K, Onishi T. J Catal 1989;120:337. Sayama K, Tanaka A, Domen K, Maruya K, Onishi T. Catal Lett 1990;4:217. Sayama K, Tanaka A, Domen K, Maruya K, Onishi T. J Catal 1990;124:541. Domen K, Kudo A, Tanaka A, Onishi T. Catal Today 1990;8:77. Sayama K, Tanaka A, Domen K, Maruya K, Onishi T. J Phys Chem 1991;95:1345. Domen K, Ebina Y, Sekine T, Tanaka A, Kondo J, Hirose C. Catal Today 1993;16:479. Domen K, Ebina Y, Kondo J. Res Chem Intermed 1994;20:895. Takata T, Furumi Y, Shinohara K, Tanaka A, Hara M, Kondo JN, Domen K. Chem Mater 1997;9:1063. Takata T, Shinohara K, Tanaka A, Hara M, Kondo JN, Domen K. J Photochem Photobiol, A: Chemistry 1997;106:45. Ikeda S, Hara M, Kondo JN, Domen K, Takahashi H, Okubo T, Kakihana M. Chem Mater 1998;10:72. Ikeda S, Hara M, Kondo JN, Domen K, Takahashi H, Okubo T, Kakihana M. J Mater Res 1998;13:852.

K. Domen et al. / Applied Energy 67 (2000) 159±179

179

[21] Takata T, Tanaka A, Hara M, Kondo JN, Domen K. Catal Today 1998;44:17. [22] Hara M, Kondo T, Komoda M, Ikeda S, Shinohara K, Tanaka A, Kondo JN, Domen K. Chem Commun 1998;357. [23] Ikeda S, Takata T, Kondo T, Hitoki G, Hara M, Kondo JN, Domen K, Hosono H, Kawazoe H, Tanaka A. Chem Commun 1998;2185. [24] Ikeda S, Tanaka A, Hosono H, Kawazoe H, Hara M, Kondo JN, Domen K. Overall water-splitting on Cu(I)-containing oxides, CuMO2 (M=Fe, Ga, Al) with delafossite. Stud Surf Sci Catal, 121 301± 4 1999. [25] Ikeda S, Takata T, Komoda M, Tanaka A, Hosono H, Kawazoe H, Hara M, Kondo JN, Domen K. Phys Chem Chem Phys 1999;1:4485. [26] Hara M, Hasei H, Yashima M, Ikeda S, Takata T, Kondo JN, Domen K. Investigation of mechanocatalytic overall water-splitting (II) Na®on-deposited Cu2O, Appl. Catal. A, in press. [27] Hara M, Komoda M, Hasei H, Yashima M, Ikeda S, Takata T, Kondo JN, Domen. J Phys Chem B, submitted.