Introduction

Hydrogen is considered as next generation fuel to propel airplanes, automotive vehicles and virtually any stationary power system using fuel cells. The non-polluting by-product ‘water’ upon hydrogen combustion has attracted world attention to save ever polluting earth environment for sustainable future. Currently, the hydrogen is derived from fossil fuels. The smallest molecule of universe sees highest demand due to its non-polluting end product as well as its remarkable chemical and physical properties. There are number of chemical transformation in which hydrogen is used as hydrogenating or reducing agent. Moreover, present trend to harvest CO2 into useful chemicals demands hydrogen. Many scientists around the world are pessimistic about CO2 hydrogenation since they see raising demand for hydrogen and currently there are no real alternatives to fulfill other than fossil fuels. Researchers have been looking at different possibilities to generate hydrogen by biological and chemical means. Electrolysis of water is one of the easy and greener route to generate hydrogen only if electricity comes from wind, tidal, photovoltaics, geothermal or hydropower. The other greener routes are photoelectrochemical water splitting [1], by direct splitting of water [2, 3] and solar thermochemical cycles. It is hoped that combination of several technologies can fulfill future hydrogen demand.

Water splitting by low valent metal oxides at high temperature is one of the clean way of hydrogen production since the temperature needed to perform chemical reaction comes from concentrated solar thermal heat. Though the technology is known since more than three decades commercial realization is yet to happen due to numerous challenges in this technology. The off-sun hours, cloudy and rainy seasons are main drawbacks for commercial realization. Moreover, technology cannot be implemented in geographically poor sun receiving regions.

This review summarizes the work done in high-temperature hydrogen production via two-step redox processes using various metal oxides (Table 1). Only the high-temperature experiments demonstrated either in solar furnace or in laboratory fixed-bed reactor have been considered. This review does not cover the hybrid technologies or other forms of hydrogen production technologies.

Bilgen et al. [4] have demonstrated the possibility of splitting water directly at high temperature. The theoretical calculation for the said reaction is depicted in Fig. 1. It was found that the amount of hydrogen produced decreases with increase in H2O partial pressure. Figure 1 gives the compounded results for partial pressure of water equal to 0.1 bar between 1,500 and 5,000 K. Bilgen [4] experimentally demonstrated that between 2,273 and 2,773 K formation of about 2–3 % H2 when mixture of steam and argon was passed in the crucible at the focus of the solar furnace.

Fig. 1
figure 1

Theoretical composition of the different products in the dissociation of water at high temperature; total pressure is 1 bar and partial pressure of H2O is 0.1 bar

Table 1 Summary of potential two-step water-splitting reaction systems reported in the literature

The dissociation of metal oxide to their respective metal is written as follows [14]:

$$ {\text{M}}_{x} {\text{O}}_{y} = x{\text{M}} + \frac{y}{2}{\text{O}}_{2} . $$
(1)

The temperature required for few metal oxides conversion to their metallic form is given in Table 2. Except for ZnO, achieving temperature needed to reduce metal oxide to their metallic form is practically impossible due to the high temperature required. Concentration ratios of up to 10,000 suns have been achieved by researchers which translate to 3,800 K. But high-temperature operation, reactor material thermal stability and radiation heat losses makes the process almost impossible. The temperature required to attain ∆G° of the reaction (1) equals zero can be substantially brought down by the use of hydrocarbon as reducing agents, for example graphite or methane, which can be written as follows:

$$ {\text{M}}_{x} {\text{O}}_{y} + y{\text{C}}\left( {\text{gr}} \right) \, = \, x{\text{M}} + y{\text{CO}} $$
(2)
$$ {\text{M}}_{x} {\text{O}}_{y} + y{\text{CH}}_{4} = x{\text{M}} + y(2{\text{H}}_{2} + {\text{CO}}). $$
(3)
Table 2 Approximate temperature required for which ∆G° of the reaction (1) equals zero

Two-step cyclic redox processes are simplest way of producing hydrogen by utilizing metal oxide. The solar reduction step is endothermic and can be written as shown in (4),

$$ \tfrac{1}{\Updelta \delta }{\text{M}}_{x} {\text{O}}_{{y - \delta_{ox} }} \to \tfrac{1}{\Updelta \delta }{\text{M}}_{x}^{{}} {\text{O}}_{{y - \delta_{\text{red}} }} + 0.5{\text{O}}_{2} $$
(4)
$$ \tfrac{1}{\Updelta \delta }{\text{M}}_{x}^{{}} {\text{O}}_{{y - \delta_{\text{red}} }} + {\text{H}}_{2} {\text{O}} \to \tfrac{1}{\Updelta \delta }{\text{M}}_{x} {\text{O}}_{{y - \delta_{ox} }} + {\text{H}}_{2} $$
(5)
$$ \tfrac{1}{\Updelta \delta }{\text{M}}_{x} {\text{O}}_{{y - \delta_{\text{red}} }} + {\text{CO}}_{2} \to \tfrac{1}{\Updelta \delta }{\text{M}}_{x} {\text{O}}_{{y - \delta_{\text{ox}} }} + {\text{CO}} $$
(6)

(where δ is non-stoichiometric coefficient and Δδ is change in non-stoichiometric coefficient).

The reaction (4) takes place at temperature above 1,000 K and many metal oxide systems have been studied over the past four decades. Several two- and three-step H2O-splitting thermochemical cycles based on metal oxides redox reactions have been reported in the literature. Nakamura [5] first proposed the two-step redox cycle in 1977 for Fe3O4/FeO redox cycle; interest then diminished for the next two decades and thereafter a spurt of interest resulted in investigation of several other oxide systems for thermochemical redox cycle for hydrogen generation. The high temperature required for reduction reaction can be supplied by either concentrated solar energy or fossil fuels. The solar reduction is usually carried out in the presence of an inert gas, if a reducing gas is used the reduction temperature can be brought down substantially. The reduced metal oxide can be oxidized back to the original state by oxidants like H2O or CO2. If H2O is used H2 can be produced and if CO2 is used CO can be produced as shown in Eqs. (5) and (6), respectively. If H2O and CO2 are used to oxidize the redox material alternatively or together one can produce the synthesis gas (CO + H2) from totally renewable sources (CO2 and H2O) [30].

Sibieude et al. [10] demonstrated reduction of Fe3O4 to FeO in a solar furnace by heating the material 300 °C above its melting point. They could obtain up to 40 % conversion in air and 100 % conversion in argon atmosphere. Figure 2 gives the conversion rate of Fe3O4 to FeO as a function of temperature under argon flow (20 l/h).

Fig. 2
figure 2

Conversion rate of magnetite versus temperature

As observed by many researchers, they also experienced that quenching the reduced oxides is very important. Table 3 summarized the FeO yield with various quenching rates. It can be seen that in presence of air up to 40 % conversion can be obtained with 373 K/s cooling rate.

Table 3 Influence of cooling rate on FeO yield in air and argon atmosphere

As per literature the reduction of Mn3O4 to MnO occurs above 1,773 K [31]. Sibieude et al. [10] have studied reduction of Mn3O4 to MnO in a solar furnace. They could obtain about 80 % conversion at 2,173 K under atmospheric pressure of air with a cooling rate of 373 K/s. The quenching of the MnO is very important to stop backward reaction, i.e., formation of Mn3O4 again. They have not performed more extensive work on this system.

Ehrensberger et al. [11] have studied non-stoichiometric FeMn oxides for two-step water-splitting reaction. They calculated ∆GR values for two-step Nakamura cycles FeO–Fe3O4 and MnO–Mn3O4 and the plotted results are shown in the Figs. 3, 4. Figure 3 indicates that ∆G0 equals to zero for the reduction of Mn3O4 to MnO is at least 500 K less than that of Fe3O4 to FeO system. However, Fig. 4 indicates that FeO can produce hydrogen between 873 and 1,073 K but MnO system is unable to produce hydrogen in significant levels. This led the authors to think of the possibility of combining Fe and Mn oxides to reduce spinel at lower temperature as well as produce hydrogen in significant amount in oxidation step. Authors demonstrated the oxidation of Fe1−yO and (Fe1−xMn x )1−yO (x ≤ 0.3) in a laboratory tubular furnace and monitored gaseous products using mass spectrometer. At atmospheric pressure, water with a partial pressure of about 4,200 Pa in nitrogen was able to oxidize (Fe1−xMn x )1−yO (x = 0.0, 0.1, 0.3) to (Fe1−xMnx)3O4 with x′ < x forming molecular hydrogen. The substitution of iron with 10–30 % Mn in the wuestite phase did not lower the total amount of hydrogen formed, but it changed the kinetics of the process significantly. It was also found that the process is thermodynamically controlled at high temperature. The rate of water splitting decreased with increase in manganese concentration.

Fig. 3
figure 3

Gibbs free enthalpy ∆GR for the decomposition of Fe3O4 (A, B) and Mn3O4 (C) to FeO (A), Fe0.947O (B) and MnO (C) as a function of temperature

Fig. 4
figure 4

Gibbs free enthalpy ∆GR for the water splitting reaction of FeO (A′), Fe0.947O (B′) and MnO (C′) as a function of temperature

They also found that during water-splitting reaction (Fe1−xMn x )1−yO forms manganese-rich rock salt phase and an iron-rich spine phase due to phase segregation processes [12].

Tamura et al. [13] extended the work to ‘NiMnFe’ system, as shown in the reaction schemes 7 and 8.

$$ {\text{Ni}}_{0.5} {\text{Mn}}_{0.5} {\text{Fe}}_{2} {\text{O}}_{4} \mathop{\longrightarrow}\limits_{\,{{\text{at}} > 1,073\;{\text{K}}}}^{\text{activation}}{\text{Ni}}_{0.5} {\text{Mn}}_{0.5} {\text{Fe}}_{2} {\text{O}}_{4 - \delta } + \tfrac{\delta }{2}{\text{O}}_{2} $$
(7)
$$ {\text{Ni}}_{0.5} {\text{Mn}}_{0.5} {\text{Fe}}_{2} {\text{O}}_{4 - \delta } + \delta {\text{H}}_{2} {\text{O}}\mathop{\longrightarrow}\limits_{{{\text{at }} < 1,073\;{\text{K}}}}^{\text{water-splitting}}{\text{Ni}}_{0.5} {\text{Mn}}_{0.5} {\text{Fe}}_{2} {\text{O}}_{4} + \delta {\text{H}}_{2} $$
(8)

They performed the above two-step reaction in a solar reactor at 1,073 K. In the first endothermic step, Ni0.5Mn0.5Fe2O4 was thermally activated to get oxygen-deficient compound, in the second step the oxygen-deficient compound was oxidized using H2O to produce H2. Since, O2 and H2 were produced in two different steps, high-temperature separation of those gases can be eliminated in the proposed method. They have demonstrated the workability of two-step water-splitting reaction with NiFe2O4, Ni0.5Mn0.5Fe2O4 and Ni0.5Zn0.5Zn2O4 systems using thermogravimetric experiments. They found that NiFe2O4 system needs lower reactivation rate (conducted after the water-splitting reaction) compared to Ni0.5Mn0.5Fe2O4 system. The oxygen released during reduction step in NiFe2O4, Ni0.5Mn0.5Fe2O4 and Ni0.5Zn0.5Zn2O4 systems were 0.2, 0.3 and 0.4 %, respectively. They also demonstrated the workability of the two-step hydrogen production in solar reactors. They performed two redox cycles to prove the oxygen and hydrogen evolution in activation (reduction) and reactivation (oxidation) processes. The activation was conducted at 1,373 K in presence of Ar and reactivation was conducted in presence of (steam + Ar) flow at 573 K. In the case of ZnFe2O4, reduction follows two pathways [16] as shown in Eqs. 9 and 10.

$$ 3{\text{ZnFe}}_{2} {\text{O}}_{4} = 3{\text{Zn}} + 2{\text{Fe}}_{3} {\text{O}}_{4} + 2{\text{O}}_{2} $$
(9)
$$ 6{\text{ZnFe}}_{2} {\text{O}}_{4} = 6{\text{ZnO}} + 4{\text{Fe}}_{3} {\text{O}}_{4} + {\text{O}}_{2} $$
(10)

The reduction and oxidation steps have been demonstrated using Xe beam experiment and solar furnace experiments. It took less than 60 s for the Zn-ferrite to release the expected amount of O2 from the lattice at 1,750 K. Authors have seen deposition of Zn on the reactor walls during reduction step and have measured O2 released using mass spectrometer.

Abanades et al. [19] examined CeO2/Ce2O3 redox pairs and demonstrated the feasibility in a solar reactor featuring an inert gas atmosphere at T = 2,273 K, P = 100–200 mbar. It consists of two chemical steps: (1) reduction, 2CeO2 → Ce2O3 + 0.5O2; (2) hydrolysis, Ce2O3 + H2O → 2CeO2 + H2. The reduction step is endothermic and takes place at T = 2,273 K, P = 100–200 mbar; however, oxidation step takes place at 673–873 K resulting in pure hydrogen which can be directly used in fuel cells application. The main advantages of the process are low cost material which is abundantly available in nature and the process uses non-corrosive chemicals. The reduced phase is very stable at ambient temperature and nonreactive to moisture and oxygen which makes this material ideal for on-site hydrogen generation which in turn overcomes problem associated with transportation. However, this technology has few drawbacks, a maximum heat input temperature slightly higher than 2,273 K, the cycle working temperature of the endothermic step must be optimized to be compatible with dish or tower technologies, and to reduce sample vaporization. High molecular weight of cerium oxides poses problem in the flow of solids in the process.

Chueh et al. [20] have extended the work on CeO2 system. They demonstrated the O2 evolution during reduction step, CO and H2 generation during oxidation step using the solar cavity-receiver reactor over 500 cycles. They could achieve solar-to-fuel efficiencies of 0.7–0.8 % and concluded that efficiency is limited by the system scale and design rather than by chemistry. However, Rager [32] pointed out that the efficiency 0.7–0.8 % efficiency refers to ‘peak instantaneous efficiency’ but after averaging the efficiency over 80 % of the fuel production, the actual efficiency is just 0.4 %. He recalculated the solar thermal process efficiency and found that the value is still lower than that of reported by Chueh et al. [20], mainly because the later authors did not consider the energy need for large amount of purge gas used in redox processes. Purge gas takes up lot of solar heat and hence results in lower solar thermal efficiency.

Kang et al. [23] have extended the work on CeO2 system. They synthesized Ce x Zr1−xO2 (x = 0.6, 0.7, 0.8, 1.0) solid solutions and tested for redox reactions. They found that the reduced Ce x Zr1−xO2−δ (x = 0.5, 0.6, 0.7, 0.8, 1.0) samples exhibited higher hydrogen production ability for water splitting due to improved oxygen diffusion through the bulk. Kaneko et al. [22] have extended the work on Ce x Zr1−xO2−δ solid solution system. They introduced Zr4+ on various ratios in CeO2 lattices and found that the oxygen releasing capacity or extent of CeO2 reduction increases with the increase of Zr4+ ions similar to Kang et al.’s [23] observations. The highest oxygen release was found at x = 2 (Ce0.8Zr0.2O2) at 1,773 K in air and the amount of reduced cerium was found to be about 11 % which is seven times higher than just with bare CeO2. The enhancement of the O2-releasing reaction with CeO2–ZrO2 oxide is found to be caused by an introduction of Zr4+, which has smaller ionic radius than Ce3+ or Ce4+ in the fluorite structure.

Le Gal and Abanades [24] doped trivalent lanthanides, viz. La, Sm and Gd in CeO2 to form binary oxides and used in hydrogen production by solar thermal redox cycles. They found that trivalent lanthanide-doped material improves the thermal stability of the material during consecutive redox cycles, but hydrogen production remains the same as ceria. They also doped trivalent lanthanides in CeO2–ZrO2 to form ternary oxides. They found that with 1 % gadolinium to ceria–zirconia solid solutions nearly 338.2 μmol of hydrogen per gram during one cycle with the O2-releasing step at 1,400 °C and the H2-generation step at 1,050 °C. This quantity of hydrogen is more than with CeO2–ZrO2 system. They also found that the addition of lanthanum enhances the thermal stability of ceria–zirconia solid solution similar to as observed in cases of lanthanum-doped CeO2 binary oxides.

Lipinski et al. [21] applied first and second laws of thermodynamics to analyze the potential of applying heat recovery for realizing high efficiency in solar-driven CeO2-based non-stoichiometric redox cycles to split H2O or CO2. They found that at 2,000 K, with 80 % solid phase heat recovery, advanced materials can only increase efficiency from 16 to 20 %, while, at 1,850 K, advanced materials can improve efficiency from 14 to 23 %, a higher maximum value because of decreased re-radiation and gas heating at the lower value of Tred.

Inoue et al. [33] demonstrated effectiveness of a ZnO/MnFe2O4 system in a lab furnace at 1,273 K. When H2O was contacted with ZnO/MnFe2O4 at 1,273 K H2 formation happens with the expense of oxidation of ZnO/MnFe2O4. The later forms spinel kind of material containing ZnII, MnII, MnIII and FeIII ions. The reaction happens by incorporation of ZnII ions into MnFe2O4 crystal structure, accompanied by the partial oxidation of MnII in MnFe2O4 to MnIII. The second step, oxygen releasing can be carried out using solar thermal route but this is not demonstrated experimentally by authors. Similarly, they have also demonstrated H2 production using CaO (or Na2CO3) and MnFe2O4 by passing steam at 1,273 K [29]. The mechanism of H2 formation is similar to that explained earlier, i.e., oxidation of MnII to MnIII to form spinel kind of material (Ca32+Fe 3+2.02 Mn 2+0.96 Mn 3+0.02 O7.02).

Roeb et al. [34] used monolith coatings for redox system. They noticed that the potential of the monolith coatings to absorb oxygen from steam and to release hydrogen decreased with the number of completed cycles which is due to sintering of the material which increases with the redox cycles.

Lundberg [6] performed computer model calculation for various systems for two-step solar hydrogen productions, the systems considered were CoO/Co3O4, MnO/Mn3O4, FeO/Fe3O4, NbO3/Nb2O5 and the halide systems FeX2/Fe3O4 where X = F, Cl, Br and I. In his calculation he found that the ratio of H2/H2O is controlled by the temperature and oxygen partial pressure generated by the redox system. The yield of the hydrogen is defined as follows:

$$ Y\;(\% ) = \frac{{{\text{H}}_{2} ({\text{formed}})}}{{{\text{H}}_{2} (\max )}} \times 100 $$
(11)

where H2max is the maximum amount of hydrogen that can be formed as per the formula:

$$ {\text{MO}}({\text{red}}) + {\text{H}}_{2} {\text{O}}({\text{g}}) = {\text{MO}}({\text{ox}}) + {\text{H}}_{2} ({\text{g}}). $$
(12)

Calculations showed that FeO–Fe3O4 and NbO2–Nb2O5 systems give more H2 yield at lower temperature and that of MnO–Mn3O4 and CoO–Co3O4 systems give >1 % H2 yield at any temperature. In reduction step, in order to reduce thermally oxidized metal oxide needs to be heated up to a temperature where its oxygen partial pressure is higher than in atmosphere (0.21 atm). It was found that though FeO–Fe3O4 and NbO2–Nb2O5 give higher yield they need to be heated above their melting point to reduce them. On the other hand, MnO–Mn3O4 and CoO–Co3O4 systems can be reduced below their melting point but hydrogen yield in these systems are very low (Table 4). Therefore, none of the systems studied are suitable to fulfill both desired conditions for the redox reactions.

Table 4 The yield of H2 at 900 K for the different metal oxide systems together with the enthalpy of the reaction, the reduction temperature in air and the melting points of the system

It was also tried to combine metal oxide which yields higher H2 with metal oxide which can be reduced below its melting point to find out whether this fulfills the need of redox cycle. Considering the spinel phase composition of (Fe0.85Co0.15)3O4 the H2 yield obtained was 45 %, but during the oxidation of the (Fe0.85Co0.15)O system the equilibrium oxygen pressure of the redox system will successfully increases and the yield of the H2 will gradually decreases down to about 3 %. The opposite effect was found during the reduction step, the spinel phase with composition (Fe0.85Co0.15)3O4 will start to be reduced at 2,020 K, but while reduction of the spinel the wuestite phase will become rich with iron and the oxygen partial pressure will decrease leading to gradual increase in the reduction temperature of 2,135 K by the time the initial composition is reached.

An yttrium-stabilized cubic zirconia material coated with iron oxide was proposed to split water in the temperature range 1,273–1,673 K [35, 36]. Kodam et al. [17] studied supported Fe3O4–FeO system. Various amount of iron oxide was supported on yttrium-supported ZrO2 for cyclic redox study. It was found that the Fe3O4 reacts with YSZ to produce Fe2+-containing ZrO2 phase by releasing oxygen molecules in the first step. It was also found that the Fe2+ ions enters into the cubic YSZ lattice. In the second step, the Fe2+-containing YSZ generated hydrogen via steam splitting to reproduce Fe3O4 on the cubic YSZ support. The system showed good reproducibility. It was found that when the Fe3O4 content was increased up to 30 wt% on the Fe3O4/YSZ sample [17], the sample became denser and harder mass after the thermal reduction step, similar to the unsupported Fe3O4. This is due to the fact that the limitation of Fe2+ solubility in the YSZ exists close to the 25 wt% Fe3O4 content in the Fe3O4/YSZ. When raising the Fe3O4 content above 25 wt%, excess Fe2+ ions would form FeO crystals on the ZrO2 surface, which in turn melts at 1,713 K. Therefore, the Fe3O4 contents should be limited to <25 % to avoid sintering of redox material and its cyclic reproducibility.

The disadvantage of mixed iron oxide cycles where oxides are partially reduced and oxidized is their low molar ratio of released oxygen to the total oxygen present in the system. The major drawback of all systems using reactive coatings is their low ratio of hydrogen mass generated to support structure mass. Considering the properties of the above problems, the cycle based on the ZnO/Zn redox pair [79] is of special interest since no cyclic heating and cooling is required and a pure metal state is achieved. It consists of the solar endothermal dissociation of ZnO(s) into its elements; and the non-solar exothermal steam-hydrolysis of Zn into H2 and ZnO(s), and represented by Eqs. 13 and 15.

$$ 1{\text{st step }}({\text{solar ZnO-decomposition}})\!: { }{\text{ZnO}} \to {\text{Zn}} + 0.5{\text{O}}_{2} $$
(13)
$$ 2 {\text{nd step }}({\text{non-solar Zn-hydrolysis}})\!: { }{\text{Zn}} + {\text{H}}_{2} {\text{O}} \to {\text{ZnO}} + {\text{H}}_{2} $$
(14)

H2 and O2 are derived in different steps, thereby eliminating the need for high-temperature gas separation. This cycle has been proposed to be a promising route for solar H2 production from H2O because of its potential of reaching high-energy conversion efficiencies and thereby its economic competitiveness [37, 38].

The first step of the two-step ZnO/Zn water-splitting cycle was first demonstrated in a solar furnace in 1977 by Bilgen et al. [4]. They have demonstrated the decomposition of ZnO in a solar furnace. They also found that Zn yield increases if ZnO is diluted with other refractory materials like ZrO2 and Y2O3 and if the reaction is carried out in the presence of argon inert gas. Table 5 summarized the Zn yield found in the different experiments and different conditions.

Table 5 Mol% zinc content of condensed vapors from ZnO and mixed oxides ZnO–Y2O3, ZnO–ZrO2 samples heated at the focus of 2 kW solar concentrator

Recently, the solar thermal ZnO dissociation was demonstrated by Lédé et al. [39] in a quartz vessel containing sintered ZnO, by Haueter et al. [40] in a rotating cavity reactor type, and by Perkins et al. [41] in an aerosol reactor type. Perkins et al. reported the O2 measurement, which is the only clear indicator of the ongoing thermal ZnO dissociation. The maximum net Zn yield was 17 % [41]. However, to-date there is no report in the literature which claims continuous dissociation of ZnO monitored by product gas analysis for more than few minutes.

Palumbo et al. [25] have studied TiO2 system for two-step solar production of Zn from ZnO, the primary reaction schemes can be written as shown in reactions (15) and (16).

$$ {\text{TiO}}_{2} ({\text{l}}) = {\text{TiO(l)}} + (1 - x/2){\text{O}}_{2} \quad T \ge 2,300\;{\text{K}} $$
(15)
$$ {\text{TiO}}_{x} ({\text{s,l}}) + (2 - x){\text{ZnO(s)}} = (2 - x){\text{Zn(g)}} + {\text{TiO}}_{2} ({\text{s}})\quad T > 1,200\;{\text{K}} $$
(16)

But the authors have not tried water splitting using partially reduced TiO x . The minimum values of x that the authors obtained experimentally were 1.91, 1.86 and 1.83 for temperatures of 2,300, 2,500, and 2,700 K, respectively, in an Ar atmosphere at 1 bar. They used the latter material to reduce ZnO to produce Zn as indicated in reaction (16). It is to be noted that the higher the degree of decomposition, the greater the vaporization of TiO2, this limits the efficiency of the water-splitting cycle using TiO2 system.

Sibieude et al. [10] have used CdO for two-step water-splitting reaction. They demonstrated reduction of CdO to Cd in a solar furnace at high temperature. The reaction scheme is shown in Eqs. 1719.

$$ {\text{CdO}} \to {\text{Cd}} + \raise.5ex\hbox{$\scriptstyle 1$}\kern-.1em/ \kern-.15em\lower.25ex\hbox{$\scriptstyle 2$} {\text{O}}_{2} \quad T > 1,200\;^\circ {\text{C}} $$
(17)
$$ {\text{Cd}} + {\text{H}}_{2} {\text{O}} \to {\text{Cd(OH)}}_{2} + {\text{H}}_{2} $$
(18)
$$ {\text{Cd(OH)}}_{2} \to {\text{CdO}} + {\text{H}}_{2} {\text{O }}\quad T > 375 \;^\circ {\text{C}} $$
(19)

They observed that subjecting CdO alone to solar radiation did not reduce the oxide, but when CdO was mixed with refractory material, in their case 20 %mol ZrO2, resulted in the formation of Cd metal in the stream of Ar. The amount of Cd metal in the deposited condensate at different temperatures is shown in Fig. 5.

Fig. 5
figure 5

Cd (metal) content of condensates versus temperature after thermal decomposition of CdO, flow rate of argon was A* 3.4 cm3/s and B* 10 cm3/s

Quenching of evaporated metal was very important in this reaction. When CdO was dissociated into Cd(g) and O(g) the recombination will also takes place very fast. When CdO alone was heated strong vaporization produces large amount of dissociated vapors which is insufficiently quenched by the argon flow on a water cooled wall of the condenser. The problem was overcome by mixing the CdO with ZrO2; in this case vaporization rate of Cd metal was lowered by its dispersion in the refractory metal oxide matrix which permits the effective quenching of vaporized metal. It is to be noted that partial pressure of oxygen plays a main role in the Cd yield. Figure 6 gives the %Cd metal recovered in various O2 partial pressures.

Fig. 6
figure 6

Dependence of Cd (metal) content of condensates on the oxygen concentration of the argon flow

Abanades et al. [15] have studied SnO2 = Sn + 1/2O2 cycle which consists of a solar endothermic reduction of SnO2 into SnO(g) and O2 followed by a non-solar exothermic hydrolysis of SnO(s) to form H2 and SnO2(s).The thermal reduction occurs under atmospheric pressure at about 1,873 K and over. The solar step encompasses the formation of SnO nanoparticles that can be hydrolyzed efficiently in the temperature range of 500–600 °C with a H2 yield over 90 %. A preliminary process design is also proposed for cycle integration in solar chemical plants. They also compared their system with literature reported ‘Sn-Souriau’ [42] three-step cycles and inferred that the reaction (22) producing hydrogen from the Sn/SnO2 mixture produced from reaction (21) is slow and partial at 600 °C which results in low H2 yield of <45 %. The three-step cycling process proposed by them is as follows:

$$ {\text{SnO}}_{2} \to {\text{SnO}} + 1/2{\text{O}}_{2} $$
(20)
$$ 2{\text{SnO}} \to {\text{Sn}} + {\text{SnO}}_{2} $$
(21)
$$ {\text{Sn}} + 2{\text{H}}_{2} {\text{O}} \to {\text{SnO}}_{2} + {\text{H}}_{2} $$
(22)

Fan et al. [43] have studied steam to hydrogen conversion using six different metals. It is interesting to note that only Fe and Sn are found to generate reasonable hydrogen at 873 K as shown in Table 6. Other metals did not show a good amount of hydrogen production at 873 K. Considering melting point of different metallic and their oxides states (as shown in Table 7) of Fe and Sn it can be inferred that Fe is very suitable for given application unless there is a provision to handle liquid metal in the solar reactor similar to Zn–ZnO2 case. The steam to H2 value (γa) of 40.82 % is lower compared to one reported by Abanades et al. [15] which is equivalent to 90 % at similar conditions. If solar reactor is designed to handle liquid metals, then both Zn and Sn seems to be better candidates for two-step redox reactions with good hydrogen yield and at low-temperature operation.

Table 6 Maximum per-pass conversion of H2O to H2 in the regeneration reactor and the stable phase obtained at 873 K for countercurrent gas–solid operation
Table 7 Melting points [44] of various phases of Fe and Sn

Recently, Cho and Kim [27] reported production of H2 using solid acids such as silica gel, activated Al2O3, CaCO3, TiO2 and ZnO. This is very interesting study as it reports on liberation of hydrogen gas at very low temperature. They have demonstrated the possibility of H2 production using a laboratory plug flow reactor. Figure 7 gives the concentration and total amount of H2 liberated at various temperatures using activated Al2O3.

Fig. 7
figure 7

H2 concentration in the product gas stream at a reaction time of 1 h and total amount of H2 produced versus reaction temperature resulted from using a wetted Al2O3. For the experiments, 60 g of Al2O3 (5.5 wt% H2O) in a stainless steel reactor and 2 ml/min of CO2 carrier gas were used at atmospheric pressure

The concentration of hydrogen produced in product gas stream using five different oxides at ≈610 K is shown in the Table 8. Though CaCO3 shows highest concentration of H2, i.e., 1,590 ppm in 1 h reaction time the activated Al2O3 produces highest amount of total H2 at 1,073 K.

Table 8 H2 concentration in the product gas stream at a reaction time of 1 h

Though the authors demonstrated the workability of the producing hydrogen from solid acids on a laboratory-scale fixed-bed reactor, replication of the results in a solar reactor needs to be performed to know the feasibility of the process.

One of the main problem to tackle is overcoming sintering of redox material. Agglomeration due to sintering brings down the recyclability over multiple redox cycles. The key properties of the redox material should include a high oxygen carrying capacity, good mechanical properties and cheap and easy synthetic procedures. If redox material do not fulfill any one of these key properties it would not be a suitable material for commercial-scale operation.

Particle size or grain size effect on rate of oxidation

It is generally accepted that smaller the particle size easier is to oxidize or reduce. In case of two-phase alloys the rate of oxidation may significantly improve with grain size reduction because both mutual solubility and diffusivity among the system will enhance [45, 46]. But this is not always the case, during the oxidation if the top layer acts as a protective layer then the further oxidation of the metals will be hampered.

Figure 8 shows the oxidation kinetics of three different alloy systems with two different grain sizes at 1,073 K. The grain size reduced Cu–Cr alloy showed very slow oxidation kinetics compared to As-cast alloy. But in case of Cu–Fe and Cu–Co oxidation kinetics found to be much faster when nano-crystals (20–30 nm) were used compared to the As-cast alloy. This is because Cr2O3 scale formed on alloy prevents further oxidation. This is similar to in case aluminum where external layer forms Al2O3 and prevents further oxidation or corrosion of the aluminum metal.

Fig. 8
figure 8

Oxidation kinetics of As-cast and grain size reduced Cu–M alloys at 1,073 K. 1 Grain size reduced Cu–Cr alloy, 2 As-cast Cu–Co alloy, 3 As-cast Cu–Cr alloy, 4 grain size reduced Cu–Fe alloy, 4 As-cast Cu–Fe alloy, 6 grain size reduced Cu–Co alloy. [45]

The reduction kinetics of metal oxides depends on many factors such as whether they are supported or unsupported, particle size, gas atmosphere, kind of metal oxides and whether single or mixed metal oxides. There are not many reports available on high-temperature reduction of metal oxides in an inert atmosphere as in the case of solar thermal reduction, but there are plenty of studies available in the literature on reduction of metal oxides using H2 or CO as reducing agents. For example, Syed-Hassan and Chun-Zhu [47] have studied the particle size effect on reduction of NiO in H2 atmosphere. The reduction profiles for NiO particles of size 20 and 24 nm are very different from that of particle size of 3.3 nm (please refer Fig. 5a in Ref. [47]). The profiles for 20 and 24 nm are almost similar and very much resemble to that of 55 nm NiO supported on SiO2 substrate. Authors concluded that reduction kinetics is independent of supported or unsupported NiO, but merely depends on the particle/crystal size. The Ea versus %NiO converted trends are very different for first reduction (particle size 3.3 nm) to that of second (particle size = 20 nm) and third reduction (particle size = 24 nm) (please refer Fig. 5b in Ref. [47]). The Ea for first reduction remains almost unchanged (indicating single rate-limiting step) throughout the whole reduction process, but for second and third reduction steps the Ea profiles continuously increased till the complete reduction of NiO. The main reason for the difference could be due to the difference in surface to bulk atoms in different particle size crystals. The size of the metallic island which forms during initial stages of reduction is bigger than particle size, i.e., 3.3 nm, or the whole particle surface can be instantaneously covered by the metallic layer without requiring a significant growth of islands; therefore, the reduction happens immediately in those particles size crystals. But for a big particle with significant atoms in the interior (e.g., 20 nm) it might take quite some time for the growth of island to cover the whole particle’s surface. Once the surface has been covered completely by the metal product, the Ni–NiO boundary would then progressively advance inward. As reduction continues, the hydrogen radicals moves towards Ni–NiO boundary on the surface and the diffusion of atoms (i.e., movement and rearrangements of atoms in the interior of the particle) takes place. In general, solid-state diffusion requires higher activation energy [48]. The number of steps of diffusion in the solid state would appear to increase with NiO conversion, resulting in continuous increases in the activation energy. This is the reason in general the reduction of bigger particles crystallites needs higher activation energy than smaller one.

Economic evaluation

The US DOE has established a target of $2 to $3 per kg hydrogen by 2025 to make it economically affordable. The short term, i.e., 2015, DOE target is $6/kg hydrogen. Any competitive technology to produce hydrogen considers this figure as a reference for their process efficiency and economic evaluation.

DOE in collaboration with TIAX (TIAX is a laboratory-based technology development company with a focus on clean energy) led the effort of cost calculation for solar thermochemical hydrogen (STCH) in many US national laboratories. They considered eight promising technologies for cost calculations, viz., hybrid-sulfur (HyS), copper chloride (CuCl), thin-film nickel ferrite (“ferrite”), sulfur-ammonia (S–A), zinc oxide (ZnO), manganese oxide (MnO), sulfur-iodine (S–I), and cadmium oxide (CdO).

Five out of eight technologies mentioned in Table 9 appear to meet DOE’s short-term target (by 2015) of $6/kg hydrogen but meeting long-term target seems quite difficult. Only thin film ferrite is very close to DOE’s long-term requirement. Even in this case achieving the long-term targets require significant technological advances in multiple dimensions. The primary cost driver for all the processes that were analyzed was heliostats costs. Reducing the heliostats cost or increase in plant efficiency will bring down the CAPEX and OPEX. Most probably mass production of heliostats and improving its efficiency will help to reduce the CAPEX and help to meet DOE hydrogen cost targets.

Table 9 Hydrogen cost estimation (per kg H2) in different processes

Conclusions

Several metal oxides have been proposed to apprehend redox cycles. To-date, the solar-to-fuel efficiencies of prototype reactors are low, on the order of <1 %. The main problem in large-scale solar application would be an efficient fast quenching system to hinder the reoxidation of the reduced metal in liquid or in vapor phase. When molten metal is made to react with steam, an oxide layer will form on the surface and it floats on top of the melt, which prevents further oxidation reaction.

So far, there are no reports demonstrating good repeatability of the cyclic two-step reaction to satisfy the practical use of the process. This would be one of the most difficult achievements in this technology because the high-temperature process will cause significant sintering of the metal oxide, which severely deactivates metal oxide for repeated cyclic reactions.

Though there are hundreds of publication available in the literature on metal oxide redox cycle for hydrogen production, only few of them are practically demonstrated in prototype solar reactor. The challenges while conducting redox experiments in a solar reactor are very different than in a laboratory-scale plug flow reactor. The commercial realization of redox technology for hydrogen production seems still far away.