**Optimization Design of Rib Width and Performance Analysis of Solid Oxide Electrolysis Cell**

#### **Meiting Guo 1,\*, Xiao Ru <sup>2</sup> , Zijing Lin 1,2,3 , Guoping Xiao <sup>4</sup> and Jianqiang Wang <sup>4</sup>**


Received: 15 September 2020; Accepted: 13 October 2020; Published: 19 October 2020

**Abstract:** Structure design is of great value for the performance improvement of solid oxide electrolysis cells (SOECs) to diminish the gap between scientific research and industrial application. A comprehensive multi-physics coupled model is constructed to conduct parameter sensitivity analysis to reveal the primary and secondary factors on the SOEC performance and optimal rib width. It is found that the parameters of the O<sup>2</sup> electrode have almost no influence on the optimal rib width at the H<sup>2</sup> electrode side and vice versa. The optimized rib width is not sensitive to the electrode porosity, thickness, electrical conductivity and gas composition. The optimal rib width at the H<sup>2</sup> electrode side is sensitive to the contact resistance at the interface between the electrode and interconnect rib, while the extremely small concentration loss at the O<sup>2</sup> electrode leads to the insensitivity of optimal rib width to the parameters influencing the O<sup>2</sup> diffusion. In addition to the contact resistance, the applied cell voltage and pitch width also has a dramatic influence on the optimal rib width of the fuel electrode. An analytical expression considering the influence of total cell polarization loss, the pitch width and the contact resistance is further developed for the benefit of the engineering society. The maximum error in the cell performance between the numerically obtained and analytically acquired optimal rib width is only 0.14% and the predictive power of the analytical formula is fully verified.

**Keywords:** solid oxide electrolysis cell; multi-physics; optimal rib/pitch ratio; parameters sensitivity; analytical expression

## **1. Introduction**

Renewable energy resources, including solar, wind, tidal, and biomass, are of great significance as the fossil energy crisis is becoming increasingly serious. However, their intermittence leads to an undesirable imbalance between demand and supply [1]. An energy storage device is required so that energy can be stored and released as needed. Among various candidates for energy storage, solid oxide electrolysis cells (SOECs) have gained popularity for higher efficiency and lower pollution and even no pollution. SOEC is an energy conversion device that can convert electrical energy and heat to chemical energy, by splitting H2O/CO<sup>2</sup> to produce H<sup>2</sup> and CO. The products H<sup>2</sup> and CO can be used as fuel in a solid oxide fuel cell (SOFC) to produce electricity or be stored as raw materials for the synthesis of hydrocarbons via the Fischer–Tropsch reaction.

SOEC technology is of great superiority and prospect. Among three main electrolysis configurations, it has been reported that the efficiency of hydrogen production by high temperature SOEC is more than twice of that by an Alkaline electrolysis cell, and is 1.5 times of that by proton exchange membrane electrolyzer [2]. There are mainly three kinds of SOEC according to reaction gas species: high temperature H2O electrolysis, CO<sup>2</sup> electrolysis, H2O and CO<sup>2</sup> co-electrolysis cell. Steam electrolysis can produce H2, which is a completely environmentally friendly fuel. In addition, H2O is rich in nature. CO<sup>2</sup> electrolysis is advantageous in that it can consume CO<sup>2</sup> and relieve the greenhouse effect, and the product CO is easier to store and transport than H2. However, it has a potential carbon deposit risk. H2O and CO<sup>2</sup> co-electrolysis can produce H<sup>2</sup> and CO mixtures, and by adjusting the inlet H2O/CO<sup>2</sup> ratio, it can produce applicable hydrocarbon synthesis. The steam electrolysis has the highest electrolysis efficiency while CO<sup>2</sup> electrolysis has the lowest efficiency, and the efficiency of H2O/CO<sup>2</sup> co-electrolysis is between them. Due to the environmental friendliness and higher efficiency of steam electrolysis, high temperature H2O electrolysis attracts increased attention. With the increase of temperature, the electricity needed to electrolyze H2O decreases, while the low quality heat needed increases. Moreover, high temperature SOEC is not only thermodynamically beneficial but also kinetically favorable. Hence, high temperature SOEC has a more promising application perspective.

Materials, performance and degradation issues are still three challenges of SOEC technology to be settled. Extensive research about SOEC concentrates on the optimization of material microstructure [3– 5], geometrical [6] and operating parameters [7] and the analysis to improve SOEC performance. Notice, however, the structural size choice is also of great significance. For example, the authors of [8] studied the effect of cathode thickness on CO2/H2O co-electrolysis performance under various operating conditions by experiment, which reveals that SOEC performance can be substantially improved by decreasing cathode thickness. For the time being, most of the SOEC researches were conducted experimentally [2,9–12]. Unfortunately, the experiment is expensive and time-consuming, so research about structure optimization by experiment is rare. Simulation is an efficient alternative to help the design of SOEC/SOFC to improve performance, especially when exploring a large combination of operating and structural parameters space. For example, Reference [1,11–13] studied the influence of pressure on SOEC. Ni et al. [14] researched electrode thickness, support type, electrode porosity and pore size and operating pressure on SOEC performance. It is concluded that anode-supported SOEC has the best output performance. Kong et al. [15] examined the impacts of different electrode-rib contact resistances, fuel compositions, electrode porosity, electrode thicknesses and electrode conductivity on the optimal anode and cathode ribs of SOFC independently.

The development of SOEC is later than that of SOFC. As the reverse process of SOFC, SOEC has basically the same materials system as SOFC. The research of SOFC is enlightening to the development of SOEC. In the last 20 years, during the simulation of SOFC, it found out that rib width design is of great significance to improve the cell performance [15–18], which has also been experimentally confirmed [19]. Actually, as early as 2003, Lin [16] had provided a phenomenological model and analytical expressions to estimate the rib effects on the concentration and ohmic polarization of anode-supported SOFC stacks. Jeon et al. [20] described a microstructure model and examined the influence of the rib and pitch widths and the electrode-interconnect contact area specific resistance (ASR) on the stack-cell performance. The authors of [17] investigated the effect of ASR between the electrode and the rib on the performance of SOFC, and conducted the rib width optimization by 2D SOFC multi-physical modes. In [18], the authors primarily compared the optimal rib width result attained by 2D and 3D multi-physical models, and revealed that the optimization result of 2D and 3D models are in good agreement. The authors of [15,21] optimized the anode and cathode rib width for anode-supported and cathode-supported SOFCs, respectively. An analytical expression of optimal rib width is deduced to help the engineering design of SOFC. In [22], the authors conducted the optimization of the cylindrical interconnect rib width of SOFC, and it found out that anode and cathode rib width should be optimized separately, which is in consistent with the conclusion in Reference [21]. Reference [23] studied the influence of rib size on the performance of a reversible solid oxide cell. It takes into account the efficiency of a solid oxide fuel cell and electrolysis cell comprehensively.

However, the polarization process of SOEC is vastly different from SOFC. During past years, to our knowledge, it is still not clear how to choose the rib width when fabricating the SOEC interconnector characterized by the rib-channel structure, i.e., the optimization of rib width has not been clearly addressed. This paper conducts a comprehensive parameter sensitivity analysis on the rib width optimization of SOEC by 2D multi-physics simulations. Analytical expressions for the optimal cathode rib width design are obtained to provide an easy-to-use guide for designing the rib-channel layout of SOEC.

#### **2. Model Description**

A typical unit of SOECs contains a dense electrolyte sandwiched by a porous anode and cathode, channels and inter-connector. To obtain better electrochemical performance, a relatively denser and thinner porous layer are generally added between the electrode and electrolyte, which is called the function layer. The working principle of SOEC is illustrated in Figure 1a. Figure 1b shows a 2D cross section of the SOEC repeating unit, displaying the rib-channel design parameters: the pitch width, *d*pitch, and the rib width, *d*rib. The multi-physics coupled modeling is applied to a pitch unit of SOEC and considers the mass, species, momentum transfer processes, the current conduction and the electrochemical reaction.

**Figure 1.** (**a**) Schematic of solid oxide electrolysis cell (SOEC) working principle; (**b**) a cross section of a SOEC repeating unit.

*2.1. Species Transfer Process*

The species conservation equation can be expressed as:

$$\vec{\nabla} \cdot \vec{\mathbf{N\_i}} = \nabla \cdot (-\mathbf{D\_i^{eff}C\_i} + \mathbf{C\_i}\vec{\mathbf{u}}) = \mathbf{S\_i} \tag{1}$$

$$\mathbf{S\_{H2}} = \frac{\mathbf{j}}{2\mathbf{F}} \tag{2}$$

$$\mathbf{S}\_{\rm H2o} = -\frac{\mathbf{j}}{2\mathbf{F}}\tag{3}$$

$$\mathbf{S\_{O2}} = \frac{\mathbf{j}}{4\mathbf{F}} \tag{4}$$

Here, N<sup>i</sup> is the total molar flux vector, which includes the convective flow and the diffusion flow. Di eff , C<sup>i</sup> , and S<sup>i</sup> are effective diffusion coefficient, molar concentration and source or sink of species

i, respectively. S<sup>i</sup> is source or sink of species i by electrochemical consumption and production in the electrode.

The species diffusion is modeled by a multi-component Dusty–Gas model [24], which is proved to be the most accurate model to simulate gas diffusion in a porous electrode of SOFC [25]. A binary molecular diffusion coefficient is generally used directly in channels, where gas convective transfer rather than the diffusion transfer is dominant, and the binary molecular mutual diffusion coefficient is expressed as [18]:

$$\mathbf{D}\_{\mathbf{\dot{j}}} = \frac{2.198 \times \mathbf{T}^{1\%}}{\mathbf{P} \left(\mathbf{V}\_{\mathbf{i}}^{1/3} + \mathbf{V}\_{\mathbf{j}}^{1/3}\right)} \left(\frac{1}{\mathbf{M}\_{\mathbf{i}}} + \frac{1}{\mathbf{M}\_{\mathbf{j}}}\right)^{0.5} \tag{5}$$

where T, v<sup>i</sup> and M<sup>i</sup> is the temperature, diffusion volume and mole mass of species i. P is the total gas pressure.

In the electrode, the influence of porosity and tortuosity should be considered to describe the effective binary molecular diffusion coefficient:

$$\mathbf{D}\_{\mathrm{ij}}^{\mathrm{eff}} = \frac{\varepsilon}{\pi} \mathbf{D}\_{\mathrm{\mathbb{I}}} \tag{6}$$

ε, τ are porosity and tortuosity of the electrode, respectively.

In the electrode, molecular mutual diffusion dominates when the electrode pore is larger than the molecular mean free path, while Knudsen diffusion becomes primary when the pore size is small and the collision between the species and pore wall increases. The effective Knudsen diffusion coefficient is expressed as:

$$\mathbf{D}\_{\rm ik}^{\rm eff} = \frac{2\varepsilon}{\Im\pi} \mathbf{r}\_{\rm g} \sqrt{\frac{8\mathbf{RT}}{\pi \mathbf{M}\_{\rm i}}} \tag{7}$$

Here r<sup>g</sup> is pore radius, and r<sup>g</sup> is expressed as:

$$\mathbf{r\_g} = \frac{2}{3} \ast \frac{1}{1 - \varepsilon} \ast \frac{1}{\varphi\_{\rm el}/\mathbf{r\_{el}} + \varphi\_{\rm io}/\mathbf{r\_{el}}} \tag{8}$$

Considering both molecular diffusion and Knudsen diffusion simultaneously, the effective diffusion coefficient in the electrode is formally expressed as [18]:

$$\mathbf{D}\_{\mathbf{i}}^{\text{eff}} = \frac{\mathbf{D}\_{\mathbf{i}\mathbf{j}}^{\text{eff}} \mathbf{D}\_{\mathbf{ik}}^{\text{eff}}}{\mathbf{D}\_{\mathbf{i}\mathbf{j}}^{\text{eff}} + \mathbf{x}\_{\mathbf{i}} \mathbf{D}\_{\mathbf{j}\mathbf{k}}^{\text{eff}} + \mathbf{x}\_{\mathbf{j}} \mathbf{D}\_{\mathbf{ik}}^{\text{eff}}} \tag{9}$$

#### *2.2. Electrochemical Reaction Model*

The electrochemical reactions include the H2O reduction reaction at the fuel electrode and the oxidation reaction at the O<sup>2</sup> electrode, which are expressed as:

$$\text{H}\_2\text{O} + 2\text{e}^- \rightarrow \text{H}\_2 + \text{O}^{2-} \text{fuelelectrode} \tag{10}$$

$$\text{O}^{2-} \text{--} 2\text{e}^{-} \rightarrow \frac{1}{2} \text{O}\_{2} \text{aire electrode} \tag{11}$$

The electronic current and ion current are governed by charge continuity equations:

$$\mathbf{V} \cdot \mathbf{\bar{i}}\_{\rm el} = \nabla \cdot (-\sigma\_{\rm el}^{\rm eff} \nabla \phi\_{\rm el}) = \begin{pmatrix} \mathbf{j}\_{\rm TPB} \lambda\_{\rm TPB}^{\rm eff} & \text{fuel electrode} \\ -\mathbf{j}\_{\rm TPB} \lambda\_{\rm TPB}^{\rm eff} & \text{air electrode} \end{pmatrix} \tag{12}$$

$$\nabla \cdot \mathbf{\bar{i}}\_{\rm io} = \nabla \cdot (-\sigma\_{\rm io}^{\rm eff} \nabla \phi\_{\rm io}) = \begin{pmatrix} -\mathbf{j}\_{\rm TPB} \lambda\_{\rm TPB}^{\rm eff} & \text{fuel electrode} \\ 0 & \text{electrolyte} \\ \mathbf{j}\_{\rm TPB} \lambda\_{\rm TPB}^{\rm eff} & \text{air electrode} \end{pmatrix} \tag{13}$$

*Energies* **2020**, *13*, 5468

$$\mathbf{j}\_{\rm TPB} = \mathbf{j}\_0 [\exp(\frac{\mathbf{n} \mathbf{a} \mathbf{F} \boldsymbol{\eta}}{\mathbf{R} \mathbf{T}}) - \exp(-\frac{\mathbf{n} \boldsymbol{\beta} \mathbf{F} \boldsymbol{\eta}}{\mathbf{R} \mathbf{T}})] \tag{14}$$

where n is the number of electrons transferred in the electrochemical reaction, iel and iio are the electronic and ionic current density vector, φel, φio are the electronic and ionic potential. λTPB,eff is the effective three phase boundary (TPB) density per unit. jTPB is the current density at TPB.

Nernst potential is the minimum cell voltage needed to drive an electrochemical reaction in SOEC:

$$\mathbf{E\_{nrest}} = \frac{-\Delta \mathbf{G}}{2\mathbf{F}} + \frac{\mathbf{RT}}{2\mathbf{F}} \mathbf{Ln}(\frac{\mathbf{P\_{H2}}}{\mathbf{P\_{H2O}}}) + \frac{\mathbf{RT}}{4\mathbf{F}} \mathbf{Ln}(\frac{\mathbf{P\_{O2}}}{\mathbf{P\_{O}}}) \tag{15}$$

Here, ∆G is the Gibbs free energy change of electrochemical reaction. PH2, PO2, PH2O are partial pressures of H2, O<sup>2</sup> and H2O. P<sup>0</sup> is the standard atmospheric pressure.

The applied voltage in an operating SOEC is:

$$\mathbf{V\_{cell}} = \mathbf{E\_{nert}} + \eta\_{\text{ohmic}} + \eta\_{\text{act}} + \eta\_{\text{con}} \tag{16}$$

where ηohmic, ηact, ηcon are ohmic, activation and concentration loss, respectively. Ohmic loss is induced by the conductivity resistance of electron and ion transfer in solid components and is calculated according to Ohm's law:

$$\mathfrak{m}\_{\text{ohm}} = \mathbf{j} \* \mathbf{ASR}\_{\text{ohm}} \tag{17}$$

Here, ASRohm is total area specific resistance.

The occurrence of the electrochemical reaction needs to overcome the reaction activation energy barrier, leading to irreversible activation loss. The relationship between current density and the activation loss is described by the Butler–Volmer equations:

$$\begin{cases} \mathbf{j}\_{0, \text{H}\_{2}} = \mathbf{j}\_{0, \text{H}\_{2}} \overset{\text{ref}}{\exp} \left( -\frac{\text{E} \mu\_{2}}{\text{R}} \left( \frac{1}{\text{T}} - \frac{1}{\text{T}\_{\text{ref}}} \right) \right) \left( \frac{\text{P}\_{\text{H}\_{2}} \text{P}\_{\text{H}\_{2} \text{O}}}{\text{P}\_{\text{H}\_{2}, \text{ref}} \text{P}\_{\text{H}\_{2} \text{O}} \text{P}\_{\text{ref}}} \right) \\\quad \mathbf{j}\_{0, \text{O}\_{2}} = \mathbf{j}\_{0, \text{O}\_{2}} \overset{\text{ref}}{\exp} \left( -\frac{\text{E}\_{\text{O}\_{2}}}{\text{R}} \left( \frac{1}{\text{T}} - \frac{1}{\text{T}\_{\text{ref}}} \right) \right) \left( \frac{\text{P}\_{\text{O}\_{2}}}{\text{P}\_{\text{O}\_{2} \text{ref}}} \right) \end{cases} \tag{18}$$

The concentration polarization is induced by a change of species concentration when SOEC is in operation:

$$\eta\_{\text{concen,a}} = \frac{\text{RT}}{\text{2F}} \text{Ln}(\frac{\text{P}\_{\text{H}\_2}\text{P}\_{\text{H}\_2\text{O}\_{-}}\text{ref}}{\text{P}\_{\text{H}\_2\text{-ref}}\text{P}\_{\text{H}\_2\text{O}}}) \tag{19}$$

$$\eta\_{\text{concen,a}} = \frac{\text{RT}}{4\text{F}} \text{Ln}(\frac{\text{P}\_{\text{O}\_2}}{\text{P}\_{\text{O}\_2-\text{ref}}}) \tag{20}$$

EO2 and EH2 are, respectively, the activation energies for the O<sup>2</sup> and H<sup>2</sup> electrode electrochemical reactions.

#### *2.3. E*ff*ective Material Property Model*

The intrinsic material conductivity is temperature dependent [1,6,15,26]:

$$
\sigma\_{\rm Ni}^{0} = 3.27 \times 10^{6} - 1065.3 \text{T} \tag{21}
$$

$$
\sigma\_{\rm YSZ}^{0} = 6.25 \times 10^{4} \times \exp^{\left(-10, 300/\text{T}\right)}\tag{22}
$$

$$
\sigma\_{\rm LSCF}^{\rm el} = 22.591 - 1.6 \times 1^{06} \times \exp^{(-6024/\text{T})} \tag{23}
$$

$$
\sigma\_{\rm LSCF}^{\rm iso} = 1.1 \times 10^9 \times \exp^{\left(-181,000/\text{R/T}\right)}\tag{24}
$$

$$
\sigma\_{\rm GDC}^{0} = 3.5 \times 10^3 \times 10^{(-6471/\text{T})} \tag{25}
$$

where R is the universal gas constant.

For the porous electrode, the relationship between the material macro property and microstructure can be expressed by coordination number theory and percolation theory [27]:

$$
\sigma\_{\mathbf{k}}{}^{\rm eff} = \sigma\_{\mathbf{k}}{}^{0} \* \left(\frac{\varphi\_{\mathbf{k}} - \varphi\_{\mathbf{k}}{}^{\rm t}}{1 + \varepsilon/(1 - \varepsilon) - \varphi\_{\mathbf{k}}{}^{\rm t}}\right)^{2} \tag{26}
$$

where σ<sup>k</sup> 0 is the intrinsic electric conductivity of the kth phase material in the dense solid. φ<sup>k</sup> is the volume fraction of the kth phase particles in the composite material, φ<sup>k</sup> t is the percolation threshold volume fraction of the k phase particles, which is determined by

$$Z \frac{\psi\_{\rm el}^{\rm t}/\rm r\_{\rm el}}{\psi\_{\rm el}^{\rm t}/\rm r\_{\rm el} + \left(1 - \psi\_{\rm el}^{\rm t}\right)/\rm r\_{\rm io}} = 1.764\tag{27}$$

$$Z \frac{\psi\_{\rm{io}}{}^{\rm t}/\rm r\_{\rm{io}}}{(1-\psi\_{\rm{io}}{}^{\rm t})/\rm r\_{\rm{el}}+\psi\_{\rm{io}}{}^{\rm t}/\rm r\_{\rm{io}}} = 1.764\tag{28}$$

where Z is the average coordination number for each particle and set as six for a random packing of spheres [15]. rel and rio are, respectively, the electronic conductivity particle radius and the ionic conductivity particle radius.

LSCF is a material that can conduct electrons and ions simultaneously. For the LSCF-GDC composite electrode, the effective electronic conductivity is expressed as:

$$
\sigma\_{\rm el}^{\rm eff} = \sigma\_{\rm LSCF}^{\rm el} [(1 - \varepsilon)\varPhi\_{\rm LSCF} \mathbf{P}\_{\rm LSCF}]^\gamma \tag{29}
$$

The effective ionic conductivity can be considered as the parallel current conduction of LSCF and GDC. The ionic conductivity of the composite electrode is expressed as:

$$\sigma\_{\rm io}^{\rm eff} = \sigma\_{\rm LSCF}^{\rm io} [(1 - \varepsilon)\varphi\_{\rm LSCF} \mathbf{p}\_{\rm LSCF}^{\rm t}]^Y + \sigma\_{\rm GDC}^{\rm io} [(1 - \varepsilon)\varphi\_{\rm GDC} \mathbf{p}\_{\rm GDC}^{\rm t}]^Y \tag{30}$$

where γ is a Bragg factor and is usually set as 1.5. P<sup>k</sup> is the percolation probability of phase k. Both the percolation probabilities of LSCF and GDC are assumed to be 1.

The TPB density per unit volume of a composite electrode with a binary mixture is expressed as [27]:

$$
\lambda\_{\rm TPB} = 2\pi \text{min}(\mathbf{r}\_{\rm el}, \mathbf{r}\_{\rm io}) \sin(\frac{\theta}{2}) \mathbf{n} \mathbf{n}\_{\rm io} Z\_{\rm io-el} \mathbf{P}\_{\rm el} \mathbf{P}\_{\rm io} \tag{31}
$$

The relevant physical quantities are expresses as follows,

$$\mathbf{P\_k} = (\mathbf{1} - (\frac{3.764 - \mathbf{Z\_{k,k}}}{2})^{2.5})^{0.4} \tag{32}$$

$$\mathbf{Z\_{k,k}} = \mathbf{Z} \ast \frac{\mathbf{q\_k/r\_k}}{\mathbf{q\_{el}/r\_{el}} + \mathbf{q\_{io}/r\_{io}}} \tag{33}$$

$$\mathbf{n} = \frac{1 - \varepsilon}{\frac{4\pi}{3}\mathbf{r}\_{\rm el}\mathbf{s}^3(\mathbf{n}\_{\rm el} + (1 - \mathbf{n}\_{\rm el})\boldsymbol{\gamma}^3)}\tag{34}$$

$$\mathbf{m\_{el}} = \frac{\varrho\_{\rm el} \boldsymbol{\gamma}^3}{1 - \varrho\_{\rm el} + \boldsymbol{\gamma}^3 \varrho\_{\rm el}} \tag{35}$$

$$\mathbf{n}\_{\rm io} = 1 - \mathbf{n}\_{\rm el} \tag{36}$$

$$Z\_{\rm iso-el} = \frac{Z}{2} \Big( 1 + \frac{\mathbf{r\_{io}}^2}{\mathbf{r\_{el}}^2} \Big) \frac{\varphi\_{\rm el} / \mathbf{r\_{el}}}{\varphi\_{\rm el} / \mathbf{r\_{el}} + \varphi\_{\rm io} / \mathbf{r\_{io}}} \tag{37}$$

More details about the specific symbols are referred to in [27].

The multi-physics simulations are conducted with the commercial finite element software COMSOL 4.3b. Simulations were first carried out to reproduce the performance of an experimental cell that is designated as the standard cell. The parameters used to reproduce the experiment are listed in Table 1, and are referred to as the standard parameter set. Unless explicitly stated otherwise, the parametric analysis was conducted by varying one parameter as the variable while all the other parameters are kept as the standard case.


**Table 1.** Basic model parameters used in the standard cell modeling.

#### *2.4. Model Verification*

To verify the multi-physics model of SOEC, simulated I–V data are compared with our experimental results. The experimental cell contains a cathode support layer (CSL), a cathode function layer (CFL), an electrolyte, a diffusion barrier layer and the anode. CSL and CFL are mixtures of nickle oxide (NiO) and yttria stabilized zirconia (YSZ), and the electrolyte is a dense YSZ layer. The composite anode is made of La0.6Sr0.4Co0.2Fe0.8O3-<sup>δ</sup> (LSCF) and Ce0.9Gd0.1O1.95 (GDC). The diffusion barrier layer (GDC) is sandwiched between anode and electrolyte to avoid formation of the insulating phase. The SOEC is operating at 750 ◦C with the inlet gas H2O:H<sup>2</sup> <sup>=</sup> 0.9:0.1 at the cathode side and O2:N<sup>2</sup> <sup>=</sup> 0.21:0.79 at the anode side. Furthermore, the pitch width and rib width are 2 and 1 mm, respectively. The contact resistance is set at 0.056 Ωcm<sup>2</sup> as deduced by our impedance measurement. The other parameters used to reproduce the experiment are listed in Table 1. The micro-structure parameters are taken from the literature experiments [29–32].

As can be seen from Figure 2, the experimental and simulation results agree well, demonstrating the validity of the multi-physics model. Naturally, some discrepancies between the experimental and theoretical I–V curves are observed. Nevertheless, the theoretical and experimental difference is less than 0.02 V for any given current densities.

**Figure 2.** Comparison of experimental and simulated I–V curves.

#### **3. Results**

#### *3.1. Parameters Sensitivity Analysis for Rib Width Optimization at the H<sup>2</sup> Electrode Side*

As seen in Figure 3, the current density first increases to a peak value then decreases with the increase of rib width. This is quite understandable as the concentration loss increases and the ohmic loss decreases when the rib width increases. Therefore, there is an optimized rib width drib-m where the current density reaches its maximum value. When drib < drib-m, with the increase of rib width, the decrease of ohmic loss is higher than the increase of concentration loss, the current density tends to increase. When drib > drib-m, with the increase of rib width, the decrease of ohmic loss is smaller than the increase of concentration loss, and the current density decreases. The competition between the concentration polarization and the ohmic polarization determines the final optimized rib width.

**Figure 3.** *Cont.*

**Figure 3.** The effect of rib width on the SOEC performance. Parameter on the O<sup>2</sup> electrode side: (**a**) the rib/pitch ratio, (**b**) the rib-electrode ASR. Parameter on the H<sup>2</sup> electrode side: (**c**) Inlet H2O content, (**d**) Electrode porosity, (**e**) Electrode conductivity, (**f**) Electrode thickness.

As shown in Figure 3a, when the rib/pitch width ratio at the O<sup>2</sup> electrode side increases from 0.25 to 0.5 then to 0.75, the optimized H<sup>2</sup> electrode rib width is 0.752, 0.720, 0.710 mm, respectively. Notice, however, although the optimized rib widths are different, the width of 0.72 mm is sufficiently optimal for all the O<sup>2</sup> electrode rib/pitch ratio conditions. The maximum current density difference between optimal rib width and 0.72 mm are only 0.05%, 0.007% for the rib/pitch ratio of 0.25 and 0.75, respectively. Therefore, it is concluded that the optimized rib width at the H<sup>2</sup> electrode side is not sensitive to the rib/pitch ratio at the O<sup>2</sup> electrode side.

The experimentally measured ASR of SOFC is in the range 0.01–0.05 Ωcm<sup>2</sup> [33]. SOEC has the same electrode and inter-connector materials. However, the oxidation atmosphere of both the electrodes of SOEC means potentially larger oxidation risk of inter-connector than SOFC, so the maximum ASR studied here is 0.08 Ωcm<sup>2</sup> . ASR of both the electrodes in the range of 0.04–0.08 Ωcm<sup>2</sup> is used to conduct parameter sensitivity analysis to optimize rib width. To study the parameters' influence on the rib width optimization at one electrode side, ASR of another electrode side is set as 0.03 Ωcm<sup>2</sup> . As displayed in Figure 3b, when ASR between the O<sup>2</sup> electrode and rib increases from 0.01 to 0.05 Ωcm<sup>2</sup> , the current density decreases gradually because of the increased ohmic loss. The optimized H<sup>2</sup> electrode rib width is 0.689, 0.705, 0.720, 0.734, 0.746 mm when ASR at the O<sup>2</sup> electrode side are 0.01, 0.02, 0.03, 0.04, 0.05 Ωcm<sup>2</sup> , respectively. However, when choosing 0.72 mm as the optimal rib width for the other four cases, the differences between the maximum current density at the optimal rib width and the current density with a rib width of 0.72 mm are 0.0756%, 0.0165%, 0.0128%, 0.04% for ASR 0.01, 0.02, 0.04, 0.05 Ωcm<sup>2</sup> , respectively. Hence, the optimized rib width on the H<sup>2</sup> electrode side is not sensitive to the ASR on the O<sup>2</sup> electrode side.

Along the fuel flow direction, H2O is consumed gradually, the molar fraction of H2O changes correspondingly. Furthermore, for practically operating SOEC, high H2O conversion rate is favorable, so the research about the rib width optimization should take into account different H2O molar fractions. Figure 3c shows, for a fixed rib width, the current density increases with the increase of H2O molar fraction. This is because the Nernst potential decreases and the electrochemical reaction rate increases with the increased H2O molar fraction, and the current density increases when the applied voltage is fixed. The optimal rib width is 0.720, 0.687, 0.648, 0.602 mm when H2O molar fractions are 90%, 80%, 70%, 60%, respectively. Because with the increase of H2O molar fraction, current density increases, concentration polarization and ohmic polarization both increase simultaneously, which leads to a slight change in the optimized rib width. This result is also in accordance with that of SOFC [15]. Nevertheless, 0.65 mm width is sufficiently optimal for SOEC with different H2O molar fraction, because the difference between the maximum current density and the current density at the rib width of 0.65 mm is only 0.0002%, 0.0004%, 0.0001%, 0.0028% for SOEC with the inlet H2O molar fraction of 90%, 80%, 70%, 60%, respectively. The optimized rib width can be considered independent of the H2O fraction.

The SOEC H<sup>2</sup> electrode porosity is in the range of 0.3–0.5 [34]. As shown in Figure 3d, the optimal rib width increases with the increase of electrode porosity due to the reduced gas diffusion resistance. For the porosity of 0.3, 0.4, 0.5, the optimized rib width is 0.616, 0.720, 0.800 mm, respectively. Even the optimized rib width for the three porosities seems to be quite different, the maximum current densities for porosity 0.3 and 0.5 are only 0.98% and 0.35% higher than the current density with a rib width of 0.720 mm. Therefore, the optimized rib width is not sensitive to the H<sup>2</sup> electrode porosity.

As shown in Figure 3e, when the H<sup>2</sup> electrode electrical conductivity increases from 717.28 to 7172.8 S/m, the current density increases relatively drastically. However, when H<sup>2</sup> electrode electrical conductivity increases from 7172.8 to 71,728 S/m, the current density has almost no change. It means that the electrical conductivity of 7172.8 S/m is large enough, electrical conductivity larger than 7172.8 S/m has no evident improvement for SOEC performance. The optima rib width for the electrode electric conductivity of 717.28, 7172.8 and 71,728 S/m is 0.741, 0.722 and 0.719 mm, respectively. However, using a rib width of 0.720 mm, the current density differs from the maximum only by 0.03% for the conductivity of 717.28 S/m, and virtually zero for the conductivity of 71,728 and 7172.8 S/m. Hence, the optimal rib width is basically independent of the electrode conductivity.

As seen in Figure 3f, the performance of SOEC decreases with the increase of H<sup>2</sup> electrode thickness. The increase of electrode thickness means longer gas diffuse path from the channel-electrode interface to the active three phase boundary. The diffusion process becomes more difficult, and the concentration loss increases accordingly. Moreover, a thicker electrode inevitably leads to larger ohmic loss for current in the thickness direction. Hence, the overall SOEC performance decreases when electrode thickness increases. However, as the electrode thickness increases, the optimized rib width changes only slightly. The optimized rib width for the electrode thickness of 0.2, 0.4 and 0.6 are 0.719, 0.720 and 0.646 mm, respectively. However, the width of 0.720 mm is sufficiently optimal. The current density for the rib width of 0.72 mm differs only 0.006% and 0.41% from the maximum for the electrode thickness of 0.2 and 0.6 mm, respectively. Therefore, the optimal rib width is insensitive to the electrode thickness. The reason behind this is that both the concentration and ohmic polarizations increase comparatively at the same time when the electrode thickness increases.

As displayed in Figure 4a, fixing the rib width, the current density decreases significantly with the increase of rib–electrode ASR due to directly the increased ohmic polarization loss. Moreover, the optimal H<sup>2</sup> electrode rib width increases, because the increased ohmic loss due to the increase of ASR can be partially offset by increasing the rib width. As seen in Figure 4b, the optimized rib width increases almost linearly with the increase of ASR. When ASR increases from 0.01 to 0.05 Ωcm<sup>2</sup> , the optimized rib width increases from 0.511 to 0.840 mm, or an increase of 64.4%. This ASR effect

is quite dramatic. With the increase of ASR, the current density decreases, the concentration loss decreases passively. The increased ohmic loss and the decreased concentration loss collectively leads to the dramatic increase of the optimal rib width.

**Figure 4.** (**a**) The effect of ASR of the H<sup>2</sup> electrode and rib interface on SOEC performance and the optimized rib width; (**b**) The relationship between optimal rib width and ASR.

## *3.2. Parameters Sensitivity Analysis for the Rib Width Optimization on the O<sup>2</sup> Electrode Side*

As can be seen in Figure 5a, with the increase of rib/pitch ratio on the H<sup>2</sup> electrode side, the current density decreases, which can be attributed to the increased concentration polarization in the cathode due to the increased rib width. When the rib/pitch ratio increases from 0.25 to 0.5, the SOEC performance decreases slightly. However, the when rib/pitch ratio increases from 0.5 to 0.75, the SOEC performance decreases drastically. It can be inferred that when the rib/pitch ratio exceeds 0.5, the negative effect of the concentration loss increase is far greater than the positive effect of the ohmic loss decrease as induced by the increase of the H<sup>2</sup> electrode rib width. However, when the rib width on the O<sup>2</sup> electrode side increases, the current density increases all the way, independent of the rib/pitch ratio on the H<sup>2</sup> electrode side, indicating the optimized rib width on the O<sup>2</sup> electrode side is not sensitive to the rib/pitch ratio on the H<sup>2</sup> electrode side.

**Figure 5.** *Cont.*

**Figure 5.** The effect of parameters on SOEC performance and the optimal O<sup>2</sup> electrode rib width. H<sup>2</sup> electrode parameters: (**a**) rib/pitch ratio on the H<sup>2</sup> electrode side; (**b**) rib-H<sup>2</sup> electrode area specific resistance (ASR); O<sup>2</sup> electrode parameters: (**c**) porosity; (**d**) electrical conductivity; (**e**) thickness; (**f**) rib-O<sup>2</sup> electrode ASR.

As shown in Figure 5b, when ASR on the H<sup>2</sup> electrode side increases from 0.01 Ωcm<sup>2</sup> to 0.05 Ωcm<sup>2</sup> , the current density decreases gradually. For a fixed ASR, the current density increases with the increased rib width on the O<sup>2</sup> electrode side, independent of the ASR change on the H<sup>2</sup> electrode side. Hence, the optimized O2—electrode rib width is not sensitive to the ASR of the H<sup>2</sup> electrode.

Figure 5c shows the current density increases extremely slightly with the increase of O<sup>2</sup> electrode porosity from 0.3 to 0.4 then to 0.5, indicating the polarization loss induced by O<sup>2</sup> diffusion inside the anode is rather small. This is also ascribed to the thinness of the O<sup>2</sup> electrode. Furthermore, the current density increases with the increase of rib width, for all the anode porosities examined. It can be concluded that the optimal O2-electrode rib width is independent of the O<sup>2</sup> electrode porosity.

Figure 5d shows the current density increases with the increase of the O<sup>2</sup> electrode electrical conductivity, due to the decreased ohmic polarization. The current density increases more dramatically when the conductivity increases from 280.6 to 2806 S/m, but increases slightly when the conductivity increases from 2806 to 28,060 S/m, indicating the ohmic polarization is no longer a major factor limiting the cell performance when the anode conductivity is above 2806 S/m. For a fixed conductivity, the current density increases with the increase of rib width. The increase of current density with the rib width is quite large for the low conductivity of 280.6 S/m, but only moderate for the conductivity of 2806 and 28,060 S/m, indicating a conductivity of 2806 S/m is adequately high for the anode.

Figure 5e shows the current density increases with the increase of O<sup>2</sup> electrode thickness, which is opposite to the trend of the cathode shown in Figure 4f. The opposite trends imply that the major polarization factors in the anode and cathode are different. The concentration loss is more influential than the ohmic polarization in the cathode, while the ohmic polarization is much larger than the concentration polarization in the anode. The thicker anode increases the difficulty of both the O<sup>2</sup> diffusion and the current conduction along the electrode thickness direction. However, the current passage through the narrow cross section of the electrode to the rib is the main ohmic polarization loss. The increase of electrode thickness can reduce the major ohmic loss and improves the SOEC performance, as observed experimentally [19]. For all the anode thicknesses considered, the current density increases with the increased O<sup>2</sup> electrode rib width, due to the reduced ohmic polarization by the shorter conduction path.

As shown in Figure 5f, fixing the rib width, the SOEC performance decreases with the increased anode ASR due to the increased ohmic loss. Meanwhile, the current density increases continuously with the increase of rib width, independent of the ASR value, confirming the concentration loss is far lower than the ohmic loss in the O<sup>2</sup> electrode. The optimal rib width can be quite close to the whole pitch size.

#### *3.3. Analytical Expression of the Optimal Rib*/*Pitch Ratio on the H<sup>2</sup> Electrode Side*

From Figure 6, it can be seen that with the increase of pitch width, the optimal rib width increases, while the optimal rib/pitch ratio (R) decreases. Furthermore, for a SOEC with pitch width 2 mm and ASR 0.01 Ωcm<sup>2</sup> , the optimal rib width is 0.387 and 0.603 mm when loaded with applied voltage 1.6 and 1.2 V, increasing by 56%, so the influence of the applied voltage on optimal rib width cannot be ignored. The specific voltage is meaningless, the possible influence factor is the total polarization loss η = Vcell−Enerst. However, here the explored total polarization change is induced only by applied voltage change. Combining the parameters sensitivity analysis above, it can be distinguished that ASR, polarization loss and pitch width from all the parameters studied as the major factors affecting optimal rib width. Here we take the rib/pitch ratio into consideration, and analytically express the relationship between optimal rib/pitch ratio with pitch width, rib-electrode ASR, total polarization loss. Finally, we get the expression for the applicable voltage range 1.2–1.6 V for SOEC. The optimal rib/pitch ratio is denoted as R1.2V, R1.4V, R1.6V.

$$\mathbf{R}\_{12V} = (-0.4294 \times \text{ASR} - 0.03264) \times \mathbf{d}\_{\text{pitch}} + (5.2165 \times \text{ASR} + 0.33022) \tag{38}$$

$$\mathbf{R}\_{\rm LAV} = (-0.3804 \times \text{ASR} - 0.02809) \times \mathbf{d}\_{\rm pitch} + (4.7373 \times \text{ASR} + 0.27819) \tag{39}$$

$$\mathbf{R}\_{\text{L\#V}} = \left(-0.2994 \times \text{ASR} - 0.01684\right) \times \mathbf{d}\_{\text{pitch}} + \left(4.3912 \times \text{ASR} + 0.19651\right) \tag{40}$$

**Figure 6.** *Cont.*

**Figure 6.** The optimal H<sup>2</sup> electrode rib width and rib/pitch ratio for different pitch width for the cell voltage of (**a**,**b**): 1.2 V; (**c**,**d**): 1.4 V; (**e**,**f**): 1.6 V.

R can be expressed generally as R = (a1\*ASR + a2) + (a3\*ASR + a4). The total polarization for SOEC with applied voltage 1.2, 1.4, 1.6 is 0.34, 0.54, 0.74, respectively. The four sets of coefficients are listed in Table 2 and the relationship between the coefficients and η is displayed in Figure 7.


**Figure 7.** The relationship between coefficient a1, a2, a3, a4 and η.

By simplifying the relationship between a1, a2, a3, a<sup>4</sup> and η, a<sup>1</sup> = 0.325η − 0.54523, a<sup>2</sup> = 0.0395η − 0.04719, a<sup>3</sup> = −2.06325η + 5.89582, a<sup>4</sup> = −0.33428η + 0.4488 are obtained to describe the relationship between η and a1, a2, a3, a<sup>4</sup> linearly. The synthetic expression of R considering the influence of pitch width, rib—electrode ASR, and the total polarization is formulated as:

$$\begin{array}{l} \text{R} = [(0.325 \times \eta - 0.54523) \times \text{ASR} + (0.0395 \times \eta - 0.04719)] \times \text{d}\_{\text{pitch}} + \\\ [(-2.06325 \times \eta + 5.89582) \times \text{ASR} + (-0.33428 \times \eta + 0.44882)] \end{array} \tag{41}$$

It should be pointed out that in Equation (41), the units of ASR, η, dpitch are Ωcm<sup>2</sup> , V, mm, respectively. Subsequently, the optimal rib can be formulated as:

$$\begin{array}{l} \mathbf{d\_{rib,a}} = \left[ (-2.06325 \times \eta + 5.89582) \times \text{ASR} + (-0.33428 \times \eta + 0.44882) \right] \times \mathbf{d\_{pitch}}\\ + \left[ (0.325 \times \eta - 0.54523) \times \text{ASR} + (0.0395 \times \eta - 0.04719) \right] \times \mathbf{d\_{pitch}}^2 \end{array} \tag{42}$$

Here, drib,a (mm) is the optimal rib width obtained analytically. It should be noted that even the drib,a (mm) is the quadratic function of dpitch, the coefficient for the first order term is an order of magnitude larger than the quadratic term. Therefore, the relationship between optimal rib width and pitch width is mainly linear.

To verify the reliability of Equation (42), two accuracy parameters are defined as

$$
\lambda\_1 = \frac{\mathbf{i}\_{\text{max}-\mathbf{n}} - \mathbf{i}\_{\text{max}-\mathbf{a}}}{\mathbf{i}\_{\text{max}-\mathbf{n}}} \tag{43}
$$

$$
\lambda\_2 = \frac{\mathbf{d}\_{\rm rib-n} - \mathbf{d}\_{\rm rib-a}}{\mathbf{d} \mathbf{rib-n}} \tag{44}
$$

where imax-n is the maximum current density corresponding to the optimal rib width drib-n obtained numerically. imax-a is the maximum current density for the analytically predicted optimal rib width drib-a. λ1, λ<sup>2</sup> are defined to evaluate the effectiveness of the analytical expression for predicting optimal rib width. Even the maximum margin of error for optimal rib width (λ2) is about 8.575%, the current density for the analytically predicted optimal rib is only 0.14% different from the maximum current density for the numerically attained optimal rib width. Hence, the cathode rib width predicted by Equation (42) is sufficiently optimal.

#### **4. Conclusions**

We have developed a comprehensive mathematical model for the performance research of SOEC. The impacts of the electrode rib widths on SOEC performance are systematically examined by varying the contact resistance, fuel composition, electrode porosity, electrode thickness and electrode electric conductivity. Different from rib width at the H<sup>2</sup> electrode side, the optimal rib width at the O<sup>2</sup> electrode side is not so sensitive to ASR and other parameters. With the increase of rib width, current density almost increases all the way, because the concentration polarization at the O<sup>2</sup> electrode side is too small, the ohmic loss induced by ASR is overwhelmingly greater than concentration loss. Nevertheless, the current density increase rate decreases when rib width is large enough. More importantly, it finds out three main factors affecting optimal rib width at the H<sup>2</sup> electrode side: ASR at the rib–electrode interface, the pitch width, and the applied cell voltage. Giving the pitch width and contact resistance, the optimal rib width decreases with the increase of cell voltage. Giving the cell voltage and contact resistance, the optimal rib width increases and the optimal rib/pitch ratio decreases with the increase of pitch width. When ASR at H<sup>2</sup> electrode side is in the range of (0.01, 0.05) Ωcm<sup>2</sup> , the optimal rib width at the cell voltage of 1.4 V is in the range (0.511, 0.840), (0.637, 1.065), (0.736, 1.248) and (0.818, 1.410) mm for the corresponding pitch width of 2, 3, 4 and 5 mm, respectively. Finally, an analytical expression is proposed to formulate the relationship between the optimal rib width and total polarization loss, ASR, and pitch width. The prediction error between the maximum current density obtained by numerically and analytically optimal rib width is within 0.14%, proving the predictive ability of the analytical expression.

**Author Contributions:** Conceptualization, M.G.; methodology, M.G. and Z.L.; validation, M.G. and X.R.; formal analysis, M.G. and Z.L.; investigation, M.G., X.R. and G.X.; data curation, M.G. and G.X.; writing—original draft preparation, M.G.; writing—review and editing, M.G. and Z.L.; supervision, Z.L.; project administration, Z.L. and J.W.; funding acquisition, Z.L. and J.W. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the National Natural Science Foundation of China (11774324 & 12074362), the Strategic Priority Research Program of the Chinese Academy of Sciences (XDA 21080200) and the Youth Innovation Promotion Association of the Chinese Academy of Sciences (2018298).

**Acknowledgments:** The financial support of the National Natural Science Foundation of China (11774324 & 12074362), the Strategic Priority Research Program of the Chinese Academy of Sciences (XDA 21080200) and the Youth Innovation Promotion Association of the Chinese Academy of Sciences (2018298) are gratefully acknowledged.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **Hydrolysis-Based Hydrogen Generation Investigation of Aluminum System Adding Low-Melting Metals**

**Zeng Gao 1,\*, Fei Ji <sup>1</sup> , Dongfeng Cheng <sup>1</sup> , Congxin Yin <sup>1</sup> , Jitai Niu <sup>2</sup> and Josip Brnic <sup>3</sup>**


**Abstract:** In this age of human civilization, there is a need for more efficient, cleaner, and renewable energy as opposed to that provided by nonrenewable sources such as coal and oil. In this sense, hydrogen energy has been proven to be a better choice. In this paper, a portable graphite crucible metal smelting furnace was used to prepare ten multi-element aluminum alloy ingots with different components. The microstructure and phase composition of the ingots and reaction products were analyzed by X-ray diffraction (XRD), scanning electron microscopy (SEM), and differential scanning calorimetry (DSC). The reaction was carried out in a constant temperature water bath furnace at 60 ◦C, and the hydrogen production performance of the multi-element aluminum alloys in different proportions was compared by the drainage gas collection method. The experimental results show that the as-cast microstructure of Al–Ga–In–Sn aluminum alloy is composed of a solid solution of Al and part of Ga, and a second phase of In3Sn. After the hydrolysis reaction, the products were dried at 150 ◦C and then analyzed by XRD. The products were mainly composed of AlOOH and In3Sn. Alloys with different compositions react at the same hydrolysis temperature, and the hydrogen production performance is related to the ratio of low-melting-point metal elements. By comparing two different ratios of Ga–In–Sn (GIS), the hydrogen production capacity and production rate when the ratio is 6:3:1 are generally higher than those when the ratio is 7:2:1. The second phase content affects the hydrogen production performance.

**Keywords:** low melting metal; Al-based alloy; metal smelting; hydrogen production

#### **1. Introduction**

With progress in science and technology, energy comes into focus for society in terms of quality of life. As the carrier of carbon-free energy, hydrogen is not only the lightest element but also the most abundant resource in nature. Hydrogen has a very high calorific value of combustion and is a clean and efficient ideal energy source [1–9]. The hydrolysis of aluminum is an environmentally friendly reaction, and the products are pollution-free. However, it is very easy to form a compact oxide film on the surface of aluminum. Breaking the oxide film becomes a key breakthrough point for hydrogen production [10–17]. Common methods include dissolving the oxide film in an acid alkaline and neutral solution, and preparing an aluminum alloy by ball milling and by activating it [18–22]. A common chemical hydrogen production method is to store the hydrogen in a hydrogen storage tank and to then transport it. The quality of hydrogen accounts for 5–7% of the quality of the storage tank [23]. Hydrogen production from a metal ingot reaction is not only more efficient but also more convenient for transportation and storage. As one of the most common metal elements, aluminum has many advantages such as low cost, abundant reserves, and good preservation. In particular, the alumina hydroxide generated after an aluminum hydrolysis reaction not only is pollution-free but also can

**Citation:** Gao, Z.; Ji, F.; Cheng, D.; Yin, C.; Niu, J.; Brnic, J. Hydrolysis-Based Hydrogen Generation Investigation of Aluminum System Adding Low-Melting Metals. *Energies* **2021**, *14*, 1433. https://doi.org/ 10.3390/en14051433

Academic Editor: Samuel Simon Araya

Received: 7 February 2021 Accepted: 2 March 2021 Published: 5 March 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

be reused [24]. Therefore, metal aluminum is the preferred raw material for hydrogen production by hydrolysis.

A.V. Ilyukhina et al. [16] used a series of low-melting-point alloys based on the metal gallium, such as Ga70-In30, Ga70-In25-Zn3, and Ga62-In25-Sn13, in an aluminum powder alloying treatment. When the content of the liquid alloy in the alloy was 5–10 wt.%, the hydrogen production performance of aluminum powder in 25 ◦C water had a small relationship with the contents of the activator. However, the hydrogen production rate decreased significantly when its content continued to decrease. The hydrolysis rate of aluminum powder depends on the hydrolysis temperature. Fan et al. [18] prepared a type of Al-Li powder alloy by mechanical ball milling. The maximum hydrogen production rate of the alloy at room temperature was 233 mL/(min·g) and the maximum hydrogen production was 743 mL/g. After that, the Al-5.3Ga-5.4Sn-2In-7.3Zn alloy was prepared by ball milling. The hydrogen production of aluminum alloy powder reached 770 mL/(g·Al) within 7 min, and the hydrogen production rate reached 77.3%. Gai et al. [22] studied the reaction of pure aluminum with different particle sizes and water at different temperatures. For a certain reaction temperature, the smaller the particle size, the greater the possibility of reaction. M.C Roul [19] proposed an activation mechanism of Al-X alloy (X = Zn, Hg, or In) that is the well-known aluminum alloy dissolution–redeposition mechanism, which became the theoretical basis of aluminum alloy activation mechanisms.

This experiment mainly uses alloying to treat metal aluminum. This method is based on adding low-melting-point metals, such as Ga, In, Sn, Ca, Mg, Zn, Bi, etc. The main reason for choosing an alloying method is that this method can hydrolyze metal aluminum in neutral solution or aqueous solution with a pH value close to neutral, which can significantly improve the activity of aluminum. For alloying methods, common treatment methods are ball milling and the smelting method; this experiment chooses the smelting method because the smelting method has the following advantages over the ball mill method: 1. The operational method is simple. 2. It has a small material loss during the experiment. 3. The precision of alloy composition is easy to control. 4. It is easier to control the hydrolysis speed during the hydrolysis process. 5. An alloy produced after ball milling is not easy to preserve and even has safety risks [25–28]. The alloy block after smelting and casting is easier to preserve. Only aluminum itself participates in the reaction, and the low-melting-point metal can be collected and reused after the reaction. This method greatly reduces the cost of preparing hydrogen, which is of great help to the development of hydrogen production by aluminum hydrolysis and has more scientific and practical value.

This article is improved based on the above research. The experiment uses a portable graphite crucible metal melting furnace, and continuously inert gas is introduced into the melting furnace to prevent oxidation. In such experimental conditions, to achieve a high rate of hydrogen production and to obtain ideal hydrogen production, a multielement aluminum alloy was formed by adding low-melting-point metals (Ga, In, and Sn) in different proportions. Then, the content of aluminum in the alloy is changed to compare the influence of alloy composition on hydrogen production. Then, X-ray diffraction (XRD), differential scanning calorimetry (DSC), scanning electron microscopy (SEM), and other characterization methods are used for correlation analysis and to further study the phenomena involved.

#### **2. Materials and Methods**

#### *2.1. Alloy Preparation*

This study used a portable graphite crucible metal smelting furnace to prepare multiple aluminum alloys. The raw materials were industrial pure Al plates (99.99%), Ga blocks (99.99%), In particles (99.99%), and Sn particles (99.99%). The melting points of the metals are shown in Table 1.

**Table 1.** The melting points of the metals.


In the experiment, 10 types of Al-Ga-In-Sn aluminum alloy ingots with different composition ratios were prepared using the metal smelting method (sample numbers 1–10). We weighed a total of 20 g of different alloying elements and mixtures of different mass ratios into a custom-sized cylindrical quartz crucible and then put the quartz crucible into the melting furnace, continued to pass CO<sup>2</sup> into the furnace, and set the melting temperature of the melting furnace to 850 ◦C. The smelted alloy ingots were placed in sealed sample bags, and these sample bags were placed in a large amount of discolored silica gel particles to reduce oxidation. If necessary, we cut the ingot appropriately to obtain the appropriate size for later experiments. The alloy chemical compositions for experiment are shown in Table 2.


**Table 2.** Alloy compositions used for experiments.

#### *2.2. Observation of Phase Structure and Microstructure*

The Merlin Compact scanning electron microscope (SEM) and the OXFOFD energy spectrometer (EDS) attached to a microscope were used to analyze the microstructure and composition of the alloy ingot and the product after the hydrolysis reaction. In addition, SmartLab (9 kW) X-ray diffraction for phase analysis was used, with Cu Kα as the radiation source, while other details were as follows: the scanning speed was 10–80◦ , the step size was 0.2◦ , and the acquisition and scanning speed was 10◦/min. The thermodynamic monitoring and analysis of alloy ingots were analyzed by a Setaram Evolution 2400 thermal analyzer (TG-DSC). The measurement temperature range was 23–615 ◦C, and the scanning speed was 5 ◦C/min.

#### *2.3. Test on Hydrolysis Performance of Aluminum Alloy*

The test can be described as follows. We put 200 mL of tap water into a three-necked flask with a volume of 500 mL, placed it in an electronic constant temperature water bath furnace, and set the temperature of the water bath furnace to 60 ◦C. We cut out a 1 g sample and put it in the flask, then used the drainage method to calculate the amount of hydrogen generated, used an electronic weighing accuracy of 0.01 g to weigh the collected water, and used Equation (1) to convert the volume of hydrogen. The proportion of the sample was measured 3 times under certain conditions, and the final average value was taken. A schematic diagram of the hydrogen production reaction device is shown in Figure 1.

**Figure 1.** Schematic diagram of hydrogen production performance test device.

In reaction Equation (1), V is the volume of hydrogen generated, m is the mass of discharged water, and ρ is the density of water. We used Equation (2) to calculate the hydrogen production conversion rate of the alloy at different ratios. In the formula, R1 is the hydrogen production conversion rate, V is the actual hydrogen production volume, and VT is the theoretically calculated hydrogen production volume. The volume of 1 mol H<sup>2</sup> in standard state is 22.4 L, and the volume of H<sup>2</sup> produced by 1 g of aluminum is 1245 mL. The experiment was carried out at room temperature and atmospheric pressure (1 atm and 25 ◦C), and the volume of 1 mol H<sup>2</sup> under this condition was 24.45 L. The theoretical volume was 1358.4 mL of H<sup>2</sup> produced by 1 g of aluminum.

$$\mathbf{V} = \mathbf{m}/\mathfrak{p}\_{\prime} \tag{1}$$

$$\text{R1} = \text{V/VT} \times 100\%\_{\text{t}} \tag{2}$$

The hydrogen production performance data were taken from the average of three experimental data, and the changes in hydrogen production and hydrogen production rate of multi-element aluminum alloys under different proportions were explored and rationally analyzed. After the reaction, the reactant obtained was dried in a drying oven at 150 ◦C before proceeding to the next step of analysis.

#### **3. Results and Discussion**

#### *3.1. SEM Observation and Analysis*

In order to study the microstructure of the alloy ingots, scanning electron microscopy and energy spectrum tests were carried out on the multi-element aluminum alloy ingots with different proportions. The sample was highly active and easily oxidized, so it needed to be quickly put into the sample table and vacuumed. It can be seen from Figure 2 that, under the microscopic conditions, when the proportion of Ga-In-Sn (GIS) is 50 wt.%, the surface structure appears granular. As the proportion of low-melting-point metals in the alloy decreases, the alloy surface becomes less grainy and the surface becomes smoother dense and slatted. As the alloy is solidified and formed after natural cooling in the molten state, a large amount of internal stress in the alloy leads to a fracture of the alloy during the nucleation process, resulting in a large number of voids and cracks in the alloy.

**Figure 2.** Scanning electron microscopy (SEM) image at 5000× of the aluminum alloy when the ratio of Ga-In-Sn is 7:2:1 and 6:3:1: (**a**) 50 wt.%Al-50 wt.%Ga-In-Sn (GIS) (7:2:1), (**b**) 50 wt.%Al-50 wt.%GIS (6:3:1), (**c**) 80 wt.%Al-20 wt.%GIS (7:2:1), and (**d**) 80 wt.%Al-20 wt.%GIS (6:3:1)).

It can be seen from the EDS surface scan results in Figure 3 that a large amount of off-white low-melting alloy phases are scattered on the grain boundary surface of the alloy. Its main component is composed of low-melting-point metal Ga, followed by a small amount of In, Sn, and Al. Combined with the EDS surface scan, it can be observed that the distribution of elements in the alloy is relatively uniform, but there is still a certain degree of segregation. One of the main reasons for this phenomenon is that the solubility of the alloy decreases in the solid state. According to the alloy phase diagram, the degree of intermetallic compound formation is limited. Therefore, segregation occurs in a local area of the alloy. The second reason is that only a small amount of low-melting-point metal forms a solid solution with Al when the temperature drops. Large amounts of Ga, In, and Sn exist in the α-Al phase as segregation. According to Figure 2d, in addition to the spherical low-melting-point alloy phase, there are other alloy phases with different sizes. There are also a lot of low-melting metals in the gap. The main reason for the above phenomenon is that the metal aluminum solidifies in the form of dendrite during solidification. At the same time, the low-melting-point metal has limited solid solubility in aluminum, which leads to the liquid low-melting-point metal being squeezed into the cracks of aluminum grain. With the continuous decrease in temperature, the gap phases of different sizes are solidified and precipitated out. The size and shape of the gap phase are related to the proportion of low-melting-point metal in the alloy. The larger the proportion is, the more brittle the alloy, the more easily it is broken, and the larger the size of the gap phase.

**Figure 3.** Scans of 80 wt.%Al-20 wt.%GIS (6:3:1) EDS surface of the alloy: (**a**) SEM diagram of the aluminum alloy at 10,000×, (**b**) EDS hierarchical image, (**c**) Al layer, (**d**) Ga layer, (**e**) In layer, and (**f**) Sn layer.

The microstructure of the reaction product after the hydrolysis reaction is shown in Figure 4. Observation at 1000× times shows that the morphology of the hydrolyzed product is lamellar, agglomerating together in a massive form. Compared with the alloy particles before the reaction, the degree of fragmentation is increased and a large amount of the internal structure of the particles is dispersed due to progress in the hydrolysis reaction, showing the shape of needles and phosphorus flakes. At a high magnification of 10,000×, it can be observed that the hydrolyzed product has a large number of pores, which may be due to the release of a large amount of hydrogen from the aluminum–water reaction.

**Figure 4.** SEM images of the 50 wt.%Al-50 wt.%GIS (6:3:1) reaction product at (**a**) 1000× and (**b**)10,000×.

#### *3.2. XRD Examination*

Figure 5 is the XRD pattern of two groups of ingots with different GIS ratios. It can be seen from the figure that there are three characteristic peaks, and the four sharper characteristic peaks are characteristic peaks of Al, the strength of which is significantly higher than that of other phases. There is no obvious characteristic peak of Ga because Ga enters into the lattice of Al to form a solid solution. The characteristic peak is covered by the characteristic peak of Al. With the increase in Ga content, the characteristic peak of Al has an obvious phenomenon of left deviation. The peaks of In3Sn and In are relatively weak. When the ratio degree of In-Sn is 3:1, it is concluded that there are more second phases on the alloy surface according to SEM diagram observation and EDS component analysis, and the second phase is In3Sn combined with the XRD results. When the ratio of In-Sn is 2:1, the characteristic peak of In3Sn cannot be detected but the characteristic peak of weak In can be detected. After hydrolysis reaction, some spherical droplets can be observed after the hydrolysate is dried. Therefore, the hydrolysate was further analyzed by X-ray diffraction, and the results are shown in Figure 6. The obvious characteristic peak of In3Sn can be seen in the figure, which proves that the liquid alloy phase does exist in this aluminum alloy. It is because of this liquid alloy phase that aluminum can be continuously solvated in liquid phase. Finally, it can diffuse freely and be transported to the surface of the alloy to make contact with water to produce hydrogen by hydrolysis reaction.

**Figure 5.** Alloy ingot X-ray diffraction (XRD) with different GIS contrasts.

**Figure 6.** GIS XRD map of hydrolysate at 6:3:1.

Different drying products were generated from aluminum alloy hydrolyzed products at different drying temperatures. Al(OH)<sup>3</sup> was generated when the drying temperature was lower than 72 ◦C. The drying product is AlOOH in the range of 72–172 ◦C. When the drying temperature is greater than 172 ◦C, the hydrolysis product is Al2O3. In this experiment, the drying temperature was 150 ◦C, so the characteristic peak detected by X-ray diffraction was AlOOH.

#### *3.3. DSC Analysis of Alloy Ingot*

Figure 7 shows the DSC heating curve of Sn alloy samples with the ratio of 80 wt.%Al-12 wt.% Ga-6 wt.%In-2 wt.%Sn alloy. The test temperature range was 23–615 ◦C, and the heating rate was 20 ◦C/min. During the heating process, the alloy has an endothermic peak due to melting at the melting point. In the figure, a small endothermic peak can be observed around 30 ◦C. According to the Al-Ga binary phase diagram, the eutectic temperature of the Al-Ga binary alloy is 26.6 ◦C. The temperature here is close to the eutectic temperature of the Al-Ga binary alloy. The phase transition occurred in 46 ◦C alloy, and it is speculated that the liquid phase is eutectic formed by gallium, indium, and tin alloy with a low melting point. After that, there is a weak characteristic peak at 142 ◦C. According to a In-Sn binary phase diagram, the melting point range of In3Sn is relatively large, which is about 120–143 ◦C. Combined with EDS component analysis and XRD analysis, it is concluded that the characteristic peak should be caused by the formation of intermetallic compound In3Sn. As the temperature rises, there is no obvious change from 200 ◦C to 500 ◦C until an obvious endothermic peak appears at 605 ◦C. A more sharp peak corresponds to the melting point value of the alloy, which should be the melting point of the aluminum-based solid solution.

#### *3.4. Analysis of Alloy Hydrogen Production Performance*

Woodall et al. [29] first studied the optimization of the hydrogen production performance of aluminum alloys using low-melting point metals and proposed the diffusion activation mechanism of aluminum alloy ingot hydrolyzed to produce hydrogen. The essence of the mechanism is the eutectic reaction between the low-melting-point metal, with aluminum as the driving force. The aluminum atoms at the grain boundaries are wrapped by the liquid metal, resulting in the rupture of the dense oxide film, which can contact water and undergo a hydrolysis reaction. The low-melting-point metal does not participate in this process. The hydrolysis reaction produces a concentration difference with the continuous consumption of aluminum, and the unreacted aluminum atoms continue

to diffuse from the aluminum lattice into the liquid alloy until the aluminum is basically consumed by the hydrolysis reaction. The quantity and rate of hydrogen production are two important indexes to measure the hydrogen production performance of alloys.

In order to test the effect of changing the ratio degree of low-melting alloy and increasing the content of the second-phase In3Sn on the hydrogen production performance, the hydrolytic hydrogen production test was carried out on the multi-component aluminum alloy with different contents in a constant temperature water bath at 60 ◦C. Figure 8 is the comparison diagram of the hydrogen production performance of aluminum alloy under different metal ratios at low melting points.

**Figure 7.** Temperature rise curve of differential scanning calorimetry (DSC) of alloy samples.

**Figure 8.** Comparison of the hydrogen generation performance of Al-Ga-In-Sn alloys: (**a**) hydrogen production comparison with GIS ratio 7:2:1, (**b**) hydrogen production comparison with GIS ratio 6:3:1, (**c**) hydrogen production rate comparison with GIS ratio 7:2:1, and (**d**) hydrogen production rate comparison with GIS ratio 6:3:1.

Figure 8a–d show the hydrogen yield and hydrogen production rate curves of different Al contents hydrolyzed in water at 60 ◦C when the ratios of GIS (Ga-In-Sn) are 7:2:1 and 6:3:1. As can be seen from Figure 8a,c, when GIS is 7:2:1, hydrogen production and hydrogen production rate are the highest when Al content is 80%, and only when Al content is 80% and 90%, the reaction is basically complete within 20 min while the reaction time of other contents is relatively long. In Figure 8b,d, the GIS is 6:3:1. According to previous detection and analysis, when In and Sn exist in the alloy at a ratio of 3:1, the possibility and content of the second-phase In3Sn are improved. It can be clearly observed in the figure that, although the change in Al content in the alloy affects the proportion of low-melting-point alloy in the multi-alloy, the alloy basically reacts completely within about 20 min. Compared with the influence of the proportions of two different low-melting metals on Al content of 90%, the low-melting metals only accounted for 10% at this time. When the GIS was 7:2:1, the hydrogen production was reduced and the hydrogen production rate was only 80.96% due to the decrease in low-melting metal content. The maximum hydrogen production rate was up to 157 mL/g min and the hydrogen production rate was up to 97.99% when GIS was 6:3:1.

Figure 9a,b are the comparison diagrams of hydrogen production and hydrogen production rate when the Al content is 50% and the Al content is 90% under different GIS ratios. It can be seen from the figure that the hydrogen production, hydrogen production rate, and maximum hydrogen production rate when the GIS ratio is 6:3:1 are significantly higher than those when GIS ratio is 7:2:1, regardless of the proportion of low-melting point metal in the alloy being the highest (50 wt.%) or the lowest (10 wt.%). Considering the improvement in hydrogen production performance and the reduction in production cost, In and Sn can be used to share the cost of expensive Ga when the GIS ratio is 6:3:1. Under these conditions, aluminum can be hydrolyzed sufficiently even if the content of low-melting-point metal is reduced, so that the whole reaction can reach a faster reaction rate and can obtain the ideal hydrogen production rate.

**Figure 9.** Comparison of hydrogen production performance of Al50 and Al90 at different GIS ratios: (**a**) hydrogen production comparison chart and (**b**) hydrogen production rate comparison chart.

#### **4. Conclusions**

It is a safer and more environmentally friendly hydrogen production technology to hydrolyze aluminum after alloying. High purity hydrogen is not only a good alternative to fossil fuels but also an ideal hydrogen source for fuel cells. It is an important research direction to produce hydrogen immediately and to supply hydrogen on demand. In this work, multi-element aluminum alloy was smelted in a portable metal smelting furnace with CO<sup>2</sup> continuously introduced, in which the ratios of low-melting-point metals Ga, In, and Sn were 7:2:1 and 6:3:1. The alloy ingot was hydrolyzed in a constant temperature water bath furnace at 60 ◦C within 24 h after melting and casting. The hydrogen production

properties of the alloys with different proportions were compared. Combined with SEM, EDS, XRD, and DSC for further analysis, the observation results are as follows:


**Author Contributions:** Conceptualization, Z.G.; methodology, F.J. and C.Y.; investigation, Z.G., F.J., and D.C.; writing—original draft preparation, F.J.; writing—review and editing, Z.G. and J.B.; supervision, J.N. and J.B. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the National Natural Science Foundation of China (grant No. 51245008) and by the Science and Technology Project of Henan Province, China (No. 202102210036).

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Not applicable.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**

