Multi-Stage Catalyst to Prevent Hydrogen Explosions in Liquefied Hydrogen Leakage and Self-Ignition by Reaction Heat

2.1. Catalyst preparation

The hydrogen oxidation catalysts used in this study were prepared by supporting precious metals (Pt) and Fe on catalyst supports. Two types of supports were employed: a cerium–zirconium–yttrium composite oxide (Ce_0.49Zr_0.46Y_0.05O_2-δ, hereafter referred to as CZY), developed by Daihatsu Motor Co., Ltd for automotive applications, and γ-Al₂O₃. Prior to precious metal loading, the supports were coated onto a designated base material (honeycomb-structured ceramics) using a slurry coating method, followed by calcination in air to ensure thermal stability (Figure 1).

Figure 1

For CZY, the coated support was calcined in air at 1000°C for 5 hours, resulting in a specific surface area (SSA) of 26.3 m^–2 ∙ g^–1. This post-treatment material is referred to as CZY-L. The SSA of γ-Al₂O₃ was 138.6 m^–2 ∙ g^–1, while that of untreated CZY was 109.1 m^–2 ∙ g^–1. Precious metal loading was carried out by dropwise addition of dinitrodiammine platinum nitrate [Pt(NO₂)₂(NH₃)₂] and ferric nitrate [Fe(NO₃)₃ ∙ 9H₂O] onto the supports. The catalysts were then dried and calcined to complete the preparation process. The specifications of the supported metals are shown in Table 1. The values in parentheses indicate the weight percentages of the supported metals. The precious metal concentrations were varied according to the configuration of each catalyst layer.

2.2. Experimental setup and environment

Catalyst performance evaluation was conducted using the REKO-1 test facility installed at Forschungszentrum Jülich in Germany (Reinecke et al., 2011). REKO-1 introduces hydrogen and air (oxygen) from the bottom of a stainless-steel cylindrical chamber with a 70 mm inner diameter. This generates an upward flow simulating a hydrogen leakage environment. We developed a reactor system with three layers of vertically stacked honeycomb catalysts, installed it inside the chamber, and conducted tests.

The honeycomb catalysts have a diameter of 65 mm, a thickness of either 5 mm or 10 mm, and a cell density of 30 cpsi (cells per square inch). They were arranged in three layers with a 20 mm spacing. To distinguish catalyst thermal runaway from gas-phase self-ignition, temperatures were monitored independently at the catalyst beds and in the gas phase. Thermocouples inserted into or near each catalyst layer were used to evaluate local heat generation caused by catalytic hydrogen oxidation, whereas separate thermocouples placed in the inlet and outlet gas streams were used to detect rapid gas-phase heating.

For the evaluation of catalytic activity, a thermal conductivity detector (TCD) and a paramagnetic oxygen analyzer were installed at the gas outlet downstream of the catalyst to monitor hydrogen and oxygen concentrations, respectively.

The test gases were introduced at room temperature, and the hydrogen concentration in the gas mixture was gradually increased to monitor the catalyst’s reaction response and temperature rise behavior.

2.3 Catalytic activity evaluation

In the activity evaluation, the hydrogen concentration was incrementally increased every two minutes under atmospheric conditions, while monitoring the temperature responses of each catalyst layer and the changes in hydrogen concentrations in the exhaust gas (Figure 2). The hydrogen flow rate was adjusted according to the mixing ratio with air, and the gas flow velocity was kept constant at 0.25 m ∙ g^–1. The reaction proceeded in an upward direction from the bottom of the reactor, and the hydrogen concentration at which each catalyst layer initiated activation, along with the corresponding temperature rise trends, was carefully recorded. This condition was not intended to reproduce the near-field jet velocity immediately after LH2 leakage. Instead, it was selected as a controlled forced-flow condition representative of hydrogen-containing gas entering a PAR catalyst section after dispersion and air entrainment. At this velocity, the residence time in each honeycomb segment was very short, approximately 0.02 s for a 5 mm catalyst and 0.04 s for a 10 mm catalyst. Furthermore, it should be noted that these values are based on the inlet condition at room temperature. In reality, the intense reaction heat of hydrogen oxidation causes rapid gas expansion, which locally accelerates the flow velocity and makes the actual residence time even shorter. This makes the condition extremely demanding in terms of reaction completion and heat management (Krenz et al., 2025). The velocity was therefore kept constant in order to isolate the effects of hydrogen concentration, catalyst composition, and layer arrangement on temperature rise and self-ignition behavior.

Figure 2

In this study, the onset of self-ignition was defined as the point when the inlet gas temperature exceeded 200°C and showed a sharp, nearly vertical increase. It should be noted that when ignition occurs spontaneously during the experiment, it produces a sound resembling that of a low-pitched woodwind instrument and causes the gas pipe to vibrate, making it easily recognizable. Our previous research demonstrated that self-ignition in multilayer catalysts originates at the chamber inlet side (Tanaka et al., 2025b). To address this issue, this research evaluated the safety of a multi-layer catalyst configuration in which the precious metal concentration in the front layer was minimized to promote a gradual and controlled progression of the oxidation reaction. It should be noted that direct optical flame visualization was not performed in this study. Therefore, the term self-ignition refers to an operationally defined gas-phase ignition event based on the independent gas-temperature measurement and its sharp transient behavior.

Through this procedure, the minimum hydrogen concentration at which spontaneous ignition occurred for each catalyst configuration was clearly determined, aiming to establish safety design guidelines for catalyst structures.

3.1. Comparison of reaction dispersion based on identical catalyst structures with different support designs (Al₂O₃ vs. CZY-L)

In Set #01, the rear layer contained the highest platinum loading (1.0 wt%) and the greatest thickness (10 mm), making it the most reactive component within the structure. The temperature of the rear layer began to increase rapidly at approximately 100 seconds, corresponding to a hydrogen concentration of 2.0%, and was accompanied by a sharp decrease in outlet hydrogen concentration. This indicates that hydrogen oxidation initiated predominantly in the rear layer. In contrast, the middle (2nd) layer exhibited a delayed temperature increase, suggesting that its activation was thermally induced by the reaction heat generated in the rear layer, rather than by its own intrinsic catalytic activity (Figure 3).

Figure 3

On the other hand, although Set #02 used the same platinum loading and layer thickness in the rear, it utilized CZY-L as the support in all three catalyst layers. In this case, while the rear layer activated at 100 seconds, the middle and front layers were activated later, around 1000 and 1100 seconds, respectively (Figure 4).

Figure 4

It has been confirmed that the specific surface area of CZY-L (26.3 m^–2 ∙ g^–1) is smaller than that of Al₂O₃ (138.6 m^–2 ∙ g^–1). At the same metal loading, a larger specific surface area means more effective active sites directly involved in the reaction. When reaction heat from the downstream rear layer propagates upstream, the Al₂O₃-based middle layer quickly activates. It initiates the hydrogen oxidation reaction at a relatively low thermal energy level. Thus, it can be inferred that the middle layer begins to share the reaction load of the entire system at an early stage. In contrast, in the CZY-L configuration, unreacted high-concentration hydrogen is considered to pose a direct load on the front layer. Consequently, high metal dispersion was demonstrated to promote a smooth, stepwise transition of the reaction zone.

3.2. Verification of a stepwise reaction configuration with rear-dominant layer and Fe-assisted middle layer

In this experiment, the rear catalyst layer was the first to activate at approximately 200 seconds, corresponding to a hydrogen concentration of 2.0%. This activation was marked by a sharp temperature increase and a significant drop in outlet hydrogen concentration, indicating that hydrogen oxidation was initiated predominantly in the rear layer. It served as the primary reaction zone and the main source of heat generation during the early stages of the experiment.

The middle layer, which contained only Fe and no precious metals, was initially assumed to function as a thermally inert buffer without catalytic activity. However, the experimental data revealed a distinct secondary event: at around 1000 seconds, the middle layer exhibited a sharp rise in temperature, accompanied by a further notable decrease in outlet hydrogen concentration. This simultaneous change suggests that the Fe-loaded middle layer contributed to hydrogen oxidation under elevated temperature conditions (Figure 5).

Figure 5

A possible mechanism is that the heat generated in the rear layer gradually raised the temperature of the entire system. Once the middle layer reached a critical threshold, Fe may have started to participate in thermochemical redox reactions. While Fe has limited activity at low temperatures, it can act as an oxygen carrier under high-temperature and hydrogen-rich conditions (Gamisch et al., 2022). Through reversible oxidation and reduction, Fe may have supplied reactive oxygen to support hydrogen oxidation. In this way, Fe could have played a supplemental role in sustaining the reaction without the use of precious metals.

3.3. The influence of inlet-side catalyst composition on hydrogen oxidation and auto-ignition behavior

To clarify the influence of the placement of Pt–Fe composite catalysts and Fe-only catalysts on hydrogen oxidation behavior and auto-ignition concentration, catalyst configurations in Sets #04 and #05 were compared. The two sets differ only in the placement of catalysts between the front and middle positions.

In the temperature profiles, Set #04 showed dominant activity in the rear layer, while temperature rise in the front layer was suppressed until the final stage (Figure 6). In contrast, Set #05 exhibited a stepwise temperature increase not only in the rear but also in the middle and front layer (Figure 7). Notably, the front layer showed a rapid temperature rise at hydrogen concentrations above 6.0%. As a result, the auto-ignition concentration was 9.5% in Set #04 and 8.5% in Set #05. This indicates that placing a Pt–Fe composite catalyst with a high noble metal loading at the front position tends to promote earlier reaction initiation as hydrogen concentration increases. In Set #05, the front layer acted as an active site and caused a sharp temperature increase. This behavior is also supported by comparison with Set #03. On the other hand, in Set #04, the front layer consisting of Fe(6.0)/Al₂O₃ showed little to no activity even under high hydrogen concentration. As a result, heat transfer was suppressed near the inlet, and the reaction mainly progressed toward the outlet side.

Figure 6

Figure 7

These results demonstrate that not only the type and loading of the catalyst but also its placement significantly affects the auto-ignition behavior. For designs aiming to suppress auto-ignition, placing highly active catalysts near the inlet may pose a risk. Conversely, to promote early ignition, placing Pt–Fe composite catalysts with high metal loading near the inlet is effective.

3.4. Effect of different impregnation methods (comparison with 2023 results)

In the 2023 experiments, Set #06 used a sequential impregnation technique for the middle and front layers, wherein Pt (0.015 wt% and 0.001 wt%, respectively) was first impregnated and calcined, followed by dropwise addition of Fe solution (0.985 wt% and 0.999 wt%). The rear layer, composed of Pt–Fe/CZY-L (10 mm thick), was the first to activate at approximately 40 seconds, marked by a rapid temperature increase. This behavior was attributed to the cooperative effect of Pt and Fe. Subsequent temperature rises in the middle and front layers at 400 seconds (H₂: 4.0%) and 700 seconds (H₂: 7.0%), respectively, indicated a clearly staged, spatially distributed reaction progression across all three layers (Figure 8).

Figure 8

In contrast, Set #01 utilized a co-impregnation method in which Pt and Fe were simultaneously supported from a mixed solution onto all three layers, with Al₂O₃ as the support. This configuration led to early activation of the rear layer at a hydrogen concentration of 2.0%, followed by limited thermal response in the middle layer and negligible activity in the front layer. The reaction predominantly occurred within the rear layer, indicating a highly localized and concentrated reaction pattern.

Inagawa et al. (2023) reported that alloying Pt with Fe promotes the rapid switching of adsorbed species on Pt nanoparticle surfaces. This effect strongly accelerates both oxidation and reduction reactions. Although this alloy configuration exhibits extremely high catalytic activity, it inherently induces excessive localized heat generation when exposed to high-flow, high-concentration hydrogen streams. Furthermore, Kim et al. (2012) reported that the order of Fe introduction in PtPd/Al₂O₃ oxidation catalysts strongly affects the interactions among noble metal particles, iron oxide species, and the Al₂O₃ support. In particular, Fe introduced after noble metal loading promoted the formation of iron oxide species with high oxygen mobility near the noble metal phase, whereas Fe introduced before noble metal loading preferentially interacted with Al₂O₃ to form Fe/Al₂O₃, resulting in a decrease in surface area.

Based on these findings, the co-impregnated Pt–Fe catalyst in the present study is considered to favor closer contact or stronger interaction between Pt and Fe species, possibly leading to an alloy-like or highly interacting state (Ma et al., 2025). In contrast, sequential impregnation may allow Pt and Fe species to remain more spatially separated, enabling Fe oxide species to act as oxygen donors or redox promoters during hydrogen oxidation (Figure 9).

Figure 9