Preparation of Pt/ γ -Al2O3 catalyst coating in microreactors for catalytic methane combustion

A catalyst preparation method, consisting of slurry washcoating with γ -Al 2 O 3 followed by impregnating pla- tinum on the microreactor walls, has been investigated. The effect of various factors in the preparation procedures on the adhesion of the washcoated γ -Al 2 O 3 was studied, including the slurry property (i.e., the binder type, concentration and molecular weight, γ -Al 2 O 3 concentration and particle size, pH), and the (micro)reactor substrate and channel shape. The results show that the adhesion of γ -Al 2 O 3 washcoat strongly depended on the slurry rheological characteristics. A good adhesion on FeCrAlloy substrates was obtained using the slurry with polyvinyl alcohol as the binder (typical concentration at 3–5wt% and molecular weight of 57,000–186,000), 20wt% γ -Al 2 O 3 (particle size being around 3µm) and pH=3.5. FeCrAlloy as the substrate exhibited an excellent coating adhesion in rectangular or round channels, primarily due to the formation of alumina film over the surface during thermal pretreatment. The aluminum-free stainless steel as the substrate only showed a good adhesion in a round channel. Well-adhered Pt/ γ -Al 2 O 3 catalysts were then applied in a microreactor comprising parallelized microchannels made of FeCrAlloy under the optimized coating procedures, and investigated in terms of its performance in the catalytic methane combustion. It is shown that the reaction temperature has a greater influence on the methane conversion than the flow rate, and a favorable coverage of methane and oxygen on the catalyst surface is essential to obtain a good catalytic performance. These reaction results are in line with the temperature behavior measured along the microreactor.


Introduction
Microreactors, with characteristic channel dimensions on the order of ca. 1 mm or below, have shown great potentials in intensifying reaction processes in comparison to conventional reactors (e.g. fixed-bed reactors for solid-catalyzed gas phase and multiphase reactions) [1][2][3]. The specific surface area of microreactors could reach 10,000 to 50,000 m 2 m −3 [4,5], compared to around 100 m 2 m −3 for traditional reactors, thereby leading to substantially enhanced mass and heat transfer rates [6,7]. The improved mass transfer can lead to enhanced reaction rates (e.g., in the presence of multiphase fluids or solid catalysts). The heat transfer enhancement avoids the presence of temperature hot spots due to the local accumulation of reaction heat in the microreactor, a feature especially favorable for handling strongly exothermic reactions or coupled exothermic and endothermic catalytic reactions [8]. Moreover, the modular, flexible and compact microreactor design allows an easy upscaling from the laboratory to the industrial scale aiming at mass production [4]. One challenge is how to deliver reaction fluids, as well as heat transfer fluids (if present), uniformly across a multitude of microchannels in order to ensure that the optimal microreactor performance is not lost during the numbering-up process (i.e., a replication of basic microchannel units) [9][10][11].
Solid catalysts (e.g., noble metal-based) can be incorporated into microreactors as wall coatings, and γ-Al 2 O 3 represents one of the most common catalyst supports in use. Thus, a successful application of microreactors for solid-catalyzed catalytic reactions strongly depends on the formation of a well-adhered and uniform catalyst layer on the microreactor substrate [12,13]. Based on the properties of the catalyst and substrate surface, various methods have been used to deposit catalysts onto the microreactor wall, as summarized in some reviews [12,14,15]. Different coating methods and the corresponding preparation procedures have been investigated, including suspension [16][17][18], sol-gel deposition [19,20], electrophoretic deposition [21,22], chemical vapor deposition [23,24], physical vapor deposition [25,26], etc. The suspension technique (or slurry washcoating) is one of the most commonly used methods for depositing catalysts in microreactors, which typically consists of the following steps: microreactor surface pretreatment, slurry preparation and deposition, impregnation of the catalytically active component (e.g., noble metal), heat treatment (e.g., drying and calcination). Sometimes, a primer coating is used before the slurry deposition to further enhance the coating adhesion. Alternatively, the catalytically active component can be mixed with the prepared slurry prior to the coating step, allowing to attach the entire catalyst onto the microreactor wall in one step. It has been reported that the slurry properties (e.g., the binder nature, viscosity, pH and particle size) could greatly influence the rheological properties of the washcoated layer as well as its adhesion [27]. A brief literature survey of key factors in these steps is presented below, including the substrate pretreatment, nature of binder, pH and particle sizes used in the slurry for washcoating, and primer coating.
The thermal pretreatment of the (microreactor) substrate is the most commonly used method to increase the substrate surface roughness. In the case of Al-containing substrates, such pretreatment also causes the migration of aluminum species over the substrate surface to form a thin film, effectively enhancing the adhesion between the coating layer and substrate. For instance, the thermal pretreatment of FeCrAlloy material could be conducted in a range of 800 to 1000°C for 5 to 10 h [28][29][30]. It has been reported that alumina crystals appeared at about 700°C. When the calcination temperature was increased to 900°C, the needle-like α-Al 2 O 3 was substantially formed and tended to cover the whole surface. This is beneficial for anchoring the deposited washcoated layer. The rapid growth of α-Al 2 O 3 grains and the agglomeration occurred at 1000°C, forming the globular morphology. Other pretreatment methods such as the anodic oxidation on aluminum or aluminum alloy plates [31,32] and the chemical pretreatment on substrates made of aluminum, FeCrAlloy or stainless steel [33,34] were also reported.
The suspension method was characterized typically by adding the binder into a slurry (e.g., of γ-Al 2 O 3 ), in order to finally form a smooth and well-adhered layer onto the walls of microreactors [35,36]. Binders are generally divided into two categories: organic binder (e.g., polymers) and inorganic binder (e.g., boehmite and colloidal alumina). The effect of different binders on the coating adhesion was investigated by Germani et al. [37]. A compromise between the shrinking behavior and particle packing of the coating could be reached using methyl 2-hydroxyethyl cellulose (Tylose) as the binder which has a better thickening effect compared to polyvinyl alcohol (PVA). Besides the binder structure, its molecular weight and concentration may also have a noticeable effect on the slurry viscosity and subsequently the prepared coating adhesion. An excessive addition of organic binders could result in cracks during calcination due to the polymer thermal decomposition [37]. Details on the binder characterization may be found elsewhere [38][39][40].
For washcoating of γ-Al 2 O 3 , the slurry stability also depends on whether γ-Al 2 O 3 particles are well dispersed in the suspension. The viscosity and stability of the slurry greatly increase with the increasing pH value. This is because the net surface charge of particles is changed by adjusting the pH [41]. The initial particle size of γ-Al 2 O 3 is also a key factor that affects the adhesion strength. Particle sizes of 2-5 µm were reported to be a suitable range to obtain a stable slurry [27,39,42], because a smaller particle size improved the affinity and interlocking between particles in the coating.
Apart from directly applying slurry onto the microreactor wall, an intermediate layer of primer coating can be deposited in a previous step, acting as a link to enhance the adhesion between the substrate and the slurry deposition. Moreover, the difference in thermal expansion coefficients between the washcoated layer and the substrate could be decreased to some extent by applying primer coating in between. Various primer compositions have been investigated and optimized [17,43]. In order to further improve the coating adhesion on Al-free substrates like stainless steel, the substrate could be treated by the pack aluminisation technique before the primer coating step [44,45]. A thin alumina layer is expected to form on stainless steel by the diffusion of metal atoms over the surface.
Microreactors with wall-coated catalysts are promising for carrying out the catalytic combustion of methane efficiently. Compared with the conventional flame combustion, the catalytic methane combustion has been performed towards achieving a lower light-off or working temperature, less exhaust emission and more stable combustion [46]. Catalyst development, reactor design and operational conditions should be thus well addressed for obtaining an optimal process performance. Noble metal-based catalysts with high specific surface areas were reported to have a better catalytic performance than perovskites or  [47,48]. Pd, Pt, Rh and Au as the active component have been widely studied [49]. Among them, Pd and Pt-based catalysts were reported as the most active one by far. The base/acid properties of the support affect the catalytic activity by interacting with the oxidized/metallized state of noble metals. It was reported that the decreased acidity strength of Al 2 O 3 support (e.g., with Pd as the active component) could enhance the catalytic performance of methane combustion [50]. Pt/Al 2 O 3 catalyst has also shown to be more active when the O 2 /CH 4 molar ratio was varied from the oxygen-rich to methane-rich conditions [48]. So far, many studies have been performed to characterize the performance of catalytic methane combustion, but mostly in conventional fixed-bed reactors [51,52] or monolithic reactors [53,54]. Fewer studies were devoted to wall-coated microreactors that are promising owing to their enhanced mass and heat transfer rates, leading to a better process control in terms of high reaction rates and suppressing temperature excursion [55][56][57]. Especially, plate-type multi-channel microreactors, with appropriate catalyst coatings, seem very attractive for use in the catalytic methane combustion, allowing a modular and compact coupling with heat transfer (for the utilization of the combustion reaction heat) or with an endothermic reaction (e.g., steam reforming), which becomes a research hotspot in the recent decade for the foreseeable industry benefits [58][59][60]. From the above literature survey, one may find that the research on how to obtain a good adhesion and dispersion of the catalytic layer onto the walls of microreactors is still needed, especially when it comes to γ-Al 2 O 3 supported noble metal catalysts for use in the catalytic methane combustion. Thus, the main objective of this study is to investigate and characterize the slurry washcoating method of γ-Al 2 O 3 onto microreactor walls, so as to ensure the formation of a stable, homogeneous and well-adhered catalyst layer. The effect of various factors in the preparation procedures was systematically studied, including the slurry property (i.e., the binder nature and its concentration and molecular weight, the pH value, the γ-Al 2 O 3 content and particle size), and the (micro)reactor substrate material and channel shape. Then, washcoating of γ-Al 2 O 3 under the optimized preparation conditions, followed by the impregnation of platinum, was used to prepare Pt/γ-Al 2 O 3 catalyst on the walls of a laboratory-scale multi-channel microreactor made of FeCrAlloy. The microreactor performance was tested in the catalytic methane combustion, where the influence of operating conditions and the temperature distribution along the microreactor have been particularly addressed. 99.97% on metals basis), PVA (98-99% hydrolyzed), acetic acid, tetraammineplatinum (II) nitrate (Pt(NH 3 ) 4 (NO 3 ) 2 , 99.99% on metals basis) were purchased from Alfa Aesar. Tylose MH300 (methyl 2-hydroxyethyl cellulose, molecular weight (MW) of 95,000), 2-HEC (2-hydroxyethyl cellulose, MW of 90,000), PEG (polyethylene glycol, MW of 20,000) were obtained from Sigma-Aldrich. γ-Al 2 O 3 with other particle sizes (20 nm − 180 µm) were also studied and obtained from Sigma-Aldrich. Boehmite (AlO(OH)) and dispersal P2 were obtained from Sasol.

Materials
The multi-channel platelets used for the coating adhesion test were made of FeCrAlloy (Kanthal A-1, 22% Cr, 5.8% Al and Fe for balance) and 316 L stainless steel, having an overall dimension of 50 mm (length) × 22 mm (width) × 3 mm (height). 10 parallel microchannels of rectangular shape with a dimension of 50 mm × 1.5 mm × 1 mm were machined on each platelet. Round capillaries made of FeCrAlloy and 316 L stainless steel having a dimension of 10 mm (outer diameter; o.d.) × 8 mm (inner diameter; i.d.) × 100 mm (length) and square capillaries made of 316 L stainless steel with a dimension of 8 mm (width) × 8 mm (height) × 100 mm (length) were used for coating adhesion test as well.
The catalytic methane combustion was performed in plate-type multichannel microreactors. The microreactor basically consisted of two multichannel platelets (made of FeCrAlloy or 316 L stainless steel) back-to-back and enclosed by two additional blind stainless steel plates (350 mm × 80 mm × 5 mm). Each platelet (317.5 mm × 50 mm × 3 mm) has 16 parallelized straight channels of 275 mm × 1.5 mm × 1 mm, the total number of machined microchannels in the microreactor being 32 (Fig. 1). Each platelet has an inlet fluid distributor and an outlet fluid collector, having a bifurcated tree-like geometry ( Fig. 1) that guarantees a uniform fluid flow distribution across parallel microchannels [61,62]. Only the two sides and bottom of the rectangular microchannel on the platelets  were coated with Pt/γ-Al 2 O 3 catalyst, the fluid distributor and collector being not coated.

Pretreatment
The platelets and capillaries made of FeCrAlloy were first immersed with acetone for 30 min in the ultrasonic bath at 45°C, in order to remove oil, grease and other dirt [63]. Thermal pretreatment was subsequently performed at 900°C (ramp from room temperature: 20°C min −1 ; 10 h at final temperature), to generate a thin alumina layer over the substrate surface which could be a strong bonding between the coating layer and the substrate [28,43]. In the case of platelets and capillaries made of 316 L stainless steel, sandblasting was first applied before the calcination pretreatment to remove the anti-corrosion layer on the surface as well as to increase its roughness.

Washcoating procedures
The γ-Al 2 O 3 slurry was prepared by mixing γ-Al 2 O 3 powders, binder and acetic acid [18]. Various binders (including Tylose, PVA, 2-HEC, PEG or boehmite) with different concentrations and/or molecular weights have been added in the slurry. γ-Al 2 O 3 concentration was varied from 10 to 40 wt%, with its initial particle size ranging from 0.02 to 180 µm. pH of the slurry was changed from 1.5 (by adding acetic acid) to 9.35 (by adding NH 4 OH). The slurry was then heated up to 65°C for 2 h under 300 rpm stirring and stored at room temperature for at least 2 weeks to remove the inside bubbles before use.
Thus prepared slurry as the catalyst support precursor was first deposited on the walls of the parallelized microchannels on the microreactor platelets (made of FeCrAlloy and stainless steel) using syringe injection. The excessive suspension outside the microchannel was immediately removed with a razor blade. The platelets were then dried at room temperature overnight for at least 8 h, dehydrated at 120°C for 8 h and finally calcined at 600°C (ramp from room temperature: 2°C min −1 ; 2 h at final temperature). Multiple layers of coatings were obtained by repeating the same washcoating procedure as described above. A similar washcoating procedure was applied for coating inside circular or square capillaries, except that after the slurry washcoating, a rotary motor held one end of the capillary with a fixed rotation speed (ca. 30 rpm), keeping it at room temperature overnight. The same drying and calcination processes were then followed (see Appendix A for more details about washcoating in such capillaries).
The primer coating was only tested on the 316 L stainless steel multi-channel platelet (with 10 parallel microchannels; dimensions shown in Section 2.1) before coating the slurry. The alumina sol-gel composed of 10 wt% dispersal P2, 3% PVA (MW of 146,000-186,000) and 1% HNO 3 was coated on the platelet by using syringe injection. Subsequently, the substrate was dried at room temperature overnight (for at least 8 h) and then dehydrated at 120°C for 8 h. Then, calcination at 600°C for 2 h took place before applying the same slurry deposition procedure as describe above.

Impregnation of catalytically active component
The incipient wetness impregnation was performed by immersing the coated multi-channel platelet in a Pt(NH 3 ) 4 (NO 3 ) 2 solution with a certain concentration. The impregnated coating was dried at room temperature overnight for at least 8 h, and then calcined at 450°C (ramp from room temperature: 2°C min −1 ; 2 h at final temperature). The nominal Pt loading over the catalyst is about 3.8 wt% for the FeCrAlloy platelet, and about 3.5 wt% for the 316 L stainless steel platelet.

Washcoat adhesion test
The adhesion test of the prepared γ-Al 2 O 3 washcoat was firstly performed by immersing the coated substrate in a glass beaker (containing acetone) that was placed in the ultrasonic bath (PCE- UC 20) for typically 3 h, the frequency of ultrasonic bath being 40 kHz. The weight loss of the substrate (i.e., multi-channel platelets or capillaries) was calculated based on its weight difference before and after the ultrasonic treatment combined with drying at room temperature. Another method used to test the adhesion is the thermal shock test, by rapidly heating the coated substrate to 800°C (ramp: 20°C min −1 ) and cooling it down quickly to room temperature in the air, and repeating the cycle for 5 runs. The weight loss (i.e., the weight difference before and after the thermal shock test) was also measured.

Catalytic methane combustion in microreactors
Two mass flow controllers (MFC, Brooks SLA5850) were used to adjust the flow rates of methane and the synthetic air for the experiment (Fig. 1). The total gas flow rate was typically adjusted from 110 to 880 mL min −1 (based on ca. 20°C and 1 atm). Methane-air mixture (concentration of methane at 2-10 vol%) first passed the inlet stainless steel tube (i.d.: 3.7 mm, o.d.: 6 mm). A part of this tube (15 cm in length) was kept inside an oven to preheat the mixture gas to the required reaction temperature. Then, the gas mixture was introduced into the multi-channel microreactor (i.e., with parallel microchannels on the platelet made of FeCrAlloy or 316 L stainless steel) placed in the oven. It has to be mentioned that the microreactor was first heated up (ramp: 10°C min −1 ) to the targeted reaction temperature under the nitrogen atmosphere to avoid any reaction occurrence during this startup process, and then the gas was switched to the methane-air mixture for the reaction to start. The reaction temperature tested ranged from 300 to 450°C, and refers to the measured oven temperature by a thermocouple located in its right corner (Fig. 1). Four thermocouples (T1-T4) were placed on the top surface of the microreactor at different axial positions (i.e., at 20%, 40%, 60% and 80% of the microchannel length from the inlet), touching the top of the middle microchannel (Fig. 1). The product gas first flew through a condenser at the outlet of the microreactor to remove water, and was collected and analyzed using a gas chromatography (GC).

Analytical procedure
Gas product was collected by a Tedlar sample bag (SKC 3 L, 9.5 × 10 in.) with a single polypropylene septum fitting, and then analyzed by GC (Hewlett-Packard 5890 Series II) equipped with a thermal conductivity detector. A Porablot Q Al 2 O 3 /Na 2 SO 4 column (length: 50 m; i.d.: 0.5 mm) and a CP-Molsieve 5 Å column (length: 25 m; i.d.: 0.53 mm) were used. The concentrations of the reference gas used were 20.7% CH 4 , 17.9% CO 2 , 2.99% CO, 1.5% C 2 H 6 , 1.49% C 3 H 8 , 5088 ppm C 2 H 4 , 5122 ppm C 3 H 6 and H 2 for the rest. The GC oven temperature was heated from 40°C up to 90°C (ramp: 20°C min −1 ) and maintained for 7.5 min. The detector temperature was kept at 200°C.
The rheological characteristics of slurries used for washcoating were evaluated by a rheometer (HAAKE Mars III, Thermo Scientific). The diameter of cone-and-plate geometry is 60 mm, the angle 2°C. The slurry viscosity was measured based on a shear rate in a range of 0.05-1500 s −1 at 20°C.
Zeta potential of the slurry as a function of pH was measured on a ZetaPALS instrument (Brookhaven) by the use of an electrophoretic technique. The mobility of charged particles can be determined via this technique by the phase analysis light scattering (PALS).
Thermogravimetric analysis (TGA) of the γ-Al 2 O 3 slurry was performed on a PerkinElmer TGA 4000 thermogravimetric analyzer. The slurry was heated up in nitrogen from 30°C to 1000°C (at 10°C min −1 ). Data were recorded by Pyris thermal analysis software.
Scanning electron microscopy (SEM) was performed on an XL30 ESEM (Philips) operating at 20 keV, to characterize the smoothness and thickness of the washcoated γ-Al 2  was measured using a Micromeritics ASAP 2420 apparatus by nitrogen physisorption at −196°C. The coating was scraped from the substrate and ground to a particle size below 25 µm. The samples were firstly degassed in the vacuum at 200°C for 8 h before nitrogen adsorption. The surface area and pore size were evaluated using the Brunauer-Emmett-Teller (BET) method, and the micropore area was quantified by the t-plot method. Transmission electron microscopy (TEM) analysis of the coated Pt/ γ-Al 2 O 3 catalyst was performed using an electron microscope CM12 (Philips) at 120 keV. The catalyst was scraped from the substrate and ground to a particle size below 25 µm, and prepared by ultrasonication in acetone.

Definitions
The loading (unit: g m −2 ) of γ-Al 2 O 3 or Pt/γ-Al 2 O 3 deposited on a multi-channel platelet is calculated as its weight gained on the substrate (i.e., after calcination) divided by the microchannel surface area subject to coating.
Here F means the molar flow rate. The subscripts i and o indicate the microreactor inlet and outlet, respectively.
The normalized length (L*) is used to infer the thermocouple location in the multi-channel microreactor (Fig. 1). It is defined as the distance from the thermal couple to the beginning of the parallel microchannel in which it was inserted (x) divided by the entire length of the microchannel (L).
The residence time (τ) in the multi-channel microreactor is defined as the total volume of the parallel microchannels (V tot ) divided by the total volumetric flow rate of the gas entering the microreactor (Q tot ).

Effect of preparation procedures on the washcoat adhesion
Many factors can have an important effect on the adhesion and thickness of the washcoated γ-Al 2 O 3 layer over the metal substrate surface, including among others the binder characteristics and the slurry properties. The influence of these key factors is discussed in the following sub-sections.

Effect of binder
The multi-channel FeCrAlloy platelets were initially coated with slurries containing different binders. The slurry composition is listed in Table 1. Fig. 2 shows that the results from the ultrasonic test are in good agreement with the thermal shock test results in Table 1, with regard to the influence of binder addition on the washcoat adhesion strength: PVA (5 wt%) ≈ Tylose (1.6 wt%) > PVA (5 wt%) + boehmite (1 wt %) > 2-HEC (1.6 wt%) > PEG (15 wt%) > without binder. Typically, well-adhered and stable washcoated layers were obtained on the parallel microchannels of the platelets by employing PVA or Tylose as the binder. No weight loss of the washcoat was observed for the case with PVA in the ultrasonic test after 3 h and only 1.2 wt% weight loss for the case with Tylose, with no weight loss found in the thermal shock test. A higher weight loss of 8 wt% according to the ultrasonic test was caused by adding boehmite with PVA in the slurry. In this case, a strong shrinkage behavior was observed during drying overnight at room temperature, which resulted in a higher washcoat detachment. PEG or 2-HEC as the binder presented a much higher detachment from the microchannel wall, finally resulting in 78.9 wt% or 69.5 wt% weight losses in the ultrasonic test, respectively (Fig. 2). Fig. 2 also shows that the weight loss curves with all binders tested reach a plateau after 1 or 2 h ultrasonic treatment, indicating that at least part of (if not all) the washcoated layer was strongly anchored to the microchannel wall in all cases.
The slurry viscosity can be adjusted by changing the binder and its concentration (Fig. 3). The slurry viscosity exponentially rises with the increasing binder percentage, especially for PVA. It was reported that the hydroxyl methyl cellulose could act as an associative function group in the acid environment [64]. The high amount of hydroxyl group is able to aid the adhesion and the bond between particles [65]. Thus, binders (like PVA and Tylose) with a large amount of hydroxyl group on their surface could form a stronger bridge between particles, enhancing effectively the adhesion. However, the hydroxyl group of PEG is located only at the end of polymer chains, which is not enough to form a firm bonding with alumina, thus causing a poor washcoat adhesion as shown in Table 1 and Fig. 2 [5,66].
As shown in Fig. 3, Tylose and 2-HEC carrying methyl and hydroxyethyl groups have a more significant effect on the slurry viscosity increase even at a lower concentration than the case with PVA. As for PEG, in order to obtain a similar viscosity, a factor of 15 in terms of concentration is required than for Tylose or 2-HEC. The slurry viscosity is not always directly related to the washcoat adhesion strength. However, the viscosity reflects the interaction of the entangled longchain binder with γ-Al 2 O 3. Hence, an enhanced adhesion is expected to be obtained by using the slurry with a suitable viscosity.
During the drying process of the washcoated layer, the binder particles were attached to the surface of alumina and strong bridges could thus be formed. Different types of hydrogen bridges could be formed in terms of different surface groups of γ-Al 2 O 3 [67]. On one hand, PVA carrying the hydroxyl group (-OH) may serve as the anchoring site for the formation of hydrogen bridges with γ-Al 2 O 3 . Al O H OH or Al O OH long chain could be formed in the suspension [67], which could effectively enhance the coating adhesion. On the other hand, partly dissolved alumina could form Al O Al group, which may enhance the particle cohesion during drying. Furthermore, γ-Al 2 O 3 particles of small sizes can be easily filled into the space between the large polymer binder particles. Alumina particles could migrate with the shrinking process to the contact points between particles during drying, leading to a remarkably increased interlocking between the polymer and alumina. Tylose and PVA have been evaluated to be the best binder considering the negligible weight loss percentage from the ultrasonic and thermal shock tests (Table 1 and Fig. 2). A smaller concentration is required for Tylose than for PVA to achieve the same viscosity (Fig. 3). Also, Tylose is much easier to dissolve in water than PVA. However, the SEM image in Fig. 4a shows some obvious cracks between the side walls and the bottom of the microchannel. The strong shrinkage behavior of Tylose during drying hindered the formation of a smooth and crack-free washcoated layer at the two corners of the rectangular microchannel. In contrast, PVA could be a good option to form a smooth and crack-free washcoated layer, as shown in the SEM image of Fig. 4b. Although the specific surface area of the obtained washcoat did not vary significantly in the case of adding different binders (Table 1), it was obviously decreased compared with the case without binders. This could be attributed to the modification of the washcoat pore structure to some extent associated with the thermal decomposition of binders during calcination [68,69]. Moreover, binders (e.g. PVA, Tylose, 2-HEC) were completely removed during calcination at ca. 600°C, as confirmed by TGA results (cf. Fig. B1 in Appendix B). The choice of this calcination temperature throughout the current work for washcoating was also based on the consideration of maintaining an appreciable specific surface area of the γ-Al 2 O 3 washcoat (cf.

Effect of concentration and molecular weight of PVA
As described above, PVA as the binder used in the slurry deposition presents the best washcoat adhesion onto the microchannel wall. The slurry characteristics depend not only on the chemical structure of the binder, but also on its concentration and molecular weight (MW). Higher MW results in an increased slurry viscosity even at low concentrations of PVA ( Table 2). The results in Fig. 5 indicate that a stable washcoated layer on the FeCrAlloy platelet with negligible weight loss during the ultrasonic treatment was obtained when using a PVA concentration of 3-5 wt% with the MW at 57,000-186,000, or of 1 wt%, but only with the MW at the highest range tested (146,000-186,000). Upon increasing the MW, longer PVA polymer chains could better disperse alumina particles and better prevent particle agglomeration in the slurry, thus it is possible to provide a better spatial stabilization of alumina in the slurry. However, when the PVA concentration is higher than 5 wt%, the weight loss of the washcoat in the ultrasonic treatment increased obviously with the increasing MW. This could be explained by the fact that a higher MW with a higher concentration commonly results in a stronger shrinkage behavior, which could easily generate cracks on the surface of the finished washcoated layer. Moreover, the excessive polymer binder could occupy a large fraction of the spatial structure in the washcoat than was required, and was not able to form a bond bridge with alumina. The structure was thus much easier to break under high temperature calcination. As for the MW of 13,000-23,000, Fig. 5 shows that the weight loss exhibited an obvious reduction as the PVA concentration was increased from 1 to 11 wt%. The high weight loss at lower PVA concentrations is mainly due to the insufficient amount of polymer and thus less hydrogen bridge links to alumina, especially when considering the fact that a good dispersion of alumina particles in the slurry could not be promoted with such low MW PVA.
In addition, it is seen from Table 2 that the BET surface area of the prepared washcoat layer slightly increased with the PVA MW and concentration. A high surface area might be attributed to a better particle dispersion and less agglomeration using PVA (5 wt%) with a higher MW [70,71]. However, a lower pore volume and average pore size appeared when using a higher PVA MW, possibly due to the disruption of pore structure order with long-chain PVA to some extent [72]. A higher BET surface area obtained at a higher PVA concentration (i.e., at the MW of 88,000-97,000; Table 2) could be explained by the fact that a larger amount of spatial structure was occupied by PVA, which inhibited the close packing of alumina particles. Accordingly, a higher surface area with a larger pore volume and pore size was observed thereof. Thus, the washcoat adhesion and specific surface area could be varied by adjusting the concentration and MW of PVA. Typically, the PVA MW of 57,000-186,000 with the concentration in a range of 3-5 wt% could be considered as an appropriated range for the slurry preparation.

Effect of pH
The pH value determines the fluidity and stabilization of the slurry. PVA with the MW of 146,000-186,000 and concentration of 5 wt% was selected as the optimized binder composition during the slurry preparation. The influence of pH (1.5-9.35) of the slurry was investigated on the weight loss of the finally obtained γ-Al 2 O 3 washcoat on FeCrAlloy platelets after 3 h in the ultrasonic bath, as shown in Fig. 6. The weight loss first decreased with the increasing pH value, but then remarkably increased with the pH value reaching 9.35. A distinctive  Table 1.  Table 1. feature may be seen, namely no weight loss at pH = 3.5. Fig. 7 shows the variation of the slurry viscosity and the zeta potential as a function of the pH value. The viscosity obviously rises, whereas the zeta potential gradually decreases with the increasing pH from 1.5 to 9.35. The larger zeta potential (higher than that of the isoelectric point) indicated that the (negative) particle surface charges increase with the increasing pH. Under a strong base environment, alumina particles in the slurry carry a negative electric charge High viscosity values could be explained by the increased adsorption rate of free ions of the negatively charged particles (Al(OH) 4 − ) over the alumina surface at high pH values [73,74]. The higher amount of free ions carrying more current could remarkably reduce the movement speed of the particles. Thus, an unevenly washcoated layer could be caused by the formation of particle  Table 1.   agglomerations and poor particle dispersion at pH = 9.35, as confirmed by the SEM image shown in Fig. 8. This eventually resulted in a high weight loss under the ultrasonic treatment (Fig. 6). In comparison, low slurry viscosities at lower pH values (< 3.5) resulted in a higher weight loss (Fig. 6). This observation is in good agreement with the tendency of the zeta potential curve (Fig. 7), an important factor to reflect the mobility and stability of the suspension. At lower pH values (e.g., 1.5), the dissolution of alumina particles took place in the acidic solution ( [75]. Al 3+ dissolution and the surface charging became the main reaction [76]. The mutual electrostatic repulsion forces between particles were enhanced due to the addition of H + ions [73,77]. It resulted in a high fluidity of the slurry with well-dispersed alumina particles. Thus at pH < 3.5, a worse coating adhesion was observed when decreasing pH. It is worth noting that the original pH of the prepared slurry was around 5 and the sedimentation took place during storage, and the obtained washcoated layer was also much easier to peel off. Therefore, a stable and uniform washcoated layer of alumina can be formed by preparing slurry at a slight acid environment of pH ≈ 3.5 under which the suitable zeta potential (not too high) and viscosity (not too low) values were ensured.

Effect of initial γ-Al 2 O 3 particle size
The weight loss curves of the washcoat subject to the ultrasonic treatment, when using different γ-Al 2 O 3 particle sizes in the slurry for deposition on the FeCrAlloy platelet, are presented in Fig. 9a. In the case of particle sizes between 45 and 180 μm, 100 wt% weight loss was found within 1 h, and most of the weight loss took place within 5 min. It seems that the larger the particle size, the higher the weight loss for a given short ultrasonic test time. Here, a serious particle sedimentation in the slurry is expected to have occurred, which could be attributed to the significant liquid-solid density difference. For instance, the slurry prepared with 100 µm alumina exhibited an obvious sedimentation with the appearance of two separate phases within 10 h (Fig. 9b). Thus, the slurry composed of larger particles is generally less adhered onto the substrate than that of smaller particles. Larger γ-Al 2 O 3 particles in the suspension are thus not favorable to form the extensive mechanical interlocking due to fewer contact points for anchorage.
As the alumina particle size decreased to 3 µm, more substantial contact points could be created between smaller particles, leading to a significantly improved anchorage and interlocking on the surface of the substrate [5]. Brownian motion usually overcomes the effect of gravity for smaller particles (e.g., with sizes around the sub-micron scale). Thus, no separate phases occurred for alumina with 3 µm even after 3 months (Fig. 9b). However, for the particle size in a range of 0.02 to 3 µm, the washcoat weight loss increased with the decreasing particle size in use (Fig. 9a). The maximum weight loss is 18 wt% after 3 h ultrasonic test in the case of using a particle size of 0.02 µm. As shown in Fig. 9b, for this particle size the sedimentation of the slurry slowly occurred within 4 days (the reason of which is not clear yet). In this case, alumina particles seemed to be loosely bounded due to the weaker interlocking formed between particles in the deposited coating, and eventually could be easily peeled off under the ultrasonic vibration.
Furthermore, smaller particles are more capable of penetrating into the surface cavities, forming a strong interlocked with the substrate. The tighter packing of particles of smaller sizes (e.g., 3 µm), the stronger the mechanical/interfacial force formed between particles and the substrate. However, much larger particles are not favorable to form the extensive mechanical interlocking, due to a looser particle packing that can be easily carried away from the ultrasonic vibration [39].
Interestingly, it has been reported that a good adhesion of CeO 2 / CuO washcoat on ceramic monoliths could be achieved starting from a slurry with a ceria particle size range from ca. 1 to 3 µm [78,79]. This size range is similar to our findings, suggesting that the optimal particle size in the slurry for a good adhesion might poorly depend on the chemical composition of the washcoated layer.
All the above observations indicate that the initial γ-Al 2 O 3 particle size is an essential factor that affects the adhesion characteristics of the washcoated layer on the substrate. A higher mechanical stability of the washcoated layer can be obtained with relatively smaller initial particle sizes. A γ-Al 2 O 3 particle size of about 3 μm seems appropriate for obtaining a well-adhered washcoated layer with endurance.

Effect of γ-Al 2 O 3 content
The adhesion test results on the FeCrAlloy platelet under the ultrasonic treatment of washcoats prepared with various γ-Al 2 O 3 concentrations in the slurry are presented in Table 3. Almost no weight loss was observed of the washcoat when the γ-Al 2 O 3 concentration used was in a range of 10-20 wt%. The weight loss sharply increased to 10.03 wt % and 32.56 wt% when the γ-Al 2 O 3 concentrations were at 30 wt% and 40 wt%, respectively. Moreover, the coating surface appeared to have irregularities in the latter cases (Fig. 10). This may be explained by the stronger attractive force caused by a closer distance between alumina particles at high concentrations. The surface cracks consequently  appeared during calcination due to the particle agglomeration. Also, the slurry viscosity was found to exponentially increase with the increasing γ-Al 2 O 3 concentration. Increasing the slurry viscosity (or γ-Al 2 O 3 concentration) and/or the number of washcoated layers tended to increase the γ-Al 2 O 3 loading on the microchannel (Table 3). SEM images (Figs. 10a and b) also visually confirm that the γ-Al 2 O 3 loading on the microchannel was remarkably increased in the cases of using 30 wt% and 40 wt% γ-Al 2 O 3 in the slurry compared with the case with 20 wt% γ-Al 2 O 3 (Fig. 4b). The formation of an uneven washcoated layer was observed, however, in the cases with higher γ-Al 2 O 3 concentrations (> 20 wt%).
The desired amount of the washcoat to support the catalytically active component on a microchannel commonly depends on the reaction type/conditions and the targeted application. In terms of the single layer loading, it could be increased from 28.01 to 123.97 g m −2 when using 10 to 40 wt% γ-Al 2 O 3 in the slurry for deposition. Similar observations were reported by several researchers [27,[80][81][82]. To reach a higher amount of the washcoat deposited on the microchannel, successive deposited layers are often needed.

Effect of reactor material and channel shape
Based on the optimal washcoating conditions on FeCrAlloy multichannel platelets found from the experiments discussed above, γ-Al 2 O 3 washcoating has been applied onto two different reactor materials (FeCrAlloy and 316 L stainless steel) in the form of capillaries or multichannel platelets with different channel shapes (round, rectangular and square). The coating conditions are characterized by using 20 wt% γ-Al 2 O 3 (initial particle size at 3 µm), PVA as the binder (MW of 146,000-186,000), pH = 3.5 in the slurry. No weight loss under the ultrasonic treatment was observed when depositing the washcoated layer on FeCrAlloy substrates with (micro)channels of both round and rectangular shapes (Table 4). In contrast, 0.66 wt%, 8.24 wt% and 11.49 wt% weight losses in the washcoating have been observed for (micro)channels made of stainless steel material with round, rectangular and square shapes, respectively. Under such circumstances, the washcoat detachment mostly occurred at two bottom corners of the rectangular or square channels after the ultrasonic treatment (Fig. 11). The weight loss difference between the two materials is firstly due to that the surface of FeCrAlloy was much rougher than that of stainless steel. Thus, more anchoring sites existed over the FeCrAlloy surface, resulting in a stronger interlocking between the washcoated particles and surface irregularities over the substrate. More importantly, a thin layer of alumina was formed on the FeCrAlloy surface after calcination, which effectively improved the adhesion strength. For the stainless steel case, iron oxide was probably formed on the surface after calcination. It is thus assumed that the washcoated layer possibly has a stronger affinity with the alumina thin layer (e.g., via Al O Al bonds) on FeCrAlloy than with iron oxide (e.g., via Al O Fe bonds) on stainless steel.
In order to improve the coating adhesion on 316 L stainless steel, the primer coating (cf. details in Section 2.2.2) was employed before depositing alumina coating [17,83]. According to the literature [17,70], a strong bond between the primer coating and the peroxidized 316 L surface could be formed. However, the test results indicate that the use of the current primer coating could not reduce the washcoat weight loss during the ultrasonic treatment (cf. Table C.1 in Appendix C), where 23.84 wt% weight loss was observed on the stainless steel multi-channel platelet for 1 h ultrasonic test. This is even worse than the results without applying primer coating ( Table 4). The reason for this is not clear yet, which indicates that the primer coating method deserves further investigation. Thus, the primer coating was not used in our following experiments.

Effect of viscosity during the drying process
The washcoated layer (i.e., as tested on multi-channel platelets of this study) should have a uniform and constant thickness along the  microchannel, for ensuring its good catalytic performance when impregnated with catalytically active components. The platelets were initially fully filled with the slurry. The excess slurry around the top of parallel microchannels was then scraped off. The slurry adhered to the bottom and the side walls of microchannels during drying. The drying process was characterized by the particle packing and the shrinkage degree, which is supposed to rationalize the washcoat shape as shown in Fig. 12. The water evaporation in the washcoated layer is accompanied by the shrinkage behavior at the same time. During the experiment, it was observed that drying of highly viscous suspensions was faster to reach the finally stagnant film (due to less volatility) compared with suspensions with low viscosity. That is, in the latter case, a longer drying time was needed to reach the stagnant film for the same γ-Al 2 O 3 loading. The film thickness further shrunk until a homogeneous solid washcoated layer was formed. For less viscous suspensions, part of the suspension at the top section of side channel walls may slowly slip  Table 3. tions: 20 wt% γ-Al 2 O 3 with its initial particle size at 3 µm, 5 wt% PVA (MW of 146,000-186,000), pH = 3.5. b Tested on parallel microchannels in a multi-channel platelet. c Tested in capillaries.  Table 4. down and accumulate at the bottom of the channels during drying. The coating layer thus appeared to be thicker on the bottom than on the two side walls. In contrast, the washcoated layer in the case of using highly viscous suspensions was much thicker at the lower part of the side walls than at the upper part. Thus, a suitable slurry viscosity is required for a uniform coating around the microchannel, which is in qualitative agreement with SEM images shown in Figs. 4b and 10).

Catalytic methane combustion over the coated Pt/γ-Al 2 O 3 catalyst in the multi-channel microreactor
The catalytic methane combustion was studied over the coated Pt/γ-Al 2 O 3 catalyst in the multi-channel microreactor (containing the microchannel platelet made of FeCrAlloy or 316 L stainless steel). The coating was applied using the optimized slurry preparation conditions characterized by using 20 wt% γ-Al 2 O 3 (3 µm particle size), 5 wt% PVA (MW of 146,000-186,000) at a slurry pH of 3.5, followed by the impregnation of Pt. The TEM picture confirms the successful dispersion of platinum particles in the catalyst (cf. Fig. D.1 in Appendix D). The reaction performance as a function of the key operational conditions was investigated, including the temperature, flow rate and O 2 /CH 4 molar ratio. The temperature distribution along the microreactor was also studied under selected reaction conditions.

Effect of operating conditions on the methane conversion
The methane conversion as a function of the reaction temperature under different total flow rates (i.e., the sum of air and methane flow rates, based on ca. 20°C and 1 atm) is shown in Fig. 13a, for the multichannel microreactor composed of FeCrAlloy platelet. At a total flow rate of 110 mL min −1 , the conversion presented a remarkable increase from 12.7% to 95.75% when the temperature was increased from 300 to 450°C at an inlet O 2 to CH 4 molar ratio (Φ) of 2. The light-off was observed with a sharp conversion increase starting at ca. 350°C. This light-off phenomenon is likely due to the local heating of the catalyst given the highly exothermic nature of the combustion reaction (the heat of the reaction, H R being ca. −810 kJ mol −1 ) and the favorable fractional coverage of the catalyst surface by the adsorbed methane and oxygen under such conditions [47,48], leading to a significant increase of the catalytic activity. The smaller flow rate rendered a longer residence time (τ; cf. Eq. (4)) in the coated microchannel. For instance, the residence time was increased by a factor of 8 from 1.8 s (880 mL min −1 ) to 14.41 s (110 mL min −1 ) at 450°C, and thus the methane conversion was increased from 78.53% to 95.75%. Fig. 13a further reveals that the reaction temperature has a greater influence than the flow rate, due to the remarkable increase of the intrinsic kinetic rate especially at the catalytic ignition temperature (where the light-off occurs) or above. Fig. 13b shows that carbon monoxide was formed increasingly when prolonging the residence time (τ > 1.8 s) at a reaction temperature of 450°C and Φ = 2 (stoichiometric ratio for a complete combustion), however, its selectivity is low (< 1.2%). One possibility for CO formation is that the adsorbed oxygen over the localized areas of the catalyst was not sufficient [48], giving rise to the incomplete combustion ( + + CH 1/2O CO 2H ) 4  2 ). It seems from Fig. 13b that the longer the residence time, the higher the chance for such reactions for CO formation to occur. It should be noted that no carbon monoxide was detected (i.e., the CO 2 selectivity being 100%), when Φ was set higher than 2, indicating a more sufficient coverage of the adsorbed oxygen on the catalyst surface. Fig. 13c reveals that the methane conversion tended to decrease when increasing Φ from 2 to 10 (oxygen-rich). The lower conversion for the oxygen-rich case could be explained by the competitive adsorption between oxygen and methane over the catalyst surface. Since the methane adsorption energy is higher than that of oxygen [84], a competitive adsorption of oxygen prevented further oxidation by inhibiting the weakly adsorbed methane on the active sites [85]. Thus, a favorable surface coverage by the adsorbed methane and oxygen over the catalyst surface is essential for achieving a desired conversion. Another possibility is that the bond between the absorbed oxygen and the oxidized platinum is stronger than that with metallic platinum [86], which is not favorable for methane and/or oxygen adsorption (rate determining step). Commonly, its metallic state presents a higher catalytic activity compared to its oxidized state (PtO 2 ), despite the high-dispersed phase of the latter state [87,88]. Thus, the platinum catalyst itself might be less active under the condition of the predominant oxygen-covered surface and thus not be able to fully catalyze methane oxidation especially when the present methane amount was high. Consequently, a low methane conversion would be expected when using an oxygen-rich mixture due to the insufficient adsorbed methane and/or less available metallic state of platinum on the catalyst surface. Fig. 13d presents a comparison of the methane conversion in the multi-channel microreactors composed of FeCrAlloy and 316 L stainless steel reaction platelets. A slight lower conversion was obtained in the stainless steel microreactor under otherwise the same conditions, possibly due to a somewhat lower Pt loading. Considering the coating adhesion strength, FeCrAlloy microreactors appear to be a better option for long-term uses, whereas 316 L stainless steel microreactors represent a more cost-effective alternative (especially if the coating adhesion is better addressed).

Effect of operating conditions on the microreactor temperature distribution
The temperature distribution along the multi-channel microreactor (with FeCrAlloy platelets) was measured during the catalytic methane combustion, using the thermocouples that touched the top of the middle microchannel on the platelet. A blank experiment for temperature measurement was also conducted with only air flowing through the microreactor, as a reference case for comparison.
It was observed in Fig. 14a that for a reaction temperate (i.e., the oven temperature) at 450°C, the temperature along the microchannel was close to uniform in the reference case without reaction occurrence. Temperature values close to the middle of the microchannel appeared to be slightly higher than those at both ends, with a difference below ca. 5°C. This is mainly a result of the oven temperature profile. Once the reaction started, such a (close to) uniform temperature profile was no longer sustained in the microchannel. For a total flow rate at 440 mL min −1 and Φ = 2 (Fig. 14a), the temperature in the front part of the microchannel increased more significantly (e.g., at the normalized length of the microchannel, L* = 0.2), due to the released combustion reaction heat and the location of the reaction front in the middle section or further upstream (vide infra). At a given axial location, the temperature was gradually increased upon increasing the reaction test time on stream until 50 min, after which the temperature reached a stable state showing the thermal inertia of the microreactor. Moreover, the temperature was increased mainly in the front part, and then was gradually decreased to the trailing end of the microchannel. It may be concluded that the majority of methane was converted at the front section of the microreactor. And (much) less reaction took place at the latter section of the microreactor under these conditions, resulting in a somewhat insignificant temperature increase therein.
The effect of flow rate on the temperature distribution along the microreactor is illustrated in Fig. 14b. More heat would be released at a higher total flow rate due to the more involved methane, despite a slightly lower methane conversion (Fig. 13a). Thus, the temperature was noticeably raised along the microreactor at higher flow rates (Fig. 14b), and the axial location with the peak temperature seems to gradually move downstream along the microreactor. This is due to a shorter residence time at a higher flow rate, implying that more methane was converted at a longer distance from the microchannel inlet. This is in line with the literature observation during the catalytic combustion of methane over monolithic catalysts based on (Pt-)LaMnO 3 that the reaction front shifted upstream along the channel at a longer residence time [89,90]. At the reaction front, the fluid has reached the catalytic ignition temperature resulting a large reaction heat release and consequently a sharp increase in the fluid/catalyst temperature [91]. The remaining length of the reactor after the reaction front experienced either no reaction or an insignificant conversion, thus this part acted mainly as a heat exchanger and the fluid/catalyst was cooled down due to external heat losses [89][90][91]. With the increasing flow rate, it tends to take a longer distance to heat the reactant mixture to the ignition temperature, resulting a downstream shift of the reaction front.
Similarly, the concentration variation of the adsorbed methane and oxygen over the catalyst surface significantly affected the temperature distribution. The obvious temperature gradient over the microreactor in this case can be observed in Fig. 14c. Both the methane conversion and the temperature increase could achieve a maximum at a stoichiometric ratio of 2 (cf. Fig. 13c as well). With the increasing inlet O 2 /CH 4 molar ratio, the temperature in the microreactor at a given axial location was obviously decreased, an observation in line with the reduced methane conversion shown in Fig. 13c. Under such circumstances, there is less thermal power provided via combustion, and possibly less favorable surface coverage over the catalyst surface due to the competitive adsorption between oxygen and methane (as discussed in Section 3.2.1), leading to a temperature or conversion decrease. Apart from that, another reason for the lower temperature under the oxygen-rich conditions (at Φ > 2) is that more balance gas (nitrogen) was involved. Thus, a part of the released heat was diluted and absorbed by the balance gas.
The actual maximum value present in the temperature profile and its position in the reactor (as shown in Figs. 14a-c) depend significantly on the pre-heating temperature, the thermal power of the reactant mixture (i.e., related to its composition and flow rate), as well as the heat losses of the reactor (realized in the current work via the conduction through the microreactor wall to the oven atmosphere). Moreover, in our experiments the temperature was only measured on the top surface of the middle microchannel (without coating), since the tips of thermocouples (o.d.: 3 mm) are not possible to touch the inner catalytic wall (Fig. 1). Thus, the temperature level and profile reported in this work are only indicative of that on the catalyst surface. Considering the good thermal conductivity of FeCrAlloy platelets and the reaction heat released, the actual surface temperature of the catalyst should be much higher than these measurements. To illustrate, the adiabatic temperature rise ( T ad ) can amount to ca. 2000°C at Φ = 2 (calculations not shown for brevity) [92]. Thus, it is reasonable to estimate that the actual temperature on the catalyst surface could be somewhat high under the present reaction conditions, which might accelerate the catalyst deactivation as well [93]. The high temperature level on the catalyst surface could also activate and thermally sustain the homogeneous combustion to a certain extent, as addressed in many literatures [89,90,[93][94][95]. Investigation of the above-mentioned aspects would contribute to optimized operating conditions and microreactor design for the catalytic methane combustion. These findings, when coupled with heat transfer studies (e.g., in the presence of the integrated heat exchanging modules), can provide insights into the potential uses of such wall-coated microreactors in energy-related applications (e.g., for heat supply in a small-scale boiler system [96]).

Preliminary results of catalyst deactivation
Catalyst life is an important indicator for its performance, especially relevant to the long-term operation in the commercial applications. Regarding the coated Pt/γ-Al 2 O 3 catalyst in the multi-channel microreactors, a detailed in-situ life time study has not been performed yet. However, the coated Pt/γ-Al 2 O 3 catalyst scrapped from the stainless steel substrate has been tested in the powder form in a fixed bed reactor for 100 h on stream. A somewhat significant decrease of the catalyst activity was already noticed (see Appendix E for more details). This deactivation is possibly due to the catalyst sintering and the growth of Pt particles (e.g., as a result of the high temperature level generated by the released heat at the reaction front), or the formation of platinum species in the oxidized state (PtO 2 ) with a lower activity compared to its metallic state [97][98][99]. These preliminary results imply that the catalyst coating preparation in microreactors still needs to be further optimized (e.g., by improving the catalyst structural stability and the metal-support interaction), together with the catalyst regeneration, in order to maintain a stable catalyst activity.

Conclusions
In this work, the slurry washcoating followed by the incipient wetness impregnation has been used to prepare Pt/γ-Al 2 O 3 catalytic coatings in microreactors. The effect of various factors in the slurry preparation procedures on the adhesion properties (tested typically via the ultrasonic treatment) of the washcoated γ-Al 2 O 3 layer on the microreactor substrates with different channel shapes was investigated. The initial γ-Al 2 O 3 particle size and pH were found as two important factors affecting the slurry stability and the washcoat adhesion strength. A strong particle attractive force could be maintained by using a particle size of around 3 µm in the slurry with a slightly acid environment (pH ≈ 3.5). PVA (typically at a concentration of 3-5 wt% and a MW of 57,000-186,000) was identified as a suitable binder to better disperse alumina particles and form sufficient hydrogen bridges with alumina, thus effectively enhancing the washcoat adhesion. FeCrAlloy as the (micro)reactor substrate exhibited an excellent coating adhesion with negligible weight loss during the ultrasonic treatment in either rectangular or round channels, primarily because of the formation of alumina film over the surface during thermal pretreatment. In contrast, 316 L stainless steel as the substrate only showed a relatively good adhesion in a round channel, presumably due to the lack of a strong chemical bonding between the washcoated layer and the pretreated substrate surface.
The catalyst performance in the catalytic methane combustion was further examined in microreactors comprising parallel microchannels made of FeCrAlloy or 316 L stainless steel. A higher total flow rate (of methane and air) was found to render higher heat generation rate and lead to a more significant temperature increase in the front part of the microreactor, despite a slightly decreased methane conversion due to a shorter residence time. Under the oxygen-rich conditions, the predominant adsorption of oxygen over methane species on the catalyst surface could lead to a decreased methane conversion and thus a lower heat generation. Thus, an appropriate fractional coverage of the catalyst surface by these species and a proper reaction condition selection are needed for achieving the favorable methane conversion, as well as the sufficient heat release for the potential uses of such microreactors for energy-related applications. Somewhat significant deactivation of the coated Pt/γ-Al 2 O 3 catalyst over a time scale of 100 h, as observed in an ex-situ test in a fixed-bed reactor, represents a further direction of improvement to be addressed in our future work.

TGA analysis of the slurry
The amount of slurries (prepared under the conditions mentioned in Table 1) for TGA analysis is 3-10 mg. The results of analysis in the case of using different binders are shown in Fig. B.1, together with the first derivation of the TGA curve (DTG). The first weight loss (< 100°C) is ascribed to the desorption of the physically adsorbed water in the slurry. The second weight loss (at 200-300°C) and third one (at 400-500°C) are associated with the thermal decomposition of binders. This indicates that binders were completely decomposed when being calcined at 600°C. However, the specific surface area of the obtained washcoat decreased with the increasing calcination temperature due to the structure damage of alumina (e.g., pore collapse), as shown for the case of PVA as the binder in Table B.1. Considering both the specific surface area and the binder removal, 600°C was chosen as a suitable calcination temperature.   Table 1. a Slurry preparation conditions: 20 wt% γ-Al 2 O 3 (initial particle size at 3 µm), 5 wt% PVA as the binder (MW of 146,000-186,000), pH = 3.5. Table C.1 shows the weight loss in the ultrasonic treatment of the γ-Al 2 O 3 washcoat obtained via applying the primer coating (details shown in Section 2.2.2), followed by the slurry deposition (other preparation conditions are the same as shown in Table 4) on the 316 L stainless steel multichannel platelet. The weight loss was found to increase to a somewhat significant level upon increasing the treatment time from 5 to 60 min.

TEM analysis of Pt/γ-Al 2 O 3 catalyst
3.5 wt% Pt/γ-Al 2 O 3 catalyst was prepared on the FeCrAlloy platelet (with 10 parallel microchannels; see Section 2.1 for dimensions), according to the procedures described in Section 2.2. Herein, the slurry for coating was prepared using 20 wt% Al 2 O 3 (3 µm), 5 wt% PVA (MW of 146,000-186,000) and pH at 3.5. Then, the catalyst was scrapped off from the substrate and ground to a particle size below 25 µm. After ultrasonication in acetone for 10 min, the sample was dropped onto the carbon coated copper grid (400 mesh) for TEM analysis. The TEM image is shown in Fig. D.1. Pt particles (black dots) seemed to be well dispersed over the alumina support by using the incipient wet impregnation method.

Appendix E
Catalyst life time test in a fixed-bed reactor 3.5 wt% Pt/γ-Al 2 O 3 catalyst was prepared on the 316 L stainless steel platelet (with 16 parallel microchannels; see Section 2.1 for dimensions), using the same protocol as mentioned in Appendix D. The fresh Pt/γ-Al 2 O 3 catalyst was then scraped from the platelet and ground to a particle size of about 25 µm for the catalyst life time test in a fixed-bed reactor made of glass (U shape; i.d.: 5 mm) in a similar oven and experimental setup as shown in Fig. 1. A full methane conversion was achieved at 450°C with 16 h on stream (Fig. E.1). However, the catalyst gradually deactivated, with the methane conversion being dropped to 98.12% at 22 h, and 84.61% at 100 h. The CO 2 selectivity also seems to drop slowly over time. The measured BET surface areas of the fresh and used (after 100 h on stream) catalysts are listed in Table E.1. An obvious reduction in the specific surface area of the used catalyst was observed, possibly due to the catalyst sintering after a long-term reaction. Meanwhile, the growth of Pt particles due to agglomeration could also have occurred, which would result in a catalyst activity loss [100][101][102]. Hence, the current catalyst preparation method still needs to be further improved to maintain a stable catalyst activity.  Table 4.   Fig. E.1).