Supplementary Materials for Grain-boundary-rich Pt nanoparticle assembly for catalytic hydrogen sensing at room temperature

The The and ratio by X-ray Photoelectron Spectroscopy (XPS, ThermoFisher NEXSA UV and X-ray Photoelectron Spectrometer). XPS peaks were calibrated according to the C 1s peak at 284.8 eV and fitted using a composite function (30% Lorentzian + 70% Gaussian) using the Avantage software. Steady-state X-ray absorption spectroscopy (XAS) measurement was performed at 12-BM beamline at the Advanced Photon Source (APS), Argonne National Laboratory. The XAS data were collected under room temperature with fluorescence mode using a 13-element germanium solid-state detector. Three ion chambers were used. One placed before for The two ion and are foil is placed between and third chambers and used for collecting XAS spectrum for energy calibration.

using Nano Measurer 1.2 software. The high-resolution XRD data were collected at the 11-BM-B beamline at the Advanced Photon Source (APS), Argonne National Laboratory. The chemical state and elementary ratio were analyzed by X-ray Photoelectron Spectroscopy (XPS, ThermoFisher Scientific NEXSA UV and X-ray Photoelectron Spectrometer). XPS peaks were calibrated according to the C 1s peak at 284.8 eV and fitted using a composite function (30% Lorentzian + 70% Gaussian) using the Avantage software. Steady-state X-ray absorption spectroscopy (XAS) measurement was performed at 12-BM beamline at the Advanced Photon Source (APS), Argonne National Laboratory. The XAS data were collected under room temperature with fluorescence mode using a 13-element germanium solid-state detector. Three ion chambers were used. One of them was placed before the sample for the incident X-ray flux reference signal. The other two ion chambers (second and third chambers) are placed after the sample. The Pt foil is placed between the second and third ion chambers and used for collecting Pt metal XAS spectrum for energy calibration.
Analysis of GBs and d spacings HRTEM images were processed using ImageJ software. After obtaining a Fast- To compare d-spacing distributions between Pt NPs and Pt NP assembly, we took images with the same magnification and same microscope so that the pixel sizes of the images for both samples were identical. 22 Pt NPs (9 cases at pixel size = 0.0149 nm; 13 cases at pixel size = 0.020 nm) and 23 Pt NP assemblies (12 cases at pixel size = 0.0149 nm; 11 cases at pixel size = 0.020 nm) were analyzed. The histograms of d-spacing distribution for {111} and {200} planes of both Pt NP and Pt NP assembly were plotted with the bin size of 0.015 nm (Fig.  3G)  C) were used to prepare catalytic H2 sensors. A wet slurry containing 1 mg Pt NP assembly (prepared by mixing 1 mg Pt NP assembly with 100 μL ethanol) was drop-casted on the sensing region of a thermometer followed by drying in air.
In the control experiments using other Pt catalysts, 1 mg of Pt NPs (precipitated by centrifuge), Pt NPs without ligand (precipitated by adding NaOH aqueous solution until pH reaches14), Pt black (Sigma Aldrich), Pt powder (Sigma Aldrich), and Adams' catalyst (Sigma Aldrich) were dispersed in 100 μL ethanol to prepare their corresponding slurries. The obtained slurries were drop-casted on the tip of a J-type thermocouple wire followed by drying in air.

Gas sensing tests
The standard H2 sensing tests in this study were carried out using sensors prepared from a J-type thermocouple wire (abbreviated as J-H2 sensors) unless otherwise specified.
Standard H2 sensing test protocol. The H2 sensing tests were carried out using a home-built apparatus at room temperature, which includes two gas cylinders (synthetic air and H2 from Airgas Co., Ltd) and mass flow controllers, a data acquisition meter (Keysight/Agilent 34972A LXI) for real-time recording of the temperature readout, and PC Computer (Hewlett-Packard Co., Ltd) for storing data (3). A J-H2 sensor was placed directly in front of the gas outlet during the test ( Fig.  2A insert). Before exposure to H2, the J-H2 sensor was first stabilized in synthetic air (21% O2+79% N2) at a flow rate of 1000 sccm at room temperature and relative humidity of 0%. Then, H2 gas was mixed with synthetic air to achieve various H2 concentrations of 4%, 3%, 2%, 1%, 0.5%, 0.25%, 0.1%, and 0.05%. These H2/air mixtures sequentially flowed over the J-H2 sensor at a constant rate of 1000 sccm to obtain the response/recovery curves of J-H2 sensors.
First-principles calculations. All DFT calculations were performed using the Vienna ab initio simulation package (VASP) (4,5). Core electrons were described using the projected-augmented wave (PAW) method (6). The Kohn-Sham wave functions were expanded on a plane-wave basis with a kinetic energy cutoff of 400 eV to describe valence electrons. The generalized gradient approximation using the Perdew-Burke-Ernzerhof functional was employed to evaluate the exchange-correlation energy (7).
The Pt surface was modeled as a four-layer (√3 × 2) Pt (111) slab. The bottom two layers of Pt (111) surface were kept frozen, while the other layers and adsorbed molecules were set free to relax. A vacuum space of 14 Å was added to all surface models to ensure no appreciable interaction between periodic images. The Monkhorst-Pack scheme was employed to sample the Brillouin zone using a 5 × 5 × 1 k-point grid for atomic structure optimization (8). For Pt201 NP and Pt383 Σ3 [110] (111) GB models, a vacuum space of > 15 Å was added to prevent any appreciable interaction between periodic images. The Brillouin zone was sampled only on Gamma point for the geometry optimization. All geometries were considered optimized when the force on each atom was < 0.03 eV/Å. The location and energy of transition states were calculated with the climbingimage nudged elastic band method (9, 10).

Strain analysis.
The change of lattice matching near the GB leads to a substantial strain on the Pt-Pt bonds. Strain analysis of Pt Σ3 [110] (111) GB is carried out based on the DFT optimized models. A number of studies have shown that Pt (111) surface is more active than Pt (100) and Pt (110) for the hydrogen oxidation reaction (11,12), and is the dominant facet of Pt NP (13,14). Thus, strain analysis and subsequent investigation of reaction activity in this work are mainly focused on Pt (111) facets.
The strain of surface atoms at the Pt GB was calculated by the following equation: , where is the average Pt-Pt bond length between the atom i and its neighboring surface atoms, ,(111) is the average Pt-Pt bond length of (111) facet in Pt NP. Micro-kinetic model for H2 oxidation reaction over Pt There are two primary hydrogen oxidation mechanisms on Pt (111) surface: associative and dissociative pathways (15). We have considered both mechanisms and built the corresponding micro-kinetic models.
1. Associative pathway For associative pathway, the H2 oxidation reaction on Pt (111) surface involves the following steps: Based on the rate-limiting step approximation (R3), we can derive the coverage of H, O2, H2O, OH and OOH as follows: where the * is the coverage of open sites. After applying the site conservation rule (Eq. 7), * + we derive the coverage of open sites as: * = The rate of H2 oxidation is then written as: where 3 + is the forward rate constant for R3, which can be obtained as Eq. (10); is the equilibrium constants for Ri, which are calculated using Eq. (11); 2 , 2 and 2 are the partial pressures of H2, O2, and H2O, which are listed in Table S3; the approach to equilibrium can be obtained by using Eqs. (12) and (13).
where ,3 is the activation energy barrier of R3 (Table S1), ∆ is the reaction energy of Ri (Table  S1), ∆ and ∆ are the correction of zero-point energy and entropy of Ri, respectively (Table  S4).

Dissociative pathway
For the dissociative pathway, we assume the following elementary steps: Based on the rate-limiting step approximation (R4), we can derive the coverage of H, O2, O, H2O, and OH as follows: The rate of H2 oxidation can be written as: where 4 + is the forward rate constant for R4, which can be obtained as Eq. (22); is the equilibrium constants for Ri, which are calculated as Eq. (11); 2 , 2 and 2 are the partial pressures of H2, O2, and H2O, which are listed in Table S3; the approach to equilibrium can be obtained by using Eqs (23) where ,4 is the activation energy barrier of R4 (Table S2), ∆ is the reaction energy of Ri (Table  S2).  Table S2) and OH * formation (0.75~1.02 eV, Table S2) are considerably higher than that of OOH * formation (0.38~0.42 eV, Table S1). Thus, the associative pathway is the dominant reaction path on NP (111) facet (Fig. 4C). Micro-kinetic modeling also suggests that the coverage of H * ( ) is close to 1. In other words, (111) facet is mostly covered by H*.

Rate calculation of Pt
However, owing to the greatly reduced barrier of O 2 * dissociation (0.008 eV, Table S2) and OH * formation (0.60 eV, Table S2)), the GB area is mostly covered by O * ( ~ 1). Diffusion of H * on Pt (111) surface fast enough, so H * on Pt NP (111) facet can easily diffuse to the GB region and react with O * , creating bifunctional reaction sites near the GB (fig. S22). The maximum rate near Pt GB can be approximated as below:    Thirty-five aliquots of diluted 500 mL Pt NP solution (17.5 L in total) were concentrated by ~17 fold to a final volume of ~1000 mL at 75 o C using a rotary evaporator in batches. The concentration of the concentrated Pt NP solution was estimated to be ~3 μM. Titanium foils (Sigma Aldrich) were utilized as the counter electrode (120 cm 2 ) and the working electrode (75 cm 2 ). An electrode potential of -2 V was applied at the working electrode for twelve hours until the supernatant color turned clear, and other conditions were the same as milligram-scale synthesis. As shown in TEM images, the gram-scale synthesized gel exhibits similar morphology and size to the milligramscale ones. Subsequently, the supernatant was carefully removed without disturbing the wet gel at the bottom. The wet gel was washed by DI water ten times and then completely exchanged with acetone in two weeks. Finally, the wet gel immersed in acetone was subjected to Critical Point Drying to prepare dry Pt NP assemblies. Notably, the gel was partially contaminated with titanium oxide due to the oxidation of titanium foil at the anode during the electrochemical synthesis.  were prepared by sonication a mixture of 1 mg Pt material, 4 mg Vulcan carbon, and 1 mL 3:1 (v:v) H2O-isopropanol (IPA) with 0.05% Nafion for 30 min. The preparation of Pt NP assembly ink was slightly different. Briefly, 1 mg Pt NP assembly powder was added to 1 mL 3:1 (v:v) H2O-IPA and sonicated for 15 min. Next, 2 mg Vulcan was added to the above suspension and sonicated for another 10 min, followed by adding 10 μL 5% Nafion and another 5 min sonication. Finally, 3 μL of ink was drop-cast by a pipette onto the GCE and dried in air, resulting in a mass loading of 3 μg Pt on the electrode or ~42.5 μg/cm 2 geo. Cyclic voltammograms of Pt NPs, Pt NP assembly, and Pt black were recorded at a potential window between -0.25 V and 1.0 V at a scanning rate of 50 mV/s. All the measurements were performed three times to obtain the average values with standard deviations.

Pt NP
Pt NP assembly Pt black Surface area (cm 2 /mg) 340 ± 24 70 ± 1 144 ± 5  . S1), Pt NP assembly, Pt black (Sigma Aldrich), Pt bulk powder (Sigma Aldrich), Adams' catalyst (Sigma Aldrich) were separately mixed with ethanol to prepare a catalyst slurry which was dropcasted on the tip of J-type thermocouple, followed by drying under the ambient condition to prepare the catalytic H2 sensors. where the J-type thermocouple was housed, while the ultra-low temperature of -30 o C was achieved by placing dry ice around the testing chamber (C). The catalytic H2 sensor was exposed to a step-wise decreasing H2 concentration from 4% to 0.05%. For the humidity test, the relative humidity was adjusted by regulating the flow rate ratio of dry air and wet air (bubbling through DI water). The sensor was exposed to H2 concentrations from 4% to 0.05% at the humidity level of 0%, 25%, 50%, 75% and 98% at room temperature.

Fig. S10.
Interference/cross-sensitivity tests. Thirty-six interference gases were tested. Their concentrations were set at 4% with only a few exceptions, including 0.4 ppm for H2S, 4 ppm for NO2, 4 ppm for SO2, and 40 ppm for ethylene. The first panel illustrates the test protocol using CO as an example. Specifically, a J-type thermocouple coated with 1 mg Pt NP assembly was first exposed to three pulses of 4% CO in air to test whether the sensor responds to CO at room temperature, then three pulses of 4% CO but in the presence of 4% H2 to access the crosssensitivity, finally three pulses of 4% H2 to test if there is irreversible poisoning.

Fig. S11.
Long-term stability test. In each cycle, the Pt NP assembly sensor was exposed to H2 concentrations from 4% to 0.05% at room temperature. 10 cycles were conducted per day for one month to complete 288 cycles with a total testing time of 281 h. For all experiments, the thermometer was loaded with 1 mg Pt NP assembly and exposed to 1% H2 in air at room temperature. (G) The responses of the above six catalytic H2 sensors at H2 concentrations from 0.05% to 4% at room temperature. 1 mg Pt NP assembly costs less than $1, and the liquid-in-glass thermometer is ~$3 (Newark.com, #13AJ1664, $2.38 each), making the total cost of a sensor < $5 per unit.             Movie S1. Demonstration of H2 leak detection using a drone sensing platform.

Movie S2.
Demonstration of detection of 2% H2 in air.