Activating low-temperature diesel oxidation by single-atom Pt on TiO2 nanowire array

Supported metal single atom catalysts (SACs) present an emerging class of low-temperature catalysts with high reactivity and selectivity, which, however, face challenges on both durability and practicality. Herein, we report a single-atom Pt catalyst that is strongly anchored on a robust nanowire forest of mesoporous rutile titania grown on the channeled walls of full-size cordierite honeycombs. This Pt SAC exhibits remarkable activity for oxidation of CO and hydrocarbons with 90% conversion at temperatures as low as ~160 oC under simulated diesel exhaust conditions while using 5 times less Pt-group metals than a commercial oxidation catalyst. Such an excellent low-temperature performance is sustained over hydrothermal aging and sulfation as a result of highly dispersed and isolated active single Pt ions bonded at the Ti vacancy sites with 5 or 6 oxygen ions on titania nanowire surfaces.


Supplementary Methods
Catalyst preparation TiO 2 nanowire arrays are grown on ceramic monoliths via a nonpolar solvent/hydrophilic solid substrate interfacial reaction under hydrothermal conditions. 1, 2 Typically, the washed and TiO 2 seeded cordierite honeycomb substrates with size up to 7.5 cm × 7.5 cm × 5 cm are placed within a sealed Teflon reactor (1 L), containing 500 mL of a non-polar solvent, 50 mL of a Ti (IV) alkoxide precursor, 5 mL of titanium tetrachloride (1 M in toluene), and 50 mL of hydrochloric acid (37 wt.%). At room temperature, titanium (IV) precursors and water are separated since the Ti precursors are dissolved in the nonpolar solvent. Under hydrothermal conditions, water diffuses away from the high-energy water/nonpolar solvent interface to the hydrophilic TiO 2 nuclei on ceramic wall, where water reacts with Ti precursors, resulting in growth-crystallization of TiO 2 . The presence of Clanions is crucial for the anisotropic growth, as they tend to absorb on the rutile (110) plane, thus inhibiting further crystal growth of this plane. After being taken out for the reactor, all the samples are sonicated in acetone, ethanol, and water for 3 hours (h) to remove organic compounds and Clresidues from the synthesis before Pt loading.
Microwave-assisted dip-coating and Na-promoted wet incipient impregnation methods are employed to load Pt on the TiO 2 nano-arrays. For the dip-coating method, the substrate is first submerged into the diluted Pt precursor solution (0.5 mg Pt precursor per mL). It is taken out and blow-dried using compression air stream. The substrate is then dried in a microwave oven for 1-3 minutes. These steps are repeated until appropriate amount of the metal precursor is absorbed. Finally, the sample is calcined in air at 500 o C for 2-4 h with a ramp rate of 2 o C min -1 . In addition, a following 1-hour stay at a different calcination temperature such as 700 o C, 800 o C, and 900 o C has been adopted to look into the induced porosity and surface area change and its effect on the single atom loading and the catalyst stability. For impregnation, the substrate is submerged into diluted Pt precursor (50 g mL -1 ) and NaOH (atomic ratio Na/Pt is 10/1), and aged for 12 h at 80 o C. The solution is then evaporated in open air at 50 o C until dry. During evaporation, the substrate is rotated every 15-20 minutes to enhance uniformity. Finally, the sample is dried at 150 o C for 12 h and calcined in air at 500 o C for 4 h with a ramp rate of 2 o C min -1 .
To evaluate their hydrothermal stability in the probe reactions (CO and C 3 H 6 oxidation) and the simulated exhaust (clean diesel combustion, CDC; and low temperature combustion-diesel, LTC-D) tests, the monolithic catalysts are aged in a flow of 12 % O 2 , 6 % CO 2 , 6 % H 2 O in N 2 at 700 o C for 100 h. In the transient engine tests, the sample is aged in a flow on 10 % H 2 O in air at 650 o C for 100 h. These aging conditions are chosen to represent a North America heavy-duty diesel application.

Catalytic activity evaluation
The catalytic activity measurements were performed in a continuous flow reactor equipped with a gas chromatography. The monolithic catalysts (5 × 5 channels × 2.5 cm long) were placed in a tubular quartz reactor with a surrounding electrical heating coil. The inlet temperature was measured using a 0.16 mm K-type thermocouple, which was fixed at < 5 mm in front of the monolith. The utilization of the small thermocouple (0.16 mm) is to prevent disruption of gas flow. Typically, the catalytic activity of each sample was evaluated through total 6 reaction cycles, in which the inlet temperature was ramped up from 100 o C to 525 o C at 2 o C min -1 and naturally cool down to 100 o C before next testing cycles. The feed gas mixtures were controlled by mass flow controller before entering the blender and then the reactor at a gas hourly space velocity of 30,000 h -1 . The feed gas always contains 12% O 2 to mimic the O 2 concentration in the CDC and LTC-D simulated exhaust, but varies CO and C 3 H 6 concentrations (also were chosen to reflect the CO and total hydrocarbon concentration in the CDC and LTC-D simulated exhausts) in each cycle such that: cycle 1 and 2 contain 2000 ppm CO and 1000 ppm C 3 H 6 ; cycle 3 contains 2000 ppm CO; cycles 4 contains 1000 ppm C 3 H 6 ; cycle 5 contains 500 ppm C 3 H 6 ; and cycle 6 contains 500 ppm CO and 500 ppm C 3 H 6 ; all are balanced by N 2 . Sample Array-50-D was evaluated in an additional sequence (cycle 345612), showing negligible difference in the catalytic activity, thus confirming the sequence of the tests does not affect the catalytic activity of TiO 2 nano-array supported Pt catalysts.
The catalytic diesel oxidation activity of catalysts under simulated exhaust conditions was evaluated according to the protocol developed by the Advanced Combustion and Emission Control (ACEC) Technical Team of USDRIVE. The light-off measurements were conducted on a customized plug-flow reactor system. Micro-cores of catalyst samples were cut and loaded into cylindrical quartz tubing using a wrapping in order to ensure no gas could bypass the catalyst channels. The catalyst temperature was monitored using two thermocouples, one measuring the inlet temperature while located ~2 cm from the catalyst front, the other measuring the mid-catalyst temperature and located from the rear into the middle of the central channel of the micro-core. Space velocity was kept at 60,000 h -1 throughout all the tests. Gas concentrations were determined per ACEC Tech Team (USDRIVE) protocol. The composition of the "LTC-D" simulated exhaust is [ 2 , and H 2 O components of the gas stream at 700 °C for 4 h for LTC-D and 550 o C for 4 h CDC protocols, followed by an evaluation step from 100 °C to 500 °C at a rate of 2 °C min -1 . Besides the regular simulated exhaust test, catalyst samples were also treated using the following simulated exhaust test treatment protocol to look into the single Pt atom species fate before and after experiencing the treatment on the Pt/TiO 2 NA samples with both fresh state and hydrothermal aged condition. CDC simulated gas treatment: 12% O 2 + 6% H 2 O + 6% CO 2 +100ppm H 2 + 500ppm CO + 200ppm NO + 1400ppm HCs, with a ramp rate of 5 o C min -1 from room temperature to 500 o C, and then keeping at 500 o C for 1h.
Sulfur poisoning effects on the catalyst performance were evaluated following USDRIVE's protocol. The catalyst was exposed to 5 ppm SO 2 added to the full simulated exhaust at a space velocity of 30,000 h -1 and 300 o C catalyst inlet temperature for 5 h. A total sulfur exposure level of 1 g sulfur per liter of catalyst was estimated following this exposure condition. Poisoning was conducted following the pretreatment in which the catalyst was annealed at 600 o C for 20 minutes before cooled down to 300 o C in 12 % O 2 , 6% H 2 O, 6 % CO 2 , and N 2 balance. After sulfur exposure, SO 2 was removed from the feed and the sample cooled to 100 °C at which point the catalyst performance was evaluated. For de-sulfation, the catalyst was annealed at 700 o C in 3% H 2 and 1% CO in N 2 for 30 mins, followed by calcination in air at 500 o C for 2 h.
The Pt/TiO 2 nano-array sample with a Pt loading of 0.53 g L -1 and dimensions of Ф 2" × 3" were tested under highly transient feed gas conditions mimicking a heavy-duty diesel (HDD) federal test procedure (FTP) as running on a HDD certified 2010 Cummins ISB (6.7 L) 320 hp engine. The characteristic of the transient feed gas is provided in Supplementary Fig. 19. The Pt/TiO 2 nano-array and reference samples were aged at 650 o C for 100 h in 10% steam/air flow to a representative end of life condition for a North American HDD application.

Catalyst characterization
The X-ray absorption fine structure data was collected at the BL14W1 station at the Shanghai Synchrotron Radiation Facility (SSRF). The storage ring of SSRF was operated at 3.5 G eV with a maximum current of 260 mA. Pt foil and PtO 2 were used as reference samples. Using Si (111) double-crystal monochromator, the data collection was carried out in fluorescence mode for Pt L 3 -edge with a 7-element detector Ge solid state detector. The energy was calibrated according to the absorption edge of Pt foil (edge energy: 11564eV).
The EXAFS data was analyzed and fitted using the ATHENA module in the IFEFFIT software packages. Each k 2 -weighted EXAFS spectrum was processed by a two-step procedure: i) subtraction of post-edge background from overall absorption; and ii) normalization with respect to the edge-jump step. The k 2 -weighted χ(k) data of Pt L 3 -edge were then Fourier transformed to real (R) space through a hanning windows (dk=1.0 Å -1 ) in order to separate the EXAFS contributions from different coordination shells. To determine the quantitative structural parameters around central atoms, the ARTEMIS module of IFEFFIT software packages was employed for least-squares curve parameter fitting.
Coordination number of model samples (Pt foil) was fixed as the nominal value. The obtained S 0 2 of Pt foil was 0.80. It was fixed in the subsequent fitting of Pt L 3 -edge date. The fitted ranges for k and R spaces were selected to be k =3-10.3 Å -1 (k 2 weighted) with R=1.0-2.2 Å (Pt 1 /TiO 2 NW) or 1.4-3.2 Å (Pt NP/TiO 2 NA).
To conduct the H 2 temperature programmed reduction (TPR), the samples were degassed in helium at ambient temperature for 20 minutes. TPR condition: 10% H 2 , balance Ar, 50 ml min -1 , 10 °C min -1 . To determine the surface area and pore size distribution of the catalysts, N 2 adsorption-desorption isotherms and H 2 chemisorption were carried out using a Micromeritics ASAP 2020 Physisorption analyzer at liquid nitrogen temperature. Brunauer-Emmett-Teller (BET) method and Barret-Joyner-Halenda (BJH) method were used to calculate the specific surface area and pore size, respectively. . The samples to be measured were degassed at 350 o C for 4 h in vacuum before the measurement of N 2 adsorption-desorption isotherm.
The morphology and structure of catalysts were characterized by electron microscopies. Scanning electron microscopy images were taken using FEI Teneo low vacuum SEM and a JEOL 6335F field emission SEM, operating at 10-20 kV. Transmission electron microscopy (TEM) including both bright field and high angular annual dark field (HAADF) images and energy-dispersive X-ray spectroscopy for composition distribution were taken using a FEI Talos STEM and a Tecnai F30 STEM. The aberration corrected HAADF STEM images were obtained on two STEM/TEM systems: 1) a JEOL JEM 2200FS, attached with a CEOS (Heidelburg, Germany) probe corrector, with a nominal image resolution of 0.07 nm; and 2) a JEOL JEM-ARM200F, equipped with a CEOS probe corrector, with a guaranteed resolution of 0.08 nm. The chemical states of the TiO 2 and Pt/TiO 2 nanowire samples were characterized using a Thermo K-Alpha X-ray photoelectron Spectrometer with an Aluminum K-Alpha 1.486 KeV source. X-ray diffraction (XRD) patterns of the samples were collected by a Bruker D2 Phaser using Cu Kα radiation (λ = 0.15418 nm) operated at 40 kV and 40 mA. The XRD data were recorded for 2θ values from 10° to 80° with an interval of 0.02°. H 2 chemisorption was used to determine the platinum dispersion (D Pt ) at 35 o C on a Micromeritics ASAP 2020 analyzer. H 2 reduction treatment of samples was conducted at 300 o C for 2 h, followed by the outgassing at 150 o C for 2 h prior to chemisorption experiment. A double isotherm method was used to determine the H 2 uptake: first isotherm revealed the total hydrogen consumed (uptake), (HC) T , while the second one, acquired after evacuation for 2 h showed the ''reversible'' or weakly adsorbed hydrogen, (HC)r. The amount of ''irreversible'' or strongly adsorbed hydrogen, (HC)i, was determined by calculating the isotherm difference. The pressure range of isotherms was controlled at 0-12 kPa, where the extrapolation to zero pressure was used to measure gas uptake. The D Pt was finally calculated by an atomic (HC)i/Pts = 1 ratio, where Pts implies a Pt atom on surface 3 .
To evaluate the mechanical stability of TiO 2 nano-array integrated cordierite honeycomb samples, a sonication experiment was conducted on both TiO 2 and Pt/TiO 2 nano-array samples before and after hydrothermal aging at 700 o C for 100 h. The samples were sonicated in water at 40 kHz for 3 h and weighed before and after the sonication. Little morphology change (in SEM) was observed and the weight loss after the sonication process is less than 1 wt. % for both fresh and aged samples, indicating the sound honeycomb structure integrity and the superior adhesion of TiO 2 and TiO 2 /Pt nano-arrays on the honeycomb monolith channel walls.
Density Functional Theory (DFT) calculations were performed as implemented in the Vienna Ab initio simulation package 4, 5 , employing the Perdew, Burke and Ernzerhof exchange-correlation functional 6 . Models for rutile TiO 2 (110) surfaces contain two layers along the y direction followed by a vacuum slab of more than 20 Angstrom. The (110) surface was constructed by repeating the cell of dimension 13.16Å  11.88Å along the x and z directions, respectively. With this model, a k-point mesh of 212 and a plane-wave energy cutoff of 400 eV were used. As systems to be modeled involve isolated molecules like CO 2 and O 2 , the long-range van der Waals interactions was included into the calculations using the scheme of Grimme 7 . Reaction pathways were searched using the nudge-elastic band method by Henkelman and Jónsson 8 .  found (a, b), the majority Pt was distributed as single atoms (c, d) and several sub-nm clusters (e, f). Supplementary Fig. 6 ac-HAADF STEM images of Pt/TiO 2 nano-array prepared by microwave-assisted dip coating. Pt nanoparticles were mainly deposited on the adjacent SiO 2 area while very few were visible on TiO 2 (a, b). On TiO 2 , Pt was dominantly distributed single atoms (d-f) and several subnanometer clusters (c).  Fig. 10 ac-HAADF STEM images impregnation Pt 1 /TiO 2 nano-array after hydrothermal degreening at 700 o C for 4 h. Pt single atom dispersion is revealed on TiO 2 (a-b), while Pt is distributed as nanoparticles on SiO 2 (a) and sintered after degreening. Supplementary Fig. 11 ac-HAADF STEM images of dip-coating Pt 1 /TiO 2 nano-array hydrothermal degreening at 700 o C for 4 h. Pt single atom dispersion is revealed on TiO 2 (a-b), while on the adjacent SiO 2 , Pt is distributed as big nanoparticles (c). Fig. 12 Different configurations of isolated Pt atom supported on rutile TiO 2 (110) and their corresponding binding energies. The binding energy is defined as the energy needed to remove Pt atom from the surface to vacuum (distance of at least 10 Å from the surface). Supplementary Fig. 13 Arrhenius plot for CO oxidation on 0.2 wt.% Pt 1 /TiO 2 NR. The feed gas is 2000 ppm CO, 12% O 2 , and N 2 balance to reflect the CO concentration in the LTC-D simulated exhaust.

Pre-Sulfation
Supplementary Fig. 19 The characteristics of the transient feed gas.  a) S 0 2 was fixed as 0.8 during EXAFS fitting, based on the known structure of Pt foil. b) CN is the coordination number. c) Interatomic distance is the bond length between Pt central atoms and surrounding coordination atoms. d) Debye-Waller factor is a measure of thermal and static disorder in absorber-scattering distances. e) ΔE 0 is the difference between the zero-kinetic energy value of the sample and that of the theoretical model. The cordierite substrate etched in the similarly acidic conditions but without additional titanium precursor displays high surface area, but loses mechanical integrity in the form of the structural collapse due to the excess porosity induced by overetching. Evidently, the formation of densely packed TiO 2 nano-array layer may have protected the cordierite substrate from aggressive acidic etching during the solvothermal growth.

Supplementary Note 2
The core Pt 4f spectra can be deconvoluted to Pt-II and Pt-IV peaks from the XPS data collected from the Pt/TiO 2 NW samples ( Fig. 2c and Supplementary Fig. 21). No metallic Pt (Pt-0) peaks were detected in all the samples, as summarized in the Supplementary  Table 3). This result is consistent with the Pt L 3 -edge EXAFS fitting result shown in the Supplementary Table 2.
Oxygen vacancies in TiO 2 can be monitored through monitoring Ti 3+ since the electrons left in the vacancy are transferred to the neighboring Ti atoms, forming two Ti 3+ cations per one oxygen vacancy. As shown in Supplementary Fig. 22, the core Ti2p XPS spectra of both TiO 2 NW powder and Pt 1 /TiO 2 NW powder show only peaks at 464.3 eV and 458.4 -458.5 eV, corresponding to the Ti2p1/2 and Ti2p3/2 of Ti 4+ . Since XPS is a surface sensitive technique with a penetration depth of ~2-10 nm, this means that there is no surface Ti 3+ and oxygen vacancies. This result is expected since the synthesis process of Pt 1 /TiO 2 NA and NW involves a calcination process at high temperature, typically at 500 o C, in which surface Ti 3+ , if existed, can be easily oxidized to Ti 4+ . This observation is consistent with the results obtained in the NA and NW samples calcined at the different temperatures (700 o C, 800 o C, and 900 o C, respectively).