Journal of Catalysis 302 (2013) 58–66

Contents lists available at SciVerse ScienceDirect

Journal of Catalysis

journal homepage: www.elsevier .com/locate / jcat

Photocatalytic synthesis of oxygenated hydrocarbons from diesel fuel for mobiledeNOx application

Jae Yul Kim a, Yeon Ho Kim a, Suenghoon Han a, Sun Hee Choi b, Jae Sung Lee c,⇑a Department of Chemical Engineering, Division of Advanced Nuclear Engineering, Pohang University of Science and Technology, 77 Cheongam-ro, Nam-Gu, Pohang 790-784,Republic of Koreab Pohang Accelerator Laboratory, Pohang University of Science and Technology, 77 Cheongam-ro, Nam-Gu, Pohang 790-784, Republic of Koreac School of Nono-Bioscience and Chemical Engineering, Ulsan National Institute of Science & Technology, 50 UNIST-gil, Ulsan, 689-798, Republic of Korea

a r t i c l e i n f o a b s t r a c t

Article history:Received 29 September 2012Revised 1 March 2013Accepted 5 March 2013

Keywords:Photocatalytic partial oxidationOxygenated hydrocarbonsTiO2–SiO2 mixed oxidesNOx reductionDiesel vehicle

0021-9517/$ - see front matter � 2013 Elsevier Inc. Ahttp://dx.doi.org/10.1016/j.jcat.2013.03.003

⇑ Corresponding author. Fax: +82 54 279 5528.E-mail address: jlee@postech.ac.kr (J.S. Lee).URL: http://www.ecocat.tistory.com (J.S. Lee).

Photocatalytic partial oxidation of dodecane has been studied as a model reaction of diesel fuel conver-sion to oxygenated hydrocarbons (OHCs) as an effective nitrogen oxide (NOx) reductant in selective cat-alytic reduction (SCR) systems. Thus, TiO2-based photocatalysts produced OHCs composed mainly of C1–C6 aldehydes under UV irradiation, and TiO2–SiO2 mixed oxide photocatalysts showed higher selectivityand yield to OHCs than pristine TiO2 by diluting contiguous Ti sites and suppressing complete oxidationto CO2. The effects of reaction variables were studied in detail. A novel scheme of NOx after-treatmentsystem for diesel engine exhaust line was proposed involving the new photocatalytic reaction, whereon-board photocatalytic partial oxidation of a small amount of diesel fuel produced OHCs that were sup-plied to the deNOx system as NOx reductant. Although further improvement is needed in the selectivityfor OHCs as well as the system operability, the proposed scheme could be a more environment-friendlyoption than the reduction by urea, currently considered the most promising technology.

� 2013 Elsevier Inc. All rights reserved.

1. Introduction

Diesel engine vehicles are the major source of harmful nitrogenoxides (NOx) emission on the road. Especially NOx emission fromheavy duty diesel vehicles accounts for almost 30% of total emis-sion despite the number of vehicles is limited to only 2% in Califor-nia, USA [1]. The fact that all diesel cars must satisfy Bin 5(NOx 6 0.07 g/mi) [2] and Euro VI (NOx 6 80 mg/km) [3] compli-ancy to be sold in the USA and Europe markets drives diesel carmanufactures to develop more efficient NOx after-treatment tech-nology than the past [4].

There have been a huge number of reports concerning leanNOx selective catalytic reduction (SCR) technology categorizedby reducing agent such as urea (or ammonia) [5–7], hydrocarbon(HC) [8], oxygenated hydrocarbons (OHCs) [2,9–12], and hydro-gen [13]. These reports show highly practical potential of leanNOx SCR technology for application to the exhaust line of realdiesel engine vehicle fleet with more than 50% NOx conversionat relatively low temperatures of 200–250 �C, similar to diesel ex-haust temperatures under lean conditions. These technologies areproposed on the assumption that those highly effective reduc-tants such as NH3, HCs (short and long chain alkanes and

ll rights reserved.

alkenes), and OHCs (alcohols and aldehydes) are sufficiently pro-duced somehow and delivered to the exhaust gas after-treatmentconvertor filled with a NOx reducing catalyst. In reality though,the reducing gases in the engine exhaust stream are insufficient;NH3 (none), HCs–OHCs (0–600 ppm carbon, varied by drivingcondition) [14,15] and H2 (almost none). Thus, additional supplyof reducing agents to the NOx SCR convertor is necessary forpractical application. Currently, urea is considered the mostpromising reductant because it generates NH3, the strong reduc-tant of NOx at low temperatures. Yet, ammonia is toxic andcarrying a urea tank on board is neither convenient nor safe. Alsoconversion of urea into ammonia does not reach 100%. Diesel fuelitself can be directly used as a convenient reductant, but it is notreactive enough at the diesel exhaust temperatures under leanconditions [16,17].

As an alternative technology to supply NOx reductants onboard, herein we report photocatalytic partial oxidation of dode-cane, a model compound of diesel fuel, into OHCs by using TiO2–SiO2 mixed oxide photocatalysts in a simple continuous flowreaction system that could be installed in the vehicle (Scheme 1and Fig. S1 of Supporting information, SI). The performance ofthe photocatalytic reactions that we report here, especially selec-tivity to OHCs, is still rather low for practical applications.Further, installation of the photocatalytic reactor with a UV lampinside the vehicle may not be convenient or energy-efficient. Butif further improvement is achieved in the reaction selectivity and

J.Y. Kim et al. / Journal of Catalysis 302 (2013) 58–66 59

the system operability, it could be more environment-friendlythan carrying the urea tank on board. The system produces OHCsbelow 100 �C, which are efficient NOx reductants and safer thanNH3. By using a very small amount of diesel fuel, the systemdoes not generate a significant fuel penalty (e.g., 6 mol of acetal-dehyde can be produced from 1 mol of dodecane). In a scientificpoint of view, the number of study regarding photocatalytic par-tial oxidation of organic compound is rather scarce than that ofphotocatalytic mineralization or total oxidation process [18,19].To the best of our knowledge, this work is the first experimentalresearch of photocatalytic partial oxidation of a diesel-like mol-ecule at low temperatures (<100 �C) without water.

2. Experimental

2.1. Catalyst synthesis

The TiO2–SiO2 mixed oxide was synthesized under typical hydro-thermal conditions. (Better be termed ‘‘solvothermal’’ with almostnon-aqueous solutions accompanied by heating in a Teflon autoclavereactor.) Thus, titanium (IV) sopropoxide (TTIP, 97%, Sigma–Aldrich),tetraethyl orthosilicate (TEOS, 98%, Acros Organics), tetrabutylammonium hydroxide solution (TBAOH, 1 M in methanol, Sigma–Al-drich), 2-propanol (100%, JT baker) were used without further purifi-cation. Mixed oxide of Ti/Si = 0.11 was made as follows:3 � (TTIP + TEOS) mole of 2-propanol, 0.0334 mol of TEOS,0.5 � (TTIP + TEOS) mole of TBAOH, 0.00371 mol of TTIP, 3 � (TTIP + -TEOS) mole of distilled H2O were dropped into the Teflon autoclavebottle in this order. This transparent yellowish solution was hydro-lyzed at 170 �C for 15 h followed by filtering with 500 ml of distilledH2O. Sample powder was dried overnight at 100 �C and calcined at500 �C for 10 h. Other samples with different ratios were synthesizedin a similar way with different precursor mole ratios. Details of thissynthesis procedure are illustrated in Table S1 and Scheme S1 of SI.Commercial TiO2 in anatase form (98.5%) was obtained from Junsei.

2.2. Catalyst characterization

The textural properties such as surface area, total pore volume,pore size distribution, and microporosity of TiO2–SiO2 mixed oxidewere determined using N2 physisorption measurements. The N2

adsorption–desorption isotherms were obtained at 77 K using ananoPOROSITY-XQ nanoporosity and surface area analyzer (Mirae-si, Korea). The surface areas were calculated by using the Bru-nauer–Emmett–Teller equation in a relative pressure range (P/P0)of 0.05–0.20 assuming a cross-sectional area of 0.162 nm2 for N2

molecule. The total pore volume and pore size distribution weredetermined from the amount of N2 adsorbed at the highest relativepressure of P/P0 = 0.995. The pore diameter and pore size distribu-tion plots were determined by applying the Barrett–Joyner–Halen-

Scheme 1. A photocatalytic partial oxidation system attached to exhaust line indiesel engine vehicle.

da (BJH) model to the desorption isotherm. Microporosity wasassessed by adapting t-plot method setting 3.5–7.0 Å as a statisti-cal thickness range.

The crystalline phase of the products was determined using X-ray diffractometer (Mac Science Co., M18XHF) with monochro-matic Cu Ka radiation at 40 kV and 200 mA. UV–Vis DRS were ta-ken by UV-2401PC Shimadzu). The X-ray absorption fine structure(XAFS) measurements were conducted on beam line 7D of PohangAccelerator Laboratory (PAL) in Korea and beam line 7C of PhotonFactory (PF) in Japan. The spectra for K-edges of Ti (E0 = 4966 eV)were taken in transmission mode at room temperature. The inten-sity of incident beam was monitored with He-filled IC Spec ioniza-tion chamber and that of transmitted beam with N2-filledchamber. The obtained data were analyzed with Athena in theIFEFFIT 1.2.9 suite of software programs [20].

Images of HRTEM, EELS, and SAED pattern were taken by HR-STEM-2200FS (JEOL JEM-2200FS with Image Cs-corrector) at Na-tional Center for Nanomaterials Technology, Korea. The XPS spec-tra were obtained with a VG-Scientific ESCALAB 220 IXLspectrometer equipped with a hemispherical electron analyzerand a Mg Ka (ht = 1253.6 eV) X-ray source. The FTIR analysiswas performed with a Perkin Elmer Spectrum 2000 Explore ma-chine. Each sample was made in 30 mg pellet diluted with KBr.The ICP analysis was conducted with ICP AES (Spectro-Vision).

2.3. Photocatalytic partial oxidation of dodecane

Photocatalytic reactions were performed in a continuous flowsystem shown in Fig. S1. The carrier gas (Ar) flew through a dode-cane saturator immersed in a constant temperature silicon oil bath.A 450 W Hg lamp (7825-34, Ace Glass) without cutoff filter wasused as a UV–Vis source. Light intensity was 0.81 W/cm2 measuredby a photometer (ORIEL 70260 with 70282 head). Tubular quartzglass was used as a cooling jacket and a light window, in whichcooling water was circulated. Ethanol solution of photocatalystpowders was sprayed over the glass rod on a 300 �C hot plate forcoating. Product compounds were analyzed by an on-line gas chro-matograph (HP 6890) equipped with a DB-5 column (Agilent Tech-nology, 125-5532) and 2 auto sampling valves for TCD and FIDanalysis. The products were identified with a GC (HP 6890)–MS(HP 5973) equipped with a DB-5 ms column (Agilent Technology,122-5563).

3. Results and discussion

3.1. Identification of products from photocatalytic partial oxidation ofdodecane over TiO2

Since photocatalytic partial oxidation of a diesel-like moleculein the anhydrous flow system has never been studied before, wefirst tried to understand the reaction behavior depending on oper-ating variables. Thus, at the beginning, we used commercial ana-tase phase TiO2 (Junsei) as the photocatalyst. Of course, TiO2 isthe best known photocatalyst for total oxidation of organic chem-icals to carbon dioxide or pollutants mineralization in various envi-ronmental applications [21–23]. Yet, photocatalytic partialoxidation is also possible although examples are rare [24–26]. Atfirst, we performed the photocatalytic reaction with 21% O2 in Arto mimic air-like atmosphere at 160 �C using TiO2 loaded on thequartz beads or rods as a photocatalyst. During the reaction, ther-mal decomposition of dodecane was not observed at these lowtemperatures (<160 �C) when light was turned off. We observedproduction of a large number of oxygenated hydrocarbons (OHCs)from photocatalytic partial oxidation of dodecane. Because of thiscomplicated product profile and since there was no previous re-ports on the reaction, we conducted a careful product analysis.

60 J.Y. Kim et al. / Journal of Catalysis 302 (2013) 58–66

Fig. 1 shows OHC products profiles of dodecane partial oxida-tion obtained by gas chromatographic (GC) analysis with a flameionization detector (FID) and GC–MS (inset). Small peaks of lessthan 12 a.u. between the large aldehyde peaks correspond to ke-tones and carboxylic acids. These compounds were identified bymass spectroscopy (Fig. S2 of SI) as listed in Table 1. Aldehydeswere the main OHC products (>93%) with a typical selectivity orderof C2 > C3 > C4 > C5 > C6 > C1. Aldehydes with more than 6 carbonswere also detected by GC–MS as shown in Table 1. The carboxylicacids and ketones (especially 2-ketones) were also formed. In theGC–MS analysis, carboxylic acids had relatively longer retentiontimes because of their large polarity, whereas aldehydes and ke-tones of the same carbon number showed similar retention times.Only a very small amount of H2 (<10 ppm) was also detected due tofurther oxidation to H2O under this oxidizing reaction condition. InTCD analysis, CO2 was found to be the other main product of thephotocatalytic oxidation of dodecane by total oxidation, but COwas not detected at all.

3.2. Effect of reaction variables on photocatalytic partial oxidation ofdodecane on TiO2

We varied several reaction variables such as space velocity, O2

concentration, and reaction temperature in order to find conditionsthat could give high yields of OHCs. The reaction reached the stea-dy state after ca. 1 h on stream, and selected data are presented inTable 2. As WHSV increased from 8.34 to 44.0 h�1 by reducing theamount of photocatalyst at the fixed total flow rate of 100 cc/min,conversion of dodecane decreased from 87% to 69% as expected.When 21% of O2 was flowed, conversion of dodecane was 4 timeshigher and selectivity of OHCs was about two times lower thanthose of 4% O2 at two WHSV conditions. As the reaction tempera-ture decreased from 160 �C to 80 �C at O2 concentration of 4%,the conversion was similar but OHCs selectivity increased from37% to 52%. Under the same temperature and O2 concentration,WHSV was halved by cutting the flow rate by half (50 cc/min), thenthe conversion increased from 24% to 47%. In this case, the OHCsselectivity was reduced somewhat, but the yield of OHCs improvedfrom 12.5% to 20.7% due to increased dodecane conversion. Sincethe reaction behavior was most significantly affected by O2 con-centration, it was reduced to even lower levels to 2.45% and 1%.The OHCs selectivity increased to as high as 70%, but conversionof dodecane decreased so much that there was no gain in OHCsyield.

Apparent turnover frequency (TOF) of this photocatalytic reac-tion was calculated based on surface titanium atom. The numberof surface atoms was calculated from the TiO2 surface area andthe area per Ti atom calculated from Ti–O bond length (1.939 Å)[27]. The TOF values in Table 2 vary around the order of 10�3

Fig. 1. Products profile from GC FID signal and mass chromatogram in GC–MS inthe inset. The noted C1–C6 stands for corresponding aldehydes.

and 10�2 s�1, which are not much different from those of usualthermal catalytic reactions. The OHCs selectivity data in Table 2and those obtained in other conditions were plotted in Fig. 2.Although widely different reaction conditions make the data pointsscattered, there is unambiguous trend that the OHCs selectivity de-creases while CO2 selectivity increases with increasing conversionof dodecane. Thus, it appears that reaction proceeds via a typicalseries reaction; dodecane ? [OHCs] ??? CO2.

3.3. TiO2–SiO2 mixed oxide photocatalysts

To increase OHCs selectivity and yield in the series reaction ofdodecane partial oxidation, we have to stabilize the intermediateOHCs against total oxidation to CO2. We conjectured that dilutionof active Ti sites could give better selectivity and employed TiO2–SiO2 mixed oxides as photocatalysts. When pure TiO2 was usedas a photocatalyst, all the carbon atoms of dodecane could contactthe surface titanium atoms, so that they are easily oxidized all theway to CO2 (Scheme 2a). If we dilute the active sites by inactiveSiO2 by forming small TiO2 nano-domains surrounded by SiO2,the high reactivity of TiO2 could be controlled and OHCs’ selectivitycould improve (Scheme 2b).

To validate this conjecture, TiO2–SiO2 mixed oxides with threeTi/Si atomic ratios were synthesized by a solvothermal methodwith almost non-aqueous solutions. To make a well-mixed stateon a molecular level in the liquid precursor mixture, it is importantto control the water content such that the precursors remain un-hydrolyzed before the temperature rises to 170 �C. If it is notcontrolled properly, the solution becomes opaque pale-yellowishdue to preferential hydrolysis of the Ti precursor.

Textural properties of the three mixed oxide samples with dif-ferent Ti/Si ratios (0.11, 0.33, and 3.00) were studied based on N2

physisorption at 77 K and compared with anatase TiO2. As shownin Fig. 3a, the N2 isotherms belonged to type-IV. The maximumadsorption was attained at some pressure between 0.8 and 1.0 P/P0, where P0 denoted the vapor pressure of N2. During the desorp-tion process, H2 hysteresis loop associated with the capillary con-densation took place, indicating the presence of mesopores usuallyformed in interparticle voids. Various depths of pores with wideand narrow holes, confirmed by BJH pore size distribution curveat Fig. 3b, brought about a disparity between adsorption anddesorption branches at the P/P0 range of 0.4–0.8 of the isotherms.As Ti/Si ratio increased, the magnitude of this disparity decreaseddue to the smaller pore volume. Increase in Ti composition causeda significant reduction of total pore volume from 0.7972 to0.1014 cm3/g as Ti/Si increased from 0.11 to 3.00 as shown inFig. 3b and Table S2. These pore volumes of the mixed oxides werelarger than that of TiO2 except the one with Ti/Si = 3.00. The resultsof BET surface area were consistent with these pore structures. TheBET surface area of TiO2–SiO2 mixed oxides were an order of mag-nitude higher than TiO2 due to the contribution of SiO2 of high sur-face area. As Ti/Si ratio increased, it decreased dramatically from754.8 to 193.6 m2/g. To investigate microporosity of samples, t-plot method was applied as shown in Fig. 3c. All data of the mixedoxides with different Ti/Si ratios were estimated from interceptpositioned at y axis by extrapolation. Almost none of the microp-ores was found in the sample with Ti/Si = 0.11, whereas two othersamples showed about 42% of micropores. The number of Ti atomson the catalyst surface was calculated from the assumption that ahalf of Ti atom per ð

ffiffiffi

2p� 1:939Þ2Å

2was exposed [27], and that

TiO2 surface area of the mixed oxide was proportional to the ratioof Ti/(Ti + Si) obtained either by ICP or XPS analysis.

The XRD patterns of the prepared photocatalysts are comparedwith pure anatase TiO2 and amorphous SiO2 in Fig. 4a. As Ti/Si ra-tio increases, intensity of amorphous SiO2 peak at 2h = 12–38� de-creases and the (101) peak of anatase TiO2 at 25.4� appears and

Table 2Photocatalytic partial oxidation of dodecane over TiO2 under UV light.a

WHSVb (h�1) O2 (%) Temp (�C) Conversion (%) Selectivity (%) OHCs yield (%) TOF (s�1) Loading materiald

OHCs CO2

44.0 21 160 66 31 69 20.5 1.40 � 10�2 Q.B.16.8 21 160 72 30 70 21.6 1.57 � 10�2 Q.B.8.34 21 160 91 26 74 23.7 2.09 � 10�2 Q.B.8.34 21 160 91 26 74 23.7 2.09 � 10�2 Q.B.8.34 4 160 21 46 54 9.66 4.28 � 10�3 Q.B.16.8 21 160 89 17 83 15.1 1.85 � 10�2 Q.R.16.8 4 160 25 39 61 9.75 4.62 � 10�3 Q.R.16.8 4 160 25 39 61 9.75 5.33 � 10�3 Q.R.16.8 4 80 24 52 48 12.5 5.12 � 10�3 Q.R.8.39c 4 80 47 44 56 20.7 8.03 � 10�3 Q.R.8.39c 2.4 80 37 49 51 18.1 9.33 � 10�3 Q.R.16.8 1 80 <1 70 30 – – Q.R.8.34 1 80 8.7 61 39 5.31 1.83 � 10�3 Q.R.

a Initial dodecane concentration: 450 ppm. Reaction was terminated at 6 h on stream.b Total flow rate: 100 cc/min.c Total flow rate: 50 cc/min.d Q.B.: quartz bead, Q.R.: quartz rod.

Table 1The OHC products identified by GC–MS.a

Aldehydesb Carboxylic acidsc Ketonesc

RT (min) Compound RT (min) Compound RT (min) Compound

4.107 Formaldehyde 8.787 Acetic acid 9.560 2-Butanone5.000 Acetaldehyde 12.653 Propanoic acid 13.305 2-Pentanone6.773 Propanal 18.058 Butanoic acid 19.948 2-Hexanone9.892 Butanal 24.313 Pentanoic acid 25.926 2-Heptanone

14.485 Pentanal 30.630 Hexanoic acid 33.145 2-Octanone21.365 Hexanal 37.248 Heptanoic acid 38.895 2-Nonanone27.315 Heptanal 41.316 Ocatanoic acid 42.586 2-Decanone34.787 Octanal 44.357 Nonanoic acid40.054 Nonanal43.594 Decanal

a Analyzed at a split ratio of 20b or 10c. Each product at a given retention time (RT) was identified by mass spectroscopy signals.

Scheme 2. Photocatalytic oxidation of dodecane: (a) total oxidation of dodecane isdominant when pure TiO2 is used. (b) Degree of partial oxidation of dodecaneincreases when TiO2–SiO2 mixed oxide is used. Size of TiO2 particle is less than3 nm.

Fig. 2. Relationship between dodecane conversion and OHCs selectivity. All pointswere acquired from different reaction conditions.

J.Y. Kim et al. / Journal of Catalysis 302 (2013) 58–66 61

gradually grows [28–32]. But the diffuse XRD patterns of mixedoxides do not provide clear structural information concerningcrystalline phase of TiO2 clusters and their dispersion in silica.Thus, we employed X-ray absorption near edge structure (XANES)analysis to investigate their short range local structure. In Fig. 4b,a strong single pre-edge peak at 4971 eV is dominant when theTi/Si ratio is low. This peak represents Ti in a tetrahedral symme-try forming Ti–O–Si bond network [33,34]. From the sample withTi/Si = 0.11, a weak but observable peak next to main pre-edgepeak starts to grow at 4975 eV. This indicates that Ti of an octa-hedral symmetry begins to form in the Ti–O–Ti bond network.

The three weak pre-edge peaks representing octahedral symme-try found in anatase phase TiO2 are detected in the sample withTi/Si = 3.00 together with the main pre-edge peak of tetrahedralTi sites [20,35].

We also performed quantitative analysis of Ti in Ti–O–Si and Ti–O–Ti states from pre-edge fitting by linear combination of XANES(Fig. S3, Table S3) [36]. XANES of the sample with Ti/Si = 0.02 andanatase TiO2 were selected as reference spectra of maximum Ti–O–Si and Ti–O–Ti, respectively. The normalized concentration of

Fig. 3. (a) N2 adsorption isotherms, (b) BJH pore size distributions, and (c) microporsity configuration from t-plot method for TiO2–SiO2 mixed oxides with different Ti/Siratios.

62 J.Y. Kim et al. / Journal of Catalysis 302 (2013) 58–66

Ti–O–Si varied from 0.696 to 0.194 and that of Ti–O–Ti from 0.304to 0.806, as Ti/Si increased from 0.11 to 3.00. The fact that 19.4% ofTi atoms are located in the tetrahedral symmetry that has to beformed only at boundary of TiO2 cluster interfaced with SiO2 ma-trix suggests that the sizes of TiO2 clusters are still small even atthe highest Ti/Si ratio of 3.00. As shown in Fig. S4, their actual sizewas less than 8 nm. From this result, we can confirm that nano-sized TiO2 clusters surrounded by SiO2 are formed and well distrib-uted even at high Ti/Si ratios as confirmed again by HRTEM imagesbelow.

Fig. 5 shows UV–Vis DRS of the samples. Blue shifts of band gapabsorption edge were observed as Ti/Si ratio decreased as expectedfor small TiO2 particles of less than 3 nm [37–41]. The maximumshift observed was about 22 nm (0.22 eV) with respect to bulkTiO2, when Ti/Si is 0.11. This shift is expected for TiO2 crystallineclusters with domain sizes below 5 nm.

To directly confirm the size and extent of dispersion of TiO2

clusters, TiO2–SiO2 mixed oxide of Ti/Si = 0.11 was examined byHRTEM, SAED, and EELS mapping images as shown in Fig. 6. TheTiO2 domains of 1–4 nm was identified as anatase TiO2 latticeimages surrounded by amorphous silica network. The SAED pat-tern in the inset shows spots superimposed on diffused ring, whichis consistent with crystalline TiO2 domains, dispersed in amor-phous silica network. The EELS mapping image also shows

well-dispersed Ti domains in Si and O [42,43]. The images ofTiO2–SiO2 mixed oxide with Ti/Si = 3.00 exhibited a lot more indi-vidual TiO2 crystal domains with larger sizes, but they were stillless than 8 nm on the average (Fig. S4) [44].

The TiO2–SiO2 mixed oxide photocatalysts were further charac-terized by XPS and FTIR as shown in Fig. 7. Binding energy (BE) of Ti2p3/2 for commercial TiO2 was 458.5 eV. As Ti/Si ratio decreasedfrom 3.00 to 0.11, the BE progressively increased(458.94 eV ? 459.2 eV ? 459.38 eV), while peak intensity de-creased. As the Ti content increased in the mixed oxides, BE of Si2p (Fig. 7b) shifted to smaller values. The O 1s peak (Fig. 7c andd) showed the clear shift from the position of TiO2 to that of SiO2

as Ti/Si ratio decreased. This could be understood by the formationof Ti–O–Si bonds by atomic infiltration of Ti atoms into SiO2 latticeand at the TiO2–SiO2 interface. Because Si atom is more electroneg-ative and less polarizable than Ti, decrease in effective negativecharge around Ti atom in Ti–O–Si makes BE of Ti 2p3/2 larger thanthat of pure TiO2 [45–47]. For the same reason, BE of Si 2p for themixed oxides decreased from that of SiO2 as Ti/Si ratio increased.The deconvolution of O 1s peak in Fig. 7d indicates that the fractionof oxygen bound to Ti increases as Ti/Si ratio increases as expected.

In FTIR spectra of Fig. 7e, three peaks in the mixed oxide sam-ples at 1080, 940, and 800 cm�1 are assigned to asymmetricm(Si–O–Si) stretching vibration, m(Ti–O–Si) stretching vibration,

Fig. 5. UV–Vis DRS of TiO2–SiO2 mixed oxides with different Ti/Si ratios.

Fig. 4. (a) XRD and (b) K-edge XANES spectra with Ti/Si ratio between 0.02 and3.00. (c) The ratio of Ti in Ti–O–Si and Ti–O–Ti from pre-edge fitting of TiO2–SiO2

mixed oxides of different Ti/Si ratios. Pure SiO2 was synthesized with the sameprocedure without Ti precursor.

J.Y. Kim et al. / Journal of Catalysis 302 (2013) 58–66 63

and symmetric m(Si–O–Si) stretching vibration, respectively[29,48–51]. The fact that intensity of m(Ti–O–Si) increased as Ti/Si ratio decreased (see inset) indicated that fraction of Ti atom en-gaged in Ti–O–Si bond increased. At the same time, m(Si–O–Si) sig-nal increased because of the higher fraction of Si in the catalyst.Thus, the highest m(Ti–O–Si) intensity was observed for the sampleof Ti/Si = 0.11 sample, although total amount of Ti in this samplewas the lowest. The catalyst of Ti/Si = 3.00 showed the lowestintensity of m(Ti–O–Si) and m(Si–O–Si) due to the largest TiO2 clus-ters surrounded by the relatively small amount of SiO2 matrix asconfirmed by the TEM image in Fig. S4.

3.4. Photocatalytic partial oxidation of dodecane over TiO2–SiO2 mixedoxide photocatalysts

Photocatalytic partial oxidation of dodecane was performed onthese catalysts, and the results are depicted in Figs. 8 and S5. Reac-tion was performed at 80 �C under UV–Vis illumination with a total

flow rate of 100 cc/min (285 ppm of dodecane, 11% O2 with Ar bal-ance) for 0.4 g of catalyst coated on quartz rods. Carbon balance be-tween reacted dodecane and all calibrated products was more than90%. The mixed oxide catalyst of Ti/Si = 3.00 showed the highestconversion (�56%) followed by pure TiO2 (�48%) and other mixedoxides (�46% for Ti/Si = 0.33, �41% for Ti/Si = 0.11). Most conspic-uous in Fig. 8b was much enhanced selectivity (�36%) for catalystof Ti/Si = 0.11, followed by Ti/Si = 0.33 (�29%) and TiO2 (�22%). Thecatalyst of Ti/Si = 3.00 showed the lowest selectivity (�14%). If wecount all OHCs composed of more than 7 carbons that weneglected because of their low concentrations, the real OHCs selec-tivity should become slightly larger. Order of OHCs yield was alsosimilar to selectivity trend; Ti/Si = 3.00 < TiO2 6 Ti/Si = 0.33 < Ti/Si = 0.11. Thus, dodecane conversion, OHCs selectivity, and yieldof TiO2–SiO2 mixed oxide samples do have correlation with Ti/Siratio. As Ti/Si ratio decreases, conversion of dodecane decreasedbut OHCs selectivity and yield increased. This could be attributedto the well-dispersed TiO2 clusters [52] that control the excessivecarbon–carbon bond fission of dodecane on contiguous Ti sites pre-valent on pure TiO2 leading to total oxidation to CO2.

There is one exception for the favorable effect of TiO2–SiO2

mixed oxide on OHCs yield. The mixed oxide with Ti/Si = 3.00showed higher dodecane conversion, but lower OHCs selectivitythan those of TiO2. The photocatalyst had more surface Ti atoms(Table S2) than TiO2, which would lead to higher dodecane conver-sion. But their properties including the large size (>8 nm) andmostly Ti–O–Ti character (�80%) would not suitable for the dilu-tion effect of the contiguous Ti sites that offer the improved OHCsselectivity. Hydrogen was also produced in less than 10 ppm onlywhen mixed oxide was utilized because almost all of H2 were oxi-dized to H2O under the present conditions (Fig. S5).

The highest OHCs yield (dodecane conversion � OHCs selectiv-ity) was obtained when TiO2–SiO2 (Ti/Si = 0.11) was used and itsvalue was 1.5 times larger than that of pure TiO2 (15% vs. 10%).Since the selectivity is a sensitive function of conversion, the selec-tivity-conversion data were plotted in Fig. S6 for various amountsof the catalysts. The excellence of the catalyst of Ti/Si = 0.11 wasobvious in the plot as its points were all located above the valuesfor other photocatalysts. If we compare the amounts of producedOHCs per Ti mass, the mixed oxide with Ti/Si = 0.11 produced13.03 times larger moles of C in OHCs than the commercial TiO2.

The OHCs selectivity and yield obtained here in this preliminarystudy are still low to be applied to a practical deNOx system fordiesel exhaust. Total oxidation to CO2 is still prevalent. But defi-nitely there is a lot of room for improvement from further optimi-zation of catalyst composition and reaction conditions. Ofparticular interest is the effect of diluting contiguous Ti sites bydispersing them on SiO2 matrix, which improves the OHCs selec-tivity significantly. Further enhancement of OHCs selectivity is

Fig. 7. X-ray photoelectron and FTIR spectra. (a) Ti 2p XPS spectra, (b) Si 2p XPS spectra, (c) O 1s XPS spectra, (d) deconvolution results of O 1s, and (e) FTIR of TiO2–SiO2 mixedoxides for different Ti/Si ratios.

Fig. 6. (a) HRTEM image and (b–e) EELS mapping image of TiO2–SiO2 mixed oxide with Ti/Si = 0.11. White circles in (a) indicate TiO2 cluster. Inset of (a) presents an SAEDpattern.

64 J.Y. Kim et al. / Journal of Catalysis 302 (2013) 58–66

Fig. 8. Reaction results of photocatalytic partial oxidation of dodecane; (a)conversion of dodecane; (b) selectivity of OHCs and (c) yield of OHCs.

J.Y. Kim et al. / Journal of Catalysis 302 (2013) 58–66 65

anticipated if Ti is distributed in smaller sizes or even in atomicscale. Research along this line is in progress. From the stabilitypoint of view, reaction lasted more than 20 h without deactivationas observed from Fig. 8. In addition, the color of photocatalyst sur-face remained white during the period, indicating no carbondeposition.

In this work, we propose a novel scheme of NOx after-treatmentsystem of diesel engine vehicle exhaust line involving the newphotocatalytic reaction, where on-board photocatalytic partial oxi-dation of a small amount of diesel fuel produced OHCs that weresupplied to the NOx treatment system as NOx reductant. The per-formance of the photocatalytic reactions that we report here, espe-cially selectivity to OHCs, is still rather low for practicalapplications. Further, installation of the photocatalytic reactor witha UV lamp inside the vehicle may not be convenient or energy-effi-cient. But if we can improve the selectivity of the photocatalyticpartial oxidation for desired OHC production, the proposed schemeis more environment-friendly than the reduction by urea, currentlyconsidered the most promising technology that has to carry the

urea tank on board. The system produces OHCs which are efficientNOx reductants below 100 �C and safer than NH3. By using a verysmall amount of diesel fuel, the system does not generate a signif-icant fuel penalty.

4. Conclusion

In the present report, a new reaction has been proposed to sup-ply reagents for on-board NOx reduction in diesel engine exhaustline; photocatalytic partial oxidation of diesel molecules to oxy-genated hydrocarbons made mostly of highly reactive C1–C6 alde-hydes. To enhance selectivity of desired aldehydes, we made aneffort to disperse active TiO2 clusters in SiO2 matrix to dilute thecontiguous Ti sites that are responsible for non-selective completeoxidation of dodecane molecules. As a result, TiO2–SiO2 mixedoxide of Ti/Si = 0.11 showed significantly improved selectivity overthe pure TiO2 particles. Further enhancement of OHCs selectivity isanticipated if Ti is distributed in smaller sizes or even in atomicscale.

Scientifically, the reaction has never been studied before and is arare example of photocatalytic partial oxidation for cracking andfunctionalizing a long chain organic alkane into smaller OHC mole-cules. In the process, the reactivity of TiO2, widely recognized as a to-tal mineralization photocatalyst, can be tamed by dispersing it in aninactive SiO2 matrix. By further research along the line, significantimprovement could be possible with careful catalyst design, optimi-zation of the reaction conditions, and reactor design.

Acknowledgments

We thank for M. Nomura of Photon Factory, KEK, Japan, forXANES analysis. This work was supported by the Hydrogen EnergyR&D Center, Korean Centre for Artificial Photosynthesis (NRF-2011-C1AAA0001-2011-0030278), and Basic Science Research Pro-gram (No. 2012-017247) funded by the Ministry of Education, Sci-ence, and Technology of Korea. It was also supported by the BrainKorea 21 and WCU (R31-30005) Programs.

Appendix A. Supplementary material

Supplementary data associated with this article can be found, inthe online version, at http://dx.doi.org/10.1016/j.jcat.2013.03.003.

References

[1] T. Huai, S.D. Shah, J. Wayne Miller, T. Younglove, D.J. Chernich, A. Ayala, Atmos.Environ. 40 (2006) 2333–2344.

[2] S.J. Schmieg, B.K. Cho, S.H. Oh, Appl. Catal., B 49 (2004) 113–125.[3] F. Willems, D. Foster, American Control Conference, 2009, ACC ‘09, 2009, pp.

3944–3949.[4] T. Johnson, Platinum Met. Rev. 52 (2008) 23–37.[5] M. Koebel, M. Elsener, M. Kleemann, Catal. Today 59 (2000) 335–345.[6] C. Ciardelli, I. Nova, E. Tronconi, D. Chatterjee, B. Bandl-Konrad, M. Weibel, B.

Krutzsch, Appl. Catal., B 70 (2007) 80–90.[7] A. Grossale, I. Nova, E. Tronconi, D. Chatterjee, M. Weibel, J. Catal. 256 (2008)

312–322.[8] K.I. Shimizu, A. Satsuma, T. Hattori, Appl. Catal., B 25 (2000) 239–247.[9] J. Szanyi, J.H. Kwak, R.A. Moline, C.H.F. Peden, J. Phys. Chem. B 108 (2004)

17050–17058.[10] J.H. Lee, S.J. Schmieg, S.H. Oh, Appl. Catal., A 342 (2008) 78–86.[11] L. Valanidou, C. Theologides, A.A. Zorpas, P.G. Savva, C.N. Costa, Appl. Catal., B

107 (2011) 164–176.[12 N. Popovych, P. Kirienko, S. Soloviev, S. Orlyk, Catal. Today. 191 (2012) 38–41.[13] S. Sitshebo, A. Tsolakis, K. Theinnoi, Int. J. Hydrogen Energy 34 (2009) 7842–

7850.[14] Y. Youn, J.W. Park, C. Kwon, J. Lee, G. Yeo, SAE [Tech. Pap.], 2006-01-1370,

2006.[15] E.R. Fanick, P.M. Merrit, SAE [Tech. Pap.], 2008-01-0067, 2008.[16] K.I. Shimizu, J. Shibata, H. Yoshida, A. Satsuma, T. Hattori, Appl. Catal., B 30

(2001) 151–162.[17] R. Brosius, K. Arve, M.H. Groothaert, J.A. Martens, J. Catal. 231 (2005) 344–353.[18] C. Minero, V. Maurino, E. Pelizzetti, Mar. Chem. 58 (1997) 361–372.

66 J.Y. Kim et al. / Journal of Catalysis 302 (2013) 58–66

[19] M. Sturini, F. Soana, A. Albini, Tetrahedron 58 (2002) 2943–2950.[20] W.B. Kim, H. Sun Choi, J. Lee Sung, J. Phys. Chem. B 104 (2000) 8670–8678.[21] J. Zhao, X. Yang, Build. Environ. 38 (2003) 645–654.[22] H. Kim, W. Choi, Appl. Catal., B 69 (2007) 127–132.[23] T. Kudo, Y. Kudo, A. Ruike, A. Hasegawa, M. Kitano, M. Anpo, Catal. Today 122

(2007) 14–19.[24] C. Giannotti, S. Le Greneur, O. Watts, Tetrahedron Lett. 24 (1983) 5071–5072.[25] J.M. Herrmann, C. Duchamp, M. Karkmaz, B.T. Hoai, H. Lachheb, E. Puzenat, C.

Guillard, J. Hazard. Mater. 146 (2007) 624–629.[26] Y. Hu, N. Wada, K. Tsujimaru, M. Anpo, Catal. Today 120 (2007) 139–144.[27] A. Fahmi, C. Minot, Surf. Sci. 304 (1994) 343–359.[28] C. Anderson, A.J. Bard, J. Phys. Chem. 99 (1995) 17963.[29] D.C.M. Dutoit, M. Schneider, A. Baiker, J. Catal. 153 (1995) 165–176.[30] S.M. Jung, O. Dupont, P. Grange, Appl. Catal., A 208 (2001) 393–401.[31] J. Ren, Z. Li, S. Liu, Y. Xing, K. Xie, Catal. Lett. 124 (2008) 185–194.[32] G. Calleja, D.P. Serrano, R. Sanz, P. Pizarro, Micropor. Mesopor. Mater. 111

(2008) 429–440.[33] S. Bordiga, S. Coluccia, C. Lamberti, L. Marchese, A. Zecchina, F. Boscherini, F.

Buffa, F. Genoni, G. Leofanti, G. Petrini, G. Vlaic, J. Phys. Chem. 98 (1994) 4125–4132.

[34] J.D. Grunwaldt, C. Beck, W. Stark, A. Hagenc, A. Baiker, Phys. Chem. Chem. Phys.4 (2002) 3514–3521.

[35] Z. Liu, R.J. Davis, J. Phys. Chem. 98 (1994) 1253–1261.[36] J.S. Lee, W.B. Kim, S.H. Choi, J. Synchrotron Radiat. 8 (2001) 163–167.[37] C. Kormann, D.W. Bahnemann, M.R. Hoffmann, J. Phys. Chem. 92 (1988) 5196–

5201.[38] L.E. Brus, J. Chem. Phys. 79 (1983) 5566–5571.[39] L. Brus, J. Phys. Chem. 90 (1986) 2555–2560.[40] H. Uchida, S. Hirao, T. Torimoto, T.S. Susumu Kuwabata, H. Mori, H. Yoneyama,

Langmuir 11 (1995) 3725–3729.[41] R. Castillo, B. Koch, P. Ruiz, B. Delmon, J. Catal. 161 (1996) 524–529.[42] E.Y. Kim, C.M. Whang, W.I. Lee, Y.H. Kim, J. Electroceram. 17 (2006) 899–902.[43] A. Hanprasopwattana, T. Rieker, A.G. Sault, A.K. Datye, Catal. Lett. 45 (1997)

165–175.[44] T.P. Ang, C.S. Toh, Y.F. Han, J. Phys. Chem. C 113 (2009) 10560–10567.[45] A.Y. Stakheev, E.S. Shpiro, J. Apijok, J. Phys. Chem. 97 (1993) 5668–5672.[46] S.M. Mukhopadhayay, S.H. Garofalini, J. Non-Cryst. Solids 126 (1990) 202–208.[47] X. Gao, I.E. Wachs, Catal. Today 51 (1999) 233–254.[48] C. Beck, T. Mallat, T. Bürgi, A. Baiker, J. Catal. 204 (2001) 428–439.[49] B. Ding, H. Kim, C. Kim, M. Khil, S. Park, Nanotechnology 14 (2003) 532–537.[50] A. Teleki, M.K. Akhtar, S.E. Pratsinis, J. Mater. Chem. 18 (2008) 3547–3555.[51] T. Nakayama, J. Electrochem. Soc. 141 (1994) 237–241.[52] G. Cernuto, S. Galli, F. Trudu, G.M. Colonna, N. Masciocchi, A. Cervellino, A.

Guagliardi, Angew. Chem., Int. Ed. 50 (2011) 10828–10833.