Fuel 89 (2010) 1221–1229

Contents lists available at ScienceDirect

Fuel

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

Combined pre-reforming–desulfurization of high-sulfur fuels for distributedhydrogen applications

Nazim Muradov a,*, Karthikeyan Ramasamy a, Clovis Linkous a, Cunping Huang a, Ibrahim Adebiyi a,Franklyn Smith a, Ali T-Raissi a, James Stevens b

a Florida Solar Energy Center, University of Central Florida, Cocoa, FL 32922, USAb Chevron Technology Ventures, LLC, Houston, TX 77042, USA

a r t i c l e i n f o

Article history:Received 15 June 2009Received in revised form 23 October 2009Accepted 28 October 2009Available online 14 November 2009

Keywords:HydrogenDiesel fuelReformingDesulfurizationCatalyst

0016-2361/$ - see front matter � 2009 Elsevier Ltd. Adoi:10.1016/j.fuel.2009.10.030

* Corresponding author. Fax: +1 321 638 1010.E-mail address: Muradov@fsec.ucf.edu (N. Murado

a b s t r a c t

A major challenge facing the future Hydrogen Economy is the issue of hydrogen fuel delivery and dis-tribution. In the near term, it may be necessary to deliver high-density hydrocarbon fuels (e.g., dieselfuel) directly to the end-user (e.g., a fueling station) wherein it is reformed to hydrogen, on demand.This approach has the advantages of utilizing the existing fuel delivery infrastructure, and the fact thatmore energy can be delivered per trip when the tanker is filled with diesel instead of liquefied orcompressed hydrogen gas. Reforming high-sulfur hydrocarbon fuels (e.g., diesel, JP-8, etc.) is particu-larly challenging due to rapid deactivation of conventional reforming catalysts by sulfurous com-pounds. A new on-demand hydrogen production technology for distributed hydrogen production isreported. In this process, first, the diesel fuel is catalytically pre-reformed to shorter chain hydrocar-bons (C1–C6) before being fed to the steam reformer, where it is converted to syngas and further tohigh-purity hydrogen gas. In the pre-reformer, most sulfurous species present in the fuel are con-verted to H2S. Desulfurization of the pre-reformate gas is carried out in a special regenerative redoxsystem, which includes an iron-based scrubber coupled with an electrolyzer. The integrated pre-refor-mer and sulfur-scrubbing unit operated successfully for 100 h at desulfurization efficiency of greaterthan 95%.

� 2009 Elsevier Ltd. All rights reserved.

1. Introduction

Considerable research and development efforts are underway,worldwide, to substitute hydrogen for gasoline and diesel fuel intransportation sectors. Due to the relatively low volumetric energydensity of hydrogen, its delivery to fueling stations and storagepresent major challenges facing the future Hydrogen Economy. Inthe near term, it may be economically more feasible to deliver highenergy density hydrocarbon fuels, such as diesel, gasoline, etc., di-rectly to the end-user wherein it is reformed to hydrogen, on de-mand. This approach has the following advantages:

(1) no change to the existing fuel delivery infrastructure isnecessary,

(2) more energy is delivered per trip when the tanker carriesdiesel instead of liquid or compressed hydrogen, and

(3) the fuel dispensing station would be able to service both inter-nal combustion and fuel cell powered vehicles at the same time.

ll rights reserved.

v).

Reforming of liquid hydrocarbon fuels to hydrogen has been afocus of intensive worldwide R&D efforts with a particular empha-sis on vehicular (on-board) applications (e.g., [1–4]). The objectiveof on-board reforming is to convert liquid fuels into hydrogen-richgas for use in an internal combustion engine (ICE) or a fuel cell (FC).In many respects, liquid hydrocarbon fuels represent a moreattractive means of carrying hydrogen than compressed H2 itself,promising greater vehicle range, shorter refueling times, increasedsafety, and perhaps most importantly, utilization of the currentfuel distribution infrastructure. The drawbacks of on-boardreformers include their inherent complexity, weight, high cost,the need for an elaborate purification of hydrogen from impurities(e.g., CO, H2S) that could degrade FC performance (although, H2

purity requirements for the ICE applications are much less strin-gent than those required for FC [5]).Three main fuel reformingstrategies for H2 production from hydrocarbons are: steam reform-ing (SR), partial oxidation (POx) and autothermal reforming (ATR).A generalized chemical reaction for ATR of hydrocarbon fuel(CnHm) is shown below:

CnHm þ xO2 þ ð2n� 2xÞH2O! nCO2 þ ð2n� 2xþm=2ÞH2 DH � 0 ð1Þ

1222 N. Muradov et al. / Fuel 89 (2010) 1221–1229

POx and ATR have a number of advantages over SR, namely,shorter startup time, better transient behavior, and less weight.They can reform a wide range of fuels, including gasoline anddiesel fuel. However, the POx and ATR reformers suffer from someshortcomings compared to SR: (i) the reformate gases from POxand ATR become diluted with N2 (since air is used in the process)resulting in lower FC performance compared to SR, (ii) H2 in FC an-ode exhaust is not easily integrated with the POx or ATR system,(iii) due to high exothermicity of the POx process, the reformermay be subject to greater thermal loses, etc.

Different fuels impose different constraints on the reformer de-sign, catalysts used and operating conditions. For example, it ismore difficult to reform diesel fuel than gasoline due to a numberof factors, e.g., the former has lower H/C ratio (thus, the potentialfor soot formation), higher energy consumption due to lower vola-tility of diesel fuel, higher temperatures in the reformer, higher sul-fur content, etc. The effect of fuel composition on the fuel processorperformance was investigated by Borup et al. [6]. It was demon-strated that short-chain aliphatic hydrocarbons tend to have favor-able reforming characteristics for catalytic ATR compared tolonger-chain and aromatic components. The Argonne National Lab-oratory researchers have investigated the reactor characteristicsand the efficiency of a catalytic autothermal reformer using surro-gates of diesel fuel (dodecane and hexadecane) as feedstock [7].The catalyst used was 1 wt.% Pt supported on cerium and gadolin-ium oxides. The reforming of these hydrocarbons was examined atthe range of oxygen-to-carbon ratios of 0.18–0.5 and steam-to-car-bon ratios of 1–3 and space velocities ranging from 10,000 to100,000 h�1 with H2 selectivity reaching up to 86%.

Reforminghigh-sulfur(orhigh-S)hydrocarbonfuels(e.g.,diesel,JP-8,etc.) isparticularlychallengingduetorapiddeactivationofconven-tionalreformingcatalystsbycokedepositsandsulfurouscompounds.Moreover,ifsulfurouscompounds(mostly,H2S)arenotremovedfromthereformategas,theycouldpoisonthecatalystsintheanodecompart-ment of FC substantially decreasing its performance or permanentlydamagingit.Thus,R&Deffortsinthisareahavebeenfocusedeitherondesulfurization of liquid hydrocarbon fuels before reforming or onthe development of sulfur-tolerant reforming catalysts. Fukunagaetal.havereportedanefficientNi-basedadsorbentthatloweredsulfurimpuritiesfromhigh-Skerosenedowntolessthan1 ppm[8].Ithasbeenshown in a number of publications that traditional Ni-based steamreforming catalysts rapidly deactivate during processing of high-S li-quid fuels. Although noble metal based catalysts (e.g., supported Ru,Pd,Rh,etc.)showedsomeimprovedtolerancetowardsulfurpoisoningandcokingcomparedtoNi-basedcatalysts,theirlongtermstabilitystillwas not acceptable. For example, the authors [9] reported that sup-ported Rh catalyst initially demonstrated a relatively good perfor-mance in steam reforming of high-S jet fuel, but over a longer periodofoperationshowedsignsofenhanceddeactivation.

Florida Solar Energy Center (FSEC), in collaboration with theChevron Technology Ventures (CTV), has developed a new on-de-mand hydrogen production technology for distributed applica-tions. The objective of this work is to catalytically convert high-Sliquid hydrocarbon fuels to an essentially sulfur-free gaseous feed-stock for steam reformation, while consuming only a fraction ofhydrogen product. The experimental results on the developmentand performance testing of the catalytic pre-reformer and desul-furization systems are presented in this paper.

2. Experimental

2.1. Reagents and catalysts

Hexadecane with greater than 98% purity was obtained fromFisher and used without further purification. Thiophene (99%)

was obtained from Alpha Aesar and used as received. Commercialdiesel fuel was purchased from a local Chevron gas station. Sincecommercial diesel fuel contains low levels of sulfur (approximately5.5 ppmw), the fuel was supplemented with thiophene (C4H8S) toelevate the sulfur content to 3180–5240 ppmw level. This thio-phene-spiked diesel was then used in all experiments as a surro-gate high-S fuel. The same was carried out with hexadecane.

Alumina-supported Ni–Mo hydroprocessing catalyst in theform of 1 mm by 5 mm extruded pellets was provided by HaldorTopsoe. The composition of the Ni–Mo catalyst was (%w/w): NiO(2–5), MoO3 (12–18), Al2O3 (68–80) and AlPO4 (5–11). In order toproduce catalytically active form of the Ni–Mo catalyst it was sul-fided using 6% dimethyl disulfide, (CH3)2S2, (DMDS) in n-heptaneas a sulfiding agent according to the following procedure. TheNi–Mo catalyst (115 g) was placed inside a stainless steel reactor(2.5 cm diameter). H2 was introduced at pressure of 6.7 atm andtemperature of 235 �C for 1 h. Then, 6% DMDS in heptane waspumped to the reactor at a flowrate of 25 mL/h for 0.5 h. The cata-lyst bed temperature was increased to 235 �C and maintained for1 h at that temperature. H2 pressure was increased to 54.5 atmand catalyst remained under that pressure for 1 h. The catalystbed temperature was then increased to 340 �C and held there foradditional 1 h before curtailing the pumping of the DMDS solutioninto the reactor. The catalyst bed temperature was decreased to175 �C, and H2 flow was maintained overnight at 15 L/h and54.5 atm pressure. A proprietary zeolite catalyst in the form of1.5 by 5 mm extruded pellets was supplied by CTV.

2.2. Analysis

Analysis of the products of pre-reforming was performed bymeans of gas chromatography (GC) as follows: the permanentgases, such as H2, CO, CO2 and CH4 were analyzed using an SRI-8610A GC (thermal conductivity detector, argon carrier gas, silicagel packed column); gaseous and liquid hydrocarbons were ana-lyzed using a Shimadzu GC-14B (flame ionization detector, heliumcarrier gas, capillary column). Analysis of sulfurous compounds(H2S, thiophene, methyl-mercaptan) was conducted using PerkinElmer GC (flame photometric detector, capillary column). Addi-tionally, H2S content in the pre-reformate gases was determinedwith the use of a set of Sensidyne� and RAE Systems gas-detectiontubes in the range of 0.2–5, 1–60 and 25–250 ppmv H2S.

X-ray diffraction (XRD) analysis of the Ni–Mo/alumina catalystsamples before and after the pre-reforming experiments was con-ducted using a Rigaku D-MaxB diffractometer. X-ray photoelectronspectroscopic (XPS) analysis of the catalysts samples was con-ducted using a Physical Electronics 5400 XPS instrument applyingnon-monochromatic Mg Ka X-rays for excitation (pass energy forsurvey 44.75 eV, step size 0.5 eV; for high resolution spectra passenergy 35.75 eV, step size 0.1 eV). The surface morphology of thecatalysts before and after pre-reforming were examined by scan-ning electron microscopy (SEM) (JEOL 6400F).

3. Results and discussion

3.1. Description of the concept

Fig. 1 depicts a simplified block-diagram of the process for con-version of a high-S fuel to hydrogen. A high-S liquid hydrocarbonfuel (e.g., diesel fuel) is first catalytically processed in a pre-refor-mer 1 to shorter chain hydrocarbons (C1–C6, preferably, C1–C4) inthe presence of hydrogen (this process is similar to a hydrocrackingprocess, therefore, hereafter we will also refer to this process ‘‘hy-dro-reforming”). Assuming (for the sake of simplification) the die-sel fuel empirical formula as C12H23 and propane as the main

1

5

2 3

Desulfurization unit Diesel fuel

Sulfur

Steam

H2

H2, C1-C6, H2S

H2,C1-C6

Fe2+

Fe3+

H2 4

Reforming unit

Fig. 1. Schematic of the concept for conversion of high-sulfur diesel fuel tohydrogen. (1) Pre-reforming reactor, (2) H2S scrubber, (3) electrolyzer, (4) steamreforming reactor, (5) gas conditioning/separation unit.

N. Muradov et al. / Fuel 89 (2010) 1221–1229 1223

product of the reaction, then, the diesel hydro-reforming processcan be described as follows:

C12H23 þ 4:5H2 ���!catalyst

4C3H8 ð2Þ

In the pre-reformer, most sulfurous species present in the fuel arecatalytically converted to hydrogen sulfide (H2S):

RSþ 2H2 ���!catalyst

H2Sþ RH2 ð3Þ

where R refers to an organic moiety.The sulfur removal from the pre-reformate gas is required in or-

der to avoid deactivation of the Ni-based catalyst in the steam re-former 4. Desulfurization of the pre-reformate gas is carried out ina special regenerative redox system, which includes a Fe-basedscrubber 2 coupled with an electrolyzer 3. In particular, ferrous/ferric (Fe2+/Fe3+) redox couple is used for oxidizing H2S to elemen-tal sulfur as follows:

2Fe3þðaqÞ þH2SðgÞ ! 2Fe2þ

ðaqÞ þ SðsÞ þ 2HþðaqÞ ð4Þ

The resulting ferrous ion is electrochemically oxidized back to theferric state, releasing H2, in a closed loop process:

2Fe2þðaqÞ þ 2HþðaqÞ ! 2Fe3þ

ðaqÞ þH2ðgÞ ð5Þ

The overall reaction is decomposition of H2S to H2 and elementalsulfur:

H2SðgÞ ! H2ðgÞ þ SðsÞ ð6Þ

Desulfurized pre-reformate gas is fed to the main reformer 4,where it is mixed with steam and processed over a Ni-catalyst tosyngas (again, assuming propane as the main product of pre-reforming):

C3H8 þ 3H2O! 3COþ 7H2 ð7Þ

The syngas is further conditioned (or shifted) and purified (viapressure swing adsorption process) to high-purity H2 via conven-tional steam methane (or naphtha) reforming processes.

A fraction of the output hydrogen product is recycled to the pre-reformer and used to conduct hydro-reforming of the high-S feed-stock. The required amount of recycled hydrogen depends on thecomposition of the feedstock and hydro-reforming products. Forexample, assuming an empirical formula of C12H23 for diesel, thefraction of hydrogen product needed in the pre-reforming processfor generating propane and butane would be about 16% and 13%,respectively.

3.2. Pre-reforming of high-S hexadecane and diesel fuel

Figs. 2 and 3 show, respectively, a schematic diagram and aphotograph of the bench-scale experimental unit employed forpre-reforming and desulfurization of high-S hydrocarbon fuels. Inthis section, the results of catalytic pre-reforming experimentsusing hexadecane and commercial diesel fuel spiked with thio-phene in the amount corresponding to 3180–5240 ppmw sulfurare described (the data on the desulfurization sub-unit will pre-sented in the next section).

The pre-reformer sub-unit consists of four zones: (1) a fuelmetering and delivery zone, (2) a reaction zone, (3) a products sep-aration zone, and (4) a products metering and analysis zone. In thefuel delivery zone, diesel (or hexadecane) is pumped to the reactorat 13.6–14.3 atm pressure using high performance liquid chroma-tography pump (Lab Alliance) at a flowrate of 0.2–0.4 mL/min.Hydrogen flow is controlled by a high-pressure mass flow control-ler (Parker) and varied in the range of 0.72–3.50 L/min. The pre-reforming reactor (ID = 25 mm, wall thickness 2 mm, length45 cm) was fabricated out of 316 stainless steel with a maximumpressure rated at 200 atm at 540 �C. Three quarters of the reactorwas filled with catalyst. The hydrogen entered at the bottom ofthe reactor and the diesel vapor flowed in at the top of the catalystlayer within the reactor. The reactor had an adjustable catalystsupport system. The adjustable catalyst support system providedflexibility in varying the length of the catalyst zone, which allowedcapability to vary reactants residence time in the reactor. Fluidpressures and temperatures were monitored, in real time, at the in-let as well as at the outlet of the reactor using a LabView dataacquisition system. The temperature in the upper section of thereactor maintained at 400–450 �C and in the lower part (wherediesel entered the reaction zone) at 450–500 �C. In the productsseparation zone, the product mixture passed through a shell-and-tube type condenser (kept at a temperature around 15–20 �C)where most of C6 + hydrocarbons condensed and refluxed back tothe reactor and the gaseous hydrocarbons (predominantly, C1–C4), H2S and excess H2 entered a knock-out vessel where the es-caped high hydrocarbons from the condenser were trapped andcollected. The effluent gas flowed through in-line filters that re-moved any particulates or aerosols that may have been producedin the reaction zone, passing through a back-pressure regulatorto reduce pressure to near atmospheric. The pre-reformate thenentered into a metering and analysis zone wherein most of the gas-eous product passed through a gas meter (Shinagawa W-NK-2),that metered the volume of the gas generated. For safety reasons,several pressure relief valves were installed, and corrosion resis-tant stainless steel 316 tubing and fittings were used throughoutthe system.

Three series of experiments with different catalytic systemswere conducted including: (i) zeolite catalyst, (ii) Ni–Mo/aluminacatalyst and (iii) the mixture of zeolite and Ni–Mo/alumina cata-lysts. The rational for selecting these particular catalysts is as fol-lows. In order to convert long-chain hydrocarbons and sulfur-organic compounds present in diesel fuel to short-chain hydrocar-bons (e.g., C1–C6) and H2S, catalysts with hydrocracking andhydrotreating activities are needed. Zeolite and Ni–Mo/aluminabased catalysts are widely being utilized in oil refining and chem-ical industries for these particular applications. The compositionof the effluent gas from hexadecane hydro-reforming over zeolitecatalyst at 450 �C and 14 atm is depicted in Fig. 4 (left). It can beseen that C1–C7 saturated hydrocarbons (paraffins) were pro-duced in the reaction, with propane being a main component inthe mixture followed by butanes. The stability of the zeolite cat-alyst, however, was not adequate, and after about 20–30 h ofoperation there were signs of deactivation manifested by theappearance of a liquid product in the knock-out vessel. Indeed,

Vent Line

Gas out

Liquid in

To analysis

To desulfurizationunit

4

8

19

20 6

7

10 9

5

3

2

1

18

17

12

11 13

14

16

15

Fig. 2. Schematic of the bench-scale experimental unit for high-sulfur diesel pre-reforming. (1) Diesel fuel storage, (2) nitrogen tank, (3) hydrogen tank, (4) HPLC pump, (5)mass flow controller, (6) two-way valve, (7) in-line filter, (8) check valve, (9) pressure relief valve, (10) pressure gauge, (11) three zone furnace, (12) reactor, (13) hydrogeninlet, (14) diesel inlet, (15) thermocouple, (16) catalyst, (17) condenser, (18) knock-out vessel, (19) back-pressure regulator, (20) gas meter.

Fig. 3. Photo of the combined diesel (hexadecane) pre-reforming–desulfurization experimental unit.

1224 N. Muradov et al. / Fuel 89 (2010) 1221–1229

zeolite catalyst dislodged from the reactor after the experimenthad turned black, apparently due to coke deposition. The amountof coke accumulated on the catalyst surface was measured to beabout 10 wt.% of the original catalyst weight. The coked catalystcould be easily reactivated by burning coke off the catalyst sur-face by air at 450 �C for 4 h.

Fig. 4 (right) shows the product distribution of hexadecane hy-dro-reforming over Ni–Mo/alumina catalyst at conditions similarto that of the zeolite catalyst. In contrast to the zeolite-producedgas, the product gas from hexadecane hydro-reforming was richin higher hydrocarbons, in particular, C4–C7 paraffins, with butanesbeing the main products, followed by heptanes. Similar to the zeo-lite, the Ni–Mo catalyst activity was not satisfactory, deactivatingmuch faster than the zeolite catalyst. After about 5 h of run time,liquid product appeared in the condenser, indicating the loss of hy-dro-reforming activity of the Ni–Mo/alumina catalyst. At the end ofthe experiment, the catalyst color had turned from light green toblack.

In the following series of experiments, the mixture of zeolitecatalyst and Ni–Mo/alumina catalyst in the weight ratio of 2:1was utilized. The experimental conditions were as follows: tem-peratures in the upper and lower parts of the reactor were 400and 470 �C, respectively, pressure of 13.7 atm, thiophene-addedhexadecane feed flow 0.3 mL/min. The results revealed that thecomposition of the pre-reformate gas changed with time duringfirst 25–30 h of the experimental run, reaching a steady state afterabout 50 h. Fig. 5a and b illustrates the results of a typical experi-ment presenting the data in two forms: (a) the time dependence ofthe molar fraction of the individual (or the family of) hydrocarbons,and (b) the composition of the product gas after 4, 20, 32 and 56 hof operation. The figure shows that at the beginning of the process(while the catalyst is fresh), methane is by far the main componentof the pre-reformate gas, followed by ethane, propane and butanes.There are practically no higher hydrocarbons in the effluent gas. Asthe experiment progresses, the concentration of methane gradu-ally decreases and the concentrations of higher hydrocarbons

Mol

ar fr

actio

n

0.0

0.1

0.2

0.3

0.4

CH4C2H6 C3H8

ΣC4ΣC5ΣC6ΣC7ΣC8+

Zeolite Ni-Mo/Al2O3

CH4CH4ΣC8+

ΣC8+

Fig. 4. Hexadecane hydro-reforming products distribution using zeolite (left) andNi–Mo/alumina (right) catalysts. T = 450 �C, P = 14 atm.

A B C D

Mol

ar fr

actio

n

0.0

0.2

0.4

0.6

0.8

1.0CH4C2H6C3H8

ΣC4ΣC5ΣC6+

Carbon number1 2 3 4 5 6

Mol

ar fr

actio

n

0.0

0.2

0.4

0.6

0.8

1.0

after 4 hrsafter 20 hrsafter 32 hrsafter 56 hrs

a

b

Fig. 5. Hexadecane hydro-reforming over zeolite/Ni–Mo catalyst. T = 400–470 �C,P = 13.7 atm. Hexadecane and thiophene feeding rates are 0.3 mL/min and 1.21 mg/min, respectively. The gas samples were analyzed after: (A) 4, (B) 20, (C) 32 and (D)56 h.

Mol

ar fr

actio

n

0.0

0.1

0.2

0.3

0.4

CH4C2H6 C3H8ΣC4ΣC5ΣC6ΣC7ΣC8+

Zeolite/Ni-Mo (450oC) Zeolite/Ni-Mo (500oC)

CH4

CH4

ΣC8+ ΣC8+

Fig. 6. Diesel hydro-reforming over zeolite/Ni–Mo catalyst at 450 �C (left) and500 �C (right). P = 13.7 atm. Diesel and thiophene feeding rates are 0.3 mL/min and0.81 mg/min, respectively.

N. Muradov et al. / Fuel 89 (2010) 1221–1229 1225

increase. After about 50 h a steady state regime is reached charac-terized by production of the pre-reformate gas consisting of C1–C6

hydrocarbons with propane being the main component of theproduct gas. Overall, the experiment was run for 80 h, and only asmall amount of liquid hydrocarbon was collected from theknock-out vessel, which corresponded to 98% conversion of hexa-decane to gaseous products.

A similar experiment was conducted by passing the thiophene-containing diesel fuel feedstock over mixed zeolite and Ni–Mo/alu-mina catalysts. The reaction was operated at two temperature re-gimes: 450 and 500 �C (temperature measured in the middlesection of the reactor) with other parameters remaining the sameas in hexadecane experiment. As in the case with hexadecane, ini-tially, the pre-reformate gas was rich with methane; however, afterabout 12–16 h the steady state regime was achieved which wasmaintained for the duration of the experiment (100 h). Averageconversion of diesel to gaseous hydrocarbon products during theexperiment was 95–97%. Fig. 6 depicts the distribution of thepre-reformate products at 450 (left) and 500 �C (right) after reach-ing a steady state regime. It is noteworthy that the distribution ofC1–C6 products of diesel and hexadecane hydro-reforming over thesame catalytic system and at the same temperature range showedsome similarities: in both systems, propane was the main productfollowed by butanes. In contrast to hexadecane, the diesel-derivedgas contained a slightly higher content of heavier hydrocarbons(C6–C8+). It is evident from Fig. 6 that increase in temperature re-sulted in an increase in methane concentration in the pre-refor-mate gas.

A relatively stable operation of a binary mixture of catalystswith different types of catalytically active sites demonstrated inthe above experiments is an interesting observation, which is yetto be fully understood. A similar observation was made by Chevronresearchers [10], who pointed out that a physically intermixed cat-alyst system comprising hydrodesulfurization and hydrocrackingcatalyst particles of the same size demonstrated a surprisinglygood stability against fouling when used in combined hydrotreat-ing and hydrocracking applications. In agreement with this obser-vation, in our binary zeolite + Ni–Mo/alumina catalytic system, theboth catalysts particles are substantially of the same size, whichmay explain their improved stability against coking.

3.3. Desulfurization of pre-reformate gas

The analysis of the pre-reformate gas produced by the zeolitecatalyst revealed that the thiophene-to-H2S conversion yield wasvery low. On the other hand, the combined zeolite/Ni–Mo catalystalmost quantitatively hydrogenated thiophene to H2S using thio-phene-laden hexadecane and diesel feedstocks. The latter pre-reformate gas, containing H2, H2S and light hydrocarbons (C1–C6),was directed to the desulfurization sub-unit.

1226 N. Muradov et al. / Fuel 89 (2010) 1221–1229

Fig. 7A provides a schematic diagram of the desulfurization sub-unit for the continuous removal of hydrogen sulfide from the pre-reformate gas and its splitting to hydrogen and sulfur. A gaseousmixture from the pre-reformer enters into a scrubbing unit, whichincludes an absorption column filled with aqueous ferric sulfatesolution, Fe2(SO4)3(aq) (total volume of solution 2.3 L). H2S whendissolved in the aqueous solution is oxidized by ferric ion (Fe3+)to elemental sulfur, and simultaneously ferric sulfate is reducedinto ferrous sulfate, FeSO4 (see Eq. (4)). (Note that the kinetics ofthe Fe3+ reaction with saturated hydrocarbons and hydrogen pres-ent in the pre-reformate gas is negligible, whereas the reactionwith sulfide ion is almost diffusion controlled.) Elemental sulfuris removed from the system by a filter, and the remaining ferroussulfate solution is fed to an electrolyzer, where ferrous sulfate isoxidized back to ferric sulfate, and protons are reduced to hydro-gen gas at the cathode. The regenerated ferric sulfate solution isthen fed back to the absorption column of the scrubber for scrub-bing hydrogen sulfide, forming a closed cycle with the net reactionbeing hydrogen sulfide decomposition to elemental sulfur andhydrogen gas.

Electrolysis of acidic FeSO4 aqueous solutions was carried outusing a modified proton exchange membrane (PEM) electrolyzeras shown in Fig. 7B. Platinum catalyst was spray-deposited ontothe cathode side of a Nafion� film to form a membrane electrodeassembly (MEA). The cathode section consisted of a stainless steel

H2

Iron (III) sulfate

Sulfur

To analysis

Iron (II) sulfatePre-reformate gas

6

5

3

4

2

1

Desulfurized pre-reformate gas

A

B

Fig. 7. Schematic of desulfurization unit (A) and electrolyzer (B). (1) Hydrogensulfide scrubber, (2) electrolyzer, (3) heating coil, (4) filter, (5) peristaltic pump, (6)three-way valve.

plate used as current collector in a contact with water for hydrogenevolution. This configuration eliminated the need for a carrier gasto sweep hydrogen from the cathode side of the electrolyzer. Anelectrolyzer potential of 0.80–1.03 V was necessary for the electro-chemical process to regenerate the scrubber solution at a rate suf-ficient to match the H2S flow rate into the scrubber. Theelectrolytic system can be operated at the range of temperaturesfrom ambient to 90 �C. In our experiments, desulfurization of thepre-reformate gas occurred optimally at the following process con-ditions: iron sulfate (total) concentration of 0.1 M, pH of 1.7–1.8(adjusted by addition of 5 N sulfuric acid), and electrolyte temper-ature of 50 �C.

It should be noted that no Pt catalyst is needed for the oxidationof ferrous ions in the anodic section of the electrolyzer. Experimen-tal results depicted in Fig. 8 imply that oxidation of ferrous to ferricions is not affected by the lack of Pt catalyst at the anode. A plaincarbon cloth can be used at the anode to allow distribution of bothcurrent and electrolyte. The experimental results demonstratedthat both H2SO4 and FeSO4 concentrations have a significant effecton the electrolytic hydrogen production rate. In particular, it wasfound that while the hydrogen evolution rate increased almost lin-early with the increase in H2SO4 concentration (within the range of[H2SO4] = 0.25–0.75 N), the same dependence on the FeSO4 con-centration goes through a maximum with respect to hydrogen pro-duction rate corresponding to about 0.15 M (Fig. 9). This wasattributed to the generation of polysulfate species at higher con-centrations that were less active toward charge transfer at the car-bon anode surface.

During initial testing of the combined scrubber–electrolyzersystem, the bench-scale desulfurization unit was continuouslyoperated for more than 300 h using a model H2S–N2 mixture with2500 ppmv H2S at the inlet and practically no H2S at the outlet ofthe scrubber (which corresponded to several turnovers of the Fe2+/Fe3+-redox system). In the diesel pre-reforming experiment usingthe zeolite/Ni–Mo catalytic system, diesel with 3180 ppmw of thi-ophene produced the pre-reformate gas with H2S content of185 ppmv (determined by GC-PFD method and confirmed by gas-detection tubes). The chromatogram showed the presence of H2Sas the only product of hydrogenation of thiophene; no unreactedthiophene or other sulfurous compounds were detected. On aver-age, about 3 ppmv of H2S was detected exiting the ferric sulfatescrubber during the experiment. A similar experiment using thezeolite–Ni–Mo mixed catalyst and hexadecane containing5240 ppmw of thiophene produced largely desulfurized pre-refor-

Time, min0 20 40 60 80 100

Hyd

roge

n vo

lum

e, m

L

0

100

200

300

400

500

600

one-side Pt-loaded membranetwo-sides Pt-loaded membrane

Fig. 8. Electrolysis of acidic FeSO4 solution using one-side and two-sides Pt loadedMEA. Pt loading: 1.8 mg/cm2, current density: 30–50 mA/cm2, electrolyte: 0.5 NH2SO4 + 0.18 M FeSO4, potential: E = 0.95 V.

Binding energy, eV0100200300400500600700

N (E

)

0

1000

2000

3000

4000

5000

O1s

Mo3p C1s Mo3dAl2s Al2p

Fig. 11. XPS spectrum of Ni–Mo/alumina catalyst.

[FeSO4], mol/L0.0 0.1 0.2 0.3 0.4 0.5 0.6

Hyd

roge

n ev

olut

ion

rate

, mL/

min

0

1

2

3

4

Fig. 9. Hydrogen evolution rate as a function of FeSO4 concentration. One-side Pt-loaded MEA, Pt loading: 1.8 mg/cm2, 0.325 N H2SO4, E = 0.95 V.

N. Muradov et al. / Fuel 89 (2010) 1221–1229 1227

mate gas with H2S content of less than 5 ppmv. The sulfur balanceof the pre-reforming reaction and the degree of the feedstockdesulfurization was determined by comparing the amount of sul-fur in feedstock (in moles) entering the reactor with the amount

Fig. 10. SEM images of Ni–Mo/alumina catalyst before (A) and after (B) high-sulfurdiesel hydro-reforming. Diesel hydro-reforming temperature: 450 �C.

of H2S (in moles) detected in the effluent gas after the scrubber.The sulfur balance for the above experiments was closed withinthe margin of error of 6–8% using two independent analyticalmethods for the gas analysis (GC and gas-detection tubes). The fueldesulfurization yield was defined as follows:

g ¼ ðC4H8SÞin � ðH2SÞout

ðC4H8SÞin� 100%

where g is fuel desulfurization yield, (C4H8S)in is number of thio-phene moles in the fuel entering the pre-reformer, (H2S)out is thenumber of H2S moles exiting the sulfur scrubber.

Binding energy, eV224226228230232234236238240

Inte

nsity

Mo 3d5/2

Mo 3d3/2

Mo4+

B

A

Fig. 12. High resolution XPS spectra of Ni–Mo/alumina catalyst before (A) and after(B) high-sulfur diesel hydro-reforming (224–240 eV range).

y 200

250

300

1228 N. Muradov et al. / Fuel 89 (2010) 1221–1229

Based on the collected data from high-S diesel and hexadecanepre-reforming–desulfurization experiments, the fuel desulfuriza-tion yields were determined to be in the range of 96–98% (aver-aged over a multi-day operation). The value of the parasitic loadrelated to the electrolyzer operation (i.e., percentage of the heatingvalue of diesel used to operate the electrolyzer) was estimated at0.3% (LHV basis).

Degree 2θ10 20 30 40 50 60 70 80

Inte

nsit

50

100

150

Fig. 14. XRD pattern of Ni–Mo/alumina catalyst after high-sulfur diesel hydro-reforming.

3.4. Characterization of pre-reforming catalyst

Characterization of the zeolite/Ni–Mo catalyst before andafter pre-reforming of high-S diesel was carried out by SEM,XPS and XRD analyses. Fig. 10 shows SEM images of the Ni–Mo/alumina catalyst before (A) and after (B) hydro-reformingof high-S diesel (at 450 �C and time on stream of 100 h). It canbe seen that no significant changes in the catalyst surface mor-phology occurred during the hydro-reforming reaction. The XPSspectrum of the original Ni–Mo/alumina pre-reforming catalystin the region of 0–700 eV is depicted in Fig. 11. The effect ofoperating conditions during the high-S fuel pre-reforming onthe oxidation state of Mo and Ni was investigated. Fig. 12 com-pares the XPS spectra of the oxidized (original) (A) and reduced(after exposure to H2 and high-S diesel fuel) (B) forms of Mo inthe Ni–Mo catalyst. The XPS spectrum of the Mo3d5/2 andMo3d3/2 transitions for the oxidized form of catalyst (Fig. 12A)shows only Mo6+ species, corresponding to binding energies of233 and 236 eV, respectively. After exposure of the catalyst toreducing and sulfiding environment, the reduced Mo4+ speciesappeared in the XPS spectrum with Mo3d5/2 binding energy ofabout 229 eV (Fig. 12B). The peaks at 229–229.6 eV in the sulfid-

Binding energy, eV850852854856858860862864866

Inte

nsity

A

B

Ni 2p3/2

Fig. 13. High resolution XPS spectra of Ni–Mo/alumina catalyst before (A) and after(B) high-sulfur diesel hydro-reforming (850–866 eV range).

ed catalyst indicate the presence of MoS2. The spectrum showsthat after hydro-reforming, a significant amount of Mo in thecatalyst remains in its oxidized form, which may point to astrong interaction of oxidized Mo species with the aluminasupport.

Fig. 13A and B shows the XPS spectra of the oxidized and re-duced/sulfided form of the pre-reforming catalyst, respectively,in the region of 850–866 eV. Interpretation of the Ni2p region ofthe XPS spectra as it relates to oxidized and reduced forms of theNi–Mo/alumina catalyst poses some challenges, because of prox-imity of Ni2p binding energies for different Ni species [11]. Indeed,it is very difficult to distinguish between different Ni species (e.g.,NiO, Ni2O3, NiAl2O4) in the XPS spectrum, as all may appear in thesame region. The weak peak at the binding energy of 854 eV in thereduced/sulfided form of the catalyst may be attributed to NiS. Theabsence of a peak at a binding energy of 852.4 eV indicates thatnickel was not reduced to its metallic state, and most likely, re-mained in its Ni2+ oxidation state.

The XRD pattern of the Ni–Mo/alumina catalyst after high-Sdiesel pre-reforming is shown in Fig. 14. Two most prominent dif-fraction peaks appearing at 2h values of 46� and 67� can be as-signed to c-Al2O3. The XRD pattern does not indicate XRD-observable Ni oxides; however, it is commonly observed that NiOand Ni2O3 may also exist in amorphous or microcrystalline phases[12]. Identification of a NiAl2O4 phase is rather difficult because ofproximity of its reflection to that of alumina. Peaks correspondingto Mo compounds are apparently too weak to be distinguishable onthe XRD pattern of the catalyst.

4. Conclusions

A novel process for converting high-sulfur diesel to hydrogenthat employs a catalytic pre-reformer coupled with an efficient sul-fur-scrubbing unit suitable for distributed hydrogen productionand dispensing applications has been developed. A robust bi-func-tional catalytic system for pre-reforming high-sulfur fuels (sulfurcontent up to 5240 ppm) to short-chain hydrocarbons (predomi-nantly, propane) at an average yield of 97% has been developedand demonstrated. A regenerable Fe2+/Fe3+-redox/electrolyzer sys-tem capable of scrubbing H2S from the pre-reformate gas and itssplitting to H2 and elemental sulfur has also been developed andtested for 300 h of continuous operation. Electrolysis of acidic

N. Muradov et al. / Fuel 89 (2010) 1221–1229 1229

FeSO4 aqueous solution was shown to be highly efficient with acolumbic efficiency approaching 100% at applied voltage of 1.0 Vor lower. The effect of reaction conditions, such as pH, FeSO4 con-centration, and temperature on the desulfurization efficiency hasbeen determined. It was shown that the electrolytic process canbe conducted with a Pt-free anode capable of oxidizing ferrous toferric ions, thereby reducing the cost of the electrolytic system.The integrated pre-reformer and sulfur-scrubbing unit operatedsuccessfully for 100 h, while achieving the desulfurization efficien-cies of 95% or greater for removing sulfur (i.e., [H2S] < 5 ppmv inthe pre-reformate gas).

Acknowledgements

Financial support for this work was provided by the FloridaHydrogen Initiative (FHI) and Chevron Technology Ventures(CTV). Authors thank Haldor Topsoe and CTV for providing thesamples of catalysts used in the experiments, and Kirk Scammon(UCF Materials Characterization Facility) for conducting analysesof the catalysts. Authors also wish to acknowledge contributionsof Ms. Pam Portwood (FHI), Stephen Adams (FHI) and Dr. DavidBlock (FSEC) toward completion of this work.

References

[1] Thomas C, James B, Lomax F, Kuhn I. Fuel options for the fuel cell vehicle:hydrogen, methanol or gasoline. Int J Hydrogen Energy 2000;25:551–67.

[2] Avchi A, Onsan I, Trimm D. On-board fuel conversion for hydrogen fuel cells:comparison of different fuels by computer simulations. Appl Catal A: General2001;216:243–56.

[3] Petterson L, Westerholm R. State of the art of multi-fuel reformers for fuel cellvehicles: problem identification and research needs. Int J Hydrogen Energy2001;26:243–64.

[4] Sopena D, Melgar A, Briceno Y, Navarro R, Alvarez-Galvan M, Roza F. Diesel fuelprocessor for hydrogen production for 5 kW fuel cell application. Int JHydrogen Energy 2007;32:1429–36.

[5] Jamal Y, Wyszynski M. On-board generation of hydrogen-rich gaseous fuels. Areview. Int J Hydrogen Energy 1994;19:557–72.

[6] Borup R, Inbody M, Semelsberger T, Tafoya J, Guidry D. Fuel compositioneffects on transportation fuel cell reforming. Catal Today 2005;99:263–70.

[7] Liu D, Kaun T, Liao H, Shabir A. Characterization of kilowatt-scale autothermalreformer for production of hydrogen from heavy hydrocarbons. Int J HydrogenEnergy 2004;29:1035–46.

[8] Fukunaga T, Katsuno H, Matsumoto H, Takahashi O, Akai Y. Development ofkerosene fuel processing system for PEFC. Catal Today 2003;84:197–200.

[9] McCoy A, Duran M, Azad A, Chattopadhyay S, Abraham M. Performance ofsulfur tolerant reforming catalysts for production of hydrogen from jet fuelstimulants. Energy Fuels 2007;21:3513–9.

[10] Habib M, Winslow P, Moore R. Catalyst system for combined hydrotreatingand hydrocracking and a process for upgrading hydrocarbonaceous feedstocks.US Patent No. 5,439,860, August 8; 1995.

[11] Wang X, Ozkan U. Characterization of active sites over reduced Ni–Mo/Al2O3

catalysts for hydrogenation of linear aldehydes. J Phys Chem B2005;109:1882–90.

[12] Liu F, Xu S, Cao L, Chi Y, Zhang T, Xue D. Comparison of NiMo/Al2O3 catalystsprepared by impregnation and coprecipitation methods forhydrodesulfurization of dibenzothiophene. J Phys Chem C2007;111:7396–402.