Fuel 89 (2010) 2536–2543

Contents lists available at ScienceDirect

Fuel

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

Effects of the operating variables on hydrotreating of heavy gas oil:Experimental, modeling, and kinetic studies

M. Mapiour a, V. Sundaramurthy a,1, A.K. Dalai a,*, J. Adjaye b

a Catalysis and Chemical Reaction Engineering Laboratories, Department of Chemical Engineering, University of Saskatchewan, Saskatoon, SK, Canada S7N 5A9b Syncrude Edmonton Research Centre, Edmonton, AB, Canada T6N 1H4

a r t i c l e i n f o

Article history:Received 7 July 2009Received in revised form 19 February 2010Accepted 23 February 2010Available online 5 March 2010

Keywords:H2 partial pressureHydrotreatmentH2 purityHeavy gas oil

0016-2361/$ - see front matter Crown Copyright � 2doi:10.1016/j.fuel.2010.02.024

* Corresponding author. Tel.: +1 306 966 4771; faxE-mail addresses: mlm499@mail.usask.ca (M. M

sk.ca (V. Sundaramurthy), ajay.dalai@usask.ca (A.K. D1 Present address: Saskatchewan Research Council, S

a b s t r a c t

The effects of H2 purity, pressure, gas/oil ratio, temperature, and LHSV on hydrotreating activities wereinvestigated in a micro-trickle bed reactor using a commercial NiMo/c-Al2O3 catalyst. Heavy gas oil(HGO) from Athabasca bitumen was used as feed. Due to their significant effects on H2 partial pressure,H2 purity, pressure, and gas/oil ratio were chosen and used in a central composite design (CCD) method.Experimental conditions used were H2 purity, pressure, and gas/oil ratio were: 75–100 vol.% (with therest methane), 7–11 MPa, and 400–1200 mL/mL, respectively. The effect of LHSV (0.65–2 h�1) and tem-perature (360–400 �C) were studied in a separate set of experiments. Vapor/liquid equilibrium (VLE) cal-culations were performed to determine the inlet and outlet H2 partial pressures. It was observed that theenhancing effects of H2 purity on hydrodenitrogenation (HDN) and hydrodearomatization (HDA) activi-ties were greater than that of gas/oil ratio; however, it was comparable to pressure. Hydrodesulphuriza-tion (HDS) activity was not considerably affected by H2 purity, pressure, or gas/oil ratio. Increasing LHSVled to a decrease in HDS, HDN, and HDA activities while increasing temperature resulted in an increase inHDS and HDN; HDA had maximum activity at about 385 �C. Kinetic fitting of the data to a pseudo-first-order power law model suggested that conclusions on hydrotreating activities’ responses to a changingH2 pressure could be equally drawn from either inlet or outlet H2 partial pressure. However, from the cat-alyst deactivation standpoint, it is recommended that such conclusions are drawn from the outlet H2 par-tial pressure.

Crown Copyright � 2010 Published by Elsevier Ltd. All rights reserved.

1. Introduction

Hydrotreating is a catalytic process in which oil feedstock is re-acted with hydrogen to remove contaminants such as sulfur (S)and nitrogen (N) and saturate aromatics and olefins [1]. The pro-cess of removal of S is known as hydrodesulfurization (HDS), theremoval of N as hydrodenitrogenation (HDN), and the reductionof aromatics as hydrodearomatization (HDA). If these contami-nants are not removed, they may have detrimental environmentaland health effects.

Commonly in the literature, hydrotreating is discussed in termsof the following operating variables: temperature, pressure, LHSV(Liquid hourly space velocity), and gas/oil [2–4]. There is rarelyany mention of H2 purity even though it is known that H2 purityis an important variable [5]. H2 purity may have been ignored inliterature as very pure H2 streams (>99.9%) were used in the re-search. However, in an industrial setting, H2 purity can be as low

010 Published by Elsevier Ltd. All r

: +1 306 966 4777.apiour), sundaramurthy@src.alai).askatoon, SK, Canada S7N 2X8.

as 70 vol.% [5]. It is therefore necessary to conduct research at con-ditions similar to those found in practice.

The effluent stream of a hydrotreater contains unreacted H2. Foroptimal economics this unreacted H2 must be recovered and recy-cled. The recycle H2 stream contains several impurities, mainly CH4

an H2S [6]. CH4 is produced by cracking and other reactions, andH2S is produced as a result of HDS. Methane is an important con-cern because it is difficult to remove and often builds up to highlevels in the recycle stream [7]. In our previous work [8] it wasfound that methane was inert toward the commercial NiMo/c-Al2O3 catalyst and its presence hindered hydrotreating activitiesby reducing the H2 partial pressure in the reactor. H2S is knownto inhibit hydrotreating activities yet its presence is crucial tomaintain the sulfided active phase of the catalyst [9].

In our previous work [8], the effect of H2 purity on hydrotreat-ing activities was studied. However, the interacting effects of H2

purity with other hydrotreating operating variables were notinvestigated. Moreover, no models were developed to correlatehydrotreating activities, namely: HDS, HDN, and HDA, with theoperating conditions: namely H2 purity, pressure, and gas/oil ratio.The objective of this work therefore is to develop models for hydro-treating activities as functions of operating variables, specifically

ights reserved.

M. Mapiour et al. / Fuel 89 (2010) 2536–2543 2537

H2 purity, pressure, and gas/oil ratio, and to optimize these threevariables for maximum hydrotreating activities. H2 purity, pres-sure, and gas/oil ratio were specifically chosen because their effectson H2 partial pressure are greater than those of temperature andLHSV especially for heavier feedstock. The effects of temperatureand LHSV on hydrotreating activities will also be presented in thispaper.

2. Experimental

Heavy gas oil (HGO) derived from oil sands bitumen supplied bySyncrude Canada Ltd. was used in this study as the feedstock. Someproperties of this HGO are summarized in Table 1. The catalyst em-ployed was a trilobe-shaped commercial NiMo/c-alumina (C424)from Criterion Catalysts with a diameter of �1.5 mm. The catalysthad a pore volume of 0.5 mL/g and a surface area of 160 m2/g.

The experiments were carried out in a micro-trickle bed reactorusing 5 mL of catalyst. A schematic of the experimental setup is gi-ven elsewhere [8]. The reactor had an internal diameter of 10 mmand a length of 240 mm. The catalyst, diluted with silicon carbide(SiC), was loaded into the reactor. This was to ensure that problemsof hydrodynamics, poor catalyst wetting, wall effects, and backmixing were mitigated [3]. For loading the catalyst, the bottomend of the reactor was sealed with a Swagelok 60 micron stainlessfilter (Solon, OH, USA) and then packed from bottom to top in threeparts. The bottom part was loaded with 22 mm of glass beads ofsize 3 mm diameter followed by 25 mm, 10 mm, and 10 mm of16 mesh, 46 mesh, and 60 mesh SiC, respectively. In the middlepart of the reactor, 5 mL of catalyst and 12 mL of 90 mesh SiC wereloaded alternately; small quantity of each at a time, for a totalnumber of 10–12 layers. Finally, the top part was loaded with8 mm of SiC of 60 mesh followed by 8 mm, 8 mm, and 20 mm of46 mesh SiC, 16 mesh SiC, and 3 mm diameter glass beads, respec-tively. The top 20 mm of the reactor was kept empty.

The sulfiding solution was made of 2.9 vol.% butanethiol, a com-monly used sulfiding agent [10], in electrical insulating oil (VOLT-ESSO 35). The catalyst was introduced into the reactor in an oxideform, but is required to be activated by converting into a sulfidedform (hence the sulfiding agent); the electrical insulating oil actsas a carrier liquid. 100 mL of the sulfiding solution was pumpedinto the reactor at a high flowrate (�2.5 mL/min) to wet the cata-lyst. Subsequently, the flowrate was adjusted to 5 mL/h and main-tained. Gas/oil ratio was operated at 600 mL/mL. The catalyst bedwas initially heated to 100 �C and gradually increased from 100to 193 �C. The reactor temperature was then maintained for 24 h.Next, the temperature was further increased to 343 �C in stepsand the reactor was kept at this temperature for another 24 h.

The catalyst was equilibrated (precoked) for seven days aftercatalyst sulfidation by flowing HGO into the reactor at the rate of

Table 1Properties of heavy gas oil feedstock.

Physical propertiesBoiling range, �C 258–592Density at 20 �C, g/cm3 0.988Sulfur content, ppm 42,310Nitrogen content, ppm 3156Aromatics content, wt.% 31.4

Simulated distillationFraction Boiling range (�C) Amount (wt.%)

Gasoline IBP to 205 0Kerosene 205–260 1Light gas oil 260–315 4Heavy gas oil 315–425 40Vacuum gas oil 425–600 55

5 mL/h. The temperature of the reactor was increased to 375 �C.The purpose of catalyst precoking was to stabilize its activity to en-sure uniform activity across the catalyst surface before the exper-iments were conducted [11]. Liquid products were collected every24 h, stripped with nitrogen gas to remove dissolved NH3 and H2S,and analyzed for sulfur and total nitrogen contents. After the cata-lyst stabilization, the experiments were carried out as designed.Each experiment was run for three days, and product withdrawnevery 24 h. A transient period of 24 h was allowed after a changein process conditions and samples taken within this period werediscarded. Products collected in the second and third days wereanalyzed for sulfur, nitrogen, and aromatics conversions. To detectif there were any changes in the catalyst activity during a run, anexperiment at the precoking condition (100% H2 purity, tempera-ture of 375 �C, pressure of 9 MPa, gas/oil of 600 mL/mL, and LHSVof 1 h�1) was intermittently repeated before and after each exper-iment. It was assumed that the catalyst did not undergo deactiva-tion if the hydrotreating conversions, at the abovementionedprecoking condition, ‘‘before” and ‘‘after” an experiment were notsignificantly different.

Sulfur contents of the feed and the liquid products were deter-mined using a combustion/fluorescence technique according toASTM 5463 procedure. Total nitrogen contents of the feed andthe liquid products were measured using a combustion/chemilu-minescence technique following ASTM D4629 procedure. Aroma-ticity, defined as the mole percent of carbon in a sample that ispresent as part of an aromatic ring structure, of the feed and theliquid products was determined by the carbon-13 nuclear mag-netic resonance (13C NMR) spectroscopy. The spectra were ob-tained in the Fourier Transform mode operating at a frequency of500 MHz. The instrumental conditions were as follows: a pulse de-lay of 2 seconds, a sweep width of 27.7 kHz and gated decoupling.Overall time for each sample was 56 minutes for 1056 scans. Deu-terated chloroform, CDCl3, was used as a solvent.

The HDS and HDN conversions were calculated as follows:

%conversion of speciesðiÞ ¼ speciesðiÞ in feed� speciesðiÞ in productspeciesðiÞ in feed

� 100

ð1Þ

where species(i) is sulfur, nitrogen, or aromatics.The concentrations of sulfur and nitrogen were measured by an

S/N Analyzer; and the results directly substituted in Eq. (1). ForHDA, the results were obtained from 13C NMR spectra. The spec-trum is composed of two distinct zones separated by the solventbar. Total saturated hydrocarbons were located between 0 and50 ppm; whereas the total aromatics were observed between 120and 150 ppm [12]. Eq. (2) was then used to determine the aromat-ics content (%) of each sample. The integrals of the saturatedhydrocarbons zone, Isat, and the total aromatics zone, Iar, weredetermined using XWIN-NMR 3.5 software. The values from Eq.(2) were substituted in Eq. (1) to determine HDA conversions

Car ¼Iar

Iar þ Isat� 100 ð2Þ

where Car, aromatics content; Iar, the integral of total aromatics; Isat,the integral of total saturates.

3. Results and discussion

3.1. Effects of pressure, H2 purity, and gas/oil ratio on hydrotreatingconversions

Often H2 purity is not included as an operating variable inhydrotreating studies even though, along with pressure and

Table 3Results of test of the significance of factors and interactions for HDS, HDN, and HDAmodels.

Factor or interaction p-Value of factor or interaction

HDS HDN HDA

Purity 0.2916 <0.0001 <0.0001Pressure 0.6136 <0.0001 <0.0001Gas/oil 0.8961 0.0199 0.0005(Purity)2 – 0.0030 –(Pressure)2 – – –(Gas/oil)2 – – –Purity x Pressure – – –Purity x Gas/oil – – –Pressure x Gas/oil – – –Model 0.6937 <0.0001 <0.0001

Table 4R-Squared statistics for the develop models of HDS, HDN and HDA.

Model R2 Adjusted R2 Predicted R2

HDS 0.0843 �0.0875 �0.4553HDN 0.9262 0.9065 0.8570HDA 0.9125 0.8961 0.8496

2538 M. Mapiour et al. / Fuel 89 (2010) 2536–2543

gas/oil ratio, it has a significant effect on H2 partial pressure. Forthis reason, H2 purity, pressure, and gas/oil ratio, were used in anexperimental design using a central composite design method(available in Expert design 6.0.1) in an effort to study their effectson HDS, HDN, and HDA activities. H2 purity, pressure, and gas/oilratio were varied within the range of 75–100 vol.% (with the restmethane), 7–11 MPa, and 400–1200 mL/mL, respectively. Temper-ature and LHSV were kept constant at 380 �C and 1 h�1, respec-tively. The experimental results obtained at conditions specifiedby the experimental design are summarized in Table 2. Analysisof the experimental results was carried out using DESIGN-EXPERT6.0.1 to optimize the considered operating conditions with respectto HDS, HDN, and HDA conversions. Regression analysis of experi-mental data generated the following regression equations for HDS,HDN and HDA:

HDS ¼ þ89:77468þ 0:046751� Purityþ 0:13798� Pressure

� 1:77717� 10�4 � gas=oil ð3Þ

HDN ¼ þ241:68506� 5:87945� Purityþ 4:87321� Pressure

þ 6:70444� 10�3 � gas=oilþ 0:037787� Purity2 ð4Þ

HDA ¼ þ2:07305þ 0:32246� Purityþ 1:95801� Pressure

þ 4:98039� 10�3 � gas=oil ð5Þ

where HDS, HDN, HDA are in percentage conversions (%); purity,pressure, gas/oil are in vol.%, MPa, and mL/mL, respectively. Theequations are valid within the operating conditions studied.

Tables 3 and 4 contain the test of significance (f-test) and R2 testresults, respectively, of HDS, HDN and HDA models. Table 3 showsthat the operating variables have no significant effects on HDS, i.e.p-value > 0.05 [13]. Moreover, Table 4 shows that R2 value for HDSmodel is very small, 0.0843, which means that HDS developedmodel poorly represents the experimental data. The f-test ofHDN and HDA data shows that pressure, H2 purity, and gas/oil ratiohave significant effects on HDN and HDA activities. Moreover, itshows that these three factors do not interact as they affect HDNand HDA activities, meaning that the effect of each factor is inde-

Table 2Experimental results at the conditions specified by the CCD experimental design.

Run Experimental conditions Response

# Pressure(MPa)

Gas/oil(mL/mL)

Purity(%)

HDN(%)

HDS(%)

HDA(%)

Inlet H2

partialpressure(MPa)

OutletH2

partialpressure(MPa)

1 9 800 75 63.2 95.8 48.9 6.7 5.32 7.8 1038 80 59.1 95.4 48.1 6.2 5.43 10.2 562 80 67.4 93.6 51.2 8.1 5.84 7.8 562 80 55.1 94.9 46 6.2 4.65 10.2 1038 80 67.6 93.2 52.6 8.1 76 9 400 88 59.9 96 47.9 7.8 5.27 9 1200 88 65 95.3 53 7.8 6.98 7 800 88 54.4 94.5 46.1 6.1 5.69 9 800 88 66.7 95.5 51.8 7.8 6.8

10 9 800 88 65.7 94 52.4 7.8 6.811 11 800 88 72.8 95 54.9 9.6 8.312 9 800 88 68.2 95.6 52.4 7.8 6.813 9 800 88 66.6 92.9 53.1 7.8 6.814 9 800 88 69.6 95.8 52.6 7.8 6.815 9 800 88 66.3 95 51.8 7.8 6.816 7.8 1038 95 69.3 93.8 53.9 7.3 6.817 7.8 562 95 65.7 94 52.3 7.3 6.218 10.2 562 95 78.5 96 55.5 9.5 8.119 10.2 1038 95 83.9 96.7 58 9.6 8.920 9 800 100 78.8 96.6 55.4 8.9 8.2

pendent of the values of the other two factors. R2 values of HDNand HDA developed models were 0.9262 and 0.9125, respectively.

Eqs. (4) and (5) do not show straightway dependence of HDNand HDA conversions on the operating variables. Therefore, toclearly illustrate the dependence, surface response plots weredeveloped, and are presented in Figs. 1 and 2 for HDN and HDA,respectively. These figures show that increasing pressure, H2 pur-ity, and gas oil ratio led to increases in HDN and HDA conversions.As interpreting the 3-D surface response can be difficult, the per-turbation plots of the effects of the variables on HDS (no surface re-sponse of HDS is shown), HDN and HDA are provided in Figs. 3–5.When interpreting a perturbation plot, one needs to be cautioussince it looks only at one-dimensional paths through a multifactorsurface. Therefore, it is recommended that perturbation plots areused in conjunction with the 3-D surface responses. Nonetheless,it is a powerful method of comparing the relative influences of fac-tors [14].

A perturbation plot shows the effect of each individual variableas the others are held constant [14]. Fig. 3 shows that effects ofpressure, H2 purity, and gas/oil ratio on HDS are not considerablysignificant. Nonetheless, Fig. 3 does show that the effects of pres-sure and H2 purity on HDS are slightly greater than that of thegas/oil ratio. Figs. 4 and 5 show that the effects of the pressure,H2 purity, and gas/oil ratio on HDN and HDA are noticeably signif-icant, and that increasing these variables led to increases in HDNand HDA conversions. Moreover, these figures show that the ef-fects of pressure and H2 purity on HDN and HDA are more signifi-cant than that of the gas/oil ratio.

In Figs. 4 and 5, it can also be observed that effects of the vari-ables are greater on HDN than on HDA. This correlatively impliesthat the effect of H2 partial pressure is far greater on HDN thanon HDA. However, by considering the mechanisms of HDN andHDA, one may expect the contrary. HDA reaction proceeds throughhydrogenation, whereas HDN reaction proceeds through hydroge-nation followed by hydrogenolysis, and hydrogenolysis is not af-fected by H2 partial pressure [1]. One explanation may be that atan operating temperature of 380 �C HDA activity is close to opti-mum due to equilibrium thermodynamic limitation [1,15]. Thus,the effects of the other variables are not as significant as theywould have been at lower operating temperatures. A second expla-nation may be the fact that the overall rate of HDN is frequently

60.80 64.32 67.85 71.37 74.89

HD

N, %

562.16681.08

800.00918.92

1037.84

80.07 83.78

87.50 91.22

94.93

Gas/oil, mL/mLH2 purity, vol. %

60.80 64.32 67.85 71.37 74.89

HD

N, %

562.16681.08

800.00918.92

1037.84

80.07 83.78

87.50 91.22

94.93

Gas/oil, mL/mLH2 purity, vol. %

56.60 62.22 67.85 73.47 79.09

HD

N, %

80.0783.78

87.5091.22

94.93

7.81 8.41

9.00 9.59

10.19

H2 purity, vol.%Pressure , MPa

56.60 62.22 67.85 73.47 79.09

HD

N, %

80.0783.78

87.5091.22

94.93

7.81 8.41

9.00 9.59

10.19

H2 purity, vol.%Pressure , MPa

Fig. 1. Surface response of the effects of pressure, H2 purity, and gas/oil ratio onHDN activity.

47.16 49.53 51.89 54.25 56.62

HD

A, %

80.0783.78

87.5091.22

94.93

7.81 8.41

9.00 9.59

10.19

H2 purity, vol. %Pressure , MPa

47.16 49.53 51.89 54.25 56.62

HD

A, %

80.0783.78

87.5091.22

94.93

7.81 8.41

9.00 9.59

10.19

H2 purity, vol. %Pressure , MPa

48.31 50.10 51.89 53.68 55.47

HD

A, %

562.16681.08

800.00918.92

1037.84

80.07 83.78

87.50 91.22

94.93

Gas/oil, mL/mLH2 purity, vol.%

48.31 50.10 51.89 53.68 55.47

HD

A, %

562.16681.08

800.00918.92

1037.84

80.07 83.78

87.50 91.22

94.93

Gas/oil, mL/mLH2 purity, vol.%

Fig. 2. Surface response of the effects of pressure, H2 purity, and gas/oil ratio onHDA activity.

Deviation from Reference Point

HD

S co

nver

sion

-1.000 -0.500 0.000 0.500 1.000

92.6

93.625

94.65

95.675

96.7

A

A

BBC

C

Deviation from Reference Point

HD

S co

nver

sion

-1.000 -0.500 0.000 0.500 1.000

92.6

93.625

94.65

95.675

96.7

A

A

BBC

C

Fig. 3. HDS perturbation plot: (A) is H2 purity, (B) is pressure, and (C) is gas/oilratio.

M. Mapiour et al. / Fuel 89 (2010) 2536–2543 2539

determined by the hydrogenation rate rather than by hydrogenol-ysis [1], and hydrogenolysis is not affected by H2 partial pressure.Thus, pressure, H2 purity, and gas/oil ratio can only affect hydroge-nation processes in HDN and HDA reactions. Since, hydrogenationof aromatic rings with heteroatoms is easier than of those whichlack a heteroatom, HDN is more affected than HDA. A possibleand plausible third explanation is that the initial concentration ofaromatics is about 100 folds that of the nitrogen concentration inthe feed, indicating that no fair comparison can be made by lookingat HDN and HDA conversions.

Difference among the HDS, HDN, and HDA mechanism may of-fer an explanation as to why there are dissimilarities in the effectsof the variables on HDN and HDA versus those on HDS. As previ-ously mentioned in this section, HDN reaction takes place viahydrogenation followed by hydrogenolysis, and HDA reaction oc-curs via hydrogenation. Increasing H2 purity, pressure, and gas/oil ratio result in increases in H2 partial pressure. This increase inH2 partial pressure directly affects the hydrogenation process. Asa result, changes in HDA and HDN conversions were observed asH2 purity, pressure, and gas/oil ratio were varied. On the otherhand, HDS reaction can proceed via two pathways: (1) hydrogena-tion followed by hydrogenolysis or (2) direct hydrogenolysis[4,16]. Consequently, HDS conversions were only very slightly af-fected by H2 partial pressure since HDS reaction has the optionof taking place directly via hydrogenolysis. Consequently, no sig-nificant effects of pressure, H2 purity, and gas/oil ratio on HDS con-versions were observed. The rate of the HDS two pathwaysdepends on a number of factors: catalyst’s properties, feed’s type,and reaction conditions. It was suggested that the hydrogenation

step of the HDS took place at the S-edge while the S–C scission stepoccurred at both S-edge and Mo-edge; where S-edge and Mo-edgeare active sites. At lower H2 partial pressure, S-edge is inactive andthe catalytic reactions therefore occur at Mo-edge; hence, S–C is fa-vored [17].

The optimal operating conditions were calculated based on con-straints in which HDS, HDN, and HDA conversions were to be

Deviation from Reference Point

HD

N c

onve

rsio

n

-1.000 -0.500 0.000 0.500 1.000

54.4

61.775

69.15

76.525

83.9

A

A

B

B

C

C

Deviation from Reference Point

HD

N c

onve

rsio

n

-1.000 -0.500 0.000 0.500 1.000

54.4

61.775

69.15

76.525

83.9

A

A

B

B

C

C

Fig. 4. HDN perturbation plot: (A) is H2 purity, (B) is pressure, and (C) is gas/oilratio.

HD

A c

onve

rsio

n

-1.000 -0.500 0.000 0.500 1.000

46

49

52

55

58

A

A

B

B

C

C

Deviation from Reference Point

A

A

B

B

C

C

Fig. 5. HDA perturbation plot: (A) is H2 purity, (B) is pressure, and (C) is gas/oilratio.

70

80

90

100

sion

s, % 56.0

60.0

sion

s, %

HDS HDN HDA

70

80

90

100

HDS HDN HDAHDS HDN HDA

2540 M. Mapiour et al. / Fuel 89 (2010) 2536–2543

maximal within the ranges of the operating variables studied. Thecollective optimum operating conditions for HDS, HDN, and HDAwere determined to be: pressure of 10.2 MPa, H2 purity of95 vol.%, and gas/oil ratio of 1037 mL/mL. Experiments were con-ducted under these conditions, and the experimental data wascompared to those predicted (see Table 5). As shown in Table 5,the percentage differences of HDS, HDN, and HDA for the experi-mental results versus the predicted results were 0.7%, 2.3%, and0.2%, respectively. It may be noted that the results for HDS maynot be reliable since the R2 of the developed model was only0.0843.

Table 5Comparison between the predicted and observed values of hydrotreating at optimalconditions: pressure of 10.1 MPa, H2 purity of 95 vol.%, and gas/oil ratio of 1037 mL/mL. Temperature and LHSV were 380 �C and 1 h�1, respectively.

Reactions Predicted bymodels (%)

Observedexperimentally (%)

Percentagedifferences (%)

HDS 95.4 96.1 0.7HDN 80.1 82.0 2.3HDA 57.8 57.9 0.2

3.2. Effects of temperature and LHSV on hydrotreating conversions

Effects of temperature and LHSV on the hydrotreating conver-sions were also studied. Temperature and LHSV ranges were360–400 �C and 0.65–2 h�1, respectively. H2 purity, pressure andgas/oil ratio were kept constant at 100%, 9 MPa, and 800 mL/mL,respectively. LHSV was kept constant at 1 h�1 when the effect oftemperature was studied. Temperature was kept constant at380 �C when the effect of LHSV was studied.

LHSV, which is the inverse of residence time, is an indication ofthe time spent in the reactor by the reactants [4]. It was observedthat decreasing LHSV led to increases in HDS, HDN, and HDA con-versions (see Fig. 6). However, one needs to bear in mind that forHDA this observation is only true for the conditions employed inthis work, especially temperature. For example, Mann et al. [18]found that HDA is independent of LHSV (between 0.5 and 4 h�1)at the temperature of 450 �C and pressure of 6.99 MPa. The reasonis that HDA maximum conversion is achieved between 370 and400 �C (usually 375–385 �C) due to the interrelation between ther-modynamic equilibrium and reaction rates [15]. The authors ob-served similar results for HDS and HDN to those found in this work.

Increasing the temperature generally leads to increases inhydrotreating conversions. Nevertheless, excessive temperaturemay impose thermodynamic equilibrium limitations leading to de-creases in hydrotreating conversions. In the cases of HDS and HDN,this hindering effect of temperature is observed at temperatureshigher than those used in practice (i.e. >425 �C) [1]. In the case ofHDA, the hindering effect of temperature is observed at lower tem-peratures than that of HDS and HDN. As mentioned earlier in thissection, the maximum HDA conversion usually occurs at tempera-ture range of 370–385 �C [15]; this range could be little higher ifthe H2 partial pressure is substantially increased.

In this work it was found that both HDS and HDN conversionsincrease with increasing temperature, however, HDN shows supe-rior increases than that of HDS (see Fig. 7). This superior effect oftemperature on HDN could not be because HDN has higher reac-tion rate than HDS. The bond energy of C@N {147 kcal/mol} ishigher than that of C@S {114–128 kcal/mol} (i.e. C–N scission inmore difficult than that of C–S). Moreover, N (0.75 Å) has smalleratomic radius than S (1.09 Å), therefore is more difficult to removeN than S [19]. The radius effect may be explained as follow: thesurface reaction steps can be simplified as follow: (i) adsorptiononto the active site, (ii) reaction, and (iii) desorption. The adsorp-tion is via lone pairs or election clouds. For the nitrogen atom

20

30

40

50

60

0 0.5 1 1.5 2 2.5

LHSV, h-1

Con

ver

48.0

52.0 Con

ver

20

30

40

50

60

Fig. 6. Effect of LHSV on hydrotreating conversions. Pressure, temperature, H2

purity, and gas/oil ratio were 9 MPa, 380 �C, 100%, and 800 mL/mL, respectively.

HDS HDN HDA

20

30

40

50

60

70

80

90

100

350 360 370 380 390 400 410Temperature ,°C

HD

S an

d H

DN

Con

vers

ions

, wt.

%

48

49

50

51

52

53

54

55

56

57

HD

A C

onve

rsio

n, w

t.%

HDS HDN HDA

Fig. 7. Effect of temperature on hydrotreating conversions. Pressure, LHSV, H2

purity, and gas/oil ratio were 9 MPa, 1 h�1, 100%, and 800 mL/mL, respectively.

M. Mapiour et al. / Fuel 89 (2010) 2536–2543 2541

the radius is smaller thus valence electrons have higher interactionwith the nucleus; consequently, nitrogen has lesser adsorptionstrength in comparison to sulfur.

Thus in theory, HDS should be more significantly affected bytemperature than HDN; however, this is not the case. An explana-tion may be that, at the temperature range under study, the effectof temperature on HDS starts to subside, while the effect of tem-perature on HDN starts to become more pronounced. In a studyby Mann et al. [18] using NiMo/c-Al2O3 as a catalyst and HGO asa feed, it was found that in the temperature range of 300–350 �C,HDS and HDN percentage conversions per �C (degree Celsius) were0.25% and 0.10%/�C, respectively. However, in the temperaturerange of 350–400 �C HDN has a higher percentage conversion per�C (0.36%/�C) than HDS (0.22%/�C).

Fig. 7 also shows that HDA conversion passes through a maxi-mum with respect to the temperature, which is in agreement withthe literature. By taking the first derivative the empirical equation(see Eq. (6)), maximum HDA conversion was determined to haveoccurred at 385 �C. From the foregoing discussion it seems thattemperature is most critical of all of the variables.

HDA ¼ �0:0089Temperature2 þ 6:8447Temperature

� 1258 ð6Þ

3.3. Kinetics of HDS, HDN, and HDA

The effects of the operating variables on hydrotreating perfor-mance can be predicted by a suitable kinetic expression or model[16,20]. H2 partial pressure often appears as a variable in manyhydrotreating kinetic expressions but one must decide whetherto use inlet or outlet H2 partial pressure in the calculations. McCul-loch and Roeder [21] suggested that more meaningful results areattained when outlet H2 partial pressure is used especially from

Table 6Power law model kinetic parameters for HDS, HDN, and HDA (temperature range is: 360–

Parameter Using inlet H2pp

HDS HDN HDAa HDAb

ko 171 2564 37.2 0.01E 24.3 56.8 27.7 �56.5m 0.21 1.33 0.57 0.58c �0.58 �1.16 �0.21 �0.22R2 0.74 0.90 0.82 0.81Adjusted R2 0.71 0.88 0.79 0.74

E = kJ/mol.a Temperature 6 380 �C.b Temperature P 380 �C.

the catalyst’s deactivation standpoint, however, the authors didnot support this suggestion with experimental evidences.

A possible way to validate this suggestion is to first determinethe parameters in a kinetic expression, containing H2 partial pres-sure as a variable, using both inlet and outlet H2 partial pressure.Next, the two resulting kinetic expressions, one with inlet H2 par-tial pressure and another with outlet H2 partial pressure, can thenbe compared through statistical analysis to determine which has abetter prediction ability with higher R2 and adjusted R2. Finally,predicted data from these two expressions can be comparedagainst experimental data generated at conditions different thanthose used in the models’ development.

In an attempt to evaluate the above suggestion, experimentswere conducted under the following conditions: temperaturerange of 360–400 �C, pressure range of 7–11, LHSV range of0.65–2 h�1, and gas/oil range of 400–1200 mL/mL. Inlet and outletH2 partial pressures were determined using HYSYS and their valuesin Table 2. The kinetic parameters in Eq. (7) were then determinedusing the inlet and outlet H2 partial pressure data along with therest of the variables. Eq. (7) is the power law model [16], andEqs. (7.b) and (7.c) are the solutions for Eq. (7) depending on thevalue of n (reaction order)

� dCdt¼ ki � Pm

H � Cn ð7Þ

where kiðTÞ ¼ ko � e�E=RT ð7:aÞ

For Eq. (1), the solution is:

lnCf

Cp¼ ko � eð�E=RTÞ � Pm

H

LHSVC ; n ¼ 1 ð7:bÞ

1n� 1

1Cn�1

p

� 1Cn�1

f

" #¼ ko � eð�E=RTÞ � Pm

H

LHSVC ; n > 1 ð7:cÞ

where Cf and Cp, initial and final concentrations of S, N, or aromatics,respectively, wt.%. LHSV, liquid hourly space velocity, h�1; R, gasconstant, 8.314; T, temperature, K; ki, apparent rate constant; ko,pre-exponential factor; PH, inlet or outlet H2 partial pressure; n,reaction order; E, activation energy; m and c, empirical regressionfactors.

The experimental data was analyzed using non-linear regres-sion model in Polymath software [20]. The parameters for HDS,HDN and HDA are presented in Table 6. HDS, HDN, and HDA wereassumed to be pseudo-first order, i.e. n = 1 [1], and consequentlyEq. (7.b) was used in the determination of the kinetic parameters.The R2 and adjusted R2, presented in Table 6, suggest that for HDSand HDN kinetic interpretation of the data using either inlet or out-let H2 partial pressure will generate models with the same predic-tive ability. However for HDA, kinetic interpretation of data usingoutlet H2 partial pressure generates models with better predictiveability than using inlet H2 partial pressure. For example, for the

400 �C).

Using outlet H2pp

HDS HDN HDAa HDAb

26 164 11012 150 0.007523.3 60.1 33.6 �20.1

0.16 0.99 0.46 0.46�0.59 �1.21 �0.24 �0.24

0.74 0.90 0.90 0.900.71 0.89 0.89 0.88

Table 8Results of the catalyst deactivation testing (temperature, gas/oil ratio, LHSV were380 �C, 800 mL/mL, and 1 h�1, respectively).

Activity For H2 partial pressure of8.1 MPa

For H2 partial pressure of4.5 MPa

Before After Before After

HDN (%) 75 75 74 68HDA (%) 53 54 53 50

2542 M. Mapiour et al. / Fuel 89 (2010) 2536–2543

temperature range of 360–380 �C, the R2 and adjusted R2 for themodel generated using inlet H2 partial pressure are 0.82 and0.79, respectively, whereas for that generated using outlet H2 par-tial pressure they were 0.90 and 0.89, respectively.

After models were developed new experimental data were usedto validate the models. Three experiments were conducted inwhich pressure, temperature, LHSV, gas/oil ratio were kept con-stant at 9 MPa, 380 �C, 1 h�1, and 800 mL/mL, respectively, andH2 purity was varied as follows: 50, 80, and 90 vol.% (with the restmethane). None of these chosen conditions were used in the devel-opment of the kinetic models. Also, note that H2 purity of 50 vol.%is an extrapolated condition, i.e. falls outside the range of the con-ditions originally used to develop the models. The experimental re-sults from these experiments were then compared against thosepredicted by the models and the results are presented in Table 7.

Table 7 shows that the predictions from the developed kineticmodels for HDS and HDN were in close agreement with the exper-imental values regardless of whether inlet or outlet H2 partial pres-sure was used. For HDA, the kinetic model developed using outletH2 partial pressure produced ever-so-slightly better predictionsthan those produced using inlet H2 partial pressure. Thus, it seemsthat it does not matter whether hydrotreating conclusions weredrawn from inlet or outlet H2 partial pressure. However, these con-clusions do not address catalyst deactivation.

Catalyst deactivation is the most important concern in any cat-alytic process. In hydrotreating it is well known that increasingH2 partial pressure results in a decrease in deactivation rate. Dueto H2 consumption, the reactor’s outlet H2 partial pressure can beconsiderably lower than its inlet H2 partial pressure, especially forheavy feedstock (see Table 2). Therefore the portion of the catalystbed at and near the reactor outlet may to experience higher deacti-vation rate as a results of lower H2 partial pressure environment.Botchwey et al. [22] and Alvarez and Ancheyta [23] have shownexperimentally that the largest portion of hydrotreating conver-sions take place in the first �30% of the catalyst bed’s length. Con-sequently this is also where most of the hydrogen consumptiontakes place, leaving a large portion of the catalsyt’s bed at a H2 par-tial pressure level significantly lower than that at the inlet. Anothernoteworthy point that can be deduced from Botchwey et al. [22]and Alvarez and Ancheyta [23] findings is that for about �70% ofthe catalyst bed’s length the H2 partial pressure level is closer in va-lue to the outlet H2 partial pressure than it is to the inlet H2 partialpressure. In other words, outlet H2 partial pressure level resemblesthat experienced by most parts of the catalyst’s bed. Therefore, it isvery critical that the outlet H2 partial pressure is determined somore complete and meaningful conclusions are drawn.

To show the relationship between the catalyst deactivation andH2 partial pressure levels, a commercial NiMo/c-Al2O3 catalyst bedwas subjected to two inlet H2 partial pressures of 4.5 MPa and8.1 MPa for a period of three days while hydrotreating experimentswere being carried out. The temperature and LHSV were 380 �Cand 1 h�1, respectively for all experiments. Before the catalystbed was subjected to either of the two H2 partial pressure levels

Table 7Comparison of the prediction results using inlet versus outlet H2 partial pressure(temperature, pressure, gas/oil ratio, LHSV were 380 �C, 9 MPa, 800 mL/mL, and 1 h�1,respectively).

Purity(vol.%)

Determinedexperimentally (%)

Determined from the models (%)

Using inlet H2

partial pressureUsing outlet H2

partial pressure

HDS HDN HDA HDS HDN HDA HDS HDN HDA

50 92 38 38 93 42 41 93 38 3980 95 64 48 95 64 50 95 63 5090 96 71 49 95 69 53 95 70 53

a designated experiment with inlet H2 partial pressure of 9 MPa,named ‘‘control”, was conducted. The same ‘‘control” experimentwas then repeated after the catalyst bed had been subjected to in-let H2 partial pressure of 4.5 MPa or 8.1 MPa. The hypothesis wasthat if the ‘‘before” and the ‘‘after” hydrotreating conversions ofthe ‘‘control’ experiment were different it can be concluded thatthe catalyst underwent some deactivation. The results are pre-sented in Table 8.

Table 8 shows that for the experiement at 4.5 MPa inlet H2 par-tial pressure, the ‘‘after” HDN and HDA conversions of the ‘‘control”are lower than those of the ‘‘before” conversions, meaning that thecatalyst suffered deactivation due to low H2 partial pressure levels.No significant differences were observed between the ‘‘after” and‘‘before” HDN and HDA conversions in the experiment conductedat H2 partial pressure of 8.1 MPa. Thus, it can be seen that lowerH2 partial pressure may cause severe catalyst deactivation in a veryshort time. It is therefore important that the entire catalyst bed ismaintained at a high enough H2 partial pressure to avoid deactiva-tion. One way of ensuring that the entire catalyst bed is at suffi-cient H2 partial pressure level is to maintain high outlet H2

partial pressure, the reactor point with the lowest H2 partial pres-sure, to avoid untimely catalyst deactivation.

4. Conclusion

Within the range of the experimental conditions considered inthis study the following conclusions were made:

� Increasing pressure, H2 purity, and gas/oil ratio led to increasesin HDN and HDA activities. Effects of these variables on HDSactivity are not considerably significant.� The positive effects of H2 purity on HDN and HDA activities

were greater than those of gas/oil ratio and comparable to thoseof pressure.� The optimal conditions for HDS, HDN, and HDA are: pressure of

10.1 MPa, H2 purity of 95 vol.%, and gas/oil ratio of 1037 mL/mL.This is achieved at temperature and LHSV of 380 �C and 1 h�1,respectively.� Decreasing LHSV led to increases in HDS, HDN, and HDA activ-

ities, while increasing temperature led to increases in HDS andHDN. HDA passed through a maximum of 380 �C as the temper-ature was varied.� A low H2 partial pressure environment can accelerate the cata-

lyst’s deactivation� Information on H2 partial pressure effects on hydrotreating

activities can be equally satisfactorily obtained using eitherinlet or outlet H2 partial pressure. However, from the catalystdeactivation standpoint it is vital to use outlet H2 partial pres-sure, since it is the reactor point with the lowest H2 partialpressure.

References

[1] Girgis Michael J, Gates Bruce C. Reactivities, reaction networks, and kinetics inhigh-pressure catalytic hydroprocessing. Ind Eng Chem Res 1991;30:2021–2.

M. Mapiour et al. / Fuel 89 (2010) 2536–2543 2543

[2] Speight JG. The desulfurization of heavy oils and residua. New York: MarcelDekker Inc.; 1981.

[3] Bej K, Ajay Dalai, John Adjaye. Comparison of hydrodenitrogenation of basicnonbasic nitrogen compounds present in oil sands derived heavy gas oil. EnergFuel 2001;15:377.

[4] Christian Botchwey, Ajay Dalai, John Adjaye. Product selectivity duringhydrotreating and mild hydrocracking of bitumen-derived gas oil. Energ Fuel2003;17:1372.

[5] Adrian Gruia. In: David SJ Jones, Peter R Pujadó, editors, Handbook ofpetroleum processing. The Netherlands: Springer; 2006 [Chapter 8].

[6] Peramanu S, Cox BG, Pruden BB. Economics of hydrogen recovery processes forthe purification of hydroprocessor purge and off gases. Int J Hydrogen Energ1999;24:405.

[7] Turner J, Reisdorf M. Consider revamping hydrotreaters to handle higher H2

partial pressures. Hydrocarbon Process 2004;March:61–70.[8] Mapiour M, Sundaramurthy V, Dalai AK, Adjaye J. Effect of hydrogen purity on

hydroprocessing of heavy gas oil derived from oil-sands bitumen. Energ Fuel2009;23(4):2129–35.

[9] Christian Botchwey, Ajay Dalai, John Adjaye. Two-stage hydrotreating ofAthabasca heavy gas oil with interstage hydrogen sulfide removal: effect ofprocess conditions and kinetic analyses. Ind Eng Chem Res 2004;43(18):5854.

[10] Andari Mounif K, Abu-Seedo Fatima, Stanislaus Anthony, Qabazard Hassan M.Kinetics of individual sulfur compounds in deep hydrodesulfurization ofKuwait diesel oil. Fuel 1996;75(14):1664–70.

[11] Speight JG. The desulfurization of heavy oils and residua. New York: MarcelDekker Inc.; 2000.

[12] Owusu-Boakye, Abena. Two-stage aromatics hydrogenation of bitumen-derived light gas oil. Master Thesis, University of Saskatchewan; 2005.

[13] Montgomery DC. Design and analysis of experiments, 4th ed. USA: John Wiley& Sons; 1997.

[14] Mark J Anderson, Patrick J Whitcomb. RSM simplified: optimizing processesusing response surface methods for design of experiments. New York, NY:Productivity Press; 2005.

[15] James H Gary, Glenn E Handwerk, Mark J Kaiser. Petroleum refining:technology and economics, 5th ed. New York, NY: CRC Press; 2007.

[16] Knudsen KG, Cooper BH, Topsoe H. Catalyst and process technologies for ultralow sulfur diesel. Appl Catal 1999;189:205.

[17] Poul Georg Moses, Berit Hinnemann, Henrik Topsøe, Jens K Nørskov, Thehydrogenation and direct desulfurization reaction pathway in thiophenehydrodesulfurization over MoS2 catalysts at realistic conditions: a densityfunctional study. J Catal 2007;248:188–203.

[18] Mann RS, Sambi IS, Khulbe K. Catalytic hydrofining of heavy gas oil. Ind EngChem Res 1987;26:410.

[19] Landau MV. Deep desulfurization of diesel fuels: kinetic modeling of modelcompounds in trickle-bed. Catal Today 1997;36:393.

[20] Deena Ferdous, Ajay Dalai, John Adjaye. Hydrodenitrogenation andhydrodesulfurization of heavy gas oil using NiMo/Al2O3 catalyst containingboron: experimental and kinetic studies. Ind Eng Chem Res 2006;45:544–552.

[21] Donald C McCulloch, Roeder RA. Find hydrogen partial pressure. HydrocarbonProcess, 1976;February:81.

[22] Christian Botchwey, Ajay Dalai, John Adjaye. Simulation of a two-stage microtrickle-bed hydrotreating reactor using Athabasca bitumen-derived heavy gasoil over commercial NiMo/Al2O3 catalyst: effect of H2S onhydrodesulfurization and hydrodenitrogenation. Int J Chem React Eng2006;4:A20.

[23] Anton Alvarez, Jorge Ancheyta. Modeling residue hydroprocessing in a multi-fixed-bed reactor system. Appl Catal A: General 2008;351(2):148.