Comparison of the HDS DBT reaction using bulk and supported catalysts

The paper describes in detail the procedure for the preparation of a granular bulk NiMoW catalyst and a supported reference NiMo/Al 2 O 3 catalyst. Mention is made of investigations of the supported and bulk catalysts by various physico-chemical methods (nitrogen adsorption-desorption method, X-ray photoelectron spectroscopy, TPD-NH 3 , HRTEM and X-ray diffraction analysis). The experiments to estimate catalytic activity and compare rate constant of hydrodesulfurization of dibenzothiophene using both cata-lysts have been carried out. It is shown that textural properties of the cata-lysts significantly differ. The supported catalyst has more developed specific surface area and pore volume than the bulk catalyst. TPD-NH 3 showed an increased acidity of the supported catalyst in comparison the bulk catalyst. It is shown by the X-ray photoelectron spectroscopy method that in both samples Mo on the surface is present exclusively in the form of Mo 4+ ion. However, the bulk catalyst differs from the supported catalyst in that it contains a larger amount of Ni as part of the active NiMo(W)S phase. The catalytic activity tests demonstrated that the bulk catalyst is more active at 240, 250 and 260 °C , it is discovered that the rate constant in hy-drodesulfurization of dibenzothiophene for the bulk NiMoW catalyst is twice higher at 240 ° C than that of the supported NiMo/Al 2 O 3 catalyst.


Introduction
Since 1930-th, bi-and/or trimetallic Co(Ni)Mo(W) catalysts supported on porous materials like alumina have been used [1][2][3].However, the existing trends towards increasing the depth of oil refining, which require introduction of heavier secondary fractions in the processing along with straight-run fractions, makes it necessary to improve the catalytic systems or tighten the process con-ditions.One of the solutions of this task for hydrotreating process is the use of bulk catalysts [4][5][6][7].
Bulk hydrotreating catalysts are characterized by the increased number of active sites in the catalyst volume compared to the supported catalysts, which contain more than 50 wt.% of the support that is inert in target reactions [8,9].
There are many preparation techniques of the bulk hydrotreating catalysts, and most of them are based on the preparation of the precursor powders [6,[10][11][12].These methods do not consider the necessity of making granulated bulk catalysts for industrial application, and the existing works primarily focus on developing new preparation techniques of bi-or trimetallic powders and studies of their physico-chemical and catalytic properties.
Recently, we compared the most used preparation techniques of bulk catalysts: the precipitation technique, the hydrothermal treatment of a precursor solution, and the spray drying of a precursor solution [13].It was established that the best synthesis procedure included spray drying of the solution of the Ni-Mo-W precursor.Moreover, it was further demonstrated that this Ni-Mo-W powder can be easily granulated to trilobe shape after plasticizing with the binding agent (pseudoboehmite) and used in the fixed bed hydrotreating reactor [14].
Considering the prospects of the developed preparation technique for granulated bulk catalysts, the question of catalytic activity in the target hydrotreating reactions remains open.Of course, with the right choice of the synthesis procedure, which provides access of the feedstock molecules to the active sites of the bulk catalyst, the activity of the bulk catalyst may be many times higher than that of the supported systems [12].However, most studies on bulk hydrotreating catalysts do not compare the properties of newly developed bulk catalysts, even in powder form, with supported systems, and the question of the superior activity of bulk catalysts is not fully understood.
In the literature, there is information on similar catalytic system of Albemarle company -"the high-dispersed polymetallic (NiMo or NiMoW) bulk catalysts of NEBULA series synthesized by coprecipitation of the corresponding salts of metal precursors" [4,15].According to information provided in open sources, the activity of this system is many times higher than that of supported catalysts [4,9].The rest of the information is a trade secret.
In the present work, to verify the suitability of the newly developed method for the synthesis of a bulk catalyst, a comparison is made between the physicochemical and catalytic characteristics of a bulk Ni-Mo-W catalyst and those of a supported Ni-Mo/Al2O3 catalyst.Moreover, the supported catalyst was obtained using a technique developed specifically for the production of highly active catalysts for the hydrotreating [16][17][18].The similarity of the systems also lies in the fact that the bulk catalyst contains pseudoboehmite as a binder, which is also used to prepare the support in Ni-Mo/Al2O3.
The assessment of catalytic activity in this work was carried out in accordance with the methods accepted in this field [9,19,20].The catalysts were tested in a model feedstock with dibenzothiophene, followed by analysis of residual sulfur content and calculation of the reaction rate constant.The conditions used correspond to those previously selected in [19] to ensure the minimal effect of internal and external diffusion in the study of this type of reaction.

Preparation of catalysts
Preparation of the supported NiMo catalyst consisted of the preparation of alumina support and impregnation solution with active metals following by impregnation of the support with the solution.The support was prepared from pseudoboehmite, which was synthesized according to the procedure described in [21].The preparation procedure of the pseudoboehmite is one of the most eco-friendly among all known and usually used techniques for the synthesis of aluminum hydroxides.The pseudoboehmite was plasticized with the solution of nitric acid with the acid modulus (ν(HNO3)/ν(Al2O3)) equal to 0.03.The plastic paste was extruded by VINCHI extruder with a spinneret with trilobe holes of 1.3±0.1 mm.Then, extrudates were dried at 120 °С for 4 h and calcined at 550 °С for 4 h.
The impregnating solution for catalysts preparation was prepared by sequential dissolution of citric acid, nickel carbonate and ammonium heptamolybdate.Then, the support was impregnated with the prepared NiMo solution in a rotary evaporator.After impregnation, the catalyst was dried at 120 °C for 12 h.It should be mentioned that the catalyst preparation technique provides the yields highly-active hydrotreating catalyst.The Ni and Mo contents in the dried catalyst were 3 wt.%and 10 wt.%, respectively.Hereinafter, this sample is referred to as Sup-Cat.
To prepare bulk trimetallic catalyst, the Ni-Mo-W solution with molar Ni:Mo:W ratio 1:0.5:0.5 was used.Nickel hydroxide, ammonium heptamolybdate and ammonium metatugstate were used as precursors of active metals.Citric acid was used as a chelating agent.The obtained solution was spray dried to obtain fine trimetallic powder, which was calcined at 300 °C.Granulated bulk NiMoW precursor preparation was carried out according to the procedure described in [14].The same pseudoboehmite as for the supported catalyst was used as a binding agent for preparing bulk catalyst.The granules of the bulk NiMoW catalyst were calcined at 300 °С.The Ni, Mo and W contents in the dried catalyst were 17, 15 and 28 wt.%, respectively.Hereinafter, this sample is referred to as BulkCat.

Sulfiding and testing of catalysts
Testing conditions in present work fully agree with the ones in [19], which were shown as the best for HDS catalyst testing to minimize the influence of external and internal diffusion.Catalytic experiments were carried out using a fraction of catalysts with a size of 0.1-0.25 mm.Each catalyst was previously sulfided in H2S flow at 220 °C for 2 h and 400 °C for 2 h.Sulfided samples of the catalysts (0.2 g) were mixed with inert SiC to obtain total volume of 4 cm 3 .Then, the samples were loaded into the fixed-bed reactor.The model feedstock contained dibenzothiophene (DBT) and undecane with the sulfur content of 2500 ppm.The testing conditions were as follows: P = 4.0 MPa, H2/feedstock = 300 Nm 3 /m 3 , T = 240, 250, 260 °С, weight hourly space velocity (WHSV) = 80 h -1 .Stationary state at each temperature was achieved after 12 h on-stream.Sampling of liquid products was carried out every 4 h.Liquid products and model feedstock were analyzed for the sulfur content using Xplorer-NS Analyzer (TE Instruments).The DBT conversion (XDBT) was calculated from analysis data using equation (1).Inaccuracy of the method is ±5%.
where CDBT0 and CDBT -initial and residual DBT content in the model feedstock and hydrotreated liquid products, respectively.The calculation of rate constants for DBT hydrodesulfurization was done with the assumption of pseudo first order DBT transformation using equation ( 2).The formula is commonly used for calculating the rate constants of HDS DBT [20].

Textural properties
Textural properties of granulated catalysts were studied by nitrogen adsorption-desorption method using ASAP 2400 (USA) equipment.Each sample was previously maintained for 1 day in N2 flow at 150 °C.Specific surface area (SSA) was calculated by Brunauer-Emmet-Teller method (BET) at relative pressures of 0.05-0.30.The amount of nitrogen adsorbed at a relative pressure close to 1 (p/p0 = 0.995 in practice) was used to calculate total pore volume (Vp, cm 3 /g).Pore size distribution was defined by the Barret-Joyner-Holenda method (BJH) using desorption isotherm.Inaccuracy of the analysis was 5 %.

X-ray photoelectron spectroscopy
The measurements were carried out on an X-ray photoelectron spectrometer SPECS (Berlin, Germany) using nonmonochromatic irradiation M gKα (hν= 1253.6 eV).The binding energy scale (Eb) was pre-calibrated by the position of photoelectron lines of the core levels of gold, silver and copper: Au4f7/2 -84.0 eV, Ag3d5/2 -368.3 eV and Cu2p3/2 -932.7 eV.The samples were applied to double-sided conductive copper tape.The charging effect that occurs during electron photoemission was taken into account using the internal standard method.The method used the Al2p line from aluminum in the support composition (Eb = 74.5 eV) [22].

X-ray diffraction
X-ray diffraction (XRD) studies were carried out using a STOE STADI MP diffractometer (STOE, Germany) equipped with a one-dimensional silicon strip MYTHEN2 1K detector (Dectris AG, Switzerland).The powder XRD patterns were acquired in a transmission geometry using Mo Kα radiation (λ = 0.7093 Å).XRD phase analysis was performed using the ICDD PDF-2 database.

TPD-NH3
Thermoprogrammed desorption of ammonia (NH3 TPD) was made on a ChemBET Pulsar TPR/TPD instrument (Quantachrome Instruments, United States) equipped with a thermal conductivity detector.100 mg of a sample (the 0.25−0.50mm fraction) were loaded to a quartz Ushaped tube.The position of the sample in the tube was fixed with quartz wool.The sample was blown with a 20 cm 3 /min He flow while heating to 550 °C with a rate of 20°/min and kept at this temperature for 30 min to remove moisture.Then, the temperature was lowered to 100 °C and a blank experiment was carried out.For this, the sample was heated to 550 °C with a rate of 10°/min in the 20 cm 3 /min He flow and kept at this temperature for 30 min.Then the sample was cooled to 100 °C and purged with ammonia for 15 min with a rate of 20cm 3 /min.The sample was purged with the 20 cm 3 /min helium flow at 100 °C for 30 min in order to remove physisorbed ammonia.For NH3 TPD, the sample was heated to 550 °C with a rate of 10°/min in the 20 cm 3 /min helium flow and kept at this temperature for 30 min.The amount of desorbed ammonia was calculated by integrati ng the signal area and using instrument calibration with a standard ammonia amount.

HRTEM
The structure and microstructure of the catalysts were studied by high-resolution transmission electron microscopy (HRTEM).For imaging in TEM and STEM modes, a ThemisZ two-corrector transmission electron microscope (USA, Thermo Fisher Scientific) with an accelerating voltage of 200 kV and a limiting resolution of 0.07 nm (TEM) was used.The micrographs were recorded using a Ceta 16 CCD matrix (Thermo Fisher Scientific, USA).

Textural properties
The obvious difference between supported and bulk catalysts lies in their textural properties.Texture of the supported catalyst is always provided by porosity of the support.In most cases, HDS supported catalysts are prepared from gamma-alumina supports, which relate to mesoporous systems.In the studied SupCat sample, alumina is also used as a support.According to the nitrogen adsorption-desorption data, the isotherms of the SupCat sample correspond to IVa type specific for mesoporous materials (Figure 1a) [23].The presence of the hysteresis loop is characteristic for capillary condensation in mesopores.Its type is similar to H1, which indicates the pores of cylindrical shape.The BulkCat samples also contains small amounts of alumina, 30 wt.%, used as a binding agent.However, its content is not so high as to impact the catalysts texture.Nitrogen adsorption-desorption isotherms of the BulkCat sample are also similar to those for the mesoporous materials, but hysteresis loop type is H3, which is specific for slit-like pores (Figure 1a).
Pore size distribution curves of the SupCat and BulkCat samples given in Figure 1b significantly differ from each other.The SupCat sample has monomodal pore size distribution with most of the pores in the range of 5-15 nm.The BulkCat sample contains the peak in the range of <5 nm and the wide pore size distribution in the range of 5-20 nm.In spite of the small amounts of mesopores, we cannot claim the absence of pores that are necessary for catalysis and availability of the active metals for the feedstock molecules.Moreover, we can suggest the presence of macropores due to the incomplete filling of the mesopores seen in the isotherms at high P/P0 values.
Speaking in numerical terms, the SSA value of the Sup-Cat sample (196 m 2 /g) is much higher than that of the BulkCat (122 m 2 /g) (Table 1).Total pore volume and average pore diameter are also higher for the SupCat sample.But it should be emphasized that nitrogen adsorptiondesorption method measures only mesopores from 4 to 50 nm and does not take into account macropores.Therefore, such great difference in texture can be compensated by macroporosity of the bulk sample.

TPD-NH 3
The difference in catalyst activity can be significantly affected by acidity of the catalyst.Considering NiMo(W) catalysts, supported or bulk, the acidity is also provided by active metals, especially by Ni and Mo(W) sulfides or Ni-Mo(W)S phase or their oxide precursors.Alumina support by itself contains preferentially weak and strong Lewis acid sites [24].When active metals are supported on alumina, the acidity increases by 1.5 times [25].However, it was shown in [26] that in case of TPD-NH3, ammonia preferentially interacts with the promoter atom.Then, how different is the total acidity of SupCat and BulkCat is not obvious.
Figure 2 shows TPD-NH3 profiles for the supported and bulk catalysts in the oxide state.Total amount of desorbed ammonia value of the SupCat is 723 µmol/g, while that of the BulkCat -583 µmol/g.The increased amount of desorbed ammonia for the SupCat compared to the BulkCat indirectly indicates an increase in the acidity of the supported catalyst, which is probably due to the predominance of aluminum oxide in the supported catalyst, because aluminum oxide does not play a significant role in target hydrotreating reactions.

X-ray photoelectron spectroscopy
Activity of the catalysts is significantly influenced by active metals state.The active component of hydrotreating catalysts consists of metals sulfides, especially Mo or W, decorated with Co or Ni atoms.To define the state of the active metals, the Mo3d, W4f and Ni2p spectra were deconvoluted to the most possible variants.The Mo3d, W4f, Ni2p and S2p spectra of the studied catalysts are given in Figures 3-4.The spectra of the SupCat were magnified 3 times due to much lower concentration of active metals (Table 2).The Mo3d spectra (Figure 3) contain the peaks with binding energies ~226.2 eV and ~233.2 eV related to sulfur (region S2s), which are specific for sulfur in sulfides (S 2- and S2 2-) and sulfates (S 6+ ), respectively.The Mo part of the spectra is presented as a doublet with the binding energy Mo3d5/2 ~228.9 eV being specified for molybdenum in Mo 4+ state [27][28][29][30][31] in MoS2.According to deconvolution, both samples contain 100% of Mo 4+ .
Figure 3 also demonstrates Ni2p and S2p XPS spectra.There are two peaks with the binding energies ~854.0 and 856.1 eV.The peak with the binding energy ~854.0 eV can be assigned to Ni 2+ in NiMoS phase [32][33][34][35][36] and/or in trimetallic NiMoWS phase in the case of the BulkCat sample [13].The peak with the binding energy 856.2 eV is due to Ni 2+ in NiO.Nickel content in NiMoS composition in the catalysts is 82.9% in BulkCat and 77.3% in SupCat (Table 3).The intensity of S2p spectra of the SupCat is magnified 3 times for clarity.There is a peak with the binding energy S2p3/2~168.4 eV characteristic of S 6+ in sulfates SO4 2- [30].The BulkCat catalysts contain much lower amounts of sulfur compared to the SupCat (Table 3).According to deconvolution data, the fraction of S 6+ in the BulkCat is 2.7% and that in the SupCat is 12.3%.In the range of lower binding energies, there are two doublets with the binding energies S2p3/2 ~161.8 eV and ~163.2 eV corresponding to sulfides S 2- and S2 2-, respectively [37][38][39].
The W4f spectrum of the BulkCat is depicted in Figure 3.The spectrum consists of two doublets -W4f7/2 ~32.5 and 35.9 eV.The peak with the binding energy ~32.5 eV can be assigned to tungsten in W 4+ cation [13,40].Its formation is caused by the transition of WO3 to WS2 after sulfidation [33].The content of tungsten in sulfide form is 75.7%.The peak with the binding energy ~35.9 eV is corresponds to tungsten in W 6+ [41] and is related to tungsten oxide.
The XRD pattern also contains additional strongly widened peaks in the range of 4-8º, 13-17° and 25-27° at 2θ.The reflections in this range of angles are specified for crystal phase of molybdenum sulfide MoS2 (PDF#00-037-1492, a = b = 3.161 Å, c = 12.299 Å, α = β = 90°, γ = 120°).The observed peaks are significantly widened, indicating high dispersity and poor crystallinity of the sulfide phase.The peaks from nickel sulfide were not detected due to overlapping with reflections from alumina and MoS2.Therefore, we made a simulation of the XRD pattern of the BulkCat with Debye method.The methodology for determining crystallite size and amount of stacks in crystallites by modeling is described in [42].The calculations were made in terms of models of MoS2 crystallites with hexagonal shape (Figure 6).The varied characteristic of MoS2 crystallites with plane-like shape were maximal length along the slab (Lmax), slab length along crystallographic direction {110} -D and slab number.The best agreement between the calculations and experimental data was obtained for the XRD pattern of the mixture of onelayered slabs (70%) and double-layered slabs (30%) with lateral sizes Lmax = 2.5 nm and D = 2.0 nm (Figure 6).A number of narrow reflections belong to the crystalline phase of NiS2 (PDF#00-011-099).The determined average size of CSR is D = 25 nm.

HRTEM
The HRTEM data for SupCat and BulkCat samples are shown in Figures 7-8.Differences in the morphology of the particles of the active component certainly exist.The slab length distribution and the stacking number distribution for the supported catalyst were calculated.In this case, the particles of the active component Mo(Ni)S2 are uniformly distributed on the support surface with the stacking number from 1 to 4 and the particle lengths from 1 to 6 nm (Figure 7).In the bulk catalyst, sulfides of active metals are localized in the form of bulk formations due to a small amount of aluminum (Figure 8).Therefore, it is not possible to calculate the stacking number and the particle length.However, we can observe mostly multilayer slabs in bulk part of the catalyst and 1-2 layered slabs over alumina binder.Then, there is the transition of some active metals from its bulk constituent to the alumina binder.

Catalytic properties
The obtained DBT conversion values at three different process temperatures are presented in Table 4.The BulkCat catalyst showed the superior HDS activity at all process temperatures.The greatest difference in activity (11.8%) between the catalysts is observed at 240 °C.As the process temperature increases, the difference decreases and reaches less than 4% at 260 °C.
Then rate constants of HDS DBT were calculated using equation (2).The results are given in Table 4.The bulk catalyst converts DBT much easier than the supported sample without considering the influence of various diffusion restrictions.Summarizing all properties of the BulkCat, the most valuable characteristics are the concentration of active sites and the morphology of active particles.The other properties, which may even be better in SupCat than in BulkCat, are not so significant.

Conclusion
The present report compares properties of bulk and supported hydrotreating catalysts.Both catalysts were prepared in laboratory scale using the procedure that can be easily reproduced in industry.
The difference in physico-chemical properties was shown in terms of textural characteristics, TPD-NH3, XRD, HRTEM and XPS.The supported catalyst has a developed specific surface area (196 m 2 /g) and large pore volume (0.41 cm 3 /g).The bulk catalyst is characterized by 1.5 times lower specific surface area and 4 times lower pore volume.It is caused by small textural values of Ni-MoW precursor of the catalyst and low content of the binding agent (alumina).Then, the diffusion of the feedstock can be restricted in the case of the bulk catalysts.
Based on the results of TPD-NH3 the total amount of desorbed ammonia was higher for the supported catalyst (723 µmol/g) compared to the bulk catalyst (583 µmol/g).
The XRD data of the supported sample contain reflections from MoS2 crystallites and alumina support, while patterns of the bulk catalyst cannot reveal precise structure.Simulation of the XRD patterns allowed us to assume that bulk catalyst contains single-(70%) or doublelayered (30%) particles of MoS2-like particles.According to the results of processing of HRTEM images, in the supported catalyst the fraction of single-layer particles was 60%, bilayer -29%, and the rest of the particles were multylayer.
The XPS data showed that the bulk catalyst contains the highest amounts of active metals in their most active state -Ni in NiMo(W)S phase composition and Mo in Mo 4+ state, as well as that it contains high amounts of W in W 4+ state, which is not presented in the supported catalyst.Due to higher surface concentrations of active metals seen in XPS spectra, it is obvious that the surface concentrations of the most active states are also higher in the bulk catalyst.
It was shown from the calculation of rate constants normalized per gram of catalyst that the bulk catalyst was superior in activity to the supported catalyst at all process temperatures.The greatest difference between the samples was observed at 240 °C (the rate constant of HDS DBT of the bulk catalyst was twice higher than that of the supported catalyst).

Figure 1
Figure 1 Nitrogen adsorption-desorption isotherms (a) and pore size distribution curves (b) of the SupCat and BulkCat catalysts.

Figure 3
Figure 3 Mo3d, S2p and Ni2p XPS spectra of the SupCat and BulkCat samples in sulfide form.

Figure 5
Figure 5 Powder XRD patterns of the SupCat (the position and relative intensity of diffraction lines characteristic of the MoS 2 , and γ-Al 2 O 3 phases are shown) and of the BulkCat sample (in comparison to the calculated XRD pattern for highly-dispersed MoS 2 particles.The position and relative intensity of diffraction lines characteristic of the MoS 2 , WS 2 and NiS 2 phases are shown).The XRD pattern of the BulkCat is shown in Figure 5.There are strongly broadened maxima in the range of 4-8°, 13-22.5°and 25-27° of 2θ.The reflections in these areas correspond to the crystal phases of molybdenum sulfide MoS2 (PDF#00-037-1492, a = b = 3.161 Å, c = 12.299 Å, α = β = 90°, γ = 120°) and tungsten sulfide WS2 (PDF#00-008-0237, a = b = 3.154 Å, c = 12.360 Å, α = β = 90°, γ = 120°).The observed peaks are strongly widened; this indicates high dispersity and poor crystallinity of the sulfide phase, which is similar to the SupCat sample.However, it is not possible to accurately define the composition of the sulfide phase.Therefore, we made a simulation of the XRD pattern of the BulkCat with Debye method.The methodology for determining crystallite size and amount of stacks in crystallites by modeling is described in[42].The calculations were made in terms of models of MoS2 crystallites with hexagonal shape (Figure6).The varied characteristic of MoS2 crystallites with plane-like shape were maximal length along the slab (Lmax), slab length along crystallographic direction {110} -D and slab number.The best agreement between the calculations and experimental data was obtained for the XRD pattern of the mixture of onelayered slabs (70%) and double-layered slabs (30%) with lateral sizes Lmax = 2.5 nm and D = 2.0 nm (Figure6).A number of narrow reflections belong to the crystalline phase of NiS2 (PDF#00-011-099).The determined average size of CSR is D = 25 nm.

Figure 6
Figure 6 Graphic representation of model MoS2 crystallites of hexagonal shape indicating varying size characteristics.Atomic color codes: S: yellow, Mo: dark blue.

Table 1
Textural characteristics of the supported and bulk catalysts.

Table 2
Elemental composition of the surface of studied catalysts (at.%).
Figure 4 W4f XPS spectra of the BulkCat sample in sulfide form.

Table 3
Content of elements in different state (at.%).

Table 4
Data on DBT conversion and rate constant values for the SupCat and BulkCat samples at 240, 250 and 260 °C.