You are on page 1of 75

Adv.

Catal.

47 (2002)

65 ~139

Hydrogen and Synthesis Gas by Steam- and CO* Reforming

and

Steam reforming reactions will play a key role in new applications ofaynthesrs gas and in a future hydrogen economy. The aim of this review is to provide a coherent descriptton of the catalysis of the reforming reacttons. The review is not comprehensive. The tirst section deals with the applications of synthesis gas and hydrogen and the various synthesis

Abbreviations: bpd. barrels per day; ASAXS. Anomalous Small Angle X-ray Scattering; ATR, Autothermal Reforming; CPO. Catalytic Partial Oxidation; CSTR. Continuously Stirred Tank Reactor: DFT, Density Functional Theory: DME. Dimethyl ether: F-T, Fischer-Tropsch Synthesis; GTL. Gas to Liquids: MCFC, Molten Carbonate Fuel Cell; MTBE. Methyl rert-Butyl Ether; MTPD. Metric Ton Per Day; MTG, Mobil Methanol to Gasoline Process; LNG, Liquified Natural Gas; PAFC, Phosphoric Acid Fuel Cell; POX. Partial Oxidation; PSA, Pressure Swmg Adsorption; PEMFC, Polyelectrolytc Membrane Fuel Cell; PROX. Preferential Oxidation; SMR, Steam Methane Reforming: SNG. Substitute Natural Gas: SOFC. Solid Oxide Fuel Cell: SPARG. Sulfur Passivated Reforming; TEM, Transmission Electron Microscopy: TIGAS, Topsoe Integrated Gasoline Synthesis; TPD, Temperature Programmed Desorption: TGA. Thermogravimetric Analysis; XRD. X-ray Diffraction: A,v,, Nickel area: C. Concentration; D,,, Diffusion coefficient for nickel particles: /II,. Surface c,, . Weight of carbon; D. Diameter: diffusion coefbcicnt of nickel atoms; Ex, Excrgy: E. Activation cncrgy; F, Molar flow rate; G. Free energy; H, Enthalpy; K,,, Adsorption constant; K,,, Equilibrium constant; !V(,, Turn-over frequency; Nx. Nickel particle radius: M = (H2 CO~)~(CO+CO2): P. Pressure; S. Entropy; T. Temperature; T(.,&T, Catalyst temperatut-c: T,J,. Carbon limit temperature for methanc decomposition, fixed gas composition, Carbon for T > T,J~; QH~ Reaction quotient; I/. Overall heat transfer coefficient; T,, . Temperature of outer tube wall: TC$,f. Mean catalyst temperature: Z. Axial distance: ( ,,, Heat capacity; (/,,, Particle diameter: ti,, Internal tube diameter: I;, Rate constant; k,., Rate constant for carbon formation; II. Exponent: (1, Heat flux: I, Reactton rate: 1.1. Rate of forward reaction: I~. Reaction rate per reactor volume; t. Time; t,,, Induction period; II,. Superficial veloctty; (I. Kinetic order; /I=QKIK,); /i( :/jz, Energy states; pn, Gas density: ;, Stoichiomctric number; H,. Surface coverage of sulfur relative to saturation.

66

J.R. ROSTRLJP-NIELSEN

et al.

11

gas technologies. The optimum choice of technology depends on the requirements of the gas composition and the scale of operation. Two examples arc included for illustration: synthesis gas for gas-to-liquid plants and hydrogen for fuel cells, The steam reforming process is described in the second section with emphasis on the role of the catalyst and problems related to carbon formation. The third section is a summary of the empirical evidence of the catalysis of the reforming reactions. The methods of characterization are discussed, and data representing sintering, activity trends, and promotion are summarized. The fourth section is a description of the mechanism based on a combination of empirical knowledge with recent data from studies of well-defined surfaces, in-situ high-resolution electron microscopy, and calculations based on density functional theory. The central concept is the role of surface dcfccts as the source of reactivity and the nucleation centers for whisker carbon formation. IC 2002 Elsevier Science.

I. Synthesis A.
APPLICATIONS

Gas

Synthesis gas (syngas), a key intermediate in the chemical industry, is a mixture of hydrogen, carbon monoxide and carbon dioxide, which can be used in a number of highly selective syntheses of a wide range of chemicals and fuels, and as a source of pure hydrogen and carbon monoxide. The present use is primarily for the manufacture of ammonia (in 2000, 120x lo6 ton/y) and of methanol (in 2000, 30 x 1O6ton/y) followed by the use of pure hydrogen for hydrotreating in refineries. Other applications are for synthesis of higher alcohols (by hydroformylation), hydrogenation of unsaturated compounds, and direct reduction of iron ore. Methanol by itself is a precursor for a number of products such as formaldehyde, acetic acid, dimethyl ether (DME), methyl-tert-butyl ether (MTBE), olefins, liquid hydrocarbons, etc. Synthesis gas can be produced from almost any carbon source ranging from natural gas and oil products to coal and biomass. Hence it represents a key for creating flexibility for the chemical industry and for the manufacture of synthetic fuels (synfuels), as became evident during the oil crisis in the 1970s and leading to the notions of Ct -chemistry, synfuels, and synthetic natural gas from coal or remote natural gas and later from renewable carbon sources. It created flexibility in what may be called the synfuel cycle (I), allowing the conversion of one carbon-containing feedstock to another, as illustrated in Fig. 1. Today, natural gas and other hydrocarbons are the dominant feedstocks for production of syngas (and hydrogen) because the investments are about one third of that for a coal-based plant. Natural gas is available in large quantities, but often in remote areas. The options for gas conversion are

11

HYDROGEN

AND

SYNTHESIS

GAS

BY STEAM-

AND

CO?

REFORMING

67

FK,. 1. The synfuel cycle (I ): ATR. autothermal reforming (2): F-T, FischerPTropsch synthesis (3): TIGAS. Topsne integrated gasoline synthesis (4): MTG, Mobil methanol to gasoline process (5): DME, dimethyl ether (6); methanation for SNG (substitute natural gas) (7).

Natural gas

_ \I +

Direct convewm LNG

Assoaated

gas _ I

W
W

Reinjection
(Flaring)

FIG. 2. Options

for conversion

of remote

natural

gas (8)

listed in Fig. 2. The conversion of natural gas into liquid products has been discussed for decades as a solution for validation of stranded natural gas, but in most situations projects involving liquefied natural gas (LNG) have proven to be more feasible. The gas conversion has suffered from the small price differential between product and natural gas. However, a number of factors have changed. It is becoming unacceptable to flare associated gas from oil fields. Moreover, synfuels may have a premium value because they are sulfur-free and because they have high cetane or octane numbers. An important challenge in Cl-chemistry is to circumvent the syngas step by a direct conversion of methane into useful products. Still, yields are far from being economical (Y, IO). The methane molecule is very stable, with a C-H bond energy of 439 kJ/mol; hence methane is resistant to many reactants (I I). Electrophilic attack requires superacidic conditions, and radical

68

J.R. ROSTRUP-NIELSEN

et al.

[I

abstraction of a hydrogen atom by a reactant Q requires that the Q-H bond exceed 439 kJ/mol (I I): CH4 + Q- = CH3- + QH. (1)

This is feasible when Q is an oxidizing agent. However, the product often has much weaker C-H bonds than methane, which implies that it is difficult to eliminate further reaction leading to complete oxidation. For example, the direct conversion of CHd into methanol may have a high selectivity (about 80%), but at a low conversion per pass, corresponding to a yield per pass less than 7% (12). This low conversion per pass results in a large recycle ratio and a difficult separation associated with a low partial pressure of the product. For comparison, commercial methanol production based on synthesis gas is typically characterized by a conversion per pass of 50% and a recycle ratio of four. The selectivity exceeds 99%, implying a yield per pass close to 50% (1.3). Methane is easily activated by Group-8 metals and oxidized to give syngas (14). As a result, large-scale industrial conversion of natural gas into products such as ammonia and methanol takes place via syngas: Methane + steam + air + syngas + products + heat. (2)

In most plants, the heat is utilized for running the plant, requiring that it is available at sufficiently high temperatures to generate high-pressure steam. Present trends in the use of syngas are dominated by the conversion of inexpensive remote natural gas into liquid fuels (gas to liquids, GTL) and by a possible role in a future hydrogen economy (15) possibly associated with the use of fuel cells. These trends imply, on the one hand, the scaleup to large-scale GTL plants (more than 500 000 Nm3 syngas/h) (16) and, on the other hand, the scale-down to small, compact syngas units for fuel cells (5-l 00 Nm syngas(or HZ)/h) (I 7). This expectation creates new challenges for the technology and the catalysis. B. MANUFACTURE OF SYNGAS B. I. $vngus Technologies The synthesis gas routes are highly efficient, but they are capital intensive because they involve exchange of energy in the reformers and heatrecovery units, as illustrated in Fig. 3 (JO,I#). The practical efficiencies are approximately 80% of the ideal, thermodynamically achievable [the ideal efficiency defined as the ratio of the lower heating values of the product to

11

HYDROGEN

AND

SYNTHESIS

GAS

BY STEAM-

AND

CO2

REFORMING

69

Heat input

Heat recovery

Syngas 60
FIG. 3. Indirect conversion of natural gas (IX).

Synthesis 25
Numbers

Separation 15
indicate relative investments

that of the feed (Is)]. The ideal efficiency increases with the H/C ratio of the product (20). Over the years, improvements in the practical efficiencies have been achieved by a combination of chemical engineering and catalyst development (I). Syngas manufacture may be responsible for ca. 60% of the investments of large-scale gas conversion plants based on natural gas (18). Therefore, there is great interest in optimizing process schemes based on steam reforming (SMR) and autothermal reforming (ATR) as well as in exploring new routes for synthesis gas. The most important reactions are listed in Table I. Steam reforming of hydrocarbons is the preferred process for hydrogen and syngas today (18). It results in a H2/CO ratio close to 3. Steam can be replaced by COz, resulting in a H*/CO ratio close to I. The HZ/CO ratio can be varied over a wide range, as illustrated in Fig. 4, as the reforming reactions are coupled to the shift reaction. In the manufacture of hydrogen, the reforming process is followed by water gas shift carried out in the presence of a copper catalyst at low temperatures (210-330C) to ensure complete conversion of carbon monoxide (1.5). The steam and CO* reforming reactions are the main subject of this article. Partial oxidation represents an alternative. It can be carried out in three ways. The non-catalytic partial oxidation (POX) requires high temperature to ensure complete conversion of methane and to reduce soot formation. Some soot is normally formed and is removed in a separate scrubber system downstream of the partial oxidation reactor(21). The thermal processes typically result in a product gas with Hz/CO ratio in the range of 1.7- 1.8. Gasification of heavy oil fractions may play an increasing role as these fractions are becoming more available because of falling demand(22). The autothermal reforming process (2,23,24) is a hybrid of partial oxi-

70

J.R. ROSTRUP-NIELSEN TABLE I Gus Reactions

et al.

5)wthesi.s Process

-AHzgg

(kJ/mol)

Steam

Reforming: -206 $z)Hz -1175 41

I. CH4+H20=C0+3Hz 2. C,H,,,+nHzO=nCO+(n+ 3. CO+H20=C02+Hz CO2 Reforming: + 2H2 (ATR):

4. CH3 + CO2 = 2C0 Autothermal

-247

Reforming

5. CH4+l;02=C0+2H20 6. CH4+H20=C0+3H2 7. CO+H20=C02+H2 Catalytic Partial Oxidation (CPO):

520 -206 41

8. CH4+;02=CO+2H2 For n-C,Hlc,.

38

SMR 2-step reforming ATR

POX

0.0
Natural 0 Naphtha

1.o
Gas

3.0
F

4.0 H&O

5.0 Ratio

FIG. 4. Hz/CO

ratios

from

various

syngas

processes.

dation and steam reforming, using a burner and a fixed catalyst bed for equilibration of the gas. This design allows a decrease in the maximum temperature, and hence the oxygen consumption can be lowered. Soot formation can be eliminated by addition of an appropriate amount of steam

11

t1YDROGEN

AND

SYNTHESIS

GAS

BY STEAM-

AND

cOz

REFORMING

71

to the feedstock and by a specific burner design. Steam can hardly be added to the thermal processes without the risk of increased soot formation because of the resulting lower temperatures. The autothermal reformer is a simple piece of equipment with a specifically designed burner and a fixed catalyst bed in a brick-lined reactor (24). Irrespective of whether the burner is thermal or catalytic or whether a fixed or a fluidized catalyst bed is used. the product gas will be determined by the thermodynamic equilibrium at the exit temperature, which in turn is determined by the adiabatic heat balance. In catalytic partial oxidation (CPO). the reactants are premixed, and all the chemical conversions take place in a catalytic reactor without a burner (25, 26). The direct CPO reaction (Table I) appears to be the ideal solution, as it provides a H2/CO molar ratio of 2 and has a low heat of reaction (38 k.I/mol). However, in practice. the reaction is accompanied by the reforming and water gas shift reactions, and, at high conversions, the product gas will be close to thermodynamic equilibrium (26). Up to 40% of the cost of a syngas plant based on ATR (CPO or POX) is related to the oxygen plant (16.18). Consequently, routes based on air and eliminating the cryogenic air separation plant have been suggested. The use of air in the process stream is possible only in once-through synthesis schemes; otherwise, huge accumulations of nitrogen are unavoidable. Attempts to use air instead of oxygen result in large gas volumes and consequently in large feed/effluent heat exchangers and compressors (27). These are hardly feasible for large-scale plants. The use of air-blown autothermal reforming (or catalytic partial oxidation) is considered for large-scale manufacture of hydrogen combined with CO? sequestration (15). Its implementation will depend strongly on imposed legislation. Cheaper technology for oxygen manufacture may be one route to cost reduction in syngas manufacture. One attempt involves eliminating the oxygen plant and including a reactor concept with oxygen addition through a membrane (28). With the large driving force for transport across the membrane, there is no need for compression of air to the pressure of the process. The reported oxygen ion diffusivities (29) make a syngas unit possible, but the feasibility of this scheme is yet to be demonstrated (10. IN). The choice of syngas technology is dictated by the need for high conversion, the requirements of the syngas composition, and by the scale of operation. B.2. Tl?L~I-nzo~~~nami~,s and $~xgu:cls Compositiorz The conversion of methane is restricted by the thermodynamics of the reforming reactions. The endothermic steam (and CO:) reforming reactions

12

J.R. ROSTRUP-NIELSEN

et al.

60%
60%

II
/ 1

.-

.-----

s/c = 5.0 W-2.5 S/C=l.O

IO Reforming eqdtbrium temperature, C

FIG. 5. Steam

reforming

and methane

conversion

must be carried out at high temperature and low pressure to achieve maximum conversion, as illustrated in Fig. 5. Modern hydrogen plants are normally designed for low steam-to-carbon ratios, although high steam-to-carbon ratios (4-5 molecules of Hz0 per C atom) would result in higher conversion of the hydrocarbons. However, a low steam-to-carbon ratio (typically 2.5 or less) reduces the mass flow through the plant and thus the size of equipment (15.30). The lowest investment is therefore generally obtained for plants designed for low steam-to-carbon ratios. Furthermore, low steam-to-carbon ratios result in a more energyefficient plant and thus a lower operating cost. In principle, a low steam-tocarbon ratio increases the amount of unconverted methane from the reformer (Fig. 5), but this can be compensated for by increasing the reformer outlet temperature, typically to 920C. Unconverted methane is normally removed (and recycled to the reforming unit) in a pressure swing adsorption (PSA) unit, which provides pure hydrogen. In synthesis plants, the unconverted methane flows downstream with the synthesis gas. Unconverted methane thus implies a larger syngas unit and results in restrictions on recycle ratios in the synthesis because of

11

IIYDROGEN

AND

SYNTFIESIS

GAS

BY SThAM-

AND

<Oz

REFORMING

73

Process Ammonia Methanol DME High-tulip. Low-temp. Acetic Higher Industrial Hydrogen Reducing acid alcohols hydrogen for PEMFC ox) Fischer Fischer Tropsch Tropsch

Optimum Hz/N2 = 3

Composition

Co-reactants

methanol olefins

gas (iron

accumulation of the inert methane in the synthesis gas, which reduces the partial pressures of the syngas components. In practice, the content of unreactive components should be less than 2.5 ~0190. Stoichiometric reforming according to Eq. (1) and Eq. (3) of Table I at H20ICHd or C02/CHa ratios of 1 is rarely feasible, because it would result in incomplete conversion at the pressures that are economical for industrial syngas plants (20-40 bar). Other thermodynamic constraints are related to the risk of carbon formation when too little oxidant is present. The composition requirements of the syngas vary with the synthesis in question, as shown in Table II. This point is illustrated by two examples: syngas for large-scale GTL plants and hydrogen for fuel cells. B.3.

Syngus,fi,rGTL Plmnts

The size of the syngas plant is directly related to the carbon efficiency of the synthesis. Unconverted syngas and light synthesis products may be recycled to the syngas unit or used as fuel (16) as illustrated in Fig. 6. Both situations imply a larger syngas unit per ton of product. Hence, it is important that the syngas composition be ad.justed for maximum conversion per pass in the synthesis. The required properties of the syngas are different for different syntheses (I (i,IN). In the synthesis of methanol, dimethyl ether, and gasoline by TIGAS and in the high-temperature Fischer-Tropsch synthesis, CO1 and CO are both reactants linked by the CO-shift reaction. Consequently, the synthesis

J.R. ROSTRUP-NIELSEN
Flue Gas

et al.

Off-gas for Fuel

FIG. 6. Flow

scheme

for natural-gas-based

low-temperature

Fischer-Tropsch

synthesis

(I 6).

gas ideally has a composition corresponding to the stoichiometry of the final product. This is expressed by the ratio A4 = (H2 ~ COz)/(CO + COl), which should be close to 2. The expression for M is derived from simple stoichiometric calculations. In contrast, CO2 is not a reactant in the lowtemperature Fischer-Tropsch synthesis for production of wax and diesel fuel, because the cobalt catalyst is not active for the shift reaction (3) (reaction 3 in Table I), and the optimum syngas Hz/CO ratio is close to the value of 2. Syngas with M = 2 cannot be produced directly by steam reforming of natural gas. However, combined steam and CO1 reforming can provide the appropriate stoichiometry (31): $H4 + $02 + ;H20 = CO + 2H2 = CH30H. (3)

Natural gas and liquid hydrocarbons have lower H/C ratios than CH4, which implies the need for less CO2 to meet the criterion for the ratio A4 = 2. The feasibility of CO2 reforming depends strongly on the price (and pressure) of the CO1 source (32). Many natural gas resources as well as biogas contain large quantities of CO2. Direct production of syngas with A4 = 2 cannot be achieved by autothermal reforming either. It requires adjustment of the gas composition by addition of Hz and/or removal of CO* (18,27). The appropriate value of the ratio A4 (= 2) is obtained by a combination of steam reforming followed by autothermal reforming (two-step reforming). Syngas for low-temperature Fischer-Tropsch synthesis with Hz/CO = 2 can in principle be obtained by combined CO2 and steam reforming, but, as shown in Fig. 7, it results in contents of non-reactive CO* and CHd in the range of about lOvol%, which are unacceptable in the synthesis loop. With autothermal reforming, direct production of a gas with a H>/CO ratio of 2 is achieved only at very low steam-to-carbon ratios (~0.6) and high exit temperatures (16,24). At higher steam-to-carbon ratios, the ratio of 2 is best

11

HYDROGEN

AND

SYNTHESIS

GAS

BY STEAMC02/CH,

AND

(02

REFORMING

75

0.44 0.50 0.57 0.64 0.70 0 I I I I I I I I I I

1.0

1.2

1.4 H,OICH,

1.6

1.8

2.0

FIG. 7. Combined T ex,t = 050"c).

steamiC0~

reforming

for

syngas

with

H2/CO-2

(P=

25 bar absolute,

obtained by partial recycle of CO* or CO?-rich tail gas (Fig. 6). The goal can also be met by combining the partial recycle of CO2 with removal of HI from the syngas, and this option could be an attractive route for producing H2 for the hydrocracking of wax from the Fischer-Tropsch synthesis to manufacture diesel fuel. Operation at low steam-to-carbon ratios is in any case attractive, because it reduces the formation of CO1 and thus the amount of gas to be recycled. The choice of technology for manufacture of syngas depends on the scale of operation (1X). For methanol production, tubular reforming would be the most economical at capacities below 1000-1500 metric tons per day (MTPD), whereas the autothermal reforming is favored at capacities exceeding about 6000 MTPD. The change is explained by the different economies of scale for the tubular reformer and the oxygen plant. For the intermediate range representing the capacity of world-scale methanol plants, two-step reforming is the optimum. Some industrial examples are listed in Table III.
B.4. Hvhgtw

,fiwFtlel

Cells

In most fuel cells, electrochemical oxidation of hydrogen takes place at the anode. In high-temperature fuel cells, it is possible to convert the fuel to hydrogen inside the cell by utilizing the heat from the electrochemical reaction, but otherwise it is necessary to convert the primary fuel outside the stack into a hydrogen-rich gas that is fed to the anode (I 7). The coupling of fuel processing with the fuel cell operation is essential to achieve high plant

16

J.R. ROSTRUP-NIELSEN

et al.

Plant Statoil, Norway NPC, Iran

Capacity methanol: methanol: 2400 t/d 3030tid diesel:

Feed a Associated Natural Associated gas: gas + CO2: gas: 60000 86000 Nm3/h Nm3/h Nm/h

Technology Two-step reforming Ff2O!CO2reforming Autothermal reforming

Escravos. Nigeria Sasol/QP, Quatar

Fischer-Tropsch 3 1 000 barrels/day (z 3900 MTPD) Fischer-Tropsch 34 000 barrels/day (= 4300 MTPD)

2 x 160000

diesel:

Natural

gas:

2x 170000Nm/h

a The volume

is that at 273 K and

I bar abs. TABLE Chaructrri.rtics IV

qfFue1 Cells
H2

Fuel cell type a PEMFC PAFC MCFC SOFC

T,,II

External reformer + +

Direct

fuel conversion

purification 50ppm 0.05% none none CO co CH4, CH4, CO, methanol CO, methanol Acid Fuel Cell; MCFC, methanol?

80 200 600-650 700~1000

a PEMFC, Polyelectrolyte Membrane Fuel Cell; PAFC, Phosphoric Molten Carbonate Fuel Cell; SOFC, Solid Oxide Fuel Cell.

efficiency. It is necessary to integrate the fuel processing system and the fuel cell stack to use the waste heat generated in the fuel cell stack itself. The aim is to reduce the amount of heat generated in the fuel cell power plant, because this can be transferred into electricity only via the Carnot cycle. In the application of a phosphoric acid fuel cell (PAFC) unit, shift reactors are required for conversion of carbon monoxide (to give contents of 0.05% CO or less). The CO purification of carbon monoxide for a polymer electrolyte membrane fuel cell should give a stream with less than 50 ppm CO, and so further clean-up is required. This may be achieved by catalytic preferential oxidation (PROX) or methanation, or by the use of a palladium membrane that allows transport of Hz but almost no CO. The high-temperature fuel cells listed in Table IV (MCFC and SOFC) are

11

HYDROGEN

AND

SYNTHESIS

GAS

BY STEAM-

AND

CO2

REFORMING

77

Combustor

Air

FK;. 8. Solid

oxide

fuel cell and gas turbine

cycle (34).

capable of converting methane, carbon monoxide, and alcohols in the anode chamber by internal reforming, thereby giving higher electric efficiency (33) and allowing the external reformer and the water-gas shift system to be eliminated. Large-scale processing schemes have been proposed (34) in which hightemperature fuel cells with internal reforming are integrated into advanced gas turbine cycles, as illustrated in Fig. 8. The SOFC exhaust gases provide a hydrogen-containing lean fuel gas mixture for the combustion chamber, allowing a more efficient control of the maximum temperature than is achievable by adding surplus air. Such schemes are claimed to have electric efficiencies close to 70%. They require that the fuel cell operate at pressure. The fuel cell is able to achieve higher methane conversions than in the schemes for chemical recuperation in which the heat in the exhaust gas is used just for the reforming reaction (35). The strong endothermicity of the reforming process is also of interest for energy transfer systems based on nuclear or solar energy (35). The choice of technology for small-scale fuel cell units (5-50 kW) is dictated by criteria such as simplicity of design and quick response to transients, and these criteria are important for automotive applications. Although steam reforming may be the most energy-efficient for these applications, catalytic partial oxidation or autothermal reforming with air may be preferred because of the compactness of the design (/ 7). In this review, we focus on the steam and CO1 reforming process, which remains the most widely used technology. Oxygen and air-blown processes may play an increasing role in future applications, but the steam reforming reactions are involved in these processes as well.

78

J.R. ROSTRUP-NIELSEN

et al.

[II

II. The Steam

Reforming

Process

A. THE REFORMER The steam and CO2 reforming reactions are strongly endothermic, as shown in Table I. Nevertheless, the overall heat of reaction may be positive, zero, or negative depending on the processing conditions. At low steam-to-carbon ratios and low temperatures, the steam reforming of higher hydrocarbons (reaction 2 of Table I) followed by the exothermic shift and methanation reactions may result in a methane-rich product gas and a slightly positive overall heat of reaction (14). The manufacture of syngas requires a high temperature to achieve a high methane conversion (Fig. 5) and the overall heat of reaction becomes strongly negative, which implies that heat must be supplied to the reaction. For example, adiabatic steam reforming of methane (carried out with H20ICH4 = 2.5 at a pressure of 20 bar absolute and a temperature of 500C) will result in a temperature drop of approximately 12C for each 1% of methane converted. In industry, the reforming reactions are typically carried out in a heated furnace in the presence of a nickel catalyst (14,36,30). The catalyst is placed in a number of high-alloy reforming tubes placed in a row along the furnace. The outer diameter of the tubes ranges typically from 100 to 150 mm and the length from 10 to 13 m. Typical inlet temperatures to the catalyst bed are 450~650C and product gas leaves the reformer at 800-950C depending on the application. Tubular reformers are designed with a variety of tube and burner arrangements (36). Such reformers are built today for capacities up to 300000Nm3 of Hz(or syngas)/h. The furnace consists of a box-type radiant section including the burners and a convection section to recover the waste heat of the flue gases leaving the radiant section, as illustrated in Fig. 9.
CO2(optional) I

FIG.

9. Flow diagram of process with tubular reformer and prereformer (37)

111

HYDROGEN ANDSYNTHESISGAS

BY STEAM-AND('02

REFORMING

79

In recent years there has been progress in steam reforming technology resulting in less costly plants, in part because of better materials for reformer tubes, better control of carbon limits, and better catalysts and process concepts with high feedstock flexibility (8). This progress has been accompanied by a better understanding of the reaction mechanism, the mechanisms of carbon formation and sulfur poisoning, and the reasons for tube failure (38). The tubular reformer is an expensive piece of equipment, and there have been strong efforts to reduce the size by improving the heat transfer and hence reducing the number of tubes. It has been normal practice to use the average heat flux as a measure of operating severity in reformers, but it appears that the most critical parameter is the maximum temperature difference across the tube wall. Tubular reformers today are designed for operation at average heat fluxes exceeding 100 000 kcal/m*/h (0.12 M W/m), almost two times higher than what was industrial practice 20 years ago. The thermal efficiency of the tubular reformer and waste heat recovery section approaches 95% (24,36), as most of the heat that is not transferred to the process (ca. 50%) is recovered from the flue gas. This heat is used for steam production and for preheating of the reformer feed, combustion air, etc. The same is true for the heat contained in the hot product gas exiting the reformer. The amount of heat transferred in the tubular reformer may be reduced by increasing the preheat temperature, but the preheater may then produce olefins from higher hydrocarbons in the feed, which easily form carbon in the reformer. Apart from the pressure, the conditions in the tubular steam reformer and in the preheater are not far from those of a steam cracker in an ethylene plant. The problem of carbon deposition has been solved by introduction of an adiabatic prereformer, as illustrated in Fig. 9 (37,39). All higher hydrocarbons are converted in the prereformer in the temperature range of 350-55OC, and the methane reforming and shift reactions are brought into equilibrium. With the use of a prereformer, it is possible to preheat the feed to the tubular reformer to temperatures around 65OC, thus reducing the size of the tubular reformer. The prereformer may also allow for feedstock flexibility ranging from natural gas and refinery off-gas to liquid fuels (40). The high efficiency of the tubular reformer is paid for by a large exchange of energy through the steam reformer and heat-recovery units with large heatexchange surfaces. This point is evident from an exergy analysis. Table V shows the principal main exergy losses from an ammonia plant compared with the enthalpy changes (41). Generally, the figures are comparable to each other, but the reformer shows a large loss of exergy because of creation of hot flue gas. This is compensated for by very low losses in the turbines and compressors using the high-pressure steam recovered from the waste heat. An ammonia plant is optimized for maximum use of the waste heat, but

80

J.R. ROSTRUP-NIELSEN

et al.

Enthalpy

and Exergv

TABLE V Analysis of an Ammonia Enthalpy (W/NH? liq)

Plant (41) a Exergy (kJ/NHj 30.7 h liq)

Total Losses:

consumption

(feed + fuel)

29.4

Reforming Steam generation

0.4 0.3 6.5 9.9 17.1 27.6 (%) 58

4.9 2.4 0.5 2.3 20.1 32.4 66

Turbines/compressors Other Energy Total process in NH3 losses steps product

Efficiency

a The exergy loss is an expression of the loss of energy (useful work) through irrcversibilities in the process. The wasted energy (heat) can be transferred into work only via the Carnot cycle. The cnthalpy losses (expressed as LHV) reflect the total energy balance. The exegy is AEw = AH ~ AFT (surroundings).

this is linked to the application of steam turbines and based on the assumption of autonomy of the plant. For many applications, such as, for example, the manufacture of hydrogen, there is no need for the energy produced. It is possible to increase the amount of heat transferred to the process gas in the reformer from about 50% to about 80% of the supplied heat when using a convective heat-exchange reformer (8,36,42), in which the flue gas as well as the hot product gas are cooled by heat exchange, with the process gas flowing through the catalyst bed. In all types of heat-exchange reformers, the heat exchange is by convection, and this generally leads to lower heat fluxes than in reformers with radiant heat transfer. B. CONSTRAINTS OF THE REFORMING PROUS The steam reforming process as practiced today faces a number of constraints (42,43). First, thermodynamics demands high exit temperatures to achieve high conversions of methane. In contrast, the catalysts are potentially active even at temperatures below 400C (Fig. 10). Consequently, there have been efforts to circumvent the constraints by the use of a hydrogen-selective membrane installed in the catalyst bed (44,4.5). Hydrogen is continuously re-

111

HYDROGEN

AND

SYNTHESIS I 20

GAS

BY STEAM-

AND

CO2

REFORMING

XI

o Ni - catalyst

cm)

Temperature (C)
FK;. IO. Steam P = I bar. reforming at low tcmpcratures (43). Conditions: HzO/CHd =4. t120iHl ~ 10.

moved from the reactant stream, thereby driving the equilibrium toward higher conversion at lower temperature. Reactor simulations and experiments (46) have shown that the reformer exit temperature can indeed be reduced to below 700C while the same conversion is maintained. However, the hydrogen produced according to this concept is at low pressure and must be compressed to the usual delivery pressure of 20 bar. This limitation renders the process uneconomical except when very low electricity prices prevail (1.5), or when hydrogen is used as a feedstock for a fuel cell or as a low-pressure fuel. A low-temperature membrane reformer would be able to use low-temperature heat. which would imply a lower loss of exergy. A similar effect is obtained in fuel cells with internal reforming (33). The heat for the reforming reaction is created by the electrochemical reaction and, hence, there is no need for heat created at a temperature above the working temperature of the fuel cell. In a reformer, the tube diameter is selected on the basis of mechanical considerations and the heat flux on the basis of materials considerations or restrictions in convective heat transfer, leaving the space velocity (catalyst volume) fixed with the corresponding low utilization of the catalyst (43,47). Several schemes have been suggested for circumventing the tubular reactor concept. These include reheat schemes (43) in which the process gas is heated in a heater followed by reforming in an adiabatic reactor. Many steps arc required to reheat the gas because of the strong endothermicity of the reaction. A variation of the reheat process scheme is the use of a circulating catalyst, with one bed for reaction and the other for heating the catalyst. This idea is also applied in other fluidized-bed petrochemical processes. However, for steam reforming, the recirculation rate would have to be very high. Moreover. catalyst dust in downstream heat exchangers would result in methanation (48).

82

J.R. ROSTRUP-NIELSEN

et al.

[II

Other attempts (43) have aimed at taking advantage of the high heat-transfer rates in fluidized beds or the use of heat pipes (49). C.
CATALYST AND REFORMER PERFORMANCE

The steam reforming catalyst normally contains nickel. Cobalt and the noble Group-8 metals are also active, but more expensive. The catalyst properties are dictated by the severe operating conditions, including temperatures of 450-950C and steam partial pressures of up to 30 bar. The catalytic activity depends on the nickel surface area. The activity of the catalyst is rarely a limiting factor. Most industrial catalysts have high activity for the reforming reactions. As shown in Fig. 10, nickel and ruthenium catalysts may be able to convert methane even at 300C (43). A typical industrial catalyst is characterized by a turnover frequency of -0.5 ss at 450C (under the following conditions: H20/CH4 = 4, HzO/H* = 10, pressure = 1 bar), corresponding to a conversion of 10% at a space velocity of 80 000 vol CH~/vol of catalyst/h. As illustrated in Fig. 11, a consequence is that equilibrium conversion at high reforming temperatures is achieved at very high space velocities (exceeding 1O6vol CHdivol catalyst/h), as determined by extrapolation of the intrinsic rates. In practice, however, the utilization of the intrinsic catalytic activity (as expressed by the effectiveness factor) is less than 10% because of transport restrictions (14,5(I). The effectiveness factor is so low that the activity is roughly proportional to the external surface area of the catalyst. The shape of the catalyst pellet should be optimized to achieve maximum activity with

1 ,E+03

1 ,E+04

1 ,E+05 Space velocity

1,E+06 Nm3/m3 h

1 ,E+07

1,E+08

FIG. I I. Methane conversions determined based on reaction kinetics of Xu and Froment

from intrinsic (51).

rates (H20/CH4

=2.5).

Simulation

minimum

pressure drop. The pressure drop depends strongly on the void

111

HYDROGEN

AND

SYNTHESIS

GAS

BY STEAM-

AND

CO2

REFORMING

83

*t

63 t
6

100000

200000

300000

Average

heat

flux q, Kcal/mVh based on the assumptions of constant

FIG. 12. Reformer catalyst performance. Simulation catalyst volume, outlet tcmpcraturc, and outlet pressure.

fraction of the packed bed and decreases with increasing particle size. Hence, the optimum is a catalyst bed of pellets with large external diameter and with high void fraction, as achieved with rings or cylinders with several holes. Other solutions may be based on the use of catalysts consisting of ceramic foams, monoliths ~ and even the hardware might be made to be catalytically active (52,53). Industrial tubular reformers operate at space velocities of 20004000 h- , dictated by the heat transfer and the tube design. Simulations show that the catalyst is not the limiting factor for the operation of a tubular reformer. An increase of the heat flux and the load at a given exit temperature by a factor of two results in an increase in methane leakage by only 10% (24) as illustrated in Fig. 12. Although there is a huge surplus of activity for conversion of hydrocarbons into equilibrated gas, high catalytic activity is nevertheless important in the complex coupling of catalytic reaction and heat transfer. This is expressed by the simple equation (14) Transferred heat = 1 U( Tb,, - Tc.,,,,) = reaction heat + AH I; + sensible heat. up<& j I (4)

For given heat flux or temperature driving force (Tu, ~ TcM), a higher catalytic activity, I+, can be used to absorb the same reaction heat (AH) at a lower catalyst temperature rc, and consequently at a lower tube wall temperature, Tcr. The result could be a substantially longer tube life. Equation (4) is based on a one-dimensional reactor model. A twodimensional reactor model is required to simulate the strong radial gradients in the reformer (14,50). Figure 13 shows the characteristic temperature,

84

J.R. ROSTRLJP-NIELSEN

et al.

Inlet FIG. 13. Typical profiles

Tube length in a tubular steam reformer(42)

ia I% Outlet

1000.

Outertubewall

. . . . -mm . . . . .a*

80

Catalyst

temperature
-40 $2

500 400 0 I 2 I 4 I 6

. +. --aI 6 I 10 ' 12 0 -10

Axial distance

(m) (H20ICH4 x2.5, model (53) and with equal firing

FIG. 14. Typical profiles in tubular reformer: hydrogen plant with prereformer P = 30 bar). Simulation based on a two-dimensional heterogeneous reactor reaction kinetics modified from Xu and Froment (51). Side wall fired reformer on all burners.

conversion, reaction rate, and catalyst effectiveness factor profiles along the tube (42). Figure 14 shows data from a simulation of a tubular reformer by means of a two-dimensional heterogeneous reactor model; the furnace chamber is also modeled(H). Most of the catalyst volume will work close to equilibrium, as expressed by the term /? (= Q&C,). The rates of reversible reactions are usually expressed

by

Y = q( 1 - P).

(5)

111

HYDROGEN

AND

SYNTHESIS

GAS

BY

STEAM-

AND

C02

REFORMING

85

As illustrated in Fig. 13, the value of /I is greater than 0.7 in most of the reformer tubes. The methane content decreases with increasing temperature. The driving force is the change of methane concentration at equilibrium with temperature. It can be shown (14,5.5) that d(Ccrr,/~fZ) is close to d(Cc,u,(eq))/dZ. This is also indicated in Fig. 13 for high temperatures. Hence, I^ = k(&, - CcuI(eq)) = F

d(G,(eq)) = Fd(Gpj4(eq))dT dZ
dT

dZ

(6)

which implies that the rate becomes proportional to the slope (dT/dZ) of the imposed temperature profile (14.55). In an adiabatic prereformer, dT/dZ approaches zero as the reaction approaches equilibrium. This point is reflected in the increase in the value of/j through the reactor. The value approaches 1 (55). In Eq. (5) y is the stoichiometric number for the rate-determining step. This was found to be close to unity for steam reforming of methane (56). Although Eq. (5) satisfies the thermodynamic requirements, it is in principle not clear that an expression such as Eq. (5) should adequately account for the observed rates close to equilibrium (55). Close to equilibrium, the rate of reaction is first order with respect to any parameter that expresses the distance from equilibrium. Most kinetics investigations deal with initial rates. far from equilibrium, and the data may reflect the situation in only a small part of an industrial reactor. D. SULFUR POISONING The Group-S metal catalysts are highly susceptible to sulfur poisoning (14, 57,58), and it is therefore important that any sulfur in the feed gas be removed. The desulfurization of natural gas is easily accomplished by reaction with zinc oxide. Liquid hydrocarbons require hydrodesulfurization with CoMo catalysts (59). Under the reforming conditions, all sulfur compounds will be converted into hydrogen sulfide, which is chemisorbed on transitionmetal surfaces: H:S + Me = Me-S + HI. (7) This reaction takes place at HzS/Hz ratios less than those required for the formation of bulk sulfides. The surface layer has a well-defined structure like that of a two-dimensional sulfide (57). Nickel is more sensitive to sulfide formation than other Group-S metals. For nickel at 500C H, = 0.5 corresponds to the ratio of partial pressures HzS/Hz = 1.6x I O-l2 (I 4). A consequence is that sulfur is quantitatively withheld by the nickel until saturation

86

J.R.ROSTRUP-NIELSEN et al.

[II

and that there is essentially no sulfur content of the feed below which stable operation of the catalyst can be achieved. It is evident that a nickelcontaining prereformer catalyst will remove essentially any trace of sulfur, thus protecting the tubular reforming catalyst and other downstream catalysts. During the operation of an adiabatic prereformer, progressive deactivation takes place, mainly by sulfur poisoning. The sulfur/nickel system is discussed in detail in Section 1II.B. In principle, it is possible to regenerate the poisoned catalyst by treatment with hydrogen [the reverse of reaction (7)], but the driving force is extremely small. Sulfur may be removed by oxidation and controlled re-reduction of the catalyst (14).

E. STEAMREFORMING OF LIQUID HYDROCARBONS In many situations when natural gas is not available, higher hydrocarbons become the preferred feedstock for the reforming process (40). Many refineries benefit from flexibility in feedstock, taking advantage of the surplus of various hydrocarbon streams in the refinery. Steam reforming of liquid hydrocarbons is also considered for hydrogen generation for fuel cells, with diesel and jet fuel considered as logistic fuels (15,Z7,60). As shown in Table I, the required heat per carbon atom is less for n-heptane than for methane. Consequently, the heat input to the tubular reformer will be slightly less when naphtha is used instead of natural gas under similar operating conditions. The higher hydrocarbons are also more reactive than methane; aromatics show the lowest reactivities, approaching that of methane (14). The conversion of higher hydrocarbons on nickel takes place by an irreversible adsorption on the surface, with only C i -components leaving the surface (14). With proper desulfurization, it has been possible to convert light gas oils and diesel into syngas with no trace of higher hydrocarbons in the product gas (38). These reforming reactions may be accompanied by thermal cracking (pyrolysis) of the hydrocarbons at temperatures exceeding 600-650C (40). The pyrolysis of hydrocarbons follows the thermal cracking mechanism (61). Apart from the pressure, the conditions in the tubular steam reformer and in the preheater are not far from those of a steam cracker in an ethylene plant. When the activity of the catalyst is low, the pyrolysis route may take over. This is the situation in the case of severe sulfur poisoning or in attempts to use non-metal catalysts, which so far have shown very low activities (14).

111

tIYDROGEN

AND

SYNTHESIS

GAS CARBON

BY STEAMFCIRMATKIN

AND

COz

REFORMING

X7

F. F. I. Vclviocrs Routes

Steam reforming generally involves the risk of carbon formation. The main reactions are listed in Table VI. Carbon formation may take place mainly by three routes (14), summarized in Table VII. At low temperatures, adsorbed hydrocarbons may accumulate on the nickel surface and slowly be transformed into a polymer film (gum) blocking the nickel surface. Figure 15 shows an example from a prereformer operating on heavy feedstock (60). With kerosene (jet fuel), the catalyst is gradually poisoned by sulfur, resulting in a shifting of the temperature profile. When the process is operated with gas oil (diesel), the sulfur poisoning is accompanied by poisoning resulting from gum formation. The formation of gum can be retarded by hydrogen (14). At high temperatures, ethylene from the pyrolysis of higher hydrocarbons may lead to pyrolytic coke, which may encapsulate the catalyst pellets. Whisker carbon is the principal product of carbon formation in steam reforming (14). The whiskers typically grow with a nickel crystal at the top. The whiskers are formed as follows: adsorbed hydrocarbon or carbon monoxide dissociates on the metal surface to give adsorbed carbon atoms, which are dissolved in the metal particle. Carbon diffuses through the particle and nucleates into the fiber at the rear interface. The nickel crystal changes shape into a pear-like particle, leaving small fragments of nickel behind in the whisker. The carbon whiskers have high mechanical strength, and the catalyst particle is destroyed when the whiskers hit the pore walls. This process may result in increasing pressure drop and hot tubes, which impedes the operation.

Reaction

9. 10. II. 12. 13. 14.

CHA=C+2H2 2co-(+CO~ CO+HZpC+HzO C,,H,,, =nC+mi2 C,,H,,, C,,H,, 2 olcfins =((Hz),, Hz = coke -gum

75 172 131 -1X8

For /K7HIC,.

88

J.R. ROSTRUP-NIELSEN TABLE VII

et al

Routes to Curhor~
Carbon type Reaction (Table VI) 14 Phenomena Critical parameters

Gum

Blocking

of Ni surface

low HIOIC ratio. absence low temperature, presence of aromatics

of Hz.

Whisker Pyrolytic

carbon coke

9-12 13

Breakup

of catalyst

pellet

low HzO/C ratio, high temperature, presence of olefins, aromatics high temperature, residence presence of olefins, sulfur poisoning time,

Encapsulation of catalyst pellet, deposits on tube wall

520 I-

I
NV

Jet-Fuel . . . . . Jet-Fuel

- 38h - 303h 1 .o

0 Relative

0.5 Axial Distance


Temperature profiles

FIG. 15. Adiabatic catalyst (60).

prereforming

of logistic

fuels.

with

Topsne

RKNGR

Nickel carbide is not stable nucleation of the carbon whisker reflected by the kinetics (14). After rate: dC, ~ dt

under steam reforming conditions. The takes place after an induction period (to), nucleation, the carbon grows at a constant = kL.(t ~ to). (8)

The rate of dissociation depends strongly on the type of hydrocarbon, with olefins and acetylene being the most reactive (14,62). The growth mechanism appears to be the same, irrespective of the type of hydrocarbon or whether it results from the endothermic dissociation of

111

HYDROGEN

AND

SYNTHESIS

GAS

BY STEAM-

AND

CO2

REFORMlNG

89

FK;. 16. In-situ high resolution transmission electron The image indicates whisker carbon formation on a step. of 40 s with the formation of 3 layers/s. It is evident that the step to the right rather than being formed uniformly

micrograph (resolution: 0.34 mn (65)). The growth was followed over a period the graphite layers were created from on the nickel surface.

methane or the exothermic dissociation of carbon monoxide (14). However, the resulting morphology and degree of graphitization depend on parameters such as type of hydrocarbon, metal, particle size, and temperature. Hence, there might not be a unique growth mechanism for the formation of carbon fibers and nanotubes (63). The importance of step sites for the nucleation of carbon was observed in early studies (64) and confirmed in recent in-situ investigations by highresolution transmission electron microscopy that indicate the segregation of carbon when formation of the whiskers takes place at specific sites on the nickel crystal (Fig. 16) (65). The mechanism of formation of the carbon is discussed in detail in Section NC. Undergoing rapid dissociation from olefins and acetylene, carbon diffuses through the nickel crystal, and the diffusion has been suggested to be rate determining, as the activation energy is close to that for the diffusion of carbon through nickel (14,62). In contrast, with methane, the dissociation of the molecule becomes the rate-determining step (66,67). The addition of potassium results in a significant increase of the induction time for methane decomposition, as shown in Fig. 17 (ha), which implies a retarding effect of potassium on the dissociation of methane and the nucleation of carbon (Section ND). The nickel particle size has an impact on the nucleation of carbon. The smaller the crystals, the more difficult is the initiation of carbon formation. This result was demonstrated in TGA experiments with two catalysts having the same activity but different dispersions of nickel (69). The catalysts were heated at a fixed rate, and as shown in Fig. 18, the onset temperature was approximately 100C higher for the catalyst with small nickel crystals (-7 nm)

90

J.R. ROSTRUP-NIELSEN

et al.

-0.9

-0.7 wpH*2RJiJ

-0.5

-0.3

FIG. 17. Decomposition potassium on the induction

of methane on supported nickel .. time. The units of tu are hours

at 500C

and 1 bar: the influence

of

0.92% NilMgA1204 dNi=7nm . -I

,.

99
300

I
350

I
400

I
450 Temperature

I
500 (C)

I
550

I
600

Fro. 18. Results of TGA studies (H20/C=0.7). Relative weight increase and temperature for two catalysts with different nickel crystal sizes and same nickel surface area and reforming activity (hydrocarbon feed: 3% nCdHtu, 7/0 Hz, 90% He), H20/C ~0.7, I bar abs.

than for that with large crystals (- 100 nm). This result is discussed further in Section NC. The whisker carbon has a higher energy than graphite, which is reflected in lower equilibrium constants for the reversible decomposition reaction of methane and carbon monoxide (14,7(I), as shown for methane decomposition in Fig. 19. It was found (14) that the deviation from graphite thermodynamics depends on the nickel particle size and that the deviation could be explained by the extra energy required by the higher surface energy, the elastic energy, and the defect structure of the carbon filaments (71). The rate of carbon formation was found to be far less on noble metals

111

HYDROGEN

AND

SYNTHESIS

GAS -

BY STEAMT(C)
600

AND

(02

REFORMING

9I

900 800

700

500

IO 45 y"

0
A x . 0.8

90

-x

75 50 20 1 .o 103/r(K-'1 1.2

FIG. 19. Decomposition for NiBi and NiCuiSiOl equilibrium constant (711).

of methane on supported nickel catalysts. Equilibrium constants catalysts. The copper content of the catalyst does not affect the

than on nickel (72). This result has been explained by the fact that the noble metals do not dissolve carbon (73). The carbon formed on the noble metals was observed to be of a structure that was difficult to distinguish from the catalyst structure. On a ruthenium catalyst, high-resolution electron microscopy revealed a structure which looked like a few atomic layers of carbon covering most of the surface (72). The whisker growth mechanism is also blocked by sulfur poisoning of the nickel surface. When formed, the carbon has a typical octopus structure with several fibers growing from one nickel crystal (74). A similar structure is formed (70) on Ni-Cu catalysts with low nickel contents (20 wt%). F .2. Curhon Limits The formation of whisker carbon cannot be tolerated in a tubular reformer. The important problem is whether or not carbon is formed, and not the rate at which it may be formed. In terms of the growth mechanism, this means that the induction period (to in Eq. 8) should be extended to infinity. This is achieved by keeping the steady-state activity of carbon smaller than unity (14). (This is discussed in Section 1II.E.) The carbon formation depends on the kinetic balance between the surface reaction of

92

J.R. ROSTRUP-NIELSEN

et al.

<
C

More B

critical AA

conditions

C - &In&y equilibra;ted

in gas

No C - affinity

in actual

gas

FIG. 20. Carbon limits in steam reforming (75): A: no potential for carbon in actual gas; A: critical steam to carbon ratio (carbon formation on Ni); B: thermodynamic limit potential for carbon in equilibrated gas; C: carbon limit with ensemble control or noble metals.

the adsorbed hydrocarbon with oxygen species and the further dissociation of the hydrocarbon into adsorbed carbon atoms, which can be dissolved in the nickel crystal. Higher hydrocarbons show a greater tendency for carbon formation on nickel than methane. The risk for carbon formation depends on the type of hydrocarbon (Id), with the unsaturated character of the hydrocarbon being critical. Liquid fuels containing high contents of aromatics are more difficult to reform than paraffinic feedstocks. Ethylene formed by pyrolysis results in rapid carbon formation on nickel. Naphtha can be processed directly in the tubular reformer when using special, promoted catalysts (14). This conversion is practiced in many industrial units, but control of the preheat temperature and heat-flux profile may be critical. These constraints are removed when a prereformer is used (37,39), as illustrated in Fig. 9. At a given temperature and for a given hydrocarbon feed, carbon will be formed below a critical steam-to-carbon ratio (Id), indicated by the carbon limit A in Fig. 20 (75). This critical steam-to-carbon ratio increases with temperature. By promotion of the catalyst, it is possible to push this limit to the thermodynamic carbon limit B, reflecting the following principle of equilibrated gas (14):
Carbon
affinity for

formation
carbon

is to be expected
after the establishment

on a nickel
of the

catalyst if the gas shows


methane reforming and the

shift equilibria.

This principle is justified by the low effectiveness factor of the catalyst for the reforming reaction, which implies that the gas inside most of the catalyst particle is nearly at thermodynamic equilibrium.

III

IIYDROGEN

AND

SYNTHESIS

GAS

BY

STEAM-

AND

cOz

REFORMING

93

Tests I a Full-size Monotube Reformer k Nicat. B, NI cat. C Noble mat. cat. 0 SPARG - Eq carbon limit /

1 o/c
FIG. 21. Carbon limit diagram (43); P=25.5 bar; the whisker cat, catalyst: carbon

data were

obtained

for

nickel particles with diameter The ratios arc molar.

250nm.

Abbreviations:

met, metal.:

Ey., equilibrium.

The thermodynamic carbon limit B is a function of the composition of the feed gas (atomic ratios O/C and H/C) and total pressure. An example of thermodynamic limits (43) is given in the diagram in Fig. 21. It was shown (63) that the thermodynamic limit (Carbon Limit B) could be shifted by using a catalyst with smaller nickel crystals, thus allowing operation under conditions not otherwise possible. By using noble metals and sulfur passivation (72,74), it is possible to push the carbon limit beyond limit B to carbon limit C. For practical design, a conservative guideline for carbon-free operation would be to require that at no position in the reactor would there be a thermodynamic potential for carbon formation (14,5(j), corresponding to carbon limit A in Fig. 20. This approach does not apply for the steam reforming of higher hydrocarbons, because the decomposition into carbon is irreversible. It means that it is important to control the parameters influencing

94

J.R. ROSTRUP-NIELSEN

et al.

3 M from Inlet

700 690 0 0.25 0.5 0.75 1.0

Relative

Distance

from

Tube

Axis

FIG. 22. Steam reforming of methane (H20/CH4 = 3.5, P = 33 bar): radial gradients and carbon limits, 3 meters from tube inlet. TM is the carbon limit temperature for methane decomposition. There is potential for carbon formation when the catalyst temperature, TCAT z TM. For the high-activity catalyst, TCAr remains below TM, i.e., there is no potential for carbon formation. For the low-activity catalyst, the mean catalyst temperature (not shown) is below TM, but the actual catalyst temperature, TCAT, becomes higher than TM close to the tube wall. indicating the risk for carbon formation (14).

carbon limit A. For practical design, it means that the actual steam-to-higherhydrocarbon ratio should exceed the critical ratio. The analysis of the risk for carbon formation should take account of the strong radial temperature gradients in the reformer tube (14,5(I). As there are almost no radial concentration gradients, the situation is equivalent to heating a gas with fixed composition to a temperature close to that at the tube wall. This means a higher potential for the endothermic decomposition of methane and a higher critical steam-to-carbon ratio for higher hydrocarbons when approaching the tube wall as compared to the values based on the average gas temperature. The potential for carbon formation can be analyzed in detail by means of a two-dimensional reactor model as illustrated in Fig. 22. The carbon limits A and B can be neglected in sulfur-passivated reforming of natural gas, as practiced in the SPARG process (76). This means operating to the left of the carbon limit curve in Fig. 21. This operation is accomplished by ensemble control, which means that the surface sites for carbon formation are blocked, although sufficient sites for the reforming are maintained (74). The effect is obtained by adding sulfur to the process feed to give a coverage, H,, exceeding about 0.7. The retarding effect of sulfur is a dynamic phenomenon. This means that carbon may be formed under certain conditions in spite of sulfur passivation ~ although at markedly reduced rates and by a mechanism different from that

111

HYDROGEN

AND

SYNTHESIS 10.000

GAS

BY STEAM-

AND

CO?

REFORMING

95

3 z .I z i3 5 z E g f

1000

r SP
100 0

;B
rsp/U 10 SP \,A 300 1
0.8 1 .o 1.2 700 500

-%I

o 400C
1.4 103/TK-

FIG. 23. Sulfur passivation and specific activity (74) in reforming at I bar absolute pressure with a Ni/Al~O~ catalyst. Sulfur-passivatcd catalyst: H*O/CHd : I, HlO/Hl = 5. HlSiH2 = activity of 2.8x IO-. Sulfur-free catalyst: H~O/CHJ = 0.94, II~O& ~2.5; #sp IS the specific non-poisoned catalyst; rzp is the specific activity.

leading to the formation of whisker carbon (74). The thermodynamic potential (-30 kJ/mol) required to initiate carbon formation was found to depend on the sulfur coverage (14). This energy is much higher than that (-2 kJ/mol) required to initiate carbon formation on sulfur-free catalysts. The rate of carbon formation decreases more with sulfur coverage (74) than does the reforming rate. In a simple Maxted model, this reflects the fact that the ensemble of surface atoms needed for the reforming reaction is smaller than that required for the nucleation of the carbon whisker (77). This model also explains the increase of activation energy and change in reaction orders for the reforming reaction, as illustrated in Fig. 23 (14). It was possible to operate at heat fluxes close to 80000 kcal/m2/h in spite of the sulfur passivation (78). The catalyst temperature increases quickly to more than 750-8OOC, whereupon the reforming rates become sufficient for conversion of methane even at high heat flux (space velocity). The high reaction temperature implies that higher hydrocarbons must be converted in an adiabatic prereformer to eliminate the risk of carbon formation by thermal cracking. The ensemble control allows operation at (Hz0 + CO2)/CH4 ratios close to the stoichiometric, as illustrated in Fig. 2 1. Operation with mixtures of CO1 and methane without steam is also possible, except for the steam required

96

J.R. ROSTRUP-NIELSEN

et al.

[III

for prereforming of the higher hydrocarbons in the feed. Similar results were obtained when a supported ruthenium catalyst was used(4.3). F.3. Metal Dusting The decomposition of carbon monoxide (reactions (10) and (11) of Table VI) may take place without catalyst on the surfaces of the equipment (e.g., heat exchangers). This may also lead to metal-dusting corrosion, in which reaction ( 11) of Table VI appears to be involved (79). For a given gas composition and pressure, there is a temperature below which there is a tendency for CO decomposition and initiation of metal dusting. However, the limits for metal-dusting corrosion are dictated by several parameters and in a complex way. The risk of corrosion is strongly influenced by the selection of construction materials and passivation methods. The sulfur present in the SPARG product gas inhibits carbon formation and metal-dusting corrosion in the outlet system, even when the operating conditions predict that carbon formation is possible. This is not surprising because a major process in metal dusting (79) is the formation of carbon on the free Fe-Ni surface of the construction material. The mechanism is equivalent to that described for formation of whisker carbon on the reforming catalyst, and chemisorbed sulfur blocks the nucleation of carbon.

III. Catalysis A. A. 1. Kinetic Studies

of Steam
REACTION

Reforming

RATE

Because of the thermodynamic constraints of the reforming reaction, the process is carried out at temperatures at which the activity of the catalyst is very high. But the reaction is also characterized by a high, negative heat of reaction. Consequently, the reforming process is subject to significant massand heat-transport restrictions, which makes it difficult to investigate the reaction fundamentally at the laboratory scale; laboratory results can easily be misleading (43,5(J). The effective diffusion coefficients of molecules in catalyst pores at the high pressures of industrial reformers are dominated by the bulk diffusion coefficient, whereas the Knudsen diffusion has significant influence at atmospheric pressure (14). Consequently, low-pressure laboratory tests of large catalyst particles can be misleading for evaluation of the activity of reforming catalysts. Hence, low-pressure tests must aim at determination of the intrinsic activity with relatively fine catalyst powders.

III]

HYDROGEN

AND

SYNTHESIS

GAS

BY STEAM-

AND

CO>

REFORMING

91

Rcaction/rcactor property

Reactor Gradicntlcss microreactor I (C) (mm) (m) vol. ( m3 ) 500 5 0.0 I2 0.2x10 I 0.3-0.5 50x10~ 60 000 (kgimlh) 2500 12

type Internal recycle reactor 20 500 50 0.0 I fi 20x I I6 mm rings 20 150000 8000 IO00 IOF 600 Bench-scale Pilot Industrial

Pressure Temperature Reactor Reactor Total Number Catalyst diameter Feed flow

(bar)

35 500-800 20 0.3 IOOX lo- I 33 0.6 1450 I500 52 I5 000

35 500--800 100 I2 0. I I I6 mm rings 550 l4SO 86 000 0500 73 000

35 500~~800 I00 12 I 1 200 I6 mm rings 110000 1450 86 000 9500 73 000

diameter length reactor

of tubes particle (mm) rate (Nmi/h)

Space velocity (vol CH$vol!h) Mass velocity

Reynolds Averacre (kcal&/h)

number heat tlux

The scale-down to laboratory reactors involves a change of a number of reaction conditions. Data for typical reactor sizes are listed in Table VIII. It is evident that neither heat fluxes nor mass flows can be reproduced in laboratory bench-scale units. Industrial reactors operate at high Reynolds numbers (Re= lOOO-lOOOO), but laboratory reactors typically operate at much lower Reynolds numbers, so that the resistance to convective heat and mass transfer is significantly greater. In laboratory experiments, the resistance to transport in the gas phase near the catalyst particle may be so great that the gas at the external surface of the catalyst pellet will be very close to equilibrium. Consequently, in investigations of reforming of higher hydrocarbons, hardly any higher hydrocarbons will be left at the catalyst surface. At the same time, the catalyst temperature may be significantly lower than that of the bulk gas phase. The higher the catalytic activity, the lower the catalyst temperature and hence the risk for carbon formation. This can easily lead to fallacious results, for

98

J.R. ROSTRUP-NIELSEN

et al.

[III

200

100

0.1 mm 0.05 mm

9 z 7 .E 2 9 t cl

50

*o-

10

*500 650 600

Gas temperature

FIG. 24. Temperature gradients in steam reforming reactors. HlO/CHd =4; HzO/Hz = 10, P= I bar abs. The calculations were made for a heterogeneous one-dimensional reactor model for various catalyst particle diameters, D,, (43). AT represents the temperature drop from the bulk gas phase to the surface of the catalyst particle.

instance, in TGA experiments with low gas velocities (80). It is essential to know the catalyst temperature or be able to estimate it. Even in investigations of small catalyst particles designed for measurements of intrinsic rates, temperature gradients can be substantial (43), as shown in Fig. 24. Evidently it is difficult to achieve gradientless reactor performance at temperatures above 600C. The temperature gradients may easily lead to the measurement of apparent activation energies that are lower than the actual values. At the high gas velocities required for measuring intrinsic rates at low conversions (Fig. 1 I), the lack of back-diffusion of product gases may lead to oxidation of the nickel catalyst, as methane behaves as an inert in the Ni/NiO equilibrium. The problem is solved by addition of hydrogen to the feed (e.g., with a H20/H2 ratio of 10). Axial dispersion also plays a significant role in determining the stability of carbide catalysts (Section 1II.C). A more direct approach is to characterize pellet kinetics by using industrial-size pellets in internal recycle reactors, which at high recycle rates approach the character of a perfectly mixed continuously stirred tank reactor (CSTR) (50). By varying the feed rate, it is possible to determine the kinetics at the high conversions encountered in large-scale reformers. The

III]

HYDROGEN

AND

SYNTHESIS

GAS

BY STEAM-

AND

CO1

REFORMING

90

mass velocity in the reactor is still one order of magnitude lower than that in an industrial reformer (Table VIII). Therefore, temperature gradients caused by the reforming reaction cannot be eliminated, and it is necessary to measure the catalyst pellet temperature in addition to the gas temperature. For this purpose, a thin thermocouple may be fitted in a drilled hole in one of the pellets. A.2. Reaction Kinetic.\ Table IX illustrates how various approaches (55) have been applied to establish intrinsic kinetics of steam reforming of hydrocarbons.

Authors Bodrov Khomenko Rostrup-Nielsen Tlrttrup (8.3) (5/) et 01. (XI) et ~1. (HZ) (K(l)

Form

of kinetics -Hinshelwood identity kinetics, power power law law

Langmuir Temkin Two-step

Pellet kinetics, Langmtnr~ Microkinetic

Xu and Froment Aparicio (84)

Hinshelwood analysis

Early work on the kinetics of the steam reforming of methane (81) was based on the assumption that the methane adsorption was rate determining, in agreement with the general assumption of a first-order dependence on methane concentration. Later work by Khomenko et al. (HZ) avoided the discussion of a rate-determining step; instead, the researchers invoked the quasi steady-state approximation in terms of the Temkin identity (X5), and the following rate expression was obtained for the temperature range of 47067OOT: k.PCH;P(l-/3) (9) = ,f( P H+PH?(l +KH~o(PH&%I$ where ,~(P,+o, P,,>) is a polynomial in P/I,, P HzO. Equation (9) includes five temperature-dependent constants. Xu and Froment (51) established a complex Langmuir-Hinshelwood expression, using a classic approach, on the basis of 280 measurements made with a NiiMgAl204 catalyst, but it is restricted to a narrow range of parameters: temperatures of 500-575C pressures of 3315 bar, and molar

100 100

J.R. ROSTRUP-NIELSEN

et al.

[III

0.1

1.1

1.15

1.2

1.25

1.3

1.35

1.4

1000/T

(Km)

FIG. 25. Steam reforming of CH4 in a plug-flow reactor with H20iCH4 =4, HlO/H> = IO, and 0.2 g of catalyst (with particle diameters in the range of 0. I GO.3 mm); space velocity is 1.7x 106(vol. total feed gasivol. cat. bed/h) h-,

H2OICH4 ratios of 3-5. A number of rate equations were established on the basis of a mechanism involving 13 steps, assuming one of the steps to be rate determining. It was shown that CO2 is produced not only by the shift reaction, but also by steam reforming. Hence, rates for the three reactions were found to give the best agreement with the measurements: a: b: CH4+H20=CO+3H2: CO+H20=COz+H2: ~1 = Q= Q=
kl PCH~ PHzO

P'Hz '22

(1 -IQ3

(10)
(11) (12)

kz . Pco . Ptl>o ( 1 - IQ> PHI z2


k3 PCHI p&O

c: CH4+2H20=COz+4H2: in which

p3.5 . 22 HZ

(1 -IQ,

PHzO z= 1 +&co .Pco+&.H~PH~ +K~,,~H~,PcHI+K~.H:o~H_. ,

Because the three reactions are not independent, it is necessary to combine the three rate equations into two: one for conversion of methane and one for production of COl:
YCHl YCOZ = = r1 + /3, 12 + i-3.

(13) (14)

These two expressions include five temperature-dependent constants. They show a small negative reaction order in the overall pressure, in agreement with the data of Figs. 2.5 (68) and 26 (86). The negative reaction order in

Ill]

HYDROGEN

AND

SYNTHESIS

GAS

BY STEAM-

AND

CO,

KEFORMING

101

1.4 ~1

I 2

I 10 20 30 Pbarabs

FK;. 26. Steam reforming of CH4 reactor; r,,, = 6OoC, H?O;CH4 -3, cylinder with 7 holes (diam.,helghtlholes: vol. cat. bed).!h.

at various pressures HlO;ltz =5, mass 16/8/7x3 mm):

(86). Data were obtained in a Bcrty of catalyst was 20 g, catalyst pellet: space velocity was IO(~ol. total feed!

Parameter in rate equation

Xu and Froment

(5/)

Avetniso\

ct trl.

(X7)

4.225. IO 1.995.10 1.020.105

exp( -~240. I:RT) exp(-67.l!RT) exp(-24.3.9iRT)

I .97- IO 2.43.10 3.99.10 3.35.10 2.06.10 6.74.10- 9.48.101

eup(-24X.O:RT) exp(p54.7;RT) cxp(p278.5;RT) 4 exp(h5.5iRT) ) exp(%.SiRT) exp(34. URT)

8.23. 10m5 exp(70.65/RT) 6.12~10~ 6.65.10-. exp(XZ.OO/RT) cxp(38,28!RT)

I .77. IO exp(-88.6XiRT)

exp(p74.9!R7) Rate constants are in are in units of k.l/mol.

a Ni/MgAllO4 catalyst, Ni surface arca estimated to be 3 n&g. units of mol g- hK Activation energies and heats of adsorption Pressures in bar.

total pressure was confirmed by measurements with industrial-size catalyst pellets in an internal recycle reactor (Berty reactor), as shown in Fig. 26. An alternative analysis of the data of Xu and Froment was carried out by Avetnisov et al. (87) involving more data and resulting in better agreement with calculated and measured values of partial pressures of CO. The kinetics constants are compared in Table X. The decrease in the reaction order in steam with temperature is reflected in the model of Xu and Froment by an increase in the coverage by oxygen atoms. This implies an unlikely negative heat of adsorption of steam. Furthermore, the model includes a significant surface coverage by methane. However, as discussed in Section WA, surface science studies show that methane dissociation does not proceed via an adsorbed precursor state.

102

J.R. ROSTRUP-NIELSEN

et al.

[III

The inconsistency may result because of the narrow range of the HzOICH~ ratio and the temperature in the investigation of Xu and Froment and, further, because the mechanism of the steam reforming reactions apparently cannot be represented by a single rate-determining step over a wide range of conditions. Bodrov et al. (88) found that COZ-reforming of methane was described by the same reaction kinetics as used for steam reforming. This result was confirmed by other investigators (89). The change in mechanism in operation with carbon dioxide instead of steam would have little practical impact on reforming, because steam will be present not far from the inlet, and also at the centers of the catalyst particles as a consequence of the low effectiveness factors of catalysts in industrial reformers. The support is probably involved in the reforming reaction by influencing the activation of steam. This inference is reflected by the terms Ka,~?o or ouzo in power-law kinetics and by the overall pressure dependency of the reforming rate (14), as discussed further in Section II1.D.

B. B.1. Nickel
swfizce area

SURFACE

CHARACTERIZATION

The activity of a nickel catalyst is related to the nickel surface area. Several methods have been used to determine nickel surface areas, including chemisorption, X-ray diffraction (XRD), and electron microscopy, as discussed in this section. XRD, with data interpreted on the basis of the Scherrer equation, provides a quick and easy method to determine the volume-averaged nickel particle size. Generally smaller nickel particle sizes are indicated by XRD than by other methods (43,90-92). There are two reasons for this disagreement: XRD is normally performed with passivated catalysts that have been exposed to air, and the technique gives the size of the crystalline nickel phase. During passivation, a surface oxide layer is formed on the nickel particle, which is not detected by XRD, causing the XRD estimate of the crystallite size to be too low. Another reason is that nickel particles are polycrystalline: XRD gives an average size of the crystalline domains, which are smaller than the total particle size estimated by transmission electron microscopy (TEM) and chemisorption methods. The presence of twin planes in the TEM images (Fig. 27) of nickel particles further supports this conclusion. Anomalous small angle X-ray scattering (ASAXS) data have recently been used to determine the particle sizes and not just the sizes of the single-crystal domains. This method detects a full distribution of nickel particles supported on silica (93).

III]

HYDROGEN

AND

SYNTHESIS

GAS

BY STEAM-

AND

CO,

REFORMING

103

FIL. 27. TEM image of a 14.8wt% I IlO:H2 = 10: 1 at 30 bar and at 500C in smtered nickel particles (WI).

Ni/MgAlzOd spine1 catalyst after 770 h of sintcring in (773 K). The arrows show the presence of internal faults

TEM provides size distributions of the nickel particles and information about the structure of the catalyst that can hardly be obtained by any other technique. The potential problems with the technique include the following: (a) only projections of the particles are obtained, (b) only a very small part of the sample is analyzed, and (c) it is difficult to distinguish the nickel and the carrier particles, and (d) investigation of passivated particles may exclude detection of the smallest particles. Nonetheless, TEM is an extremely powerful technique for characterization of catalytic surfaces. The most widely used method for measurements of the nickel surface area is chemisorption. In the application of this method, it is crucial that multi-layer adsorption at the reactive surface or adsorption on the carrier be avoided and that monolayer adsorption on the nickel surface be obtained. Furthermore, the coverages of all facets of the nickel crystals should be similar, so that the chemisorption capacity can be converted into a nickel surface area. Presumably, chemisorption detects only exposed nickel metal atoms whereas nickel atoms at the metal-support interface are normally not measured. B.2. Chtwisorption
ofH~&ogen

Chemisorption of hydrogen has been used often as a convenient method to determine the surface area of nickel in supported catalysts (14,94). TPD spectra of nickel single crystals show that a low energy state of hydrogen,

104

J.R. ROSTRUP-NIELSEN

et al.

[III

PI, is formed reversibly at room temperature. The fraction of reversibly held hydrogen at room temperature on supported nickel catalysts is also observed to vary between IO-90%. The saturation coverage of hydrogen at -153-(-93C) (120-l SOK) (/I t and /;2 states) corresponds to 0.9-l .O monolayers on Ni(ll1) and Ni(lOO) (Y4-96). On Ni( 1 lo), a saturation coverage of 1.5 ML can be obtained(97). From an extrapolation of the rate of desorption obtained at a temperature of 220K measured for a change in coverage from 1.5 to 1.0 ML, and assuming a sticking coefficient of hydrogen of 0.1 at room temperature, we conclude that Ni( 110) will probably be covered with 1.5 ML of hydrogen atoms at the hydrogen pressures used for chemisorption measurements. However, most chemisorption measurements are made with catalysts having large nickel particle sizes (>lOO A); hence Ni( 110) and other high-index planes constitute only a small fraction of the total nickel area. The additional hydrogen adsorbed associated with coverages >l ML of hydrogen is probably small. The rate of adsorption of hydrogen at room temperature and pressures of 100400 Torr is observed to proceed in three stages (94): (i) 50-80% of the adsorption occurs during the first 5 min; (ii) most of the remaining adsorption takes place in the following 45 min; (iii) a small part (335%) of a monolayer is adsorbed later. It was observed (98) that chemisorption of hydrogen on a Ni/MgAlzOd catalyst continued to increase linearly with time for 65 h. The stoichiometry of hydrogen adsorption on nickel was addressed by Pannell et al. (99), who obtained a H:Ni ratio of 1.1 by comparing the chemisorption of hydrogen at room temperature to BET measurements of the surface area of nickel powder. Good correspondence was also observed between nickel crystallite size determined by hydrogen chemisorption at room temperature and that determined by XRD and TEM for Ni/SiOz and NilAl (100). Hydrogen chemisorption seems to be able to give a quantitative value of the total nickel surface area in supported catalysts when the following criteria are met: the procedure recommended by Bartholomew (Y4) is used, the carrier does not interact strongly with the metal phase, and the nickel particles are not too small.

B .3. Chemisorption

c~fSulfur

Dissociative chemisorption of hydrogen sulfide is another method used to measure the nickel surface area(9H,I4). The method is attractive because sulfur cannot be removed from the nickel surface once it is adsorbed, unless high temperatures and pressures of steam and/or hydrogen are applied. The adsorption energy of sulfur is high, approximately 90-290 kJ/mol(101), and

Ill]

HYDROGEN

AND

SYNTHESIS

GAS

BY STEAM-

AND

cOz

REFORMING

105

1.8 1.6 -:

:
m l

0.6 0.4

- ': : 300 I 400 I I 500 600 Temperature ("C) I 700


I

800

FK,. 2X. Steady-state sulfur coverages determined as a function of the sulfidation with a Ni!MgAlzOd catalyst and a concentration of H2S in Hz of I I ppm contacting for 7 days.

temperature the catalyst

coverage dependent, and so contamination of the nickel with promoter or carrier atoms does not seem to influence the sulfur adsorption capacity (14). The sulfur chemisorption measurement is performed as follows: a flow of HIS/HZ over the catalyst is continued until saturation, and the sulfur uptake of the catalyst is then determined. The saturation coverage of sulfur determined with a radiochemical technique, LEED, and Auger spectroscopy was characterized (l02-104), and values corresponding to 440 ppm of S/m were obtained for Ni( 11 l), Ni( loo), and Ni( 110): the S:Ni ratio was found to be 0.54. Nickel is nearly saturated with sulfur at 550C (823 K) and values of PHJS/PHI > 1 ppm (Y&80). If it is assumed that 0s = 0.995 at 500C (773 K) and PII~s/P~,~ is in the range 7.546 ppm, the coverage of sulfur can be expressed as follows (101):

where As = -19 J/K/mol and AH = -289 kJ/mol. The isostere is plotted in Fig. 28 together with the sulfur uptake of a NilMgA1204 catalyst normalized to that at 550C. The coverage at temperatures of 550C (823 K) and higher is in excellent agreement with the isostere determined by Alstrup et al. (101),

106

J.R. ROSTRUP-NIELSEN

et al.

[III

but the measured sulfur uptake is significantly greater than predicted by the isostere at lower temperatures. The nickel crystallite size determined by XRD does not change with temperature in the temperature range 350-550C (623-823 K), and sulfidation of the MgA1204 carrier at 446 (7 19 K), 5 17 (790 K), and 526C (799 K) led to no significant sulfur uptake. The additional uptake of sulfur on the catalyst observed at temperatures below 550C (823 K) is therefore not an uptake by the carrier or a sintering phenomenon, but could be an indication of spillover of sulfur from nickel onto the carrier or an indication of the presence of sulfur in lower layers of nickel or bulk nickel as the temperature approaches the bulk sulfidation limit. The work of McCarty and Wise (10.5) supports the idea that bulk sulfidation becomes important, as their data also show a strong increase in the sulfur uptake as the bulk sulfidation limit is approached, but their experiments were done under conditions different from those considered here. In early work, Rostrup-Nielsen (98) compared the nickel surface areas determined by hydrogen and sulfur chemisorption following Beecks suggestion (106) to represent the non-activated hydrogen adsorption as a monolayer. The hydrogen chemisorption was determined at -72C (201 K) and 0.073 bar for a nickel catalyst. As discussed later by Weatherbee and Bartholomew (107), part of the nickel area in supported nickel catalysts adsorbs hydrogen only by an activated process. Using 10 kJ/mol as an estimate of the activation energy, as was observed for Ni/AlxOs, we infer that the activation of hydrogen is more than seven times faster at 27C (300 K) than at -72 (201 K) and that hydrogen, which adsorbs after 30-45 min at room temperature, will be completely adsorbed only after 3.5-5.5 h at -72 (201 K). The uptake of hydrogen after one hour was used for estimates of the nickel area (98). The H:S ratio estimated by Rostrup-Nielsen (98) from sulfur and hydrogen chemisorption was 1.35, which is significantly lower than the expected value of 1.85 estimated on the basis of a H:Ni stoichiometry of 1: 1 and a S:Ni stoichiometry of 0.54. This difference indicates that chemisorption of hydrogen may not have been completed under the conditions applied. Oliphant et al. (108) also determined the H:S ratio by sulfur and hydrogen chemisorption and obtained values in the range 0.96-l .37 for a 3% Ni/A1203 catalyst and nickel powder. The reason for this low H:S value may be that Oliphant et al. measured the sulfur capacity close to the bulk sulhdation limit. The data of Fig. 28 show that the uptake of sulfur increases dramatically as the bulk sulfidation limit is approached. To shed further light on the ratio of sulfur to hydrogen chemisorption on nickel catalysts, the hydrogen and the sulfur chemisorption capacities of a series of nickel catalysts supported on MgA1204 were measured. The hydrogen chemisorption capacity was determined by the following procedure:

III]

HYDROGEN

AND

SYNTHESIS

GAS

BY STEAM-

AND

COz

REFORMING

107

5 Ni Area from

10 the Sulfur

15 Capacity

20 (m g)

25

FIG. 29. Nickel surface areas of various catalysts calculated from hydrogen data compared with the nickel surface areas determined by sulfur chemisorption. fitted to a straight line through the origin. The slope of the line is 0.98.

chemisorption The data arc

First, the catalyst was reduced at 600C (873 K) in hydrogen for one hour. Then the remaining hydrogen gas was pumped off and the sample cooled to 25C (298 K). At this temperature, the sample was exposed to small amounts of hydrogen gas. A new pulse of hydrogen was not allowed into the reactor before the pressure over the sample was stable for 5 min. The typical time between two hydrogen pulses was 20min, and the chemisorption experiment typically lasted for 4 h. The hydrogen chemisorption was determined at 0.39 bar and 25C (298K). The hydrogen chemisorption capacity of the carrier determined by the same method is small (< 10% of the lowest Hz chemisorption). The hydrogen chemisorption capacity is converted to a nickel surface area by use of the stoichiometry H:Ni = 1: 1 and the assumption that one Ni atom occupies an area of 6.5 A (14). The nickel areas determined by this procedure for a series of catalysts are compared to those determined by sulfur chemisorption in Fig. 29. The slope of the straight line in this figure is 0.98, which is so close to unity that it seems reasonable to conclude that hydrogen and sulfur chemisorption capacities both give good values of the nickel surface area.

It has been suggested that step sites on nickel surfaces are important for steam reforming (69,cYO). If this is true, it is important to find a method to probe only these sites and not the total number of surface nickel atoms.

108

J.R. ROSTRUP-NIELSEN TABLE XI potcrssium-promoted BET surface area Cm* gm ) 71.1 76.4

et al.

[III

Datu Catalyst name Comment

chamcterizing

Ni/A12@ N2 adsorption capacity (ml kg~ ) 196 13

uta!v.st.s

(80)

Nickel surface area (m2 gg ) 3.62

Rate of CzHb reforming at 500C (773 K) h (molg-~ h--l) 0.82 0. I9

C3
CIO C3+2.1%K

4.19

a Both catalysts: support y-Al203, nickel content 20 wt%. H20iClHb = 8, HzO/Hz = IO, I bar abs. Catalyst Cl0 wab prepared from catalyst C3 by addition of potassium

van Hardeveld et al. (109-I I I) suggested that the number of step defects or B5 sites can be determined by nitrogen adsorption, and this method was later used to characterize a series of catalysts (80). Several authors (112) challenged the method for determination of B5 sites. However, Arumainayagam et al. (I 13) and Tripa et al. (114) recently showed that adsorption of nitrogen on platinum is restricted to step defect sites. Adsorption of nitrogen on Ni( 111) is observed at low temperatures. Quick et al. (115) and Yoshinobu et al. (116) investigated adsorption of nitrogen on Ni( 111) at -190C (83 K) and -184-(-158C) (89-l 15 K), respectively. Extrapolation of the absorption isotherms for nitrogen reported by Yoshinobu et al. to room temperature shows that nitrogen does not adsorb on Ni( 111) at room temperature and pressures below 1 bar. Hence, it seems reasonable to conclude that the adsorption of nitrogen on nickel catalysts at room temperature takes place only on step defect or BS sites, as proposed by van Hardeveld et al. Determination of the number of step defect sites using nitrogen adsorption, however, remains only semi-quantitative because of the presence of unoccupied B5 sites and physically adsorbed nitrogen; the number of BS sites is estimated to be between one and two times the number of nitrogen molecules adsorbed. The effect of potassium promotion of Ni/AlIOs on the nitrogen adsorption capacity is illustrated by the data in Table XI. The data show that potassium promotion of catalyst C3 does not change the surface area of the active phase or the total BET area. However, potassium has a strong effect on the nitrogen adsorption capacity, suggesting that the number of step defect sites is strongly reduced by potassium. In Fig. 30, the rate of ethane reforming is plotted as a function of the nitrogen adsorption capacities of the Ni/Al*Ol catalysts. There is a good correlation between the reaction rate and this capacity, including data for both potassium-promoted and unpromoted catalysts. The correlation between

1111

HYDROGEN

AND

SYNTHESIS

GAS

BY STEAM-

AND

CO2

RF.FORMIN(;

I OS

200 NZ cap.

400 (ml/kg)

600

800

FIG;. 30. Rate of steam reforming of ethane [H2O/C2/11(,=8, 500C (773 K)] on nickel catalysts supported on alumina (X0). the nitrogen adsorption capacity.

The

Il~OiH~ = IO, tcmpcraturc horizontal axis represents

the reaction rate and the nickel surface area is not good when data for the potassium-containing samples are included. One explanation for this result is that potassium may preferentially bond to the highly uncoordinated nickel atoms at the step defect sites and that these sites largely determine the activity of the catalyst. This explanation is consistent with the DFT calculations presented in Section IV, which show that potassium bonds strongly only to step defect sites and therefore blocks these sites preferentially.

C.

SINTERING

Sintering is an important cause of deactivation of nickel-containing steamreforming catalysts. A good understanding of the sintering mechanisms is crucial, both for prediction of the extent of deactivation and for design of catalysts that maintain high activity. Several investigations of the sintering of nickel particles have been reported (14,90,91,117-123). Sintering is a complex process, influenced by many parameters, including temperature, chemical environment, catalyst composition and structure, and support morphology. The most important parameters are the temperature and the atmosphere in contact with the catalyst. Increasing the temperature results in a significantly faster sintering, and the presence of water greatly accelerates sintering (120). Sintering tends to be faster for narrow particle size distributions (117). The support can affect the sintering in various ways. It has been proposed that the pore structure, the morphology, and phase transitions of the support determine the final particle size of the nickel.

110 12 10 8 z 5 s !?? LL 8-

J.R. ROSTRUP-NIELSEN

et al.

[III

64 20

20

40

60

80

Particle size in nm
FIG. 3 I. Particle size distribution of a sintered spinel, sintered for 770 h in a mixture of HlO:Hl supported nickel catalyst (22 wt% Ni/MgAllOd = 10: I at 30 bar and 500C (773 K)) (90).

To describe the sintering of metal particles, two mechanisms are generally proposed: atom migration (Ostwald ripening) and crystallite migration and coalescence. Ostwald ripening refers to the process whereby metal atoms are emitted from one metal particle and captured by another. In the coalescence process, the crystallites themselves move over the support and collide to form larger particles. The driving force for both processes is the difference in surface energy, which varies inversely with particle size. In both of these processes, sintering slows down with time and results in a semi-stable state, with characteristic asymptotic particle size distributions. Granqvist and Buhrman (124-126) have shown that Ostwald ripening results in a particle size distribution with a tail towards small particle sizes and a steep slope towards larger particle sizes. A recent simulation(l27) confirmed these characteristics of the size distribution from Ostwald ripening. The particle size distribution resulting from the coalescence model is a log normal distribution, which has the tail towards the large particle side and falls steeply towards the small particle side. The use of particle size distributions to deduce the sintering mechanism was challenged by Wanke (12H), who proposed that the particle size distribution in a sintered catalyst depends on the initial distribution in the fresh catalyst. Although Wankes (128) argument has merit, it is useful to consider the assumptions made by Granqvist and Buhrman (125) in their derivation of their statistical model for coalescence of islands: They assumed that islands have undergone many collisions and it is only after a large number of such collisions that the distribution approaches a log normal distribution. Sehested et al. (90) investigated the particle size distributions of various catalysts by electron microscopy. An example of a particle size distribution is shown in Fig. 3 1. The smooth line is a log-normal fit of the data. The excellence of the

Ill]

HYDROGEN

AND

SYNTHESIS

GAS

BY

STEAM-

AND

CO2

REFORMING

III

fit strongly indicates that particle migration and coalescence was the dominant sintering mechanism in this experiment. Fuentes and Gamas (129) recently proposed that Ostwald ripening can mathematically also give rise to particle size distributions similar to log-normal distributions. More work is needed to verify whether this is also observed experimentally. Other information used to distinguish between the two models of sintering is the reaction order of the loss of metal surface area as a function of time (117,llK. 124,130). Most of the reported sintering data were analyzed on the basis of power-law kinetics:

in which ANi is the nickel surface area, a function of time and the rate constant k. A generalized power-law expression is also used by some authors (120). The value of n in Eq. (16) is determined by the sintering mechanism. For particle migration and coalescence, II = 8; Ostwald ripening is characterized by n = 5 (130). Values of n in the range of 3-l 5, with an average of 8.6, have been reported for supported nickel catalysts, and values of 2214, with an average of 7.2, have been reported for supported platinum catalysts (120). These values are close to the value of 8 predicted for particle migration and coalescence. Facetting of the metal particles may lead to values of n > 8. Generally, high values of n are observed for sintering at low temperatures and the reverse at high temperatures (117,118.120). The change in n is attributed to a change in the sintering mechanism from particle migration and coalescence at low temperatures to atom migration at high temperatures. The influence of the nickel loading and the carrier surface area on the nickel surface area after sintering has been investigated (90). The nickel surface areas observed after long sintering times are shown in Fig. 32. The nickel loading is important in determining the nickel surface area after sintering. Supports with two different surface areas were used, and it is evident from Fig. 32 that there is only a small difference in the nickel surface areas. These two observations are consistent with sintering by particle migration and coalescence, which was inferred from an analysis of the particle size distributions. In this mechanism, the diffusion of nickel atoms over the nickel crystallites determines the rate of sintering. An effective diffusion constant for migration of spherical nickel particles is given in ref. (124):

where D, is the surface diffusion coefficient of nickel atoms in cm*/s, a is the atomic diameter, and R is the radius of the nickel particles.

112
8 765432-

J.R. ROSTRUP-NIELSEN

et al.

[III

10

15

20

25

30

35

Ni (wt%)

FIG. 32. Nickel surface areas after long sintering times (~600 h) for catalysts supported on MgA1204 with various surface areas; the data are plotted as a function of the weight percent nickel in the catalyst. The sintering conditions were 500C (773 K), HzO:Hz = IO: 1. and pressure = 30 bar (90).

Increasing the surface area of the carrier also increases the distance between the nickel particles, but because the mobility of the particles depends so strongly on the particle size, the increase in distance between the particles results in only a small reduction of the nickel particle size. On the other hand, an increase in the nickel loading results in a decrease in the distance between the particles. A small increase in the particle size will decrease the mobility so much that it compensates for the decreased distance between the particles. Equation (17) states that the particle diffusion ceases when the particle size exceeds a given size. This maximum size increases with temperature. Results of sintering tests with NiiMgAl204 (80) are shown in Fig. 33. At 550C (823 K), the sintering was insignificant. Even the non-sintered catalyst had large nickel particles (R,,,, = 61 nm). This observation, rather than the rule of Tamman, may explain the lack of sintering at 550C (823 K). Values of n in the power-law expression can be obtained from the data in Fig. 33, the values being 7.7 and 6.9 at temperatures of 700 (973 K) and 800C (1073 K), respectively. According to Hughes (130), this value is close to what is expected for spherical particle migration and coalescence.

D.

ACTIVITY

TRENDS

The rate of steam reforming of methane is correlated linearly with the rate of steam reforming of ethane and of liquid hydrocarbons. Hydrogenolysis of ethane shows a trend similar to that characteristic of the reforming reactions (14).

III]

IIYDROGEN

AND

SYNTHESIS

GAS

BY STEAM-

AND

CO1

REFORMING

II3

0.1 1 1

I
10

,
100 11 IO

Time

(hours)

FIG;. 33. Sintcring of supported nickel (14.80). Catalyst: Ni!MgAllOA, 8 wt% Ni, K,,,,,,, = 61 nm. BET surface area 4m/g, H>O:IIz =3, P= I bar. Ihc lines were Wed using the power law expression Ah,,/AL, = kt. The exponents, II. are -0.007. 0.15. and -0.17 at 550 (X23), 700 (073), and 800C (1073 K). respectively.

Activity for the reforming reactions is roughly proportional to the metal surface area for a given metal on any of a wide selection of supports (I 4). The turnover frequency No (at HlO/CHJ =4.0, H20/H2 = 10, temperature 450C (723 K), and pressure 1 bar) is typically in the range 0.5-I .O s- The activity per unit metal surface area (the specific activity) increases (14) with the metal crystal size. Rhodium and ruthenium show No values about ten times higher than nickel, platinum, or palladium. The low activity of cobalt is probably attributable to the processing conditions, with the HlO/Hl value being close to that causing oxidation of the metal. The replacement of steam by carbon dioxide results in a decrease in activity to an extent that depends on the metal (72). The effect of CO2 is less for nickel than for noble metals. This result corresponds to the observations of Bodrov and Apelbaum (88) showing identical rates on nickel for steam and CO? reforming (Section Ill.A.2). Hence, the superiority of rhodium and ruthenium compared to nickel is less pronounced for CO1 reforming than for steam reforming(72). The change in behavior of the metals when steam is replaced by carbon dioxide may be related to the magnitude of the heat of chemisorption of carbon monoxide (72). Nickel is characterized by a lower heat of chemisorption (143 kJimol) of carbon monoxide than ruthenium ( 163 kJimo1) or rhodium ( I74 kJimo1). On the other hand, the value for platinum ( I33 kJ/mol) is comparable to that for nickel (131). The addition of potassium to a nickel catalyst may result in a decrease of No by more than one order of magnitude (14). as illustrated in Section 1II.B (Table Xl). In practice, this means that even trace amounts (-1000 ppm by wt) of alkali may bias activity tests. The effect of potassium on a

114

J.R. ROSTRUP-NIELSEN TABLE XII

et al.

[III

Catalyst content Ni (9. I ) Ni (7.1), Ni (8.3), Ni (8.5). Ru (0.4) Ru (0.5), Ru (0.3), Ru (0.3), Rh (0.5) Rh (0.4), Rh (0.4). Rh (0.4), Pt (0.5) Pt (0.3),

metal (wt%)

Relative

Rate

Activation Energy (kJ/mol) 109 100 88 105 96 71 105 100 121 (100) 96 (142)

1.00 Li (1.6) Na (1.3) K (1.0) 0.6 0.3 0.2 5.0 Li (1.1) Na (1.3) K ( 1.3) 2.0 2.0 0.3 17 Li (0.2) Na (I .3) K (I .O)

(16)
8.0 0.5 1.8

Na (1.4)

(0.2)
= 10;

a Catalyst support was MgAl204; HlO/CHd x4.0, HzO/Hl Temperature 500C (773 K); Pressure I bar abs (33).

number of group 8 metals has been observed, as illustrated in Table XII. The large decline in activity is not accompanied by any significant change in the activation energy. It is remarkable that the decline in activity upon promotion with alkali is also observed for hydrogenolysis of ethane with nickel catalysts (14) a reaction occurring in the absence of steam. The influence of alkali is stronger for catalysts on less acidic supports, which suggests that the alkali partial pressure over the catalyst is important (14). A less acidic support is characterized by a weaker bonding of alkali, resulting in more rapid transport (via the gas phase) from the support to the metal. As shown in Fig. 30, the effect may be related to a blockage of step sites on the metal surface by potassium. Difficulties in desulfurizing heavy feedstocks have led to attempts to use non-metallic catalysts for steam reforming (14). Claridge et al. (132) reported that molybdenum carbide and tungsten carbide are catalysts for steam and CO1 reforming and catalytic partial oxidation. However, in synthesis gas the

Ill]

HYDROGEN

AND

SYNTHESIS

GAS

HY STEAM-

AND

CO2

REFORMING

115

carbides were stable only at elevated pressures (ca. 8 bar), and they were transformed into the oxides at ambient pressure. CO* reforming of methane catalyzed by MolC was investigated (133) at 8 bar and 1.6 bar pressure with a plug flow reactor and a flow reactor with external recycle operated as a CSTR, respectively. In the plug flow reactor, the deactivation of the catalyst started from the inlet of the reactor and downward, while the catalyst was stable in the CSTR at high conversions. Thermodynamic calculations of the stability of Mo2C catalysts under CO2 reforming conditions show that Mo2C is stable only at high product concentrations. This was later confirmed by Tracy et al. (134). However, Tracy et al. and Tsuji et ul. (135) claimed that Mo2C supported on other carriers had greater stability than pure MolC. The activity for CO2 reforming of MozC is more than two orders of magnitude lower than that of a 1.8% Ru/MgAlzOd catalyst on the basis of weight (133). The resistance of MolC to carbon formation is significantly greater than that of nickel. E. CATALYSTPROMOTION The catalysts are promoted to reduce the risk of carbon formation. Carbon is avoided when the steady state activity of carbon is less than one (14). In terms of a very simplified sequence shown in Table XIII, this means

(18)
The steady state carbon activity can be decreased by the following: enhancing the adsorption of steam or CO2 (indicated by the equilibrium constant K,,.) enhancing the rates of surface reactions (indicated by the rate constant kg) decreasing the rate of hydrocarbon activation (indicated by the rate constant k,d) Kinetics experiments indicated that the adsorption of steam was enhanced by active magnesia and alkali and that spill-over of adsorbed steam to the metal surface may play a role (14). This was reflected by negative reaction orders with respect to steam. The method of preparation of the Ni/MgO catalyst plays an essential role in the promoting effect, as illustrated in Table XIV. Similar effects of La203 and Ce203 on CO1 reforming and steam reforming have been observed (136-138). However, little fundamental work has been done to clarify the detailed role of enhanced adsorption of steam and CO2 on the catalyst.
l l l

II6

J.R. ROSTRUP-NIELSEN TABLE


Steam Rejimning: Simpl$ied

et al.

[III

XIII
Reaction Sequrncc

Reactant

Rate constant or equilibrium constant

Product

b k,

CnH>.*z
C* whisker carbon

kg

C* + OH* H20+* 64, e

gas OH* + ;H2

Preparation No. (X0) A20 A21 A22 A23 A25 a Activity b Coking

Preparation method 2 2 I I 3 data: HlO/CzHe data: H20/nC,H16

Calcination temperature

Activity 1.,/t., 8.21222

Coking

parameters

(Eq.

8)

k,. (Ilgimin) 20 I2 I.4 17 1.8

41 (min) 33 27 91 27 152 r,s mol/m2 Ni/h.

high

2.0/183 I I .)I287

high

2.01282 10.5/165

= 8, H20iH2 = IO, 1 bar, 500C (773 K), Y, mol/g/h, =2/7, HzO/Hz = IO, I bar, 500C (773 K).

Steam adsorption on various supports commonly used for nickel-catalyzed steam reforming catalysts was investigated by micro-calorimetry (7.5). In contrast to what would be expected from the simple model sketched above, the magnesia support showed the lowest amount of adsorption of steam, followed by magnesium-aluminum spine1 and the alumina. However, by isotope exchange experiments (79, it was shown that the magnesia-based catalyst is more active for dissociation of the adsorbed steam. The result demonstrates that the improved adsorption of steam on magnesia supports resulting in improved resistance to carbon formation is a dynamic effect. In terms of the reaction sequence in Table XIII, K,. cannot reflect a true equilibrium constant. This would violate the principle of microscopic reversibility, because steam is

1111

HYDROGEN

AND

SYNTHESIS

GAS

BY

STEAM-

AND

((I2

REFORMING

117

also adsorbed directly on the nickel surface, as illustrated by a more detailed reaction sequence:
H20 + *sup = H20*sup,

(19)
+ H*sup>
+ *mp,

H20*sup

+ *sup
aup + *Ni

=
=

OH*,,,,,
OH*Ni

(20) (21) (22)

OH*

H20 + 2*Ni = OH*Ni + HANK.

This sequence implies that the spillover of steam probably involves OH species instead of molecular water. This inference is in agreement with many recently published results. Bradford and Vannice (139) developed a kinetic model for Ni/MgO and Ni/Ti02 and concluded that surface OH groups, possibly situated on the support, react with CH, intermediates absorbed on the nickel. Work by Efstathiou et al. (140) indicated that spillover of lattice oxygen from yttrium-stabilized zirconia was involved in the reforming reaction. Bitter et al. (141) found for CO2 reforming of methane on Pt/ZrOl catalysts that the reaction rate was proportional to the length of the metalsupport perimeter. A number of recent investigations have dealt with the impact of changing the catalyst composition on the activation of methane. Osaki et ul. (142) investigated the degree of dehydrogenation of CH, species on various catalysts and found indirectly that x was larger for nickel than for cobalt and larger for MgO-supported than for SiOI-supported catalysts. Zhang and Verykios (143) claimed a similar double effect (i.e., methane activation as well as enhanced adsorption of COz) to be responsible for the promoting effect when using La203 as support for a nickel catalyst. Other investigators have shown similar promoter activity of CelOi-containing catalysts for steam reforming of butane (43). Borowiecki et ul. (143,144) reported retarding effects of molybdenum and tungsten on the coking rate. Later work suggests that it is molybdenum oxide that causes the reduced rate of carbon formation (145). Bradford and Vannice (146) who studied PtiTiO., and PtiZrOl catalysts for CO2 reforming, found strong evidence for TiO, layers on the platinum surface suppressing carbon deposition, probably by ensemble control. More work is required to explain these promoting effects of various oxides and to clarify whether the promoters act by decorating the nickel surfaces as illustrated in Fig. 34. A direct blockage of surface nickel atoms with resulting ensemble control was observed for partly sulfur-poisoned nickel catalysts, as discussed in Section I1.E. A similar mechanism was claimed for the promoting effect of bismuth addition to nickel (148). Alloying of nickel with copper (70) can also decrease the rate of carbon formation, but it is not possible to achieve the

J.R. ROSTRUP-NIELSEN

et al.

[III

FIG. 34. In-situ HRTEM characterization of La203 is located on the nickel crystal

of a lanthana-promoted (147).

nickel

catalyst.

A cr) istallite

450

475

500 Temperature/C

525

550

FIG. 35. TGA data showing the accumulation of carbon on Ni/MgAlIOd (I 7 wt% Ni/Au catalysts (I .85 at% Au). The reaction conditions were as follows: H2OiC4H H20=5; the feed gas contained 69~01% Nz; the pressure was I bar abs (75).

Ni) and 1 = I .5;

required high surface coverage of copper atoms as it is with sulfur atoms to eliminate carbon formation. Although nickel and copper form a stable random alloy, this is not the case for the Ni-Au system (149,150). Nickel and gold do not mix in the bulk, but may form stable alloy-like structures in the outermost layer. As a result, it was possible to control the surface coverage and to eliminate carbon formation, as illustrated by the TGA data in Fig. 35 (71).

IV]

HYDROGEN

AND

SYNTHESIS

GAS

HY STEAM-

AND

(02

REFORMING

II9

Alloying of platinum with rhenium (1.51) or tin (1.52) was claimed to reduce carbon formation.

IV. Reaction

Mechanisms

The reforming reaction takes place on the nickel surface. A better understanding of the mechanism of reforming can be obtained from experiments in which nickel single crystals are studied. In this section, the activation of CHJ, CO, CO?, HzO, and H2 on nickel single crystals and the activation of CH? on catalysts are discussed, followed by a detailed evaluation of the mechanisms of reforming and coking and the effects of promoters in light of recently published DFT calculations. A. ADSORPTWN ot REKTANTS A. 1. Methune The dissociation of methane on nickel surfaces has been investigated extensively, and several details of the reaction pathway are known. From the first investigations of methane activation on Ni( 1 I I), Ni( loo), and Ni( 1 IO), activation energies in the range 27756 kJ/mol were obtained (153,154). These values are probably too low, because the methane pressure was not sufficient to allow full thermal equilibration between the gas phase and the single crystal. Nielsen et uf. (155) characterized dissociation of methane on Ni( 100) as a function of the pressure of methane and temperature and in the presence of a thermal finger allowing full thermal equilibration between the gas phase and the single crystal. The data show that methane dissociation proceeds via a direct mechanism rather than via a CH4 intermediate on the nickel surface, with an activation energy of 59f 1.5 kJimo1. Holmblad et al. (156) investigated the sticking of methane to Ni( 100) using a molecular beam technique and concluded that the dissociation of methane is dominated by the first vibrationally excited stretching mode. Good agreement was obtained between the data determined by a thermal technique and with a molecular beam. Bengaard et uf. (1.50) studied the effect of potassium promotion (in the absence of oxygen) on the dissociative sticking of methane to Ni( 11 1) and observed that 0.1 ML of potassium reduces the sticking coefficient by more than two orders of magnitude. Egebjerg et ul. (157) investigated the effect of steps on the dissociation of CHQ on Ni( 11 1) and Ru(0001). However, the step density was comparable to the uncertainty of the carbon coverages, making it difficult to establish any increased reactivity of the steps. DFT calculations

120

J.R. ROSTRUP-NIELSEN

et al.

[IV

show that carbon atoms at nickel step sites are 130 kJ/mol more stable than those at planar nickel sites; the result indicates that carbon effectively blocks step sites in these experiments. Egebjerg et al. (157) determined the activation energy for dissociation of methane on Ni( 1 11) to be 74f 10 kJ/mol. Because of the fast saturation of the more reactive step sites by carbon, the measured sticking coefficients most likely represent values for terrace sites rather than steps. The products of the dissociation of methane on Ni( 111) at 150 K determined with a molecular beam technique were shown to be CH3 and H (158). Formation of CH and 2H has been observed to result from heating 0.05 ML of CHs from 80 to 220 K and cooling again to 80 K (159). At higher a temperature, 320 K, CH radicals form CzH2 species, and at 400 K and high coverages, CbHh is formed. The results of these experiments suggest that the barrier for dissociation of CHs radicals to give CH2 + H is ca. 50 kJ/mol and that the barrier for dissociation of CHz radicals is significantly less. Qualitative information about the thermochemistry of CH,y species can also be obtained from the work of Yang et al. (159). The relative stabilities of the CH, species were determined to be CHs < CH+2H < OSC2H2 +2H < 0.167C6H6+2H and CHz+H<CH+2H. A.2. Curbon Monoxide Numerous investigations have been concerned with the adsorption energy of CO on nickel single crystals (160-164). The initial adsorption energies for CO on various nickel facets are approximately 130 kJ/mol, and the adsorption energy at high CO coverages is in the range 95-101 kJ/mol. Al-Sarraf et al. (163) reported that the adsorption energy of CO on nickel increases dramatically, up to 300 kJ/mol, as a result of potassium promotion. Sinniah et al. (164) determined the effect of steps on the CO adsorption energy, finding the energy difference between a terrace and a step site on Ni(911) to be 2.5f0.8 kJ/mol. However, in the presence of small amounts of hydrogen atoms, the efficiency for CO trapping at the step sites relative to terrace sites increases dramatically. Early work showed that dissociation of CO is highly structure sensitive. For example, flash desorption (at a rate of temperature increase of 16 K/s) of CO from Ni( 111) results in a single peak at 433 K attributed to molecular desorption; in contrast, a stepped surface resulted in an additional peak at 823 K, interpreted as evidence of the associative desorption of CO (165). The interaction of CO with other nickel facets was also characterized (166-169); the dissociative sticking of CO was found to be of the order of 1Om2-1O- for Ni( 11 l), Ni( loo), and Ni( 110) in the temperature range 400-500 K. Steinrtick et al. (I 70) found a dissociative sticking coefficient of CO on a sputter-

tv1

HYDROGEN

AND

SYNTHESIS

GAS

BY

STEAM-

AND

CO2

REFORMING

121

damaged surface that was 20 times the value determined for Ni( 111). This observation suggests that the CO dissociation at step defects sites dominates the overall reaction, in support of the conclusions of Erley and Wagner (16.5) that CO essentially only dissociates at step sites. In all these studies, the overall activation energy for the dissociative sticking of CO on nickel is negative, a result that can be rationalized as a consequence of the binding energy of CO being larger than the activation energy for dissociation of CO bonded to the surface. Goodman et al. (171) reported methanation of CO on Ni( 100) and concluded that the observed reaction rate agreed well with those observed for supported catalysts. The activation energy for CO methanation on Ni( 100) was found to be 103 kJimo1. The roles of sulfur and potassium were also investigated. Sulfur is a strong poison for CO methanation but does not affect the activation energy (I 72). Potassium also decreases the rate of methanation without affecting the activation energy, but potassium is reported to increase the rate of CO dissociation, increase the surface coverage by carbon during the reaction, and increase the content of higher hydrocarbons in the product (173). The experimental results show that CO adsorbs strongly on nickel and the coverage of CO could be significant during reforming catalysis. Experimental evidence also indicates that dissociation takes place at step or defect sites and supports the activation and binding energies obtained by DFT calculations presented below.

A.3. Curhon Dio.xi& Activation of carbon dioxide by nickel has been studied less intensively than activation of CO. CO1 activation on Ni( 100) and on Ni( 110) has been reported (I 74-l 78), as has methanation of CO1 on Ni( 100) (I 7Y). Molecular adsorption of CO:, on Ni( 110) has been observed at low temperatures (e.g., 80 K). Upon heating to temperatures above 100 K, a bent and ionized form of CO:- arises (I 77,178). At temperatures above 230 K, this species dissociates into CO and 0. A dissociative sticking coefficient for CO1 on Ni( 110) at room temperature of ca. 0.1 was determined (I 74). Heating of a Ni( 100) single crystal exposed to CO2 to 300 K gave a peak for desorbed CO at 420 K, and CO1 exposure at 150 K resulted in a CO2 peak at 250 K and a CO peak at 420 K. The binding energy of CO1 is therefore inferred to be so small that the coverage of CO2 is expected to be negligible under reforming conditions. The sticking of CO1 to Ni( 100) was investigated by DEvelyn et al. (176) by a molecular beam technique; the activation energy for the sticking of CO* to Ni( 100) was estimated to be approximately 12 kJ/mol.

122

J.R. ROSTRUP-NIELSEN

et al.

[IV

Peebles et al. (I 79) investigated the methanation of CO2 on Ni( 100) and determined the activation energies for formation of CH4 and CO to be 89 kJ/mol and 73382 kJ/mol, respectively. Sulfur was shown to be a strong poison for the CO2 methanation; the rate of CH4 formation was markedly reduced. Good agreement was found between the rate of methanation of CO2 on a single crystal and on supported nickel catalysts. A.4. Steam The reactivity of nickel surfaces towards water depends on the specific facet of the surface and the number of step sites on the surface. The reactivity ranges from that of the relatively unreactive close-packed (Ill) and (100) surfaces on which water is adsorbed reversibly, to that of the more open (110) and stepped surfaces on which water can chemisorb without clustering and dissociate under some conditions (180-182). Kasza et al. (180) found that the dissociation of water is much faster on Ni(760) than on Ni(ll0) and that step sites act as sinks for OH radicals, indicating that the binding energy of OH at steps is significantly higher than at terraces. Benndorf et al. (181,183) studied the adsorption and reactions of H20 on Ni(l1 l), Ni(221), and Ni(665). On the more open surfaces they observed new desorption states at higher desorption temperatures, and these were not present on Ni( 111). Two of these new states were related to recombination of products of the water decomposition. The effects of sodium addition to Ni(775) on water adsorption and reactivity to were investigated by Mundt and Benndorf (181). At low sodium coverages (ha < 0.03) the step-induced desorption states disappeared. The authors suggested that this observation was evidence of sodium decoration of step sites. At higher sodium coverages, an cy HIO-desorption state was observed (at 225 K), and at even higher sodium coverages, the terrace desorption state disappeared until only a sodium-induced state remained. The coverage of nickel surfaces by water molecules is not expected to play an important role, but OH radicals and oxygen atoms may have significant coverages. A.5. Hydrogen Hydrogen is known to react dissociatively with all nickel facets and steps, even at room temperature (96,167,184-188). Initial hydrogen sticking coefficients in the range of 0.1 to 1 have been observed for nickel at gas temperatures between 100 and 700C (or the equivalent molecular beam normal kinetic energies) (96,167,186,188). The activation energy of dissociative adsorption of hydrogen was determined by molecular beam

IV1

HYDROGEN

AND

SYNTHESIS

GAS

BY STEAM-

AND

CO:

REFORMING

123

experiments with Ni( 1I l), Ni( loo), Ni( 1 lo), and Ni(445) to be 9.6 (I&S), 5-5.9, (187), 3.3 (186,187) and 0 (185) kJimo1, respectively. The activation energy for associative desorption of hydrogen can be determined by modeling of the Hz TPD spectra reported by Christman et al. (184) and Rendulic et al. (18.5). Assuming a pre-exponential factor of IO s , the activation energies of desorption of Hz from Ni( 11 l), Ni( loo), Ni( 1 lo), and Ni(445) were determined to be 96, 90, 89, and 87 kJ/mol, respectively. On the basis of these values, it seems reasonable to conclude that the binding energy of hydrogen to nickel is close to 86 kJ/mol, in satisfactory agreement with the values 90-96 kJ/mol determined by Christmann et al. (184). This adsorption energy is probably sufficient to make hydrogen coverages significant during reforming. A.6. Methane Activation in Mrthanc~-H?~ultl Mixtures

The dissociation of methane on nickel catalysts was characterized with isotopic tracers (69, IHY, 190.84). The reactant gas fed to a backmixed flow reactor was 5% CH4 in D2 and the steady-state concentrations of CHd, CH3D, CH~DZ, CHD3, and CD4 in the product were measured. Data are shown in Fig. 36 for CHa conversion in the presence of a NiiMgAI204 catalyst and a 1.2% K/Ni/MgA1203 catalyst. It is evident from these data that potassium promotion reduces the activity dramatically. The concentrations of CHq, CHjD, CHzDz, CHD3, and CD4 as a function of temperature were analyzed on the basis of a microkinetics model to estimate rate constants for

350

400

450

Temperature

(C)

PI<,. 36. Results of experiment with -200 mg of a Ni/MgAlzOA catalyst or a K;NiiMgAl>Od catalyst in a CSTR with a flow of 5 mlmin of 5% Cl IJ in Dz. The steady-state concentration of CH4 in the product is shown as a function of the reaction temperature. The total pressure was I .4 bar.

124

J.R. ROSTRUP-NIELSEN

et al.

[IV

the dehydrogenation was used:

of methane on the nickel surface. The following

model

cx4+2*

2 k-1

cxj*+x*,

(23)

cxj*+*
cx2*+*

k2 ++ cx2* +x*,
k2 k3

(24)

cx* +x*,
2x*,

(25)

k-3 x*+2* where (27) ; (26)

k-4

(28)

k-2 = 2x 102sm exp

77.5 kJ/mol RT

(30)

k3 = 1.4x 1013smexp(-28:Fo)

(31)

km3 = 3.7~10~~~ exp -

75.5 kJ/mol RT

>

(32)

k4 = 2n10s~bar~exp(-5k~~),

(33)

kPq = 103sm exp(-91

kJJo)

(34)

Four parameters were adjusted to fit the model to the data. They were the two parameters describing the rate constant k, and the ratio of the pre-exponential

IV

HYDROGEN

AND

SYNTHESIS

GAS

BY STEAM-

AND

COz

REFORMING

125

factors of the back reaction (23) and the forward reaction (24) A-i/2, and the difference between the activation energies for these two reactions, Em I - E2. All the other parameters were estimated a priori. The experimental results could be modeled successfully only when it was assumed that the H:D ratio in the *CX2 species is the same as the H:D ratio in the hydrogen/deuterium gas in contact with the catalyst. This result implies that the rates of the forward reactions (25) and (26) are high compared to that of the reverse reaction (25). The rate constants of the reverse reaction (24) the forward reaction (25), and the back reaction (25) determined by Watwe et al. (191) meet the criteria and are used here. The pre-exponential factor for reverse reaction (26) was assumed to be 10 s- , and the average of the values of E-4, discussed previously (9 1 kJimo1) was used. The pre-exponential factor for the forward reaction (26) was assumed to be 2 x 10 s bar, and Ea was chosen as the average of the values discussed previously, 5 kJ/mol.

Catalyst

A, (bar-

SC )

El (kJ!mol)

A-,/A2

E-1 -E? (kJhol) IO 3 IO 7 17.0 -12.X

Ni/MgAl104 K!Ni/MgAI,O1

2x107 3x IO

53.1 X6.7

2.4x 5.Xx

The lines in Fig. 36 are results of the microkinetics model with the values given in Table XV. The table shows that the activation energy for sticking of methane increases by 34 kJ/mol as a result of promoting the catalyst with 1.2% K. Interestingly, the pre-exponential factor of reaction (23) also increases by a factor of 15 as a result of potassium promotion. The increased pre-exponential factor can be explained by the DFT calculations presented below. These calculations show that potassium will block the step sites preferentially and that the activation energy is higher by 27 kJ/mol for reaction on Ni( 11 1) than for reaction on Ni(2 1 l), in good agreement with the increase in the activation energy observed here. Furthermore, when potassium blocks the few highly active step sites, which normally determine the activity of the catalyst, the less active sites on the nickel planes determine the overall activity. The increased number of active sites will give a higher pre-exponential factor. Experiments therefore support the idea that step sites are much more reactive for reforming than terrace sites.

126

J.R. ROSTRUP-NIELSEN

et al.

[IV

B. THE MECHANISM OF STEAM REFORMING The many experiments concerning steam reforming and the individual elementary steps entering into the reaction have provided detailed insights into the reaction, but it was not until the detailed density functional theory (DFT) calculations of Bengaard et al. (69) were reported that a consistent picture of the process began to emerge. In this work, the full potential energy diagram of the reaction on two different nickel surfaces was considered. The two surfaces are shown in Fig. 37. One is the most close-packed Ni( 111) surface. This is the lowest-energy surface and therefore the most abundant. Thus, it is also the most studied, and a number of researchers have considered the first step in the steam reforming, methane activation, on this surface (/91&19Y). The other surface considered is Ni(21 l), which in addition to (111) facets exposes (100) steps. It was chosen to investigate the possible influence of steps and other defects on the reactivity.

Fir;. 37. Structure of the (left) Ni( I 1I ) and (right) Ni(2 1 I ) surfaces used in density functional theory studies of the reaction mechanism and energetics on flat and stepped surfaces.

The energetics of the full reaction are shown in Fig. 38. This figure shows the energies of intermediates on the surface and activation barriers separating the intermediates along the reaction path. The structures of the reaction intermediates for both surfaces are shown in Figs. 39 and 40. On both the close-packed surface and the steps, the highest barriers are found for the reaction step in which CO is formed (C* + O+ + CO* + *). Since the attempt frequency in an Arrhenius expression for the rate constant k = ~8~~~~ for the first dissociation step is much lower than that for a surface reaction such as CO formation, CH4 dissociation may also be the slowest step in the process under some conditions (200). The steps are much more reactive than the close-packed surface. All intermediates are also much more strongly bound at the steps than on the terraces. There should therefore be (at least) two different reaction channels, one with a low activation barrier, which is associated with steps, and another associated with terraces. The latter channel is characterized by a higher barrier, but there will be more free active sites in this channel, both because

IV1

HYDROGEN 300

AND

SYNTHESIS

GAS

BY STEAM-

AND

CO2

REFORMING

127

200 z f c w 100
7+4

Ni(1

-100

Ni(211)

OH,C,5H

6H steam reforming on the Ni( I I I) and in Figs. 39 and 40. All cnegies are gas phase far from the clean surface. is found to be 292 kJ/mol. If this is kJimol, which is in good agreement

FIG. 38. Calculated energies along the reaction path for Ni(21 I) surfaces. The corresponding structures arc shown given relative to a situation in which all reactants arc in the The calculated reaction energy (the final point in the figure) converted into a reaction enthalpy at 298 K, it becomes 230 with the experimental value of 206 kJimol(69).

CH,

(gas)

CH,

CH,

CH,

CH,

CH,

CH

C+H

CH

CH,

CH

CH,

FIG. 39. processes in surface. The including the by use of the the gas-phase

Calculated structures of intermediates and transition states for all elementary the transformation of methane into adsorbed carbon and hydrogen on a Ni( I I I) structure of the intermediates is the result of a complete total energy minimization adsorbate and surface degrees of freedom. The transition states have been located nudged elastic band method (ZOO). In all cases except the first picture representing molecule and the clean surface. the (I I I ) surface is viewed from above.

128

J.R. ROSTRUP-NIELSEN

et al.

[IV

CH, (gas)

CH,

CH,

CH3

CH,

CH,

CH

C+H

CH

CH,

CH

CHz

FIG. 40. Calculated structures of intermediates and transition states for all elementary processes in the methane activation on the step sites of a Ni(21 I) surface. The structures of the intermediates are the results of a complete total energy minimization including the adsorbate and surface degrees of freedom. The transition states have been located by use of the nudged elastic band method (200). In each case the surface has been rotated for the best view of the structure of the intermediate or the transition state.

there will usually be more terrace sites than step sites and because the coverage with intermediates is smaller on the terrace sites. C.
CARBON FORMATION

It is clear from Fig. 38 that adsorbed atomic carbon is much more stable at the steps than at the terraces. Consequently, the steps should be better nucleation sites for graphite formation than the terraces. Adsorbed carbon can form at the step, and a graphene layer can then grow from the step, as illustrated in Fig. 41. In Fig. 38, the energy of a graphene layer on a Ni( 111) surface is included as a dashed line (the graphene is most stable on the (111) surface because there is an almost perfect registry between the graphene layer and the nickel atoms on the surface). Because this line is below the energy of the most stable form of adsorbed carbon on the surface (at the step), there is a driving force for graphene formation. The energy gain evident from Fig. 38 is calculated by assuming an infinitely large graphene island. An island of finite size will be less stable as a result of the energy cost associated with the perimeter of the island (in fact the edge towards the step will be more stable, but most of the perimeter will be away from the step and will be less stable). The island must be large enough that the energy gain

IV1

HYDROGEN

AND

SYNTHESIS

GAS

BY STEAM-

AND

CO? REFORMING

129

FK,. 41. Illustration

of a graphene

island

nucleated

from

a step on a Ni( I I I ) surf&x

(6Y)

FIG. 42. Schematic nickel particle during

illustration of the process steam reforming.

by which

graphite

tubes

are formed

around

per carbon atom outweighs the perimeter energy. As the diameter D of the island increases, the number of perimeter carbon atoms relative to the total number of carbon atoms decreases as l/D. Graphene formation is therefore energetically advantageous only if an island of sufficient size is nucleated. Given the energetics indicated by the DFT calculations, the critical island size above which the graphene island is more stable than carbon atoms adsorbed along the step is of the order of 25A (200). The implication is that only particles of a certain size can support graphene nucleation, in agreement with the experimental observation that small nickel particles (less that about 50 A in diameter) do not show graphite formation, as discussed in Section 1I.F. After a graphene island is nucleated, the growth may continue by transport of carbon atoms to the island. The graphene may eventually cover the whole crystallite. New layers may nucleate at the step by pushing the already formed layer(s) further out (Fig. 42). There is an extra small energy gain associated with the formation of graphite-like layers associated with the weak interlayer interaction. The carbon atoms are so much more stable at a step than on

130

J.R. ROSTRUP-NIELSEN

et al.

[IV

the flat (111) surface that it is energetically advantageous to form new steps. New nucleation sites can therefore be formed spontaneously during the steam reforming process, and the system has the freedom to rearrange the nickel particles to best form graphene layers. These arguments suggest that the availability of step sites is important both for a high turnover rate and for graphite formation. This raises the question of where promoters such as potassium, sulfur, and gold are located on the surface during turnover of the catalytic reaction.

D.

PROMOTION

Bengaard et al. (69) used the DFT calculations to evaluate the stability of various promoters such as potassium, sulfur, and gold on nickel surfaces. They found that both sulfur and gold are considerably more stable at a step than on a terrace, and that potassium is more stable when it is bound to oxygen along the step (Table XV). The potassium was found to be most stable as -K-O-K-Orows along the steps. Similarly, it was found from the calculations and experiments that gold in the nickel surface is attracted to step edges. Thus, it was suggested that the major carbon-preventing effect of these promoters is to block the steps and hence remove the nucleation sites for graphite formation. The potassium and sulfur promotion result as these adsorbates physically block the surface (but there may also be longerrange effects along the surface (201,202)). In the case of gold, it segregates preferentially to the step sites (Fig. 43), where it destroys the ability of neighboring sites to adsorb carbon, as discussed by Besenbacher et al. (149). The steps also adsorb nitrogen considerably more strongly than the terraces (Table XVI). The calculated energetics suggest that at room temperature only step adsorption will be possible. N2 adsorption should therefore titrate the most active reaction sites, in accord with the experimental observations discussed in Section 1II.B. Addition of the promoters should decrease the activity of the catalyst (cf. the discussion in Section I1I.E). The decrease should be determined by the promoter coverage at the steps. The promoters need not cover all step sites to prevent graphene nucleation, because a graphene island of a finite size is needed for it to be stable. Promotion can therefore hamper graphite formation without destroying the activity completely. In this respect the sulfur and gold decoration are more effective than potassium decoration, because sulfur and gold will spread out along the step, whereas the interaction with the oxygen in the case of potassium leads to attractive interactions between the potassium atoms. The addition of less than an almost complete step coverage will lead to a fraction of the steps being completely covered while the rest are free and

IV

HYDROGEN

AND

SYNTHESIS

GAS

BY STEAM-

AND

CO1

REFORMING

131

FIG. 43. Snapshot of a nickel particle segregate to the steps at the surface where nucleation.

containing 5% gold. The gold atoms preferentially they prevent adsorption of carbon and thus graphcnc

Surface

Kio

Adsorption

cncrgy S -162 -210

(kJ/mol) N2 -8 49

Au cncrgy, kJ/mol Y 0 -36

Ni( I I I) Ni(2I I)

-131 -206

The energy shown for sulfur adsorption HlS(g) + Ni* - Ni-S + Hz(g). The last column the surface at a step rclativc to that in a terrace represent adsorption at the step site (6Y).

is that for the reaction shows the energy of gold in site. The results for Ni(21 I)

open for synthesis. This means that promotion will decrease the activity, but the activation energy should be unaltered because the nature of the active site does not change (see Section 1I.F and Table XII). It is interesting that there is a universal explanation of the effect of several different promoters, including sulfur, alkalis, or gold, which all suppress graphene formation by blocking the nucleation sites. The inhibition will work as long as the promoter is present in sufficient quantities at the steps. If the chemical potential of carbon is high enough, the promoter may be displaced

132

J.R. ROSTRUP-NIELSEN

et al.

[V

from the step and graphite formation may start. The promotion will therefore appear as a shift in the carbon limit as discussed in Section 1II.E. The competition between the promoter and carbon at the nucleation sites can be illustrated simply for the case of sulfur promotion. The inhibition of carbon formation by sulfur works only at high sulfur coverages (0s z 0.7) corresponding to a HzS/H2 ratio of 2.6x 1O-6 at 850C (14), far in excess of what is expected to completely saturate step sites. However, the competition between adsorbed carbon and sulfur atoms is reflected by the overall reaction: CH4+S*=C++H2S+H2. (35)

It is interesting to compare an order of magnitude estimate of the experimental value of the equilibrium constant for this reaction with those obtained from the work of Bengaard et al. (69) at nickel step defect sites and at nickel planes. Experimentally, the onset of carbon formation was found to be at partial pressures of HZ, HxS, and CH4 of ca. 0.32, 8.3x lo-, and 0.36 bar, respectively, at 850C. Assuming that &/& z 1, it is estimated that Kexp (the equilibrium constant for Eq. (35)) zz 7x 10m7bar. Using the adsorption energies determined in ref. (69) and standard reaction entropies for the following two reactions of 96 and 50 kJ/mol, respectively, it can be estimated that Ksteps= 1.1 x 1O- bar and K,,,,, = 1.4x IO- bar, respectively, at 85oOC: CH4 + 5* = C+ + 4H+, H2S + * = S* + HZ. (36) (37)

The estimate of the equilibrium constant obtained for step sites is close to that determined experimentally, whereas a much lower value is expected at planar sites. On the basis of these calculations, it seems reasonable to conclude that the onset of carbon formation is determined by the relative coverage of steps by carbon and sulfur.

V. Conclusions Steam reforming has been a well-established process in industry for more than 70 years, and it will play an important role in future applications related to gas-to-liquids conversion and to a possible new hydrogen economy. The process may appear simple from an overall consideration, as the product composition is determined by simple thermodynamics, but in reality it is a complex coupling of catalysis, heat transfer, and mechanical design. There are still major tasks ahead to improve reforming technology, as it

HYDROGEN

AND

SYNTHESIS

GAS

BY STEAM-

AND

CO2

REFORMING

133

remains capital intensive. It remains a challenge to take advantage of the huge surplus of catalytic activity in present reformer designs, and there is still room for developing catalysts to better withstand the risks of carbon formation and sulfur poisoning. Recent studies of the fundamentals of the steam and CO1 reforming reactions have led to a more consistent understanding of the mechanism of the main reactions and the competing reactions leading to carbon formation. This knowledge can be related to the structure of the catalyst and the function of promoters. In-situ high-resolution electron microscopy has provided new information about sintering mechanisms and the importance of steps in nucleation of whisker carbon. DFT calculations have quantified the energetics of methane activation and shown that activation barriers are smaller on surface steps where carbon is the most stable surface species. The better understanding of the mechanism should result in a stronger microkinetic description of the steam reforming reaction. The steam reforming process is well documented on the basis of industrial experience and through numerous experimental investigations of the catalysts and the process. The empirical knowledge forms a jigsaw puzzle, and it has been illustrated how this can be put together through a closely coupled experimental and theoretical effort. The key to success is the identification of the active sites for both the process and the formation of carbon whiskers. With this knowledge, one can not just describe the process, one can also develop a universal understanding of the effect of many different kinds of promoters. The central concept is the role of surface defects as the source of both reactivity and as the nucleation center for whisker carbon formation. This work has illustrated the benefit of close interaction between industrial catalysis and basic research. In general, we believe that this interaction, often represented by industry/university collaboration, should remain in balance. It should be a two-way process, and industry needs scientists involved in basic and exploratory research. We need the courage to formulate new concepts of catalysis and accept the risks of failure.

References
1. Rostrup-Nielsen, J.R.. Cural. rOdq 71, 243 (2002). ,?. Christensen, T.S., Sew Christensen, P., Dybk.jax. 1.. Bak Hansen, J.-H.. and Primdahl. S/d swfe Sri., Cu/ul. 119, 883 ( 1998). 3. Schulz, H.. AppI. Curul. A 186, 3 (1999). 4. Topp-Jargensen. J., Stud. Szw$ Sri. Cutal. 36, 293 ( I988 ). 5. Yurchak, S.. SILK/. .%/I$ Sci. Cutul. 36, 25 I (I 988). 6. DybkjEr, I.. and B&id Hansen. J., Stuck. Surf Sci. C~rtal. 107, 99 (1997). 7. tlarms, A., Hiihlein, B., Jsrn, E.. and Skov, A.. Oil Gas 1 78, ( 15) I20 (1980). I.].,

134

J.R. ROSTRUP-NIELSEN

et al.

8. Aasberg-Petersen, K., Bak Hansen, J.-H., Christensen, T.S., Dybkjrer, I., Seier Christensen, P, Stub Nielsen, C., Winter Madsen, S.E.L., and Rostrup-Nielsen, J.R., App/. Cutul. A 221, 379 (2001). 9. Rostrup-Nielsen, J.R., in Proc. 15h World Petroleum Congr. 1998, Vol. 11, p. 767. Wiley, New York, 1998. IO. Lange, J.P., Ind. Eng. Chem. Res. 36, 4282 ( 1997). II. Crabtree, R.H., Stud. Surf: Sci. Cutul. 81, 85 (1994). I2. Yarlagadda, O.S., Morton, L.A., Hunter, N.R., and Gessar, H.D., Comhust. Flume 79, 216 ( 1990). 13. Rostrup-Nielsen, J.R., in NATO AS1 Combinatorial Catalysis and High Throughput Catalyst Design and Testing (E.G. Derouane et ul., Eds.), p. 337. Kluwer, Dordrecht, 2000. 14. Rostrup-Nielsen, J.R., in Catalysis, Science and Technology (J.R. Anderson and M. Boudart, Eds.), Vol. 5, ch. 1. Springer, Berlin, 1984. 15. Rostrup-Nielsen, J.R., and Rostrup-Nielsen, T., Cuttech 6(4), 150 (2002). 16. Rostrup-Nielsen, J.R., Dybkjax, I., and Aasberg-Petersen, K., ACS Preprints Petr: Chem. Diu. 45 (2), 186 (2000). 17. Rostrup-Nielsen, J.R., Phvs. Chem. Chem. Phys. 3, 283 (2001). IX. Rostrup-Nielsen, J.R., C&al. Toduy 63, 159 (2000). 19. Rostrup-Nielsen, J.R., Stud. Surf: Sci. Cut&. 21, 257 (1984). 20. de Jong, K.I?, Cutal. Todq 29, 171 ( 1996). 21. Marion, C.I?, and Slater, W.L., in Proceedings of the 6h World Petroleum Congress, Frankfurt 1969, Vol. 3, p. 159. 22. Pitt, R., &r/d R+zing 11(l), 6 (2001). 23. Ernst, W.S., Venables, S.C., Christensen, P.S., and Berthelsen, A.C., Hvdrocurbon Process. 79(3), 100-c (2000). 24. Christensen, T.S., and Primdahl, I.]., filvdrocurbon Process. 73(3), 39 (1994). 25. Bodke, AS., Bharadwaj, S.S., and Schmidt, L.D., J. Cutul. 179, 138 (1998). 26. Basini, L., Aasberg-Petersen, K, Guarnoni, A., Aragno, A., and Ostberg, M., Cutul. 7ijdu.v
64, 9 (2001). 27. 28.

Dybkjcr. Mazanec,
136,

I., and Christensen, T.S., Stud. Surf Sci. Cutul. 136, 435 (2001). T.J., Prasad, R, Odegard, R, Steyn, C., and Robinson, E.T.. Stud. SW/~ Sci. Cutul.

147 (2001).

29.
30.

31. 32. 33. 34. 35.


36.

Udovich, C.A., Stud. Suet .Sci Curul. 119, 417 (1998). Dybkjax, I., and Winter Madsen, S., lnt. 1 Hydrocarbon Eng. 3(I), 56 (1997/1998). Holm-Larsen, H., Stud. Su$ Sci.Cutul. 136, 441 (2001). Rostrup-Nielsen. J.R., Stud. Swf Sci. Cutal. 81, 25 (1994). Rostrup-Nielsen, J.R., and Christiansen, L.J.. Appl. Cutul. A. 126, 381 (1995). Selimovic, A., Palson, J., and Sjunneson.L., in Abstracts, Fuel Ccl1 Seminar 1998, Palm Springs, CA., p. 88. Rostrup-Nielsen, J.R., Aasberg-Petersen, K., and Schoubye. l?, Stud. Su$ Sci. Cut&. 107,
473 (1997).

37. 3X. 39.


40.

41.

Rostrup-Nielsen, J.R., Dybkjax, I., and Christiansen, L.J., in NATO AS1 Chemical Reactor Technology for Environmentally Safe Reactors and Products (H.J. de Lasa et al., Eds.), p. 249. Kluwer, Dordrecht, 1992. Rostrup-Nielsen, J.R., Stud. Surf: Sci. Cutul. 36, 73 (1988). Mohri, T., Takemura, K., Shibasaki, T., Ammoniu Plant. SqjI 33, 86 (1993). Christensen, T.S., Appl. Cut&. A. 138, 285 (1996). Rostrup-Nielsen, J.R., Dybkjrer, I., Christensen, T.S., Stud. Surfi SC;. Cutul. 113,81 (1998). Dybkjax, 1. in Ammonia (A. Nielsen, Ed.), p. 199. Springer, Berlin, 1995.

HYDROGEN

AND

SYNTHESIS

GAS

BY STEAM-

AND

CO,

REFORMING

135

32. Rostrup-Nielsen, J.R.. and Christiansen, L.J., i/f Chemical Reaction and Reactor Design (H. Tominaga and M. Tamaki, Eds.), ch. 5.2, p. 247. Wiley. New York, 1998. 43. Rostrup-Nielsen, J.R., Bak Hansen. J.-H.. and Apariclo, L.M., J: J/x?. /+f/: /n.~/. 40, 366 ( 1997). 44. Lzgsgaard-Jsrgenscn. S.. llc?jlund Nielsen. PE.. and Lchrman. P.. Cut~/. 7&k+, 25, 303 ( 1995). 4-i. Kikuchi, E., Cutal. 7&/ux 25, 333 (1995). 46. Aasbcrg-Petersen, K., Stub Nielsen, C., and L.regsgaard-JPrrgenscn. S.. (u&/l. To&~. 46, 193 (199X). 47. Rostrup-Nielsen, J.R.. Catal. Gx$ 18, 305 (1993). 48. Olsbyc, O., Dahl, I.M. Slagtern. A.. and Blom. R., i/l Proceedings of the First European Congress on Chemical Engineering, Firenrc, May 1997, Vol. I, p. 367. European Federation of Chemical tngincers (EFCE). 1997. 4Y. Richardson, J.T.. Paripatyadar, S.A.. and Shcn. J.C.. AIChE ./. 34, 743 (198X). 50. Rostrup-Nielsen, J.R.. Christiansen. l..J., and Bak Hansen. J.-H., .3&. Cc/tu/. 43, 287 (1988). j/. Xu. J., and Froment, G.F.. AIC/lE J. 35, 88 (1989). Testing of a High s:. Udcngaard, N. R.. Christiansen, L. J.. and Summers, W. A., Endurance Efficiency Steam Reformer for Fuel Cell Power Plants. IzPRI AP-6071, Project 2192-l (198X). 53 Richardson, J.T.. and Tudigg, M.V., St&. Sw/. Gi. Ccrrc~l. 91, 345 (1995). 54. Christianscn. L.J. (unpublished results). 55. Rostrup-Nielsen, J.R.. ./. Mol. Cutcrl. il 163, I57 (2000). jh. Khomcnko, A.A., Apelbaum. L.O.. Shub. F.S., Snagovskii. Y.S.. and Temkin. M.I.. Ki/?c,f. CutuI. 11, 1233 (1970). 57. Wise. H.. McCarty, J.. and Oudar, J. it7 Deactivation and Poisoning of Catalysts (J. Oudar and H. Wise, Eds.). Marcel Dckkcr. New York. 1985. .jcI. Bartholomew, C.H.. Agrawal. PK.. and Katzcr. J.R.. Ah. Cc~tul. 31, I35 (1982). SY. Topsw, H., Clausen, B.S., and Massoth, F.E.. in Catalysis. Science and Technology (M. Boudart and J.R. Anderson, Eds.). Vol. I I. Sprlngcr. Berlin 1996. 60. PiwctL. M., Larsen. J.S.. and Chrlstenscn, T.S.. in Proceedings of the 1996 Fuel Cell Seminar. p. 780. I Y96. 61. Zdonik. S.B., Green, E.J.. Haller. L.P.. Oil Grrv .I 65 (26), 96 (1967). h-7. Alstrup, 1.. Tavares, M.T., Bernardo. C.A.. Sarrensen. 0.. Rostrup-Nielsen. J.R.. I24otr: (or/: 49, 367 (IY9X). 63. Rostrup-Nielsen, J.R., St& &r/: %i. Chl~l/. 139, 1 (2001 ). 64 Grenga, H.E., and Lawless. K.R., J A&. Ph?:s. 43, 1508 ( 1972). hi. llanscn. P.L., Wagner, J.B.. and Helweg. S. (unpublished results). 66. Snocck. J.-W.. Froment. G.F., and Fowlcs. M.. J Ctrtul. 169, 250 ( IY97). 67. Alstrup. I.. and Ta\ares. M.T.. J. Chltr/ 135, 147 (1992). 6X. Schcsted. J. (unpublished results). hc). Bcngaard. H.S.. Norskov. J.K., Sehcsted, J.S., Clausen. B.S.. Nielsen. L.P.. Molenbrock, A.M., and Rostrup-Nielsen. J.R., .I Ctrrcll. 209, 365 (2002). 70. Bernardo, C.A.. Alstrup, 1.. and Rostrup-Nielsen. J.R.. J. Cutal. 96, 5 I7 (19X5). 7/. Alstrup, I.. J: (tr/o/. 109, 241 (1988). 72. Rostrup-Nielsen. J.R., and Bak Hansen, J.-H.. .I Cutal 144, 3X (1993). 73. Lobo. L.S.. and Trimm. D.L.. .I Cirtrrl. 29, 15 ( 1973). 74. Rostrup-Nielsen, J.R., ./. Catal. 85, 31 (19X4). 75. Alstrup, I.. Clausen. B.S., Olsen, C., Smits. R.H.II., and Rostrup-Nielsen, J.R., Sllrtl. .S/r~-f .SC~i. Cuttr/. 119, 5 ( 199X).

136

J.R. ROSTRUP-NIELSEN

et al.

76. Udengaard, N.R., Bak Hansen, J.-H., Hanson, D.C., and Stal, J.A., Oil Gas J 90(10), 62 (1992). 77. Andersen, M.T., Topsme, F., Alstrup, I., and Rostrup-Nielsen, J.R., 1 C&II. 104,454 (1987). 78. Dibbern, H.C., Olesen, P., Rostrup-Nielsen, J.R., Tottrup, PB.. and Udengaard, N.R., Hydrocarbon Proces.~. 65( l), 7 1 ( 1986). 79. Grabke, H.J., Mater. Corr: 49, 3 17 ( 1998). HO. Rostrup-Nielsen, J. R., Steam Reforming Catalysts. Danjsh Technical Press, Copenhagen, 1975. 81. Bodrov, I.M., Apclbaum, L.O., and Temkin, M., Kim/. Cat&. 5, 614 (1964). 82. Khomenko, A.A., Apelbaum. L.O., Shub, F.S., Snagorskii, Y.S., and Temkin, ML., Kin&. Catal. 12, 367 (1971). 83. Tottrup, PB., Appl. Catal. 4, 377 (1982). X4. Aparicio, L.M., 1 Catal. 165, 262 (1997). 85. Boudart, M., and Djega-Mariadassou, G., Kinetics of Heterogeneous Catalytic Reactions, p. 78. Princeton University Press, Princeton, 1984. 86. Bak Hansen, J.-H. (unpublished results). 87. Avctnisov, A., et al. (to be published). 88. Bodrov, I.M., and Apelbaum, L.O., Kinet. Catal. 8, 326 (1967). 89. Bradford, M.C.J., and Vannice, A.M., Appl. Catal. A. 142, 73 ( 1996). 90. Sehcsted, J., Carlson, A., Janssens, T.VW., Hansen, PL., and Datyc, A.K., .1. Cafal. 197, 200 (2001). 91. Tcixeira, A., and Giudici, R., Chm. Eng. Sci. 54, 3609 (1999). 92. Borowiecki, T.. Denis, A., Nazimek, D., Grzegorczyk, W., and Barcicki, J., Chem. Siosow: 27, 229 (1983). 93. Rasmussen. F.B., Molenbroek, A.M., Clausen, B.S., and Feidenhansl, R., 1 Catal. 190, 205 (2000). 94. Bartholomew, C.H., Cutal. Left. 7, 27 (1990). 95. Christmann, K., Z. Nutwfbrsch. A 34, 22 (1979). 96. Winkler, A., and Rend&, K.D., Surf: Sci. 118, 19 ( 1982). 97. Christmann, K., Penka, V, Behm, R.J., Chehab, F., and Ertl, G., Solid Srute Commurz. 51, 487 (1984). 9X. Rostrup-Nielsen, J.R.. 1 Catal. 11, 220 ( 1968). 99. Pannell, R.B., Chung, K.S., and Bartholomew, C.H., 1 Cutal. 46, 340 (1977). IOU. Bartholomew, C.H., and Pannell, R.B., 1 Catal. 65, 390 (1980). 101. Alstrup, I., Rostrup-Nielsen, J.R., and Roen. S., Appl Catal. 1, 303 (198 I ). 102. Perdereau, M., C.R. Acad. Sci. Paris 267, I 107 ( 1968). 103. Perdereau, M., and Oudar, J., Surf: Sci. 20, 80 (1970). 104. Perdereau, M., Smf: Sci. 24, 239 (197 I ). 105. McCarty. J.G., and Wise, H., 1 Chem. Phys. 72, 6332 ( 1980). 106. Beeck, O., Adu. Catal. 2, 151 (1950). 107. Weatherbee, G.D., and Bartholomew, C.H., 1 Catal. 87, 55 (1984). 108. Oliphant, J.L., Fowler, R.W., Pannell, R.B., and Bartholomew, C.H., J. Caral. 51, 229 (1978). IOY. van Hardeveld, R., and van Montfoort, A., Surf Sci. 4, 396 (I 966). 110. van Hardeveld, R., and Hartog, F., Surf: Sci. 15, 189 ( 1969). 111. van Hardeveld, R., and van Montfoort, A., Surf: Sci. 17, 90 (1969). 112. Nieuwenhuys, B.E., and Sachtler, W.M.H., 1 Colloid Infr+~e Sci. 58, 66 (1977). 113. Arumainayagam, C.R., Tripa, C.E., Xu, J., and Yates, J.T., Surf Sci. 360, 121 ( 1996). 114. Tripa, C.E., Zubkov, T.S., Yates, Jr., J.T., Mavrikakis, M., and Norskov, J.K., 1 Chem. Phyx 111, 8651 (1999).

IIYDROGEN

AND

SYNTHESIS

GAS

BY

STEAM-

AND

CO?

REFORMING

137

I IS.

Quick.

A.,

Browne,

V.M.,

Fox,

S.G..

and Hollins,

P, Szrr$ SC;. 221,

4X (1989).
P/7ys.

/ 16. Yoshinobu, I/ 7. RIchardson, I/K. /I). /20. Kuo, H.K.. Bartholomew, Bartholomew.

J., Zenobi, R., Xu. J.. Xu, Z.. and Yates, Jr., J.T., .1. Chem. J.T., and Grump, J.G., J. Ccltul. 57, 4 I7 ( 1979). Gancsan, C.H.. C.H., P.. and De Angelis, and Sorensen, W.L.. A/>/,/. Cutul. A 107,

95,9393

(1991

).

R.J., ./. Ca,tr/. 64, 303 (1980). .I Cutul.81, 131 (1983). 1 (1993). 17, 567 379

IZI. Young, D.J.. Udaja. P., and Trimm, D.L.. Stutl. Slr$ S<,i. Cutrrl. 6, 33 I (I 980). 1-73. Borisova. M.S., Fenclov, V.B.. Drisko. VA.. and Simonova, L.G., Ktnrt. Crrrtrl. ( 1976). 123. Slagtern. (1997). A., Olsbye, C.G.. C.G., U.. Blom. R.. Dahl. R.A.. R.A.. I.M., and Fjcllv& H.. &/J/.

Ccrttrl. .4 165,

124. Granqvist. 1175. Granqvist, 116. /.?7. 12X. 129. 130. /3/. Granqvist, De Smet. Wanke. Fuentes.

and Buhrman, and Buhrman,

Appl
./I

PhJ:v. Left. 27, 693 (1975).


477 (1976). (1997). (C.fI. Bartholomew and 46, 23X (1977). Lur7puiu 13, 6884 Deactivation

Cutal. 42,

C.G.. and Buhrman, R.A.. J Cutrrl. Y., Deriemaekcr, L.. and Finsy. R.. S.E., .1. (utul. 46, 234 (1977). G.A.. and Gamas, E.D.. 1~1 Catalyst

J.B. Butt. Eds.). p. 637. Elsevier. Amsterdam, 1991. Hughes. R.. Deactivation of Catalysts. Academic Press, Hammer, B., Hansen. L.B.. and Ncrrskov. J.K.. P/gx. Kru. York. A.P.E., Brungs, A.J.. ./I C&a/. 180, 85 (1998). Marquez-Alvarez.

London. 1984. B. 59, 7413 ( 1999). c.. Sloan, J.R., 5-O J., Tsang. .I S.C.. 201, and 206

1.33. Claridge, J.B., Green. M.L.H..

1.13. Schested. J., Jacobsen. C.H.J.. Rokni. S., and (2001). /34. Tracy, D.. Xuc. E.. Ross, J. H. R., EUROPACAT

Rostrup-Nielsen, 5, Abstract

(h/a/.

IO (2001).

135. Tsuji. M.. Toshihiro, M., and Naito, S.. Cutul. /.etr. 69, I95 (2000). 1.36. Zhang. Z., and Verykios. X.E., Ctrttrl. Left. 38, I75 ( 1996). 137 Horiuchi, T., Sakuma. K. Fukui. T. Kubo, Y. Osaki. T.. and Mori. /38. /3Y. I-/O. /4/. 1 I I ( 1996). Seshan. K., ten fsargc. H.W.. tlally. M.A.. W.. van Keulen. &/I/. A.N.J.. 142,

T. .4pp/ J.II.R.,

Cutul. 4

144,

and Ross. 97 (1996).

LStud. .%I,%
65 ( 1996).

sci. Cuttr/. 81, 2x5 (1994). Bradford. M.C.J., and Vannice.

Cutd. A

Efstathiou, A.M., Kladi, A.. Tsipouriari. Bitter. J.ll., Scshan. K.. and Lcrcher,

VA.. and Vcrykios. X.E., ./. J.A.. .I Cutul.171, 279 (1997).

Catul. 158,

1317. Osaki. T.. Masudan, H., Horiuchi, 143. Boro%ieckl. T.. and Golebio\vski. /44. 14-i. 146. /47. 14X. IIY. 150. /.(l /.(2. li.7. /i4

T.. and Mori, T., Crrtul.l~tt. 34, 59 ( 1995). A.. Crrtrrl. Left. 25, 309 (1994). 141 (1997).

Chcng, Z.. Wu. Q., 1.1, J., and Zhu, Q.. Curd. fix/+ 30, 147 (1996). Borowiccki. T., Golcbiowski, A., and S&sinska. B., &/I/. (u/u/. A 153, Bradford. M.C.J.. and Vannicc, M.A., 1 (u/u/. 173, I57 (1998). fianscn. P.L.. Wagner. J.B.. and Helweg. S. (unpublished result\). Irimm, D.L., Bcscnbacher. Nmsko\. Bengaard.

CutuI. Tot/g,
F., Chorkcndorff.

37, 233 I..

(1997). Clauscn,

B.

S.

Hammer.

B..

Molenbrock, J.R..

A.M.. and

J.K.. and Stensgaard, H.S.. Alstrup. I.,

I.. .Scie/x~c, 279, 19 I3 ( 1998). Chorkcndorff: I. Ullmann. S..

Rostrup-Niclscn.

Norcko\. J.K., J. Richardson. J.I.. Stag, S.M.. and Beebe. Jr.. T.P. (19X7). Chorkcndorff.

(trttrl. 1X7, 23X ( 1999). Jung, J-K.. and Zhao. J.. S/M/..SI//% SC.;. Ctrtal. 136. 203 (2001 ). Rcsasco. D.b.. Stlrtl .Sl//./ .G?. Cuttrl. 1 11, 543 ( 1997). Goodman, D.W.. Kay. B.D.. and Yates. Jr.. J.l.. J (11en1. f/~j,.v. X7, 2305 I., and Ullmann. S.. S10.f. S?i. 227. 291 (1996).

I., Alstrup.

138 155. 156. 157. /SK. 15Y. 160. 161. 162. 163. 164. 165. 166. 167. 168. 169. 170. 17/. 172. 173. f74. 175. 176. 177. 178.

J.R. ROSTRUP-NIELSEN

et al.

179. 180. 181. 182. 183. fX4. /X5. 186. 187. 188. IHY. 190. IY/. 192. IY3. 194.

Nielsen, B., Luntz, A.C., Holmblad, P.M., and Chorkendorff, I., Cutul. Lrtt. 32, 15 (1995). Holmblad, P.M., Wambach, J., and Chorkendorff, I., 1 Chem. Phyx 102, 8255 (1995). Egebjerg, R.C.. Ullmann, S., Al&up, I., Mullins, C.B., and Chorkendorff, I., Szrrf: Sri. 497, 183 (2002). Lee. M.B., Yang, Q.Y., Tang, S.L., and Ceyer, S.T., 1 Chum. Ph~s. 85, 1693 (1986). Yang, Q.Y., Maynard, K.J.. Johnson, A.D.. and Ceyer, S.T., 1 Chem. Php. 102, 7734 (1995). Christmann, K., Schober, O., and Ertl, G., J Chem. Phgx 60, 4719 (1974). Takagi, N., Yoshinobu, J., and Kawai, M., Chem. Phw. Lett. 215, 120 (1993). Stuckless, J.T., Al-Sarraf, N., Wartnaby, C.E., and King. D.A., .L Chew. Phw. 99, 2202 (1993). Al-Sarraf, N., Stuckless, J.T., Wartnaby, C.E., and King, D.A., Snrf: Sri. 283,427 (1993). Sinniah. K., Reutt-Robey, J.E., Robinson Brown, A.. Doren, D.J., $1. Chem. Ph?;s. 101, 764 (1994). Erley, W., and Wagner, H., Surf: Sci. 74, 333 (1978). Rosei, R., Ciccacci, E, Memeo, R., Mariani, C., Caputi, L.S., and Papagno, L., J. Cutul. 83, 19 (1983). Stcinriick, H.l?, Rendulic, K.D., and Winkler. A., Su$ Sci. 154, 99 (1985). Astaldi. C., Santoni, A., Della Valle, F., and Rosei, R.. S~rr/. Sci. 220, 322 (1989). Nakano, H.. Kawakami, S., Fttjitani, T., and Nakamura, J., Surf. Sci. 4544456, 295 (2000). Steinrtick, H.P., DEvelyn, M.P., and Madix, R.J., Surf: Sci. 172, L561 (1986). Goodman, D.W., Kelley. R.D., Madey, T.E., and Yates, Jr., J.T., SL~$ Sci. 63, 226 ( 1980). Goodman, D.W., and Kiskinova, M., Surf: Sci. 105, L265 ( 198 1). Campbell, C.T., and Goodman, D.W.. Su$ Sci. 123, 413 (1982). McCarty, J., Falconer, J., and Madix, R.J., J: Cut~zl. 30, 235 (1973). Benziger. J.B., and Madix. R.J., Str$ Sri. 79, 394 ( 1979). DEvelyn, M.P., Hamza, A.V, Gdowski, G.E.. and Madix, R.J.. Szr$ Sci. 167, 451 (1986). Bartos, B.. Freund, H.-J., Kuhlenbeck, H., Neumann, M., Lindner, H., and Miiller, K.. Strrf: Sri. 179, 59 (1987). llling, G.. Heskett, D., Plummer, E.W., Frcund, H.-J.. Somers, J., Lindner, Th., Bradshaw, A.M., Buskotte, U., Neumann, M.. Starke, U., Heinz, K., de Andres, PL., Saldin, D.. and Pendry, J.B., Sut$ Sci. 206, 1 ( 1988). Peebles, D.E., Goodman, D.W., and White. J.M., J. Phw. Chem. 87, 437X (1983). Kasza, R.V, Griffiths. K., Shapter, J.G., Norton, P.R., and Harrington, D.A., Sutyf: Sci. 356, 195 (1996). Benndorf, C., and Mundt, C., J. Lhc. Sci. Technol. A 10, 3026 (1992). Mundt, C.. and Benndorf, C., Surf: Sci. 287/288, 1 19 (1993). NobI, C., and Benndorf, C., Swf. Sri. 182, 499 (1987). Christmann, K., Schober, O., Ertl, G., and Neumann. M.. J. Chem. PhJs. 69,452s ( 1974). Rendulic. K.D.. Winkler, A.. and Steinrtick, H.P., S&. SC;. 185, 469 (1987). Rendulic, K.D., Winkler, A., and Karner, H., 1 J+zL.. Sci Techno/. A 5, 488 (I 987). Hamza, A.V, and Madix, R.J., J: PhJ:c. Clwn. 89, 5381 (1985). Rcndulic. K.D.. Anger. G., and Winkler, A., Surf: Sci. 208, 404 (1989). Kemball, C., Puoc. R. Sot. A207, 539 (1951). Leach, H.F., Mirodatos, C., and Whan, D.A., 1 Curd. 63, 138 (1980). Watwe, R.M., Bengaard. H.S.. Rostrup-Nielsen, J.R., Dumesic, J.A.. and Nerskov, J.K., J. CutuI. 189, 16 (2000). Burghgraef, H., Jansen A.P.J., and van Santen, R.A., &I:$ Sei. 324, 345 (1995). Burghgraef, H., Jansen A.P.I., and van Santen, R.A., J. Circ,m. PhJs. 101, 11012. (1994). Yang, H.. and Whitten, J.L., Surf. Sci. 25.5, 193 (1991).

HYDROGEN 195. IY6. IY7 IYX. IYY. 200. 201. 20-7.

AND

SYNTHESIS

GAS

BY STEAM-

AND

CO:

REFORMING

139

Yang. H., and Whitten, J.L.. J. Ckem. P/I):,. 96, 5529 ( 1992). Siegbahn. P.E.M.. and Panas. I., Surf: .%,i. 240, 37 (1990). Kratzer. P., Hammer, B., and Nsrskov. J.K.. J. Clwm. PIg,.s. 105, 5595 (1996) Au. C.-T.. Liao, M.-S.. and C.-F. Ng,, .I P!~J. Clzeni. /I 102, 3959 (1098). Au. C.-T., Ng. C.-F., and Liao. M.-S., ./ CLIILII. 185, 12 (1999). Bengaard, H.. Thesis. Technical University of Denmark. 2002. Lang. N.D., Holloway. S.. and Nmskov, J.K.. &I/;/: Sci. 150, 24 (1985). Mortensen. J.J.. Hammer. B.. and Norskov. J.K., .%/I$ .Qi. 414, 3 I5 ( 199X).

You might also like