Steam reforming is industrially an important chemical process and is one of the most widely used technologies for converting primary fossil fuels (natural gas, alcohols or liquid"oils") into hydrogen. Essentially steam is mixed with hot gas or vaporised liquid fuel and passed over a supported metal catalyst at an elevated temperature. In the case of methane, the chemical reaction produces a mixture of carbon monoxide and hydrogen (equation 1). Inevitably, some degree of shift reaction (equation 2) occurs within the reformer, yielding a reformate which comprises a mixture of hydrogen and carbon oxides (CO and CO2) together with steam and unconverted methane.

CH4 + H20oC0 + 3H2 (1) steam reforming CO + H2O #CO2 + H2 (2) Water gas shift reaction The product from the reaction is governed by the thermodynamics of the system, that is to say the product species are in thermodynamic equilibrium. The product gas composition is therefore dictated by the steam/hydrocarbon ration of the reactants, by the operating temperature (outlet temperature) and the pressure. The reaction of most hydrocarbons with steam is endothermic and so heat needs to be supplied to the reactor if the outlet temperature is to be similar to the inlet. Since the aim in most cases is to produce a hydrogen-rich gas, and this is favoured by a high reactor outlet temperature, the reactor needs to be externally heated. Such a situation arises in the steam reforming of natural gas, and this is an application of growing importance for fuel cell systems.

Another reaction, which has been referred to as dry reforming or C02 reforming, may be carried out if steam is not available. The reaction in this case is slightly more endothermic than the steam reforming reaction (1) and the proportion of hydrogen in the product is less :- CH4+C022CO+2H2 (3) C02 reforming Note that the product gas from such a reaction is relatively rich in carbon monoxide.

There are many designs of fixed bed catalytic steam reforming reactors which have been proposed for fuel cell systems. These have evolved over the years from conventional externally fired reformers (e. g. for ammonia manufacture) and are preferred for large scale industrial hydrogen production. In nearly all cases, heat for the reforming reaction is provided by a burner, which is located within the reactor assembly. For small scale systems, this has the advantage of providing a relatively compact arrangement with a short distance from the heat source to the reactor wall, see for example EP-A-615949. Nevertheless, in all cases, the burner has to operate at a relatively high temperature compared with the temperature within the catalyst bed, in order to overcome losses in heat transfer through the gaseous medium and reactor wall. Operation of the burner at a relatively high temperature requires a relatively large amount of fuel, potentially produces excessive nitrous oxides and may require expensive materials in the construction of the burner to withstand the high temperatures.

It is an object of the present invention to provide a method of performing a steam reforming reaction and a reactor in which the burner or heater does not have to operate at such high temperatures.

According to a first aspect of the present invention, there is provided a method of performing a steam reforming reaction comprising:- providing a heat source; providing a reaction chamber with a catalyst around the heat source; supplying reactants to the reaction chamber; and wherein the heat source generates heat by the catalytic combustion of fuel.

The use of catalytic combustion enables the reactor to be run at relatively low temperatures, generally below 700°C thus reducing the running costs of the heat source.

Proton exchange membrane (PEM) fuel cells have a particularly tough demand in that they cannot tolerate CO concentrations above a few parts per million (ppm). The reason for this is that the platinum metal anode catalysts are sensitive to poisoning by CO, which adheres to the catalytically active sites, thereby inhibiting hydrogen electro-oxidation. Many fuel processors for proton exchange membrane (PEN) fuel cells therefore comprise additional chemical conversion steps to convert CO in the reformer product to CO2. Usually this is achieved through the use of one or more shift reactors followed by preferential oxidation or methanation of the remaining traces of CO. For alkaline fuel cells (AFC), it is necessary that the inlet to the fuel cell anodes are essentially devoid of both carbon dioxide and carbon monoxide.

In the present invention, a hydrogen selective membrane is preferably provided around the reaction chamber for hydrogen produced in the reaction chamber to pass through to produce a hydrogen steam which is relatively free of CO. This has the additional advantage of changing the nature of the steam reforming reaction, since hydrogen once it is formed in the reactor is removed through the membrane as well as leading to simplification of the design of the fuel processor through the elimination of shift reactors.

According to a second aspect of the present invention, there is provided a reactor for performing steam reforming, the reactor comprising:- a heat source; a reaction chamber arranged in use around the heat source, the reaction chamber being arranged to receive reactants and provide catalyst for the reaction; and wherein the heat source is a catalytic combustor.

The various aspects of the invention may be embodied in many ways, but some specific embodiments will now be described by way of example with reference to the accompanying drawings, in which:- Figure 1 shows a reactor for performing a steam reforming reaction; Figure 2 shows a typical longitudinal temperature profile through a catalytic steam reformer ; and Figure 3 shows a perforated supply tube for supplying an air/fuel mixture for combustion.

The hybrid reactor shown in Figure 1 consists of a central combustion tube 10 in which heat is generated by catalytic combustion of a suitable fuel, e. g. natural gas or anode exhaust gas from a fuel cell system in a suitable supply of oxygen such as air.

In this example, the combustion takes place over a combustion catalyst 11 coated on the inside surface of the central combustion tube 10. However, the combustion catalyst 11 could be provided in any suitable manner such as by the provision of combustion catalyst particles in the combustion tube 10. Reforming catalyst 20 is located around the central combustion tube 10, and it is in this annular region 21 occupied by the reforming catalyst 20 that the reforming reaction takes place on methane and steam which is supplied to the annular region 21. Around the reforming catalyst is a tubular membrane 30 made from a palladium silver (Pd/Ag) alloy through which hydrogen generated by the steam reforming reaction diffuses to the outside annulus of the reactor 40. The pressure in the reformer annulus 21 is maintained at a higher value than the outer product hydrogen annulus of the reactor 40. This pressure difference provides a driving force for hydrogen to migrate to the outer product hydrogen annulus 40. Hydrogen produced elutes from the outer annulus 40. Any unconverted methane and hydrogen may be recirculated to the central combustion tube 10 to provide fuel for the catalytic combustion. In addition, fuel for the catalytic combustor could be natural gas, or depleted anode exhaust gas from the fuel cell stack, if the reformer is incorporated into a fuel cell system. The central combustion tube 10 and the outside wall 41 of the outer product hydrogen annulus 40 may be made from stainless steel.

The steam reforming reaction is carried out over a conventional supported metal catalyst, designed to operate at relatively low temperatures. This could comprise, for example, supported nickel catalysts, using A1203, MgO, La203, etc. as supports.

Other catalysts may also be used, such as supported precious metal catalysts (incorporating Ru, Rh, Pt for example). Reforming catalyst is subject to deactivation during use, caused by poisoning of the active sites with sulphur from the feedstock, by carbon deposition or coking, and by sintering of the ceramic support. Although steps can be taken to minimise these effects through the design of the catalyst together with choice of suitable operating conditions, it is inevitable that in practice some deactivation will occur. In this example of the present invention, reforming catalyst could be removed from the reactor preferably by the provision of the reforming catalyst in pellets which are removable from the reactor once the activity has fallen to an unacceptable level. This has an advantage over some designs of compact reformer in which the reforming catalyst is in the form of a thin film deposited on the surface of the reactor and so is difficult to replace on a regular basis.

We have demonstrated (see Figure 2) the use of a coated catalyst for promoting the combustion of methane in such a reactor in which pelletted steam reforming catalyst is employed. In the example, the reactor was a single annular arrangement as shown in Figure 1. In practice many such tubes would be needed for a practical fuel processor, and it is possible that multiple combustion tubes could be inserted into a catalyst bed.

In a tubular reforming reactor such as described, the temperature along the axis of the tube varies. With fresh catalyst that has uniform activity through the reactor, the temperature profile takes the form shown in Figure 2. As catalyst deactivates, there is some change in the temperature profile, the nature of which depends on the mechanism of deactivation.

As the endothermic reforming reaction takes place through the catalyst, the temperature dips, and then rises as heat is absorbed from the combustion tube. Large temperature variations are generally not desirable since they can exacerbate deactivation-low temperatures promoting carbon formation, and high temperatures promoting sintering of the support. In internal-reforming fuel cells, one approach to even out the temperature profile within the reforming stack is to grade the reforming catalyst. In other words, catalyst at the front of the catalyst bed where most of the reaction occurs is designed to be of lower activity than that at the end of the bed.

However, this can be difficult to achieve and control. This approach could also be taken in the reactor described here. An alternative is that the heat generated in the combustion tube 10 could be graded, with more heat being generated at the inlet of the catalyst bed, and less heat at the exit of the bed. This can be achieved by admitting the air and fuel in the combustion tube via a suitably perforated supply tube 12. This is shown in Figure 3 with more perforations 13 in the end of the tube 12 at the front of reactant inlet of the catalyst bed 20 than at the other end.

The deposition of the combustion catalyst will now be described. It is difficult to put down layers of supported metal catalyst greater than a few microns thickness using conventional sol-gel technology. Layers of much greater thickness can be deposited using washcoats, as in the production of automotive three-way catalysts, and this technique was adopted in this study.

In the example given below, the aim was to deposit 10-501lm layers of palladium- doped alumina combustion catalyst onto sheets, tubes and channels of VDM Nicrofer (equivalent to Incolloy 800) using washcoats.

The catalyst preparation route shown below was investigated:- Prepare Nicrofer Surface Coat with aluminium sol Sinter to form oxide priming layer # Coat with washcoat of either reforming or combustion catalyst Sinter to form catalytic surface Each stage has a number of alternatives. Experimentation was performed at each stage to identify the best combination of alternatives to give a catalyst layer of the target thickness. At each stage, coupons of the reformer tube were examined by optical microscopy and then metallographically prepared. The metallographical prepared coupons were examined by optical microscopy, scanning electron microscopy (SEM) and energy dispersive x-ray microanalysis (EDX). The washcoats were studied by x-ray diffraction (XRD) and their surface area was measured by BET analysis. In the example preparation reported here, grinding was found to be the best method of preparing the metal surface to accept a sol. Coupons of the required material were ground to 60 grit on one side, and left in the as-received state on the other side. The as-received coupons grew an oxide layer under the sol after coating. Pickling the coupons induced intergranular cracking. Surface preparation by chemical means, such as pickling, has many advantages for a manufacturing process and subsequent refurbishment of heat exchangers when the catalyst has degraded.

Sols were prepared from dispersible alumina powders 10/2, P2 and DA supplied by Condea GmbH. After dipping and sintering, half of the coupons were thermally cycled 50 times at 5 minute intervals between ambient and 850°C.

For the final stage, two sets of washcoats were prepared: alumina washcoats and combustion catalyst washcoats:- 1. The alumina washcoats were made from the non-dispersible aluminium hydroxide Martinel Trihyde 01-107 (Martinswerk) mixed with either of two dispersible hydrated alumina sols Disperal 10/2 or Disperal P2 (Condea Chemie).

2. for the combustion catalyst formulation washcoats palladium was included as the combustion catalyst and lanthanum was added to help prevent long term high temperature degradation of the catalyst coating. Four alumina washcoats were prepared, with palladium added plus lanthanum at two different concentrations.

Another four washcoats were prepared with lanthanum added alone, the intention being to prefire the lanthanum plus alumina washcoats and then dope with a palladium nitrate solution.

Two test coupons, previously prepared and coated with sol, were dipped in each wash coat.

Optical examination of the initial alumina coatings revealed that some mudcake cracking appeared on surfaces after subsequent calcinations. SEM of metallographically prepared samples showed the coating thickness ranging from 2, um to 7 um. Some oxide growth was seen on some coupons, but was absent on others. In general, a good surface was achieved which could be used as a basis for subsequent deposition of the combustion catalyst.

After the final stage of preparation, optical microscopy showed that all four alumina washcoats formed coherent, but mudcracked coatings on the test coupons. SEM revealed coating thicknesses between 10 and 25, um. All the alumina washcoats had an apparently porous structure, with needle-shaped particles of approximate dimensions 0.5 x 0.5 x 2, um. Two of the Pd + La combustion washcoats had remained coherent. These had formed layers of 25-801lm thickness.

XRD of the oxides derived from the alumina washcoats showed that the material present is composed either of chi alumina (a variant of gamma alumina) or gamma alumina, which is the preferred phase as a catalyst support. In addition, the wide diffraction peaks showed poor crystallinity, hence high surface area, which is also preferred. This evidence is supported by BET studies which showed that the surface areas of the washcoat oxides are of the order of lOOm2g~l and above.

As well as the palladium/lanthanum combustion catalysts described above, the combustion catalyst could be formed from any suitable materials.

The prepared combustion catalysts were found to reduce the temperature at which combustion was required to take place in the steam reforming reactor of Figure 1 thus reducing fuel consumption, producing less nitrous oxides and enabling less expensive materials to be used for the burner material.