Reactor Design

Initially we were interested in relatively small rates of hydrogen production, which entailed using a small reactor, and therefore we knew that heat losses would be a problem. In an effort to overcome this we placed some reduced copper on silica catalyst inside a domestic vacuum flask and passed the methanol/air vapours in through a dip tube. The presence of a small quantity of platinum or palladium catalyst enabled the reaction to light off at room temperature, and tests showed that high conversion of methanol and an excellent yield of hydrogen were possible. However when we carried out the experiment using the same catalyst mixture in a conventional through flow reactor much poorer yields were obtained and these were not improved even with external heating of the tube. Using a through flow reactor in which the reactants were injected directly into the middle of the catalyst bed again gave selective results. Clearly point injection had a major beneficial effect and it is interesting to consider possible reasons for this.

Upon ignition of a highly exothermic reaction such as this the reaction increases in intensity and the reaction zone moves upstream towards the incoming reactants. With a conventional fixed bed reactor the reaction front stabilises at the leading edge of the catalyst bed. With point injection, however, the reaction front again moves towards the incoming reactant flow but as it approaches the inlet the reactant velocity increases to such an extent that the front remains within the catalyst bed where selective oxidation can be maintained. This very high gas velocity also means that it is much more difficult for the desired hydrogen product to back-diffuse and be consumed in the inlet region where oxygen partial pressure is higher.

In the case of point injection, heat utilisation is also improved as heat transferred either by convection or radiation from the combustion zone is absorbed by the catalyst, and is then available to promote subsequent endothermic reactions. With a planar reaction front, heat is lost much more readily to the reactor walls and this loss must be made up by the consumption of desired hydrogen, thus reducing the overall hydrogen yield.

A further interesting question that arises is whether steam reforming might give higher yields of hydrogen than partial oxidation. Certainly this is true in conventional plants in which external heat is supplied. However, the addition of water to the feed not only adds to the weight penalty but also increases the amount of heat required to vaporise the feed and bring the vapour to the reaction temperature. For an autothermal process this heat must arise from the combustion of part of the fuel or indirectly perhaps from burning some of the desired hydrogen end product. Thus in such a situation, some product would be consumed in order to make more product. Clearly the net effect will depend upon the precise details of the reactor design and the amount of water required. Conventional methane/steam reformers customarily use a large excess of steam, 5 to 7 mole ratio, which assists the heat flux and helps to maintain catalyst activity by limiting carbon formation. We have evaluated this new type of reactor both with and without water and have shown that at a 1:1 mole ratio with methanol there is little net benefit to be gained in hydrogen yield, although the carbon monoxide yield was significantly reduced by the addition of water.

Results of Small Reactor Studies

Three series of experiments were carried out, all with methanol charged at 10g/h and using a 3 weight per cent copper/silica catalyst in an upstream zone and a physical mixture of this catalyst and a 5 weight per cent palladium/silica catalyst, in the ratio of 9:1, in a downstream zone which comprised 20 per cent of the total catalyst bed volume. For one series of experiments water was injected along with the methanol at a rate of 6g/h. The reactor consisted of 50g of catalyst held in a tube of 4cm internal diameter, and into which the methanol/air/water feed was injected 4cm below the catalyst bed surface through a 6mm tube. For comparison purposes a conventional tubular reactor was used, made from 1cm internal diameter tubing. The temperature was that measured by a thermocouple icm downstream from the reactant inlet.

The manner in which the reaction temperature varied as the air rate was increased, is shown in Figure 1, for the three sets of experiments. Temperatures were somewhat lowered when water was added, reflecting the additional heat requirement, but were substantially lower in the case of the tubular reactor where the heat losses became so much greater. Figure 2 shows how the methanol conversion paralleled the changes in reaction temperature. The methanol conversion reached 90 per cent when the oxygen:methanol ratio was 0.5, in line with the stoichiometry given in Equation (i).

The carbon monoxide and hydrogen yields as the methanol conversion increased are shown in Figures 3 and 4, respectively. The carbon monoxide yields were significantly lower from the novel reactor, especially when water was added. This reflects the influence of the carbon monoxide shift reaction in the cooler parts of the reactor. The hydrogen yields were similar with and without added water, but again much superior for this reactor compared to the tubular reactor. This reflects the greater heat losses from the tubular reactor and the need to compensate by burning some extra feed or desired hydrogen product. At essentially complete methanol conversion the gas composition was 35 volume per cent hydrogen, 20 volume per cent (carbon monoxide+carbon dioxide) and 45 volume per cent nitrogen, with a net hydrogen flow of 12 l/h, quite sufficient for the applications which we have discussed.

Reactor Scale-Up

After the successful demonstration of the novel properties of the concept on a bench scale reactor it was necessary to scale up the operations in order to:

To this end a stainless steel reactor with a capacity of two litres was constructed, as shown in Figure 5. The reactor shell was blanked off at the base and a one inch diameter pipe in the bottom wall acted as a take off pipe for the make gas. The top of the reactor was flanged with a centrally mounted injector pipe which penetrated midway into the catalyst bed. The catalyst was mounted on a gauze base positioned to give clearance above the gas take off pipe. Temperatures in the catalyst bed were measured at nine points by thermocouples mounted in drillings through the reactor shell. By this means both axial and radial temperature distributions were monitored.

The reactor was supplied with feedstock air and purge nitrogen which were metered by mass flow controllers supplying up to 500 Normal litres/minute (N 1/min) each. Methanol was delivered by a calibrated diaphragm pump at rates up to 100 ml/min. The methanol and air reactants were only brought together in the injector inside the reactor. After reaction over the catalyst the product gases were cooled in an efficient condenser and the condensate collected. The dry product gas was then analysed by gas chromatography and the flowrate measured by mass flowmeter before venting.

Scale-Up Data

The temperatures in the catalyst bed are shown as a function of time, that is from the start of air injection, in Figure 6. The reaction was initiated from ambient conditions and commenced within a few minutes; steady state temperatures were achieved within 0.75h. Analysis showed that hydrogen gas was generated within 30 minutes of start. A noticeable feature of Figure 6 is the eventual formation of a hot spot in the region adjacent to the injector. A temperature of about 450°C was observed in this region, while temperatures in the remainder of the reactor were much cooler, generally in the range 200 to 300°C.

The thermal conditions in the bed, and hence temperature distributions, are the result of numerous factors including feedrate, extent and selectivity of the reaction, thermal capacity of the bed and reactor design. Consequently neither temperature nor pressure (close to ambient in all cases) can be considered as process variables useful for reactor evaluation in this strongly adiabatic, thermally heterogeneous environment. It is useful to assess reactor performance in terms of the effect of the oxygen:methanol molar ratio on the reaction parameters, since this ratio will have the greatest bearing on the extent of the reaction (conversion) and the reaction selectivity.

The exit gas composition is shown as a function of the oxygen: methanol ratio in Figure 7. It can be seen that the gas composition was largely independent of this ratio, the average gas composition over the range depicted being: hydrogen 42, nitrogen 37 and (carbon monoxide+carbon dioxide) 21 volume per cent. Since the extent of methanol conversion was strongly dependent on the oxygen : methanol ratio, as expected, the reaction selectivity was largely independent of methanol conversion.

Under these conditions selectivity to hydrogen, defined as the fraction of hydrogen bound up in converted methanol which appears as molecular hydrogen in the product, was in excess of 85 per cent. Alternatively expressed, the molar ratio hydrogen:methanol converted reached 1.71. A limiting value of 2.0 was anticipated based on the reaction stoichiometries from Reactions (i) and (ii). A maximum hydrogen output of 3.28 N m³/h was achieved at an oxygen:methanol ratio of 0.56. The hydrogen yield reached 1.20 N m³ hydrogen per kilogram methanol fed to the reactor. A limiting value of 1.40 was anticipated, based on the expected reaction stoichiometry. Methanol conversion was strongly dependent on the oxygen : methanol ratio and under the conditions of maximum hydrogen output a methanol conversion of 94 per cent was observed.

Very low oxygen concentrations in the exit gas are also indicated in Figure 7. In fact the analytical method did not distinguish between oxygen and argon present in the atmosphere. However, by making a comparison between the ratio (oxygen + argon): nitrogen provided by experimental data and the ratio of argon : nitrogen present in the atmosphere, it was estimated that the residual oxygen concentrations in the exit gas were usually in the range 1000 to 2000 ppm. The data given in Figure 7 suggest that carbon monoxide concentrations in the exit gas are less than 10 volume per cent. The observed concentration range was 4 to 10 volume per cent depending on the conditions. Such levels may be diminished even further by the incorporation of water into the feed. The catalyst was extremely efficient for the water gas shift reaction under these temperature conditions. In fact concentrations of carbon monoxide in the exit gas as low as 2 volume per cent have been recorded under such conditions.

Conclusions from Scale-Up Data

The data presented above show that the attractive features of this novel reactor, first demonstrated on a bench scale reactor, have been proven on a scale of commercial interest. These are:

The observed exit gas compositions are very close to those expected from an overall reaction stoichiometry:

The fact that gas compositions were largely independent of the oxygen:methanol ratio suggests that oxygen had little part to play in the primary reforming step, and that the reaction sequence was:

Thus the high selectivity to hydrogen observed using this process was due to the successful competition for oxygen displayed by carbon monoxide, in preference to hydrogen, over this catalyst under these conditions:

The Use of Hydrocarbons as Fuels

With the ready availability of hydrocarbon fuels, their use as feedstocks in this novel reactor concept would be an attractive option. Methane (natural gas), propane (LPG), n-hexane, 60–80 petroleum ether, lead-free gasoline and commercial diesel have all been evaluated as feedstocks using the bench scale reactor described previously. The catalyst used in all these cases was a more active 1 per cent platinum-3 per cent chromium/silica formulation. All of these fuels have been successfully reformed although here discussion will be restricted to the liquid hydrocarbons.

None of the hydrocarbons was capable of auto-ignition with the catalyst employed. However, the reactions could be initiated readily from ambient in the usual way using either methanol or product hydrogen, and then be sustained by introducing the hydrocarbon and switching off the initiator.

With the use of liquid hydrocarbon fuels such as gasoline and diesel, where a great multiplicity of compounds are present, the use of molar oxygen:hydrocarbon ratios for reactor assessment was not possible. However, with a knowledge of the carbon analysis for the fuel the oxygen:carbon atomic ratio could be calculated and used for reactor/catalyst evaluation.

The hydrogen gas concentration in the exit gas, and the hydrogen selectivity as a function of the oxygen:carbon ratio using lead-free gasoline as feedstock are shown in Figure 8. The latter was a premium test gasoline containing a maximum 45 per cent aromatics, a maximum 20 per cent olefins with the balance saturates. A difference in performance between this and the analogous data obtained using methanol is immediately apparent. With the latter the gas composition was largely independent of the oxygen:carbon ratio, but with the gasoline feed the gas composition—as reflected by the hydrogen gas concentration—was ratio dependent. The selectivity to hydrogen reached a maximum of about 80 per cent at an oxygen : carbon ratio in the region of 2.0.

This situation is the same when using any of the other hydrocarbon fuels tested. Figure 9 gives a comparison of the four liquid hydrocarbon fuels and shows that hydrogen was abstracted from all the fuels except diesel with similar efficiency, reaching similarly high hydrogen selectivities at an oxygen:carbon ratio close to 2.0. The inference is that hydrocarbon species whether aliphatic, aromatic, naphthenic, saturated or unsaturated were all reformed with equal facility. Quite clearly, a commercial device using a hydrocarbon fuel would be required to operate at an oxygen:carbon ratio of 2.0.

Hydrocarbon conversions were uniformly high in the range 80 to 90 per cent, and were less dependent on the oxygen: carbon ratio than those using methanol feed. Hydrogen yields obtained from the liquid hydrocarbon fuels were:

n-hexane: 1.09 N m³ hydrogen/kg (limiting value 1.82 N m³/kg)

petroleum ether: 0.83 N m³ hydrogen/kg (limiting value 1.21 N m³/kg)

gasoline: 1.28 N m³ hydrogen/kg (limiting value 1.31 N m³/kg).

Commercial diesel fuel was also reformed but the results were erratic and further development work is required.

Because hydrogen is abstracted from these fuels with similar efficiency the maximum hydrogen yield observed is largely a reflection of the degree of unsaturation of the fuel at constant hydrocarbon conversion.

There was a perceived threat of coking from these highly unsaturated fuels, especially when the reactor was operated under rich conditions, that is at oxygen:carbon ratios less than 2. This could occur for example on start up, shut down or on changing conditions. Consequently a durability run was carried out using lead-free gasoline as feed at an oxygen : carbon ratio of 0.27, an extremely oxygen deficient condition.

Figure 10 shows the constancy of the six measured parameters over the course of an operation lasting for a day. On discharge parts of the reactor and the thermocouple showed signs of coking; the catalyst was free of carbon.

Conclusions from the Hydrocarbon Data

In principle the use of natural gas, LPG, naphtha, and lead-free gasoline provide options for feedstocks in the present technique although they have not been demonstrated on a commercial scale reactor. These feedstocks show little tendency to cause coking even under ultra rich conditions. The use of diesel fuel must await further development.

Auto-ignition is not achieved using hydrocarbons alone, but is readily effected using either methanol or hydrogen gas.

Gas compositions and especially hydrogen selectivity are more sensitive to the oxygen : carbon ratio than is methanol reforming. Hydrogen selectivity greater than 80 per cent, at a hydrocarbon conversion greater than 90 per cent, was achieved at the stoichiometric oxygen : carbon ratio of 2 using gasoline feedstock.

The suggestion of a stoichiometric oxygen : carbon ratio suggests the participation of oxygen in the primary reforming reaction, which is not the case with methanol. The observations argue for an overall stoichiometry for all hydrocarbon fuels: