1.

Process integration conditions selection

1.1.1 Feed pressure

Once the coke formation problem was solved, the most appropriate hydrogen production (in R1) followed by purification (in MM) process parameters were selected. In order to get the optimal feed pressure, the effect of the feed pressure on both the CPO reaction and the hydrogen separation was tested. Figure Error! No text of specified style in document..1 shows the effect of increasing pressure on the hydrogen flow produced by CPO in the catalytic reactor R1. Sweep/permeate side pressure was atmospheric and feed/retentate side pressure was increased varying the pressure difference across the membrane. As expected from the thermodynamics of the CPO reaction, the amount of produced hydrogen decreased slowly as increasing feed pressure. At around 4 bar of feed pressure the produced hydrogen had decreased by 30%. The test conditions are summarised in Table Error! No text of specified style in document..1 and an overview of the mixture composition produced in the CPO reactor is shown in Table Error! No text of specified style in document..2. The composition of the mixture changed in time with feed pressure as it affected the CPO process. Table Error! No text of specified style in document..2 shows the average composition of the process in some of the tested feed pressures, the rest of the flow till 100% was Ar.

900 800

Produced hydrogen (mLN/min)

700 600 500 400 300 200 100 0 0 2 4 6 8 10

Pressure difference (bar)

Figure Error! No text of specified style in document..1. Feed pressure effect on produced hydrogen in the CPO reactor (reactor temperature= 1073 K and methane feed flow 700 NmL/min).

Table Error! No text of specified style in document..1. Test conditions in the catalytic reactor and in the membrane module for the analysis of the optimal feed pressure. T (K) Flow in (NL/min) Sweep flow (NL/min) Feed pressure (barg) Permeate pressure Process Catalytic reactor Membrane module 1073 773 2.3 2.9 2.5 0 / 0.55 / 1.7 0-9 0-9 atmospheric Hydrogen production by Hydrogen purification with M6 methane CPO PdCu membrane

Table Error! No text of specified style in document..2. Average composition produced in the catalytic reactor, and therefore fed to the membrane module. With feed pressure = 1.0 bar(g) With feed pressure = 3.5 bar(g) With feed pressure = 5.5 bar(g) With feed pressure = 9.0 bar(g) n.d.: no detected H2 29.0 26.7 21.5 18.1 CO2 1.8 2.2 3.7 4.3 CO 17.1 16.3 12.9 12.1 CH4 3.8 4.7 9.0 9.9 O2 n.d. n.d. n.d. n.d.

A decrease in hydrogen production in the reactor R1 due to an increase in feed pressure affects directly the hydrogen recovery in the membrane module. There are two processes with an opposite effect on the membrane performance proceeding simultaneously in the system. As the feed pressure increases less hydrogen is produced in R1, decreasing the hydrogen pressure difference across the membrane hence decreasing the hydrogen recovery in the MM. However, the hydrogen pressure difference across the membrane increases as the feed pressure increases. Figure Error! No text of specified style in document..2 shows the effect of the feed pressure on the hydrogen recovery with constant permeate side pressure equal to the atmospheric pressure and feeding the variable mixture produced in R1 and shown before. The hydrogen recovery in the membrane module is calculated with Error! Reference source not found.. As expected, the hydrogen recovery increased as feed pressure

increased even though feeding the hydrogen percentage decreased, due to a higher hydrogen pressure difference across the membrane. The selected operational feed pressure was 3.5 bar(g), as it was the value with the highest production/recovery ratio.
100
Without Ar sweep With 550 mL/min Ar sweep

hydrogen

80

Hydrogen recovery (%)

60

40

20

0 0 2 4 6 8 10

Pressure difference (bar)

Figure Error! No text of specified style in document..2. Feed pressure effect on the hydrogen recovery at different sweep gas flows (M6 membrane temperature=673 K).

Figure Error! No text of specified style in document..2 also shows the effect of the sweep gas flow variation on the hydrogen recovery with M6 membrane in the MM. Sweep gas flow addition had a positive effect on the process, by reason of increasing the hydrogen partial pressure across the membrane. With 550 NmL/min of sweep gas, the hydrogen recovery was approximately doubled, but with 1700 NmL/min no further significant improvement was observed.

1.1.2 Sweep gas flow

As explained before hydrogen production is thermodynamically favoured at low pressures and hydrogen purification at high feed pressures. Apart from the feed pressure effect analysed before, the sweep gas flow is also a determining parameter. Higher sweep gas flow implies lower hydrogen partial pressure in the permeate side of the

membrane, increasing the hydrogen pressure difference across the membrane. Another disadvantage of increasing pressure to increase hydrogen pressure difference, apart from the fact that it decreases the produced hydrogen amount, is that gases different from hydrogen could go through the membrane [1]. Even though the use of sweep gas is beneficial for the overall process, it is also an expensive part of it, which makes indispensable to optimize the required flow. The hydrogen dilution in the retentate, or feed, side of the membrane is another critical factor for hydrogen permeance across the membrane. In the integrated hydrogen production plus purification process, the hydrogen concentration in the feed flow to the membrane depends on the conversion and feed dilution of the previous R1, so it cannot be controlled.

In order to obtain an optimal sweep gas flow, hydrogen permeation tests were performed with different hydrogen dilutions and using different sweep gas flows. Figure Error! No text of specified style in document..3 illustrates the hydrogen recovery tests performed at 3.5 bar(g) in the feed side and atmospheric pressure in the sweep side with the M7 membrane at 773 K. Four different hydrogen dilutions in inert gas were tested, between 20 and 70%. For each dilution hydrogen flow across the membrane was measured at different sweep gas flows. The sweep gas flow used in each measurement was calculated as a percentage of the feed gas that was added to the membrane module. It can be observed that when hydrogen percentage was high, 70%, even with no sweep gas the hydrogen recovery was very high, more than 90%. But for the lower hydrogen percentage cases, around 20-25 %, the recovery without sweep gas flow was lower than 20%. When using a sweep flow around 15-20% of the total flow fed to the MM, the hydrogen recovery reached its maximum percentage. Larger sweep gas flow, and therefore higher economical cost, will not imply higher efficiency of the process.

120

100

Hydrogen recovery (%)

80

60

40
71%H2 40%H2 60%H2 35%H2
24% H2

20
28%H2

19% H2

16% H2

10

20

30

40

50

% (sweep gas/feed gas)

Figure Error! No text of specified style in document..3. Hydrogen recovery of M7 in the membrane module with different hydrogen percentages and sweep gas flows at 3.5 bar pressure difference across the membrane and 773 K.

Hydrogen recovery versus sweep gas flow at different hydrogen percentages was also studied at 2.0 bar(g) of feed pressure and 773 K, shown in Figure Error! No text of specified style in document..4. The aim of this second test was to check the dependence of the optimal sweep gas flow percentage, calculated from the results given in Figure Error! No text of specified style in document..3, with the feed pressure. The tested hydrogen dilutions were in the range of the ones used in the previous test at 3.5 bar(g), between 16 and 70%. The negative effect of not using sweep gas in hydrogen recovery was more noticeable in this case. The hydrogen partial pressure in the feed side of the membrane was lower and therefore, also the hydrogen pressure difference across the membrane, which is the driving force for hydrogen purification. Even though this difference, the maximum hydrogen recovery was achieved again when using a sweep gas flow of around 15-20% of the flow fed to the MM. Regardless the composition of the flow fed to the MM, that could vary in time in a real processes due to catalyst deactivation or feeding composition variations, the optimal sweep gas flow

can be determined just measuring the flow fed to the membrane module and using a sweep gas flow of around a 20% of this value.
120

100

Hydrogen recovery (%)

80

60

40

20
71%H2
27%H2

60%H2
24%H2

40%H2
19%H2

35%H2
16%H2

0 0 10 20 30 40 50 % (sweep gas/feed gas)

Figure Error! No text of specified style in document..4. Hydrogen recovery of M7 in the membrane module with different hydrogen concentrations and sweep gas flows at 2.0 bar pressure difference across the membrane and 773 K.

1.2

Membrane performance in complex mixtures

Hydrogen recovery from a gas mixture is not only affected by hydrogen dilution and hydrogen partial pressure difference across the membrane, as studied before, but also by the mixture components. When producing hydrogen from methane, as in this work, hydrogen is always mixed at least with N2, CO, CO2, H2O and CH4 in the case in which sulphur compounds from natural gas are removed before. The CO is known to have an important inhibition effect in membranes behaviour [2- 5]. About water vapour contradictory information is reported in the literature [6, 7]. Regarding CO2, CH4 and argon do not affect the membrane so strongly [8, 9]. CO has been found to cause more inhibition than the same concentration of CO2 [10]. In the present work the CO

percentage was around four times the CO2 one, so the study was focused on the CO

effect on the prepared PdCu membranes behaviour, supposing negligible effect of the CO2. In this section the results of the tests where hydrogen was produced via CPO and wetCPO processes in the R1catalytic reactor and hydrogen was separated in the MM are given. When producing hydrogen via wet-CPO process different water/methane ratios were tested to analyse the effect of water addition on membrane behaviour. The catalytic reactor R1 was operated at 1073 K, the membrane module at 773 K, the feed pressure of the system was 3.5 bar (g), the sweep side was at atmospheric pressure, and the used sweep gas flow was the 20% of the flow produced in R1 and fed to MM. The effect of the gas mixture components and the water content on the membrane behaviour was analysed with three membranes: M6, M7 and M8.

With the M7 membrane hydrogen was purified from a complex mixture produced in R1 via CPO. Figure Error! No text of specified style in document..5 shows the methane conversion measured in R1 and the hydrogen recovery recorded in the membrane module during the operation time. The catalyst activity was stable during the test. The measured conversion was around 70-75% and the hydrogen yield around 1. The average H2 and CO percentages in the mixture were 27.5% and 14% respectively. Hydrogen recovery had decreasing tendency in time to 85%. Mixture components fed to the membrane seemed to have a slight negative influence on membrane behaviour, but the test was not long enough to obtain more conclusions.

100

2.0
1.8

Conversion, recovery (%)

80 60
40

1.6 1.4
1.2

1.0
0.8

0.6 20
CH4 conversion R1 H2 recovery MM H2 yield R1

0.4 0.2

0 0 2 4
Time (h)

0.0 6 8

Figure Error! No text of specified style in document..5. Hydrogen production via CPO in R1 at 773 K and purification in MM with M7 at 1073 K and 3.5 bar of pressure difference.

With the M6 membrane longer mixture tests were performed. The mixture fed to the MM was produced in R1 via methane CPO and wet-CPO alternated in time. The process, operational conditions used and the average mixture composition produced in R1 in each time section are summarized in Table Error! No text of specified style in document..3. The catalyst activity and membrane hydrogen recovery results are depicted in Figure Error! No text of specified style in document..6. Between the different testing days the system was kept in inert atmosphere at temperature, identified in the figure with thicker lines. In the figure it can be observed the positive effect of sweep gas addition in the first hours, where CPO was performed without sweep gas, section A, and then with sweep gas, section B. Methane conversion and hydrogen production in the R1 reactor were quite stable, 85-90% and 0.5-0.6 NL/min, respectively. The highest hydrogen yield in R1 was obtained at the highest S:C ratio due to higher H2/CH4 ratio in the reforming process, operating conditions F. When changing the reaction conditions in the R1 the composition of the mixture produced in the R1 and fed to the membrane module changed. The range of the average composition of each

Yield

compound in the different conditions tested in the reactor were 18-23% H2, 2-6.8% CO2, 14-19% CO, 3-5% CH4, 1-2% H2O and the rest Ar. Table Error! No text of specified style in document..3. Process, operational conditions used and mixtures composition employed for complex mixtures analyse with M6. A Process S:C Sweep gas T reactor (K) T membrane (K) CO percentage (mol%) H2 percentage (mol%) CH4 percentage (mol%) CO2 percentage (mol%) O2 percentage (mol%) Feed flow to MM (NL/min) n.d.: no detected CPO No B D E Wet-CPO 0.02 0.11 0.50 550 NmL/min 1073 673 17.8 18.1 17.5 19.4 19.7 21.6 4.6 4.6 4.4 2.7 2.7 4.7 n.d. n.d. n.d. 2.5 2.4 2.5 C F 0.90

18.0 20.4 4.4 2.4 n.d. 2.4

14.8 22.6 2.9 7.0 n.d. 2.6

100

2.0

A
CH4 conversion & H2 recovery (%)
80

B C B

F
1.6

60

1.2

40

0.8

20

0.4

CH4 conversion

H2 recovery (%)

Produced H2 flow (LN/min)

0 0 5 10 15 20 25 30

0.0

Reaction time (h)

H2 flow (LN/min)

Figure Error! No text of specified style in document..6. Performance of the CPO-membrane system. R 1 at 1073 K, M6 in membrane module at 673 K, feed pressure=3.5 bar(g), sweep gas at atmospheric pressure.

The variation in the mixture composition fed to the membrane module had no significant negative effect on the membrane performance, as the hydrogen recovery was measured to be around 60% during the test time at 673 K. Two of the compounds can affect more significantly the membrane performance: carbon monoxide and water. Carbon monoxide is known to have an important inhibition effect. When operating in B conditions, CPO without water, the hydrogen recovery seemed to be lower than at wetCPO conditions (C, D, E or F sections). The lowest carbon monoxide concentration, 14%, was produced in F conditions with the highest S/C ratio. The good performance of the system in the F compared to the others taking into account the decreasing CO percentage showed that, probably, the positive effect of water overrules the anticipated negative effect of carbon monoxide on the membrane performance. Between 60 and 70% of the produced hydrogen in the first reactor was recovered by the membrane module, regardless of operational conditions and mixtures.

The hydrogen recovery of the M6 membrane in a H2:N2 mixture, with 25% of hydrogen and 550 mL/min of sweep gas flow, was measured to be around 80%. Due to operation at complex mixture conditions the hydrogen recovery decreased to 60-70%, so there was a reduction of 10 to 20%. Similar data have been obtained in the literature. A reduction of 20% in the hydrogen recovery is shown in Li et al. work [10] when operating a Pd membrane in a simulated WGS mixture at 673 K. It has been observed that the membrane permeance can decrease when operating with carbon monoxide due to coke formation at low water/carbon monoxide ratios, as used in this work [11, 12]. Cheng et al. [13] observed a quick decrease in the hydrogen recovery after 12 h operation of a Pd membrane installed in a methane CPO reactor due to coke formation from carbon monoxide at 773 K. Whereas in other works, with water/carbon monoxide ratios 3 and 19, carbon monoxide had no significant negative effect at 573 K, the permeance decreased by 8% when operating in simulated gas mixture, but remained constant during further operation [14]. High temperature, 773 K, and water addition seemed to be the reasons for not observing coke in the tests with M6.

Complex mixture tests were also performed with the M8 membrane. Catalyst activity and hydrogen recovery results are shown in Figure Error! No text of specified style in document..7. Table Error! No text of specified style in document..4 summarizes the average CO and H2 percentages obtained in the different hydrogen producing processes performed in the R1. As it can be observed in the table, the hydrogen percentage in the mixture increased when adding water, due to additional reforming reactions. As a consequence of the reactive process modification adding water, the selectivity of the catalytic process in R1 varied, decreasing the percentage of CO in the outflow mixture.

The process in the MM needed quite a long time to stabilize. Recorded hydrogen recovery in the MM needed around three hours to arrive to steady state conditions. Once at the steady state, hydrogen recovery was around 60, 60, 67, 67, 63 and 67% respectively at each operational condition. The CO percentage did not seem to damage the membrane, and neither the rest of the gases as steam, CO2 or CH4. The obtained hydrogen recovery was around 60-67%, regardless the operational conditions and the mixture composition during more than 50 hours. As observed in the similar test performed with the M6 membrane the positive effect of water could have overruled the anticipated negative effect of carbon monoxide

Table Error! No text of specified style in document..4. Average percentages of H2 and CO in the outflow of the catalytic reactor and fed to the membrane module at different operational conditions. CPO S:C H2 (%) CO (%) 27.4 15.5 0.5 27.5 13.6 Wet-CPO 1.0 1.5 2.0 28.1 31.2 34.1 11.9 10.8 8.9

2.5 32.0 8.3

100

4.0 3.5

Conversion, recovery (%)

80 3.0 60 2.5

40

1.5 1.0

20

CPO
0 0

CH4 conversion R1

wet- CPO 0.5

wet-CPO wet-CPO MM 1.0 H2 recovery 1.5

30

wet-CPO H2 yield R1 2.0

40

10

20

50

0.5 wetCPO 2.5 0.0

Time (h)

Figure Error! No text of specified style in document..7. Hydrogen production at 1073 K in the catalytic reactor R1 and purification with M8 at 773 K in the MM and 3.5 bar of pressure difference.

From the hydrogen purification from complex mixtures tests performed in the MP it can be said that between the 60 and 70% of the hydrogen produced in R1was purified in the membrane module, regardless the used operational conditions and mixture composition. This value is close to the 70% obtained in a membrane reactor with methane reforming by Basile et al. [15] ]. As compounds adsorption on membrane surface decreases at high temperatures, 773 K could have been high enough temperature to avoid inhibitive adsorptions to occur. The combination of CO concentration decrease and water concentration increase seemed to have a positive effect in the hydrogen recovery, together with the fact that at higher S:C ratio the hydrogen percentage in the mixture was higher.

1.3

Membrane lifetime with H2:N2 and complex mixture tests

In this section an overview of the three tested membranes (M6, M7 and M8) lifetime with H2:N2 or complex mixtures is shown. Figure Error! No text of specified style in document..8 summarizes the main tests performed with the M6 membrane and the

Yield

2.0

hydrogen and nitrogen flows measured across the membrane during its lifetime. All the flow measurements were carried out with pure hydrogen and nitrogen at a feed pressure of 2 bar(g), with an atmospheric pressure on the permeate side and at temperatures of 673 to 773 K. The membrane selectivity varied along its lifetime from 1500 to 250 at 773 K, but in all the cases the nitrogen flow through the membrane was negligible compared to the hydrogen flow.

During the first hydrogen flow measurements, between 0 h and 200 h, there was a permeance increasing tendency probably due to the fact that the metal layer continued activating. The difference between hydrogen flow at 200 h and 550 h is because of temperature increasing from 673 to 773 K. The hydrogen flow through the membrane increased decreased in time from around 2.4 to 1.2 NL/min at 673 K, measured at 26 h and 2500 h respectively. From the 900 h to 1800 h and from 3300 h on, the membrane was kept in air at 298 K. The first hydrogen flow measurement shown in the mixture 773 K period, at 1850 h, is from the day before the mixture tests were started. The lower hydrogen flow at 1850 h can be an effect of the storage in air. Earlier tests have shown that it takes at least 150 hours in hydrogen atmosphere to get a stable hydrogen flow through the membrane after storage in air. During the period where the membrane was operated under complex mixtures the hydrogen flow was stable at around 1.4 NL/min. Around 2450 h there was a power failure in the system. As a consequence of the power failure the membrane was cooled without control from 773 K to room temperature in H2 atmosphere. Till this moment the permeability was slowly increasing during the mixtures tests, from 1.1 to 1.4 NL/min. But after the power failure the hydrogen flow decreased to values around 0.5 NL/min. At 3200 h the membrane was cooled to 298 K in inert gas at 1 K/min. It was heated to 673 K again at 3600 h to continue the tests, but it was not selective. 1.2 NL/min of nitrogen went through the membrane with a pressure difference of 2 bar. The membrane was taken out from the module and it was observed that the metal layer of the membrane had been peeled off as Figure Error! No text of specified style in document..9 shows.

4,5
Hydrogen flow (NL/min)

H2 inert H2 673 K 298 K 773 K

air 298 K

Power failure mixture inert 773 K 673 K

mixture 773 K
Hydrogen
Nitrogen flow (NL/min)

4 3,5 3 2,5 2 1,5

Nitrogen

4,E-03

3,E-03

2,E-03

1 0,5
0 0 500 1000 1500 2000 2500 3000 Time (hours)

1,E-03

0,E+00 3500

Figure Error! No text of specified style in document..8. Hydrogen and nitrogen flow during the lifetime of M6 (feed pressure=2 bar(g) and sweep gas at atmospheric pressure).

Figure Error! No text of specified style in document..9. M6 membrane peeling off detected at 3600 h.

Analysing the hydrogen and nitrogen flow values during the entire lifetime of the M6 membrane several conclusions can be obtained. The nitrogen flow did not increase significantly during the lifetime of the membrane, so this flow can be due to a leak in the ceramic-metal sealing. However, the hydrogen flow decreased during its lifetime. The membrane permselectivity calculated from the data given in Figure Error! No text of specified style in document..8 varied from 1500 to 250. This selectivity decrease was not due to pores formation in the metallic layer but because of a decrease in

hydrogen permeation. From this observation it can be concluded that the membrane deterioration was not because metal layer damaging but because of a change in the metal structure that entailed a worsening of the hydrogen bulk diffusion capacity.

The power failure occurred in the system at 2424 h and the subsequently uncontrolled cooling in hydrogen atmosphere had a negative effect on the hydrogen flow through the membrane. In Error! Reference source not found. and Error! Reference source not found. it was observed that controlled cooling in hydrogen atmosphere did not affect membrane permeance. In literature thermal cycling of the hydrogenated metallic layer has been observed to produce membrane failure characterized by crack formation and metal film peeling [16], as also observed with the M6 membrane. In the case of the M6 membrane controlled cooling or heating did not seem to negatively affect membrane behaviour, whereas the same thermal cycle without controlling the rate had an irreversible negative effect. However, further investigating e.g. by longer thermal cycling tests are needed to determine if uncontrolled cooling in hydrogen is the reason for membrane failure.

The peeled off M6 membrane metal layer was analysed with XRD and SEM techniques. The outer surface of the metal layer showed deep grain boundaries in SEM image, Figure Error! No text of specified style in document..10, not observed before. The oxygen observed on the membrane surface of the membranes alloyed in the third batch, Error! Reference source not found. and Error! Reference source not found., could have had a negative influence causing a morphologic change in the metallic layer. Grain boundaries and the area where morphologic changes occurred are weak places where leakage and finally peeling off the membrane can occur. As concluded from the M6 lifetime figure, with hydrogen and nitrogen flows, open pores were not created in the metal layer as N2 flow did not increase. The peeled off foil showed a lot of ceramic particles at the support side, indicating a lot of stresses at the interface. This pattern occurs normally after cooling in a hydrogen filled membrane system, which was the case with M6 membrane when the power failure occurred during operation. Most probably deep grain boundaries, but not open pores, were created due to the stress till the metal foil peeled off and the membrane completely failed. The composition of the metal layer was analysed with SEM-EDX to check if coke formation occurred when working under mixture, which could lead to metallic phase deactivation [17]. Just Pd

(89.6 wt%) and Cu were observed in the metal layer surface. The stable permeance observed in Figure Error! No text of specified style in document..6 corroborates that the metallic layer was no deactivated during operation in complex mixtures. In the EDX composition analyse of M6 performed after the alloying step there was carbon in the membrane surface, probably coming from the preparation method and removed during the operation.

Figure Error! No text of specified style in document..10. SEM image of the peeled off M6 metal layer, x18000 magnification.

The peeled off M6 membrane was analysed with XRD in order to have deeper knowledge of the metal layer. Figure Error! No text of specified style in document..11 shows the diffractogram of M6 membrane sample, in dark line, and as a reference the diffractogram of a membrane prepared and alloyed in the same way as the M6 in grey is given. The metal layer has fcc structure as expected from the composition and explained in the previous chapter. In the diffractogram of the reference sample after alloying there are some maximums that do not appear in the M6 diffractogram, at 35, 38, 4, 58 and 67 angles that correspond to the Al2O3 from the support. In the M6 diffractogram maximum peaks associated to a crystalline structure between pure Pd and PdCu are observed, suggesting the PdCu solid solution formation. This means that the copper atoms are placed in the Pd net, displacing slightly the peaks to higher angles. In the literature it has also been observed that before alloying separate Pd and Cu maximums appear in the diffractogram [18-20]. Whereas, after the alloying all the characteristic peaks of Pd were in the same position for both Pd and PdCu [21, 22]. This shows the formation of PdCu interstitial alloys, which occurs when the atoms of one component are smaller than the other and the smaller atoms fit into the spaces between the larger atoms [23].

Figure Error! No text of specified style in document..11. XRD diffractogram of the M6 membrane at the end of its lifetime.

Hydrogen and nitrogen flows across the M7 membrane measured with 2 bar of pressure difference during its lifetime can be observed in Figure Error! No text of specified style in document..12. The hydrogen flow between days 0 and 10 was measured at 663 K and the rest at 773 K. Main events occurred during M7 lifetime have been summarized in Table Error! No text of specified style in document..5. As it can be observed in the table, there were some power failures in the system the days 31, 47 and 59. In the first power failure (identified with * in the figure) the membrane was cooled at 1 K/min but in hydrogen atmosphere. In the other two power failures the membrane was cooled with no temperature control but in inert atmosphere. In the first the day 47 it was cooled to 298 K and in the second one, day 59, to 693 K.

The consequence of the three uncontrolled cooling can be directly observed in hydrogen and nitrogen flows. After the first uncontrolled power failure, day 31, the hydrogen flow increased substantially and there was a small increase in the nitrogen flow. But after the second uncontrolled power failure, day 47, the nitrogen flow increased drastically, indicating the formation of open pores or structures in the metal layer and implying the end of the membrane operational lifetime. The last power failure had a different effect on the membrane as hydrogen flow decreased to half. This indicates that this power

failure affected the hydrogen diffusion process and not the metallic layer morphology. The day 60 the membrane was measured to be no selective, most of the flow that entered the MM was detected in the permeate side. Figure Error! No text of specified style in document..13 shows a picture of the M7 broken membrane, with a thermal stress characteristic breakage, when it was taken out of the membrane module. Uncontrolled cooling together with the possible oxygen negative in the metal morphology could have enhanced defect formation along the grain boundaries of the membrane [24]. All these could have been the main reasons for the membrane breaking.

Power failure* H2 673 K 7000

Hydrogen flow (NmL/min)

Power failure

H2 773 K N2

H2 dilution 773 K

H2 dilution 773 K

inert 773 K

mixture 773 K

200 180

6000 5000 4000 3000 2000 1000 0 0

H2

140 120 100 80 60 40 20 0 10 20 30 40 Lifetime (days) 50 60 70

Figure Error! No text of specified style in document..12. Hydrogen and nitrogen flow during M7 membrane lifetime measured with 2 bar of pressure difference. Table Error! No text of specified style in document..5. Lifetime summary of the M7 membrane. Days 0 10 11 17 25 - 31 31 33 46 47 Hydrogen permeance measurements with thermal cycle at 673 K Hydrogen permeance measurements with thermal cycle at 773 K Hydrogen recovery Vs sweep gas flow tests at 773 K Power failure*. Controlled cooling to 298 K Hydrogen recovery Vs sweep gas flow tests at 773 K Power failure. Uncontrolled cooling to 298 K

Nitrogen flow (NmL/min)

160

59

Power failure. Uncontrolled cooling to 693 K

Figure Error! No text of specified style in document..13. Picture of the broken M7 membrane.

Comparing the peeling off occurred in the M6 and the M7 membranes there were some differences. In the M6 membrane it seemed that there was a unique point from which the metal layer broke. As explained in the M6 lifetime summary no open pores formation occurred and the metal layer was most probably peeled off due to stress. On the other hand, during the M7 membrane lifetime pores formation were observed as the nitrogen flow increased. When looking at the broken M7 membrane it was observed that there were more than one places where the metal layer had been peeled off. This might indicate that there were several open pores that continue opening till the metal layer peeled off from the ceramic support.

The hydrogen and nitrogen flows measured along M8 membrane lifetime with H2:N2 and complex mixtures tests are shown in Figure Error! No text of specified style in document..14. The main tests performed during this period are summarized in Table Error! No text of specified style in document..6. During the first 10 days flows were measured at 663 K and the rest of the time the membrane operated at 773 K. During the M8 membrane lifetime there were not power failures, despite variations in the membrane behaviour were observed. After the hydrogen permeance test in pure hydrogen at 773 K, shown in Error! Reference source not found., the membrane was kept in inert gas for more than 10 days, but the hydrogen permeance decreased and the leak flow increased during this time. In the period where the membrane was tested in the MM purifying hydrogen produced in the catalytic reactor, the hydrogen flow and

leaks were quite stable. As it was concluded from the hydrogen recovery test in complex gas mixtures performed with M8 membrane, Figure Error! No text of specified style in document..7, the mixture components did not have a severe influence in the membrane behaviour. The M8 membrane was then tested with a sulphur compound addition. These results and its complete lifetime are shown in the following sections.

3500
3000

H2 673 K H2 N2

H2 773 K

inert 773 K

CPO, wet-CPO 773 K

200 180 160

2500

Hydrogen flow (NmL/min)

140

2000
1500

120 100 80 60 40

1000 500 0

20

10

20

30

40

50

Lifetime (days)

Figure Error! No text of specified style in document..14. Hydrogen and nitrogen flow during M8 membrane lifetime (feed pressure=2 bar(g) and sweep at atmospheric pressure).

Table Error! No text of specified style in document..6. Lifetime summary of the M8 membrane. Day 09 13 17 18 - 31 32 43 Hydrogen permeance measurements with thermal cycle at 673 K Hydrogen permeance measurements with thermal cycle at 773 K In inert atmosphere at 773 K Hydrogen production plus purification tests at 773 K

Nitrogen flow (NmL/min)

After analysing the effect that the performed tests had on the three membranes it can be concluded that uncontrolled cooling had a strong effect on membranes lifetime. After one and three uncontrolled power failures the M6 and M7 membranes, respectively, they broke. Different effects of the stress suffered by the metallic layer due to the uncontrolled cooling were observed in the nitrogen and hydrogen measured flows. In some cases the membrane failed because open structures were formed in the metallic layer, increasing notably membrane leakage (M7). In other cases a hydrogen flow decrease was observed, indicating that the hydrogen diffusion through the metallic layer had been affected (M6 and M7). Regarding the PdCu alloy metal it was corroborated with XRD the fcc structure of the M6 membrane and it was observed that the PdCu alloy was formed by Pd crystalline net where the Cu atoms where located.

1.4

Membrane performance with sulphur addition

In the previous sections it has been analyzed the optimum operational parameters and the effect cooling and complex mixture compounds have on membranes behaviour. Two of the membranes tested with complex mixtures (M6 and M7) failed as explained before, most probably due to thermal stresses. The third membrane, M8, was tested with sulphur addition as usually together with the methane some sulphur compounds ppm are part of the natural gas. In this section sulphur addition effect, till 100 ppm, on M8 membrane behaviour is explained.

1.4.1 H2S and H2 mixture (I)

In order to study the effect of H2S impurity on M8 membrane behaviour, hydrogen permeability was sequentially measured in H2+H2S atmosphere with H2S concentration of 100 and 40 ppm and with pure hydrogen at 773 K. With these cycles it was possible to analyse if the permeability was recovered when taking out the sulphur from the system. Figure Error! No text of specified style in document..15 shows the permeability inhibition caused by the addition of sulphur compound ppm to the stream. Membrane permeability decreased by approximately 80%, but it was not a permanent poisoning. Permeability was recovered being under H2 atmosphere at 773 K. The

permeability inhibition observed due to the presence of sulphur was nearly identical at H2:H2S mixtures with 40 ppm and 100 ppm H2S. The measured hydrogen flow through the membrane with 40 ppm of H2S was slightly higher than with 100 ppm, 25% of the value in pure hydrogen, and permeability increased slowly in time. When taking out the H2S from the system the permeability was the same one measured before the sulphur addition. The selectivity of the membrane changed during the test. Nitrogen flow across the membrane varied from around 10 Nml/min in the first measurement to 64 NmL/min after the test with 100 ppm of H2S. The test with 40 ppm of H2S did not further increase the nitrogen flow across the membrane.

5,000

100

H2

40

H2

200 180

4,000 Hydrogen flow (NmL/min)

160 140 Nitrogen flow (NmL/min)

3,000
H2 N2

120 100 80 60

2,000

1,000

40 20

0 0 20 40 60 Time (h) 80 100

Figure Error! No text of specified style in document..15. Poisoning of the M8 membrane in 100 and 40 ppm H2S/H2 mixtures and recovery in pure H2 at 773 K. The decrease of the hydrogen permeability can be attributed to H2S molecules adsorption on the metallic surface. It seemed that only reversible site blocking with no irreversible bulk sulphidation occurred under the testing conditions, as the permeability was completely recovered. The H2S/H2 ratio in the tests was lower than the threshold ratio required to form thermodynamically stable sulphides (molar ratio H2S/H2=0.0011) [25]. The high temperature employed in the test, 773 K, could have enhanced reversible sulphur adsorption [26]. But the metallic PdCu surface was damaged by the sulphur as nitrogen flow increased. It seemed that sulphur atoms attacked weak areas, possibly

grain boundaries, creating open pores in the metallic layer and decreasing membranes selectivity. But sulphur-adsorbing sites are almost saturated at 40 ppm, as the hydrogen permeance was almost the same with 40 and with 100 ppm. The permselectivity of the membrane when this first analysis with sulphur addition was finished was low, 70.

1.4.2 CPO process with sulphur addition

Sulphur addition effect on membrane permeability was also analysed when operating under complex hydrogen containing mixtures produced in R1. Hydrogen production via CPO was performed in R1 with 0.35 g of 40NiMgO catalyst at 600 h-1 and 1073 K. The membrane module operated at 773 K with the M8 membrane at a pressure difference of 2 bar. The test results are given in Figure Error! No text of specified style in document..16. During the first hours the process operated without sulphur. After 3.5 h, 32 ppm H2S were added to the system and the conditions were kept constant during 6 hours. The catalyst activity in R1 seemed to be not affected by sulphur addition as CH4 conversion and H2 yield were stable. This was a good result as Ni catalysts are usually deactivated in the presence of sulphur compounds [27]. The complex mixture produced in R1 and fed to the MM contained an average of 12.4% CO, 1.3% CO2, 21.2% H2, 4.4% CH4 and 59.1% Ar.

CPO
100

CPO + 32 ppm H2S

3.0 2.5 2.0

Conversion, recovery (%)

80

60 1.5 40 1.0 20
CH4 conversion R1 H2 recovery MM H2 yield R1

0.5 0.0 0 2 4 Time (h) 6 8 10

Figure Error! No text of specified style in document..16. H2S addition effect on hydrogen production in R1 via CPO (1073 K) and purification in MM with M8 (773 K).

When H2S was added to the integrated system hydrogen recovery in the membrane dropped drastically from 80% (without H2S) to 45% (with 32 ppm of H2S) and then it slightly increased in time to 50%. This final measured hydrogen recovery in mixture+H2S test was lower than the previously observed 60-65% with complex mixtures without sulphur addition. On the other hand, in the H2:H2S test with 40 ppm the hydrogen flow across the membrane decreased a 75%, much more than the 30% observed in the mixture +H2S test comparing the permeability without and with sulphur. Membrane deactivation was not more pronounced due to the simultaneous presence of complex mixture and sulphur in the stream.

When this mixture+H2S test was finished nitrogen flow through the membrane at 2 bar of pressure difference was measured to be 200 NmL/min. Before this test the recorded leakage was around 60 NmL/min. The further leakage increase meant that open structures continued to be forming in the metallic layer. Most probably sulphur atoms attach the grain or pores boundaries forming more or bigger pores.

Yield

1.4.3 Permeance recovery try (I)

After the tests performed with H2+H2S and complex+H2S mixtures the membrane leakage had strongly increased. In order to try to decrease the leakage the membrane was maintained in H2 atmosphere for some days. Firstly the membrane was kept at 773 K and then, temperature was raised as listed in Figure Error! No text of specified style in document..17. In case there was adsorbed sulphur on the membrane, desorption is expected to become more favourable at higher temperatures resulting in more permeance recovery, since the dissociative adsorption of H2S on metals is exothermic. Different effects can be observed in the figure. As it was expected from the previous tests with H2S+H2 mixture, it did not seem to be adsorbed sulphur on the metallic surface. In case some sulphur was adsorbed, it should have been desorbed during the recovery at 773 K increasing the hydrogen permeance, but it decreased. Secondly, the nitrogen flow through the membrane did not increased during the recovery time, it decreased from 340 to around 200 NmL/min after the recovery thermal cycle.

6000

773 K

823 K

873 K

773 K 823 K 898 K

773 K

600

5000

500

Hydrogen flow (NmL/min)

4000

400

3000

300

2000

200

1000 H2 0 56 N2 58 60 62 64 66 68 70

100

Lifetime (days)

Figure Error! No text of specified style in document..17. First thermal cycle in hydrogen atmosphere attempt with M8 for hydrogen permeability and leakage recovery.

Nitrogen flow (NmL/min)

However, there was also a negative effect of the recovering thermal cycle on the membrane behaviour. The employed temperatures seemed to be too high for the membrane. The hydrogen flow through membranes usually increases with temperatures. But when checking the hydrogen flow measured at the same temperature and different times a drop is observed. At 773 K it was 4100 NmL/min the day 57 and 2100 NmL/min the day 69. This suggested that the membrane structure was seriously damaged at those temperatures. Pores could have been closed due to metal diffusion at the employed high temperatures, which lead to leakage decrease, but the metal bulk was damaged affecting the hydrogen flow through the membrane. At the end of the recovery trying the membrane selectivity was too low, 10 the day 69. Even then, more sulphur tests were performed with the membrane to further analyse sulphur effects on the membrane surface.

1.4.4 H2S and H2 mixture (II)

Even though, as explained before, membrane M8 selectivity was too low, some extra sulphur tests were done with it. First of all the H2 + H2S mixture test was carried out again. Hydrogen fluxes through M8 membrane were measured when adding 100 and 60 ppm of H2S alternated with permeability measurement in pure H2 to establish a baseline.

Hydrogen flow through the membrane in pure hydrogen atmosphere before doing the current test was 2100 NmL/min. Figure Error! No text of specified style in document..18 shows the effect of 100 and 60 ppm H2S addition in a H2:H2S mixture. The obtained results were similar to the ones obtained before with 100 and 40 ppm of H2S, Figure Error! No text of specified style in document..15. In this second H2:H2S test, hydrogen flow across the membrane decreased also significantly with the presence of H2S, a 50%. Comparing these results with the ones obtained in the 100 and 40 ppm test it can be observed that even if the membrane was damaged during the thermal recovery trying, the measured hydrogen flow when sulphur addition was performed was the same, around 1000 NmL/min. As the membrane permeability was higher before the first H2:H2S test than before the second one, the 1000 NmL/min were a reduction of 80% in the first one and 50% in the second one. It seemed that hydrogen diffusion

through the metal was affected in the same way most probably due to a similar surface blocking due to the sulphur.

In this test hydrogen permeability was also recovered to original values after being in pure H2 at 773 K, so sulphur adsorption was reversible. Membrane selectivity was not affected by the H2S, the leak did not increase during the test. The nitrogen flow was stable during the test, which can imply that H2S did not further damage the membrane.

H2
2.500

100 ppm H2S

H2

60 ppm H2S

H2
800 700

2.000

Hydrogen flow (NmL/min)

1.500

500 400

1.000

300 200

500 100
H2 N2

0 0 10 20 30 Time (h) 40 50 60

Figure Error! No text of specified style in document..18. Poisoning of the M8 membrane in 100 and 60 ppm H2S/H2 mixtures and recovery in pure H2 at 773 K.

1.4.5 Wet-CPO process with sulphur addition

The last test performed with sulphur addition was the process integration where hydrogen production and purification was carried out with the addition of some H2S ppm. In this test wet-CPO process was performed in R1 with 0.35 g of 40NiMgO catalyst at 600 h-1 and 1073 K. From previous results it was concluded that the addition of water improved hydrogen production, but in this case H2S is also added. Mixing water and H2S a very corrosive medium for the system and the membrane is created. In order to minimize the corrosion a low S/C ratio equal to 0.5 was employed in the

Nitrogen flow (NmL/min)

600

hydrogen production wet-CPO process. The membrane module operated at 773 K with the M8 membrane at a pressure difference of 2 bar.

During the first hours the process operated without sulphur, see Figure Error! No text of specified style in document..19. During the time without sulphur it can be observed that hydrogen recovery increased from 60 to 80%. This result is in accordance with the positive effect of water addition in membrane behaviour observed also in the previous chapter. After 4 hours 33 ppm of H2S were added to the system and these conditions were maintained during 5.5 hours. The catalyst activity in R1 was not strongly affected by sulphur addition as CH4 conversion and H2 yield were quite stable, although a slow activity decreasing tendency was observed. When H2S was added to the system hydrogen recovery in the MM dropped drastically from 80% (without H2S) to 36% and then, it slightly increased in time till around 45%. Results are again very similar to the ones obtained in sections 1.4.1 (H2+H2S first test) and 1.4.2 (CPO+H2S test). The hydrogen recovery decreased around a 50% when operating with sulphur addition, similar to the one observed with 40 and 100 ppm.

wet-CPO
100

wet-CPO + 33 ppm H2S

5,0 4,0 3,0 2,0 1,0

Conversion, recovery (%)

80 60 40 20 CH4 conversion R1 0 0 2 4 H2 recovery MM H2 yield R1

0,0

Time (h)

10

Figure Error! No text of specified style in document..19. H2S addition effect on hydrogen production in R1 via wet-CPO (1073 K) and on purification in MM with M8 (773 K).

Yield

After doing the wet-CPO+H2S test nitrogen flow through the membrane was measured with 2 bar of pressure difference and it had significantly increased from 200 to 500 NmL/min. This implied that a part of the measured flow in the permeate side were components other than H2. Taking out from the measured permeate flow the other gases flow, hydrogen recovery would most probably have decreased in time. Therefore, membrane deactivation during wet-CPO with sulphur addition was more aggressive to the metallic support, causing open pores or cracks in the thin metal layer.

1.4.6 Permeance recovery try (II)

A long recovery test in hydrogen atmosphere was performed after the sulphur addition in wet hydrogen production process. By this moment the membrane was much damaged, but the aim of the current test was to analyse if permeance or leakage improvement could be achieved with thermal cycle. The membrane was maintained in hydrogen atmosphere during almost 15 days. Temperature was varied and hydrogen and nitrogen flows through the membrane were measured at 773, 823 and 873 K. The high temperatures were kept during short time, as it had been observed before that strongly damage the membrane. During the recovery trying no hydrogen permeability improvement was observed. The leakage decreased slightly when increasing the temperature to 873 K. Probably some open pores were closed due to the high temperatures, but during the last days at 773 K a leakage increase tendency was observed. It seemed that the membrane structure was too damage and no improvement was observed, but it did not break or peeled off.

4.500 4.000

773 K 823 K

873 K

773 K

1000

Hydrogen flow (NmL/min)

3.000 2.500 2.000 1.500 1.000 500

H2 N2

600

400

200

0 0 50 100 150 Time (h) 200 250 300 350

Figure Error! No text of specified style in document..20. Second thermal cycle in hydrogen atmosphere attempt with M8 for hydrogen permeability and leakage recovery.

1.4.7 Complete lifetime of the M8 membrane

In this section an overview of the M8 membrane complete lifetime is presented. Hydrogen and nitrogen flows across the membrane measured at 2 bar of pressure difference are shown in Figure Error! No text of specified style in document..21. This membrane was firstly characterized with pure hydrogen permeability tests involving thermal cycles in inert and hydrogen atmosphere at 673 and 773 K (0 h to 20 days). It was then tested in hydrogen purification from complex mixtures produced from CPO and wet-CPO in the R1at 773 K (32 to 48 days). Finally, tests with sulphur addition were performed: H2+H2S mixture (50 to 55 days), CPO process in R1 and purification in MM with H2S addition (day 56), first recovery try (58 to 69 days), H2+H2S mixture test (70 to 74 days), wet-CPO process in R1 and purification in MM with H2S addition (day 79) and second recovery try (80 to 91 days).

Nitrogen flow (NmL/min)

3.500

800

CPO+H2S 773 K
H2 673 K 6000 H2 773 K inert 773 K CPO, wet-CPO H2/H2S 773 K 773 K

wet-CPO+H2 S 773 K recovery try 6000

recovery H2/H2S 773 K try

5000

5000

Hydrogen flow (mLN/min)

4000

4000

3000

3000

2000

2000

1000

1000

10

20

30

40

50

60

70

80

90

100

Lifetime (days)

Figure Error! No text of specified style in document..21. Complete lifetime summary of the M8 membrane.

Around day 44 the hydrogen permeability increased significantly from 1800 to 4100 NmL/min. Water addition during the complex mixtures production, via wet-CPO, could have caused this variation. The nitrogen flow did not increased, so the metal layer was not damaged. The first test performed with 100 and 40 ppm H2S did not irreversibly deactivate the membrane, but its structure was damaged as the nitrogen flow increased. This reversible sulphur adsorption is in accordance with results from the literature. In the results from Kim et al. no sulphide formation was observed in a fcc Pd80Cu20 membrane [28]. Most probably weak areas, as grain boundaries, were attacked by the sulphur producing open pores or regions. During the time in sulphur atmosphere the hydrogen flow decreased, as hydrogen diffusion through metal sulphides is orders of magnitude slower than in pure metals. After the CPO test with sulphur addition (day 56), the nitrogen and the hydrogen flow increased. During the first thermal recovery trying the membrane was kept in hydrogen, but too high temperatures were tested damaging the membrane performance and strongly decreasing hydrogen permeability to around 2000 NmL/min.

Nitrogen flow (mLN/min)

H2

N2

After the first permeability recovery trying, a second test with 100 and 60 ppm H2S in H2:H2S mixture was performed. Similarly it was observed that metallic layer was not irreversibly deactivated, but that the structure of the membrane was damaged and leaks increased. During the performed wet-CPO with sulphur addition test the hydrogen recovery increased, most probably due to the positive effect of water addition, and nitrogen flow was stabilized. In the last recovery trying hydrogen permeability was quite stable and the leakage had a slow increasing tendency, the membrane behaviour did not improve.

1.4.8 M8 membrane characterization

After performing the previously explained tests with the M8 membrane it was taken out from the MM. Visually some special remarks were observed on the membrane surface, illustrated in Figure Error! No text of specified style in document..22. There was a split part, leakage, which started 10 mm from the close end cap. The surrounding area of several centimetres was rough. A lot of stresses must have been present to result in those small folds and roughness. This close end cap site of the membrane was the final part of the membrane, so most of the hydrogen had diffused to the permeate side before arriving there. Therefore, sulphur concentration in this area was much higher than the ppm added to the system. This could also be corroborated by the fact that on the other side of the membrane, near the open end cap shown in Figure Error! No text of specified style in document..23, it can be observed the drying fold of the ceramics. This can be a suspected weak area of the membranes, but no problems were observed in this area in the M8 membrane. This area was located at the entrance of the membrane where the lowest sulphur concentration was.

Figure Error! No text of specified style in document..22. M8 membrane general image at the end of its lifetime (A) and in the close end cap area (B).

Figure Error! No text of specified style in document..23. M8 membrane image at the end of its lifetime near the open end cap area.

In SEM images the structural differences in the membrane area near the close and near the open end cap were observed, Table Error! No text of specified style in document..7. The close end cap area surface, where the split was located, was rougher than the open end cap area. Some open pores could be observed in the x10.000 enlargement image of the close end cap area. These open pores explain the experimentally observed leakage increasing after the sulphur tests. The surface was rougher than before the permeability tests, also observed in the literature after H2S test in PdCu membrane [29].

Table Error! No text of specified style in document..7. SEM images of both end caps of the M8 membrane at x2.000 and x10.000 enlargements. Close end cap

Open end cap

In Pd foils pinholes and crack formation have been determinate in the literature to be caused by the stress generated due to the difference in the lattice constant between Pd and bulk Pd4S after H2S exposure [30]. In the current work with PdCu membranes no sulphur compounds adsorbed were found after the tests. This result is in accordance with others from the literature, where no sulphur was observed in a bcc phase at 623 K by SEM and XRD [31] and neither in fcc phase till 700-1000 ppm of H2S [32]. From atomic modelling it has been demonstrated that irreversibly adsorption of S monolayer on most PdCu alloys is not favourable [33]. Even if no irreversible adsorbed sulphur compounds were found in the M8 membrane, during operation some Pd could have formed reversible Pd4S and generate the stress that caused the observed splits.

With EDX technique element composition was measured at three different positions of the end and open cap areas samples. The average composition of the membrane was quite heterogeneous, as shown in Table Error! No text of specified style in document..8. But it seemed that the Pd percentage in the open end cap area was higher than in the close end cap area. This could be due to the metallic deposition method employed for membranes preparation. The membranes were located vertically to perform the electroless plating and probably the amount of Cu deposited at the bottom and at the top was not exactly the same. The second approach to this composition difference can be metal diffusion during the tests performed with M8 membrane. Metal diffusion can be specially affected by components other than H2 (i.e. H2S). Cu atoms could have been forced to diffuse from the region exposed to high sulphur concentration to the area with low sulphur concentration. If the membrane composition was initially homogeneous the Cu should have been moved from the close end cap to the open end cap area, opposite to the results observed in the table. In the literature it has also been observed that S interactions with Pd70Cu30 enhance Pd surface concentration [34]. Most probably this result opposite to the expected one was due to the fact that membrane composition was not homogeneous before the sulphur test. Adsorbed sulphur was not detected with the SEM-EDX composition analysis, which supported the conclusion obtained from the experimental tests that the sulphur adsorption on the membrane surface was reversible.

Table Error! No text of specified style in document..8. Element composition measurement performed at three different positions in the M8 samples with SEMEDX. Close end cap At.% Pd 78.96 73.28 76.07 Open end cap At.% Pd 82.09 80.83 81.52 At.% Cu 19.81 23.64 18.27 At.% Cu 10.77 15.68 12.33

The phase structure of the broken M8 membrane was studied by XRD. Diffraction measurements were done in the open and close cap area and both in the external and

internal (the support side) surface, depict in Figure Error! No text of specified style in document..24. Due to the fact that the metal PdCu layer in the M8 membrane was very thin, the XRD results in both surface sites (A and B) were the same. The crystalline structure of the samples was fcc, as expected from the membrane composition. When comparing the diffraction peaks of the M8 sample and of the pure Pd, Figure Error! No text of specified style in document..25, it is observed that they were very similar. This result was also detected in the M6 membrane, Figure Error! No text of specified style in document..11. Cu atoms are located inside the Pd crystalline net when alloyed and it just modified slightly the Pd net parameters. This effect is well observed at high angles. The peak splitting that occurs at high angles is because of modifications in the surroundings of the Pd coordination net due to the presence of Cu. This means that the Cu modifies the structure and this distortion generates the formation of two signals at 2Theta values of 68 and 83. In the two X-ray diffraction measurements performed in the open cap side of the membrane, the one exposed to the lowest sulphur concentration, the splitting of the peaks is more pronounced that in the close cap side. This would involve that the metal alloy of the open cap area had more Cu atoms into the Pd net than in the close cap area, indicating a Cu diffusion from the higher sulphur concentration area to the lower sulphur concentration area. This Cu atoms diffusion was also concluded from the EDX analysis.

Close A Close B Open A Open B

Peak List

01-087-0645

Figure Error! No text of specified style in document..24. XRD diffractogram of the M8 membrane at the end of its lifetime.

Peak List 01-080-1916

01-087-0645 01-084-1770

Pd

01-080-1916
10 20 30 40 Position [2Theta] 50 60 70 80 90

Figure Error! No text of specified style in document..25. Position of the

01-084-1770

maximum diffractogram peaks observed in the M8 membrane and the characteristic maximums of pure Pd.

Another interesting conclusion from the XRD diffractogram is that no irreversibly

20 30 40 50 60 70 90 adsorbed 10 sulphur was observed. In the literature it had also been found80that XRD Position [2Theta] patterns of clean and used H2S Pd47Cu53 foil were nearly identical [35]. This

observation is in accordance with the experimental and end-life SEM-EDX results.

1.5

Discussions

During the first tests of process integration, hydrogen production in R1 and hydrogen purification in MM, coke formation was observed. This problem was solved by changing the MM tube and locating a stainless steel tube with the inside covered by hard chrome plating. With this covering the contact between the gas mixture and the stainless steel was limited and coke was not generated.

After solving the coke formation problem, process parameters were optimized to improve hydrogen production and recovery, such as feed pressure and sweep gas flow. The aim of both of them was to increase the hydrogen partial pressure across the membrane to enhance hydrogen recovery. But the hydrogen production step was unfavoured by high pressures and the use of sweep gas is expensive, so they needed to be optimized. The highest hydrogen recovery in the MM was obtained with 3.5 bar (g) of feed pressure and atmospheric pressure in the permeate side and using a sweep gas flow equal to the 15-20% of the flow produced in R1 and fed to the MM.

Membranes behaviour in complex mixtures was analysed performing both CPO and wet-CPO processes in the R1at a space velocity of 600 h-1 (ggash-1gcat-1) with 0.35 g of catalyst. The obtained hydrogen recovery in the MM was around 60-65% regardless the composition of the mixture, which depended on the hydrogen production process carried out in the catalytic reactor R1. This hydrogen recovery was around a 10-20% lower than the one obtained with a same hydrogen percentage H2:N2 mixture. Mixture components like CO, CO2 and CH4 seemed to do not have strong negative effect on the membrane hydrogen permeability at the tested concentrations, whereas water addition seemed to have a positive effect on it. No coke formation was observed in the membranes, which could have caused hydrogen permeability decrease. The operational temperature of the membrane was quite high, 773 K, which could have also minimized the compounds adsorption on the metallic surface.

From sulphur addition tests performed with 40, 60 and 100 ppm in H2:H2S mixtures it was concluded that sulphur was not irreversibly adsorbed on the metallic surface, but it did affect the metallic structure. Hydrogen permeability was recovered when taking out the sulphur compound from the system, but the nitrogen flow through the membrane increased. Open pores where detected by SEM images in the membrane area where sulphur concentration was higher than the fed ppm. This indicated that sulphur atoms firstly attacked weak areas producing open structures. Afterwards nitrogen flow did not further increase, but hydrogen permeability decreased. Two different mechanisms may be responsible for the inhibition of hydrogen transport through the membranes with H2S: poisoning of the catalytic dissociation of H2 at the surface, what was directly observed when sulphur was added to the system; and reduced permeability of hydrogen through the bulk of the membrane, which was observed in the lifetime summary. The

quite high temperature employed in the tests, 773 K, and the fact that the membrane was fcc, more resistant than the bcc phase to sulphur, could have been the reasons for no irreversible sulphur adsorption. Sulphur addition produced leaks in the membrane surface, which the thin metal layer of the membrane (0.866 m) made easier. No adsorbed sulphur compounds were detected in the metal layer after the sulphur tests by EDX and XRD analysis. The prepared PdCu membranes seemed to can operate with temporary addition of 100 ppm, or lower, H2S concentration without decreasing hydrogen permeability, but affecting its selectivity.

Two of the membranes lifetime was strongly affected by some uncontrolled cooling that occurred due to power failures in the system. Controlled cooling and heating, even in hydrogen, did not affect their permeability, but after the uncontrolled cooling leakages appear and the metal layer was peeled off. In the third membrane, no power failures occurred and its lifetime was long. In this membrane the cause for its continuous selectivity decrease was the sulphur. When the third membrane was taken out from the module a split was observed in the area where the highest sulphur concentration was.