1965, No. 1 27

Cryogenic production of ultra-pure hydrogenS. Shaievitz

In recent years the industrial demand for ultra-pure hydrogen has shown a considerableincrease. The following article describes a unit for economically producing hydrogen in anultra-pure state. This unit enables the relatively small consumer to reap benefits derivedfromconventionallarge scale processing.

Introduetion

With the growth of the electronics industry and therequirements of certain metal fabricators, the demandfor hydrogen in an ultra-pure form (99.995%) hasshown substantial expansion in recent years. The appli-cation generally involved is the brazing of componentsin an atmosphere that must exclude oxygen, water va-por, hydrocarbons, and nitrogen. For instance, electri-cal and magnetic properties of materials associatedwith the manufacture of several electronic devices arequite susceptible to impurities of the order of 1 : 106

or less. As a result, there is a constant effort to upgradeall materials connected with the manufacture of thesecomponents. Other applications of ultra-pure hydrogenas a protecting atmosphere in the electronics industryinclude the production of silicon and other types ofcrystals, and the metallizing of ceramics.Also, some specialized applications may be mention-

ed here. In the refractory metals field, where powdermetallurgy techniques are employed, an ultra-pure hy-drogen atmosphere appears to have a beneficial effect.Molybdenum rolling, a difficult operation, is said to beimproved in an atmosphere of ultra-pure hydrogen.Certain metals such as titanium, columbium, molybde-num, tungsten, and their alloys form undesirable prod-ucts with nitrogen and oxygen at elevated temperatures.Consequently, the hydrogen used in the heat treatmentof these materials must be nearly free of impurities.Ultra-high-purity hydrogen atmospheres have been ap-plied to the bright annealing of stainless steel and alsoappear to be useful in the production of tin plate [1].

For large-scale users, on-site generation using thesteam reforming and shift converter process [2] is themost economical approach, and in terms of capacityit outranks other methods. Raw materials for this pro-cess, natural gas and steam, are often readily availableand low priced.

S. Shaievitz, B.Chem.Eng., is a research worker at Philipslaboratories, Briarcliff Manor, N.Y., U.S.A.

661.96 :621.593

In general, however, the electronics and metal heat-treating industries may be considered as small scaleusers of hydrogen (under 150 m3 per hour at normaltemperature and pressure). The capital costs of thesteam reforming process are then prohibitive, and unlessthe consumer happens to be located adjacent to anelectrolytic operation, he is dependent upon purchasesof cylinder or trailer supplies. This method of supplyis not very economical however, because of the costs.involved in transporting, handling, and distributingrelatively small quantities, and because cylinder gasmay require additional processing before ultra-highpurity can be realized.

The steam reforming process consists of two steps. In the firststep natural gas (consisting principally of methane) togetherwith steam is passed over a solid catalyst at a pressure of 20 atmand a temperature of 650-950 °C to produce hydrogen and carbonmonoxide. In the second step, the carbon monoxide is furthertreated with steam to form additional hydrogen and carbon diox-ide. This second step is commonly referred to as the "shift reac-tion". Usually the shift reaction is applied in three stages to re-duce the carbon monoxide content to about 1 : 105 (10 p.p.m.).

An economical means of transporting and handlinghydrogen for application on a relatively small scale isin the form of liquid anhydrous ammonia. Since largequantities of ammonia are produced the small user is,in effect, obtaining the advantages of large scalehydrogen production and the convenient low cost ofshipping liquid anhydrous ammonia. It remains todissociate the ammonia into a 75%H2-25%N2 mix-ture and to process this mixture to obtain the desiredpurity level. The first step presents no difficulty: am-monia dissociators are manufactured by several firmsand are readily available. For the purification of the

---------- ------------------------------(1) A survey of techniques in which an atmosphere of ultra-purehydrogen is used can be found e.g. in Chemical Week, May 19,1962, pp.104-121, particularly on p. 120 and 121.(2) A description of the steam reforming process can be found inthe article cited under [1). Also, see Chem. Engng., August 7,1961, pp. 62-64, or Brit. chem. Engng. 8, 466-470, 1963.

28 • PHILIPS TECHNICAL REVIEW VOLUME 26

gas obtained, a cryogenic process has been developedat Philips Laboratories, Briarcliff Manor, U.S.A.; thatlends itself to nearly automatic operation and, as aconsequence, requires little operating labour. Thisprocess will now be described.

The cryogenic purification process

The new cryogenic purification process involvesfour steps, viz:1) cooling of the J~aseous N2-H2 mixture, resultingfrom the dissociation of ammonia, to approximately- 193 oe accompanied by the condensation of most ofthe nitrogen present in the mixture;2) the separation of the vapor from the condensedliquid nitrogen;3) the removal of the remainder of the nitrogen byadsorption at low temperatures; and4) the isothermal regeneration of the adsorbers, bymeans ofpurging with low pressure hydrogen. We shallnow give a detailed description of the process, follow-ing the flow sheetjig. 1.At point A (at the left in the figure) dissociated am-

monia is combined with recycled purge hydrogen andfed to the compressor B at a suction pressure of about0.3 atm. Feed gas (24%N2 - 76%H2) is compressed to60 atm, and, after removal of the compression heat inan aftercooler C, passes through one of two ammoniaadsorbers D.These adsorbers serve to remove any tracesof undissociated ammonia and moisture that might bepresent in the gas. The gas is cooled to a few degreesabove its dew point by the outgoing gas streams in themain heat exchanger E. It is then brought to approxi-mately -193°C by flowing through a condensing coil Fwhich is immersed in the liquid nitrogen tank G. Atthe outlet of this coil the nitrogen concentration in thevapor has been reduced from 24% to 7.5% while thecondensed liquid nitrogen contains approximately12.5% dissolved hydrogen. The vapor is separated fromthe liquid in the separator-surge vessel H. This vesselis immersed in the liquid nitrogen bath in order to pre-vent re-evaporation of the liquid nitrogen condensatewhich would result from heat leak. The vapor leav-ing H flows through one of two automatically operat-ed reversing adsorbers Ads (in the situation sketched,adsorber Ads 1). In so doing, nearly all the nitrogenimpurities are removed.At the point I (top right) the purified hydrogen is

split into two streams. The major portion passesthrough one of a set of dual adsorbers K which removesthe trace quantities of nitrogen remaining in this stream.The effluent from these adsorbers may contain lessthan 1 : 105 (10 p.p.m.) of impurities. These final ni-trogen adsorbers K are manually regenerated in theconventional way. Since the quantity ofnitrogen to be

adsorbed is rather small, the units may be convenientlysized for a daily cycle. Flowing through the heat ex-changer E, the hydrogen product is rewarmed to near-ly ambient temperature. A back pressure valve V4 onthe product line L maintains pressure in the plant.The smaller fraction of the split hydrogen stream

(30% or less) is throttled to slightly above atmosphericpressure by valve VI. This stream is used to regenerateone of the two reversing adsorbers (adsorber Ads 2 inthe situation sketched). Regeneration is accomplishedisothermally: that is, regeneration takes place at thesame temperature as adsorption and is effected bypurging the adsorber bed with pure hydrogen at areduced pressure. This method of regeneration is cho-sen because it offers the possibility of automatic ad-sorber operation and consequently of greatly reducingthe cost-price of the hydrogen produced. Of course,automatic continuous operation imposes a certain re-quirement to the ratio between the flow rates of thetwo streams (see Appendix).

Immersing the adsorbers in the liquid nitrogenbath serves two purposes. One function of the bath isto remove the heat of adsorption. Allowing the operat-ing adsorber (Ads 1 in the situation sketched) to warmup would result in a decrease in its adsorptive capacity.The other function of the nitrogen bath is to removethe refrigeration ("heat" of desorption) produced in theadsorber being purged. Cooling of this adsorber wouldresult in a decrease in its ability to be purged by lowpressure gas. Purge gas leaving adsorber Ads 2, con-taining hydrogen and the nitrogen swept off the bed,is warmed in the main heat exchanger E and recycled tothe compressor.

Liquid nitrogen is continuously removed from theseparator-surge vessel H and is automatically throttled(valve V-2) into the liquid nitrogen tank G.A portion ofthe liquid nitrogen and all of the dissolved hydrogenflash to a vapor. These vapors, plus the nitrogen boil-off, leave the tank. A gas refrigerating machine M re-condenses a portion of the nitrogen vapor and returnsit as a liquid to the tank. The remaining vapors arewarmed in the main heat exchanger and are dischargedfrom the plant. One of the objects of this process is torealize a high hydrogen recovery. The only hydrogenlost is that dissolved in the liquid nitrogen during thecondensation step at elevated pressure (60 atm). Hy-drogen recovery is in the vicinity of 95%. Fig. 2 showsa pilot plant built during the development.

Choice of pressure

The most critical process parameter to be chosen isperhaps the compressor head pressure. Several factors- besides the increased operating and fabricationcosts at higher pressures - must be considered. Fig. 3

1965, No. I PRODUCTION OF ULTRA-PURE HYDROGEN 29

L

o

s

t

Fig. I. Flow sheet of generator for ultra-pure hydrogen. Amm ammonia storage. Diss ammoniadissociator. B compressor; compresses the gas mixture to 60 atmospheres. C altercooler. Dpair of N H3 adsorbers, operating in turn. E main heat exchanger. F condensing coil. G liquidnitrogen tank. H separator and surge vessel. Ads 1 and Ads 2 reversing main nitrogen adsorbers,operating in turn. [branching point of gas stream. K pair of small Anal nitrogen adsorbers. Lproduct line. M gas refrigerating machine. The greater part of the liquid nitrogen produced byM is fed back to the liquid nitrogen tank, the remainder flows through line N, cools the gasmixture before entering the coil F, and then, in gaseous form, passes through the heat insulat-ing container (the "cold box") in which the cryogenic equipment is housed. In this way con-densables present in atmospheric air that could decrease the heat insulationareexcluded fromthe interior of the cold box. S surge tank. V valves. - Vi expansion valve; V2, 3, 4 and JOback pressure valves; V5-9 solenoid-operated switch-valves, electrically controlled by cycletimer (not drawn) for automatic switching of the main nitrogen adsorbers.

Not drawn are two liquid level sensors; one of them controls the liquid level in the tank bystarting and stopping M, the other opens V2 when the liquid level in H has reached a certainheight.

relates vapor and liquid equilibrium compositions ofthe hydrogen-nitrogen system at 80 "K as a functionof pressure. As shown, the hydrogen concentrationin the vapor phase increases rapidly with pressure atlow pressures until it reaches about 90 % at 20 atm. Itremains between 90 % and 93 % until IJ 0 atrn, and fallsoff steadily with increased pressure until, at 182 atm,

the critical pressure is reached. The hydrogen coneen-tration is then 55 %.If the removal of nitrogen were the only factor gov-

erning the choice of pressure, a value of about 25 atrnwould be satisfactory since operation at higher pressureshas only a slightly beneficial result. However, as is dis-cussed in the A ppendix, the performance of the revers-

30 PHILlPS TECHNICAL REVIEW VOLUME 26

Fig. 2. Photograph of pilot unit [31. For M in fig. I two gas refrigerating machines(Norelco "Cryogenerator" I'J) were used which are visible in the foreground.

"I:W802 V 80oK_/ --r-./ J""-,.

I "~./

k-'~ r-_-- r-""L-I.,...--

60

40

20

20 /,0 60 80 100 120 1/,0 160 180 atm_Ptot

Fig. 3. Equilibrium composition vs. pressure diagram for a mix-ture of H2 and N2 at 80 "K. V vapor part, L liquid part of the curve.

ing adsorbers is dependent upon the ratio of adsorbingand desorbing pressures: a minimum ratio is necessaryfor steady operation and higher ratios enhance prod-uct purity. On the other hand there are several factorsthat set a limit to the increase of this pressure ratio.As fig. 3 indicates, the hydrogen content of the vaporphase falls off rather rapidly above 100 atm. Further-more, the concentration of hydrogen dissolved in liq-uid nitrogen increases steadily with pressure. The lat-ter effect represents a loss of prod uct and a diminutionof recovery. Increased cornpression power is anotherfactor that must be considered when deterrnining the

adsorption-desorption pressure ratio. rn view of theabove factors, an operating pressure of 60 atm hasbeen selected.

The switching a/the adsorbers

The method by which the adsorbers are automaticallyswitched merits discussion. Assume that adsorberAds 1 is at compressor head pressure (60 atrn) and thatadsorber Ads 2 has been purged at low pressure (1.2atm). Referring to the flow sheet, the switch valvesare in the following position:V5 (open), V6 (closed),V7 (closed), V8 (open), and V9 (closed). Switching ofthe adsorbers occurs in two steps: I) equalization ofpressures, and 2) pressure build-up and blow-down.During the first step, valves V5, V6, V7 and V8 areclosed, and V9 is opened. This permits adsorbers Ads Iand Ads 2 to equalize in pressure. During the equali-zation step, pressure is being built up in the vessel Hsince the compressor continues to operate. As can bederived from fig. 3, it is desirable to minimize this build-up in pressure since the equilibrium nitrogen concen-tration in the vapor increases with pressure above 60atm. Thus, the vessel H is also designed to act as a

1*1 Norelco "Cryogenerator" is a registered trademark of NorthAmerican Philips Co. Inc.

1965, No. 1 PRODUCTION OF ULTRA-PURE HYDROGEN 31

surge vessel and thereby prevents excessive pressurebuild-up.The second step in the switching operation is placing

adsorber Ads 2 on the adsorption part of the cycle andadsorber Ads 1 on the desorption part of the cycle. Toaccomplish this, switch valves V6 and V7 are opened,and V5, VB and V9 are closed. At this instant, ad-sorber Ads 1 and adsorber Ads 2 are both at the samepressure, viz, about one-half of the head pressure. Ad-sorber Ads 1 then proceeds to blow down while thepressure in adsorber Ads 2 builds up to operating pres-sure. A surge tank installed on the compressor suction(S in fig. 1) prevents excessive pressure build-up whileadsorber Ads 1 is blowing down. A check valve pre-vents back flow into the ammonia dissociator.

Between the separator-surge vessel H and the switchvalves is a back pressure valve (V3). The function ofthis valve is to maintain the separator at or above nor-mal operating pressure. Were this valve not present,the reversing of the switch valves would cause the sep-arator-surge vessel to rapidly blow down from itselevated pressure to the equalized pressure of the ad-sorbers. As seen from fig. 3, this would result in someflashing of the liquid nitrogen within the separator anda higher nitrogen vapor concentration. In addition,foaming 'may result from the sudden effervescence ofthe liquid nitrogen. This in turn would lead to liquidcarry-over to the adsorbers and cause fouling of theadsorbers.The back pressure valve VlO has two functions. Dur-

ing adsorber switch-over it prevents the on-stream ad-sorber (Ads 2) from blowing down while it is in theprocess of pressure build-up. Italso prevents the flow ofgas to the other adsorber (Ads 1) which is being blowndown, prior to being purged. The level in the liquidnitrogen tank is automatically regulated by a liquidlevel sensor (not drawn in fig. 1) which controls on-offoperation of the refrigerator.The pilot plant built during the development of the

hydrogen generator and illustrated in fig. 2, was dissim-ilar to the ultimate design in that a somewhat differ-ent method was employed for purging the reversingadsorbers, and the final nitrogen adsorbers were notincluded. Nevertheless a product purity greater than99.92% was realized in this pilot 'plant.

Appendix: Calculation of the required purge flow rate

As mentioned, for economical reasons the removal of adsorbednitrogen from the adsorbers Ads 1and Ads 2 is effected by thedecompression method - rather than by the more conventionalheating of the adsorber bed - to enable automatic adsorberoperation. We shall now examine the purge flow rate necessaryto clean an adsorber within the time interval during which theother can continuously be in operation.

The starting point of our consideration is the formula:

PoY = e Ptot' (1)

in which y is the mole fraction of nitrogen in vapor phase, Po thevapor pressure of pure nitrogen at the temperature in question,Pu« the total pressure of the gas mixture and e a proportionalityfactor. A formula of this form can be used for the description ofa binary liquid-vapor system as well as for the adsorption of onecomponent of a binary gas mixture [41. In the former case thefactor e is the ratio of the mole fraction of condensable compo-nent (here N2) actually present in the vapour, to that predictedfor ideal behavior, so it then describes to what extent the be-havior of the mixture deviates from that of an ideal mixture; forthe latter e = I. When used for the description of an adsorptionprocess, the factor e will be provided with an index a.Assuming that the nitrogen concentration in the product is

negligible compared to the inlet concentration, the condition forsteady state operation of the reversing adsorbers - quantitypurged = quantity adsorbed in the same period of time - canbe expressed by the formula:

yr Wr = YIl WIl . (2)

In this formula:Wr = mass flow during adsorption (moles of mixture per hour),WIl = mass flow during purging (moles of mixture per hour),Yr inlet concentration of nitrogen during adsorption (mole

fraction),YIl outlet concentration of <nitrogen during purging (mole

fraction); YIl is assumed to be constant.With the help of (I), when related to the adsorption process, it

follows from formula (2) that:

WIl (ea/Ptot)rWr = (ea/Ptot)Il .

. . . . . . . (3)

The value of the ratio Wul Wr can be calculated if we assume thatthe ratio of the /ia values for the two pressures in question is thesame as that of the corresponding e values for a liquid-vapor sys-tem. As adsorption can be considered to be a type of condensationprocess, this assumption seems reasonable for making a roughestimate. Calculation of the e values in question gives e(1 atm) =1 and e(60 atm) = 4.2, leading to Wul Wr = 0.091. Therefore apurge flow rate as low as about 9% of the flow during the ad-sorption part of the cycle is theoretically sufficient for steadystate operation. Although this figure is only a rough estimate, itslow value clearly illustrates the technical feasibility of the iso-thermal adsorption-desorption process.

[31 This unit, built by Cryogenerators Division of North Ameri-can Philips Co. Inc., for Hamier Industries Inc., was in operationfor some time at the Commercial Steel Treating Co., Detroit,Mich. (U.S.A.).[41 See M. J. Hiza and A. J. Kidnay, Advances in cryogenic engi-neering 8,174-182,1963.

Summary. An economically operating unit has been developedwhich is capable ofproducing ultra-pure hydrogen (99.995 %) at arate up to 300 m3 (NTP) per hour. The design of this hydrogengenerator evolved from experimental work undertaken by Phi-lips Laboratories, Briarcliff Manor, U.S.A., and from operatingexperience gained from a pilot installation. The process consistsof two main steps, viz, 1) the dissociation of ammonia into hy-drogen and nitrogen, and 2) the separation of nitrogen from thegas mixture. This separation is mainly performed by coolingthe mixture to 80 OK at a pressure of 60 atm. A liquid con-sisting mainly of nitrogen and a vapor consisting mainly ofhydrogen are then obtained. For the final purification adsorbersare used, viz, two reversing adsorbers contained in the liquidnitrogen tank and an additional pair in the product line. Whenone of the reversing adsorbers is in operation (at 60 atm), theother is isothermally purged at low pressure and vice versa. Theother adsorbers are regenerated in the conventional manner.

32 PHILJPS TECHNICAL REVIEW. VOLUME 26

Recent scientific publications by the staff ofthe Philips laboratories and factories

Reprints of those papers not marked with an asterisk * can be obtained free of chargefrom the Philips Research Laboratories, Eindhoven, Netherlands, where a limited numberof reprints are available for distribution. Requests need only quote the reprint numbergiven in italics at the end of each entry.

W. Albers, C. Haas and H. J. Vink: Quenching effectsand the determination of the existence region of semi-conducting compounds.Philips Res. Repts 18, 372-376, 1963 (No. 4). R 486

E. Andrich: Anwendung von PTC-Widerständen. (Ap-plication of PTC thermistors; in German.)Elektron. Rdsch. 17, 132, 135, 136, 138, 1963 (No. 3).

A 73

V. Belevitch: Factorization of scattering matrices withapplications to passive-network synthesis.Philips Res. Repts 18, 275-317, 1963 (No. 4). R 481

G. Blasse: New type of superexchange in the spinelstructure; some magnetic properties of oxidesMe2+C0204 and Me2+Rh204 with spinel structure.Philips Res. Repts 18, 383-392, 1963 (No. 5). R 488

G. Blasse: Magnetic properties of oxidic spinels con-taining tetravalent manganese.Philips Res. Repts 18, 400-404, 1963 (No. 5). R 490

G. Blasse: Magnetic-garnet phases containing penta-valent antimony.Philips Res. Repts 19, 68-72, 1964 (No. I). R 500

G. Blasse and J.F. Fast: Néel temperatures of someantiferromagnetic oxides with spinel structure.Philips Res. Repts 18, 393-399, 1963 (No. 5). R 489

G. Brouwer: Graphical analysis of stationary surfaceinjection phenomena in semiconductors.Philips Res. Repts 18, 432-446, 1963 (No. 5). R 493

G. P. Brouwer: Brightness variations on tungsten rib-bon.Philips Res. Repts 18, 361-371, 1963 (No. 4). R 485

S. Duinker and J. A. Geurst: Long-wavelength responseof magnetic reproducing heads with rounded outeredges.Philips Res. Repts 19, 1-28, 1964 (No. I). R 496

R. A. Ford, E. Kauer, A. Rabenau and D. A. Brown:The electronic states of octahedral and tetrahedralMrr+ in a, f3 and y manganous sulphide.Ber. Bunsenges. phys. Chemie 67,460-465, 1963 (No. 5).

A 76

A. H. Gomes de Mesquita, C. Langereis and J. J. Leen-houts: The structure of NbSn2.Philips Res. Repts 18, 377-382, 1963 (No. 5). R 487

N. Hansen and W. Littmann: Automatisches Gerät zurBestimmung der Oberflächengröf3e feinteiliger Sub-stanzen. (Automatic device for determining the surfacearea of powders; in German.)Z. Instrumentenk. 71, 153-159, 1963 (No. 6). A 77

W. F. Knippenberg: Growth phenomena in silicon car-bide. (Thesis Leiden, June 1963.)Philips Res. Repts 18, 161-274, 1963 (No. 3). R 480J. L. Meijering: Miscibility gaps in ferromagnetic alloysystems.Philips Res. Repts 18, 318-330, 1963 (No. 4). R 482B. Okkerse: Consecutive Laue and Bragg reflexions inthe same perfect crystal.Philips Res. Repts 18, 413-431, 1963 (No. 5) .. R 492H. G. Reik: Derivation of Holstein's expression for thehigh temperature mobility of small polarons fromKubo's formula.Physics Letters 5, 236-237, 1963 (No. 4). A 72E. Roeder and M. Klerk: Untersuchungen mit demElektronenstrahl- Mikroanalysator an druckgesintertemTantalkarbid mit geringem Mangan- und Nickelzusatz.(Investigations on hot-pressed tantalum carbide con-taining small additives of manganese and nickel, bymeans of electron-probe X-ray emission microanalysis;in German.)Z. MetalIk. 54, 462-470, 1963 (No. 8). A 75K. J. Schmidt-Tiedemann: Tensor theory of the con-ductivity of warm electrons in cubic semiconductors.Philips Res. Repts 18, 338-360, 1963 (No. 4). R 484P. Schnabel: Four-point method for measuring theanisotropy of resistivity.Philips Res. Repts 19, 43-52, 1964 (No. I). R 498J. Schröder: Apparatus for determining the thermalconductivity of solids in the temperature range from20 to 200°C.Rev. sci. Instr. 34, 615-621, 1963 (No. 6). A 74F. Sellberg: Plasma resonance in a germanium rod.Philips Res. Repts 19, 53-67, 1964 (No. I). R 499N. C. de Troye: Some properties of symmetric switch-ing functions.Philips Res. Repts 18, 331-337, 1963 (No. 4). R 483J. L. Verster: On the use of gauzes in electron optics.(Thesis Delft, Oct. 1963.)Philips Res. Repts 18, 465-605, 1963 (No. 6). R 495J. Verweel: Permeability of dense nickel-zinc ferritesin polarizing fields.Philips Res. Repts. 19,29-42, 1964 (No. I). R 497K. J. de Vos: Refined replicating technique for studyingstructural changes in alnico-type permanent-magnetalloys.Philips Res. Repts 18, 405-412, 1963 (No. 5). R 491W. J. Witteman and Th. Werkman: Cylindrical high-pressure apparatus.Philips Res. Repts 18,447-463, 1963 (No. 5). R 494

Volume 26, 1965, No. 1 Published 14th January 1965pages 1-32