A closed-loop proposal for hydrogen generation using steel waste

INTERNATIONAL JOURNAL OF ENERGY RESEARCHInt. J. Energy Res. 2009; 33:481–498Published online 2 December 2008 in Wiley InterScience(www.interscience.wiley.com). DOI: 10.1002/er.1491

A closed-loop proposal for hydrogen generation using steel wasteand a prototype solar concentrator

Abdul-Majeed Azad�,y, Sathees Kesavan and Sirhan Al-Batty

Department of Chemical Engineering, University of Toledo MS 305, 2801 W. Bancroft St., Toledo, OH 43606-3390, U.S.A.

SUMMARY

An economically viable and environmental-friendly method of generating PEM grade hydrogen has been proposed andis by the reaction of certain metals with steam, appropriately called ‘metal–steam reforming’—MSR. The drawbacks ofconventional processes (hydrogen and carbothermic reduction schemes) are overcome by resorting to solution-basedreduction schemes and are made economically feasible using iron oxides from steel industry’s mill-scale waste. A novelaqueous-based room temperature technique using sodium borohydride (NaBH4) as the reducing agent has beendeveloped that produces highly active nanoscale iron particles (�40 nm). By using hydrazine as an inexpensive and,compared with NaBH4, more stable reductant, body centered cubic iron particles with�5 nm edges were obtained viasolvothermal process under mild conditions from acid digested mill-scale waste. The nanoscale zerovalent iron (nZVI)powder showed improved kinetics and greater propensity for hydrogen generation than the coarser microscale iron. Therate constants for the MSR were obtained for all the reduction schemes employed in this work and are given bykhydrogen 5 0.0158min�1 kcarbon 5 0.0248min�1 ksodiumborohydride 5 0.0521min�1 and khydrazine50:1454 min�1, assumingfirst order kinetics. Another innovative effort converted the magnetite waste directly into nZVI under solvothermalconditions, thus obviating the sluggish and time-consuming acid dissolution step. This particular aspect has significantramification in terms of time and cost of making the iron precursor. To initiate and sustain the somewhat endothermicMSR process, a solar concentrator consisting of a convex polyacrylic bowl with reflective aluminum coating wasfabricated and evaluated. This unique combination of mill-scale waste as iron source, hydrazine as reductant, mildprocess conditions and solar energy as the MSR actuator obviates several drawbacks plaguing the grand scheme ofproducing and delivering pure and humidified H2 to a PEMFC stack. Copyright r 2008 John Wiley & Sons, Ltd.

KEY WORDS: hydrogen generation; metal–steam reforming; mill-scale; nanoscale zerovalent iron; X-ray diffraction;electron microscopy; solar concentrator

*Correspondence to: Abdul-Majeed Azad, Department of Chemical Engineering, University of Toledo MS 305, 2801 W. Bancroft St.,Toledo, OH 43606-3390, U.S.A.yE-mail: [email protected]

Contract/grant sponsor: Department of Energy (DOE)Contract/grant sponsor: Edison Materials Technology Center (EMTEC)

Received 10 October 2008

Accepted 10 October 2008Copyright r 2008 John Wiley & Sons, Ltd.

1. INTRODUCTION

PEM fuel cells require hydrogen of highest purity,which is either difficult to come by, or comes at apremium price if one starts with fossil fuels [1,2].In the light of the serious economic constraintsand safety concerns associated with hydrogen inthe context of a truly global hydrogen economy,one needs to find innovative ways and means ofgenerating, storing, transporting and supplyinghydrogen to the end users. Interestingly, thequestion of its storage and supply arises only afterit is produced in large enough quantity so as tocater to the need of its use on the national andglobal scale.

Metal–steam reforming (MSR) using iron is asafe and economically viable method of hydrogengeneration from inexpensive raw materials. Theuse of iron and iron waste for hydrogen generationvia MSR is known. For example, the reaction thatoccurs around 6001C has long been known as oneof the promising ways of generating H2 [3–8]:

3Fe ðsÞ þ 4H2O ðgÞ ! Fe3O4 ðsÞ þ 4H2 ðgÞ ð1Þ

The reverse reaction could be viewed as ahydrogen storage scheme. The theoreticalamount of hydrogen being produced/stored is4.8wt% which corresponds to ca. 4211 l H2 l

�1

Fe at standard temperature and pressure. While ona mass basis, H2 has high energy content(142MJkg�1, lower heating value [LHV] of33.3 kWhkg�1); its volumetric energy density isconsiderably lower. At pressure of 680 atm, LHVof H2 is about 1.32 kWh l�1. The same number forliquid H2 is 2.35 kWh l�1. In contrast to this, thecorresponding energy density for gasoline is8.88 kWh l�1. Hence, the following on-board H2

generation scheme is proposed. In a practicallyviable design, the elemental iron is packed intocartridges that are loaded on the vehicle; thesomewhat endothermic nature of the reactionrepresented by Equation (1) necessitates that theelemental iron cartridges be preheated. Additionof steam to the preheated cartridges producessteam–hydrogen mixture of high purity that couldthen be supplied directly to PEMFC on thevehicle. After the conversion of water intohydrogen, the cartridges with the spent iron

oxide are exchanged for the new ones packedwith fresh elemental iron. The spent iron oxide isagain converted into ready-to-use active metalusing the novel solution-based reduction schemesemployed in this work. Thus, the process of H2

generation via MSR in an efficient and cyclicfashion consists of: (a) iron–steam reaction and,(b) reduction of the spent iron oxide into activeelemental iron for the next cycle.

However, the task of making this seeminglysimple process technologically viable andsustainable is challenging. Most importantly, thekinetic of metal oxidation (forward) and oxidereduction (reverse) as per Equation (1) ought to besignificantly improved, in order to mitigatesintering and coarsening of iron and iron oxideparticles during repetitive hydrogen generation-oxide conversion cycles.

With this goal in mind, we have developed aneffective technique of producing H2 by using theso-called ‘mill-scale’ from steel industry, as an ironsource, in a way that is consistent with the mostsought-after criteria: environmental-friendliness,availability and low cost. An attractive andinexpensive source of iron is the waste fromthe steel industry in the form of magnetite,provided magnetite can be reduced to metalliciron. Mill-scale is a porous, hard and brittlecoating of several distinct layers of iron oxides(predominantly Fe3O4) formed during thefabrication of steel structures. It is magnetic innature with iron content up to as high as 93%.Prior to sale or use, the steel structures must bedenuded of this oxide scale. Most of the steel mill-scale waste usually ends up in landfills. Thecurrently known art, however, requires a greatdeal of energy in the practice of making iron fromthe steel industry waste, as it necessitatesemploying high temperatures.

From a kinetic perspective, it can be envisionedthat all process parameters and experimentalconditions remaining identical, the reactioninvolving nanoscale iron would possess andexhibit higher propensity of reaction withsteam according to Equation (1) than themicron-scale or coarser Fe derived via eitherof the high-temperature processes, namely,hydrogen reduction or carbothermic reduction.

A.-M. AZAD, S. KESAVAN AND S. AL-BATTY482

Copyright r 2008 John Wiley & Sons, Ltd. Int. J. Energy Res. 2009; 33:481–498

DOI: 10.1002/er

Such processes are not efficient ways of producingiron from the oxide as they are energy intensive.Moreover, the use of high temperatures results inthe formation of coarse iron that is not as active or‘potent’ and is unlikely to yield hydrogen as perthe theoretical prediction owing to the limitationof gas–solid reaction via diffusion. In addition,one technique requires precious H2, while thesecond process generates carbon dioxide gas(stoichiometrically, 2 moles for every three atomsof iron produced). Thus, the regeneration ofelemental iron from the spent oxide via hydrogenreduction is unattractive in a commercial settingand makes the recovery of iron from steel wastemore expensive than disposing such waste into alandfill. The carbothermic reduction, on the otherhand, produces micron size iron particles by ahigh-temperature energy-intensive process, leadingto CO2 generation thus defeating the sustainabilityaspect.

We have recently carried out a systematic andthorough investigation on steel waste material toestablish the feasibility of hydrogen productionand recycling of the oxide on a laboratory scalein the least energy intensive and mostenvironmentally benign way [9,10]. It was foundthat the mill-scale samples could be quantitatively(�100%) reduced to elemental iron, both via H2

and carbothermic reduction, as has beendemonstrated by others. However, with anobjective to eliminate the use of high temperatureand H2 gas and, to mitigate the CO and/or CO2

emission, solution-based techniques have beendevised resulting in the formation of nanoscaleelemental iron particles from mill-scale powder ator near room temperature. In the present work,details of novel solution-based reductiontechniques, whereby the mill-scale is convertedinto highly active nanoscale iron powder, isreported. The spent oxide after the metal–steamreaction (1) could also be quantitatively recycledto yield highly active nanoscale metallic iron again,without the loss of its reactivity; thus, thehydrogen generation capacity of the samematerial over several cycles is maintained.Moreover, the method totally obviates the issueof sintering and coarsening of the iron/iron oxidethat is a direct result of high-temperature use.

Consequently, the possibility of deactivationduring the cyclic operation of MSR becomes anon-issue.

The methodology employed to obtain value-added high-activity nanoscale zerovalent iron(nZVI) powder from the mill-scale waste, and itsuse to produce hydrogen via metal–steam reactionis also discussed in this paper. Also discussed is thefeasibility of direct conversion of the solid wasteinto nanoscale iron powder without resorting to itsdissolution in mineral acids prior to reduction by asolvothermal process using hydrazine as thereductant. Finally, since the MSR process isslightly endothermic and requires energy to initiatethe process, and to make the scheme of hydrogenproduction attractive and viable economically on alarge scale, a heat source alternative to conventionalheating in a furnace was considered. A prototypesolar concentrator was conceived and built with thegoal to utilize it to concentrate the solar energy forachieving the required temperature (�6001C) andrunning the MSR reaction. The temperature profileof the prototype solar bowl was also established.

2. EXPERIMENTAL

2.1. Sample processing

Mill-scale samples were procured from threedifferent vendors: North American Steel (KY),Midrex (NC) and Nucor-Yamato (AR). Thesamples from North American Steel were labeledas pickled (NAS-P), sludge rinse (NAS-S) andentry loop (NAS-E) by the supplier, referring tothe chemical treatments used for stripping theoxide layer from the steel structures. Samples fromMidrex and Nucor–Yamato are hereafter referredto as MR and NY, respectively, for brevity. All thefive mill-scale samples were ball-milled followed byattrition milling in 2-propanol and sieved through325 mesh (average particle size p45 mm) prior toany processing or characterization. Hydrogenreduction of the sieved mill-scale powders wascarried out at 9001C for 8 h using high-purity dryH2 at �101 kPa. For the carbothermic reduction,the sieved mill-scale powders were mixed withactivated carbon in the molar ratio of 1:3,pelletized with small amount of polyvinyl alcohol

A CLOSED-LOOP PROPOSAL FOR HYDROGEN GENERATION 483


DOI: 10.1002/er

to aid compaction and heated at 11001C for 4 h,using nitrogen as a blanket gas. The carbothermicreaction can be represented as:

Fe3O4 ðsÞþ3C ðsÞ ! 3Fe ðsÞ þ 2CO ðgÞþ CO2 ðgÞ ð2Þ

As mentioned earlier, regeneration of elementaliron from the spent oxide via hydrogen orcarbothermic reduction is an energy-intensiveprocess and hence undesirable in a commercialsetting. To obviate these predicaments, solution-based procedures were developed. In a typicalexperiment, 6 g of the Midrex (MR) or Nucor-Yamato (NY) samples were first digested in 200mlaqua regia. The undissolved non-magnetic impuritieswere removed by centrifugation and the clear yellowsolution was made up to 250ml with distilled water.In one case, this acid solution was treated withaqueous sodium borohydride (NaBH4, Alfa Aesar)solution in the presence of sodium hydroxideas pH stabilizer [11–13]. Equation (3) representsthe borohydride reduction reaction

FeðH2OÞ3þ6 þ 3BH�4 þ 3H2O

¼ Fe0 þ 3BðOHÞ3 þ 10:5H2 ð3Þ

This resulted in immediate formation of nanoscaleiron particles (30–40nm)—a clear advantage overthe agglomeration observed during hydrogen andcarbothermic reduction.

Though the borohydride reduction of mill-scalewaste to zerovalent nanoiron has distinctadvantages, it is not an economically feasibleoption, as the reductant by itself is deemedhydrogen storage material and is quite expensive.Moreover, its aqueous solutions are rather unstable.Therefore, in a second method, iron nanoparticleswere prepared by a simple solvothermal processusing hydrazine monohydrate, which is a lessexpensive and more stable reducing agent [14,15].The reaction is represented by the followingequation:

4Fe3þ þ 12OH� þ 3N2H4

¼ 4Fe0 þ 12H2Oþ 3N2 ð4Þ

Preliminary experiments were carried out withcommercial FeCl3 � 6H2O (Alfa Aesar, ACS min97%) in order to first establish and optimize the

experimental conditions and procedural protocolsfor the mill-scale reduction. In a typical experiment,4 g of FeCl3 � 6H2O were dissolved in 20ml absoluteethanol. Sodium hydroxide pellets (5 g) were addedand the mixture was stirred for 2 h. To this, 15ml ofhydrazine monohydrate (99%+, Alfa Aesar) wereadded and the mixture was transferred to a 1-lcapacity stainless steel autoclave vessel (AutoClaveEngineers, PA) whose internal wall was lined withzirconium metal to avoid corrosion under stronglyalkaline conditions. The vessel was flushed withnitrogen to expel air and pressurized to about 5 atm.The autoclave was maintained at 1001C for 10h andthen naturally cooled to room temperature. Theproduct was centrifuged several times with waterand ethanol to remove the impurities. The size of thefinal particles was found to fall in two regimes:30–40nm and 300–450nm. This is in conformity tothe results reported earlier [14].

Reduction of the mill-scale powder dissolved inacid followed a similar procedure except that 10mlof mill-scale (30.5 g mill-scale/1000ml acidsolution) were added to the highly alkalinesolution (pH410) and 20ml of hydrazinemonohydrate were used. In this case, the size ofthe iron nanocrystals decreased to 5 nm.

For large-scale production of zerovalentnanoiron, 33ml of the mill-scale acid solution(equivalent to 1 g mill-scale) were used. Theautoclave reduction runs with hydrazineconducted at 5 and at 10 atm pressures yieldedblack magnetic particles, which were found to bemagnetite (Fe3O4) rather than the expectedmetallic iron phase, based on the XRD results.Therefore, the method was slightly modified wherethe mill-scale solution was reacted with sodiumhydroxide to precipitate Fe(OH)3 first, followed bycentrifuging and vacuum drying. In yet largerbatch conversions, it was also required to increasethe pressure from 10 atm to about 50 atm alongwith the amount of hydrazine monohydrate inorder to scale up the nano iron batch from �0.25to 4 g. 130ml of the mill-scale solution wasprecipitated with sodium hydroxide and theprecipitate centrifuged and vacuum dried at 601Cfor 8 h followed by solvothermal reduction with150ml hydrazine monohydrate and at 50 atminitial N2 pressure. This produced 4 g of nZVI.



DOI: 10.1002/er

The autoclave temperature was maintained at1001C in all the runs.

When a relatively large batch was used (33ml ofthe mill-scale acid solution, equivalent to 1 g mill-scale), the autoclave reduction runs with hydrazineconducted at 5 and at 10 atm pressures yieldedblack magnetic particles. They were found to bemagnetite (Fe3O4) rather than the expectedmetallic iron phase, based on the XRD results.Therefore, the method was slightly modifiedwhere the mill-scale waste (only sieved through325 mesh) was directly reduced solvothermallyeliminating the sluggish and expensive aciddigestion process. But this is achieved at the costof some increase in iron particles size.

2.2. MSR reaction

MSR experiments were carried out in a cylindricalquartz reactor (OD5 1 in; ID5 0.78 in) at 6001Cin a Lindberg/Blue (NC) tubular furnace with aPID controller. Helium was used as an inertbackground during the temperature ramp. Afterthe sample attained the desired temperature, a50:50 v/v mixture of steam (water preheated to4001C) and He was introduced into the reactor atambient pressure. The reaction was allowed to rununtil no hydrogen signal was detected in theexhaust stream. The effluent gas was analyzedusing a Shimadzu gas chromatograph (GC-2010)equipped with a Pulsed Discharge Helium Ioniza-tion Detector (PDHID—Model D-4-I-SH17-R)using He as the carrier gas.

2.3. Characterization

Chemical analysis of the as-received mill scalesamples was done by X-ray fluorescence (XRF)technique. The X-ray powder diffraction patternson the raw, reduced and post-MSR samples werecollected at room temperature on a Philipsdiffractometer (PW 3050/60 X’pert Pro), usingmonochromatic CuKa1 radiations (l5 1.54056 A)and Ni filter. Morphology of the samples beforeand after reduction and steam reforming wereexamined by a Philips scanning electron micro-scope (XL30 FEG). The morphology and crystalstructure of the nanoscale iron particles werestudied by transmission electron microscopy

(JEOL 3010) operated at 300 kV equipped withselected area electron diffraction and the energydispersive spectroscope (EDS). The samples forTEM analyses were prepared by ultrasonicating avery dilute suspension of the nZVI particles inacetone for 5min followed by placing a drop ontoa copper TEM grid and drying by evaporation inambient air. Hydrogen yield in the MSR reactionwas quantified by a Shimadzu GC-2010 unit fittedwith a molecular sieve 5A PLOT capillary column(30m� 0.32mm) and a PDHID detector. Class-VP software was used for the peak analysis andquantification.

3. RESULTS AND DISCUSSION

3.1. High-temperature mill-scale reduction

Table I shows the XRF analyses of the as-receivedmill-scale samples from various vendors. As can beseen, two of the three NAS samples werepredominantly rich in Cr (52–67% by weight),while the third sample contained about 18% Crand 8% Ni by weight. On the other hand, MR andNY samples contained 92–94wt% Fe and nosignificant Cr or Ni. The presence of chromium insignificant amounts has serious implications on theenvironmental aspects of the process.

Hence, if the NAS samples were to be used forthe H2 production, one has to isolate Fe, whichmeans disposing nickel and hexavalent chromium,the latter being a known health hazard in watersystem. Since Ni and Cr-ferrites (seen in the XRD

Table I. XRF analyses of the mill-scale samples used inthis study.

M

Wt%

NAS-P NAS-S NAS-E Midrex Nucor–Yamato

Fe 29.13 42.3 70.42 93.92 92.91Cr 66.57 51.59 18.11 0.16 0.168Mn 2.75 2.49 1.4 1.19 1.3Ni 0.178 0.872 8.24 0.232 0.189Ca – 0.128 – 1.33 3.14Si 0.332 0.267 0.357 1.49 1.05Mo 0.282 1.3 0.53 – –Cu – 0.337 0.366 0.608 0.515V 0.412 0.259 – – –



DOI: 10.1002/er

patterns of the as-received mill-scale specimenfrom NAS, but not shown here) are difficultto breakdown, preliminary reduction experimentsfailed to generate elemental iron quantitativelyfrom NAS samples, as revealed by XRDpatterns and corroborated by thermogravi-metric experiments. Hence, the NAS series wasabandoned in further studies.

On the other hand, hydrogen and carbothermicreduction of MR and NY samples resulted incomplete conversion of the scales into elementaliron. Gravimetric analyses of the samplesbefore and after reduction yielded an averageweight loss of 27.5 wt% (hydrogen reduction) and27wt% (carbothermic reduction) for the MR andNY samples, respectively; this agrees very wellwith the theoretical value of 27.64% weight lossfor the reduction of Fe3O4 to Fe. The morphologyof the mill-scales before and after hydrogenreduction is compared in Figure 1, while themicrostructural features of the MR and NYsamples after carbothermic reduction are shownin Figure 2.

3.2. Sodium borohydride-based reduction of themill-scale

The TEM images and the EDS spectrum of thenanoscale iron particles derived from acidicsolution of Midrex samples are shown inFigures 3 and 4, respectively; Cu and Mo tracesin the EDS originate from the Cu–Mo grid usedfor the TEM imaging. The crystallite size calcu-lated from the Scherrer equation [16] applied tothe h110i reflection in the X-ray diffractogram ofthe solution-derived nanoscale iron is �26 nm.

3.3. Hydrazine-based reduction of the mill-scale

The solvothermal process using hydrazine as thereducing agent described by Xiaomin et al. [14] forthe one-step synthesis of nanoscale iron was also usedby Li et al. [15] for nickel. It has been found that theuse of hydrazine as a reductant is particularlyadvantageous as it leads to the complete reductionof Ni2+ ions in solution to Ni0, as compared with theformation of Ni–B, Ni2B or Ni3B or Ni–P nano-particles with sodium borohydride or hypophosphate

Figure 1. SEM features of the mill-scale samples: (a) MR and (b) NY, before (1) and after (2) H2 reduction.



DOI: 10.1002/er

Figure 2. Morphology of the carbothermically reduced samples: (left) MR and (right) NY.

Figure 3. TEM images of the iron particles obtained by room temperature processing of Midrex waste.Scale bar: 20 nm.

Figure 4. TEM–EDS pattern of the solution-derived nano iron particles.



DOI: 10.1002/er

[15,17]. In the context of the present work, use ofhydrazine provides the following additional andrelevant advantages:

1. precise control of size (subject to processingconditions) and shape of nanoparticles,

2. the ultimate chemical state is zerovalent metal-lic species and not an intermetallic,

3. the process is greener as no noxious fumes orchemicals are emitted whose disposal couldpose a challenge of problem; the only bypro-duct is nitrogen as the hydrogen is utilized inthe reduction scheme,

4. the process is less energy consuming because thereaction is carried out in an autoclave (closedsystem) under moderately high pressure andmild temperature and

5. hydrazine is chemically more stable and cheap-er, compared with sodium borohydride and theprocess can be scaled up easily for large-scaleproduction.

The phase composition of the product wascharacterized by XRD and a typical pattern of thesample obtained using 10ml of the mill-scalesolution at 5 atm N2 pressure is shown inFigure 5. The most prominent diffraction peaks

at 2y5 44.69, 65.03 and 82.3 could again beindexed as those corresponding to the h110i,h200i and h211i planes, respectively, of the phasepure a-Fe (bcc); no iron oxides or hydroxides weredetected. The strong and sharp peaks revealed thatthe iron powder synthesized via hydrazinereduction under solvothermal conditions washighly crystalline.

Figure 6 shows the TEM image of the nZVIparticles obtained from the hydrazine reduction of asmall batch of mill-scale under solvothermalconditions (1001C/5atm). The average size of thenZVI is around 5nm; this is very significant frompoint of view of hydrogen generation via MSR assmaller particle size means larger surface area, whichin turn means higher activity for the desired reaction(as is evident from the hydrogen generation plotsshown later). Moreover, the particles crystallize in theinherent cubic motif. Thus, clearly, the steelindustry’s mill-scale waste is a viable source ofhigh-quality nZVI for the generation of high-purityhydrogen in a green way for PEMFCs.

3.4. Single-step direct reduction of the mill-scale

The scheme hitherto employed consists ofbringing the mill-scale waste into solution by acid

Figure 5. X-ray diffraction patterns of the nanoscale iron obtained by solvothermal process (duration: 10 h) using 10mlof the mill-scale solution at 1001C and initial nitrogen pressure of 5 atm.



DOI: 10.1002/er

dissolution followed by reduction into nanoscaleiron. This acid dissolution step, however, is asomewhat slow process and will have appreciableimpact on the scale-up scheme. This critical step inaddition to the expensive acids employed iseliminated by resorting to the direct one stepreduction of the mill-scale waste.

In a typical experiment, 10 g of the mill-scalewaste from Midrex was ball-milled and sievedthrough 325 mesh. It was mixed with 10 g NaOHdissolved in 50ml ethanol and 50ml hydrazinemonohydrate. The pH of the solution wasmaintained at 10.5, the magnetite waste wasreduced to nZVI in an autoclave at 60 atm initialN2 pressure for 10 h at 1001C. Figure 7 shows theSEM images (with inserted TEMs) of the nZVIobtained by the direct conversion of the solidwaste. The XRD pattern conforms to that of bcca-Fe shown in Figure 5 and hence is notreproduced again.

In order to optimize the conditions for theproduction and scale up of the nZVI from themill-scale waste, the solvothermal processing wasconducted in a broad domain of experimentalduration and the incumbent pressure. The resultsare summarized in Figures 8 and 9. Figure 9 showsthe X-ray diffraction patterns of the productobtained after various dwell times during

reduction. For reaction time between 1 and 2h,the product is a mixture of Fe and Fe3O4. When thereaction time is increased to 3h and above, completeconversion to nZVI is realized. The effect ofpressure on the reaction dominance is shown inFigure 10.

Hence it can be inferred that for a 5 g batchconversion, the threshold reaction time is less than3 h and the threshold pressure is �90 atm.

3.5. MSR results

Figure 10 compares the hydrogen yield from MSRreaction carried out at 6001C by using elementaliron obtained via a number of reduction techni-ques employed in this work: namely, usinghydrogen, carbothermic, sodium borohydrideand hydrazine. For a ready appreciation of therelative propensity of various iron samples forhydrogen generation via MSR reaction vis-a-visthe schemes employed for their recovery from themill-scale waste, representative microstructuralfeatures are also included.

The correlation between the particle sizeand kinetic profile of hydrogen generation isquite evident. Even though the amount ofhydrogen generated per gram of iron sample(given by the area under the respective curve) is

Figure 6. TEM image of the ZVI from mill-scale via hydrazine reduction; scale bar: 5 nm.



DOI: 10.1002/er

Figure 7. SEM images of the nZVI obtained via ‘direct’ reduction of the mill-scale by hydrazine, showingtwo size regimes: 300–450 nm and 30–40 nm (not seen clearly). Inset shows TEM images of the ZVI conforming

to the nano artifacts.

Figure 8. XRD analyses of the product obtained via solvothermal hydrazine reduction at 60 atm initial N2 pressure,1001C using 5 g mill-scale batch.



DOI: 10.1002/er

nearly the same and of PEM grade puritywith no carbon compound peaks within theprecision limits of such measurements by gaschromatography, the mode of hydrogen genera-tion is very characteristic of the morphologicalfeatures of the iron used.

For instance, the fastest hydrogen releasekinetics were observed with the cubic 5 nm nZVI

derived from solvothermal hydrazine reductionroute, followed by those with 40 nm nZVI particlesobtained via sodium borohydride reduction atroom temperature. In both these cases, about 90%of hydrogen was released within the first 10–15minof the MSR initiation. In comparison, the effect oflarger grain size (micro vs nano) of iron particlesderived via high-temperature hydrogen andcarbothermic reduction techniques is reflected inthe slower and more sluggish kinetics; in both thecases, hydrogen generation is seen to continueeven after about 2 h of initiation. However,irrespective of the technique employed togenerate iron from the mill-scale waste, nearstoichiometric amount of hydrogen is obtained ineach case. In the case of solution route, inparticular, the methodology is very attractive, asthe magnetite formed after the steam reformingcan be recycled to generate fresh batch of activenanoiron particles for subsequent MSR cyclesagain.

The SEM images of the sodium borohydride(NBH)-derived nZVI powder before and afterthe MSR reaction (at 6001C) are shown inFigure 11. The TEM and high-resolution TEMimages of the post-MSR NBH-derived nZVI

Figure 10. Comparison of hydrogen yield from MSR reaction at 6001C and the mode of hydrogen release using ironfrom mill-scale after reduction by hydrogen and carbothermic (hi-temp) and borohydride and hydrazine (lo-temp).

Figure 9. Threshold process parameter for the 5 g batchmill-scale conversion at 1001C with an initial N2

pressure of 60 atm.



DOI: 10.1002/er

powder, are shown in Figures 12(a) and (b),respectively. Slight increase in the particle sizedue possibly to the conversion of iron (bcc)

into magnetite (spinel) and some sinteringcould be seen. The XRD pattern of the post-reformed sample is shown in Figure 12(c),

Figure 12. (a) TEM (scale bar: 100 nm), (b) HRTEM (scale bar: 5 nm) and (c) XRD pattern of the post-MSR nanomagnetite residue.

Figure 11. SEM images of NBH-derived nZVI before (left) and after (right) MSR reaction.



DOI: 10.1002/er

which conforms to that of magnetite, Fe3O4

(ICDD 79-0418).

3.6. Kinetics study of the MSR reaction

Assuming that the MSR reaction is not diffusion-controlled, as steam will have instant access to theentire solid surface of the nanoscale iron particles, itcould be viewed as a surface-controlled one. Thus,the reactor could be modeled to the one approx-imating catalyst deactivation due to poisoning; inthe present case, this is a valid assumption since theconsumption of the active iron surface by steamleads to the formation of inactive iron oxide (akinto deactivation), releasing hydrogen. The deactiva-tion rate, rd (which happens to be the overallreaction rate in this case) is given by

rd ¼ �ðda=dtÞ ¼ k0d � Cmp � aq ð5Þ

where ðda=dtÞ is the rate of change of activity, k0dthe deactivation rate constant and, Cp the concen-tration of steam.

Since steam is in large excess, its concentrationcan be assumed to be constant throughout thereaction. Hence,

rd ¼ �ðda=dtÞ ¼ k0d � aq ð6Þ

Assuming first order kinetics for this process, wecan write: �ðda=dtÞ ¼ k0d � a, which uponintegration, yields: a ¼ e�kd�t or �ln a5 kt.

The normalized experimental data (hydrogengeneration rate, a) presented in Figure 11 isplotted as a function of time (�ln a vs time),whose slope gives the reaction rate constant. Thevalues of k determined for the MSR process, usingelemental iron derived from different reductiontechniques employed in this work, are summarizedbelow.

khydrogen 5 0.0158min�1

kcarbon 5 0.0248min�1

ksodiumborohydride 5 0.0521min�1

khydrazine 5 0.1454min�1

In the absence of experimental data at morethan one temperature (6001C), the activationenergy for the MSR process was estimated usingthe standard thermodynamic data and by invokingthe collision theory concept. For a pseudo first-

order reaction, the pre-exponential factor A isgiven by:

A ¼ ðkB � T=hÞ � expðDS#0 =RÞ ð7Þ

where kB is the Boltzmann constant(5 1.38066� 10�23 JK�1), h is Planck’s constant(5 6.626� 10�34 J s), T, the reaction temperature(K), DS0

#, the standard entropy of magnetite,Fe3O4 (5 151.5332 JK�1mol�1), and R, the gasconstant5 8.314 Jmol�1K�1.

This gives A5 1.49764� 1021 s�1 5 8.9858�1022min�1. Using the Arrhenius equation,k5A� exp(�Ea/RT), and substituting the valueof the pre-exponential factor A, and the reactionrates in the above equation, the activation energiesare computed and are found to be 413.71,410.44, 405.05 and 397.60 kJmol�1 for theproduction of iron via hydrogen, carbothermic,sodium borohydride and the hydrazine-basedsolvothermal reduction, respectively. As can beclearly seen, the effect of particle size isreflected more pronouncedly in the values ofreaction rate constant (differing by an orderof magnitude), than the activation energywhose value remains practically identical. Itshould, however, be pointed out that this issimply fortuitous as the MSR reaction has sofar been conducted at a solitary temperatureof 6001C.

3.7. Fabrication, evaluation and interfacing of theprototype solar concentrator

As stated earlier, the MSR process is slightlyendothermic and hence requires energy to initiatethe process. In order to make the scheme ofhydrogen production attractive and viable eco-nomically on a large scale, a heat source alter-native to conventional heating in a furnace mustbe sought. One such source could be the solarenergy. A prototype design of a solar concentratorwas conceived with an aim to utilize this design toconcentrate the solar energy to achieve therequired temperature (6001C) for running theMSR reaction. The prototype solar concentratorcomprised a convex polyacrylic sheet, 46.5 in indiameter and 36 in in focal length, coated with analuminum reflective layer, constructed by Replex



DOI: 10.1002/er

Plastics Inc., Mount Vernon, OH. The unit wasfitted with additional mounts and fittings at theUniversity of Toledo to enable capturing thesunrays at different angles during different timesof the day. A reactor support consisting of astainless steel tube of 1.25 in diameter was secured

centrally at a height of about 36 in (in the vicinityof the focal point of the concentrator) abovethe dish. A quartz tube with 1 in outer diameterwas used as the reactor in which the charge ofnZVI could be loaded for MSR. The setupis shown in Figure 13. Real-time temperature

Figure 13. Prototype solar concentrator for hydrogen generation via MSR reaction.

1230

1240

1250

1260

1270

1280

1290

1300

1310

1320

12:00 12:28 12:57 13:26 13:55 14:24 14:52

Time, (hr:min)

Tem

p,

°C

Figure 14. Temperature at the center of the reactor vs time study.



DOI: 10.1002/er

measurements were carried out with this setupon different days and different times of the dayand the results are summarized in Figures 14–16.As can be seen, the maximum temperatureattained at the center of the concentrator, in theperiod from noon to mid afternoon was as high as13101C (Figure 14). Figure 15 shows the depen-dence of the temperature attainable with theangular movement of the dish to a location so asto optimize the maximum solar capture. Here, the

angle, y is defined as the angle made by theconcentrator with the horizontal floor. When theconcentrator was kept at a fixed angle formaximum temperature at the center of thecollector and left undisturbed, the temperaturedropped drastically to the ambient in less than30min (Figure 16). The reason is that the opticschanges and requires a continuous change inangular position of the concentrator with respectto sun’s position.

0

5

10

15

20

25

30

35

40

45

50

12:00 13:12 14:24 15:36 16:48

Time, (hr:min)

An

gle

, θ

Figure 15. Angle of the concentrator with the horizontal floor vs time study.

0

200

400

600

800

1000

1200

1400

14131211

Time, (hr:min)

Tem

p,

°C

Figure 16. Temperature vs time study, at a fixed angle of the concentrator.



DOI: 10.1002/er

As demonstrated in the previous section, MSRcan be effectively carried out at 6001C and atemperature as high as 13001C is not required.Therefore, studies were made to attain a nominaltemperature of around 6001C by reducing theeffective ‘exposed’ diameter of the dish by maskingthe periphery with black paper. It was found thatwith an effective circular opening of about 30 inand 18 in, a temperature of about 1157 and 8201C,respectively, was attained. Therefore an 18 in (orsmaller) concentrator would be adequate for theMSR reaction.

A design for the actual reforming reaction usingthe solar concentrator as the heat source has beenarrived at and built. A tightly coiled 1/16 indiameter stainless steel tube is fitted into thequartz tube with rubber stoppers at the ends.Water pumped through the coil is preheated bycapturing the solar heat in the cooler section of the

reactor and generates steam. Helium is used as theblanket gas to avoid the oxidation of nZVI locatedin the zone of concentrated solar heat prior to theMSR reaction. The product (humidified hydrogen)is led either into a GC for product identification orinto a 3W 4-cell PEM fuel cell stack. Figure 17shows the schematic of the proposed solar-powered MSR interface.

The results of the hydrogen generation studiesusing this solar concentrator-nZVI interface willbe reported in a future communication.

It is important to note that although hydrazineis inexpensive compared with sodium borohydride,the current commercial generation of hydrazinerequires ammonia as raw material hence indirectlyinvolves carbon emissions. Therefore, searchfor a zero-carbon cycle, non-hydrogen bearingreductant to make zerovalent nanoscale ironparticles from mill-scale waste, is warranted.

Figure 17. Schematic of the proposed solar-powered MSR setup.



DOI: 10.1002/er

4. CONCLUSIONS

Metal-steam reforming using iron is a safe andeconomically viable method of hydrogen generationfrom inexpensive raw materials for ready PEMFCutilization. A novel solution-based technique usingsodium borohydride as the reductant yieldednanoscale iron with an average particle size of�40nm. In another case, by using hydrazine as analternate reductant and solvothermal conditions(1001C/5 atm), body centered cubic iron particles�5nm in size were obtained. This is a significantachievement in that hydrazine is more stable andmuch less expensive reductant compared withsodium borohydride and the solvothermal processis easily scalable. The hydrogen production viametal-steam reforming reaction represented byreaction (1) shows that the nanoscale iron possessesa higher propensity of reaction with steam than thatderived via either of the high-temperature reductionprocesses adopted here, under identical experimen-tal conditions. Thus, large particle size, reducedsurface area and agglomeration become non-issue inthe case of solution route of generating elementaliron of higher catalytic activity. Hydrogen gene-ration via metal-steam reforming using nanoscaleiron derived from mill-scale waste would become abenign proposition both in its energy intensivenessand the over all economics of the process. Toinitiate and sustain the somewhat endothermicMSR process, a solar concentrator consisting of aconvex polyacrylic sheet with aluminum reflectivecoating was fabricated. The correlation of theangular dependence of the temperature profile withtime and diameter of the exposed surface wasestablished. Thus, a unique combination of mill-scale waste as iron source, hydrazine as thereductant, mild process conditions for nZVI gen-eration and solar energy as the impetus foractuating MSR obviates several drawbacks pla-guing the grand scheme of producing and deliveringpure and humidified H2 to a PEMFC stack.

ACKNOWLEDGEMENTS

The financial support provided by the Department ofEnergy (DOE) and Edison Materials Technology Center

(EMTEC) for this work is gratefully acknowledged. Thanksare also due to Drs. John Mansfield and Kai Sun for accessto the scanning and transmission electron microscopes atthe EMAL Center of University of Michigan, Ann Arbor,to Mark Schuetz of Replex Plastics for providing thereflective solar bowl and to Robert Dunmyer for hisvaluable assistance in the assembly, testing and interfacingof the solar concentrator with the MSR setup.

REFERENCES

1. Turner J, Sverdrup G, Mann MK, Maness P, Kroposki B,Ghirardi M, Evans RJ, Blake D. Renewable hydrogenproduction. International Journal of Energy Research 2008;32:379–407.

2. Sørensen BA. Sustainable energy future: constructionof demand and renewable energy supply scenarios.International Journal of Energy Research 2008; 32:436–470.

3. Otsuka K, Mito A, Takenaka S, Yamanaka I. Productionof hydrogen from methane without CO2-emissionmediated by indium oxide and iron oxide. InternationalJournal of Hydrogen Energy 2001; 26:191–194.

4. Otsuka K, Yamada C, Kaburagi T, Takenaka S.Hydrogen storage and production by redox of ironoxide for polymer electrolyte fuel cell vehicles.International Journal of Hydrogen Energy 2002; 28:335–342.

5. Otsuka K, Kaburagi T, Yamada C, Takenaka S. Chemicalstorage of hydrogen by modified iron oxides. Journal ofPower Sources 2003; 122:111–121.

6. Takenaka S, Kaburagi T, Yamada C, nomura K,Otsuka K. Storage and supply of hydrogen by means ofthe redox of the iron oxides modified with Mo and Rhspecies. Journal of Catalysis 2004; 228:66–74.

7. Hacker V. A novel process for stationary hydrogenproduction: the reformer sponge iron cycle (RESC).Journal of Power Sources 2003; 118:311–314.

8. Santos DMD, Maurao MD. High-temperature reductionof iron oxides by solid carbon or carbon dissolved in liquidiron–carbon alloy. Scandinavian Journal of Metallurgy2004; 33:229–235.

9. Azad AM, Kesavan SK. Generation of high purityhydrogen from steel industry waste. In Proceedings of the2nd International Green Energy Conference (IGEC-2),Dincer I, Li X (eds), Oshawa, Ontario, Canada, June25–29, 2006; 1–9.

10. Azad AM, Kesavan SK. Enviro-friendly hydrogen genera-tion from steel mill-scale via metal-steam reforming.Bulletin of Science, Technology and Society 2006;26:305–313.

11. Schlesinger HI, Brown HC, Finholt AE, Gilbreath JR,Hoekstra HR, Hyde EK. Sodium borohydride, its hydro-lysis and its use as a reducing agent and in the generation ofhydrogen. Journal of the American Chemical Society 1953;75:215–219.

12. Forster GD, Barquin LF, Pankhurst QA, Parkin IP.Chemical reduction synthesis of fine particle FeZrB alloysunder aerobic and anaerobic conditions. Journal of Non-Crystalline Solids 1999; 244:44–54.



DOI: 10.1002/er

13. Shen J, Li Z, Yan Q, Chen Y. Reactions of bivalentmetal ions with borohydride in aqueous solutionfor the preparation of ultrafine amorphous alloyparticles. Journal of Physical Chemistry 1993; 97:8504–8511.

14. Xiaomin N, Xiaobo S, Huagui Z, Dongben Z, Danban Y,Qingbiao Z. Studies on the one-step preparation of ironnanoparticles in solution. Journal of Crystal Growth 2005;275:548–553.

15. Li Z, Han C, Shen J. Reduction of Ni21 by hydrazine insolution for the preparation of nickel nano-particles.Journal of Materials Science 2006; 41:3473–3480.

16. Cullity BD. Elements of X-Ray Diffraction (2nd edn).Addison-Wesley: Reading, 1978.

17. Shen J, Zhang Q, Li Z, Chen Y. Chemical reaction for thepreparation of Ni–P ultrafine amorphous alloy particlesfrom aqueous solution. Journal of Materials Science Letters1996; 15:715–717.



DOI: 10.1002/er

Documents

A closed-loop proposal for hydrogen generation using steel waste