solar fuel

8/13/2019 Solar Fuel

1/7

19E N G I N E E R I N G & S C I E N C E N O . 2

P L E A S E C O N T I N U E TO T H E N E X T P AGE . . .


2/7

20 E N G I N E E R I N G & S C I E N C E N O . 2

He who cannot store will have no power afterfour is a favorite mantra of Nate Lewis (BS 77,MS 77), Argyros Professor and professor of chem-istry, in a nod to the late Johnnie Cochran Jr. Butthe stakes here are even higher than the outcome ofO. J. Simpsons trial for double murderLewis istalking about saving our civilization by running iton solar energy. Solar cells make electricity, and ifyou have a secret method for storing enough juiceto keep the lights burning across Americaor eventhe Los Angeles basinwhile the sun is down,he advises you to go out and buy electricity atnight for a nickel a kilowatt-hour, sell it to all yourneighbors for 25 cents a kilowatt-hour by day, andthen fund your own solar-energy research institute

with the profits.e most promising avenue is to store solar

energy in chemical form, as a fuel that can be usedon demand. Plants do this via photosynthesis, turn-ing sunlight, water, and carbon dioxide into sugarmolecules containing lots of high-energy chemicalbonds. Lewis and many other researchers world-

wide are trying to take a leaf from photosynthesis,as it were, by finding a method for turning sunlightinto fuelwithout all the complex cellular machin-ery. Photosynthesis is beautiful, and it has manylessons to teach us, Lewis says. Birds with feathersare not good models for 747s, but we knew birdscould fly, so we were inspired to build airplanes.

e easiest fuel to make is hydrogen gas, H2.

Simply stick a couple of electrodes in a vat of water

and run the current through themthe positiveelectrode makes O2, and the negative one makes

H2. Burning the H

2in a fuel cell later gives you

water back again, plus the stored electricity. Small,pilot-scale electrolyzer plants, as water-splittingfacilities are called, already exist, but they use a lotof really expensive components, and theyre notexactly the sort of thing you could put on the roofof your house, or even in your back yard. Lewissgoal is to repackage tons of complicated equipmentinto tiny assemblages of cheap materials that could

Solar Fuel I : Rods and StonesBy Douglas L. Smith


3/7


be installed almost anywhere. Were going to makefuel from the sun with no wires, he says. To do this,the light-harvesting and water-splitting machineshave to be one and the same, not an assortment ofcomponents spread out over a solar farm.

THEHOLESTORY

A conventional solar cell uses a cheese-and-crack-er arrangement of two slabs of silicon, n-type andp-type, and the place where the cheddar meets theTriscuit is called thep-njunction. N-type siliconis doped with traces of phosphorus or arsenic,atoms of each of which have five electrons in theirouter shell instead of silicons four. Four of the dop-ants electrons pair up with electrons in adjoiningsilicon atoms to make the chemical bonds that holdthe slab together, leaving the fifth one at liberty, ifgiven a little shove, to wander through the crystal.Similarly, inp-type silicon, the dopantboron orgalliumhas only three electrons in its outer shell.e dopant atom makes three and a half bonds,if you will, to its silicon neighbors, and the absentelectron, or hole, is essentially a free-floating posi-tive charge. An electron from an adjoining atom

will readily jump over to fill it, moving the hole tothe atom the electron abandoned.

When light hits the solar cell, the photons willpenetrate some distance into the silicon beforebeing absorbed, and the longer the wavelength (orthe redder the light), the deeper they go. It takes300 microns, or three-tenths of a millimeter, ofsilicon to soak up all the wavelengths that make upvisible light, and herein lies the problem. Whena silicon atom in either layer absorbs a photon, anelectron gets excited, creating an electron-hole pair.

If the silicon atom is in the n-layer, the electron canhappily skate away to become part of the current,but then what happens to the hole? In order tomaintain the charge balance and keep the currentflowing, the hole has to make its way to the safehaven of thep-type layer, on the opposite side ofthep-njunction, running a gantlet of onrushingelectrons in the process. (Because they are out oftheir element, if you will, these much-set-uponholes are called minority carriers.) Any defect inthe crystals lattice structure makes a natural trap

where electrons and holes can recombine, dissipat-

ing their energy as wasted heat. Free electronsgenerated in thep-type layer face a similar struggle.e upshot is you need ultrapureand thereforevery expensivesilicon to make solar cells.

GROWYOUROWNASPENFOREST

But what if, instead of using a flat plate ofsilicon, you made nanorodsteeny, tiny cylindersoriented so that the light shines down their centralaxes? e minority carriers would just have toreach the cylinders skin to make it to safety, andthe odds of an electron falling into a hole dropdrastically. is is exactly what nature does,remarks Lewis. is is what leaves do. e cellsabsorb light over the long axis and transport theheavier particlesthe water vapor and the nutri-entsover the short axis. ats what the retinalrod cells in your eyes do. is principle is foundthroughout nature, and yet its never been imple-mented in solar-cell design.

Even better, you dont need a layer of the oppo-site type of siliconsimply toss the rod in a beakerof salt water, which conducts electricity quite read-ily, and the liquid-solid interface acts as the finishline for the minority carriers sprint to safety. Oncethey cross into the liquid, they are ready to split

water molecules into H2and O

2. All you have to

do is coat the silicon surface with catalysts that willfacilitate the reactions.

Harry Atwater, Hughes Professor and professorof applied physics and materials science, and Lewistested the notion by growing what Lewis calls anaspen forest of nanorods, each some three micronslong and about a quarter of a micron thick, on aconductive backing. e complex growth process

used standard technology for making silicon chips,but for reasons that will become clear momentarily,Josh Spurgeon (MS 06), a Lewis grad studentworking on this joint project, made the first nano-forest of cadmium selenium telluride instead.is material generates electricity by absorbingblue light, which takes about one microns worthof semiconductor. But such a photocell doesnt

work under red light, which takes three micronsto absorb, because the minority carriers can onlymake it about a micron before being sopped upby a lattice defect. Spurgeon shone light in all thecolors of the rainbow on a submerged nanorodarray and on a CdSeTe slab made by the same

method and compared the results. e slab andthe rods both performed about the same underblue light, but the nanorods kept on soaking upphotons all the way through the visible spectrum.But we didnt want to wait around for the rods togrow, says Lewis. His grad student James Maiolotreated a silicon slab with hydrogen fluoride, etch-ing so many holes in it that they overlapped toleave spires. We wanted to make the equivalent ofrods to see if they would work, or if the theory wasmissing something. It wasnt.

Light will travel a certain

distance into a crystal

before being completely

absorbed. When a photon

is absorbed by a solar cell,

a pair of opposite charges

in the form of an electron

(e-) and a hole (h+) are

produced. If either one

has to travel too far

through unfriendly terri-

tory, as the hole does in

this traditional slab design

(below left), it is likely

to recombine with its

opposite and no current

will flow. In the nanorod

design (below center and,

blown up, below right),

the light can penetrate to

any depth and the holes

can still escape out the

sides and into a conductive

liquid.


4/7


NOTSOFAST. . .

If making hydrogen from sunlight were thatsimple, however, you wouldnt be reading thisarticle. ere are still three big problems to solve.e first is really only a problem when you try, asLewis and Atwater are, to make fuel from the sun

without using wires: H2and O

2production need to

be kept separate. is makes collecting the hydro-gen easier, and prevents the gases from recombiningto make water as they form. All fuel-producingsolar cells to date isolate the oxygen-producing elec-trode from the hydrogen-producing electrode witheither distance or a membrane. e simplest designuses a single semiconductor to absorb the light. Inthis case, its an n-type semiconductor that acts asthe oxygen-producing electrode. e holes leapinto the drink and suck electrons out of the watermolecules in a process called oxidation, creatingO

2gas and protonsthe electron-deprived nuclei

of hydrogen atoms. e electrons go through awire to the hydrogen-producing electrode, usuallyplatinum metal or the like, where they rejoin thenewly produced protons that have diffused throughthe water (or across the membrane) to make H

2a

process called reduction.e second problem with splitting water is

fundamental chemistry: oxidation is brutal. Strongoxidants tear chemical bonds apartfire is an oxi-dative process, and think of what would happen toyour favorite shirt if you spilled concentrated bleachon it. Since the goal is to get rid of the wireand,incidentally, the very expensive precious-metal elec-trode to which its attachedthe hydrogen-produc-ing electrode needs to be ap-type semiconductorfused seamlessly with the n-type to inject electronsinto the water. Unfortunately, knownp-type semi-

conductors dont last long when exposed to strongoxidants. is includes elemental silicon itself, saysLewis. Silicon is not stable under oxidizing condi-tions in water. It makes sandSiO

2.

e third problem we will come to shortly.

FOOLSGOLDANDRUST

Proving that the nanorod idea worked allowedLewiss and Atwaters groups to consider usingnontraditional materials to collect light. Forinstance, says Lewis, iron oxide has been rejectedin solar cells because its very defective as found innature. You need three microns of rust to absorbsunlight, and the carriers can only move 20 nano-meters, so the other 2.99 microns are just wasted.But if we make rods 20 nanometers wide andthree microns long, we can still move the carrierssideways. e same thing occurs with fools gold,

which is iron pyrite, FeS2. So we are now trying to

make solar cells out of fools gold and rust, and youcan decide whether were completely crazy or onlypartially crazy.

e good news is that theres a whole class ofdefect-riddled minerals that can absorb light tomake electricity while withstanding the oxidativeassault. Some of them, such as titanium dioxide,or TiO

2, the coloring agent in white paint, are also

in abundant supply and available at rock-bottompricestwo absolutely fundamental considerationsif the technology is to be adopted on a global scale.ey are very stable materials. ese literally arerocks; theyre oxidatively inert, theyre photochemi-cally inert, says Lewis. e bad news is that they

would work better if we were inhabiting the mooninstead of the earth, because they have a band gap

where they start to absorb light that is about threeelectron volts or higher. e band gap is theenergy difference between the state that an electronis in when it is firmly attached to its atomthe so-called valence bandand the state it gets kickedinto once it absorbs a photon and gains enoughenergy to hop from atom to atomthe conduc-

tion band.Unfortunately, three electron volts is borderlineultraviolet. Earths ozone layer screens out mostsunlight at higher than that energy, so the effi-ciency of those materials is limited to a few percentat best. e light that actually makes it down tous starts at about 2.75 electron volts on the purpleend, hits peak intensity at about 2.35 electron voltsin the blue-green region, and then slowly trailsaway down through the infrared. e optimalband gap for catching everything with a singlematerial is about 1.4 electron volts, which is actu-ally in the near-infrared. It only takes 1.23 volts tosplit water, but theres a catchand thats the third

problem.e band gap not only has to be big enough, ithas to be in the right spot, energetically speaking.For example, iron oxides band gap is 2.2 electronvolts, but the electrons thus liberated arent veryenergetic. From an electrochemical point of view,making H

2from water and making O

2from water

are two entirely separate reactions; worse, theyoperate at different energy levels. If you use aniron-oxide photon absorber, you can still makeoxygen, but you cant make hydrogen any more,

In this water-split-

ting solar cell, the light

absorber is the oxygen-

making electrode, an

n-type semiconductor. The

surrounding water acts as

the p-type layer, accepting

the holes, which convert

water molecules into O2

and H+.

N-type

semicon

ductor

Metal


5/7


Lewis says. Absolute energies count. e electronsthat are promoted to the conduction band arentreducing enough to reduce water to make hydro-gen. Its like climbing a wallgoing three feet updoesnt put you a foot over the top if you start outfour feet down.

ALLTHEOXIDESTHATAREFITTOPRINT

Just because no known oxidatively stable mate-rial has a big enough band gap at the right energylevel doesnt mean that such a substance doesntexist. e quest for something that would fill thebill was taken up in 2006 by Rhodes Scholar ToddGingrich (BS 08) and grad student Jordan Katz(PhD 08), as detailed in Caltech News2008, No.

1. e idea was to test tiny samples of thousandsof candidates as fast as possible by modifying a$140 inkjet printer to spit swatches of metal saltsdissolved in water onto thin glass plates. e printhead mixes the salts in exacting proportions, justas it would faithfully render a sunset from cyan,magenta, yellow, and black inks, but the outputlooks more like a color-calibration page than avacation photoup to 250 tiny test squares ona single sheet. A 500C oven bakes the dissolvedmetals into metal oxides. We choose our metaloxides in an Edisonian educated-guess fashion,Lewis says. We see which ones give us the rightenergy levels to reduce water to hydrogen, and then

we investigate which ones stay around after oneday of operation. Were building up a database totry to give theorists a guide to help us find the nextgeneration of materials.

Lewiss lab is sticking to metal oxides because,well, theyre already as oxidized as they can get.e chemists intuition is that whatever oxidizes

water to oxygen will be sufficiently oxidizing thatif we dont have an inherently stable, nearly ionicchemical bond such as a metal oxide, were goingto eat our semiconductor alive. is tactic has

uncovered a couple of promising materials already,but Lewis is not betting the solar farm on oneprinterhes also working on a Plan B.

ings would be greatly simplified if one mate-rial didnt have to do everything. Nature uses sucha divide-and-conquer strategya plants photosyn-thetic system uses two light-absorbing componentsthat together provide the oomph to drive the reac-tion. e photosynthetic machinery within such acell straddles a membrane that isolates and protectsthe hydrogen-producing apparatus from the harshoxidants on the other side. In a membrane-based,two-part inorganic system, the hydrogen-produc-ing side could be powered by metal sulfides. Metalsulfides have chemical compositions similar to theactive catalytic sites in the enzymes used by a fam-ily of photosynthetic microbes called green sulfurbacteria to eke out a living in low-light, oxygen-freeenvironments. ese hardy bugs have even beenfound in hydrothermal vents called black smokerson the deep ocean floor, where they subsist on thedim glow of the hot vent itself. e oxygen-pro-ducing side could use good old iron oxide, and ifthe two types of nanorods could somehow makeelectrical contact through the membrane, the sys-tem wouldnt need wires.

is gives us the ability to retain the chemi-cal stability of two separate materials, says Lewis,while not also asking one single material to deliverthe full energy needed to split water. e design

This plot shows the intensity of the sunlight reaching the

earths surface versus the amount of energy per photon in

electron volts.

Top: Katz (left) and Lewis

watch a glass plate come

out of the printer.

Right: A portion of a typi-

cal plate.


6/7


is actually the inverse of a fuel cell, where hydro-gen and oxygen come in, protons go across themembrane, and out comes electricity. Instead, inour cell, in comes the energy source, protons goacross the membrane, and out comes chemical fuel.

Were driving it in reverse.

BRINGINGINTHESHEAVESis all sounds horribly complicated, but the

Lewis lab has hit on a very clever way to make theintegrated system. e process can be done on anyscale, which is vital for any technology intendedto supplant fossil fuels, and if the work with foolsgoldan iron sulfideand rust is successful, theraw materials would be dirt cheap, which is even

more important. For the moment, the people put-ting the pieces together are using silicon nanorods,because the industrial techniques for making themare well established.

e process starts with an ultrapure silicon wafer,

but it can be reused over and over again, as gradstudent Spurgeon has shownsuccessive genera-tions of rods grown on the same wafer look just asgood as the first batch. e rod-growing technique

was developed by grad student Brendan Kayes (MS04), who is coadvised by Atwater and Lewis, along

with Atwater postdoc Mike Filler and coworkers.It begins with a pattern of tiny gold dots that canbe laid down on the wafer by any of various means,and which become the growth template. eirdiameter is the size of the nascent nanorods, and

the spacing between them is the spacing betweenthe rods in the array. e rods are grown through aprocess called chemical vapor deposition, in whichthe gold, now in little molten globules, reacts withgaseous silicon chloride (SiCl

4) in a hydrogen

atmosphere at a temperature of 1,000C. Whenthe solution becomes supersaturated with siliconatoms, the excess precipitates out at the bottom ofeach droplet, becoming one with the silicon waferbelow. e gold rides on top, as the silicon startsto deposit, and you continue that growth process,Lewis explains. It literally grows like cornstalksin a field. Turn off the gas flow, and you can stopthe growth at any height you like. Even better,says Lewis, We can do this with copper. We cando this with nickel. We dont need gold as thecatalyst.

However, theres a really good reason people havebeen making solar cells out of slabs instead of rods.If you picture the layer of atoms that forms the sur-face of a slab, a silicon atom sitting there can havethree of its four bondswhich point to the fourcorners of a tetrahedron centered on the siliconatomeasily connected to other silicon atoms inthe surface layer. e fourth bond sticks out intospace. ese dangling bonds are prime locationsfor charge carriers to recombine. A slab has limitedsurface area, while a nanorod is basically all surfaceareayou couldnt ask for a better way to maxi-mize the number of dangling bonds. So if wedont have a way of tying up those dangling bonds,then we dont have much hope of actually makinggood devices out of them, says Lewis.

e nanorods end the manufacturing processwith their dangling bonds capped with hydrogenatoms. Unfortunately, SiH bonds are oxida-tively unstable in air. We do a lot of fundamental

chemistry in our group, so we developed a methodusing phosphorus pentachloride or PCl5, and zit

medicine, benzoyl peroxide, to convert the silicon-hydrogen bonds to SiCl bonds, says Lewis.

Another set of steps replaces every chlorine atomwith a methyl group, CH

3. e silicon-carbon

bonds are sufficiently strong that they fool thosesurface silicon atoms into thinking that theyre

just like the bulk material. ere are no danglingbonds left at the surfaceless than one electricaldefect in every 100,000 surface atoms. is work

was done by then-postdocs Hossam Haick andPatrick Hurley, along with collaborators Peidong

Yang of UC Berkeley and his grad student Allon

Hochbaum.

SOLARPANELSYOUCANUNROLL

For the next step, postdoc Kate Plass has devel-oped a remarkably simple process for embeddingthe nanorods in a membrane. She uses a type ofsilicone rubber called polydimethyl siloxane, orPDMSa common waterproof sealant that, saysLewis, is more affectionately known to people on

A cornfield of silicon nano-

rods grown from seeds of

gold. The scale bar is 30

microns, or millionths of

a meter.

Some day some fellow will invent a way of concentrating and storing up

sunshine . . . When we learn how to store electricity, we will cease being apes

ourselves; until then we are tailless orangutans. Thomas Edison


7/7


the street as fish-tank goop. Plass carefully poursa thin layer of PDMS over the nanorods until shealmost, but not quite, covers their tips, leavingthem exposed to make contact with the water later.She gives the surface a close shave, literally, with arazor blade, and then she peels the goop off in onesmooth, sharp motion, like doing a hot-wax treat-ment to get rid of unwanted body hair. (Howlingand jumping around like Mel Gibson in WhatWomen Wantor Steve Carell in e 40-Year-OldVirginis optional.) Just as the hairs go with the

wax, the nanorods go with the goop, shearing offneatly at their bases and leaving the silicon waferready to take another round of metal dots. eliberated nanorods retain their original parallelorientation and regular spacing, only now they arein a piece of plastic that we can roll up, Lewis says.

Plass has so far made nanorod-embedded mem-brane squares more than a centimeter on a side.Scaling the process up should not pose any greatdifficulties, but it will take something more than

just building very large tweezers.Before being embedded, thep-type silicon nano-

rods in a demonstration system would be coatedwith cobalt, which acts as a catalyst for H

2produc-

tion. Postdoc Steve Maldonado is working on thatangle. ep-type membrane would be laminatedback-to-back with a second membrane containingn-type, oxygen-producing nanorods of iron oxidecoated with a cobalt-oxide catalyst; sandwichedbetween the two would be an electrically conduc-tive layer.

e idea is that eventually this technology mightbe developed to the point where the solar-mem-brane laminate could be sandwiched between twomore layers of plastic that would allow water tobathe the nanorods. e entire assembly would

then be manufactured in large rolls that would besold at home-improvement stores, the way rolls ofinsulation are sold today. (e main difference, ofcourse, would be that youd unroll this stuff on topof your roof, instead of in the attic underneath.)

With an eye toward this, the Lewis lab is startinga collaboration with Mory Gharib (PhD 83), theLiepmann Professor of Aeronautics and profes-sor of bioengineering, to measure the membranesmechanical properties to discover how far one canbe bent without fracturing the rods.

ere is one last hitch, however. Rememberthose naked protons that got stripped off the watermolecules and that have to cross the membrane

and reunite with their electrons to make H2?Fish-tank goop is impermeable, so a more porousmembrane will have to be found. Either that,says Lewis, or were going to have to poke lots oflittle tiny holes in it.

We havent yet built a full system, says Lewis.But we have a clear path toward being able, in twosheets of plastic, to at least demonstrate a proto-type system that could take water to hydrogen andoxygen with sunlight. As Melvin Calvin said at myfirst Department of Energy meeting in 1982, It

is time to build an actual artificial photosyntheticsystem, to learn what works and what doesnt workand thereby set the stage for making it better.Now, only 25 years later, it may finally be on thebrink of happening. A Caltech-MIT initiativecalled Powering the Planet, about which you willread more in the next article, has set itself the goalof creating a solar-fuel generator that uses Earth-abundant elements, needs no connecting wires, canbe scaled up with existing manufacturing technolo-gies, and is 10 times more efficient than photosyn-thesis.

Calvin wasnt the first to have this idea. Lewis isalso fond of quoting omas Edison, who in 1910told Elbert Hubbard, in Volume 1 of the lattersseries of books called Little Journeys to the Homes ofthe Great, Some day some fellow will invent a wayof concentrating and storing up sunshine to useinstead of this old, absurd Prometheus scheme offire. Ill do the trick myself if some one else doesntget at it. (Apparently, it never quite made the topof his to-do list.) Edison went on to say, Sunshineis spread out thin and so is electricity . . . the trick[is], you see, to concentrate the juice and liberate itas you needed it. . . . is scheme of combustionto get power makes me sick to think ofit is so

wasteful. . . . When we learn how to store electric-ity, we will cease being apes ourselves; until then weare tailless orangutans. You see, we should utilizenatural forces and thus get all of our power.

Adds Lewis, In the end, by far the biggestenergy source available to humans is the sun. Now

we cant afford to use it, but in the future we cantafford notto use it. Nature figured this out, butfrom a product point of view, photosynthesis isa failed solution; the fastest-growing plants, on ayearly basis, store less than 1 percent of the total

sunlight that hits an average acre over a year. Wecan do better, we have to do better, and we will dobetter. e question is not if, but when and how.Someday, youll paint your house with solar paint[thats another article for another dayDS], andinstead of putting up solar pool heaters, we will rollout solar fuel generators.

Nathan S. Lewis (BS 77, MS 77), CaltechsArgyros Professor and professor of chemistry, gotin on the ground floor of the solar-fuel game as anundergrad working with Harry Gray (see following

article). Lewis got his PhD at MIT in 1981, andearned tenure at Stanford before returning to Caltechin 1988. Hes been working on the photochemistry ofsemiconductor-liquid interfaces (including an effort todevelop solar paint) ever since, with time out for suchside projects as developing an electronic nose and help-ing to debunk cold fusion.

Every silicon atom has four

bonds, not all of which are

shown here. The dan-

gling bonds stick out of

the surface of the crystal,

and somethinghas to cap

them off.

A piece of nanorod-

embedded plastic.

PICTURE CREDITS: 20-23, 25 Doug Cummings; 23

Bob Paz; 24 Brendan Kayes; 25 Michael Filler
http://pr.caltech.edu/periodicals/EandS/articles/LXXI2/gray.pdfhttp://pr.caltech.edu/periodicals/EandS/articles/LXXI2/gray.pdfhttp://pr.caltech.edu/periodicals/EandS/articles/LXXI2/gray.pdf

solar fuel

Documents