SUNLIGHT-DRIVEN MEMBRANE-SUPPORTED PHOTOELECTROCHEMICAL WATER SPLITTING

Size: px
Start display at page:

Download "SUNLIGHT-DRIVEN MEMBRANE-SUPPORTED PHOTOELECTROCHEMICAL WATER SPLITTING"

Transcription

1 SUNLIGHT-DRIVEN MEMBRANE-SUPPORTED PHOTOELECTROCHEMICAL WATER SPLITTING Nathan S. Lewis, Harry A. Atwater Harry B. Gray, Bruce S. Brunschwig Jim Maiolo, Kate Plass, Josh Spurgeon, Brendan Kayes, Mike Filler, Mike Kelzenberg, Morgan Putnam California Institute of Technology Pasadena, CA 91125

2 Energy Conversion Strategies Fuel Light Electricity CO 2 Sugar Fuels SC Electricity e H O 2 O H 2 2 sc SC O 2 Photosynthesis H 2 O Semiconductor/Liquid Junctions Photovoltaics

3 Fuel from Sunlight QuickTime and a YUV420 codec decompressor are needed to see this picture.

4 Photoelectrochemical Cell e - O 2 e - e - H 2 SrTiO 3 metal H 2 O h + H 2 O Solid Liquid Light is Converted to Electrical+Chemical Energy

5 Combinatorial Discovery of Nanorod Materials FTO slide printed with binary combinations of eight metals, pyrolyzed to form mixed metal oxides. Typical photocurrent map of a slide of metal oxides. The X- and Z-coordinates give each spot s identity, which is correlated with photo-catalytic activity. Here, combinations of Zn and Co in row #9 show photoanodic current

6 Lessons from Photosynthesis

7 Turner Cell

8 System Concept H 2 O 2H 2 O 2 catalyst anode Solar PV & membrane H 2 catalyst cathode O 2 4H +

9 Why Radial Junction Solar Cells? hν 1/α n-type p-type ~L n Traditional single junction solar cell Idealized radial junction nanorod solar cell ~L n Traditional device design requires long minority carrier diffusion lengths, and therefore expensive materials Radial geometry allows for high quantum efficiency despite short minority carrier diffusion lengths

10 Device Modeling Comparison of photovoltaic efficiency for (a) planar and (b) nanowire Si devices as a function of absorber thickness and minority carrier diffusion length Relatively high efficiencies possible despite low diffusion lengths if depletion region recombination can be minimized B.M. Kayes, H. A. Atwater, and N. S. Lewis, J. Appl. Phys (2005)

11 Structural Organization in Nature Stomata Leaf Root Forest

12 Enables New Materials High efficiencies possible despite low diffusion lengths => Inexpensive materials Cheaper traditional materials Silicon as a model system Non-traditional, earth-abundant materials Tandem partners with silicon

13 Relaxes Catalyst Activity Requirements Hydrogen Evolution Reaction

14 Why Macroporous Silicon? Idealized radial junction nanorod solar cell Idealized macroporous silicon model sample Macroporous silicon can be made from single-crystalline (100) silicon starting material. Bulk-limited V OC can be calculated from the Shockley diode equation: V OC ln = kt q J J ph 0 J 0 = qd L p p N n 2 i D

15 Making Macropores 5% HF + 10 mm SDS

16 Macroporous Silicon

17 I-V Measurements

18 I-V Curves 100 mw/cm 2 QuickTime and a TIFF (Uncompressed) decompressor are needed to see this picture.

19 Testing a Nanorod Array Solar Cell Characterize a working nanorod solar cell Ready fabrication of nanorod electrodes to put into a working cell CdSe, CdTe While not an ideal commercial material (toxicity, abundance), CdSe x Te 1-x is diffusion length limited has ideal bandgap (~1.4 ev) allows for an early test of the theory

20 The Alumina Template Allows for dense, uniform array of rods Carefully controlled dimensions Easily removable template

21 The Alumina Template Allows for dense, uniform array of rods Carefully controlled dimensions Easily removable template

22 CdSeTe Nanorods Porous Alumina Alumina template removed in 1 M NaOH CdSeTe Nanorods (via electrodeposition) CdSe layer to prevent shunting (via sputtering) Titanium Back Contact (via sputtering)

23 CdSeTe Nanorods Excellent control over the dimensions and composition of these rods is possible

24 CdSeTe Nanorods Very uniform over relatively large area (a few cm 2 )

25 CdSeTe Nanorods Use a liquid junction to make a photoelectrochemical cell Measure current-voltage properties in the dark and light and compare nanorod performance to an analogous planar electrode Thin sputtered CdSe layer to prevent shunting through back contact negligible contribution to performance

26 Why Radial Junction Solar Cells? hν 1/α n-type p-type ~L n Traditional single junction solar cell Idealized radial junction nanorod solar cell ~L n Traditional device design requires long minority carrier diffusion lengths, and therefore expensive materials Radial geometry allows for high quantum efficiency despite short minority carrier diffusion lengths

27 CdSeTe Nanorods Spectral response: 0.12 Planar Nanorod 0.1 ld ie Y 0.08 m tu n a u 0.06 l Q a rn te x E Wavelength (nm) As expected from IV curves, planar usually exhibits better quantum yields the above curves were chosen because their similar magnitudes allow for ease of shape comparison Shape of the curve is more significant Nanorod samples lose less quantum yield in the red Since longer wavelength light penetrates deeper into the semiconductor, this is an indication that the nanorod geometry is more effectively collecting the carriers

28 Si Nanorods Nanorod Array Growth Procedure Alumina template removed in 1 M NaOH Mounting Wax Porous Alumina Silicon Nanorods Conductive Catalyst Metal (via evaporation) Supportive Nickel Substrate (via electrodeposition)

29 Si Nanorods Silicon nanorod array after template is removed

30 Spatially resolved photocurrent λ=488 nm 5 μm Al/Ag 5 μm 10 μm (1.18 μm diameter) NSOM schematic SEM-NSOM Topography Overlay 4-Point Probe Near-field scanning optical microscopy (NSOM) can measure nanowire photoconductivity as a function of excitation location. Carrier collection lengths can be studied as a function of nanowire diameter, growth conditions, passivation, etc... R wire = 17.9 MΩ R C = 200 kω N D ~ cm -3

31 Photocurrent (norm.) Al 200 Photocurrent (pa) mv bias 200 mv bias Effective diffusion length 10 μm -1 0 µm 10 µm Wire position Energy band diagram Increasing energy Charge transport by drift and diffusion: J ( x) = qμ n( x) E( x) + qd δ n n n n dx Charge continuity for electrons with effective, single-τ recombination rate: δn t = 1 J n q x Monitor terminal current as function of injection point: J = J + J meas n x= 0 p x= 0 L n,eff = D τ = 1.9 µm τ n,eff 1 ns n n, eff L p,eff = D τ = 2.2 µm p δn τ p, eff n, eff + when np, x= x = Iph Vary τ to match measured photocurrent profile: τ p,eff 4 ns NSOM suggests minority carrier collection length is comparable to wire diameter. g n g inj

32 Surface Requirements for Semiconductor Heterojunctions hν ~L Traditional planar photovoltaic Idealized high aspect ratio photovoltaic Low Trap State Density Stable in Ambient Conditions Facile Interfacial Charge Transfer

33 Oxide Buffer Layer - Pattern Fidelity No oxide buffer layer 10 μm 10 μm 100 μm H 2 anneal 1000 o C 100 μm Oxide buffer layer 10 μm H 2 anneal 1000 o C 50 μm An oxide buffer layer is critical for maintaining pattern fidelity during growth. 3 μm array, 500 nm Au, T growth = 1000 o C, P growth = 760 Torr

34 Large Area Au-Catalyzed Si Arrays 100 μm 3 μm array, 500 nm Au, T growth = 1000 o C, P growth = 760 Torr, 30 min growth, 2 mole % SiCl 4 in H 2

35 Large Area Au-Catalyzed Si Arrays 30 μm 3 μm array, 500 nm Au, T growth = 1000 o C, P growth = 760 Torr, 30 min growth, 2 mole % SiCl 4 in H 2 Nearly 100% vertically aligned, 75 μm length microwire arrays over areas > 1 cm 2.

36 Copper-Catalyzed Si Wire Arrays 15 μm 250 μm 3 μm array, 500 nm Cu, T growth = 1000 o C, P growth = 760 Torr, 10 min growth, 2 mole % SiCl 4 in H 2 Copper produces wire arrays that are structurally equivalent to gold.

37 Si Wire Array Junctions QuickTime and a TIFF (Uncompressed) decompressor are needed to see this picture.

38 Surface Requirements for Semiconductor Heterojunctions hν ~L Traditional planar photovoltaic Idealized high aspect ratio photovoltaic Low Trap State Density Stable in Ambient Conditions Facile Interfacial Charge Transfer

39 Surface Passivation H Si PCl 5 benzoyl peroxide Cl Si CH 3 MgCl CH 3 Si or Cl 2(g) Measured through STM, T = 4 K 0.38 nm 1 nm Si(111)-CH 3

40 SEM of Chlorinated Si Nanorod Arrays Chlorination with PCl5 is known to cause etching of silicon surfaces Chlorinated 10 min, then methylated Chlorinated 45 min 2 μm 1 μm Membrane-Supported Nanorod PEC H2 Functionalized surfaces appear rough, but etching is not destructive

41 Electrical Performance of Methylated Silicon Nanowires Methylated wires (60 nm diameter) Sensitive to gate voltage Maintain higher conductance Hydrogen terminated surfaces Deteriorate Normalized average conductance Average intrinsic conductance Gate voltage (V) Time (h) Haick, H.; Hurley, P. T.; Hochbaum, A. I.; Yang, P.; Lewis, N. S. J. Am. Chem. Soc. 2006, 128,

42 Constructing the Pieces of a Solar H 2 Fuel Generator

43 Polymer Embedding of Si Rod Arrays * O Si * PDMS (polydimethylsiloxane)

44 Large Area Rod Array Removal Top-down view Side view 115 μm Large area arrays (> 1 cm 2 ) transferred in one piece. Conformal coating from top to bottom of rods

45 Embedded Rod Pattern Fidelity Inter-rod angle and spacing is constant before and after PDMS casting and removal Apparent rod diameter differences due to examination of rod tips versus base θ Before removal After removal Rod diameter (μm) Shortest inter-rod distance (μm) Angle (θ, ) 2.5 ± ± 2 92 ± ± ± 3 94 ± 2

46 Large Area Rod Array Removal Top-down view Side view 115 μm Large area arrays (> 1 cm 2 ) transferred in one piece. Conformal coating from top to bottom of rods

47 Regrown Arrays

48 Polymerization from Si Rod Surfaces Allylated Si base Methylated Si base polymer sheath grown around rods H H H CH 2 Si 1 st generation Grubbs catalyst H Ru* CH 2 Si CH 2 Si Ru* n

49 Dual Junction Nanorod Arrays 1.23 V needed to electrolyze water Requires a heterojunction or dual junction Single absorber: band gap of ev that straddles the necessary potentials Photoanode and photocathode absorbers: Band gaps can better match the solar spectrum ( ev) design Allows for greater flexibility in materials Efficiency could approach 25% Optimal structure would have several lightabsorbing layers absorbing different portions of the spectrum with an overall potential greater than 1.23 V

50 Fe 2 O 3 Nanowires

51 System Concept H 2 O 2H 2 O 2 catalyst anode Solar PV & membrane H 2 catalyst cathode O 2 4H +

52 Conclusions Without massive quantities (10-20 TW by 2050) of clean energy, CO 2 levels will continue to rise The only sufficient supply-side cards we have are clean coal, nuclear fission (with a closed fuel cycle), and/or cheap solar fuel We need to pursue globally scalable systems that can efficiently and costeffectively capture, convert, and store sunlight in the form of chemical fuels He that can not store, will not have power after four Semiconductor/liquid junctions offer the only proven method for achieving this goal, but we have a great deal of fundamental science to learn to enable the underpinnings of a cost-effective, deployable technology Nanorods, randomly ordered junctions to generate the needed potential Catalysts to convert the incipient electrons into fuels by rearranging the chemical bonds of water (and CO 2 ) into O 2 and a reduced fuel

53 CdS e S 2 2- M New Redox Couples Nonaqueous Solvents Si e A M S 2- EtOH D Four Strategies e - Load e - e e - e - A hυ I - Si H 2 O D M Surface Modification Inexpensive Semiconductor I 3 - ACN