1

Supporting Information

High Efficiency Solar-to-Hydrogen Conversion on a Monolithically

Integrated InGaN/GaN/Si Adaptive Tunnel Junction Photocathode

Shizhao Fan†, Bandar AlOtaibi

†, Steffi Y. Woo

‡, Yongjie Wang

†, Gianluigi A. Botton

‡ and

Zetian Mi†*

†Department of Electrical and Computer Engineering, McGill University, 3480 University Street,

Montreal, Quebec H3A 0E9, Canada

‡Department of Materials Science and Engineering, Canadian Centre for Electron Microscopy,

McMaster University, 1280 Main Street West, Hamilton, ON L8S 4M1, Canada

*: E-mail: zetian.mi@mcgill.ca; Phone: 1 514 398 7114

2

Supplementary Discussions and Methods

S1. On the conduction band offset between GaN and Si

The electron affinity of n-Si is 4.05 eV1. The electron affinity of n-GaN has been reported to be

in the range of 3.5 eV to 4.1 eV2-4

. Previous studies have also confirmed that the n-GaN/n-Si

heterointerface has a negligibly small energy barrier for electron transport5. In this study, both Si

and GaN are heavily n-type doped to facilitate electron transfer at the GaN/Si interface.

Therefore, under illumination, the collected photoexcited electrons of the n+-Si layer can

effectively inject into the GaN:Si nanowire segment at a small applied bias. The injected

electrons can drive proton reduction on the lateral surfaces of GaN nanowires or recombine with

holes injected from the p-InGaN nanowires in the tunnel junction.

S2. Schematics of the samples studied in this work

Figure S1. Schematic illustration of (a) the n+-p Si solar cell wafer, (b) the p-InGaN/tunnel junction/n-

GaN nanowires on n+-Si substrate, and (c) the p-InGaN/tunnel junction/n-GaN nanowires on n

+-p Si solar

cell substrate. The p-InGaN/tunnel junction/n-GaN nanowires grown on n+-Si substrate and n

+-p Si solar

cell substrate have identical structures. The p-InGaN nanowires (red region) consist of six InGaN

segments connected by ~15 nm p-type GaN/InGaN short period superlattices to minimize In phase

separation during the MBE growth process and also to facilitate hole transport along the vertical direction

of nanowires. The tunnel junction (shaded gray region) consists of n++-GaN (20 nm)/In0.4Ga0.6N (4 nm)/

p++-GaN (20 nm).

3

S3. Fabrication of n+-p Si solar cell substrates

n+-p Si solar cell wafers were fabricated using a standard thermal diffusion process. The front

and back side of a double side polished p-doped Si(100) wafers (WRS Materials, thickness: 256

– 306 µm; resistivity: 1 - 10 Ω·cm) was first covered with phosphorus and boron dopants by spin

coating, respectively. Subsequently, the sample was baked at 950 oC for 20 mins under a N2 flow

rate of 200 standard cubic center meter per minute (sccm) in a diffusion furnace. The thermal

diffusion process leads to the formation of an n+ emitter layer and a p

+ electron back reflection

layer on the front side and back side of the Si wafer, respectively. The sheet resistivity of the n+

emitter layer was in the range of 8 - 14 Ω/sq, corresponding to donor concentrations of ~5×1020

/cm3. For the p

+ back reflection layer, the sheet resistivity was in the range of 30 - 60 Ω/sq,

corresponding to acceptor concentrations of ~1.5×1020

/cm3. Such solar cell wafers were used for

the MBE growth of InGaN nanowire arrays. In addition, Si solar cell devices were realized by

depositing Ti/Au metal contact on top of the n+ emitter layer and Ni/Au metal contact on the p

+

backside, followed by an annealing at 550 oC for 2 mins.

S4. Molecular beam epitaxial growth

Catalyst-free InGaN/GaN nanowire arrays were grown on both n+-Si substrate and n

+-p Si solar

cell wafer by radio frequency plasma-assisted molecular beam epitaxy (MBE). The surface oxide

of Si was first removed using buffered hydrofluoric (HF) acid before loading into the MBE

chamber. The substrate was further degassed in situ at ~ 800 oC before growth initiation.

Subsequently, InGaN/GaN nanowire structures were grown under nitrogen rich conditions with

the following growth parameters: a nitrogen flow rate of 1.0 sccm, forward plasma power of 350

W, and Ga and In flux in the ranges of 4.5 × 10-8

to 8 × 10-8

Torr and 4 × 10-8

to 8 × 10-8

Torr,

4

respectively. The substrate temperature was varied in the range of 650 to 780 oC. Si and Mg were

used as the n- and p-type dopants, respectively. The tunnel junction consists of 20 nm n++-GaN,

4 nm InGaN, and 20 nm p++-GaN, which were grown at slightly lower substrate temperatures ~

650 oC to enhance In incorporation.

S5. Optical characterization

Photoluminescence measurement was carried out using a 325 nm Cd-He laser as the excitation

source. The photoluminescence emission was spectrally resolved by a high resolution

spectrometer and detected using a photomultiplier tube.

S6. Structural characterization

The SEM images were taken using an Inspect F-50 FE-SEM system. The STEM was performed

using a double aberration-corrected FEI Titan3 80-300 STEM operated at 200 kV. Atomic-

resolution, atomic-number sensitive (Z-contrast) STEM-high-angle annular dark-field (HAADF)

images were obtained using a detector angular range of 63.8 – 200 mrad. Elemental mapping by

electron energy-loss spectroscopy (EELS) in STEM mode was done with the N K, In M4,5, Ga

L2,3, and Pt M4,5-edges and the spectrum imaging technique. Weighted-principal component

analysis (PCA) was applied to the spectrum images for noise-reduction of the Ga-, and Pt-signals

in their respective elemental maps using the MSA plugin implemented within DigitalMicrograph

by HREM Research Inc. Geometric phase analysis (GPA) was applied to atomic-resolution

HAADF images using the plugin implemented within DigitalMicrograph by HREM Research

Inc.

5

S7. Fabrication of the nanowire working electrode, the platinized n+-p Si substrate and ITO

Photodeposition of Pt nanoparticles on as-grown GaN/InGaN nanowires was conducted under

vacuum in a sealed glass reactor with a quartz lid. A UV-enhanced 300 W Xenon lamp

(Bulbelectronics, USA) was used to shine light from the top quartz lid. The sample was

immersed in a mixed solution of methanol (12 mL) and deionized water (50 mL). 20 uL of 1mM

Chloroplatinic acid hydrate (99.9%, Sigma Aldrich) was used as Pt precursor. Under

illumination, photo-excited holes from nanowires were consumed by methanol while the Pt

precursor was reduced to form Pt nanoparticles on the surface of nanowires. A thin layer of

indium-gallium eutectic was then applied on the backside of the Si substrate, which was attached

to a copper wire by silver paste. After drying in air, the working electrode was prepared by

capsuling the sample backside and edge with epoxy, with only nanowires on the growth front

exposed in solution. The n+-p Si substrate was rinsed by buffered HF solution for 2 mins to

remove the surface oxide layer, and was subsequently loaded into an electron beam evaporation

chamber to deposit ~ 1 nm Pt. The ITO substrate (12 Ω·cm from Sigma Aldrich) was deposited

with 1 nm Pt by electron beam deposition as well.

S8. Photoelectrochemical reaction and measurement

An Interface 1000 potentiostat (Gamry Instrument) was used, and the three-electrode

electrochemical testing system includes an Ag/AgCl reference electrode, a Pt wire, and a

monolithically integrated InGaN/GaN/Si photocathode. The light source for

photoelectrochemical measurement was a 300 W Xenon lamp (Hamamatsu Photonics, Japan).

H2 evolution was measured using the three-electrode configuration at 0.26 V vs. NHE in a

vacuum sealed quartz chamber under 1.3 sun conditions. The light intensity was calibrated using

both a thermopile (818P-100-55, Newport) and a photodiode sensor (818-ST2-UV/DB, Newport)

6

with attenuator. The evolved H2 gas was sampled using an air tight syringe and analyzed by a gas

chromatograph (Shimadzu GC-8A) equipped with a thermal conductivity detector and Argon

carrier gas.

S9. Solar cell performance of the n+-p Si substrate

After the deposition of p- and n-metal contacts, the performance of the n+-p Si substrate as a

solid state solar cell was tested under AM1.5G illumination. Without the use of any surface

passivation or anti-reflection coating, the solar cell efficiency was measured to be 8.24%, with a

short circuit current of 10 mA, an open circuit potential of 0.75V, and a fill factor of 0.538,

shown in Figure S2. Considering ~35% reflection of the incident light on polished Si wafers6,7

, a

much higher efficiency could be obtained based on such n+-p Si solar cell wafers, if appropriate

surface passivation and anti-reflection coating are applied.

Figure S2. I-V curves of the n+-p Si solar cell (red circles) measured under AM1.5G illumination (49

mW) and under dark conditions (black squares).

7

S10. On the photoelectrochemical performance of the platinized Si solar cell wafer and

platinized InGaN tunnel junction nanowires on n+-p Si substrate

The n+-p Si solar cell wafer was platinized by electron beam evaporation. With a nominal

thickness of 1 nm, Pt nanoparticles were formed on the Si surface, which were sufficient for

HER but did not block the incident light8. As shown by the purple curve in Fig 3b, the platinized

n+-p Si substrate exhibits an onset potential of 0.36 V vs. NHE for a photocurrent density of -2

mA/cm2. A saturated photocurrent density of -22.7 mA/cm

2 was reached at VNHE = -0.17 V. The

relatively low saturated current density of the n+-p Si solar cell wafer, compared to the

theoretical value of -56.15 mA/cm2 under 1.3 sun illumination, is directly related to the light

reflection as well as the non-optimal Pt loading on Si surface. During the e-beam deposition

process, Pt nanoparticles may not necessarily nucleate at the active sites of HER. In addition, any

remnant native SiOx can cause bleaching of some Pt nanoparticles during the subsequent

photoelectrochemical measurements.

The equivalent circuit diagram of the double-band photocathode is shown in Fig. S3. The p-

InGaN tunnel junction nanowires generate a photovoltage at the p-InGaN/electrolyte junction

(semiconductor-liquid junction). Considering the bandgap of InGaN (~2.39 eV) and the

downward band-bending at the p-InGaN/Pt/electrolyte interface, the HER resistance of the p-

InGaN/electrolyte interface, represented by RHER,2 is supposed to be smaller than the HER

resistance at the n-GaN/Pt/electrolyte interface, represented by RHER,1. The excited carriers with

energy 2.39 eV in p-InGaN can in principle reduce H+ without external bias

9. Therefore, RHER,1

does not eliminate the contribution of p-InGaN. The poor fill factor of p-InGaN/tunnel

junction/n-GaN on n+-Si substrate (Fig. 3a in the main text) could be mainly due to the high

interface serial resistance for the transport of photoexcited holes in p-InGaN, represented by

8

RSi/GaN in Fig. S3, since metallically doped n-Si is not activated by light. In contrast, the

accumulation of photoexcited holes in the p-InGaN nanowire segment may reduce the interface

serial resistances for the injection of photoexcited electrons from n+-p Si.

Figure S3. The equivalent circuit diagram of the double-band photocathode under illumination. The

photodiodes, n+/p Si and p-InGaN/electrolyte, represent the photovoltages from the buried n

+-p Si

junction and the p-InGaN/electrolyte semiconductor-liquid junction. The current sources, Jph,1 and Jph,2,

represent the photocurrent densities generated from Si and InGaN respectively.

As discussed in the main text, the performance of the underlying Si solar cell substrate is

significantly improved with the integration of InGaN/GaN nanowire arrays. This is explained by

the enhanced light trapping effect, reduced nonradiative surface recombination, and efficient

carrier extraction by the platinized GaN:Si nanowire segment. The light trapping effect of

nanowire arrays has been explored previously for photovoltaic applications10-15

. InGaN/GaN

9

nanowire arrays have subwavelength dimensions and can strongly scatter the incident light12

,

which significantly enhances the light absorption of the underlying n+-p Si substrate. In addition,

the large valence band offset of ~2 eV at the n-GaN/n+-Si heterointerface blocks the

photoexcited holes from Si and acts as a back surface field in the solar cell, which reduces the

surface recombination of photoexcited charge carriers in Si. Therefore, the HER resistance at the

n-GaN/Pt/electrolyte interface, RHER,1, is smaller than the HER resistance at the n+-p

Si/Pt/electrolyte interface.

Our studies on equivalent InGaN/GaN nanowire arrays grown on n+-Si substrate confirmed that

such nanowires could generate a photocurrent density of -5 mA/cm2 under 1.3 sun illumination

conditions (Fig. 3a in the main text). Therefore, the photocurrent density generated by proton

reduction from the bottom n-GaN segment of nanowires, due to the injection of photoexcited

electrons from the underlying Si solar cell substrate, is estimated to be ~36 mA/cm2.

S11. On the stability of the monolithically integrated InGaN/GaN/Si photocathode

While some of the monolithically integrated InGaN/GaN/Si photocathodes were quite stable, it

was also observed that some other devices degraded rapidly. The degradation did not apparently

affect the saturated photocurrent but shifted the onset potential significantly, shown in Figure S4,

which also led to a reduction of the fill factor of the integrated photocathode. After 1 hour of

photoelectrochemical reaction in 1M HBr, the morphology of nanowires was not changed,

shown in Figure S5. The change of onset potential in solution may be explained by the increased

resistance at the GaN/Si interface during reaction, partly due to the interface etching related

problems. During the growth of InGaN/GaN nanowires on Si, a thin layer of SiNx may form at

the GaN/Si interface16-18

. However, the SiNx can react with HBr to form SiBrx during

10

photoelectrochemical measurements. Therefore, performance of the InGaN/GaN tunnel junction

nanowires on n+-p Si substrate can degrade without noticeable morphological change.

Figure S4. The change of current density with applied voltage versus NHE for InGaN/GaN nanowires

grown on Si solar cell substrate in 1M HBr solution under dark (fully filled black squares) and 1.3 sun of

AM1.5G illumination (fully filled red circles and blue triangles). The onset potential of photocurrent

shows a significant shift after 20 minutes of reaction, while the saturated photocurrent density is nearly

unchanged.

Figure S5. SEM image of InGaN/GaN nanowires after an extended period of photoelectrochemical

reaction. Bright dots on nanowires are Pt nanoparticles.

11

REFERENCES

(1) Sze, S. M.; Ng, K. K. Physics of semiconductor devices, John Wiley & Sons,

U.S.A., 2007, pp 226.

(2) Morkoç, H. Handbook of Nitride Semiconductors and Devices, Wiley-VCH

Verlag GmbH & Co. KGaA: Weinheim, 2008, vol 1, pp 18.

(3) Grabowski, S. P.; Schneider, M.; Nienhaus, H.; Mönch, W.; Dimitrov, R.;

Ambacher, O.; Stutzmann, M. Appl. Phys. Lett. 2001, 78, 2503.

(4) Wu, C.; Kahn, A.; Taskar, N.; Dorman, D.; Gallagher, D. J. Appl. Phys. 1998, 83,

4249.

(5) Guha, S.; Bojarczuk, N. A. Appl. Phys. Lett. 1998, 72, 415.

(6) Huang, Y.-F.; Chattopadhyay, S.; Jen, Y.-J.; Peng, C.-Y.; Liu, T.-A.; Hsu, Y.-K.;

Pan, C.-L.; Lo, H.-C.; Hsu, C.-H.; Chang, Y.-H.; Lee, C.-S.; Chen, K.-H.; Chen, L.-C. Nat.

Nanotechnol. 2007, 2, 770-774.

(7) Branz, H. M.; Yost, V. E.; Ward, S.; Jones, K. M.; To, B.; Stradins, P. Appl. Phys.

Lett. 2009, 94, 231121.

(8) Boettcher, S. W.; Warren, E. L.; Putnam, M. C.; Santori, E. A.; Turner-Evans, D.;

Kelzenberg, M. D.; Walter, M. G.; McKone, J. R.; Brunschwig, B. S.; Atwater, H. A. J. Am.

Chem. Soc. 2011, 133, 1216-1219.

(9) Kibria, M.-G.; Hieu P.-T.-N., Cui, K.; Zhao, S.; Liu, D.; Guo, H.; Trudeau, M.-L.;

Paradis, S.; Hakima, A.-R.; Mi, Z. ACS Nano 2013, 7, 7886-7893.

12

(10) Liu, W.; Oh, J.; Shen, W. IEEE Electron Dev. Lett. 2011, 32, 45-47.

(11) Garnett, E.; Yang, P. Nano Lett. 2010, 10, 1082-1087.

(12) Xie, W.; Oh, J.; Shen, W. Nanotechnol. 2011, 22, 065704.

(13) Callahan, D. M.; Munday, J. N.; Atwater, H. A. Nano Lett. 2012, 12, 214-218.

(14) Iyengar, V. V.; Nayak, B. K.; Gupta, M. C. Sol. Energ. Mat. Sol. Cells 2010, 94,

2251-2257.

(15) Wen, L.; Zhao, Z.; Li, X.; Shen, Y.; Guo, H.; Wang, Y. Appl. Phys. Lett. 2011, 99,

143116.

(16) Consonni, V.; Hanke, M.; Knelangen, M.; Geelhaar, L.; Trampert, A.; Riechert, H.

Phys. Rev. B 2011, 83, 035310.

(17) Ristić, J.; Calleja, E.; Fernández-Garrido, S.; Cerutti, L.; Trampert, A.; Jahn, U.;

Ploog, K. H. J. Cryst. Growth 2008, 310, 4035-4045.

(18) Furtmayr, F.; Vielemeyer, M.; Stutzmann, M.; Arbiol, J.; Estradé, S.; Peirò, F.;

Morante, J. R.; Eickhoff, M. J. Appl. Phys. 2008, 104, 034309.