Atomic layer-deposited tunnel oxide stabilizes silicon ... - Nature

ARTICLES PUBLISHED ONLINE: 19 JUNE 2011 | DOI: 10.1038/NMAT3047

Atomic layer-deposited tunnel oxide stabilizes silicon photoanodes for water oxidation Yi Wei Chen1† , Jonathan D. Prange2† , Simon Dühnen2 , Yohan Park1 , Marika Gunji1 , Christopher E. D. Chidsey2 and Paul C. McIntyre1 * A leading approach for large-scale electrochemical energy production with minimal global-warming gas emission is to use a renewable source of electricity, such as solar energy, to oxidize water, providing the abundant source of electrons needed in fuel synthesis. We report corrosion-resistant, nanocomposite anodes for the oxidation of water required to produce renewable fuels. Silicon, an earth-abundant element and an efficient photovoltaic material, is protected by atomic layer deposition (ALD) of a highly uniform, 2 nm thick layer of titanium dioxide (TiO2 ) and then coated with an optically transmitting layer of a known catalyst (3 nm iridium). Photoelectrochemical water oxidation was observed to occur below the reversible potential whereas dark electrochemical water oxidation was found to have low-to-moderate overpotentials at all pH values, resulting in an inferred photovoltage of ∼550 mV. Water oxidation is sustained at these anodes for many hours in harsh pH and oxidative environments whereas comparable silicon anodes without the TiO2 coating quickly fail. The desirable electrochemical efficiency and corrosion resistance of these anodes is made possible by the low electron-tunnelling resistance (1 µm) noble metal oxide catalyst coatings. Comparing the overpotentials measured for n-Si anodes in the light with those for p+ -Si anodes in the dark, the photovoltage was calculated to be in the range of 510–570 mV (dark ∼+350 mV, light ∼−200 mV). This photovoltage is similar to that of the best Si photoelectrochemical solar cells27 , and close to the open circuit photovoltage reported for high quality p–n junction Si solar cells (∼700 mV; ref. 28). The somewhat lower photovoltage observed here compared to Si solar cells may result from a less than optimal choice of the work function of the catalyst metal on the TiO2 (ref. 29), and non-idealities such as non-radiative carrier recombination at defects30 . As electrodes used for water oxidation are exposed to highly corrosive and oxidative environments, the endurance of the nanocomposite anodes was investigated in electrochemical life tests. Figure 2 shows the measured potential required for the illuminated anodes to give a constant current of 1 mA through the 0.196 cm2 sample in both 1 M acid and 1 M base solutions. The samples without the TiO2 layer failed under illumination in both solutions within half an hour, reaching the maximum voltage the potentiostat could supply, whereas the samples with the TiO2 layer lasted for at least 8 h, the duration of these endurance tests. Figure 3 shows the current obtained on Ir/TiO2 /p+ -Si in the dark as a function of time over a period of 24 h while being held at a constant potential of 1.7 V versus NHE

3 2 1 0

0

2

4

6

8

10 12 14 16 18 20 22 24 Time (h)

Figure 3 | Anode stability in the dark during water splitting. Constant potential lifetime tests at 1.7 V versus NHE on a 0.196 cm2 sample are in 1 M NaOH solution circulated at 120 ml min−1 on (blue solid line) Ir/TiO2 /p+ -Si and (blue dashed line) Ir/p+ -Si anodes.

with and without the TiO2 layer in the 1 M NaOH solution. The sample with the corrosion resistant TiO2 layer remained operational for at least 24 h, whereas the sample without the TiO2 layer failed within 0.5 h. It is worth noting that the nominal current densities in these lifetime measurements are achieved at higher measured or applied biases than in the cyclic voltammograms in Fig. 1 because of inefficient oxygen removal from the sample surface in the lifetime test cell (see Supplementary Information). Analysis of the samples that gave the results in Fig. 2 by X-ray photoelectron spectroscopy (XPS) depth profiling after constant current life tests (1 mA on a 0.196 cm2 area) revealed that the discrete layering of the nanocomposite remains intact for the anode containing TiO2 (Fig. 4a). Furthermore, crosssectional TEM of similarly tested samples reveals that the SiO2 thickness was comparable to that of the initial interfacial SiO2 before testing (see Supplementary Information). Samples without the TiO2 layer grew a thick, insulating SiO2 layer after the constant current experiment (Fig. 4b). These results indicate that a corrosion resistant TiO2 layer of only 2–3 nm thickness protects the underlying Si substrate under the conditions investigated, while still allowing for efficient transport of electrons and holes between the solution and the Si anode. The efficiency of electronic transport across this interface was characterized in greater detail using a benchmark electrolyte solution composed of aqueous ferri/ferrocyanide ions. The

NATURE MATERIALS | VOL 10 | JULY 2011 | www.nature.com/naturematerials

© 2011 Macmillan Publishers Limited. All rights reserved

541

NATURE MATERIALS DOI: 10.1038/NMAT3047

ARTICLES a

b Si

80 60 40 20 0

O

Ir Ti 0

2

4

6 8 Depth (nm)

100

Atomic composition (%)

Atomic composition (%)

100

10

12

80 O

60 40

Si 20 0

Ir 0

5

10

15

20 25 30 Depth (nm)

35

40 45 50

Figure 4 | Representative XPS depth profiles of samples after stability testing. Analysis of n-Si samples under 1 sun illumination that a, have the TiO2 protection layer and b, samples without the TiO2 protection layer. The elements are represented by black squares for O, red circles for Si, blue triangles for Ir, and green diamonds for Ti.

542

a

4

Ir/2 nm TiO2/n-Si light

3

Ir/2 nm TiO2/p+-Si Ir/10 nm TiO2/p+-Si

i (mA cm¬2)

2 1 0 ¬1

Ir/2 nm TiO2/n-Si

/p+-Si

2 nm TiO2

¬2 ¬3 ¬4

b

¬0.8 ¬0.6 ¬0.4 ¬0.2 0 0.2 0.4 0.6 0.8 1.0 E/V versus Ag/AgCl

1,000 100

i (A cm¬2)

exchange of electrons between this solution and metal electrodes is fast, allowing for characterization of the electronic transport across the interposed TiO2 layer by cyclic voltammetry (ref. 31). Facile electron transport was observed for samples coated with the ultrathin Ir layer (Fig. 5a). Samples without the Ir layer exhibit orders of magnitude lower current density and no observable Fe(ii)/Fe(iii) redox waves, indicating the importance of the ultrathin metal layer as a charge carrier mediator between the substrate and the solution. In place of the Ir overlayer, other metals such as Pt and Ru were deposited and found to serve the same function. The p+ -Si anodes with ALD-TiO2 exhibit relatively small peak-to-peak splitting (130 mV) and high current densities, comparable to results obtained with bulk metal electrodes. Samples with a substantially thicker TiO2 layer (10 nm instead of 2 nm) resulted in increased peak-to-peak splitting (610 mV), indicating the importance of using a thin TiO2 layer for efficient tunnelling-mediated transport of electrons. The Ir/TiO2 /n-Si anodes were analysed in both the dark and solar simulated light for electronic transfer efficiency using the ferri/ferrocyanide solution (Fig. 4a). The dark cyclic voltammetry reveals an asymmetry, with the anodic peak missing and the cathodic peak remaining. This is consistent with the presence of a sufficient concentration of electrons in the n-Si to rapidly reduce Fe(iii) and the lack of holes required to oxidize Fe(ii). This is also the cause of the low water oxidation current density observed for the Ir/TiO2 /n-Si anode without illumination (Fig. 1d). As a result, the electrolyte-solid interface behaves as a Schottky junction, giving rise to the observed cyclic voltammetry asymmetry. In contrast, the cyclic voltammetry of the illuminated sample not only recovers its symmetry, but also displays a negative potential shift and an increase in peak current density. The recovered anodic peak is due to the photo-generated holes. Comparing the Fe(ii)/Fe(iii) redox potentials for the dark p+ -Si and light n-Si samples, a negative shift of ∼550 mV is observed, which is consistent with the shift observed for water oxidation overpotentials at 1 mA cm−2 between the two samples. The intrinsic electron transport properties of the nanocomposite anodes under conditions similar to those used in dark electrolysis were probed by temperature-dependent, metal contact current– voltage (I –V ) measurements. Electron tunnelling through the TiO2 layer on the p+ -Si substrate was confirmed by varying the TiO2 thickness. The current density for thin (4 nm) TiO2 samples

1.2

Current compliance 2 nm TiO2

10 4 nm TiO2 1 0.10 0.01 20

10 nm TiO2

30

40 50 60 Temperature (°C)

70

80

Figure 5 | Electronic transport characterization. a, Cyclic voltammogram of the ferri/ferrocyanide solution for (black line) 2 nm TiO2 /p+ -Si, (red) Ir/2 nm TiO2 /p+ -Si, (blue) Ir/10 nm TiO2 /p+ -Si, (violet) Ir/2 nm TiO2 /n-Si, and (green) Ir/2 nm TiO2 /n-Si with simulated 1 sun illumination. All measurements were done in the dark except where otherwise mentioned. b, Temperature-dependent current density measurement through (red) 2 nm, (green) 4 nm, and (violet) 10 nm TiO2 films on p+ -Si substrates. The dashed horizontal line indicates the compliance of the meter. 50 nm thick Ir dots of 100 µm diameter were used as the top metal contact. Current measured in a probe station (no electrolyte present) at a Si substrate voltage versus the Ir contact of 0.5 V (flatband voltage = −0.13 V). Current–voltage data are provided in Supplementary Information.

result in an increasingly temperature-dependent current density, an indication of a more thermally-activated and bulk-limited conduction mechanism such as trap-assisted tunnelling or Frenkel– Poole conduction (Fig. 5b; ref. 33). The significant dependence on the thickness of the measured electronic conduction across the NATURE MATERIALS | VOL 10 | JULY 2011 | www.nature.com/naturematerials

© 2011 Macmillan Publishers Limited. All rights reserved

NATURE MATERIALS DOI: 10.1038/NMAT3047 TiO2 layer indicates the importance of using a deposition method such as ALD, which exhibits excellent uniformity and control of thickness at the nanoscale. The ALD-TiO2 film thickness we have investigated is much smaller than that described in a recent publication in which conformal deposition of relatively thick (∼35 nm) TiO2 was performed over n- or p-type Si nanowires4 . These coated nanowire arrays exhibited improved behaviour compared to correspondingly coated planar Si photoelectrodes; however, the current densities attributed to water oxidation were much smaller than those that we have observed. This is consistent with greater resistance to carrier transport associated with the larger TiO2 thickness, and with the absence of the Ir layer that catalyzes water oxidation and promotes electronic conduction across our nanocomposite photoanodes. In this report, we have demonstrated and characterized the operation of an efficient and dimensionally stable semiconductor anode for photoelectrochemical water oxidation. This nanocomposite anode uses a pinhole-free, corrosion-resistant, ALD-grown TiO2 tunnel oxide layer that protects an underlying Si substrate during water oxidation at an overlying catalyst layer in both dark and light conditions. The ultrathin ALD-TiO2 layer is thick enough to permit hours of continuous operation in corrosive environments (acidic or basic) without apparent structural change, while being thin enough to allow facile electronic transport via tunnelling. Compared to previously reported metal oxide photoanodes, this nanocomposite device is capable of reaching much higher saturation current densities (tens of mA cm−2 ), and of maintaining low overpotentials at moderate current densities. The measured photovoltage range of ∼510–570 mV approaches the open-circuit photovoltage for state-of-the-art silicon solar cells. The reported nanocomposite structure allows the decoupling of the electrochemical reaction at the catalyst surface from the underlying photovoltaic substrate, which should permit future improvements by further independent optimization of the different components. Therefore, this approach is quite general, and should have applications in protecting semiconductor substrates other than silicon, and in the integration of other conductive catalyst layers besides iridium. Furthermore, ALD is currently used in industrial semiconductor device fabrication34,35 , which indicates the potential of this deposition technique to be employed on a large scale in energy applications such as solar fuel synthesis.

Methods The Si wafers used in the reported experiments were degenerately doped p+ -type Si (100) wafers (0.001–0.002  cm,500 µm thickness) and n-type Si (100) wafers (0.1–0.2  cm,500 µm thickness). The wafers were used as received, with a thin (