Submitted to

1

DOI: 10.1002/aenm.201200949 Article type: Full Paper

The role of nanotubes in carbon nanotube-silicon solar cells

Daniel D. Tune, Frank Hennrich, Simone Dehm, Michael F. G. Klein, Konstantin Glaser,

Alexander Colsmann, Joseph G. Shapter, Uli Lemmer, Manfred M. Kappes, Ralph Krupke and

Benjamin S. Flavel*

D. D. Tune, Prof. J. G. Shapter Centre for Nanoscale Science and Technology, Flinders University, Bedford Park, SA 5042, Australia

E-mail: Benjamin.Flavel@partner.kit.edu Dr. F. Hennrich, S. Dehm, Prof. M. M. Kappes, Prof. R. Krupke, Dr. B. S. Flavel

Institute of Nanotechnology (INT), Karlsruhe Institute of Technology (KIT), 76021 Karlsruhe, Germany M. F. G. Klein, K. Glaser, Dr. A. Colsmann, Prof. U. Lemmer

Light Technology Institute (LTI), Karlsruhe Institute of Technology (KIT), 76131 Karlsruhe, Germany

Dr. A. Colsmann, Prof. U. Lemmer, Prof. R. Krupke DFG Centre for Functional Nanostructures (CFN), Karlsruhe Institute of Technology (KIT), 76131 Karlsruhe, Germany

Prof. R. Krupke Institute for Materials Science, Technische Universität Darmstadt, 64287 Darmstadt,

Germany Keywords: carbon nanotubes; electronic structures/processes/mechanisms; photovoltaic

devices; solar cells; thin films

The mechanism of action of nanotube-silicon heterojunction solar cells is under discussion

with literature reports suggesting either p-n or Schottky junction characteristics. The crux of

the issue is whether the nanotubes contribute to the observed photocurrent or not. In order to

further understand the mechanism of action of these solar cells, devices were fabricated using

nanotubes sorted by (n,m) species, so that the excitonic transition is well defined and is

outside the range of absorption of silicon and such that any contribution to the photocurrent

from the nanotubes should be easily resolved from that of the silicon by analysis of the

photocurrent spectrum. The devices exhibited the photocurrent spectra of silicon only,

indicating that the nanotubes do not contribute to the photocurrent. However, by changing

Submitted to

2

the back contact electrode material, results were obtained that appear to show such a

contribution.

1. Introduction

Carbon nanotube-silicon solar cells have seen rapidly increasing research interest in recent

years[1] with advances made in just the last year bringing the current maximum power

conversion efficiency to 11.5 % in the dry state,[1b] 15% with a TiO2 antireflective layer[1d]

and >18 % in the wet state.[2] In this architecture, a thin film of carbon nanotubes is placed in

contact with the surface of a silicon wafer to produce a photoactive junction. Such nanotube-

semiconductor junctions have wide application in a range of advanced materials fields

including electronics, catalysis, sensing and more. In the literature, there are two main

mechanisms suggested for how these devices convert light to electrical power; the devices

work in the manner of 1) a p-n heterojunction solar cell, with the nanotubes contributing as

photoactive material[3] or, 2) a metal-semiconductor (MS) Schottky junction solar cell or the

closely related metal-insulator-semiconductor (MIS) solar cell, where the nanotubes operate

as transparent conducting electrodes.[4] Carbon nanotubes have been shown to a) produce a

photocurrent when exposed to light,[5] b) be capable of forming photoactive p-n junctions[6]

and c) be capable of contributing to photocurrent generation in photoactive donor-acceptor

pairs.[7] Additionally, it is known that air-exposed semiconducting single walled carbon

nanotubes are p-type due to oxygen adsorption.[8] Considering this, it is not unreasonable to

assume that nanotube-silicon solar cells could function in the manner of p-n junctions.

Indeed, the literature contains reports that suggest exactly that; recent measurements on

nanotube-silicon devices by Jung preclude the MIS or MS mechanism.[1b] However, other

measurements on similar nanotube-silicon devices show greatly enhanced (500x)

performance with an interfacial oxide layer, suggesting an MIS-type mechanism.[1e]

Submitted to

3

A more direct probe of the mechanism of action is to use photocurrent spectroscopy to

measure the external quantum efficiency (EQE) of the devices. Contribution from the

nanotubes to the photocurrent would convincingly rule out an MS or MIS mechanism. In the

work of Tzolov,[3a] photocurrent spectroscopy of nanotube-silicon devices shows a broad

feature in the IR that is attributed to the nanotubes. However this feature is very similar to

the free carrier absorption profile of the silicon substrate,[9] confounding the issue. Also, in

the EQE data of Ong[3g] there is a feature that is attributed to photocurrent generation by the

nanotubes. However this small feature in the EQE, which roughly corresponds to the optical

gap of the (7,6) nanotubes used in the study (~ 1150 nm), sits right on the edge of the silicon

absorption spectrum (~ 1200 nm) thus the conclusion of device p-n junction characteristics

based solely on this feature is uncertain.

To determine more conclusively whether or not nanotubes contribute to photocurrent

generation in carbon nanotube-silicon solar cells, devices were fabricated using two different

semiconducting nanotube samples enriched in either small or large diameter nanotubes.

These were chosen since the first excitonic transition (S11) of the major (6,5) chirality in the

small diameter nanotubes lies at ~ 980 nm and overlaps the silicon spectrum similarly to the

(7,6) nanotubes used by Ong,[3g] whereas in the larger diameter nanotubes the gap is

~ 1320 nm, well above the absorption onset wavelength of silicon. Any contribution to the

photocurrent from the larger diameter nanotubes should thus be easily resolved by analysis of

the photocurrent spectrum.

2. Results

Figure 1 shows photoluminescence (PL) measurements of the semiconducting fractions of

nanotubes produced by the a) CoMoCat and b) laser ablation methods (hereafter referred to

as COM and LAB), where the excitation and emission energies correspond to the second

Submitted to

4

(S22) and first (S11) excitonic transitions, respectively. It can be clearly seen that the two

samples are composed of markedly different populations of nanotube chiralities. The smaller

diameter COM sample contains a very high proportion of the (6,5) nanotube species whilst

the larger diameter LAB material has the (9,7) species and smaller quantities of (10,5), (13,2)

and (12,4). Absorption of emitted photons by the inner filter effect of water for λ > 1400 nm

prevents any larger diameter nanotube species from being observed. Although the LAB

material contains a larger number of nanotube species it is clear that the LAB material will

still have its first excitonic transitions, S11, corresponding to wavelengths > 1250 nm, well

outside that of any of the COM material (~ 950 – 1150 nm) and, importantly for this study,

well outside the range of silicon absorption.

UV-Vis-NIR absorption spectra of nanotube films on glass (Figure 2, dash-dotted lines)

support the PL measurements, with the COM material displaying a dominant S11 absorption

peak at ~ 1030 nm and the LAB material showing a broad absorption band spanning ~ 1300 –

1800 nm. The sharp Gaussian present in the absorption spectrum of the COM material is

indicative of the (n,m) purity of the nanotube sample as also measured by

photoluminescence, whereas the broad band present in the spectrum of the LAB material is a

convolution of several peaks relating to the four major (n,m) species shown in the PL

measurements. Closer inspection reveals a substantial difference between the positions of the

S11 transitions as measured by PL and UV-Vis-NIR. The PL yields an S11 of ~ 980 nm for the

(6,5) species and ~ 1320 nm for the (9,7) species whereas the absorption spectra yield ~ 1030

nm for the (6,5) and at least ~ 1490 nm for the (9,7). This apparent shift in the energy of S11

has been reported in the literature and, in the case of the (6,5) nanotube, exactly matches that

observed elsewhere.[10]

Submitted to

5

Here it is important to note that the PL measurements are performed on highly disperse,

aqueous solutions of surfactant-wrapped semiconducting nanotube material immediately

following the sorting process. However, the transmission spectra are obtained from dry

nanotube films on glass substrates, prepared from the sorted fractions, wherein the relatively

surfactant- free nanotubes are present as bundles. Since the excitonic properties of carbon

nanotubes are greatly affected by many body interactions, Coulomb interaction and charge

transfer between adjacent nanotubes in bundles significantly distorts the thin film

transmission measurements. Additionally, in the dry transmission measurements, the filter

effect of water is not an issue thus the contribution of the larger diameter nanotubes is

observable, yielding a much larger apparent shift in S11 of the more polychiral LAB material

than for the purer COM material. Therefore, if there is a contribution from the nanotubes

then the S11 features in the transmission measurement of the dry films (not those in the PL of

the aqueous suspensions) should be represented in the EQE of solar cell devices that employ

the same dry films.

The UV-Vis-NIR transmission properties of nanotube films on silicon (Figure 2, dotted lines)

are dominated by silicon absorption due to a) the fundamental bandgap (< 1200 nm, steep

onset) and b) free carrier absorption (> 1400 nm, slow onset). Since the wavelength of light

corresponding to the first excitonic transition, S11, of the major (6,5) nanotube species in the

COM material is well below that corresponding to silicon’s fundamental bandgap, the major

absorption features due to (6,5) material are not discernible. Still, a small contribution to the

absorption can be seen in the region 1000 – 1400 nm (Figure 2(a), dotted lines) arising from

the shoulder on the higher wavelength side of the main (6,5) peak. In the case of the LAB

material, the S11 absorption energies of the major chiralities are well below silicon’s bandgap.

Consequently, the S11 absorption features of this material are clearly visible in the range 1300

Submitted to

6

– 1800 nm (Figure 2(b), dotted lines) and exactly match the corresponding features in the

absorbance spectra of LAB films on glass (Figure 2(b), dash-dotted lines).

Carbon nanotube-silicon solar cells were fabricated with semiconducting COM and LAB

nanotube material (the same thickness films as those used for the optical measurements), with

eGaIn back contact electrodes, and the EQE of as prepared and HF treated devices was

measured (Figure 3). Immediately apparent is the poor response of the as prepared devices

due to a thick oxide formed on the silicon during the nanotube film deposition. On the other

hand, the HF treated device shows a significant improvement in EQE. The high porosity of

the nanotube film means that hydrofluoric acid can penetrate through to the underlying

silicon allowing the oxide to be removed without harm to the nanotubes.[11] Indeed, the

presence of an S11 absorption feature in the UV-Vis-NIR spectra of the LAB nanotube films

on silicon show that they are still present and their electronic properties unchanged after the

HF treatment although adsorbed oxygen will have been largely replaced with fluorine.[11]

Comparison of the EQE of COM solar cell devices with the UV-Vis-NIR absorption spectra

of COM films on glass shows no correlation between the position of S11 and any features in

the EQE. Importantly, a similar comparison of the EQE of LAB solar cell devices with the

UV-Vis-NIR absorption spectra of LAB films on silicon also shows no such correlation. Any

contribution to photocurrent generation resulting from photons absorbed via the S11 transition

of the nanotubes should be clearly visible in the EQE in the range corresponding to the

observed peak in the UV-Vis-NIR. Even though the photocurrent response of the COM

nanotubes would be obscured by the silicon response, we note that (a) EQE measurements

are extremely sensitive by design, (b) the peak S11 absorption of the (thin) (6,5) films was

~10 % and, (c) the EQE of none of the devices approached 100%, leaving plenty of room for

the nanotube response to be observed over that of the silicon base. In fact, the EQE of both

Submitted to

7

COM and LAB devices are very similar to each and show no significant deviation from what

would be expected from a silicon Schottky junction solar cell, which is in apparent

disagreement with [3g]. It is possible that different mechanisms apply to devices made with

different nanotube species due to variations in band alignment with the silicon substrate thus

the lack of a response from the LAB devices cannot be interpreted to mean that there can be

no response from devices using larger bandgap material such as COM (6,5) and the (7,6)-

enriched material used in [3g]. However we observe no response from either the smaller

bandgap LAB or the larger bandgap COM devices used in this study and we note that the

LUMO of the (7,6) nanotube is only ~80 meV below (6,5). The light (AM1.5G) and dark IV

measurements for the devices shown in the EQE measurements yielded VOC = 410 mV, JSC =

9.4 mA/cm2, FF = 0.35, Rseries = 95 n = 1.46 and efficiency = 1.4 % for LAB and VOC =

395 mV, JSC = 17.4 mA/cm2, FF = 0.35, Rseries = 150 n = 1.26 and efficiency = 2.4 % for

COM and that these are comparable with the VOC = 370 mV, JSC = 14.6 mA/cm2, FF = 0.3,

Rseries = 150 n = 3.75 and efficiency = 1.7 % in [3g].

It was demonstrated that the nanotubes do not contribute to the photocurrent in these devices

but then the influence of the back contact was studied by fabricating otherwise identical solar

cell devices using silver paste as the back contact electrode instead of eGaIn, and measuring

the EQE (Figure 4). The work function of eGaIn is ~ 4.16 eV whereas for silver paste ~

4.5 – 4.8 eV and eGaIn makes a good ohmic contact to n-type silicon ( ~ 4.26 eV) whereas

silver paste does not. In the spectra of as prepared COM (silver) devices (Figure 4(a), dotted

line) there is a prominent feature in the range ~ 900 - 1200 nm that is not present in devices

where eGaIn is used as the back contact (Figure 4(a), dashed line). A similar feature is also

present in the EQE of the as prepared, silver-backed LAB devices (Figure 4(b), dotted line)

and even after HF treatment this feature remains clearly apparent as a shoulder in Figure 4(a),

Submitted to

8

solid line. Importantly, the features in the EQE, before and after HF treatment of the silicon,

appear to be independent of any of the nanotube absorption characteristics shown for

comparison in Figure 4(a) and Figure 4(b), dash-dotted lines.

The decrease in the EQE towards shorter wavelengths (1000 – 900 nm) is typical of a back-

illuminated Schottky junction solar cell.[12] Here it is important to note that the terms ‘front-

illuminated’ and ‘back-illuminated’ refer to whether the rectifying contact is illuminated

directly, or indirectly through the device substrate, respectively. To determine whether or not

this is the origin of the 900 – 1200 nm feature in Figure 4, the internal quantum efficiency

(IQE) of a back-illuminated solar cell with an ohmic contact and a rectifying contact, as

shown in Figure 5(a), has been calculated using the analytical model of Basu and Saha.[12]

Equation (9) in that work uses as fitting parameters the thickness of the semiconducting

layer t, the width of the depletion layer d, the minority carrier diffusion length L, the minority

carrier diffusion coefficient D, the surface recombination velocity s, and the

wavelength-dependent absorption coefficient α (λ). This work used t = 525 μm, L = 500 μm,

D = 10-3 m2.s-1, for 525 μm thick n-Si with a donor density of 1014 cm-3,[13] used (λ) from

Green[14] and d 10-6 m and s = 1 m.s-1 were set as best fit to the photocurrent spectra of our

as-prepared cells with silver back contacts. The IQE of a front-illuminated solar cell with a

rectifying contact and an ohmic contact, as shown in Figure 5(b), has also been calculated,

based on the analytical model of Soukoup and Akers[15] requiring only L and (λ) as

parameters, which was chosen identical to the back-illuminated case. Figure 5(c) shows the

IQE modeling results for front- and back-illuminated solar cells. The non-monotonic

behavior for the back-illuminated cell is due to the enhanced absorption of shorter

wavelength light by the substrate and the limited diffusion of the minority carrier towards the

Submitted to

9

charge separating back contact. Hence a back-illuminated device can generate photocurrent

only in the long wavelength range.

Comparing the IQE spectrum of the modelled back-illuminated silicon Schottky junction

solar cell, Figure 5(a), with the feature observed in the EQE of the silver-backed carbon

nanotube-silicon solar cell, Figure 4(a), solid line and dotted line, it seems likely that the

enhanced EQE in the wavelength range close to the absorption edge of silicon arises due to

the contribution of a rectifying back contact rather than as a result of contribution to the

photocurrent from the nanotubes. The presence of this contact leads to predictably poor solar

IV parameters of VOC = 180 mV, JSC = 0.5 mA/cm2, FF = 0.14, Rseries = 24 kn = 1.97 and

efficiency = 0.01 % for LAB and VOC = 226 mV, JSC = 3.1 mA/cm2, FF = 0.17, Rseries = 53

kn = 2.06 and efficiency = 0.12 % for COM. This is despite the fact that the silver backed

devices were otherwise identical to the eGaIn devices and the silver paste itself is highly

conductive.

The nanotube films used in this study and others are, by design, very thin and it may be that

the contribution of such a film is so small that it could not be observed. However, a device

made with a much thicker nanotube film not only does not show a contribution from the

nanotubes but, rather, shows distinct reductions in efficiency in the regions of the nanotube’s

S11, S22 and * absorptions (Figure S1). This is expected though since any exciton

produced in the nanotube film away from a nanotube-silicon junction would have no driving

force for separation and cannot percolate far through the nanotube network in the manner of

free charges. Any contribution could only come from the nanotubes in a thin region close to

the silicon surface and so a thicker film effectively just gets in the way.

Submitted to

10

If the nanotubes do not contribute to the photocurrent then it may be that they function

primarily as transparent front electrodes. This is possible even for a device made with high

purity semiconducting nanotube material as, even without the presence of fluorine resulting

from intentional acid treatment, adsorbed oxygen molecules act as p-type dopants. Indeed,

for electrode purposes, one would ideally like to use highly doped semiconducting nanotubes

(with a Fermi level pushed into the valence or conduction band) in preference to metallic

nanotubes.[16] Even if the nanotubes do not contribute to the photocurrent, this would not

necessarily mean that the mechanism of action of the devices is not that of a p-n junction.

Measurements by Jung[1b] on their carbon nanotube-silicon solar cells indicate that the

mechanism is similar to that of a single crystal silicon p-n homojunction. In those devices,

removal of the native oxide significantly improved device performance. However,

measurements by Jia[1e] on similar devices showed the opposite; that the presence of an

interfacial oxide layer improved device performance, indicating an MIS mechanism. These

differing results may be related to the structure of the nanotube films in the different devices.

In Jia’s devices the nanotube bundles were present as a porous, randomly aligned membrane

but in Jung’s the bundles were very highly aligned by design. The difference could also be

due to the different types of nanotube material, their (n,m) species purity or the effects of

treatments they received during fabrication. Such apparently conflicting data further

illustrates the complexity and diversity of carbon nanotube films and their properties and

underscores the need for more work to determine the effect of various films’ nanostructure

and component nanotube type/species.

3. Conclusion

The photocurrent spectra of carbon nanotube-silicon solar cells fabricated with small and

large diameter semiconducting nanotubes do not show any features that could be attributed to

Submitted to

11

photocurrent generation by the nanotubes. This is despite the fact that the nanotubes used in

this work were deliberately chosen so that any such contribution should be clearly

discernible. Thus the nanotubes in the devices reported herein are not contributing elements

of a p-n junction-type solar cell. Photocurrent spectra of carbon nanotube-silicon solar cells

fabricated with non-ohmic back contact electrodes do show a similar feature to that which has

been elsewhere attributed to photocurrent generation by the nanotubes. However the data

herein suggests that this feature arises due to the choice of back contact electrode material.

With these understandings in mind, future research endeavours in this field should be largely

directed towards exploiting and optimising the application of carbon nanotubes as front

contact electrodes. With very high transparency and excellent conductivity, carbon

nanotubes are well suited to this purpose and could have the potential to surpass the

performance of the current generation of front metallisation strategies via reduced optical

shading and enhanced carrier collection over the full solar cell surface.

4. Experimental Section

Two types of SWNT raw materials were used for this study: (i) CoMoCat SWNTs (Sigma-

Aldrich) and (ii) SWNTs prepared in-house by pulsed laser vaporisation in an argon

atmosphere using carbon targets doped with 1 atom% Ni and Co catalyst and an oven

operated at 1000 °C.[17] For the starting suspensions each of the SWNT material (typically 10

mg) was suspended with sodium dodecyl sulphate (SDS) (1 wt.%) in H2O (15 ml) using a tip

sonicator (Bandelin, 200 W maximum power, 20 kHz, in pulsed mode with 100 ms pulses)

applied for 2 hr at 20 % power. The resulting dispersion was then centrifuged at 100,000 g

for 1 hr (Beckmann Coulter) to remove larger agglomerates.

Gel filtration was performed as described by Moshammer[18] using Sephacryl S-200 gel

filtration medium (Amersham Biosciences) in a glass column of 20 cm length and 2 cm inner

Submitted to

12

diameter. After filling the glass column with the filtration medium, the gel was slightly

compressed to yield a final height of approximately 14 cm. For the separation, SWNT

starting suspension (10 ml) was applied to the top of the column and subsequently a solution

of SDS (1 wt.%) in H2O as eluent was pushed through the column by applying sufficient

pressure with compressed air to ensure a flow of ~ 1 ml.min-1. After ~ 10 ml of this eluent

had been added most of the metallic SWNTs had moved through the column whereas the

semiconducting SWNTs remained trapped in the upper part of the gel. After applying a total

of 20 ml of SDS solution in this fashion, the metallic nanotubes were completely removed

from the gel. By then changing the eluent from SDS (1 wt.%) in H2O to sodium cholate (2

wt.%) in H2O the semiconducting SWNTs also subsequently eluted completely from the

column and were collected separately in fractions of approximately 1 ml.

Randomly aligned nanotube membranes were prepared similarly to Wu.[19] Briefly, the

nanotubes were filtered through mixed cellulose ester (MCE) membranes (0.45 µm, HAWP,

Millipore) and rinsed thoroughly with deionised water. Circular regions (0.32 cm2) of the

resulting SWNT-MCE membrane were taken for device fabrication. Phosphorous doped n-

type silicon wafers (CZ, 5-25 Ω.cm, <100>, ABC GmbH), polished on one side and with a

thermal oxide (100 nm) were diced into pieces (2 cm2). E-beam lithography (Raith) was used

to define square regions (0.09 cm2) in the negative resist (maN-2403, micro resist technology

GmbH, Germany), which was developed and then the front metal contact (Ti/Au, 5/50 nm)

was sputtered. Following lift off the oxide in these regions was etched with a reactive ion

etcher (50 sccm CHF3, 5 sccm O2, 140 W) (Plasmalab 80 Plus, Oxord). The SWNT films

were deposited onto the substrate surfaces by placing the circular SWNT-MCE membranes,

nanotube side down, centred over the etched holes. The membranes were wetted with

deionised water then compressed and baked (130 ºC, 15 min). Upon cooling, the substrates

Submitted to

13

were immersed in acetone to dissolve the MCE leaving nanotube membranes of

approximately 35 nm in thickness[20] tightly adsorbed onto the substrate surfaces providing

0.09 cm2 square active areas surrounded by 0.23 cm2 regions where the nanotube membranes

overlap the front metal contacts. After etching of the rear oxide with HF (2 %) the devices

were completed with silver paste or gallium indium eutectic (eGaIn) back contacts and

mounted on steel plates (Figure 6).

UV-Vis-NIR spectra of films of sorted nanotube material were recorded with a Varian Cary

500 spectrophotometer. In the transmission measurements of films on silicon, the maximum

transmission was 111-116 % @ 1300nm due to the uncoated silicon used as a baseline having

a greater absorbance than the nanotube coated sample thus the spectra were normalized to

100% @ 1300 nm. A similar process was applied to the absorption spectra of films on glass,

for the same reason. Photoluminescence excitation (PL) maps of sorted nanotube material in

solution were measured in the emission range of 900 - 1700 nm and excitation range of 500 -

950 nm (scanned in 3 nm steps) using a modified FTIR spectrometer (Bruker IFS66)

equipped with a liquid nitrogen cooled Ge photodiode and a monochromatised excitation

light source as described elsewhere.[21] The EQE was measured using a setup consisting of a

450W xenon light source, an optical chopper (47.7 Hz), a 300 mm monochromator (LOT-

Oriel), a custom designed current amplifier (DLPCA-S, Femto Messtechnik) and a digital

lock-in amplifier (eLockin 203, Anfatec). A modified photoreceiver (OE-200-S, Femto

Messtechnik) with a Si/InGaAs sandwich diode was used to monitor the stability of the

monochromatic light beam. Initial calibration was carried out with reference silicon and

germanium diodes (Thorlabs, NIST traceable calibration).

Supporting Information Supporting Information is available online from the Wiley Online Library or from the author.

Submitted to

14

Acknowledgements

B.S. Flavel gratefully acknowledges the support of the Alexander von Humboldt Foundation.

M.F.G. Klein is further supported by the Karlsruhe School of Optics & Photonics (KSOP).

Received: ((will be filled in by the editorial staff)) Revised: ((will be filled in by the editorial staff))

Published online: ((will be filled in by the editorial staff))

Carbon nanotube-silicon solar cells have recently been reported with 15 % conversion

efficiency, however the mechanism of action of such devices is still uncertain. In this paper, the role of the nanotubes in the carbon nanotube-silicon design is examined, using nanotube

material highly enriched in small or large diameter semiconducting nanotubes, yielding unexpected results.

Keyword carbon nanotubes; electronic structures/processes/mechanisms; photovoltaic devices; solar cells; thin films

D. D. Tune, F. Hennrich, S. Dehm, M. F. G. Klein, K. Glaser, A. Colsmann, J. G. Shapter, U.

Lemmer, M. M. Kappes, R. Krupke and B. S. Flavel* The role of nanotubes in carbon nanotube-silicon solar cells

[1] a) X. Li, Y. Jung, K. Sakimoto, T.-H. Goh, M. A. Reed, A. D. Taylor, Energy & Env.

Sc. 2013; b) Y. Jung, X. Li, N. K. Rajan, A. D. Taylor, M. A. Reed, Nano Lett. 2012, 13, 95; c) D. D. Tune, B. S. Flavel, R. Krupke, J. G. Shapter, Adv. Energy Mat. 2012, 2, 1043; d) E.

Shi, L. Zhang, Z. Li, P. Li, Y. Shang, Y. Jia, J. Wei, K. Wang, H. Zhu, D. Wu, S. Zhang, A. Cao, Sci. Rep. 2012, 2, 884; e) Y. Jia, A. Cao, F. Kang, P. Li, X. Gui, L. Zhang, E. Shi, J. Wei, K. Wang, H. Zhu, D. Wu, PCCP 2012, 14, 8391; f) D. Kozawa, K. Hiraoka, Y.

Miyauchi, S. Mouri, K. Matsuda, Appl. Phys. Express 2012, 5, 042304. [2] See reference 10 in [1b].

[3] a) M. B. Tzolov, T.-F. Kuo, D. A. Straus, A. Yin, J. Xu, J. Phys. Chem. C 2007, 111, 5800; b) J. Wei, Y. Jia, Q. Shu, Z. Gu, K. Wang, D. Zhuang, G. Zhang, Z. Wang, J. Luo, A. Cao, D. Wu, Nano Lett. 2007, 7, 2317; c) A. Arena, N. Donato, G. Saitta, S. Galvagno, C.

Milone, A. Pistone, Microelectron. J. 2008, 39, 1659; d) Z. Li, V. P. Kunets, V. Saini, Y. Xu, E. Dervishi, G. J. Salamo, A. R. Biris, A. S. Biris, Appl. Phys. Lett. 2008, 93, 243117; e) Z.

Submitted to

15

Li, V. P. Kunets, V. Saini, Y. Xu, E. Dervishi, G. J. Salamo, A. R. Biris, A. S. Biris, ACS Nano 2009, 3, 1407; f) Z. Li, V. Saini, E. Dervishi, V. P. Kunets, J. Zhang, Y. Xu, A. R.

Biris, G. J. Salamo, A. S. Biris, Appl. Phys. Lett. 2010, 96, 033110; g) P. L. Ong, W. B. Euler, I. A. Levitsky, Nanotechnology 2010, 21, 105203.

[4] a) A. Behnam, J. L. Johnson, Y. Choi, M. G. Ertosun, A. K. Okyay, P. Kapur, K. C. Saraswat, A. Ural, Appl. Phys. Lett. 2008, 92; b) Y. Jia, J. Wei, K. Wang, A. Cao, Q. Shu, X. Gui, Y. Zhu, D. Zhuang, G. Zhang, B. Ma, L. Wang, W. Liu, Z. Wang, J. Luo, D. Wu, Adv.

Mater. 2008, 20, 4594; c) Y. Jia, P. Li, J. Wei, A. Cao, K. Wang, C. Li, D. Zhuang, H. Zhu, D. Wu, Mater. Res. Bull. 2010, 45, 1401; d) Z. Li, Y. Jia, J. Wei, K. Wang, Q. Shu, X. Gui,

H. Zhu, A. Cao, D. Wu, J. Mater. Chem. 2010, 20, 7236; e) P. Wadhwa, B. Liu, M. A. McCarthy, Z. Wu, A. G. Rinzler, Nano Lett. 2010, 10, 5001; f) Y. Jia, P. Li, X. Gui, J. Wei, K. Wang, H. Zhu, D. Wu, L. Zhang, A. Cao, Y. Xu, Appl. Phys. Lett. 2011, 98, 133115; g) P.

Wadhwa, G. Seol, M. K. Petterson, J. Guo, A. G. Rinzler, Nano Lett. 2011, 11, 2419. [5] a) A. Fujiwara, Y. Matsuoka, H. Suematsu, N. Ogawa, K. Miyano, H. Kataura, Y.

Maniwa, S. Suzuki, Y. Achiba, Jpn. J. Appl. Phys., Part 2 2001, 40, L1229; b) M. Freitag, Y. Martin, J. A. Misewich, R. Martel, P. Avouris, Nano Lett. 2003, 3, 1067. [6] J. U. Lee, Appl. Phys. Lett. 2005, 87, 073101.

[7] a) D. J. Bindl, M.-Y. Wu, F. C. Prehn, M. S. Arnold, Nano Lett. 2010, 11, 55 b) . vr ek, . Cook, . a aoui, M. ondo, J. Phys. Chem. Lett. 2011, 2, 1646.

[8] P. G. Collins, K. Bradley, M. Ishigami, A. Zettl, Science 2000, 287, 1801. [9] W. Spitzer, H. Y. Fan, Physical Review 1957, 108, 268. [10] R. M. Jain, R. Howden, K. Tvrdy, S. Shimizu, A. J. Hilmer, T. P. McNicholas, K. K.

Gleason, M. S. Strano, Adv. Mater. 2012, 24, 4436. [11] K. Kim, J. Ryu, C. Lee, J. Jang, K. Park, J Mater Sci: Mater Electron 2009, 20, 120.

[12] P. Basu, H. Saha, Phys. Status Solidi A 1976, 37, 625. [13] J. A. del Alamo, R. M. Swanson, Solid-State Electron. 1987, 30, 1127. [14] M. A. Green, M. J. Keevers, Prog. Photovoltaics 1995, 3, 189.

[15] R. J. Soukup, L. A. Akers, J. Appl. Phys. 1978, 49, 4031. [16] A. G. Rinzler, E. P. Donoghue, ACS Nano 2011, 5, 3425.

[17] S. Lebedkin, P. Schweiss, B. Renker, S. Malik, F. Hennrich, M. Neumaier, C. Stoermer, M. M. Kappes, Carbon 2002, 40, 417. [18] K. Moshammer, F. Hennrich, M. M. Kappes, Nano Research 2009, 2, 599.

[19] Z. Wu, Z. Chen, X. Du, J. M. Logan, J. Sippel, M. Nikolou, K. Kamaras, J. R. Reynolds, D. B. Tanner, A. F. Hebard, A. G. Rinzler, Science 2004, 305, 1273.

[20] D. D. Tune, B. S. Flavel, J. S. Quinton, A. V. Ellis, J. G. Shapter, ChemSusChem 2013, 10.1002/cssc.201200600. [21] S. Lebedkin, F. Hennrich, O. Kiowski, M. M. Kappes, Phys. Rev. B: Condens. Matter

Mater. Phys. 2008, 77, 165429.

Submitted to

16

Figure 1: Photoluminescence spectra of sorted nanotube material (see text): (a) CoMoCat and (b)

laser ablation. Spectra were taken in aqueous solution where photoluminescence of λ > 1400 nm is

not observable due to absorption by water.

Figure 2: UV-Vis-NIR spectra of films prepared from the semiconducting fractions of (a) COM and

(b) LAB nanotubes. Transmission spectra are shown for films on silicon and absorption spectra are

shown for films on glass.

Figure 3: External quantum efficiency of as prepared and HF treated carbon nanotube-silicon solar

cells fabricated with films of either (a) COM or (b) LAB semiconducting nanotubes with eGaIn

back contact electrodes. UV-Vis-NIR absorption spectra of corresponding films on glass are shown

for comparison.

Submitted to

17

Figure 4: External quantum efficiency of as prepared and HF treated carbon nanotube-silicon solar

cells fabricated with films of either (a) COM or (b) LAB semiconducting nanotubes with silver paint

or eGaIn back contact electrodes. Note, a log EQE scale has been used to allow visual comparison

between the shape of the small “as prepared” EQE profile and the much larger “HF treated” profile.

UV-Vis-NIR absorption spectra of the corresponding nanotube films on glass are shown for

comparison. There was no photocurrent response from any device for λ > 1200 nm.

Figure 5: The schematics in (a) and (b) show the two solar cells modelled with back- and front-

illuminated rectifying contacts, respectively, and their calculated IQE is shown in (c).

Figure 6: Schematic showing solar cell device structure.