Light harvesting with multiwall carbon nanotube/silicon heterojunctions

Light harvesting with multiwall carbon nanotube/silicon heterojunctions

This article has been downloaded from IOPscience. Please scroll down to see the full text article.

2011 Nanotechnology 22 115701

(http://iopscience.iop.org/0957-4484/22/11/115701)

Download details:

IP Address: 151.50.29.112

The article was downloaded on 05/02/2011 at 10:10

Please note that terms and conditions apply.

View the table of contents for this issue, or go to the journal homepage for more

Home Search Collections Journals About Contact us My IOPscience

IOP PUBLISHING NANOTECHNOLOGY

Nanotechnology 22 (2011) 115701 (8pp) doi:10.1088/0957-4484/22/11/115701

Light harvesting with multiwall carbonnanotube/silicon heterojunctionsPaola Castrucci1,5, Claudia Scilletta1, Silvano Del Gobbo1,Manuela Scarselli1, Luca Camilli1, Mirko Simeoni2,Bernard Delley3, Alessandra Continenza4 andMaurizio De Crescenzi1

1 Dipartimento di Fisica and Unita CNISM, Universita di Roma Tor Vergata, 00133 Roma,Italy2 Consorzio CREO, Via Pile 60, 67100 L’Aquila, Italy3 Paul Scherrer Institut, WHGA/123, CH-5232 Villigen PSI, Switzerland4 Dipartimento di Fisica and CNISM, Universita degli Studi di L’Aquila, 67100 L’Aquila, Italy

E-mail: [email protected]

Received 7 October 2010, in final form 23 December 2010Published 4 February 2011Online at stacks.iop.org/Nano/22/115701

AbstractWe report on a significant photocurrent generation from a planar device obtained by coating abare n doped silicon substrate with a random network of multiwall carbon nanotubes(MWCNTs). This MWCNT/n-Si hybrid device exhibits an incident photon to current efficiencyreaching up to 34% at 670 nm. We also show that MWCNTs covering a quartz substrate stillexhibit photocurrent, though well below than that of the MWCNTs coating the silicon substrate.These results suggest that MWCNTs are able to generate photocurrent and that the siliconsubstrate plays a fundamental role in our planar device. The former effect is particularlyinteresting because MWCNTs are generally known to mimic the electronic properties ofgraphite, which does not present any photocurrent generation. On the basis of theoreticalcalculations revealing a weak metallic character for MWCNTs, we suggest that both metallicand semiconducting nanotubes are able to generate e–h pairs upon illumination. This can beascribed to the presence of van Hove singularities in the density of states of each single wallcarbon nanotube constituting the MWCNT and to the low density of electrons at the Fermilevel. Finally, we suggest that though both MWCNTs and Si substrate are involved in thephotocurrent generation process, MWCNT film mainly acts as a semitransparent electrode inour silicon-based device.

(Some figures in this article are in colour only in the electronic version)

1. Introduction

Multiwall carbon nanotubes (MWCNTs) consist of multiplelayers of graphite sheets forming concentric cylinders. Thenumber of concentric cylinders in a MWCNT can vary fromtwo to many tens. These systems are closely related tographite layers and are generally considered to exhibit similarelectronic properties. On the other hand, the MWCNTelectronic density of states is difficult to model, due to the largenumber of shells and their mutual interactions; therefore, onlyfew theoretical predictions of MWCNT electronic properties

5 Author to whom any correspondence should be addressed.

are available [1]. Theoretical and experimental efforts havebeen mainly dedicated to double wall carbon nanotubes(DWCNTs), being the simplest form of multiwall carbonnanotubes [2–5]. In particular, DWCNTs have been foundto generate photocurrent in the infra-red and visible regionsand the peaks in the observed photocurrent spectra havebeen related to electronic transitions between van Hovesingularities in their electronic density of states [3]. Thisis consistent with energy loss spectra calculations performedfor single and multiwall carbon nanotubes, predicting thepresence of interband electronic transitions due to the reduceddimensionality of these systems [1].

0957-4484/11/115701+08$33.00 © 2011 IOP Publishing Ltd Printed in the UK & the USA1

Nanotechnology 22 (2011) 115701 P Castrucci et al

A few years ago we showed the ability of MWCNTsto generate photocurrent in an electrochemical cell with anincident photon to current efficiency (IPCE) higher [6] thanthat measured for single wall carbon nanotubes (SWCNTs) byKamat and co-workers [7]. This paved the way to consideringthe electronic properties of MWCNTs different from those ofgraphite, which does not exhibit any photocurrent effect. In thesame period and in the following years, other papers showedthat MWCNTs can generate photocurrent as well [8–15].Moreover, a few works have been devoted to investigating theability of single or multiwall carbon nanotubes to serve asenergy conversion material and semitransparent electrode incarbon nanotube–silicon heterojunctions built in the traditionalvertical Si-solar cells scheme [14–16]. Nonetheless, themechanism of carrier photogeneration in MWCNTs andtransport still remain an open question.

In this paper, we report on a relatively simple photovoltaicdevice with a planar architecture where the MWCNTs weredirectly grown onto n-Si substrate. We found that this deviceshows an IPCE of 34% at 670 nm, about one third of thatcollected in commercial poly-crystalline silicon solar cells andimproves by a factor of 104 the bare Si substrate performance.The IPCE trend as a function of the incident photon wavelengthresembles that of the silicon solar cell, extending over thewhole visible spectrum. We also report that MWCNTsgrown on a quartz substrate still exhibit photocurrent, thoughwell below that for MWCNTs grown on a silicon substrate.These results suggest that (a) MWCNTs are able to generatephotocurrent and (b) the silicon substrate plays an unexpectedfundamental role in our planar device. We interpreted thislatter finding as due to the action of the MWCNT film asboth photocurrent generator and semitransparent electrode forthe silicon substrate underneath. Finally, to understand theformer indication, namely why MWCNTs are able to generatephotocurrent, we performed ab initio calculations of theirelectronic properties. We found that the density of states ofMWCNTs can be well reproduced by a superposition of thoseof the single wall carbon nanotubes constituting the MWCNTstructure, each maintaining its one-dimensional features inthe electronic and transport properties. Moreover, metallicMWCNTs have been found to have rather long Thomas–Fermistatic screening length that may give rise to a non-perfectscreening of the e–h pairs generated within the nanotubes uponillumination, thus allowing for photo-induced currents even inmetallic MWCNTs. Since van Hove singularities, responsiblefor the e–h pairs generation, are not present in graphite (or ingraphene sheets) the carbon nanotubes can be considered asbuilding blocks for photosensors and photovoltaic devices.

2. Experimental methods

Carbon nanotubes were grown on an n doped Si(001)substrate (resistivity ρ = 0.1–1.0 � cm−1), covered atroom temperature by Fe catalyst nanoparticles and held at720 ◦C, by dynamical chemical vapour deposition in C2H2

atmosphere (200 sccm, p = 12 Torr) for 10 min. TheSi(001) substrate was degreased before insertion in an ultra-high vacuum chamber and the native oxide was not removed

before Fe nanoparticle deposition. The same procedurehas been applied to grow carbon nanotubes on a quartzsubstrate. The Fe nanoparticles were obtained by evaporating1 nm of iron from a tungsten crucible in ultra-high vacuumon the substrate kept at room temperature. The synthesismethod has been used elsewhere [6], and its peculiarityconsists in exposing Fe nanoparticles to acetylene withoutbreaking the vacuum conditions. This method prevents Fenanoparticles from oxidation and avoids the use of othergases, such as ammonia, to deoxidize the iron nanoparticlesbefore acetylene exposure. It gives rise to an almost uniformdistribution of Fe nanoparticles whose diameter depends onthe nominal thickness of the deposited Fe. As a consequence,the carbon nanotubes’ inner diameter is strongly affectedby the size of the obtained Fe nanoclusters. No post-growth purification has been performed. Optical absorbancemeasurements were performed by using a Perkin-Elmer,Lambda 35, UV/vis spectrometer. Transmission electronenergy loss spectroscopy (EELS) measurements were carriedout in a TEM apparatus operating at an accelerating voltageof 120 kV and equipped with an electron energy filter.The energy resolution of the spectra was estimated to beabout 0.8 eV measuring the full width at half-maximum ofthe elastic electron peak. More details on the experimentwere reported elsewhere [6, 17]. The photocurrent spectrawere measured using an optical set-up made up of axenon lamp equipped with a monochromator, focusing andcollecting optics, a reflecting chopper and lock-in electronics.Metallic electrodes were deposited directly on the randomnetwork of MWCNTs (figure 1(b)). The light spot size was(0.5 × 3) mm2. The photocurrent density, I (λ), wasmeasured under illumination as a function of the incidentphoton wavelength, λ. The incident photon power densitywas monitored with a calibrated silicon photodiode and datawere collected by a lock-in technique. The incident photonsto current efficiency is defined as the fraction of the incidentphotons, Nph, converted into photocurrent, i.e. the number ofthe generated e–h pairs, Ne–h, multiplied by the electroniccharge, e. The number of the incident photons is then evaluatedin terms of the power density of the Xe lamp, P(λ), sinceNph = λP(λ)/hc. Therefore, it results

IPCE (%) = electrons

photons= 100 hcI (λ)

eλP(λ).

I (λ) was measured by modulating the light by anoptical chopper and recovering the amplified current signal(converted to voltage) by a lock-in amplifier locked on thechopper frequency and the lamp power P(λ) was measuredsimultaneously in a similar way by a calibrated Si photodiode.Interestingly, the power density of our Xe lamp ranges from 0.5to 3 mW cm−2 over the investigated spectral region. Finally,using a Keithley 2602A sourcemeter, we recorded (a) thecurrent by positioning the light spot at variable distance fromthe metallic contact electrode, S, from which the signal ismeasured and to which the bias can be applied (figure 1(b))both in dark and under white light illumination; (b) the current–voltage curves.

2

Nanotechnology 22 (2011) 115701 P Castrucci et al

Figure 1. (a) Scheme of the device used to perform photocurrentmeasurements. Note the planar geometry of the metallic contacts. Ifno bias is applied to the sample, the T1 switch is open and the T2 isclosed, to exclude the voltage generator. (b) SEM image of theMWCNT random network over which the metallic electrodes havebeen deposited.

3. Results and discussion

3.1. Photocurrent measurements

Figure 1(a) illustrates the scheme of the experimental set-upused for photocurrent measurements both without and withbias application. Figure 1(b) displays the scanning electronmicroscopy (SEM) image of the nanotube sample, showingthe presence of randomly distributed nanotubes on the Sisurface with a very small amount of other carbonaceousproducts. High resolution transmission electron microscopy(HR-TEM) images (not shown here) were used to detectthe nanotube multiwalled nature and their good crystallinity.Figure 2(a) displays the photoresponse of the MWCNT/Sisample acquired at null applied voltage and with the light spotat the interface between the MWCNTs and the S electrode.Upon irradiation with visible and near ultraviolet light, weobserved photocurrent generation. In particular, the inset offigure 2 shows typical on–off cycles, these ones were recordedat λ = 500 nm. The presence of a squared and sizeable darklight signal reveals that a true photovoltaic effect occurs. Westress that a much lower (four orders of magnitude) IPCE wasmeasured from the bare n-Si substrate (figure 2(b)), where twometal electrodes were deposited on the surface so to have thesame planar configuration as our MWCNT/n-Si device. Thisresult suggests that the measured photoresponse is basicallyinduced by the presence of carbon nanotubes. The present

Figure 2. (a) IPCE of the MWCNT random network on Si substrate(black dots); inset: on–off cycles recorded at λ = 500 nm; (b) IPCErecorded from a bare n-Si substrate, note that the IPCE has beenmultiplied by a factor of 103; (c) IPCE of a commercial silicon p–nsolar cell collected in the conventional up–down configuration (blackdiamonds).

IPCE spectral response is different from that recorded fora MWCNT network in an electrochemical cell [6], whichincreases by decreasing the impinging radiation wavelength,exactly as the MWCNT optical absorbance. Figure 2(c)reports the IPCE spectrum collected on a commercial siliconp–n solar cell obtained measuring the photocurrent in theconventional vertical geometry. It is worth noting that itresembles a thick silicon slab optical absorptance and itis limited in the ultraviolet region because of the surfacerecombination. The maximum IPCE for our MWCNT/Sidevice amounts to 34%, about one third of the signal measuredin a conventional silicon solar cell. Moreover, the MWCNT/SiIPCE spectrum (figure 2(a)) extends over all the visible andnear ultraviolet wavelength range (down to 300 nm) andshows a maximum located at about 670 nm. It is alsoworth noting that the wavelength region is very similar tothat measured for the poly-crystalline silicon p–n solar cellthough the latter one exhibits a marked decrease at 350 nm.These are very intriguing results. In fact, since the silverpaint electrodes do not touch directly the silicon substrate andwe are dealing with a planar device, one should expect thatthe Si substrate would not be involved in the photocurrentgeneration process. The silicon substrate, instead, looks

3

Nanotechnology 22 (2011) 115701 P Castrucci et al

to play a fundamental role. The substrate role has beenconfirmed by other experiments we performed by coveringGe or GaAs slabs with MWCNT films (not shown here)that evidenced an IPCE spectral behaviour not far from theoptical absorptance of Ge and GaAs, respectively. Moreover,we succeeded in measuring a sizeable photocurrent intensityfrom a MWCNT film coating a quartz insulating substratewhen illuminated by a red laser spot (power density of about100 mW cm−2) but we obtained no detectable photocurrent byusing our monochromatized low power density Xe lamp. Thismeans that MWCNTs can generate a photocurrent but witha rather low efficiency. All these experiments indicated thatsome other mechanism involving MWCNTs and the substratemust be invoked to interpret the observed IPCE. Recently,similar observations have been reported in case of severalthicknesses of SWCNT/Si films arranged in planar [18] orvertical configuration devices [19] and interpreted in terms ofthe ability of the SWCNT random network to behave not onlyas photocurrent generator but also as semitransparent electrodeon the semiconducting substrate. Indeed, at the two materialinterface a large number of carbon nanotube/Si heterojunctionsare present which can allow the carriers photogenerated in thesilicon to participate to the acquired photocurrent. In thisscenario, the thickness of the carbon nanotube film plays afundamental role because it must be thick enough to havethe highest number of heterojunctions and thin enough tonot absorb too much the impinging light. Therefore, wecan conclude that in the present experiment the MWCNTfilm thickness is thin enough to allow the impinging lightto be partially transmitted into the silicon slab and generatee–h pairs. These pairs are split at the many MWCNT/Siheterojunctions (MWCNTs are mostly metallic in character)and MWCNTs transport carriers towards the electrodes. In acertain sense, we can say that the MWCNT random networkacts as a semitransparent electrode for silicon substrate.However, the extension of the IPCE spectral range down to300 nm could be ascribed to the e–h pairs generated intoMWCNTs themselves, separated at the MWCNT/MWCNTjunctions or at the MWCNT/Si heterojunctions and transportedby MWCNTs at the metal electrodes. Finally an IPCEmaximum at 670 nm and decay around 750 nm have beenalso observed for the same device contacted in a verticalconfiguration (not shown here) and could be related to the lowdiffusion length of the carriers into the n doped silicon, so thattheir recombination probability before reaching the electrodeis very high.

Interestingly, we observed that the photocurrent signalstrongly depends on the light spot position but the IPCEspectra show the same trend as a function of the wavelength,though with reduced intensity values. When the light spotwas close to the S electrode a decrease in the photocurrentsignal was measured with respect to the dark current value.However, when the light spot was close to the G electrode anincrease in current signal was recorded with respect to the darkcurrent value. This indicates that the photocurrent exceededthe dark value, changed sign and flowed in the ‘reverse’direction, though no bias was applied. When the sample isilluminated in the middle of the two electrodes, the current

Figure 3. Dynamic photoresponse upon white beam illumination(a) at a chopper frequency of 1 Hz; (b) upon manual lightswitching-on at the G electrode. (c) Dynamic photoresponse of theMWCNT random network grown on a quartz substrate upon whitebeam illumination at the S electrode after manual light switching-onand associated fitting (red solid curve). The time t0 of lightswitching-on was arbitrarily set to zero.

signal amplitude is strongly reduced. In figure 3(a) we reportthe dynamic photoresponse upon white beam illumination(chopper frequency of 1 Hz). It is remarkable that, duringillumination, there is initially a quite fast response and thesignals recorded illuminating the S or G electrodes havedifferent absolute current values. This latter effect can berelated to a geometrical inequality of the two electrodesand/or to a contact asymmetry of the two metal–nanotubejunctions [20–22], which could also be the origin of the non-negligible current revealed when the light spot is in between thetwo electrodes6. To better analyse these signals, we measuredthe dynamic response of the photocurrent when switching on(figure 3(b)) and off the light manually. The experimental datafit, in a first approximation, the exponential form I = I0(1 −exp(−(t − t0)/τ )) and I = I0(exp(−(t − t0)/τ )), respectively,where τ is the time constant, t0 is the time when the light isswitched on or off and I0 is the steady state photocurrent. Thisbehaviour is typical of the carriers’ diffusion due to gradient

6 In the case of a symmetric Schottky junction at two metal–nanotubeinterfaces, the current recorded in the middle of the two electrodes should bezero. An eventual current is the result of the light spot’s finite size and itsmanual positioning, giving rise to an imbalance in the charges flowing.

4

Nanotechnology 22 (2011) 115701 P Castrucci et al

charge concentration [23–26]. The time constant was found tobe (0.024±0.002) s. If the connections of the electrodes to theinstruments are exchanged, a negative (positive) current is stillobserved when the light spot is close to the S (G) electrode.This indicates that a built-in potential does exist, even withoutapplied bias, between the electrodes and the MWCNT/Sisystem under illumination. This effect is similar, though with areversed sign of the photocurrent, to those observed in the caseof highly entangled carbon nanotube networks on insulatingsubstrates [23–26]. In addition, the time constants arerather small compared to those reported for carbon nanotubenetworks for similar electrode distances [23–26]. However,if we perform similar measurements on a MWCNT randomnetwork deposited on a quartz substrate, we find that thephotocurrent sign behaviour is similar to that measured forsingle and multiwall carbon nanotubes on insulating substrates.Moreover, the dynamic photoresponses as a function of time fitvery well the exponential equations shown above (figure 3(c))with time constant τ = 0.60 ± 0.02 s for electrode spacingof 3.5 ± 0.3 mm; these values are in perfect agreement withthose reported for MWCNTs [23]. Thus, what we measure isa genuine photocurrent produced by MWCNTs. It is worthnoting that the maximum photocurrent intensity value, ISC, forcarbon nanotubes on quartz substrate is about 22 nA, muchsmaller than that measured for MWCNTs on Si substrate (byabout four orders of magnitude). In this case the photocurrentgeneration occurs completely through the MWCNT randomnetwork, since the quartz substrate, due to its large band gap,does not participate either in the e–h pair generation or in theirseparation and transport. Therefore, measurements performedby coating a quartz substrate with MWCNTs unambiguouslydetermine the amount of photocurrent that MWCNTs areable to generate. These findings confirm that the IPCEmeasurements reported in figure 2(a) for the MWCNT/Sisystem must necessarily strongly involve the silicon substratein the process. The role of the silicon substrate is evenmore evident when measuring current–voltage curves collectedwith a bias at the S electrode. The MWCNTs grown on thequartz substrate (figure 4(a)) show a linear behaviour with nochanges in conductance going from dark to light, as reportedby Wei et al for a MWCNT bundle suspended between twoelectrodes [12], suggesting that the multiwall carbon nanotubeshave metallic character. The open circuit voltage, VOC, is verysmall −36 μV and 43 μV, while the short circuit current,ISC, is 22 nA and −30 nA, in the case of S and G electrodeillumination, respectively. On the other hand, the MWCNTs onSi present non-linear current–voltage behaviour (figure 4(b)),giving rise to an inversion of the photocurrent sign at certainapplied bias and a VOC of 0.03 V and −0.01 V for S and Gelectrode illumination, respectively.

3.2. Four wall carbon nanotubes electronic DOS ab initiocalculations

The fact that MWCNTs coating an insulating quartz substratehave been demonstrated to generate a photocurrent isparticularly intriguing. Indeed, because of their intrinsicmultiwall nature, they should mimic the electronic behaviour

Figure 4. Current–voltage curves collected at dark (black curves)and under white irradiation impinging on the S (blue dotted curves)and G (red dash–dot–dot curves) electrodes, for MWCNTs on quartzsubstrate (a) and MWCNTs on Si substrate (b). Note the differentvoltage scales in (a) and (b).

of graphite that is known not to generate photocurrent. Tounderstand this important point, ab initio calculations havebeen performed within density functional theory (DFT) usingthe Dmol3 code [27] for single as well as multiwalled carbonnanotubes. We pointed our attention to metallic multiwallcarbon nanotubes because it is well known that in metalsexciton formation is not allowed or excitons rapidly recombinebecause of electronic screening. Moreover, for computationaltime reasons, we limited ourselves to a four wall carbonnanotube. On the other hand, we expect that MWCNTs witha high number of walls, which certainly are present in ourcarbon nanotube film, have an electronic behaviour similar tographite, and this could be one of the reasons for the observedvery low photocurrent generation efficiency of the MWCNTrandom network. It is worth stressing that this calculation hasbeen performed, independently of the substrate used, for an

5

Nanotechnology 22 (2011) 115701 P Castrucci et al

Figure 5. Electronic DOS calculated for a graphene sheet (a), a (10,10) SWCNT (b), a (5, 5) (10, 10) DWCNT (c) and a (5, 5) (10, 10)(15, 15) (20, 20) 4WCNT (d).

individual MWCNT in order to study the peculiar electroniccharacteristics of this low-dimensional system.

The calculated densities of states (DOS) are reportedin figure 5 for an arbitrary set of single, double and fourwall (4WCNT) carbon nanotubes. Moving from the simplegraphene sheet (lower panel) towards the single wall (10,10) nanotube, more complex structures appear mainly due toband-folding effects (i.e. the graphene band structure foldedinto the smaller Brillouin zone of the nanotube). As weconsider additional walls, that is, as we add larger nanotubeson top of the first one, this effect is amplified and we canobserve an increasing number of singularities (peaks), resultingfinally in a smoother DOS with larger values. This canbe understood by taking into account two main effects: thefirst one, as mentioned before, is related to band-foldingsince the 2D Brillouin zone shrinks as the wall diameter isincreased. The second one is related to the increased numberof nanotubes (walls) that smooths out the huge singularitiespresent in each SWCNT DOS. Moreover, the total densityof states of the 4WCNT structure is very close to the baresum of the DOS of each SWCNT composing the structure(figure 6) showing that there is a rather small interactionamong nanotubes belonging to the same multiwalled structure.The large van Hove singularities on both sides of the Fermilevel (Ef) together with the very small density of states at Ef

may originate the photogenerated e–h pair in MWCNTs. In

Figure 6. (a) Comparison between the electronic DOS calculated fora (5, 5) (10, 10) (15, 15) (20, 20) 4WCNT and for the sum of the fourSWCNTs constituting the 4WCNT; (b) difference between thecalculated electronic DOS for a (5, 5) (10, 10) (15, 15) (20, 20)4WCNT and for the sum of the four SWCNTs constituting the4WCNT.

addition, the calculation shows that as the walls are added,the density of states increases by about a factor of 5. Thismeans that each multiwall contributes, in principle, to thephotocurrent much more than a single wall. This result hasbeen confirmed experimentally using an electrochemical cell:the MWCNTs showed an IPCE far higher than that observedfor SWCNTs in a similar experimental configuration [6, 7].Finally, these results clearly indicate that multiwall carbonnanotubes, although with a rather small number of walls,present different electronic properties from bulk graphite sheetsystems. Figure 7 reports the transmission EELS comparingthe spectrum recorded on an individual MWCNT with thoseof highly oriented pyrolytic graphite (HOPG) very thin flake.Both HOPG and MWCNT spectra show two typical featuresdue to π and σ + π plasmons; however, the MWCNTstructures are shifted towards lower energy with respect tothose of graphite (located at 7 and 28 eV) indicating that theelectronic properties of MWCNTs are not equivalent to thoseof graphite. Moreover, it has been reported that the energydownshift of such plasmons increases as the number of wallsconstituting the individual MWCNT probed [28] is reduced,while theoretical calculations assessed that this effect increasesas the intertube or intratube interactions decrease in bundlesof SWCNTs and MWCNTs, respectively [1, 29, 30]. This isalso consistent with our DFT calculations. In addition, theMWCNT spectrum presents a shoulder at energies 2–4 eV,well below the typical plasmon π peak for HOPG. In thecase of SWCNTs, such transitions have been reported andinterpreted as excitations between localized electronic states

6

Nanotechnology 22 (2011) 115701 P Castrucci et al

Figure 7. Transmission electron energy loss spectra measured bycollecting inelastic electrons from a HOPG very thin flake and aMWCNT. Note the shoulder in the MWCNT spectra at energieslower than the plasmon π peak typical of the graphitic systems.

related to van Hove singularities [31]. These transitions havebeen considered as being responsible for the photocurrentgeneration in SWCNTs since they produce electron–hole pairsunder illumination. In the case of MWCNTs (number of wallsless than 22) this shoulder should be associated to the samekind of singularities that are still present in the electronicdensity of states of MWCNTs. A complete calculation of theseenergy loss features as a function of the increasing number ofwalls is reported in [1].

From the density of states at the Fermi level we cancalculate the Thomas–Fermi static screening length, λTF, forthe MWCNTs: we obtain λTF ranging between 15 and 50 A(14.8 A for the one reported in figure 6) corresponding toa carrier density at the Fermi level of the order of 9 ×1020 electrons cm−3, significantly smaller than the usual carrierdensities in bulk metals (about 1022–1023 electrons cm−3).These findings imply that each e–h pair created within thenanotube is far less screened than in a typical metal, thusexplaining why even metallic nanotubes may give rise tophotocurrent generation [32]. On the other hand, graphite,though having a carrier density at the Fermi level similar to thatof MWCNTs, does not show any photocurrent generation. Thiscan be due to the absence of van Hove singularities inducingthe e–h pair generation.

Moreover, our first-principles calculations showed thatit is possible to consider a MWCNT as the sum of severalSWCNTs with rather large screening lengths, thus supportingthe idea that rather large local electric fields might exist on thesame nanotube. This last effect adds up to all the other effectsconsidered. Besides, the 4WCNT’s theoretical density of statespresents many van Hove singularities that should considerablyincrease the number of channels allowing electron–hole pair

generation. Among the many MWCNTs present in realsamples, some of them would surely be metallic: however, ourcalculation pointed out that the screening power at the Fermilevel is considerably reduced due to the rather low density ofstates at EF (about one–two orders of magnitude lower than theordinary metallic density). This would result in quite a long lifetime of the e–h pair even in the metallic state, as confirmed byphotocurrent generation from metallic SWCNTs, observed byMohite et al [3]. In fact, this work clearly demonstrates thepresence of the typical M11 transition of the metallic SWCNTsin the photocurrent response, in addition to the S11 and S22

transitions due to semiconducting nanotubes. On the otherhand, the weak metallic character of MWCNTs is confirmedby the Ohmic behaviour of the current–voltage curves reportedin figure 5(a) while their peculiar metallic character couldexplain why the Schottky barrier model looks to be a goodinterpretation of: (i) the photocurrent at null bias applied,(ii) the photocurrent dependence on the position of the lightspot position between the two electrodes, and (iii) the dynamicphotoresponse of carbon nanotubes grown on quartz insulatingsubstrate, as well.

4. Conclusions

In summary, we demonstrated that MWCNTs grown on a Sisubstrate are able to enhance the IPCE of the bare substrate byroughly a factor of 104, showing a maximum IPCE of 34%.We also show that MWCNTs covering a quartz substrate stillexhibit photocurrent, though well below than the MWCNTscoating the silicon substrate. These results suggest thatMWCNTs are able to generate photocurrent and that the siliconsubstrate plays a fundamental role in our planar device. Theformer effect is particularly interesting because MWCNTsare generally known to mimic the electronic properties ofgraphite, which does not present any photocurrent generation.Theoretical ab initio calculations showed that in principle:(1) the MWCNTs’ density of states is larger than that ofSWCNTs thus enhancing the photocurrent flow generatedin each nanotube; (2) MWCNTs do not behave as idealmetals, since they show a much lower carrier density atthe Fermi level; this gives rise to a longer recombinationlifetime of the e–h pairs, and it explains how the poormetallic character could make MWCNTs good photocurrentgenerators; (3) the electronic properties of MWCNTs aredefinitively different from those of graphite, as evidenced bylocal EELS measurements, revealing the presence of a shoulderbetween 2 and 4 eV ascribed to the transitions between vanHove singularities. The experimental evidences discussed inthis paper provide insights into the unique photoabsorptionproperties of multiwall carbon nanotubes and ask for furthertheoretical and experimental investigations to fully understandthe mechanisms of e–h charge separation and transport, anecessary requirement to develop the next generation of lightharvesting devices and photosensors. This work confirmsthat despite the low intrinsic photocurrent generation abilityof these tubular nanostructures, they could play an importantrole in view of their potential development into low costdevices, for example obtained by spraying them onto any

7

Nanotechnology 22 (2011) 115701 P Castrucci et al

semitransparent semiconductor substrate in order to obtainhybrid heterojunctions without the need for extended doping.

Acknowledgments

We are very grateful to Dr Stefano Casciardi (ISPESL,Monteporzio Catone, Italy) and Dr M Diociaiuti (ISS,Roma, Italy) for their fundamental support in the TEM andtransmission EELS measurements. This work was supportedby Iniziativa Calcolo per la Fisica della Materia at CINECA-Supercomputing center. The authors acknowledge the financialsupport of the Queensland Government Smart Futures FundNational And International Research Alliances Program ‘SolarPowered Nano-Sensors For Data Acquisition And SurveyingIn Remote Areas’, ABN 83 791 724 622.

References

[1] Shyu F L and Lin M F 2000 Phys. Rev. B 62 8508–16[2] Kwon Y K and Tomanek D 1998 Phys. Rev. B 58 R16001–4[3] Mohite A, Sumanasekera G U, Hirahara K, Bandow S,

Iijima S and Alphenaar B W 2005 Chem. Phys. Lett.412 190–4

[4] Sun J L, Wei J, Zhu J L, Xu D, Liu X, Sun H, Wu D-H andWu N-L 2006 Appl. Phys. Lett. 88 131107

[5] Wei J et al 2007 Nano Lett. 7 2317–21[6] Castrucci P, Tombolini F, Scarselli M, Speiser E, Del Gobbo S,

Richter W, De Crescenzi M, Diociaiuti M, Gatto E andVenanzi M 2006 Appl. Phys. Lett. 89 253107

[7] Barazzouk S, Hotchandani S, Vinodgopal K and Kamat P V2004 J. Phys. Chem. B 108 17015–8

[8] Passacantando M et al 2008 Appl. Phys. Lett. 93 051911[9] El Khakani M A, Le Borgne V, Aissa B, Rosei F, Scilletta C,

Speiser E, Scarselli M, Castrucci P and De Crescenzi M2009 Appl. Phys. Lett. 95 083114

[10] Scarselli M, Scilletta C, Tombolini F, Castrucci P,De Crescenzi M, Diociaiuti M, Casciardi S, Gatto E andVenanzi M 2009 Superlatt. Microstruct. 46 340–6

[11] Scarselli M, Scilletta C, Tombolini F, Castrucci P,Diociaiuti M, Casciardi S, Gatto E, Venanzi M andDe Crescenzi M 2009 J. Phys. Chem. C 113 5860–4

[12] Wei J, Sun J L, Zhu J L, Wang K, Wang Z, Luo J, Wu D andCao A 2006 Small 2 988–93

[13] Sarker B K, Arif M, Stokes P and Khondaker S I 2009 J. Appl.Phys. 106 074307

[14] Li Z, Kunets V P, Saini V, Xu Y, Dervishi E, Salamo G J,Biris A R and Biris A S 2009 ACS Nano 3 1407–14

[15] Jia Y, Li P, Wei J, Cao A, Wanga K, Li C, Zhuang D,Zhu H and Wu D 2010 Mater. Res. Bull. 45 1401–5

[16] Hatakeyama R, Li Y F and Kaneko T 2010 Appl. Phys. Lett.97 013104

[17] Castrucci P, Tombolini F, Scarselli M, Scilletta C,De Crescenzi M, Diociaiuti M, Casciardi S, Rosei F andEl Khakani M A 2009 J. Phys. Chem. C 113 4848–55

[18] Le Borgne V, Castrucci P, Del Gobbo S, Scarselli M,De Crescenzi M, Mohamedi M and El Khakani M A 2010Appl. Phys. Lett. 97 193105

[19] Jia Y et al 2008 Adv. Mater. 20 4594–8[20] Fuhrer M S et al 2000 Science 288 494–7[21] Lee J U 2005 Appl. Phys. Lett. 87 073101[22] Freitag M, Martin Y, Misewich J A, Martel R and

Avouris Ph 2003 Nano Lett. 3 1067–71[23] Lu S and Panchapakesan B 2006 Nanotechnology 17 1843–50[24] Lien D H, Hsu W K, Zan H W, Tai N H and Tsai C H 2006

Adv. Mater. 18 98–103[25] Merchant C A and Markovic N 2008 Appl. Phys. Lett.

92 243510[26] Liu Y, Lu S and Panchapakesan B 2009 Nanotechnology

20 035203[27] Delley B 1990 J. Chem. Phys. 92 508–17

Delley B 2000 J. Chem. Phys. 113 7756–64[28] Bursill L A, Stadelmann P A, Peng J L and Prawer S 1994

Phys. Rev. B 49 2882–7[29] Kramberger C et al 2008 Phys. Rev. Lett. 100 196803[30] Marinopoulos A G, Reining L, Rubio A and Vast N 2003 Phys.

Rev. Lett. 91 046402[31] Pichler T, Knupfer M, Golden M S, Fink J, Rinzler A and

Smalley R E 1998 Phys. Rev. Lett. 80 4729–32[32] Mohite A, Chakraborty S, Gopinath P, Sumanasekera G U and

Alphenaar B W 2005 Appl. Phys. Lett. 86 061114

8