Solar Energy Materials & Solar Cells 117 (2013) 624–631

Contents lists available at ScienceDirect

Solar Energy Materials & Solar Cells

0927-02

http://d

n Corr

E-m

journal homepage: www.elsevier.com/locate/solmat

Direct photocurrent generation from nitrogen doped TiO2 electrodesin solid-state dye-sensitized solar cells: Towards optically-activemetal oxides for photovoltaic applications

Hussein Melhem a, Pardis Simon b, Jin Wang b, Catherine Di Bin a, Bernard Ratier a,Yann Leconte b, Nathalie Herlin-Boime b, Malgorzata Makowska-Janusik c,Abdelhadi Kassiba d, Johann Boucle a,n

a XLIM UMR 7252, Universite de Limoges/CNRS, 87060 Limoges Cedex, Franceb IRAMIS/SPAM/LFP, CEA-CNRS URA 2453, CEA de Saclay, 91191 Gif sur Yvette, Francec Institute of Physics, Jan Dlugosz University, Al. Armii Krajowei 13/15, 42-200 Czestochowa, Polandd Institut des Molecules et Materiaux du Mans, CNRS 6283, Universite du Maine, 72085 Le Mans Cedex 9, France

a r t i c l e i n f o

Available online 29 September 2012

Keywords:

Solid-state dye-sensitized solar cell (DSSC)

TiO2 Nanocrystals

Nitrogen doping

Laser pyrolysis

Electron paramagnetic resonance (EPR)

Density functional theory (DFT)

48/$ - see front matter & 2012 Elsevier B.V. A

x.doi.org/10.1016/j.solmat.2012.08.017

esponding author. Tel.: þ33 5 87 50 67 62; fa

ail address: johann.boucle@unilim.fr (J. Boucl

a b s t r a c t

Nitrogen-doped titanium dioxide (TiO2) is considered as a promising photocatalytic material due to its

optical absorption extended in the visible region compared to pure TiO2. In the field of photovoltaic

applications, dye-sensitized solar cells based on N-doped nanocrystalline titania electrodes have

demonstrated improved performance due to the beneficial effects of nitrogen on the electronic and

optical properties of TiO2. In this context, we report on the influence of nitrogen doping on the

performance of solid-state dye-sensitized solar cells, starting from TiO2 and N-TiO2 nanocrystals

synthesized by laser pyrolysis. Using an integrated approach based on experimental and theoretical

investigations, the relationship between the local electronic features of the starting metal oxide

materials and device operation is described. We demonstrate that the short-circuit current density of

the solar cells based on an N-doped TiO2 electrode increases by more than 10% compared to that of pure

anatase. This improvement is clearly associated with the extended absorption of the doped electrode,

suggesting that alternative charge generation mechanisms occur in the cells in addition to the

conventional dye absorption. Computer simulations on isolated nanoclusters, as well as electron

paramagnetic resonance (EPR) experiments, confirm that nitrogen atoms in the presence of oxygen

vacancies can explain the introduction of additional energy states near the valence band of TiO2.

Surface states associated with nitroxide radicals are also suggested to act as charge traps under

illumination. These aspects confirm the strong potentialities of optically-active metal oxides for

photovoltaic applications.

& 2012 Elsevier B.V. All rights reserved.

1. Introduction

Dye-sensitized solar cells (DSSC) recently demonstrated powerconversion efficiencies over 12% [1,2], placing them as a relevantlow cost alternative to second generation solar cells such asinorganic thin film technologies. Whereas champion cells are mainlybased on liquid electrolytes, solid-state DSSC that use solid electro-lytes also appear as promising photovoltaic systems due to easy cellassembly and improved stability [2,3]. Efficiencies as high as 7%, andeven up to 10%, have been recently demonstrated using respectivelyorganic hole transporters such as spiro-OMeTAD (2,2’,7,7’-tetrakis-(N,N-dip-methoxyphenylamine)-9,9’-spirobifluorene) [4] and

ll rights reserved.

x: þ33 5 55 45 76 49.

e).

inorganic compounds such as CsSnI3 [5]. Many aspects have been,and are still, considered to improve device performance, the mainstrategies focusing on: extending the light-harvesting region intothe near infrared using new strongly-absorbing dyes; lowering theredox potential of the electrolyte to increase the open-circuit voltage(VOC); replacing solid-state molecular glass electrolytes by light-absorbing p-conjugated polymers; modifying the TiO2/dye/electro-lyte interface; using light-trapping or co-sensitization strategies [2].Moreover, important research efforts have also been paid to identifyalternative metal oxides or alternative electrode architectures [6].However, anatase TiO2 processed from colloidal pastes remainsthe material of choice to achieve the best performance up to now. Inthe context of the nanocrystalline electrode, the demonstration ofnovel device concepts that may exploit additional charge genera-tion mechanisms will be required in order to allow higher deviceefficiencies.

H. Melhem et al. / Solar Energy Materials & Solar Cells 117 (2013) 624–631 625

In this context, several studies demonstrated that doping metaloxides using metals [7–12] or non-metal elements [13–18] can be afruitful strategy to improve photocurrent generation. In particular,Lindgren et al. demonstrated the appearance of additional statesnear the valence band of TiO2 by the incorporation of nitrogen,leading to visible light absorption [19]. Although the energy statesintroduced by nitrogen in the gap can also act as recombinationcenters for the generated photo-carriers, the photo-electrochemicalproperties of the doped electrodes clearly indicated additionalphotocurrent generation in the blue part of the spectrum comparedto un-doped electrodes. In 2005, Ma et al. demonstrated a liquidDSSC based on N-doped titania powders presenting an improvedpower conversion efficiency of 8% and an improved stability [13].Again, an improved incident photon to current efficiency (IPCE) inthe 400–532 nm region was observed, which was attributed to theinfluence of nitrogen on the optical band gap of the photo-electrode.Such benefits of doping by nitrogen were also evidenced in the fieldof photo-catalysis [20,21], starting to significantly raise the interestof researchers for metal oxide materials active in the visible. Sincethen, various strategies have been reported to reduce the opticalband gap of TiO2 for both photovoltaic applications [10,15,22–25]and photo-catalysis [26–30]. Among all these strategies, doping bynitrogen is probably one of the most reported [13,24,25,31].Although improved solar cell performance is usually reported usingN-doped TiO2 electrodes, systematic studies have not yet beenundertaken in order to clarify the exact charge transfer mechanismsoccurring in the active layer. Especially, current generation resultingfrom the absorption of photons by the doped metal oxide electroderemains to be described, as well as the influence of nitrogen dopingon solid-state DSSC operation.

In this work, we investigate the use of nitrogen-doped TiO2

electrodes based on nanocrystals synthesized by laser pyrolysisfor solid-state dye-sensitized solar cells. These nanocrystals,which have been recently demonstrated to be a relevant alter-native to commercial pastes [32], take benefit from a versatilesynthesis method that allows easy doping procedures [33]. Deviceperformance is discussed with regard to the influence of nitrogenon the nanocrystal local structure using both electron paramag-netic resonance analysis and numerical simulations. It is worthnoting that these simulations are for the first time performedusing density functional theory (DFT) based on quantum chem-istry codes on isolated TiO2 clusters with and without nitrogen.The improved photo-current generation observed in the presenceof nitrogen confirms that optically-active metal oxides are ofparticular interest in the field of hybrid photovoltaics.

2. Experimental procedure

2.1. Synthesis of TiO2 and N-TiO2 nanocrystals

Laser pyrolysis of liquid titanium tetraisopropoxide (TTIP) wasused to produce both pure and nitrogen-doped titanium oxidenanocrystals [33]. A TTIP aerosol was produced using an ultra-sonic spraying process (Pyrosol) and carried in the reaction zoneby a carrier gas where it crossed the CO2 laser beam (1 kW). Someamount of C2H4 was introduced in the reactant mixture due to thepoor absorption of the infrared laser radiation by TTIP. Thedecomposition of precursors in the laser beam is followed byparticle growth which is rapidly stopped when the mixture exitsthe reaction zone, leading to the formation of nanoparticles whichare collected downstream on collection filters. N-doped particleswere formed in a similar manner, except that gaseous ammonia(NH3) was used as source of N atoms within the reactant mixture,and no C2H4 was required in this case. After the synthesis, thenanopowders are annealed in air at 400 1C for 3 h in order to

remove free carbon atoms that result from the decomposition ofTTIP and/or C2H4.

The morphology of the particles was studied by transmissionelectron microscopy (TEM) with a Philips CM12 (120 kV) instru-ment. Diffuse reflectance spectra of the powders were measuredon a UV–vis-NIR spectrophotometer (Jasco V-570), which wasequipped with an integrating sphere assembly. A given amount ofnanopowder was uniformly pressed in a powder holder (Jasco)and placed in the sample holder on an integrated sphere for thereflectance measurements. The absorption coefficient, as well asthe optical energy gap of the samples was deduced from theabsorption data using the Kubelka-Munk equation [34].

2.2. Solar cells assembly and testing

Porous TiO2 and N-TiO2 electrodes were deposited by spin-coating on fluorinated tin oxide (FTO) substrates in ambientconditions from ethanol dispersions (nanopowders content 10wt%) containing ethyl-cellulose (EC), as already reported [32].After film deposition, a gradual sintering step up to 450 1C in airwas performed, as well as a conventional TiCl4 treatment, leadingto 2mm thick porous layers. A dense TiO2 layer of approximately150–300 nm was initially deposited by spray pyrolysis on the FTOsubstrate. The porous electrodes were then immersed at 80 1Covernight in a diluted solution of the commercial indoline dyeD102 (Mitsubishi Paper Mills, Japan) dissolved in an acetonitri-le:tert-butanol (1:1 volume ratio) mixture. The obtained dye-sensitized electrodes were rinsed and infiltrated by the molecularhole conductor spiro-OMeTAD (Merck KGaA, Germany) by spin-coating, using conventional recipes with lithium salt and tert-butylpyridine as additives. Gold top contacts were finally evapo-rated under vacuum at 10�6 mbar using shadow masks thatdefined two active areas per substrate (0.18 cm2 each).

Scanning electron microscopy (SEM) was performed on bareand infiltrated dye-sensitized porous TiO2 and N-TiO2 layers usinga JEOL MEB-FEG 7400F. UV-visible optical absorption spectrawere recorded by a SAFAS DES 200 spectrometer used in trans-mission mode. Current density–voltage (J–V) characteristics wererecorded in air using a calibrated Keithley 2400 source-measureunit, in the dark and under simulated solar emission (AtlasSolarconstant 575PV). The spectral mismatch between the emis-sion of the solar simulator and the global AM1.5G (ASTM G173-03) solar spectrum was taken into account using standardprocedures [35] and the solar simulator irradiance was correctedaccordingly to match 100 mW cm�2 on the tested cell. Theincident photon-to-current efficiency (IPCE) was measured usinga monochromated 75 W Xenon lamp (Newport) and a calibratedpicoamperemeter (Keithley 485) used in static regime in order toensure full collection of all photo-generated charge carriers. Thecalibration was performed using a calibrated silicon photo-detector of known spectral response (Newport).

2.3. Computer simulations of TiO2 and N-TiO2

The electronic properties of nanostructured (TiO2)n and(TiO2-xNx)n were calculated using the density functional theory(DFT) formalism following a cluster approach. The clusters wereconstructed by removing Ti and O atoms, from a large clusterobtained by expansion of the anatase crystal structure unit cell inall three dimensions. The anatase nanostructures were modeledwith hydrogen saturation atoms, without full geometry optimiza-tion. The single point calculations were performed using GAMESSprogram package [36] applying effective core potential basis setnamed Stevens-Basch-Krauss-Jasien-Cundari (SBKJC) [37–39].Our previous computational works proved that to reproduce theHOMO–LUMO energy gap splitting of the anatase structure, it is

H. Melhem et al. / Solar Energy Materials & Solar Cells 117 (2013) 624–631626

appropriate to use long range correlated LC-BLYP hybrid func-tional with generalized gradient approximation (GGA) [40]. TheLC functional smoothly replaces the DFT exchange by the Hartree-Fock (HF) exchange at long inter-electron distances. Using hydro-gen saturated (TiO2)17 clusters, the self-consistent field (SCF)procedure converges quickly and calculated HOMO–LUMO isclose to the energy gap of the bulk anatase structure. Thecalculated value of HOMO–LUMO energy gap splitting for mod-eled (TiO2)17 clusters is equal to 3.03 eV.

2.4. EPR experiments

Continuous-wave electron paramagnetic resonance (CW-EPR)experiments were performed on an EMX Bruker spectrometerworking in X-Band under controlled atmosphere (nitrogen, oxy-gen) and at controlled temperature from 4 to 100 K using anOxford Instruments cryostat. About 10 mg of the titanium dioxidenanopowders were stored in EPR quartz tubes under normalatmosphere or using partial pressures of oxygen, argon or nitro-gen. The EPR spectra were recorded in the dark or under UV-irradiation using a Xenon lamp with the possibility to filter the366 nm spectral line or to use the entire Xe lamp spectrum. TheEPR parameters (g-tensor components, hyperfine coupling) areextracted from the experimental spectra by a suitable fittingprocedure.

3. Results and discussion

3.1. Influence of nitrogen on the physical properties of TiO2

nanocrystals

The TiO2 and N-TiO2 nanopowders have been characterized byelemental analysis, transmission electronic microscopy (TEM),X-ray diffraction (XRD), and diffuse reflectance in order toevaluate their chemical composition, morphology, crystallinestructure and optical properties. Table 1 summarizes the corre-sponding parameters. TEM images and XRD patterns can be foundin the supplementary material provided with this article.

Both samples present mainly spherical nanoparticles withdiameters around 10 nm. In similar flow and laser power condi-tions, the synthesis in presence of NH3 leads to the formation ofparticles with smaller sizes compared to C2H4. This could beinterpreted in terms of catalytic decomposition of TTIP by NH3,thus increasing the number of germs [41]. In both cases, the maincrystallographic phase is anatase. The N-TiO2 powder is asso-ciated with a small amount of nitrogen (0.5 at%) which is foundto drastically modify the absorption onset of the nanopowders,leading to an optical band gap which is found to be red-shifted bymore than 1 eV with regard to the pure anatase powder. Suchobservations are consistent with results obtained by Asahi et al.on the optical properties of doped TiO2 materials [21], as well as

Table 1Main physical properties of the TiO2 and N-TiO2 nanopowders synthesized forthis study.

Nanopowder Mean

diametera

(nm)

Crystalline

phaseb

Nitrogen

contentc

(at.%)

Optical

band gapd

(eV)

TiO2 12.572.5 4 95% anatase – 3.2

N-TiO2 7.571.8 4 95% anatase 0.5 2.1

a Estimated from TEM images.b Estimated from X-Ray diffraction patterns.c Evaluated from chemical elemental analyses.d Estimated from Kubelka–Munk’s equation applied on diffuse reflectance

measurements (Fig. 1).

with experimental observations made on N-doped TiO2 materials[13]. We will discuss in more detail these aspects in the followingsections. Fig. 1 presents the optical properties of the nanopowderscompared to the AM1.5G standard solar irradiance.

In addition to potential changes in electronic and chargetransport properties of the metal oxides, the additional absorp-tion band introduced by nitrogen will lead to visible photonabsorption in the electrodes. The incorporation of nitrogenstrongly affects the optical properties of the TiO2 powder whichturns from white to a yellow color. In the next sections, we willtry to assess if these absorbed photons contribute to photo-current generation in solid-state DSSC, and in this case, whichprocesses are involved.

3.2. Influence of nitrogen on the performance of solid-state DSSC

Solid-state DSSC were processed starting from the TiO2 andN-TiO2 nanocrystals dispersed in ethanol solution in the presenceof ethyl-cellulose, using a recipe described in the experimentalsection, as well as in previous studies [32]. After film depositionand conventional sintering treatments, around 2mm thick porousTiO2 and N-TiO2 electrodes are obtained on FTO/dense TiO2

substrates. Fig. 2 presents scanning electron microscopy imagesof the resulting electrodes, showing that nanostructured pores ofcomparable morphology are obtained in both cases. The relativelysmaller mean diameter of the doped powder results in a relativelyless homogeneous morphology of the electrode, as no surfactantsare used during particle dispersion and film formation, althoughmore quantitative investigations are required to characterize theelectrode porosities.

The porous electrodes are subsequently sensitized by theorganic indoline dye D102 before being infiltrated by the spiro-OMeTAD solid-state hole transporter. The corresponding opticalabsorption coefficient is plotted in Fig. 3 before and after thesensitization and infiltration procedures. The pure electrodes arefound to be almost transparent, without any significant contribu-tion from scattering processes. The N-doped electrode alonedemonstrates a larger absorption coefficient than the pure TiO2

electrode for wavelengths lower than 400 nm, in accordance withthe additional optical features observed in the optical spectra ofthe nanopowders (inset of Fig. 3). After dye loading and electrodeinfiltration by the molecular hole transporter, two specific fea-tures can be observed between the un-doped and doped electro-des. First, a larger absorption coefficient is evidenced below400 nm for the N-doped sample. This region corresponds to both

Fig. 1. Optical properties of the TiO2 and N-TiO2 nanopowders with regards to the

AM1.5 G solar irradiance. The data are extracted from diffuse reflectance mea-

surements performed on the powders.

Fig. 2. SEM images of porous electrodes (2mm thick) processed from pastes based on the (a) TiO2 and (b) N-TiO2 nanopowders. Both electrodes are sintered and have been

treated by TiCl4 (see experimental section).

Fig. 3. Absorption coefficient of the FTO/dense TiO2/porous TiO2/D102/spiro-

OMeTAD active layers without and with nitrogen doping. The inset presents the

absorption coefficient of the bare TiO2 and N-TiO2 porous electrodes before

sensitization and infiltration procedures.

Fig. 4. Current density–voltage characteristics under 1 sun (AM1.5 G, 100 mW cm�2)

of solid-state DSSC based on TiO2 and N-TiO2 porous electrodes. The inset shows the

corresponding curves in the dark.

H. Melhem et al. / Solar Energy Materials & Solar Cells 117 (2013) 624–631 627

the absorption edge of the molecular glass spiro-OMeTAD and tothe intrinsic optical absorption of the additional band observedfor the doped electrode. Again, this observation is consistent withthe influence of nitrogen doping on the optical absorption featureof the starting nanopowders.

Secondly, the absorption by the dye between 425 and 600 nmis found to be enhanced by the nitrogen doping by 25%. Thisimprovement can be associated either with different electrodesurface area or with different behaviors for dye grafting betweenthe un-doped and N-doped TiO2 electrodes. Ma et al. reportedthat a better dye uptake is usually observed in N-doped titaniaelectrodes, mainly due to a larger surface area of the dopedelectrode compared to the bare TiO2 [13]. Also, the surfacepotential of the starting nanoparticles have direct influence onthe dye uptake properties [31]. The introduction of NH3 duringthe particle synthesis is likely to result in different electrostaticproperties of the particle surface with regard to pure titania(modification of the isoelectric point), leading to different rheo-logical behavior of the nanopowders during paste formulation.The isoelectric point is also a crucial parameter that drives dyeuptake [42]. All these aspects can explain the trend revealedthrough the optical properties of the sensitized electrodes.

Finally, a gold top contact is evaporated on the infiltrated dye-sensitized electrode, and the current–voltage characteristics inthe dark and under simulated solar emission (100 mW cm-2) aremonitored (Fig. 4).

In the dark, a faster current onset is observed for the deviceincorporating the nitrogen-doped electrode. This can be conse-quence of an efficient doping process which would enhance thecharge transport properties of the electrodes, leading to largercurrent in forward bias. Under illumination, the reference devicebased on pure TiO2 exhibits short-circuit current (JSC), open-circuit voltage (VOC) and power conversion efficiency (Z) of8.7 mA cm�2, 0.76 V and 3.4% respectively. The N-doped elec-trode clearly improves the photocurrent collection by more than10%, with an improved JSC of 9.7 mA cm�2, resulting in animproved overall efficiency of 3.6%. This increase is howeverlimited by a modest fill factor of 0.49 compared to 0.52 for thecontrol device. Such improvements in photocurrent and deviceperformance have been observed on multiple series of devices,confirming the beneficial effects of the nitrogen doping ondevice operation. At the opposite, the open-circuit voltageremains almost unchanged after doping with 0.75 V, whichsuggests that the conduction band of the metal oxide and theassociated electronic states are not significantly influencedby the introduction of N atoms [13,21]. The increase of photo-current with doping can result from several phenomena. It isadmitted that the incorporation of N in the TiO2 structure canhelp fill some oxygen vacancies naturally present in titania,leading to reduced trap state distributions and improved chargemobilities [16]. We will discuss such effects in the last sections ofthis article. One other possible mechanism is the direct

H. Melhem et al. / Solar Energy Materials & Solar Cells 117 (2013) 624–631628

conversion of solar photons into electrons by the doped electrode.In order to further characterize the influence of nitrogen on deviceperformance, the normalized IPCE spectra of the devices arepresented in Fig. 5.

A clear improvement in photocurrent generation for theN-doped TiO2 electrode is observed in the 375–400 nm regionof the solar spectrum. This feature is associated with the opticalmodifications introduced by nitrogen in TiO2, as discussed in theprevious sections. Thus, these results indicate that solid-stateDSSC based on nitrogen-doped titania can convert visible photonsinto electrons thanks to the metal oxide. Although the exactcharge generation paths remain to be fully investigated, they arelikely to involve either the dye and hole transporter together, orthe hole transporter alone (Fig. 6). A direct energy transfer fromthe metal oxide to the dye is also possible. Further investigationsto rationalize the charge or energy transfer processes occurringin the device are now under progress. In any case, these results

Fig. 5. IPCE spectra normalized at 500 nm for the solid-state DSSC based on TiO2

(black solid circles) and N-TiO2 (blue solid squares) porous electrodes.

Fig. 6. Schematic representation of proposed charge transfer processes occurring

in a solid-state DSSC based on an optically-active N-TiO2 porous electrode. In

addition to the conventional photon absorption by the dye (process 1), visible

photons can be absorbed by the N-doped TiO2 electrode of reduced optical gap

(process 2), leading to the formation of electron-hole pairs in the metal oxide.

Photocurrent generation from this process requires that either photo-generated

holes are transferred to the gold electrode through the dye and the spiro-OMeTAD

(process 3), or through the spiro-OMeTAD directly (process 4) if the metal oxide

surface coverage by the dye is not complete. Energy transfer from the metal oxide

to the dye is also a possible mechanism (process 5). The absolute positions of

energy levels of TiO2, D102, spiro-OMeTAD and gold are extracted from the

literature, while they are only estimations for the N-doped TiO2.

confirm that optically-active metal oxides are a promising strategyto further improve the power conversion efficiency of solid-stateDSSC.

In the following sections, we give preliminary information onthe local electronic configuration of titania nanocrystals as afunction of nitrogen doping, using both numerical modeling andexperimental electron paramagnetic resonance (EPR) experi-ments. The influence of the dopant on the nanocrystal physicalproperties is found to be highly dependent on its location in thestructure. Interstitial and substitutional nitrogen in TiO2 areknown to lead to different electronic and optical properties [43].In this context, experimental or theoretical tools that can bringrelevant information on these aspects are crucial.

3.3. Local features of TiO2 and N-TiO2 nanocrystals

3.3.1. Computer simulations of TiO2 and N-TiO2

In this work, computer simulations based on DFT calculationsusing quantum chemistry code, initially developed to model thephysical properties of isolated molecules, have been performed onisolated TiO2 clusters (see experimental section). Such approach isfound to be highly original, as conventional modeling of TiO2

nanomaterials is usually made using a plane-wave methodologyin a specific environment. In the stoichiometric anatase nanograinstructure, the valence band is dominated by O(2p) orbitals, whilethe conduction band is created by a mixing of oxygen andtitanium orbitals [44]. Asahi et al. reported that N-doped TiO2

films present optical properties showing visible light absorp-tion at wavelengths shorter than 500 nm due to the band gapnarrowing by the mixing of N(2p) and O(2p) states [21]. In orderto check these assumptions for doped nanocrystals, the structurallocation of nitrogen atoms (substitutional and interstitial) wasanalyzed [45]. The substitutional N-doping was created by repla-cing oxygen by nitrogen. The atomic percentage for all nitrogenspecies created in an N-doped (TiO2)17 cluster was equal to3.9 at%. This relatively high value compared to the experimentalnitrogen loading in the N-TiO2 nanopowder (0.5 at%) is due to thelimited number of atoms contained in the investigated cluster. Bychecking the stability of the clusters, we observe that the totalenergy of substitutionally doped nanograins is higher than that ofinterstitial ones. This observation is consistent with reportedcalculations which showed that the transition from the substitu-tional to the interstitial nitrogen doping undergoes an exothermicprocess in the order of 0.8 eV [46]. The electronic properties ofsubstitutionally and interstitially nitrogen doped clusters werealso computed. The calculated HOMO–LUMO energy gap splittingof doped cluster varies from 1.98 eV up to 2.85 eV dependingon the nitrogen position in the structure. Therefore, regardlessof the type of doping, a reduction of the energy gap of anataseis observed. However, no significant participation of nitrogenatomic orbitals is evidenced in the creation of the HOMO orLUMO levels of the doped clusters (Fig. 7a). This result disagreeswith data associated with 3D or bulk N-doped TiO2 materials.Shevlin and Woodley have seen that for small clusters, the HOMOand LUMO levels are a mixture of O(2p) and Ti(3d) orbitals [47].However, for larger clusters, one can clearly observe an influenceof the separation of oxygen and titanium atoms in the creation ofthe HOMO and LUMO orbitals. In this case, the HOMO andHOMO�1 states are composed of O(2p) states and the LUMOand LUMOþ1 states are composed of Ti(3d) states. The men-tioned calculations were performed in periodically reproducedTiO2 nanograin lattice structures, whereas our present data areobtained for isolated clusters located in vacuum. Therefore, ourresults reveal the strong influence of the environment, as well asof high surface to volume ratio, on the electronic properties ofTiO2 nano-clusters.

H. Melhem et al. / Solar Energy Materials & Solar Cells 117 (2013) 624–631 629

A significant influence of nitrogen atomic orbitals on theHOMO level is observed for N-doped clusters containing addi-tional oxygen vacancies (Fig. 7b). In this case, nitrogen possessesan acceptor character. For both clusters containing or not contain-ing oxygen vacancies, the position of the substitutional nitrogenclose to titanium (Ti–N�1.93 A) does not affect the HOMO–LUMO

Fig. 7. HOMO–LUMO energy gap splitting calculated by LC-BLYP DFT methodol-

ogy for the N-doped TiO2 anatase cluster without (a) and with (b) oxygen

vacancies.

Fig. 8. Experimental and simulated EPR spectra of the TiO2 and N-TiO2 nano-

powders under ambient conditions and for various illumination conditions:

(a) un-doped TiO2 under UV irradiation; (b) N-TiO2 without UV in the dark and

(c) N-TiO2 under UV irradiation.

Table 2EPR spectral parameters including g-tensor components and hyperfine coupling A-tens

made under UV irradiation (except for paramagnetic center IV which is observed both

Paramagnetic centers g-tensor components

Type Nature gx gy

I Superoxide radicals (On�

2 ) 2.019 2.004

II Trapped holes (O�) 2.030 2.014

III Colored center 2.003 2.003

IV Nitroxide radicals (NOn�) a 2.0038 2.002

V Substitutional nitrogen radicals (Nn) 2.0085 2.007

a Observed in the dark.b Not properly resolved.

energy gap splitting as much as the second equivalent substitu-tion position of nitrogen (Ti–N�1.98 A).

This observation is related to the charge transfer process thatcan occur from nitrogen to the adjacent titanium atom. Thischarge transfer is found to be more pronounced for nitrogenatoms placed far from the titanium. Finally, our preliminarycalculations indicate that when N-doping is located close to thesurface of the TiO2 cluster in the presence of oxygen vacancies,an additional deep donor level is introduced below the LUMOlevel, as reported in the literature [48]. In the context of thisstudy, these results illustrate the strong influence of N atomson the electronic properties of TiO2 clusters. They also suggestthat computer simulations based on cluster approaches arepowerful tools to assess in the first place the role of dopants innanocrystals.

3.3.2. Local features of TiO2 and N-TiO2 probed by EPR

The main objective of electron paramagnetic resonance (EPR)experiments deals with the characterization of active electroniccenters which can enhance or at the opposite prevent the chargetransfer mechanisms occurring in the metal oxide. Such analysesare regularly reported on TiO2 nanomaterials, and bring highlyrelevant information as long as doping procedures are involved[49–52]. In titanium oxide materials, surface or bulk trappedholes, as well as superoxide radicals can increase charge recom-bination rates [53]. Alternatively, nitrogen doping is known tobroaden the spectral sensitivity of TiO2 materials, and to favorDSSC device operation. In this section, we perform preliminaryEPR experiment on the TiO2 and N-TiO2 nanopowders to gain abetter insight on its influence on solid-state DSSC operation.

No resolved EPR signal is detected on the pure TiO2 sample inthe dark and in ambient conditions. Fig. 8a reports a characteristicEPR spectrum recorded under ambient atmosphere and under UVirradiation of the un-doped TiO2 nanopowder. Three main con-tributions are revealed through the EPR spectral parameterswhich are summarized in Table 2.

Taking into account the deviation of the g-tensor componentsfrom the Lande factor of free electrons (gx¼gy¼gz¼2.0023), thedifferent paramagnetic centers are consistent with superoxideradicals (referred as type I in Table 2), trapped holes (referred astype II in Table 2) and colored centers (referred as type III inTable 2), giving rise to a very narrow and isotropic EPR line.However, if these active electronic centers are evidenced in thesample, their absolute concentrations remain relatively small(�1013–1015 spin/g) and we can expect only minor influenceson the intrinsic electronic properties of TiO2 nanoparticles.

Fig. 8b shows the experimental and simulated EPR spectrarelated to the nitrogen-doped nanopowder without any UVirradiation. The EPR signal is consistent with the presence ofnitrogen atoms associated to oxygen to form nitroxide radicalsNO� (referred as type IV in Table 2. This seems to indicate that

or associated with TiO2 and N-TiO2 nanopowders, estimated from measurements

in the dark and under irradiation for the N-TiO2 sample).

Hyperfine coupling components

gz Ax(Gauss) Ay(Gauss) Az(Gauss)

2.004 – – –

2.014 – – –

2.003 – – –

�1.930 b 5 32 5

2.001 13 32 7

H. Melhem et al. / Solar Energy Materials & Solar Cells 117 (2013) 624–631630

some fraction of nitrogen ions show a tendency to migrate on thenanoparticles surfaces and couple with surface trapped holes.This signal is also likely to result from nitric oxide (NO) adsorbedat the particle surface [54]. It is worth noting that NO has alsobeen observed trapped inside voids of bulk TiO2 materials [55].Finally, substitutional nitrogen radicals (N*, referred as type V inTable 2) are found to be involved in the EPR spectrum of N-TiO2,but only under UV irradiation. In the dark, it remains in diamag-netic state [54,56]. Under UV irradiation, nitrogen ions involved inthe crystalline sites of TiO2 transfer one electron to the conduc-tion band of TiO2 leading to the paramagnetic species Nn, which isindeed detected by EPR (Fig. 8c).

Thus, nitrogen doping is effective in TiO2 nanopowders due tothe effective location of nitrogen in crystalline sites. However,another population of nitrogen atoms are located at the particlesurface as NO*- radicals and are observed under dark or under UV-irradiation as well. These radicals may act as traps for thephotogenerated charge carriers under light irradiation, counter-balancing the positive effect of doping on charge trappingobserved in related studies [16]. At the opposite, substitutionalnitrogen inside the crystalline sites of TiO2 contributes to intro-duce intermediate electronic levels in the gap, leading to visiblelight-active metal oxide. This feature is responsible for improvedphotocurrent collection in solid-state DSSC, and especially for thedirect conversion of blue photons into electrons. However, whencomparing the ratio between surface-like and bulk-like nitrogen,it seems that surface nitrogen is dominant. This situation isfavored by the high specific surface of nanoparticles synthesizedby laser pyrolysis, and also by the tendency of nitrogen dopants tobe more easily involved at the particle surface than in theircrystalline core. Systematic investigations by EPR spectroscopyare now under progress in order to monitor the doping processand the achievement of efficient substitutions of nitrogen intoTiO2 nanoparticles.

4. Conclusions

In summary, we demonstrate solid-state dye-sensitized solarcells using TiO2 and nitrogen-doped TiO2 electrodes, processedfrom nanocrystals synthesized by laser pyrolysis. As reported forliquid DSSC, the influence of nitrogen doping is found to bebeneficial on device performance, with a significant improvementof photocurrent compared to the pure TiO2 device. This improve-ment is found to be correlated with the electronic and opticalproperties of the starting nanopowders. In particular, the appear-ance of an additional absorption band in the N-doped TiO2 isresponsible for the conversion of additional blue photons intoelectrons, in addition to the conventional charge generationmechanisms based on the light absorption by the organic dye.Preliminary computer simulations based on innovative clusterapproaches demonstrate that nitrogen doping plays a crucial rolein the narrowing of the optical band gap of TiO2, especiallythrough substitutional N atoms dispersed in the cluster in thepresence of oxygen vacancies which results in the generation ofN(2 p) states that extend the valence band of the metal oxide.Electron paramagnetic resonance experiments also revealed thepresence of substitutional nitrogen radicals N* in the TiO2 nano-crystals. However, an additional population of nitroxide radicalsmainly located at the particle surface is also evidenced. Theseradicals, which are not directly involved in the visible lightactivity of the N-doped nanopowder, are likely to act as chargetraps under light irradiation. Such local features play a crucial rolein the charge photo-generation mechanisms occurring in theinvestigated solid-state DSSC. Although better insight is requiredon the corresponding charge generation processes, these results

suggest that optically-active metal oxide electrodes are promisingto further enhance the performance of solid-state DSSC or todemonstrate efficient photocatalytic materials.

Acknowledgments

JB gratefully acknowledges the ‘‘Program InterdisciplinaireEnergie’’ of the CNRS (COLHYBRDIE project–2010/2011), as wellas the ‘‘Region Limousin’’ and European Union (FEDER) for anELIARE granted Project (no. 11–3408). Calculations have beencarried out in Wroclaw Center for Networking and Supercomput-ing /http://www.wcss.wroc.plS, Grant no. 171).

Appendix A. Supporting information

Supplementary data associated with this article can be foundin the online version at http://dx.doi.org/10.1016/j.solmat.2012.08.017.

References

[1] A. Yella, H.W. Lee, H.N. Tsao, C. Yi, A.K. Chandiran, M.K. Nazeeruddin,E.W.G. Diau, C.Y. Yeh, S.M. Zakeeruddin, M. Gratzel, Porphyrin-sensitizedsolar cells with cobalt (II/III)-based redox electrolyte exceed 12 percentefficiency, Science 334 (2011) 629–634.

[2] B.E. Hardin, H.J. Snaith, M.D. McGehee, The renaissance of dye-sensitizedsolar cells, Nature Photonics 6 (2012) 162–169.

[3] J. Boucle, J. Ackermann, Solid-state dye-sensitized and bulk heterojunctionsolar cells using TiO2 and ZnO nanostructures: recent progress and newconcepts at the borderline, Polymer International 61 (2012) 355–373.

[4] J. Burschka, A. Dualeh, F. Kessler, E. Baranoff, N.-L. Cevey-Ha, C. Yi,M.K. Nazeeruddin, M. Gratzel, Tris(2-(1H-pyrazol-1-yl)pyridine)cobalt(III)as p-type dopant for organic semiconductors and its application in highlyefficient solid-state dye-sensitized solar cells, Journal of the AmericanChemical Society 133 (2011) 18042–18045.

[5] I. Chung, B. Lee, J. He, R.P.H. Chang, M.G. Kanatzidis, All-solid-state dye-sensitized solar cells with high efficiency, Nature 485 (2012) 486–489.

[6] Q. Zhang, G. Cao, Nanostructured photoelectrodes for dye-sensitized solarcells, Nano Today 6 (2011) 91–109.

[7] X. Zhang, F. Liu, Q.L. Huang, G. Zhou, Z.S. Wang, Dye-sensitized W-doped TiO2

solar cells with a tunable conduction band and suppressed charge recombi-nation, Journal of Physical Chemistry C 115 (2011) 12665–12671.

[8] X. Zhang, S.-T. Wang, Z.-S. Wang, Effect of metal-doping in TiO2 on fill factorof dye-sensitized solar cells, Applied Physics Letters 99 (2011) 113503-3.

[9] Y. Duan, N. Fu, Q. Liu, Y. Fang, X. Zhou, J. Zhang, Y. Lin, Sn-doped TiO2

photoanode for dye-sensitized solar cells, Journal of Physical Chemistry C 116(2012) 8888–8893.

[10] A.K. Chandiran, F. Sauvage, M. Casas-Cabanas, P. Comte, S.M. Zakeeruddin,M. Graetzel, Doping a TiO2 photoanode with Nb5þ to enhance transparencyand charge collection efficiency in dye-sensitized solar cells, Journal ofPhysical Chemistry C 114 (2010) 15849–15856.

[11] J.H. Choi, S.H. Kwon, Y.K. Jeong, I. Kim, K.H. Kim, Atomic layer deposition ofTa-doped TiO2 electrodes for dye-sensitized solar cells, Journal of theElectrochemical Society 158 (2011) B749–B753.

[12] A.M. Tripathi, R.G. Nair, S.K. Samdarshi, Visible active silver sensitized vanadiumtitanium mixed metal oxide photocatalyst nanoparticles through sol–gel tech-nique, Solar Energy Materials and Solar Cells 94 (2010) 2379–2385.

[13] T. Ma, M. Akiyama, E. Abe, I. Imai, High-efficiency dye-sensitized solar cellbased on a nitrogen-doped nanostructured titania electrode, Nano Letters 5(2005) 2543–2547.

[14] H. Tian, L. Hu, C. Zhang, W. Liu, Y. Huang, L. Mo, L. Guo, J. Sheng, S. Dai,Retarded charge recombination in dye-sensitized nitrogen-doped TiO2 solarcells, Journal of Physical Chemistry C 114 (2010) 1627–1632.

[15] H. Tian, L. Hu, C. Zhang, S. Chen, J. Sheng, L. Mo, W. Liu, S. Dai, Enhancedphotovoltaic performance of dye-sensitized solar cells using a highly crystal-lized mesoporous TiO2 electrode modified by boron doping, Journal ofMaterials Chemistry 21 (2011) 863–868.

[16] W. Guo, L. Wu, Z. Chen, G. Boschloo, A. Hagfeldt, T. Ma, Highly efficient dye-sensitized solar cells based on nitrogen-doped titania with excellent stability,Journal of Photochemistry and Photobiology A: Chemistry 219 (2011)180–187.

[17] J.C. Zhang, Z.Y. Han, Q.Y. Li, X.Y. Yang, Y. Yu, W.L. Cao, N. S-doped TiO2, Anodeeffect on performance of dye-sensitized solar cells, Journal of Physics andChemistry of Solids 72 (2011) 1239–1244.

H. Melhem et al. / Solar Energy Materials & Solar Cells 117 (2013) 624–631 631

[18] A. Subramanian, H.W. Wang, Effects of boron doping in TiO2 nanotubes andthe performance of dye-sensitized solar cells, Applied Surface Science 258(2012) 6479–6484.

[19] T. Lindgren, J.M. Mwabora, E. Avandano, J. Jonsson, A. Hoel, C.G. Granqvist,S.E. Lindquist, Photoelectrochemical and optical properties of nitrogen dopedtitanium dioxide films prepared by reactive DC magnetron sputtering,Journal of Physical Chemistry B 107 (2003) 5709–5716.

[20] S. Sakthivel, M. Janczarek, H. Kisch, Visible light activity and photoelectro-chemical properties of nitrogen-doped TiO2, Journal of Physical Chemistry B108 (2004) 19384–19387.

[21] R. Asahi, T. Morikawa, T. Ohwaki, K. Aoki, Y. Taga, Visible-light photocatalysisin nitrogen-doped titanium oxides, Science 293 (2001) 269–271.

[22] W. Guo, Y. Shen, G. Boschloo, A. Hagfeldt, T. Ma, Influence of nitrogen dopantson N-doped TiO2 electrodes and their applications in dye-sensitized solarcells, Electrochimica Acta 56 (2011) 4611–4617.

[23] T. Toyao, M. Minakata, K. Iyatani, A. Ebrahimi, P.C. Chen, C.B. Tsai, Y. Horiuchi,M. Matsuoka, M. Anpo, Development of dye-sensitized solar cells based onvisible-light-responsive TiO2 thin films with a unique columnar structure,Research on Chemical Intermediates (2012) 1–10.

[24] L. Qi, C. Li, Y. Chen, Dye-sensitized solar cells based on nitrogen-doped TiO 2-B nanowire/TiO2 nanoparticle composite photoelectrode, Chemical PhysicsLetters 539-540 (2012) 128–132.

[25] J. Zhang, Q. Sun, J. Zheng, X. Zhang, Y. Cui, P. Wang, W. Li, Y. Zhu, Thecharacterization of nitrogen doped TiO2 photoanodes and its application inthe dye sensitized solar cells, Journal of Renewable and Sustainable Energy 3(2011) 033108–033109.

[26] D. Mitoraj, H. Kisch, The nature of nitrogen-modified titanium dioxidephotocatalysts active in visible light, Angewandte Chemie—InternationalEdition 47 (2008) 9975–9978.

[27] H. Pan, X. Qiu, I.N. Ivanov, H.M. Meyer, W. Wang, W. Zhu, M.P. Paranthaman,Z. Zhang, G. Eres, B. Gu, Fabrication and characterization of brookite-rich,visible light-active TiO2 films for water splitting, Applied Catalysis B:Environmental 93 (2009) 90–95.

[28] C. Karunakaran, G. Abiramasundari, P. Gomathisankar, G. Manikandan,V. Anandi, Cu-doped TiO2 nanoparticles for photocatalytic disinfection ofbacteria under visible light, Journal of Colloid and Interface Science 352(2010) 68–74.

[29] P. Silija, Z. Yaakob, V. Suraja, N.N. Binitha, Z.S. Akmal, An enthusiastic glancein to the visible responsive photocatalysts for energy production andpollutant removal, with special emphasis on titania, International Journalof Photoenergy 2012 (2012).

[30] S.S. Soni, M.J. Henderson, J.F. Bardeau, A. Gibaud, Visible-light photocatalysisin titania-based mesoporous thin films, Advanced Materials 20 (2008)1493–1498.

[31] W. Guo, Y. Shen, L. Wu, Y. Gao, T. Ma, Effect of N dopant amount on theperformance of dye-sensitized solar cells based on N-Doped TiO2 electrodes,Journal of Physical Chemistry C 115 (2011) 21494–21499.

[32] H. Melhem, P. Simon, L. Beouch, F. Goubard, M. Boucharef, C. Di Bin,Y. Leconte, B. Ratier, N. Herlin-Boime, J. Boucle, TiO2 nanocrystals synthesizedby laser pyrolysis for the up-scaling of efficient solid-state dye-sensitizedsolar cells, Advanced Energy Materials 1 (2011) 908–916.

[33] B. Pignon, H. Maskrot, V.G. Ferreol, Y. Leconte, S. Coste, M. Gervais, T. Pouget,C. Reynaud, J.F. Tranchant, N. Herlin-Boime, Versatility of laser pyrolysisapplied to the synthesis of TiO2 nanoparticles—application to UV attenua-tion, European Journal of Inorganic Chemistry (2008) 883–889.

[34] P. Simon, B. Pignon, B. Miao, S. Coste-Leconte, Y. Leconte, S. Marguet, P. Jegou,B. Bouchet-Fabre, C. Reynaud, N. Herlin-Boime, N-doped titanium monoxidenanoparticles with TiO rock-salt structure, low energy band gap, and visiblelight activity, Chemistry of Materials 22 (2010) 3704–3711.

[35] V. Shrotriya, G. Li, Y. Yao, T. Moriarty, K. Emery, Y. Yang, Accurate measure-ment and characterization of organic solar cells, Advanced FunctionalMaterials 16 (2006) 2016–2023.

[36] M.W. Schmidt, K.K. Baldridge, J.A. Boatz, S.T. Elbert, M.S. Gordon, J.H. Jensen,S. Koseki, N. Matsunaga, K.A. Nguyen, S. Su, T.L. Windus, M. Dupuis,

J.A. Montgomery, General atomic and molecular electronic structure system,Journal of Computational Chemistry 14 (1993) 1347–1363.

[37] W.J. Stevens, H. Basch, M. Krauss, Compact effective potentials and efficientshared-exponent basis sets for the first- and second-row atoms, The Journalof Chemical Physics 81 (1984) 6026–6033.

[38] W.J. Stevens, M. Krauss, H. Basch, P.G. Jasien, Relativistic compact effectivepotentials and efficient, shared-exponent basis sets for the third-, fourth-,and fifth-row atoms, Canadian Journal of Chemistry 70 (1992) 612–630.

[39] T.R. Cundari, W.J. Stevens, Effective core potential methods for the lantha-nides, The Journal of Chemical Physics 98 (1993) 5555–5565.

[40] Y. Tawada, T. Tsuneda, S. Yanagisawa, T. Yanai, K. Hirao, A long-range-corrected time-dependent density functional theory, The Journal of ChemicalPhysics 120 (2004) 8425–8433.

[41] S.H. Jung, S.W. Kang, Formation of TiO2 thin films using NH3 as catalyst bymetalorganic chemical vapor deposition, Japanese Journal of Applied Physics,Part 1: Regular Papers and Short Notes and Review Papers 40 (2001)3147–3152.

[42] A. Kay, M. Gratzel, Dye-Sensitized Core-Shell Nanocrystals, Improved effi-ciency of mesoporous tin oxide electrodes coated with a thin layer of aninsulating oxide, Chemistry of Materials 14 (2002) 2930–2935.

[43] M. Ceotto, L. Lo Presti, G. Cappelletti, D. Meroni, F. Spadavecchia, R. Zecca,M. Leoni, P. Scardi, C.L. Bianchi, S. Ardizzone, About the nitrogen location innanocrystalline N-doped TiO2: combined DFT and EXAFS approach, TheJournal of Physical Chemistry C 116 (2012) 1764–1771.

[44] M. Harb, P. Sautet, P. Raybaud, Origin of the enhanced visible-light absorp-tion in N-doped bulk anatase TiO2 from first-principles calculations, TheJournal of Physical Chemistry C 115 (2011) 19394–19404.

[45] A. Fujishima, X. Zhang, D.A. Tryk, TiO2 photocatalysis and related surfacephenomena, Surface Science Reports 63 (2008) 515–582.

[46] C. Di Valentin, G. Pacchioni, A. Selloni, Origin of the different photoactivity ofN-doped anatase and rutile TiO2, Physical Review B 70 (2004) 085116.

[47] S.A. Shevlin, S.M. Woodley, Electronic and optical properties of doped andundoped (TiO2)n nanoparticles, The Journal of Physical Chemistry C 114(2010) 17333–17343.

[48] J. Wang, D.N. Tafen, J.P. Lewis, Z. Hong, A. Manivannan, M. Zhi, M. Li, N. Wu,Origin of photocatalytic activity of nitrogen-doped TiO2 nanobelts, Journal ofthe American Chemical Society 131 (2009) 12290–12297.

[49] T. Berger, M. Sterrer, O. Diwald, E. Knozinger, Charge trapping and photo-adsorption of O2 on dehydroxylated TiO2 nanocrystals—an electron para-magnetic resonance study, Chemphyschem: a European Journal of ChemicalPhysics and Physical Chemistry 6 (2005) 2104–2112.

[50] P. Billik, G. Plesch, V. Brezova, L. Kuchta, M. Valko, M. Mazur, Anatase TiO2

nanocrystals prepared by mechanochemical synthesis and their photoche-mical activity studied by EPR spectroscopy, Journal of Physics and Chemistryof Solids 68 (2007) 1112–1116.

[51] F. Spadavecchia, G. Cappelletti, S. Ardizzone, C.L. Bianchi, S. Cappelli, C. Oliva,P. Scardi, M. Leoni, P. Fermo, Solar photoactivity of nano-N-TiO2 from tertiaryamine: role of defects and paramagnetic species, Applied Catalysis B:Environmental 96 (2010) 314–322.

[52] H. Sun, Y. Bai, W. Jin, N. Xu, Visible-light-driven TiO2 catalysts doped withlow-concentration nitrogen species, Solar Energy Materials and Solar Cells 92(2008) 76–83.

[53] J. Weidmann, T. Dittrich, E. Konstantinova, I. Lauermann, I. Uhlendorf,F. Koch, Influence of oxygen and water related surface defects on the dyesensitized TiO2 solar cell, Solar Energy Materials and Solar Cells 56 (1999)153–165.

[54] S. Livraghi, M.C. Paganini, E. Giamello, A. Selloni, C. Di Valentin, G. Pacchioni,Origin of photoactivity of nitrogen-doped titanium dioxide under visiblelight, Journal of the American Chemical Society 128 (2006) 15666–15671.

[55] S. Livraghi, A. Votta, M.C. Paganini, E. Giamello, The nature of paramagneticspecies in nitrogen doped TiO2 active in visible light photocatalysis, ChemicalCommunications (2005) 498–500.

[56] P. Mei, M. Henderson, A. Kassiba, A. Gibaud, EPR study of nitrogen-dopedmesoporous TiO2 powders, Journal of Physics and Chemistry of Solids71 (2010) 1–6.