Analysis of photocurrent and capacitance of TiO 2 nanotube – polyaniline hybrid composites synthesized through electroreduction of an aryldiazonium salt

TiO 2 nanotube – polyaniline hybrid composites were synthesized by a multi-step electrochemical procedure as follows: (1) electrochemical oxidation of Ti foil to obtain TiO 2 ordered nanotubular electrodes, (2) functionalization of TiO 2 by electroreduction of 4-nitrobenzendiazonium salt, (3) electrochemical reduction of the nitro group to amine, (4) galvanostatic electropolymerization of aniline onto the surface of the aminophenyl-modi ﬁ ed TiO 2 electrode. Samples were also prepared by direct electropolymerization of PANI on TiO 2 nanostructures without the deposition of 4-nitrobenzendiazonium as under-layer. Di ﬀ erent times for the electropolymerization step were applied. The composite materials were morphologically and electrochemically characterized by scanning electron microscopy, cyclic voltammetry and electrochemical impedance analysis. The improvement of the most relevant properties of the PANI – TiO 2 composites for their application as photocatalysts and as supercapacitors was checked.


Introduction
TiO 2 is an n-type semiconducting material with very interesting properties which is used in many different applications.Among these properties, its chemical stability, high photocatalytic activity, nontoxicity, low cost, availability, good mechanical exibility and conductivity should be noted. 1It has potential applications in many elds, such as lightweight battery electrodes, electromagnetic shielding devices, anticorrosion coatings and sensors. 2,35][6][7] However, several drawbacks can limit its use for these emerging applications.In the rst place, the wide band gap energy of TiO 2 (E g ¼ 3.2 eV) limits it to absorb visible light (l > 380 nm).5][6][7] The relatively small specic capacitance was attributed to the poor electrochemical activity and poor electrical conductivity of TiO 2 . 8,91][12] Higher conductivity, fast electrochemical switching and high specic capacitance values as well as environmental stability may be conferred to the nal structure by the presence of the polymer. 13A very recent work, on the use of polypyrrole, reported that its nanostructure may greatly inuence the properties and overall functionality for its possible application. 14][17][18] Due to its delocalized conjugated structure in electron-transfer processes, PANI can be used as a photosensitizer of TiO 2 to remarkably improve the efficiency of its photocatalytic activity under visible light. 1,19The redox reversible processes associated with doping and undoping of counterions accompany charge storage in PANI giving rise to a pseudocapacitance; this contribution has to be added to a double-layer capacitance due to the separation of charges which takes place at PANI-electrolyte interface. 20The theoretical specic capacitance of PANI, which is a combination of pseudocapacitance and electrical double-layer capacitance, has been estimated as 2.0 10 3 F g À1 . 21owever, when electropolymerization is used for coating of surfaces, weak adhesion between the organic layer and a substrate of different nature can occur: as was reported in the literature, the lack of strong interfacial bonding may result in the collapse of the hybrid nanostructures. 22Other authors reported that, despite the very high capacitance (3407 F g À1 ) measured at ordered PANI nanorods on an ITO substrate, low stability was obtained during repeated cycles, due to poor electrical connectivity between the nanorods and the smooth surface of the ITO substrate. 23This poor cyclability has become a major obstacle for PANI to be used in supercapacitors.
In this context, the present work considers the synthesis of hybrid TiO 2 electrodes and investigates the possibility to stabilize PANI deposition on them, by means of a preliminary surface functionalization through the electrochemical reduction of 4-nitrobenzendiazonium salt.Actually, the electrochemical reduction of diazonium derivatives is an elegant and versatile technique for surface functionalization, which was rst reported in 1992 by Delamar et al. 24 The mechanism of attachment was described as the generation of aryl radicals and subsequent formation of covalent bonds between the surface and the aryl groups. 25,267][28][29][30] The electrodeposition of PANI using an under-layer obtained by electroreduction of aryl diazonium has been proposed in the literature to modify carbon nanotubes, 31 glassy carbon electrodes, 32 and gold: 33 the PANI composites obtained with this technique showed higher chemical stability, a more uniform layer and high values of capacitance.
In the present work, the electrodeposition of PANI using aminophenyl moieties as an under-layer is proposed to modify TiO 2 nanotubular structured electrodes obtained by electrochemical anodization of Ti foil.A rst layer of nitrophenyl moieties was graed on titania nanotubes by the electrochemical reduction of 4-nitrobenzendiazonium salts; nitrogroups were then electroreduced to amino-groups.Finally, PANI was electropolymerised at the modied electrodes.As a comparison, PANI was also electropolymerized on bare TiO 2 nanotubes.The electrodes were characterized by scanning electronic microscopy, cyclic voltammetry, electrochemical impedance spectroscopy and charge-discharge tests to evaluate their performance with particular regard to the enhancement in photocurrent and capacitance.

Chemicals
Ammonium uoride was purchased from Fluka; 4-nitrobenzenediazonium tetrauoroborate and tetrabutylammonium hexauorophosfate (TBAPF6) were purchased from Sigma-Aldrich.Aniline was freshly distilled and stored in the dark.Alcohols and the other solvents were purchased in rened grade from Carlo Erba or Sigma Aldrich and used without further purication.The aqueous solutions were prepared with doubledistilled water (18 MU cm).

Apparatus
All electrochemical experiments were performed at room temperature using an AMEL Potentiostat, controlled with the Junior Assist soware, coupled with an AMEL Frequency Response Analyzer, controlled with the Zassist soware.Irradiation of the samples was performed by means of a 300 W Xe lamp (LOT-Quantum Design Europe) equipped with suitable lters to select different wavelengths.
A conventional three-electrode cell was generally used in which the investigated samples worked as anodes; a Pt grid was used as counter electrode while a saturated calomel electrode (SCE) was the reference.All the potential values given in the text refer to SCE.The exposed geometrical area of the working electrodes was 1 cm 2 .A two electrode conguration was adopted for the synthesis of nanotubular structures: also in this case a grid of Pt was used as counter electrode.

Synthesis of the anodic materials
Titanium foil (Sigma-Aldrich®, 0.25 mm thickness 99.7% metal basis) was rstly degreased by sonication in acetone, then in isopropanol and in methanol (10 min each), and nally rinsed with deionized water.The pretreated foil was then submitted to electrochemical anodization.Oxidation was performed in glycerol-water (75 : 25 vol%) solutions containing 0.14 M NH 4 F: a potential ramp was applied from open-circuit voltage (OCV) to a xed potential (20 V) with a scan rate of 50 mV s À1 ; then, the applied potential was maintained at this xed value for two hours.Aer oxidation, the electrodes were rinsed in deionized water and dried in a nitrogen stream.A subsequent annealing treatment was needed in order to transform the amorphous phase into a crystalline structure.Thermal treatment was performed in air atmosphere for 1 h at 400 C.
Aer anodization a functionalization of the electrode surface with 4-nitrobenzenediazonium was performed by cyclic voltammetric runs at 100 mV s À1 in the range of potential 0.4 to À0.7 V in acetonitrile + 0.1 M (TBAPF 6 ) solution containing 2 mM of 4-nitrobenzenediazonium tetrauoroborate.Electrochemical reduction of nitro group to amino group was performed by cyclic voltammetry in water-ethanol solutions (90 : 10 vol%) containing 0.1 M KCl; the potential was varied from the open circuit potential to À1.2 V and back to 0.2 V.The scan rate was 100 mV s À1 .
Finally, PANI polymerization was carried out on either bare TiO 2 or phenylamine modied TiO 2 electrodes.To this aim galvanostatic runs were performed at a constant current density of 20 mA cm À2 from aqueous solutions containing aniline (0.1 M) and HNO 3 (1 M).Different electropolymerization steps (30, 60 and 120 minutes) were applied in order to obtain PANI lms with different thicknesses.A minimum of 30 minutes of electropolymerization was needed to obtain a measurable coverage of the polymer.According to the Faraday law, if a unit current yield is supposed, the related load of PANI may be evaluated as 69.5, 139 and 278 mg cm À2 respectively for the three indicated steps, when 0.5 electrons per mole aniline are considered to be involved in the process. 34o simplify the discussion, hereaer NT will be used to indicate a bare nanotubular structure, while modied samples will be identied with NT-PANI x , NT-DZ-PANI x , when PANI is directly linked on nanotubes, or by means a preliminary phenylamine underlayer, respectively.The PANI load of the different samples will be indicated by the subscript number x which represents the time (minutes) of duration of the electropolymerization step.

Characterization of the electrodes
Scanning electron microscopy (SEM) was used to study the morphology of the assembled electrodes prepared by the different approaches (NOVA 600, Nanolab dual beam).Focused ion beam (FIB) was also used to obtain very precise cross sections of samples for subsequent imaging via SEM.
In order to evaluate the performance of the samples, the trend of photocurrent with the potential was preliminarily investigated.Depending on the sample different polarization trends were obtained and the plateau of photocurrent saturation was achieved at a different potential from the OCV value (data not shown).Thus, the value of 1 V above the OCV was selected to measure photocurrent in order to have a reference value at which all the samples were under saturation conditions.
Most of the measurements were done in the UV range (l ¼ 365 nm); the performances were also evaluated using the lamp without lter, to better simulate solar radiation (white light).The value of photocurrent was taken as the difference between light and dark currents.Electrochemical impedance spectroscopy (EIS) measurements were carried out in the dark, at the OCV and in a frequency range from 100 kHz down to 0.1 Hz.The complex capacitance values were derived from the impedance data evaluated at each frequency.Values of global capacitance were also derived for comparison, from typical charge-discharge tests: According to eqn (1) the capacitance was calculated from the slope DV/Dt of the discharge curve obtained at a constant current density i.Typical experiments were performed in a potential range of 1.5 V with a discharge current of 0.1 mA.

Results and discussion
Fig. 1 shows an example of the nal raw NT structure obtained aer two hours of oxidation of Ti foils.As it can be observed, a regular structure of NTs was obtained, with an average tube diameter of about 90 nm, and a mean length of about 1 mm.
The NT TiO 2 electrodes were then processed to obtain the under-layer of phenylamine, suitable for anchoring the PANI, with a procedure which has been indicated as a simple way to achieve a regular distribution of the subsequent layer, with a high stability. 31,33g. 2A shows an example of the CV obtained aer two consecutive cycles of functionalization.In the rst cycle, a cathodic peak is observed at about 0 V which can be related to the reduction of the diazonium group.Further cycles exhibit a displacement of the cathodic peak towards more negative potentials, as well as a decrease of the recorded current indicating the progressive blocking of the electroactive surface due to the formation of the organic lm. 25,26 further process was performed cycling the potential from 0.2 to À1.2 V in water-ethanol solution containing 0.1 M KCl to transform the NO 2 termination into NH 2 termination, to activate the aromatic ring for the subsequent electrophilic attack.
As it is generally reported in the literature 35,36 this process occurs through an intermediate reversible step which involves the nitrous-hydroxylamine (NO-NHOH) redox couple, followed by the irreversible amination of the hydroxylamine group.According to this mechanism, data presented in Fig. 2B show the anodic peak related to the intermediate process at about À0.52 V: just a knee is observable in the direct scan of the CV, assessing the corresponding reduction, at about the same potential, due to its partial overlapping with the irreversible reduction wave, which has its maximum at around À0.8 V.
The functionalized samples were then submitted to aniline electropolymerization.The process was carried out by galvanostatic electrolysis at 20 mA for different intervals of time.For comparison purposes, the electropolymerization was carried out also at bare NTs, with the same procedure adopted for the functionalised electrodes.The nal coverage of the surface by the polymer was checked by cyclic voltammetry and SEM analyses.Fig. 3 shows an example of the trend of voltammetric runs performed in H 2 SO 4 0.1 M at an NT-DZ-PANI electrode which assesses the successful link of the polymer to the surface.Fig. 4 shows the comparison between the SEM images of NT-PANI (Fig. 4A) and NT-DZ-PANI (Fig. 4B) obtained aer 60 minutes of electropolymerization: when bare NT electrodes are coated, PANI tends to deposit irregularly inside the tubes and on their top mouth: a partial plugging of the tubes can be also observed, which may compromise further access of the electrolyte to the tubes.A different aspect of the nal surfaces is observable for the NT-DZ-PANI samples, with a more regular distribution of the polymer.Moreover, the regular deposit does not seem block the entrance on the NTs, while a thickening of the walls and a decrease in the diameter of the NTs is observed, indicating that the polymerization mainly occurs at the NT walls.
The photocurrents of the prepared samples were evaluated in aqueous Na 2 SO 4 solution, under 1 V of overpotential from the OCV, at which saturation current conditions were guaranteed for all the samples.Fig. 5A shows the results obtained at l ¼ 365 nm for different PANI loads with and without under-layer.In particular, the comparison is done in terms of the ratio between the photocurrent measured at the different samples and that measured at a bare NT electrode.Fig. 5B shows data of photocurrent density obtained for NT-DZ-PANI samples irradiated with white light.
Data shown in Fig. 5 indicate a positive effect only at low loading of PANI: only the performances of the PANI 30 samples are larger than those of NT under UV light; the situation is   slightly better under white light, under which also the sample with 60 min loading performs well.Nevertheless, in any case the performances abruptly decrease at samples highly loaded: a negative effect of PANI polymerization was observed at longer deposition times, probably due to the partial plugging of the structure previously mentioned.As highlighted by different authors [37][38][39][40][41] the combination of PANI with TiO 2 can improve their performance in relation to UV light and white light because PANI is a conducting polymer and a dye with a forbidden band gap of 2.8 eV, 42 making it a good sensitizer of TiO 2 .Based on the well-established energy band theory of PANI-TiO 2 composites, the photocatalytic mechanism under visible irradiation involves the electrons in the p-orbital of PANI which are excited to the p*-orbital by visible light and can readily inject into the CB of TiO 2 because the energy level of them matched well for the charge transfer. 43Nevertheless, an excessive PANI load may occupy the active sites of TiO 2 ; moreover, it may cause the recombination of the electrons with PANI + , which competes with the injection into the conduction band of TiO 2 . 43The maximum of visible photoactivity implies that there is an optimal load for the interfacial electrons transfer from PANI to TiO 2 .
By comparing the data in Fig. 5, it is worth noting that, PANI deposition time being the same, the gain in efficiency for NT-DZ-PANI with respect to NT samples is higher under white light, probably due to the greater absorption of PANI under visible light.Thus, for the sample with NT-DZ-PANI 30 , the percentage of current increase with respect NT sample is about 100% (Fig. 5B), whereas the analogous gain is only of 20% at 365 nm (Fig. 5A).In the absence of the under-layer the analogous comparison leads to only 10% of gain (samples NT-PANI 30 in Fig. 5A).
The presence of the under-layer also causes a lower decrease at high PANI loads: for NT-DZ-PANI 120 the residual activity at 365 nm is about 20% of the initial value, whereas just 6% of the initial activity remained for NT-PANI 120 .This result can directly be linked to the better distribution of PANI on the surface of the TiO 2 nanotubes (checked by SEM analyses) when the polymerization is carried out aer the functionalization of the structure with the benzene-diazonium salt.
Electrochemical impedance spectroscopy (EIS) was used to obtain information on the capacitive behavior of the samples.Cyclic voltammetry is oen proposed to derive the capacitance: however, due to the large potential perturbation, the CV currents may include contributions from both localized and conduction band charge transfers, thus giving overestimated capacitance values; at the same time charge transfer kinetics could also affect the current, leading to underestimated capacitance at high potential scan rates. 44As a steady state technique with small potential variation, EIS has been suggested as a more reliable method to get capacitance data, with minimized effect from non-capacitive Faradaic contributions. 44n this work all the EIS were performed at the OCV, superimposing a sinusoidal wave of 10 mV.
The effect of the PANI coating on TiO 2 is shown in Fig. 6, where the trend of Nyquist plot for bare NTs is compared with those of electrodes with different PANI load (Fig. 6A).
The general trend of the spectra always involves rst a semicircle in the higher frequency range, which may be attributed to the electrode surface properties and the charge transfer resistance.A nearly vertical trend is observed in the lower frequency range which may be connected to a capacitive component or/and mass transfer effects.The shape of the plots and the slope of the straight line depend on the PANI load.A gradual transition from capacitive to Warburg-like behavior was already reported in the literature for PANI based systems in the low frequency range, in aqueous and non-aqueous media and with various substrates. 45n the present case, when a high load of PANI is considered (as in the sample NT-PANI 60 ) the presence of the new interface TiO 2 -PANI is well highlighted by the appearance of a second circle at the intermediate frequencies (see inset of Fig. 6A).
The effect of PANI is also evident in the Bode modulus and phase angle plots (Fig. 6B and C): due to the conductive character of PANI, a clear decrease in modulus is observed when PANI load increases, and a new peak appears in the phase diagram at intermediate frequency, when PANI load is enough to clearly show its contribution.In order to better understand these results, the impedance data may be analyzed in terms of complex capacitance. 46Starting from the real and imaginary impedance values, we can express as: C 0 ðuÞ ¼ ÀZ 00 ðuÞ ujZðuÞj 2 C 00 ðuÞ ¼ Z 0 ðuÞ where Z 0 (u) and Z 00 (u) are the real and imaginary parts of the complex impedance Z(u), respectively, and u is the angular frequency u ¼ 2pf.At low frequency, C 0 (u) corresponds to the capacitance of the electrode material and C 00 (u) corresponds to the dissipation of energy by irreversible processes that lead to a hysteresis. 47ata in terms of Bode modulus plot are shown in Fig. 7: the inset reports the trend of the relative Nyquist C 00 vs. C 0 .
Data in Fig. 7 illustrate the gain in capacitance obtained due to the PANI load, with values of C, evaluated at 0.1 Hz, ranging from 813 mF cm À2 for NT to 14 mF cm À2 measured for NT-PANI 120 .
The effect of the underlayer is illustrated in Table 1, which reports the comparison in terms of ratio between C of the different samples and C of the related bare NT, with and without the under-layer.
The large PANI effect results in values of capacitance up to 15-fold the values of the bare NT.The presence of the underlayer increases this effect especially for the sample with 60 min of PANI load; the effect of the under-layer is less evident at 120 min of PANI loading, probably because of the effect of plugging of pores, previously mentioned.
Capacitance was also measured by charge-discharge tests in order to make a comparison: values of 800 mF cm À2 , 1.8 mF cm À2 , and 12 mF cm À2 were respectively obtained at NT, NT-DZ-PANI 30 NT-DZ-PANI 120 which is in good agreement with those obtained from EIS.
5][6][7] However, if TiO 2 -PANI are considered, the comparison is not so straightforward.Resistance and capacitance of PANI depend on the thickness of the lm and on the preparation procedures; moreover, the values of capacity depend on the techniques used for their measurement. 44lso of particular concern is the trend of C 00 vs. f plot (Fig. 8), because it allows a direct comparison of the resonant frequency (f 0 ) for the different samples.As is generally obtained for semiconductor materials, the trend of the imaginary capacitance with the frequency passes through a maximum at a value f 0 , which virtually separates the range of the capacitive (f < f 0 ) and resistive (f > f 0 ) behavior of the sample.Moreover, from the resonant frequency, the value of the relaxation time constant (s 0 ) can be calculated as: This parameter is another important gure of merit which determines the charge-discharge time of the condenser.Low time constants (of the order of ms) are preferred for electrochemical supercapacitors to ensure fast charge-discharge characteristics. 45ata in Fig. 8 indicate that the PANI layer over the sample surface makes the capacitive behavior active at lower frequency, and samples with higher load have slower charge-discharge process.
In Fig. 9 the C 00 vs. f trends for different samples are compared, in particular, two extreme cases are represented.On the one hand, we observe the trend related to a conducting Pt electrode, which is reported in the same gure as a comparison: a regular increasing trend without any peak is presented by this sample, as we expected due to its completely resistive behavior.On the opposite side, we observe the behavior of NTs bare electrode (see inset) which shows a capacitive behavior at a frequency lower than 310 Hz (f < f 0 ); the abrupt increase on C 00 , which is measured at very low frequency, may be due to a faradic capacitance contribution, related to possible charge transfer at the electrode-solution interface.The trends obtained with the other modied samples differ in terms of displacement of f 0 and/or faradic response contribution.The main effect of the PANI is an increase in conductivity of the surface, and aer 60 min the semiconductor character of the titania NT is almost completely masked and the trend is very similar to that of Pt.
When the interlayer is interposed in between, the capacitive behavior is maintained: the resonance frequency is shied to lower values with respect to that of NT bare sample, and the faradic capacitance contribution becomes evident at f lower than 10 Hz.However, the shiing to the right of the curve related to the sample NT-DZ-PANI 120 lets us to conclude that also in this case, the prevailing conductive character of the polymer will prevail at high PANI loads.

Conclusions
From the results presented in this work, the following conclusions can be drawn: -The morphology of PANI deposited onto the TiO 2 NT electrodes is highly improved by the under-layer of phenylamine, with a more regular distribution of the polymer.Moreover, the PANI deposit does not plug the entrance on the NTs because the polymerization mainly occurs at the NTs walls.
-The photocurrent responses indicated that the electropolymerization process should not be too much prolonged.Aer 30 min, the photocurrent obtained is higher than working with bare NTs, but it worsens when applying a higher duration polymerization process.This negative effect has been attributed to the plugging of the nanotubes.
-When light without any lter is applied, the highest photocurrent values were obtained in the samples with underlayer.This result is due to the greater absorption of PANI in the whole spectrum of visible/UV light.
-The presence of the PANI layer results in characteristic shapes in the Nyquist and Bode diagrams, related to the gain in the capacity.The hybrid composites have a capacitance 15-fold the value of bare NTs, with the highest values obtained working with the samples NT-DZ-PANI.
-Although the deposition of PANI produces an increase in the conductivity of the TiO 2 , so masking its semiconductor characteristics, with the use of underlayer the capacitive behavior prevails, even at high PANI loads.

Fig. 1
Fig. 1 SEM images of surface (A) and of FIB cross section (B) of TiO 2 nanotubes synthesized in glycerol solutions.

Fig. 2
Fig. 2 Functionalization process of the NT surface.First step (A): two consecutive CV runs performed in acetonitrile solution containing 2 mM nitro-benzendiazonium salt; second step (B): CV performed in water-ethanol solution containing 0.1 M KCl.

Fig. 3
Fig. 3 Example of the trend of CV runs performed in 0.1 M H 2 SO 4 at NT-DZ-PANI electrode.

Fig. 5
Fig. 5 Effect of PANI load on photocurrent measured at 365 nm, in 0.1 M Na 2 SO 4 at 1 V of overpotential.(A) The percentage between the photocurrent measured at the different samples and that measured at a bare NT electrode.(B) Photocurrent density measured on samples irradiated with white light.

Fig. 6
Fig. 6 Effect of the PANI load on the impedance measured for different NT-PANI samples (A): Nyquist plots; (B) phase and modulus (C) Bode diagrams.

Fig. 7
Fig. 7 Trend of capacitance modulus vs. frequency measured at NT samples modified with different PANI loads.Inset: Nyquist plot.

Table 1 Fig. 8
Fig. 8 Trend of imaginary capacitance vs. frequency measured for NT samples modified with different PANI loads.