Effect of the second chromophore energy gap on photo-induced electron injection in di-chromophoric porphyrin-sensitized solar cells

This work investigates the effect of the second chromophore energy gap on charge generation in porphyrin-based di-chromophoric dye-sensitized solar cells (DSSCs). Three di-chromophoric porphyrin dyes (PorY, PorO and PorR) containing three organic chromophores with decreasing frontier orbital energy offsets, including a carbazole-triphenylamine chromophore (yellow, Y), a carbazole fused-thiophene chromophore (orange, O) or a carbazole-thiophene benzothiadiazole thiophene chromophore (red, R), were investigated using optical and electrochemical methods, steady-state photoluminescence and photovoltaic device characterization. Energy transfer from the organic chromophore to the porphyrin was suggested in PorY and PorO as the main charge generation mechanism in DSSCs using these di-chromophoric dyes. On the other hand, electron transfer from the photo-excited porphyrin to the organic chromophore as a competing pathway leading to the loss of photocurrent is suggested for PorR-sensitized solar cells. The latter pathway leading to a loss of photocurrent is due to the lower lying lowest unoccupied molecular orbital of the additional organic chromophore (R) and suggests the limitation of the current di-chromophoric approach to increase the overall efficiency of DSSCs.


Introduction
Porphyrins are some of the most remarkable sensitizers in dye-sensitized solar cells (DSSCs), achieving some of the highest power conversion efficiencies (PCEs) to date [1][2][3]. Multi-chromophoric porphyrin dyes having more than one chromophore, either increasing the extinction coefficient or extending the photon absorption wavelength range as an alternative to the co-sensitization approach, have attracted the interest of researchers to further increase the DSSC performance [4][5][6][7][8][9][10][11]. One of the interesting aspects of the multi-chromophoric design is that the electronic communication between the chromophores can be inhibited by a twisted threedimensional structure, weakening the dispersion forces (polarizability of the molecule) and elongating the electron lifetime [9]. Furthermore, intramolecular hole transfer between the porphyrin and the side chain chromophore can lead to faster dye regeneration rate at the TiO 2 /dye/electrolyte interface without increased driving force for the electron transfer reaction [12]. However, the PCEs of DSSCs using multichromophoric porphyrin dyes are often lower than those of the best single porphyrin or porphyrin/organic dye co-sensitized solar cell approaches. In one of the best reported examples, a PCE of 6.2% was achieved by a carbazole substituted porphyrin di-chromophoric DSSC [13], which is only half the value reported using a donor-p bridge-acceptor structured porphyrin dye (SM315)-sensitized solar cell (13%) [1].
One of the main limitations in the photovoltaic performance of reported multi-chromophoric porphyrins is that the light absorption of the second chromophore is not extended far enough beyond the porphyrin absorption. In this regard, generating suitable electronic levels in the second chromophore that lead to the extension of light absorption while not introducing unwanted energy/electron transfer pathways within the chromophores has been the main challenge. In principle, there are three distinct mechanisms relating to energy transfer processes. The most frequent one is Förster resonance energy transfer (FRET), which does not require strong electronic coupling between the two adjacent chromophores but is mainly restricted to singlet-singlet energy transfer [14]. A less studied but also very important energy transfer process is the Dexter mechanism, which involves electron exchange between the donor and the acceptor [15]. Though this electron transfer mechanism requires significant molecular orbital mixing of the two chromophores, it can promote both singlet-singlet and triplet to triplet excited state energy transfer [16,17]. The third possible energy transfer process is photo-induced electron transfer (PET) which causes quenching of energy by dissipating heat. A feature of the PET quenching mechanism is the formation of exciplex (D þ †A 2 )* after photo-excitation of the donor [18,19].
Understanding charge generation mechanisms in multi-chromophoric sensitizers is quite important in the field of multi-chromophoric DSSCs [6,20]. The quantum yield of these possible energy/electron transfer reactions in a di-chromophoric dye can affect the electron injection efficiency into the semiconductor. In a TiO 2 -D1-D2 di-chromophoric dye-sensitized film system, where D1 is proximate to the surface of TiO 2 , the main route for charge generation after photo-excitation of D2 is energy transfer from D2* to D1, forming D1*, followed by electron injection from D1* to TiO 2 . Similar energy transfer pathway was illustrated in artificial light-harvesting antenna macromolecules [21]. In two Re-Ru supramolecules, it is also possible for a remote electron injection from D2* to TiO 2 [22]. However, one special case is that in which the lowest unoccupied molecular orbital of D1 lies higher than that of D2 and D1 is photo-excited. As such, the direction of energy/electron transfer is reversed under photo-excitation either due to a change in the overlap of photoluminescence (PL) and absorption spectra, or a change in the driving force for electron transfer. Energy and electron transfer from D1* to D2 may occur and decrease the electron injection yield from D1* to TiO 2 conduction band due to a kinetic competition between the two 'opposite' pathways. Here in this work, steady-state PL measurements and photovoltaic device characterization are correlated to provide information on charge transfer mechanisms in di-chromophoric DSSCs. Figure 1 displays the structures of Por, PorY, PorO, PorR and the organic chromophores. The three organic chromophores employed are with systematically lower energy gaps: PorY contains a carbazole (CbTPA) chromophore, PorO contains a carbazole-fused thiophene (CbTh) chromophore and PorR contains a carbazole-thiophene benzothiadiazole thiophene (CbBTD) chromophore. The organic chromophores Y, O and R absorb photons at different wavelength range: Y for yellow, O for orange and R for red. The organic chromophore and the porphyrin were linked via a phenylethenyl unit to prevent effective electronic communication between the two chromophores by a dihedral angle.

Experimental procedure
Differential pulse voltammetry (DPV) was conducted with a three-electrode system using an eDAQ (BVI) instrument. The working electrode was a Pt wire with the surface area of 6.6 mm 2 , the counter electrode rsos.royalsocietypublishing.org R. Soc. open sci. 5: 181218 was a slice of Pt mesh and the reference electrode was an inhouse-made Ag/AgCl wire. The Ag/AgCl potential was calibrated using 1 mM freshly prepared ferrocene (Fc, 98%, Aldrich) in the same solution.
CbTPA and PorY were measured in dichloromethane (DCM, 99.8%, Chem-supply) with concentration of 0.5 mM, while CbTh, PorO, CbBTD and PorR were measured in dimethylformamide (DMF, 99.99%, Honeywell) with concentration of 0.5 mM. Tetrabutylammonium perchlorate (99.0%, Fluka) was dissolved in solutions with concentration of 0.1 M as supporting electrolyte. Both DCM and DMF were further dried by going through a column of activated alumina and purged with argon for 30 min prior to use. DPV plots of the single porphyrin Por were recorded in both solutions for comparison.
UV-visible absorption spectroscopy was conducted using a Shimadzu UV-3600 spectrophotometer. Solutions in DMF with concentration of 1 mM were used for all compounds.
PL spectra were recorded with a fluorescence (Fluorolog w FL3-221, Horiba) spectrometer using a single photon counting interface (FluoroHub) at room temperature. Most of the compounds were measured in DMF, except for CbBTD and PorR which were measured in tetrahydrofuran (THF, 99.99%, Honeywell) with concentration of 1 mM. The slit width for emission measurement was chosen as 3 nm to get modest numbers of counts per second (CPS).
Current density-voltage (J -V) measurements were performed using a simulated 100 mW cm 22 AM 1.5G solar simulator (TriSOL, OAI) coupled to a Keithley 2400 source measure unit. A 20 min light soaking and three J-V measurements were applied to the finished devices before the last recording. All the measurements used a 6 mm Â 6 mm mask to set the illumination area. Incident photon-to-current conversion efficiency (IPCE) and light harvesting efficiency (LHE) were tested after the J-V measurements. IPCE measurement used a QEX10 quantum efficiency measurement system (PV measurements), which sets the beam dimensions to be smaller than the active area of the devices [8]. LHE was calculated from the UV-visible absorbance of the sensitizers on TiO 2 films with a small amount of electrolyte injected [6,23]. Absorbed photon-to-current conversion efficiency (APCE) was calculated according to the following equation:

Results
DPV plots of all the compounds are shown in figure 2. Those of CbTPA and PorY were reported in [13], while those of Por, CbTh and PorO were reported in [8]. The DPV plots of CbBTD and PorR are for the first time reported here. For the organic chromophores CbTPA, CbTh and CbBTD, no to two oxidation and reduction peaks are observed. CbTPA shows two oxidation reactions while no reduction in the potential window. CbTh and CbBTD both show one broad oxidation peak at a similar potential, whereas only CbBTD displays an intense reduction peak at around 21.65 V versus Fc/Fc þ . The porphyrins, on the other hand, typically show two oxidation and two reduction peaks, except for Por and PorO measured in DMF.
It is noted that different solvents have been used for the DPV measurements. DCM is better for electrochemical characterization because it is easier to get rid of moisture compared with DMF. However, PorO faces solubility problem in DCM possibly due to the more planar structure of the carbazole fused-thiophene. Therefore, the characterization of PorO, PorR and their organic counterparts used DMF as solvent.
The change of solvents from DCM to DMF resulted in a 0.1 eV variation of reduction potentials on Por (Por to Por#). This solvent effect is also observable for the di-chromophoric porphyrins, where the first reduction potential of PorO and the second reduction potential of PorR show 0.1 eV negative compared to the first reduction potential of PorY. Figure 3 displays the UV-visible absorption spectra of all the compounds in DMF. As indicated by figure 3a, the UV -visible absorptions of the three organic compounds are gradually red shifted from CbTPA to CbTh and to CbBTD, showing the onset wavelength of absorption from 400 to 450 nm and to 600 nm. Their peak molar extinction coefficients 1, however, reduce from 0.55 Â It is observed that the absorption spectra of PorY, PorO and PorR closely resemble the superposition of the absorption spectra of their individual chromophores. This suggests the non-conjugated nature of these di-chromophoric dyes, which is a different molecular engineering approach from those conjugated sensitizers with strong push-pull effect by introducing functional groups with electron donating and electron withdrawing abilities; the latter approach is often characterized by a red-shift in light absorption spectrum [24,25]. An approximately 5 nm broadened Soret band is also observed in the di-chromophoric dyes when compared to Por, which is due to the effect of geometric distortion of the porphyrins [13,26].
Steady-state PL measurements were performed in solution at three different excitation wavelengths (280 nm, 480 nm and 560 nm), the latter two selectively exciting each of the different chromophores. Figure 4 shows the PL spectra of the investigated compounds in solution. Normalized PL intensity with arbitrary unit is shown in electronic supplementary material, figure S1.

Photoluminescence of Por and organic chromophores at three different excitation wavelengths (figure 4a)
Photo-excitation of Por at 280 nm leads to two PL bands between 570 and 700 nm, which were assigned to the radiative decay from S 1 to the ground state S 0 . Such a spectrum is characteristic of a Zn porphyrin chromophore reported in the literature [27,28]. The emission from CbTPA being excited at 280 nm is seen in the wavelength range 420-700 nm. It is noted that this range overlaps with part of the Soret band and the entire Q band absorption of Por (figure 3). The peak PL emission intensity of CbTPA is about four times higher compared with that of Por, whereas the absorption of CbTPA is only two times higher compared with that of Por at 280 nm (figure 3), which indicates that the organic chromophore is a stronger emitter compared with the porphyrin. Photo-excitation of CbTh at 480 nm results in a PL band at 530 nm, which also covers the entire Q band absorption of Por. The PL band of CbBTD appears in the 580 -800 nm wavelength range when photo-excited at either 480 or 560 nm.

Photoluminescence of PorY, PorO and PorR at three different excitation wavelengths (figure 4b)
Photo-excitation of PorY leads to weak PL bands at 350-480 nm with peak intensity value of 0.05 Â 10 5 CPS assigned to the emission of Y unit. In addition, relatively strong bands with peak intensity value of 1.4 Â 10 5 CPS at 600 -700 nm assigned to the porphyrin emission are observed.  When PorO is photo-excited at 480 nm, the emission from the organic chromophore O at 530 nm is much weaker compared to that of CbTh, whereas the emission assigned to the porphyrin core at 600 -720 nm wavelength range shows a peak value of 6 Â 10 5 CPS. Note that the absorption of porphyrin core at 480 nm is marginal as indicated by figure 3 and the emission of Por at 480 nm photo-excitation is negligible ( figure 4b). The appearance of porphyrin emission in PorO suggests a substantial FRET from the organic chromophore O to the porphyrin core.
The PL band of PorR shows similar shape for photo-excitation at both 480 nm and 560 nm. It is noted that the porphyrin core in PorR only weakly absorbs photons at 480 nm; this similar PL band to the emission of PorR at two excitation wavelengths indicates that the emission of PorR originates mainly from the organic chromophore R. Compared to that of CbBTD in figure 4a, the PL intensity of PorR shows a threefold greater emission when photo-excited at 560 nm. This difference in PL intensity is attributed to the differences in the absorption coefficients of PorR (0.58 Â 10 5 M 21 cm 21 ) and CbBTD   The enhanced photovoltage of PorY-and PorO-sensitized solar cells is due to the employing of the non-conjugated linker between the organic chromophore and the porphyrin, adding steric effect thus increasing the electron lifetime [8,13]. This bulky effect might also be applicable in PorR-sensitized solar cells as compared to Por-sensitized solar cells; however, previous measurements under reduced dye loading condition suggest that the strong dispersion forces between the organic chromophore R and the redox couple in electrolyte contribute significantly to the longer electron lifetime of the former [9].
The change in photocurrent of these DSSCs has been analysed individually using the IPCE-LHE-APCE approach in our previous studies [9,13]. However, the impact of energy/electron transfer between the two chromophores on electron injection has not been studied. Since electron injection efficiency of DSSCs is roughly same as the APCE value when using thin film electrode where charge collection efficiency is uniform, APCE values of DSSCs using the investigated porphyrins are measured as shown in figure 6 and averaged APCE values at Soret and Q bands of the individual dyes are listed in table 1.
The APCE value of DSSC in the Q band absorption region when using PorY is 90%, significantly higher compared to when Por is employed (60%). The APCE value of DSSC using PorO is 60% in the Q band absorption region, which is very similar to that of DSSCs using Por. The broad shoulder between 450 nm and 500 nm is attributed to photo-excitation of the organic chromophore. PorR shows     Figure 7. Energy level diagram of (a) the second organic chromophores and (b) the porphyrins derived from the peak potentials of differential pulse voltammograms [1]. Energy levels were calculated from UV-visible absorption spectra by the equation reduction of PorR is assigned to the reduction of the organic chromophore R. The low-lying reduction potential of the organic chromophore R in PorR enables a possible intramolecular electron transfer from the photo-excited porphyrin to the chromophore with the electron acceptor character of the benzothiadiazole unit. The E Ox 1st of Por, PorY, PorO and PorR is within 0.10 eV variation of each other (25.40 eV versus Fc/ Fc þ ). In fact, all the first oxidation peaks of the di-chromophoric sensitizers are comprised of a two-electron transfer reaction, forming a broad and intense current response plot (figure 2). Spectro-electrochemical (SEC) measurements were carried out in the same solution as that of the electrochemical characterization, in order to distinguish these first oxidation peaks, as shown in electronic supplementary material, figure S2. These SEC experiments, however, suggest that for the dichromophoric sensitizers, distinguishing the origin of the E Ox 1st would be difficult. The sensitizers are more likely to change to more planar structures after oxidation when compared with those in neutral state.

Energy/electron transfer pathway at dye-sensitized films
As indicated by the PL spectra in figure 4, energy transfer from the organic chromophore to the porphyrin is possible in both PorY and PorO because the requirements of FRET are satisfied [14,30,31]. In PorY, both Por and Y chromophore are photo-excited at 280 nm. Energy transfer from Y* to Por increases the concentration of Por*, therefore enhancing emission from Por* and reducing emission from Y* to nearly unobservable. Similar to PorO, the marginal emission from O* as well as the strong emission from the Por at 480 nm photo-excitation both imply energy transfer from O* to Por. However, energy transfer from the organic chromophore R to the porphyrin core does not likely occur in PorR. When photo-excited at 480 nm, there is no observable PL band assigned to the porphyrin from PorR emission as shown in figure 4b. In addition, energy transfer from the porphyrin to the organic chromophore in PorR is also not favoured as suggested by the similar PL intensity of PorR and CbBTD after considering the absorbance of two molecules in THF at 560 nm. The emission quenching of Por in PorR when photo-excited at 560 nm, therefore, possibly originates from a PET pathway with a non-radiative back electron transfer as proposed below.
When photo-excited at 560 nm, the electron transfer from the porphyrin core to the organic chromophore in PorR may lead to an oxidized porphyrin cation radical (Por †þ ) and a reduced carbazole-thiophene benzothiadiazole thiophene moiety (CbBTD †2 ). The two ions might form an exciplex (Por †þ †CbBTD †2 )* under a PET mechanism or electron exchange (Dexter mechanism). The non-radiative back electron transfer (Por †þ $ CbBTD †2 ) results in no observation of PL from Por* in PorR in figure 4b. The distance between the absorber (Por) and the quencher (CbBTD or R) allows efficient electron transfer because they are in one molecule.
In addition, emission spectra of both PorR and CbBTD were measured in DMF solvent under the same experimental condition. Compared to that in THF solvent, PL intensities of both PorR and CbBTD in DMF are much weaker (electronic supplementary material, figure S3). Such a strong solvent effect suggests charge transfer nature of the emissive state in both molecules: in DMF with stronger polarity, charge transfer state is more stabilized leading to less emission, compared to the less polar THF [32]. Hence, PET pathway is proposed here as the main approach that quenches the Por emission in PorR.
It should be pointed that identifying the contributions of FRET or electron transfer for the PL quenching requires detailed measurements of PL lifetimes, quantum yields and understanding of electron transfer mechanisms [33,34]. While these investigations are quite demanding on both instrument and elaboration, the work presented here provides the primary study and correlates the possible energy/electron transfer mechanism to electron injection efficiency in DSSCs.
The proposed energy/electron transfer pathways at the three di-chromophoric porphyrin-sensitized TiO 2 films are shown schematically in figure 8. Energy transfer from the organic chromophore to the porphyrin, in either PorY or PorO sensitized TiO 2 , is expected to contribute to charge generation at wavelengths where the organic chromophores are selectively excited. PET from the porphyrin to the organic chromophore in PorR sensitized TiO 2 , on the other hand, may decrease the charge injection efficiency leading to lower APCE.
The more significant APCE value of DSSC using PorY as compared to that of Por has been attributed to increased injection efficiency due to the increased bulkiness of PorY and lower concentration of 'noninjecting dye' [13]. Since porphyrin, electrolyte and TiO 2 all absorb at around 400 nm, selective photoexcitation of the organic chromophore in DSSC is not feasible. Therefore, it is not possible to evaluate the effect of energy transfer from the photo-excited Y to Por. The relative values of APCE below 400 nm did not decrease for DSSCs sensitized with PorY compared with that of Por (figure 6), which rsos.royalsocietypublishing.org R. Soc. open sci. 5: 181218 suggest that energy transfer does lead to photo-generation with the same overall quantum yield (w a 0 ¼ w b , PorY in figure 8).
A broad shoulder between 450 and 500 nm in APCE of PorO-sensitized solar cell is attributed to photo-excitation of the organic chromophore O. The APCE values at this wavelength range are lower compared with that at the Soret band, but on average, higher than that at the Q bands suggesting similarly efficient photo-generation from the photo-excited organic chromophore compared to the photo-excited porphyrin (w a 0 ¼ w b , PorO in figure 8). From this, it is not possible to distinguish between an energy transfer step from the O to the porphyrin core, followed by electron injection from the porphyrin or direct electron injection from the organic chromophore. However, the injection quantum yield from the two chromophores in PorO is similar.
The low APCE at the Soret band in DSSC using PorR suggests an intense decreased electron injection efficiency, which is quite possible, as discussed above, due to electron transfer to the organic chromophore competing with electron injection into the conduction band of TiO 2 from the excited porphyrin core (decreased w b of PorR in figure 8). Additionally, the APCE values were even lower at wavelengths above 660 nm, where the organic chromophore absorbs ( figure 6). This suggests that the charge generation efficiency of PorR is lower compared to that of Por (electron transfer with efficiency of w a 0 of PorR in figure 8). The poor photocurrent of PorR-sensitized solar cell in figure 5 is also assigned to this competing electron transfer pathway.

Challenge in multi-chromophoric sensitizer design
The low APCE of PorR-sensitized DSSC is not due to some other effects such as dye aggregation [35,36], the conduction band edge potential of TiO 2 being too high (diminishing the driving force for electron injection) [37] or sub-nanosecond recombination between the oxidized dye and surface-trapped electrons [38,39]. The electron injection efficiency of a di-chromophoric dye-sensitized electrode is similar to that using chenodeoxycholic acid as a co-adsorber in Por-sensitized electrode, which suggests that rationally designed di-chromophoric dyes may intrinsically prevent aggregation because of their non-conjugated property [13]. Some other organic sensitizers with twisted structure also suggested an aggregation-free property [40]. As investigated by stepped-light induced measurements of photocurrent and photovoltage, the TiO 2 conduction band edge potential of PorR-sensitized DSSC was identical to that of Por-sensitized DSSC, which suggests that the driving forces for electron injection of the two systems are similar [9]. The fast recombination component at the dye-sensitized semiconductor surface is assigned to the strong coupling between the surface trap state and the oxidized dye [  and PorR use the same porphyrin structure with typically less than 200 fs electron injection kinetics as coupling to the TiO 2 conduction band [39] and secondly, steady-state PL studies in figure 4 suggest similar PL intensity within the three di-chromophoric dyes, which may suggest similar excited state lifetime. The observations and discussions above proposed a major challenge in designing multi-chromophoric dyes for application in DSSCs. To increase the light absorption of sensitizers, chromophores with narrow frontier orbital energy offsets are good candidates, while the position of these chromophores in multichromophoric dye molecules needs to be carefully considered. Attaching such a chromophore further away from the TiO 2 surface, like in the case of PorR dye, can significantly reduce the power conversion efficiency as this opens a competitive charge transfer pathway. This may lead to poor photocurrent generation in a DSSC due to the insufficient electron injection or low photovoltage as a result of the strong intermolecular forces at the TiO 2 /dye/electrolyte interface [42,43].
One possible approach to solve this problem is to covalently attached the low energy gap chromophore further away from the TiO 2 surface and to better couple the excited state of the low energy gap chromophore with the conduction band electrons in TiO 2 . In this way, remote electron injection from the redshifted chromophore would increase, while the recombination between the injected electrons with the electrolyte would be harder because of the larger dispersion forces between the low energy gap chromophore and the electrolyte. Another possible approach would be to design a low energy gap chromophore with high electron injection efficiency and to tailor the multi-chromophoric dyes with long-lived charge separation [44,45]. These approaches, albeit accompanied with a few trade-offs, would be interesting research areas for novel dye molecule engineering for advanced photocurrent conversion devices.

Conclusion
Three di-chromophoric dyes containing a porphyrin and an organic chromophore with gradually decreased energy levels were characterized both in solution and on TiO 2 surface. The energy/electron transfer mechanisms within the two chromophores in three di-chromophoric porphyrins have been characterized using steady-state PL spectroscopy measurements in concert with APCE measurements. Similar charge generation efficiency from the two chromophores was suggested in PorY and PorO sensitized TiO 2 electrodes, in which energy transfer occurred from the organic chromophore to the porphyrin. When a low energy gap organic chromophore was introduced to the porphyrin-based di-chromophoric sensitizer, electron transfer, rather than an energy transfer, from the excited porphyrin to the organic chromophore was strongly suggested. This electron transfer pathway in PorR-sensitized TiO 2 film competes with electron injection from the porphyrin to TiO 2 conduction band, leading to a low APCE of only 30%. These findings about the charge generation mechanism demonstrate the possible limitations in multi-chromophoric sensitizers and provide valuable information for the next generation of multi-chromophoric dye molecule structure design.
Data accessibility. All data supporting the findings of this study are included in the article and electronic supplementary material.