Molecular engineering of several butterfly-shaped hole transport materials containing dibenzo[b,d]thiophene core for perovskite photovoltaics

Several butterfly-shaped materials composed of dibenzo[b,d]thiophene (DBT) and dibenzo-dithiophene (DBT5) cores were designed as hole transporting materials (HTMs) and their properties were studied by density functional theory (DFT) computations for usage in mesoscopic n-i-p perovskite solar cells (PSCs). To choose suitable HTMs, it was displayed that both of lowest unoccupied molecular orbital (LUMO) and highest occupied molecular orbital (HOMO) energies of molecules were located higher than those of CH3NH3PbI3 (MAPbI3) perovskite as they were able to transfer holes from the MAPbI3 toward Ag cathode. Negative solvation energy (ΔEsolvation) values for all HTMs (within the range of − 5.185 to − 18.140 kcal/mol) revealed their high solubility and stability within CH2Cl2 solvent. The DBT5-COMe demonstrated the lowest values of band gap (Eg = 3.544) and hardness (η = 1.772 eV) (the greatest chemical activity) and DBT5-CF3 displayed the biggest η = 1.953 eV (maximum stability) that were predominantly valuable for effective HTMs. All HTMs presented appropriately high LHEs from 0.8793 to 0.9406. In addition, the DBT5 and DBT5-SH depicted the lowest exciton binding energy (Eb) values of 0.881 and 0.880 eV which confirmed they could produce satisfactory results for the PSCs assembled using these materials. The DBT5-SH and DBT5-H had maximum hole mobility (μh) values of 6.031 × 10–2 and 1.140 × 10–2 which were greater than those measured for the reference DBT5 molecule (μh = 3.984 × 10–4 cm2/V/s) and about 10 and 100 times superior to the calculated and experimental μh values for well-known Spiro-OMeTAD. The DBT5-COOH illustrated the biggest open circuit voltage (VOC), fill factor (FF) and power conversion efficiency (PCE) values of 1.166 eV, 0.896 and 23.707%, respectively, establishing it could be as the best HTM candidate for high performance PSCs.

www.nature.com/scientificreports/ the biggest V OC , FF and PCE values of 1.166 eV, 0.896 and 23.707%, respectively. Therefore, this material was suggested as a highly efficient candidate for PSCs fabrication which could possibly lead to higher PCEs than that measured using the pure DBT.

Methods of computations
The DFT geometry optimization computations were carried out in CH 2 Cl 2 solution on several butterfly-shaped HTMs 1-11 at B3LYP-D3/6-31G(d,p) method by Gaussian 09 software 45 . The conductor-like polarizable continuum model (C-PCM) method was adopted to estimate solvent influence 46 . It is noteworthy that B3LYP-D3/6-31G(d,p) method was employed herein as it is commonly utilized by many researchers and known as a method of high validity affording appreciated data in real-time and this makes it a reliable and economical method. As an example, electrochemical features of HTMs for solar devices were predicted using the DFT computations at B3LYP-D3/6-31G(d,p) level [47][48][49] . The infrared (IR) spectra were obtained by the freq keyword in DFT calculations until no imaginary (negative) frequencies were obtained. Results of frequency calculations on HTMs 1-11 depicted absence of imaginary frequencies. The UV-Visible (UV-Vis) and photoluminescence (PL) spectra were attained through time-dependent density functional theory (TD-DFT) calculations at the B3LYP-D3/6-31G(d,p) method in which number of states was 30, i.e. TD = (nstates = 30). GaussView 5 software was used to obtain the contour and surface maps, IR, UV-Vis and PL spectra, and molecular orbitals 50 .
Inside an organic π-conjugated molecule, the charge transference occurs by non-coherent hopping mechanism due to electron-phonon coupling is noticeably stronger than electronic coupling under ambient conditions. According to Marcus theory, Eq. (1) is used to estimate the hole hopping rate (k h ), where h, k B , T, V and λ, respectively, reveal Planck's constant, Boltzmann constant, absolute temperature, transfer integral and reorganization energy 51 .
The λ value can be achieved using Eq. (2), where E 0 and E + , respectively, signify energies of optimized neutral and cationic molecules whereas the E * 0 and E * + respectively show energies of neutral and cationic samples at cationic and neutral states 52 .
The transfer integral (V) is computed by Eq. (3), in which e ii , e jj exhibit site energies of ψ i |Ĥ|ψ i , ψ j |Ĥ|ψ j while S ij and J ij denote ψ i |ψ j overlap matrix element and ψ i |Ĥ|ψ j transfer integral, respectively 53,54 .
The hole mobility (μ h ) is calculated by Eq. (4), where D and e indicate diffusion coefficient and electron charge, respectively. The D is obtained by Eq. (5), where r m , d and m show centroid to centroid distance, spatial dimension, specific hopping route, p m = k m / m k m displays probability of hopping process 55 .
The adiabatic ionization potential (IP) was measured by IP = E + − E 0 equation, where E 0 and E + illustrated, respectively, energies of neutral and cationic optimized molecules 56 .
The crystalline structures of HTMs 1-11 were simulated by means of polymorph module within Material Studio software 57 through computing ten most important space groups of Cc, C2, P-1, C2/c, Pbcn, P2 1 , P2 1 /c, Pbca, P2 1 2 1 2 1 , Pna2 1 . To predict all crystal structures and unit cells, the optimized structure is firstly entered into the Materials Studio and then its ESP charges are computed using DMol3 module/calculation (Task: Energy, Basis set: DNP, Properties: Electrostatics and population analysis are selected). Then, the total charges are set to zero using DMol3 module/Analysis/Population analysis. Secondly, all crystal structures and 10 unit cells are predicted by Polymorph module/calculation with choosing the parameters as force field: Dreiding, charges: use current (ESP), quality: Fine, electrostatic: Ewald, van der Waals: Ewald and Space Groups: thick all items. Finally, the Polymorph module/Analysis is selected and all of the predicted crystal structures and unit cells are listed which will be sorted based on their total energies and the most stable crystal structure (of minimum total energy) is selected.

Results and discussion
Binding energy, polarizability and solubility. In this work, several butterfly-shaped HTMs were designed for mesoscopic n-i-p PSCs via addition of another thiophene ring onto the DBT core so that 11 molecules were obtained in which different substituents were placed onto the diphenylamine groups. The structures HTMs 1-11 ( Fig. 1) were optimized using B3LYP-D3/6-31G(d,p). Moreover, Fig. 2 exhibits that diverse sub-   Figure 3 presents electrostatic surface potentials (ESPs) for compounds 1-11 that shows charge distributions onto the surfaces of such butterfly-shaped structures. The ESPs demonstrate electrical charge distributions which can be a measure of molecular polarity so that a more symmetric ESP exhibits that the material has a smaller dipole moment. Furthermore, the ESP may be accounted for the structural stability as a more symmetric charge dispersion may reveal that weaker inter-/intra-molecular interactions occur and the molecule is less reactive. Thus, the DBT5-COMe indicating the most asymmetric charge distribution has the utmost reactivity but the DBT5-H, DBT5-CN and DBT5-CF 3 samples with the most symmetric charge distributions are the most stable structures. Such results are in consistent with the contour maps achieved for these materials.

Contours and surfaces.
The contour maps in Fig. 4 illustrate charge distributions over the butterfly-shaped structures 1-11 so that positive and negative charges, respectively, are displayed by yellow and red colors. It is seen that the most asymmetric charges distribution happens for DBT5-COMe which results in the biggest dipole moment of 8.5837 D for this sample, see Table 1. Indeed, molecules demonstrating more symmetric charge distributions have smaller dipole moments and vice versa. As a result, the DBT5-H, DBT5-CN and DBT5-CF 3 with the lowermost dipole moments of 0.0030, 0.0028 and 0.0017 D, respectively, illuminate the most symmetric charge distributions. Electronic properties. It is known that HOMO and LUMO energy levels of HTM influence its hole transfer property and in turn affect performance of solar cell. The energy levels diagrams of FTO, TiO 2 , MAPbI 3 perovskite, samples 1-11 and Ag cathode are presented in Fig. 5. Notably, the HOMO energy of an appropriate HTM for the PSC must be positioned upper than the valence band energy of MAPbI 3 (− 5.43 eV) 51 . As the HOMO levels of all samples except for the DBT5-CN (− 5.55 eV) are located higher than MAPbI 3 valence band, all molecules but DBT5-CN are beneficial HTMs with appropriate energy level alignments which lead to successful hole transport from MAPbI 3 toward the HTMs. Also, the deepest HOMO energies of − 5.39, − 5.36 and − 6.35 eV belong to the DBT5-COOH, DBT5-COMe and DBT5-CF 3 molecules, respectively, which may result in easier hole transfer from MAPbI 3 to the HTMs and the Ag cathode.
As the HOMO energies of DBT5-COOH, DBT5-COMe and DBT5-CF 3 samples are deeper than HOMO level of Spiro-OMeTAD (− 5.09 eV) 47   The LUMO energy of a favorable HTM of a PSC must be located higher than MAPbI 3 conduction band (− 3.93 eV) 60 in order to stop backward movement of photo-created electrons from MAPbI 3 toward the Ag cathode. Figure 5 shows that all LUMO levels are positioned at upper energies than MAPbI 3 conduction band confirming they are suitable materials which effectively inhibit the electron transport from perovskite to the cathode.  www.nature.com/scientificreports/ over the whole molecules but the central dibenzodithiophene fragment. It may be suggested that the molecules with higher HOMO distributions compared to the LUMO orbitals may indicate more effective hole transport properties than the electron transfer effects. Figure S2 shows the density of states (DOS) spectra for molecules 1-11 in which the green and blue regions depict occupied and virtual orbitals. As well, there is a gap between occupied and virtual orbitals which is called bandgap (E g ) which is HOMO-LUMO energy. The E g values in Table 2 exhibit that DBT5-CF 3 and DBT5-COMe respectively have the largest and the lowest bandgaps of 3.906 and 3.544 eV. As well, the E g value enhances from 3.550 eV (in DBT) to 3.574 eV (in DBT5) signifying addition of another thiophene ring leads to increasing the E g . Comparing the similar DBT5-OMe, DBT5-OEt and DBT5-COOH, DBT5-COMe depicts that replacement of substituent with an electron donating group decreases the E g amount.
Diagrams of HOMO and LUMO energies against Hammett para-substituent constant (σ p ) 61 are provided in Fig. S3. Apparently, almost linear lines of quite high regressions values (R 2 = 0.9271 and 0.8822) are measured for E HOMO and E LUMO diagrams versus Hammett constants confirming the latter is not a highly linear relationship and shows some deviation from linearity. Therefore, it may be stated that varying the para substituents substantially affects the HOMO levels whereas the LUMO energies are not greatly changes. Moreover, the non-linear diagram

Molecular descriptors.
To explore electronic and chemical properties of compounds 1-11, their molecular were attained and the results are provided in Table 2. Three η = (I-A)/2, χ = (I + A)/2 and µ = -χ formula were used to obtain reactivity and structural stability, where η, µ and χ represent global hardness, chemical potential and electronegativity, respectively. Besides, I = − E HOMO and A = − E LUMO , respectively, specify vertical ionization and electron affinity 62 . Usually, lower E g , µ and η amounts exhibit more chemical reactivity and accelerated charge transport for a material. Furthermore, an electrophilic molecule reveals a superior electrophilicity index which is estimated using the formula ω = µ 2 /2η. It is observed in Table 2 that the DBT5-COMe shows the lowermost E g = 3.544 eV plus η = 1.772 eV but a medium µ = − 3.585 eV. On the other hand, DBT5-CF 3 displays the utmost E g = 3.906 eV and η = 1.953 eV but a moderate µ = − 3.397 eV. Therefore, it can be proposed that the DBT5-COMe and DBT5-CF 3 have the maximum and minimum chemical activities with perovskite and Ag cathode materials, respectively. Additionally, the DBT5-OEt with the lowest µ = − 2.752 eV may display moderate reactivity. Moreover, the DBT5-OEt illustrates the smallest ω = 2.101 eV approving its lowest electron affinity which is beneficial for an efficient hole transport material.
The chemical stability of samples 1-11 was evaluated by comparing their hardness (η) values so that a material with a greater η exhibits a superior chemical stability. As can be seen, the η values vary as X = DBT5-   Table 2. As a suitable HTM for the PSC device, it should indicate the least electron affinity due to it must accept the hole from the MAPbI 3 perovskite and transport it toward Ag cathode. The lowest and the highest electron affinities of 0.950 and 1.813 eV are attained for the DBT5-OEt and DBT5-COMe, respectively. Thus, the DBT5-OEt may be chosen as the best HTM considering the A values.
A comparison of molecular descriptors obtained for compounds 1-11 allows to offer the most promising HTM. Briefly, the DBT5-COMe shows the lowermost E g = 3.544 eV plus η = 1.772 eV indicating its maximum chemical activity. The DBT5-OEt has the lowest ω = 2.101 eV which reveals it has the least electron affinity. The  IR, UV-Vis and PL spectra. The IR, UV-Vis absorption and PL emission spectra of samples 1-11 were achieved to study their functional groups and optical properties. Figure 7a demonstrates the IR spectra of compounds 1-11 which show by varying substituents, several peaks with diverse intensities are appeared. The bands at about 500-800 cm −1 correspond to bending of = C-H bonds. Besides, some peaks placed near 900, 1000, 1050, 1100, 1150 cm −1 can be attributed to vibrational stretchings of C-C, C-S, C-N, C-O, C-F bonds, respectively 63 .
The bands situated at around 1350 and 1550 cm −1 are due to asymmetric and symmetric vibrational stretchings of C=C bonds 64 . The bands located near 3700 and 1650 cm −1 are correlated to vibrational stretching and bending, respectively, of O-H bonds but the peaks at about 3000 cm −1 are owing to stretching of C-H bonds 65 . The UV-Vis spectra of all samples are shown in Fig. 7b and a summary of spectral data including oscillator strengths (f), maximum absorbance wavelengths (λ abs max ), main transitions and LHEs are presented in Table 3. It is seen that all compounds exhibit two peaks within the range of about 250 to 550 nm except for four molecules DBT5-CN, DBT5-CF 3 , DBT5-COOH and DBT5-COMe containing electron withdrawing substituents which only indicate one peak. Furthermore, three DBT5-CN, DBT5-COOH and DBT5-COMe compounds exhibit the highest peak intensity and among them the DBT5-COMe displays the highest intensity.
Evidently, λ abs max values are different for samples 1-11 and change in the range of 371.50 to 407.84 nm, respectively, for the DBT5-CF 3 and DBT, respectively, so that the main transitions associated with these peaks are H → L (97.54%) and H → L (97.77%). In addition, the LHE values of all molecules are high which confirm they have great capabilities of light absorption. The LHE of similar DBT5-OMe, DBT5-OEt and DBT5-COOH, DBT5-COMe molecules elucidate that replacing the substituents with electron donating groups increases the LHE amounts. The LHE values vary from 0.8793 (for DBT5-CF 3 ) to 0.9406 (for DBT) validating all of these materials are beneficial HTMs for PSCs. Figure 7c displays the PL emission spectra of molecules 1-11 and Table 4 lists E em max , λ em max , f em max , Stokes shifts, Exciton binding energy (E b ) and radiation lifetime (τ) of HTMs 1-11. The PL spectra of DBT5-OH, DBT5-CN, DBT5-CF 3 , DBT5-COOH and DBT5-COMe show one broad peak whereas other molecules demonstrate two maxima or a sharp peak plus a weak shoulder while a broad peak plus a shoulder. The λ em max values change from 406.18 nm (in DBT5-COOH) to 469.85 nm (in DBT). Additionally, all λ em max values are greater than their associated absorption λ abs max values. As well, greater λ em max and λ abs max are obtained for the DBT compared to that of DBT5 which lead to greater optical bandgaps for DBT5. A comparison of DBT5-OMe, DBT5-OEt and DBT5-COOH, DBT5-COMe similar HTMs approves that the λ em max and λ abs max values have red shifts by replacing the substituents with electron donating groups. These results are in consistent with the E g values measured for these samples.
Stokes shift exhibits the wavelengths difference of the UV-Vis and PL peaks. Hence, Stokes shift is larger when PL peak shows a greater red shift toward visible spectral region. The Stokes shift corresponds to losing of energy by the absorbed photons via a non-radiative mechanism that results in decreasing average energy of emitted photons. Notably, the radiative PL emission is correlated to the recombination of holes with electrons that is unfavorable in photovoltaics. Thus, superior Stokes shifts upon non-radiative PL emissions are more advantageous for PSC devices. As can be seen, the lowest Stokes shift of 24.67 nm is achieved for DBT5-COOH while the highest value of 64.64 nm is attained for both DBT5-OH and DBT5-OEt samples. Also, adding another thiophene ring to the DBT molecule decreases the Stokes shift from 62.01 in DBT to 53.29 nm in DBT5. Moreover, the Stokes shifts increase by replacement of X substituent with electron donating moieties in DBT5-OMe, DBT5-OEt as well as DBT5-COOH, DBT5-COMe.  www.nature.com/scientificreports/ The radiation lifetime (τ) values were provided to estimate lifetimes of radiative recombination between holes and electrons, i.e. greater amounts exhibit longer recombination processes but smaller values illustrate shorter recombination. Table 4   .238 ns illuminates that radiation lifetime enhances by replacement of X substituent with a more electron donating group. Also, the DBT5-H (0.097 ns) with the neutral substituent reveals a smaller radiation lifetime than those of the DBT (0.118 ns) and DBT5 (0.121 ns). Notably, the DBT5-CF 3 including electron withdrawing groups shows a small τ = 0.082 ns indicating its favorable short electron-hole recombination that is highly suitable to improve the PSC efficacy. The binding energies of Excitons (electron-hole couples) are measured and showed by E b values in Table 4. The E b amount corresponds to the PSC performance, i.e. a higher E b illustrates stronger Coulombic attraction of electron and hole that causes difficult separation of Exciton binding, less current density and lower PSC efficiency. Noticeably, the lowest and the largest E b values of 0.880 and 1.465 eV are obtained for the DBT5-SH and DBT5-COOH, respectively. In addition, the DBT5 and DBT5-H depict low E b values which confirm they can lead to satisfactory results for the PSCs assembled using these materials. Notably, substituent replacement by electron withdrawing groups in DBT5-COOH as well as DBT5-COMe enhances the E b amount. As a result, Exciton binding separation of happens the easiest in DBT5-SH which may afford the utmost current density for the PSCs.
Hole transport properties. The hole mobility of a material has a great influence onto PSC performances as it changes both the V OC and current density of photovoltaic devices. The hole mobility data of HTMs were achieved by computing ten crystalline structures for each sample to obtain the most stable (lowest total energy) molecule that was selected to calculate the hole mobility. Figure 8 illustrates the most stable predicted unit cells for crystalline structures of all samples. Besides, crystallographic data for the most stable crystal structures of samples 1-11 are provided in Table 5. The dimer structures of compounds 1-11 with the DBT5 core were used to estimate their hole hopping characteristics (Fig. S4) that demonstrate end-to-end configuration for all samples but the face-to-face configuration for 6. It is notable that in face-to-face arrangement, the π-π stacking intermolecular interactions are able to boost electronic couplings.
It should be reminded that the HOMO and LUMO energy levels in Fig. 5 showed that the five molecules including DBT, DBT5-OH, DBT5-OMe, DBT5-OEt and DBT5-CN do not have suitable band alignments and cannot transfer holes from the MAPBI 3 to the Ag cathode. However, we calculated the hole mobility data for all www.nature.com/scientificreports/ samples 1-11. The hole mobility (μ h ), hole reorganization energy (λ h ), hole mobility rate (k h ), electron coupling (V), and centroid-centroid distance (r) of samples with favorable and unfavorable band alignments are given in Table 6 and Table S1, respectively. Among samples with favorable band alignments listed in Table 6, the biggest and smallest λ h values of 0.1403 and 0.3729 eV, respectively, are measured for DBT5 and DBT5-COOH. Besides, k h and μ h exhibit alike changes by altering the HTMs and decrease in the order of DBT5-SH > DBT5-H > DBT5-COMe > DBT5-CF 3 > DBT5 > DBT5-COOH. Noticeably, the samples composed of SH and H substituents show the greatest μ h of 6.031 × 10 -2 and 1.140 × 10 -2 cm 2 /V/s whereas other molecules exhibit smaller k h and μ h amounts. Notably, the hole mobility achieved for the DBT (7.805 × 10 -2 cm 2 /V/s) is greater than the value measured for the champion DBT5-SH material but it is observed in Fig. 5 that the DBT does not display a suitable band alignment with respect to the Ag cathode electrode. In addition, the μ h values of all compounds except for the DBT5-COOH in Table 6 are greater than that of the reference DBT5 molecule containing the SMe substituent. Comparing the two similar DBT5-COMe and DBT5-COOH materials exhibits that the μ h value is almost 100 times greater for the latter. The calculated and experimental μ h values of Spiro-OMeTAD, respectively, are 5.65 × 10 -3 and 4.53 × 10 -4 cm 2 /V/s 47 . When the μ h data of the DBT5-SH and DBT5-H are compared with those measured for the Spiro-OMeTAD, it  , where FF and P in show the fill factor and incident power (100 mW/cm 2 ) while V OC and J SC respectively exhibit the open circuit voltage and short circuit current density. The PCE greatly boosts by enhancement of both V OC and J SC which are the highest voltage measured at zero current density and maximum current density at zero voltage, respectively. To estimate the PCEs of PSCs based on HTM samples with suitable band alignments, the experimentally reported J SC = 22.7 mA/cm 234 was used and the V OC values were estimated using formula VOC = ELUMO of acceptor − EHOMO of donor − 0.3/e 67 , in which MAPbI 3 and HTM are electron acceptor and donor materials, e stands for the unit electronic charge and 0.3 shows a constant signifying voltage decrease. The conduction band of MAPbI 3 (− 3.93 eV) 60 and the HOMO energies of HTMs were utilized to estimate V OC values. Hence, a HTM indicating a deeper HOMO level can produce a larger V OC . Table 7 and Table S2 demonstrates the photovoltaic performance parameters for HTM samples. It is found that the DBT5-COOH affords the maximum V OC = 1.166 eV confirming it can be the most effective HTM for the PSCs. Additionally, the V OC enhances from 0.476 eV (in DBT5) to higher values in other HTMs indicating substitution of SMe groups by other groups increases the V OC value. As well, the neutral H substituent in DBT5-H affords a greater V OC (0.675 eV) than those measured for the DBT5 and DBT5-SH HTMs. Three HTMs including DBT5-COOH, DBT5-COMe and DBT5-CF 3 exhibit highest V OC amounts of 1.166, 1.127 and 1.120 eV, respectively, which verify these materials have a high capacity for application in PSCs. Accordingly, the DBT5-COOH with the utmost V OC is the most promising material for the PSC device fabrication.
The FF values are attained using the formula FF =

Conclusion
The

Data availability
The computational data will be delivered on reasonable request. If someone wants to request the data from this study, please contact Zahra Shariatinia (shariati@aut.ac.ir).