C-N linked donor type porphyrin derivatives: unrevealed hole-transporting materials for efficient hybrid perovskite solar cells

A new series of Zn(II) and Cu(II)-based porphyrin complexes 5a , 5b doubly functionalised with carbazole units, were developed to be used as hole-transporting material (HTM) in perovskite solar cells (PSCs). Those complexes were obtained via nucleophilic substitution reaction mediated by PhI(OAc) 2 /NaAuCl 4 ·2H 2 O, or using C-N transition metal-assisted couplings. The hole extraction capability of 5a , b was assessed using cyclic voltammetry; this study confirmed the better alignment of the Zn(II) complex 5a with the perovskite valence band level, compared to the Cu(II) complex 5b . The optimised geometry and molecular orbitals of both complexes also corroborate the higher potential of 5a as HTM. Photoluminescence characterisation showed that the presence of 5a , b as HTMs on the perovskite surface resulted in the quenching of the emission, matching hole transfer phenomena. The photovoltaic performance was evaluated and compared with those from reference cells made with the standard HTM spiro-OMeTAD. Optimised 5 -based devices showed improvements in all photovoltaic characteristics; their open circuit voltage ( V oc ) reached close to 1 V and short-circuit current density ( J sc ) values of 13.79 and 9.14 mA cm -2 for 5a and 5b , respectively, disclosing the effect of the metallic centre. A maximum power conversion efficiency (PCE) of 10.01% was attained for 5a , which is 65% of the PCE generated by using the spiro-OMeTAD reference. This study demonstrates that C-N linked donor type porphyrin derivatives are promising novel HTMs for developing efficient and reproducible PSCs.


Introduction
Nowadays, it is unquestionable that to meet the rapidly growing global energy demand and avoid large amounts of greenhouse gas emissions, renewable energy sources need to be exploited to the maximum extent. Indeed, solar power is the world's most abundant energy resource, and harnessing it would yield a never-ending energy supply. 1 Nonetheless, the greatest challenge has always been the development of efficient and cost-effective approaches to solar energy conversion. Photovoltaic (PV) technology is touted as the most promising pathway and the global PV market is currently being dominated by crystalline silicon solar cells, which have demonstrated a high power conversion efficiency (PCE) of 26.1%. 2 A cutting-edge breakthrough is now perovskite solar cells (PSCs). 3,4 Progress in this emerging technology is so impressive that from the first example reported in 2009 with a PCE of 3.8%, 5 12-16% in mid-2012 and 2013, to reach already a certified PCE of 26.0% in 2023. 2c Most of those achievements are due to the presence of light absorbers based on lead perovskite materials which exhibit broad absorption in the solar spectrum combined with high charge carrier mobility and low recombination. Even so, one of the most important technical improvements came from replacing the liquid redox mediator with a p-type solid-state hole-transporting material (HTM) that is crucial for long-term stability. Moreover, even if PSC devices can work without the HTM, this active layer is essential for high performance as it facilitates charge carrier separation and extraction by transporting holes from the perovskite material to the cathode whilst reducing unwanted charge recombination. Currently, typical n-i-p structured PSCs generally use 2,2',7,7'tetrakis[N,N'-di(4-methoxyphenyl)amino]-9,9'-spirobifluorene (spiro-OMeTAD) as the HTM. Nevertheless, spiro-OMeTAD production costs are high due to the laborious and complex multistep synthesis, resulting in a low yield, combined with a complicated purification process. Furthermore, additives and dopants are needed to improve spiro-OMeTAD hole mobility capacity, and those accelerate perovskite decomposition to cause detrimental effects on the long-term stability of the whole device. Accordingly, tremendous efforts have been devoted to developing alternative HTMs. 6,7 Low-cost carbazolebased molecules, bearing a wide variety of functional groups to fine-tune electronic and optical properties, have attracted much attention due to PSC performances equalising that of spiro-OMeTAD. [8][9][10] Meanwhile, porphyrins, due to their high thermal stability and efficient light-harvesting throughout a large portion of the solar emission spectrum, easily tailored peripheral and non-peripheral substituents, and tuneable optical and electronic properties, are showing their great potential as HTMs since the pioneering work carried out in 2016. 11,12 In this first literature report, Zn(II) porphyrinethynylaniline conjugates as HTM in PSCs generated a PCE of 16.6%. 13 Later, Chen et al. reported the use of Zn(II) and Cu(II) complexes of 5,10,15,20-tetrakis{4-[N,N-di(4methoxyphenyl)amino-phenyl]}porphyrin with outstanding thermal and electrochemical stability. Under the same operating conditions, whilst the Zn(II) complex gave rise to PCE of 17.78%, comparable to the spiro-OMeTAD PCE of 18.59%, the Cu(II) complex contributed only a PCE of 15.36%. 14 In 2018, a PCE of 18.85% resulted from the utilisation of a Zn(II) porphyrin bearing fluorinated triphenylamine substituents as HTM. In the same study, efficiencies of 17.71% and 16.37% were reported for non-fluorinated versions, and all these outcomes are a consequence of incorporating F substituents in the periphery of the porphyrin core that creates an increase in charge injection/transfer of the devices due to better-matched HOMO level and improved hole mobility. 15 Besides, Chiang et al. combined triple cation perovskites with porphyrin dimers incorporating ethynylaniline moieties to achieve a PCE of 19.44%, outperforming the widely used spiro-OMeTAD (18.62%). 16 In a broader context, it has been proven that triarylamine substituents in the porphyrin cores have a deep impact on energy levels to play an important role in enhancing charge transport in the amorphous HTM film and thus in improving the PSC device performance. However, there are no reports on the effects of directly N-linking N-donors to the porphyrin core in the field of HTMs, apart from the work conducted by Cho et al. describing Zn(II)-phthalocyanines substituted with secondary amine donor groups, including diphenylamine and carbazole derivatives, as a mixture of positional isomers reaching a maximum PCE of 11.75% for the diphenylamino-substituted macrocycle vs. a spiro-OMeTADbased device attaining 16.78%. 17 According to previous studies, although the amino groups of meso-(diarylamino)porphyrins raise the porphyrin HOMO level, electronic perturbation is only moderate since the diarylamino groups twist out of the bulky porphyrin plane, minimising conjugation. 18 We thus inferred that the introduction of bulky and completely planar N-electron-donating groups might be more fruitful due to the increased donating ability of the nitrogen lone pair participating in conjugation, together with the increased steric hindrance due to their completely planar configuration. Following this line of reasoning, Cu(II) and Zn(II)based porphyrins doubly functionalised with carbazole were readily obtained and successfully used as HTMs in PSC devices. In the present study, also contributing to the fundamental understanding, the dibrominated precursor structures, as well as the meso-unsubstituted porphyrins, were used for the same purpose.

Experimental details 1. Materials
All chemical reagents and solvents employed in the synthetic procedures were of chemical or analytical grade and used without further purification. Preparative thin-layer chromatography was carried out on 20 x 20 cm glass plates coated with silica gel (0.5 mm thick). Column chromatography was carried out using silica gel (63-200 microns). Analytical TLC was carried out on pre-coated sheets with silica gel (0.2 mm thick). Commercial sources are Sigma-Aldrich, Merck, Acros, Fluorochem, Porphyrin-systems, Macherey-Nagel, and PanReac AppliChem. 5,15-bis(3,5-di-tert-butylphenyl)porphyrin 2 and porphyrins 3a,b and 4a,b were synthesised under modified conditions reported in the literature. [19][20][21] Detailed procedures are provided in the ESI.

Measurement and Characterisation
2.1. Structural Characterisation. 1 H and 13 C solution NMR spectra were recorded on Bruker Avance 300 (300.13 and 75.47 MHz, respectively) and 500 (500.13 and 125.76 MHz, respectively) spectrometers. CDCl 3 and CD 3 OD were used as solvents and tetramethylsilane (TMS) as the internal reference; the chemical shifts are expressed in δ (ppm) and the coupling constants (J) in hertz (Hz). Unequivocal 1 H assignments were made using 2D COSY ( 1 H/ 1 H), while 13 C assignments were made based on 2D HSQC ( 1 H/ 13 C) and HMBC (delay for long-range J C/H couplings were optimised for 7 Hz) experiments. ESI mass spectra were recorded on a Micromass Q-Tof 2 spectrometer (Micromass, Manchester, UK) operating in positive mode. Highresolution mass spectra were recorded on a LTQ Orbitrap XL mass spectrometer (Thermo Fischer Scientific, Bremen, Germany) using CHCl 3 as solvent. The UV-Vis spectra were recorded on a UV-2501PC Shimadzu spectrophotometer using N,N'-dimethylformamide or chlorobenzene as solvent. 2.2. Cyclic voltammetry. Autolab PGSTAT302N potentiostat was used in electrochemical experiments. Voltammograms were recorded using a three-electrode cell arrangement; a polished glassy-carbon (GC) pin (3 mm in diameter) served as a working electrode, a platinum wire as a counter electrode, and the reference electrode was acetonitrile Ag|Ag + . Acetonitrile solution of AgNO 3 (0.01 M) with the addition of 0.1 M of Bu 4 NPF 6 was used in the compartment of the reference electrode. Porphyrin complexes 5a and 5b have extremely low solubility in acetonitrile. For this reason, the solvent mixture chlorobenzene:acetonitrile in a 3:1 volume ratio was used for electrochemical studies. The concentration of 5a, 5b, and conventional spiro-OMeTAD was 1.9 mM; all solutions contained Bu 4 NPF 6 salt as a supporting electrolyte in a concentration of 0.075 M. Prior to measurements, the solutions were deaerated by purging high purity Ar during ca. 5 min. Voltammograms were recorded at a potential scan rate of 100 mV s -1 . During the measurement, the Ar flow was kept above the solution in the cell. Electrode potentials in voltammograms are quoted concerning the equilibrium potential (E 1/2 ) of Fc + /Fc redox couple in the same mixture of solvents with supporting electrolyte 0.075 M Bu 4 NPF 6 .
Steady cyclic voltammograms of GC electrode in Fc solutions, either in acetonitrile or in the mixture of solvents, are presented in ESI in Figure S1. E 1/2 of Fc + /Fc couple for each solvent was defined as the midpoint between potentials corresponding to the oxidation and reduction peak. The HOMO energy levels were estimated from the onset potentials of the oxidation (E on Ox ) using the equation E HOMO (eV) = − (5.1 + *E on Ox + **0.042), where: * E on Ox is given vs. E 1/2 (Fc + /Fc) in a 3:1 volume ratio chlorobenzene:acetonitrile mixture; ** 0.042 is the difference between E 1/2 (Fc + /Fc) in mixed solvents and in acetonitrile. The difference reflects the solvent effect on the redox potential of the internal reference (Fc + /Fc) and must be used as a correction factor. The formal potential of the Fc + /Fc redox couple of 5.1 eV in the Fermi scale was obtained using the redox potential of ferrocene in pure acetonitrile. 22 2.3. Electronic structure calculations. The initial structures were constructed and then optimised in GAMESS-US (2018.R3) 23 using the B3PW91 hybrid functional 24 with the LANL2DZ 25 basis set and Effective Core Potentials (ECPs) combination. The basis set was obtained from Basis Set Exchange. 26 As the systems have over 100 atoms, ECPs were used to replace the core electrons with an effective potential and accelerate the calculations. Eighteen core electrons were removed for zinc, ten core electrons were removed for copper, and twenty-eight core electrons for bromine due to the use of ECPs. B3PW91 functional was used as it was shown to have good results in the description of transition metal chemistry. 27 Default values for the convergence in the geometry optimisation were used in all calculations. The Hessian matrix (second-order partial derivatives of energy concerning the cartesian coordinates of all atoms) was calculated on the final geometries and the absence of negative eigenvalues showed the geometries correspond to minima. The optimised geometries in xyz format are available at https://github.com/peabreu/POR-PSK. To establish the differences between the optimised structures, all geometries that had the same metal atom were superimposed, minimising the root mean square deviation of the coordinates of the central N atoms. This is shown in Figure S2 included in the ESI. There is almost no difference in structures when the hydrogen atom is replaced by bromine (blue and red structures). The two 3,5-di-tert-butylphenyl groups are not perpendicular to the macrocycle and have a similar angle value (ranging between 50 and 53 degrees) in all structures. Structures 3 and 4 have an almost planar macrocycle (with a deviation from planarity of less than 1 degree), but structures 5a and 5b are distorted from planarity by almost 10-20 degrees (measured by calculating the dihedral angle starting in the central metal atom, followed by the nitrogen and the next two bonded carbons in the pyrrole ring). The carbazole substituents in structures 5 are also not perpendicular to the macrocycle ring with angles around 56 and 60 degrees for 5a and 5b, respectively. This rotation can be attributed to the repulsion of the hydrogen atoms in the phenyl rings, as well as electronic correlation. 28

Photoluminescence.
Room temperature steady-state photoluminescence (PL) and PL excitation (PLE) were conducted in a Fluorolog-3 Horiba Scientific set-up with a double additive grating Gemini 180 monochromator (1200 grooves mm -1 and 2×180 mm) in the excitation and a triple grating iHR550 spectrometer in the emission (1200 grooves mm -1 and 550 mm). This system is equipped with a 450 W Xe lamp as the excitation source.
ii) Ullmann coupling: porphyrin 4b (32 mg, 0.036 mmol) was dissolved in dimethylsulfoxide (1.5 mL) and the resulting mixture was purged with nitrogen for 5 min. Then, it was added carbazole (59.4 mg, 0.36 mmol, 10 equiv.), Nphenylbenzohydrazide (3 mg, 0.014 mmol, 0.4 equiv.), CuI (1.4 mg, 0.007 mmol, 0.2 equiv.), and Cs 2 CO 3 (46.3 mg, 0.14 mmol, 4 equiv.). The resulting mixture was stirred at 120 °C under a nitrogen atmosphere for 5 days. After reaching room temperature, the reaction mixture was diluted with CH 2 Cl 2 , washed first with a saturated solution of NaHCO 3, and, finally, with distilled water. Then, the organic phase was separated, and the solvent was removed under reduced pressure. The same conditions described in procedure i) were used to purify the crude mixture and compound 5b was isolated in 31% yield (12.2 mg). iii) Nucleophilic reaction mediated by PhI(OAc) 2 /NaAuCl 4 ·2H 2 O: to a solution of the porphyrin 4b (15.4 mg, 0.017 mmol) in CH 2 Cl 2 (5 mL) was added carbazole (14.2 mg, 0.085 mmol, 5 equiv.), PhI(OAc) 2 (5.5 mg, 0.017 mmol, 1 equiv.), and NaAuCl 4 ·2H 2 O (10.1 mg, 0.025 mmol, 1.5 equiv.). The mixture was stirred at room temperature for 4.5 h until the TLC control showed the full consumption of the starting material. Then, it was added CH 2 Cl 2, and the reaction mixture was washed with water, extracted with CH 2 Cl 2, and the solvent was evaporated under reduced pressure. After purification of the crude mixture as described in procedure i), compound 5b was isolated in 11% yield (2 mg).

Preparation and characterisation of perovskite solar cells
Triple cation Cs + /CH 3 NH 3 + /CH(NH 2 ) 2 + perovskites made by an anti-solvent dropping method were used to prepare complete devices as described elsewhere. 29,30 Fluorine-doped tin oxide (FTO) glass substrates (2 mm thickness, TCO-7, 7 Ω/square, Greatcell Solar) were patterned via VersaLaser (VLS 2.30, Universal Laser Systems) to create a scribing to electrically isolate the photoelectrode from the counter electrode. In the next step, substrates were mechanically cleaned using a 10 % Hellmanex (Hellma GmbH) water solution. After rinsing abundantly with distilled water, substrates were sonicated in a potassium hydroxide ethanolic solution and posteriorly in distilled water, 10 minutes for each step, to be then dried at 55 °C for 30 minutes. Before blocking layer deposition, substrates were submitted to additional cleaning by oxygen plasma treatment (Zepto, Diener) for 10 minutes. A 50-80 nm TiO 2 blocking layer was deposited via spray pyrolysis of a precursor solution containing 0.56 M acetylacetone (Sigma-Aldrich) and 0.18 M titanium diisopropoxide bis(acetylacetonate) (Sigma-Aldrich) in 7 mL of anhydrous isopropanol (Sigma-Aldrich). Substrates were preheated at 450 o C before spray application by an atomiser, using air as carrier gas. Samples were left for additional 45 minutes at the same temperature for the formation of the anatase phase and allowed to cool down to room temperature afterwards.
For the application of mesoporous TiO 2 , commercial 30 NR-D paste (Dyesol) was diluted in pure ethanol (VWR) (1:6 w/w) and spin-coated on the blocking layer at 5000 rpm for 10 seconds with a ramp of 2000 rpm s -1 to achieve a 150-200 nm thick mesoporous layer. Samples were then heated at 100 °C for 10 minutes, for pre-drying, and the film was annealed at 500 °C for 30 minutes. After cooling down to 150 °C, the substrates were immediately transferred to a nitrogen atmosphere glove box (MBraun with pressure varying from 2-3 mbar), to prevent moisture adsorption, for depositing the perovskite films. The perovskite solution precursor was prepared by dissolving 1. The devices were characterised right after their preparation, at room temperature and ambient air. The solar simulator (Newport -Oriel, LSH-7320) was calibrated using a single Si photodiode (Newport -Oriel, 91150V), and an output of 1000 W m -2 (with the reference cell held at 25 °C) with Air Mass 1.5 Global (AM 1.5G) spectral filtering is equivalent to 1 sun. The I-V curves were obtained by applying an external potential load and measuring the generated photocurrent using a Zenium (Zahner Elektrik) workstation controlled by the Thales software package (Thales XT 5.1.4). The scan speed and step potential used were 10 mV s -1 and 10 mV, respectively; the individual cells were characterised in reverse mode (from open-circuit to shortcircuit) with a black mask with an aperture of 0.196 cm 2 .

Synthesis
The synthetic strategy to obtain the new Zn(II) and Cu(II)-based porphyrin complexes doubly functionalised with carbazole 5a and 5b is outlined in Schemes 1 and 2, and it required the previous preparation of starting meso-dibrominated scaffolds 4a and 4b from 5,15-bis(3,5-di-tert-butylphenyl)porphyrin 2. As shown in Scheme 1, porphyrin 2 was obtained in 64% yield by reacting dipyrromethane 1 with 3,5-di-tert-butylbenzaldehyde in the presence of catalytic amounts of trifluoroacetic acid (TFA), followed by oxidation with 2,3-dichloro-5,6-dicyano-1,4benzoquinone (DDQ). 19 Afterwards, metalation of the porphyrin inner core with zinc(II) or copper(II) acetate, 20 followed by bromination of both free meso-positions present in complexes 3a or 3b with N-bromosuccinimide (NBS), 21 afforded the required derivatives 4a and 4b in 82% and 85% yield, respectively. The best protocol to obtain the Zn(II) complex 5a was based on the nucleophilic substitution reaction involving 4a and carbazole mediated by PhI(OAc) 2 /NaAuCl 4 ·2H 2 O 31 (Scheme 2) . This approach was considered after finding that the reaction of Zn(II) complex 4a with carbazole using either an Ullmann protocol 32 or a Buchwald-Hartwig 33 approach led mainly to the exchange of the metal at the inner core by the catalyst metal [Cu(II) or Pd(II)]. To overcome this issue, the coupling of carbazole with the Ni(II) complex of 5,15-dibromo-10,20bis(3,5-di-tert-butylphenyl)porphyrin was also performed under Ullmann conditions to afford the desired Ni(II) derivative bearing two carbazole units in 50% yield. However, all subsequent attempts to remove Ni(II) under different acidic conditions, to then introduce Zn(II) or Cu(II) ion, led to high degradation of the porphyrin and made this approach impractical. On the contrary, when the double nucleophilic substitution of 4a with carbazole (5 equiv.) was carried out in the presence of PhI(OAc) 2 (1 equiv.) and NaAuCl 4 ·2H 2 O (1.5 equiv.) in CH 2 Cl 2 at room temperature, the desired product 5a was isolated in 22% yield accompanied by the corresponding free-base in analogous amounts. Further metalation of the freebase with Zn(AcO) 2 allowed to isolate 5a in a total yield of 59%. To overcome this entire demetallation/metalation process and build up a straight synthetic route, treatment of the dibrominated porphyrin 4a dissolved in CH 2 Cl 2 with triethylamine followed by solvent removal was carried out before the oxidative nucleophilic substitution step. Compound 5a was then isolated in an overall yield of 51%. When the nucleophilic substitution approach was extended to the Cu(II) complex 4b, the desired derivative 5b was just obtained in 11% yield. However, as the exchange of the Cu(II) ion is less probable due to the higher stability of the porphyrinic complex 4b, the aforementioned Ullmann and Buchwald-Hartwig cross-couplings were again considered (see Table S1). Under Ullmann conditions, the coupling of 4b with carbazole (10 equiv.) in the presence of Nphenylbenzohydrazide (0.4 equiv.), CuI (0.2 equiv.), and Cs 2 CO 3 (4 equiv.) was carried out in dimethylsulfoxide (DMSO) at 120 °C for 5 days to afford the desired derivative 5b in 31% yield. Considering Buchwald the conditions [carbazole (20 equiv.), Pd(OAc) 2 (0.4 equiv.), t BuONa (71 equiv.), 18-crown-6 (0.3 equiv.), and rac-BINAP (0.6 equiv.), at 120 °C in a mixture of N,N'-dimethylformamide and toluene (1:2)], the yield of 5b was improved to 39% with the reaction time reduced to 2 days. It is worth referring that the desired product was not obtained when the reaction was carried out in tetrahydrofuran; the most common solvent referred to in couplings involving porphyrins. 33b,34 Scheme 1 Synthesis of scaffolds 2-4.

Structural characterisation
Structural elucidation of the free-base scaffold 2 and all the Zn(II) complexes described above involved the use of 1D ( 1 H and 13 C) and 2D [( 1 H, 1 H) COSY, NOESY, ( 1 H, 13 C) HSQC, and ( 1 H, 13 C) HMBC] NMR techniques (Figures S3, S5, S10, and S14-S18). The 1 H-NMR spectrum of the Zn(II) complex 3a, when compared to compound 2, showed the absence of the signal corresponding to the resonance of the inner core NH protons (δ -3.02 ppm). No noticeable change was observed for the resonances of the remaining protons. In 4a, the absence of the singlet at δ 10.3 ppm corresponding to the resonance of the protons at meso-positions confirmed the success of the dibromination step. When compared to the precursor, the presence of the two bromine atoms induced a deshielding of the doublet generated by the resonances of four β-pyrrolic protons from δ 9.46 to 9.62 ppm, whilst the remaining signals were slightly shifted to high field. After the introduction of the carbazole moieties, compound 5a, the two doublets generated by the resonances of the β-pyrrolic protons were shielded from δ 9.62 and 8.88 ppm to δ 8.88 and 8.58 ppm, respectively. In the aromatic region, four signals appeared ranging from ca. δ 8.5 ppm to ca. δ 6.8 ppm and correspond to the resonance of the carbazole protons.
The singlet due to the resonances of the protons from the 3,5di-tert-butylphenyl substituents remained almost unchanged, exhibiting a chemical shift around δ 1.5 ppm. The structures of all Zn(II) and Cu(II) porphyrin complexes were also supported by mass spectrometry (MS-ESI/HRMS-ESI) and UV-Vis spectroscopy. Apart from compound 4b, which displayed the m/z signal corresponding to the [M+2H] + ion, all the remaining compounds showed the presence of the peak with m/z corresponding to the [M] + or [M+H] + ion (Figures S4,  S6-S9, S11-S13, and S19-S22). As expected, both Zn(II) and Cu(II) porphyrin complexes 3a and 3b exhibited typical absorption features of metallo complexes of arylporphyrins with a Soret band assigned to allowed S 0 → S 2 transitions and two Q bands due to S 0 → S 1 transitions (Figure S23). 35 Bromination of the meso-positions induced a red-shift (ca. 18 nm) of the Soret band (compounds 4a and 4b), while a slight blue-shift (ca. 4 nm) was observed after their replacement by carbazole moieties (compounds 5a and 5b). Also, in the Q bands region, significant changes were observed after the bromination by a red-shift ranging from 20 to 27 nm, and a less noticeable blue-shift after insertion of the carbazole units (3 to 11 nm). Spin-coated films of 5a and 5b on glass substrates exhibited similar absorption spectra to those in solution with a slight blue-shift of the Soret and Q bands (ca. 10 nm) for Zn(II) complex 5a relatively to complex 5b (Figure S24).

Photovoltaic performance
To assess the performance of the synthesised porphyrins as HTMs, n-i-p PSCs were fabricated. A triple cation Cs + /CH 3 NH 3 + /CH(NH 2 ) 2 + perovskite was applied onto a TiO 2 mesoporous layer on top of an electron extraction layer. 29 The porphyrin was then applied as the HTM, over the perovskite layer, and, finally, a thin gold film was deposited to serve as a counter-electrode. For comparison purposes reference cells made of spiro-OMeTAD were used. Lithium bis(trifluoromethylsulphonyl)imide (LiTFSI), tris(2-(1H-pyrazol-1-yl)-4-tert-butylpyridine)cobalt(III) tri[bis(trifluoromethane)sulfonimide] (FK209) dopants and the 4-tert-butylpyridine (TBP) additive were tested to improve the electric conductivity of the HTM layer, either based on the prepared porphyrins or spiro-OMeTAD. The reference cell, loaded with a HTM layer of spiro-OMeTAD doped with composition A -see Table 1, displayed a typical PCE value of 14.74%, with a short-circuit current density (J sc ) of 20.08 mA cm -2 , an open-circuit voltage (V oc ) of 1.00 V, and a fill factor (FF) of 0.72. It is widely accepted that by reducing Co(III) -present in FK209 -to Co(II), spiro-OMeTAD becomes partially oxidised, rendering the HTM layer more electrically conductive. A cell without FK209 (composition B) displayed a FF of 0.58 and a PCE of 6.01%, similar to the performance of an additive-free cell (composition C), but a much lower J sc , only 7.90 mA cm -2 . To further increase the electrical conductivity of the HTM layer, hygroscopic co-dopant LiTFSI can be added. Indeed, in the absence of LiTFSI, the devices showed much lower photovoltaic parameters and a PCE < 4% (composition D). However, due to the hygroscopic nature of this additive, the stability of asprepared cells can be also compromised by water absorption during their preparation and under operation. 6,7   The role of the HTM additives was also assessed for devices loaded with the meso-unsubstituted Cu(II) porphyrin complex 3b ( Table 1). When the 3b HTM was applied without any dopant, the PCE of the corresponding device remained low,2.30%. However, when 0.5 equiv. of LiTFSI with 0.05 equiv. of FK209 and 3.3 equiv. of TBP per mol of HTM was addedoptimised composition, the PCE rose to 5.30%. In contrast, the absence of FK209 led to devices with a PCE of 4.26%. It is worth noting that the 10,20-unsubstituted porphyrinic system was selected to eliminate any effect of the substituents at this stage. Subsequently, under this optimised additive composition, the role of Zn(II) complex 3a and also the corresponding Ni(II) and Pd(II) complexes were evaluated ( Table S2 in ESI). It was found that 3a gave rise to a slightly higher efficiency -PCE = 6.39% -, whilst Ni(II) and Pd(II) complexes resulted in devices with low efficiency -PCE < 3%. This last result can be assigned to a poor alignment of the HTM HOMO level with the valence band of the perovskite layer, compared with Cu(II)-and Zn(II)-based porphyrinic HTMs. 14,36 The study proceeded then to Zn(II) and Cu(II)-based porphyrins 5a and 5b doubly functionalised with carbazole. To the best of our knowledge, no C-N direct bond design in porphyrins has been previously reported as HTMs for PSCs. Only, Pozzi and Nazeeruddin described a Zn(II)-phthalocyanine as a mixture of positional isomers reaching a PCE of 6.65% (vs. a spiro-OMeTAD-based device achieving 16.78%). 17 The feasibility of hole extraction of 5a and 5b was evaluated using cyclic voltammetry (CV) - Figure 1. The HOMO energy levels were estimated from the onset potential of oxidation (E on Ox ), assuming the formal potential of the Fc + /Fc redox couple as 5.1 eV in the Fermi scale. 22 The LUMO was obtained by adding E 0-0 = 1240/ int (determined from the interception of the normalised absorption and emission spectra, see Figure S25). The obtained data are presented Table 2, together with the HOMO level of spiro-OMeTAD also determined experimentally. As shown in Figure 2, the HOMO level of spiro-OMeTAD, 5.11 eV, and Zn(II) complex 5a, 5.54 eV, have favourable energy alignment with the perovskite layer (ranging from 5.40 eV to 5.50 eV) 37 and is expected to lead to higher V oc and J sc . The deeper HOMO level of Cu(II) complex 5b, 5.74 eV, must hinder the transportation of holes, leading to lower efficiencies. At the same time, the LUMO levels of both 5a (3.38 eV) and 5b (3.62 eV) are higher than that of the perovskite (3.8 eV), which must block the electron transport to the Au counter-electrode and hence suppressing the carrier recombination.   The optimised geometry and molecular orbitals of 5a and 5b using B3PW91/LANDL2DZ level of theory are shown in Figure 3, and also indicate the higher potential of the Zn(II) complex as HTM. For this complex the HOMO electron density distribution is mainly localised at the porphyrin core and at the twisted carbazole units, while for the Cu(II) complex 5b is more delocalised over the whole molecule. The performance of porphyrins 5a and 5b as HTM was evaluated under the optimised conditions previously established for the meso-unsubstituted complexes 3. Figure 4 shows the current density-potential difference (J-V) curves of devices with 5a and 5b HTMs; the corresponding photovoltaic characteristics are summarised in Table 3. As expected, this strong π-donor group directly connected to the macrocycle produces drastic changes. An improved FF of 0.73 for 5a and 5b (vs. 0.69 for 3a and 0.62 for 3b) and a J sc of 13.79 and 9.14 mA cm -2 for 5a and 5b, respectively, were recorded. Although significantly lower than those of the spiro-OMeTAD reference cell (ca. 14.7%), these devices displayed a PCE of 9.87% for 5a and 6.37% for 5b. These photovoltaic performances are substantially higher than those achieved with unsubstituted porphyrins 3a (PCE = 6.3%) and 3b (PCE= 5.30%), and higher than Zn(II)-phthalocyanine 17 , pointing out also the critical role of the metallic Zn(II) centre in comparison to Cu(II).
To gain insight into the role of electron-withdrawing substituents in cell performance, dibrominated porphyrins 4a and 4b were also used as the HTM layer. Figure 4 displays the J-V curves of these devices, whilst the respective photovoltaic characteristics are shown in Table 3. A PCE of 4.68% and 3.31% was obtained respectively for devices based on 4a and 4b, which are much lower compared to 5a and 5b. In general, halogens, due to their relatively large size, can induce a  conformational distortion of the porphyrin core when compared to those caused by other substituent groups. 39 However, the superimposed geometries of 3 and 4 indicate almost no difference in the structures when the hydrogen atom is replaced by bromine ( Figure S2). This points out that the inductive electron-withdrawing character of the bromine atoms has an important contribution to the lower values of Voc and Jsc observed for these devices.
To assess the reproducibility of device performance, a batch of 12 individual cells using the Zn(II) porphyrin complexes 3a, 4a, and 5a as HTMs were fabricated (cf. Tables S2-S6 in ESI). Figure  5 shows the statistical data for J sc , V oc , FF, and PCE, respectively, along with the spiro-OMeTAD reference (vide Table S3). The performance of prepared devices follows the sequence 4a < 3a < 5a for all parameters except for FF, where 3a HTM-based devices display the lowest value. This was assigned to the high solubility of perovskite in most solvents, and only a few nonpolar solvents could be used in HTM deposition, such as benzene (highly toxic), chlorobenzene, or toluene. Porphyrin 3a is highly soluble in N,N´-dimethylformamide, but poorly soluble in chlorobenzene or toluene. Therefore, the spin-coating deposition process results in a more irregular distribution of the HTM layer and less intimate contact with the perovskite, which may explain the lower mean FF value and higher standard deviation compared to devices based on 4a and 5a. Generally, the 4a < 3a < 5a sequence is fully in line with the evolution of electron-withdrawing substituents -bromine groups, to those with an electron-donor character -carbazole units.
The PCE histograms are plotted in Figure 6 showing a good reproducibility of the PSCs using porphyrin-based HTMs.  Indeed, more than 60% of 5a-based devices exhibited efficiencies superior to 9.3%, and the best-performing cell reached a PCE of 10.01%, which is 65% of the PCE displayed by the spiro-OMETAD reference. Moreover, 5a-based devices show a remarkably narrower PCE distribution. These results confirm that the use of porphyrin 5a provides better performance and, most relevantly, a substantial enhancement in reproducibility, a very relevant characteristic when aiming at the development of devices with potential commercial interest.

Photoluminescence analysis
To better understand the quality of the perovskite/HTM interface, steady-state photoluminescence (PL) and photoluminescence excitation (PLE) studies were performed. Firstly, the PL emission of the perovskite film on TiO 2 /glass substrates was recorded under the excitation wavelength of 550 nm (incidence on the perovskite face), resulting in an emission band with a maximum at ca. 755 nm, as can be seen in Figure 7. PLE recorded at 755 nm evidenced a tail of absorption whose intensity increases from 650 nm up to 500 nm. In addition, its maximum occurs at the same spectral region where porphyrins typically exhibit the maximum absorption ( Figure  S24). So, Zn(II) and Cu(II) porphyrin complexes 5a and 5b were also studied when deposited as a film on glass substrates. Figure  8 shows the results of the selective excitation of PL and PLE experiments collected from 5a when irradiating at the porphyrin side. The selected excitation wavelengths correspond to a maximum (420 nm) and a minimum (500 nm) of the PLE spectrum. Both PL spectra exhibit two maxima at 621 and 676 nm. Furthermore, the PLE recorded at 676 nm fairly matches the absorption spectrum previously observed in UV-Vis spectra ( Figure S24). In the case of 5b no emission could be detected due to the poor film quality. Hence, for comparison purposes, the powders of both samples were analysed ( Figure S26). Indeed, both emit within the same spectral region of the PL band of the perovskite, particularly 5b. Further PL analysis was performed in perovskite-HTM films based on porphyrin 5a and 5b and spiro-OMeTAD, as depicted in Figure 9, to assess the role of the HTM at the perovskite PL recombination. The photo incidence is now placed on the glass surface as for device characterisation. Moreover, by this way, it is ensured that the interfaces are the same either in the absence or the presence of the HTM, minimising possible phenomena related to different light absorption from the HTM layers studied. It was observed that PL intensity decreased when both porphyrin-based HTMs were used, being the higher effect observed for 5a. However, the strongest decrease was observed for spiro-OMeTAD. A reproducible behaviour regardless the excitation wavelength used Figure 8 Photoluminescence at different excitation wavelengths and photoluminescence excitation of porphyrin 5a deposited as a film on a glass substrate (λem = 676 nm).

Figure 9
Photoluminescence under the excitation wavelength of 562 nm of perovskite-HTM films based on spiro-OMeTAD and porphyrin 5a and 5b, with the photo incidence placed on the glass substrate.
( Figure S27), though with distinct relative PL emission intensities. These results indicate hole extraction and transfer from the perovskite to the HTM layer, which is in line with the estimated HOMO energy alignment (Figure 2). Nonetheless, the non-negligible contribution from the porphyrin PL emission can play a role in PL spectra as their emission occurs in the same spectral regions as the perovskite. Accordingly, the contribution of the porphyrin PL signal to the overall spectrum may account for a lesser PL intensity reduction when compared to spiro-OMeTAD. Therefore, a direct association between the PL intensity and the charge transfer phenomena is not straightforward. As a result of this investigation, it becomes evident how important it is to control the optical response of the HTM layer to infer the charge dynamics of hole extraction and transport.

Conclusions
A series of Zn(II) and Cu(II)-based porphyrins doubly functionalised with carbazole were successfully synthesised and used as HTM in PSC devices. The Cu(II) complex was properly obtained by C-N transition metal-assisted couplings involving the dibrominated porphyrin precursor. Nevertheless, the Zn(II) complex, due to the metal exchangeability and lability behaviour, led to the use of a nucleophilic substitution protocol. Cyclic voltammetry experiments showed the best alignment of the HOMO energy level of Zn(II) complex 5a with the perovskite valence band. The optimised geometry and molecular orbitals of both complexes also corroborate the higher potential of 5a as HTM. Photoluminescence studies showed an emission quenching, compatible with hole transfer phenomena. Photovoltaic devices fabricated with a 5a-based HTM layer displayed a reproducible maximum power conversion efficiency of ca. 10%.
These results are the first example of the use of direct C-N linkage involving porphyrins in the preparation of HTMs, and with significantly higher PCEs compared to previously reported devices based on a phthalocyanine analogue, which displayed a PCE of 6.65%. Furthermore, this work highlights the influence of the metal present in the porphyrin cavity, as well as the nature of the substituents on the PCE displayed by the resulting devices.