Doping Strategies for Tetrasubstituted Paracyclophane Hole Transport Layers in Perovskite Solar Cells

Because of its excellent hole conductivity, p-doped 2,2 ′ 7,7 ′ -tetrakis-( N,N -di-p- methoxyphenyl-amine)-9,9 ′ -spiro-biﬂuorene (spiro-MeOTAD) is commonly deployed for hole transport in organic metal halide perovskite solar cells, but its rather expensive synthesis prompts the research for alternatives. In this work, tetrasubstituted [2.2]paracyclophanes (PCPs) are synthesized and investigated for replacing spiro-MeOTAD. To enhance their conductivity, diﬀerent doping strategies are followed. Best conductivities are achieved by doping PCP thin ﬁlms with tris(2-(1 H -pyrazol-1-yl)-4- tert -butylpyridine) cobalt(III) tris(bis(triﬂuoromethylsulfonyl)imide) (FK209), matching the conductivity of state-of-the-art p-doped spiro-MeOTAD. Best performance in solar cells is leveraged by doping PCPs with the co-dopants lithium bis(triﬂuoromethanesulfonyl)imide (LiTFSI) and 4-tert -butylpyridine (tBP) which are also used to p-dope spiro-MeOTAD thin ﬁlms in solar cells. Yet, the thermal device stability is maximized upon doping PCPs with FK209 and 2,3,5,6-tetraﬂuoro-7,7,8,8-tetracyanoquinodimethane (F 4 TCNQ).


Introduction
In perovskite solar cells (PSCs) with regular device architecture, hole collection at the electrodes is most often facilitated by hole transport layers of spiro-MeOTAD. [1]Spiro-MeOTAD can be readily processed into homogeneous thin films, it exhibits good DOI: 10.1002/adfm.202402110hole conductivity when electrically doped, and has been instrumental in the evolution of PSCs. [2][5][6] Substitutions at donor groups in these molecules offer a wide range of adaptability in terms of solubility and ionization potential (IP). [6]n order to achieve appropriate conductivity, spiro-MeOTAD is commonly doped with LiTFSI:tBP.Upon exposure to oxygen, spiro-MeOTAD forms weakly bound donor-acceptor complexes, i.e., spiro-MeOTAD + O 2 − , due to the low IP of spiro-MeOTAD and the high electron affinity (EA) of molecular oxygen.The O 2 -anion is then exchanged with the TFSI − anion from Li + TFSI − to form spiro-MeOTAD + TFSI − and lithium oxide. [7]It was reported that water can also be involved in the oxidation process. [8,9]The positive charge on spiro-MeOTAD + is only weakly bound to the delocalized negative charge on the TFSI -anion, hence constituting a mobile hole. [7,10]BP is not part of the redox reaction, but it increases the solubility of LiTFSI in chlorobenzene (CB) and significantly influences the layer morphology of LiTFSI-doped spiro-MeOTAD thin films. [11]Furthermore, tBP is known to decrease recombination at surface defects, resulting in enhanced open-circuit voltages of solar cells. [12][19] Furthermore, it evaporates over time, leaving behind pinholes and aggregated LiTFSI.The hygroscopicity of LiTFSI promotes the degradation of the perovskite in case of water ingress. [11]Although innovative approaches have been reported to mitigate these effects, such as binding tBP in a complex by halogen bonding or replacing tBP with pyridine moieties in the hole transport material's (HTM) structure, [19,20] avoiding the use of LiTFSI is desirable.
Over the last few years, FK209 has become increasingly popular as a third doping component in addition to LiTFSI:tBP. [13,14,16]et, FK209 can also oxidize spiro-MeOTAD on its own and without any exposure to oxygen.In this process, Co(III) is reduced to Co(II) and one of the TFSI ions becomes the counterion to spiro-MeOTAD + . [21]Another common dopant for organic semiconductors, which has been used for spiro-MeOTAD, is F 4 TCNQ. [22,23] 4 TCNQ forms a ground state charge transfer complex with spiro-MeOTAD.Both FK209 and F 4 TCNQ are commercially available at moderate cost.5] We synthesize novel tetrasubstituted PCPs and systematically investigate their optoelectronic properties, particularly their charge carrier dynamics and p-doping.Quantum mechanical calculations demonstrate the electronic structure of these molecules and identify the optical transitions of oxidized tetrasubstituted PCPs (i.e., their cations), which aids the interpretation of UV-Vis-NIR spectra.We employ metal-insulatorsemiconductor charge extraction by linearly increasing voltage (MIS-CELIV) in order to investigate the charge carrier dynamics of neat and doped PCPs.Finally, we implement pdoped PCP layers into solar cells and test the performance and thermal stability of devices in dependence of the HTL's doping.

Synthesis of Thiophene-Bridged Tetrasubstituted PCPs
PCP was previously substituted with triphenylamine (TPA) donor groups either directly or with ethene -bridges in between.Recently, we demonstrated cost-efficient disubstituted PCPs with thiophene -bridges.Substitution of TPA in para-position with methoxy or tert-butyl groups allowed for tuning of the IP via mesomeric and inductive donor effects, respectively. [6]Both groups increased the solubility of molecules by preventing their aggregation and promoting interactions with organic solvents.
Here, we synthesized three different tetrasubstituted PCPs with TPA donor groups attached via thiophene -bridges and additional substitution of TPA in para-position.In the first target compound (TPCP-1), unsubstituted TPA was used, while the second (TPCP-2) and third (TPCP-3) target compounds feature methoxy-and tert-butyl-substitutions in the para-positions of the TPA units.
The tetrasubstituted PCPs were synthesized as depicted in Figure 1 using consecutive cross-couplings.In a first step, the desired donor moiety was attached to thiophene via Suzuki coupling.[26] These were coupled using a common Suzuki cross-coupling setup with Pd(dppf)Cl 2 and K 2 CO 3 in dimethyl sulfoxide (DMSO) at 90 °C, resulting in yields of 100%, 76%, and 84% for 5, 6, and 7, respectively.
The subsequent quadruple coupling of 5, 6, and 7 to the PCP core could not be achieved in sufficient yields via Suzuki coupling or by CH activation, which worked well for disubstituted PCPs. [6]Therefore, we followed the strategy published by Kobayakawa et al., [27] who used Negishi cross-coupling to connect two thiophene moieties with a pseudo-ortho iodinated [2.2]paracyclophane.To obtain the required Negishi reactant, Kobayakawa et al. lithiated thiophene with n-butyllithium and then added a zinc chloride solution.With the Negishi reactant ready, the [2.2]paracyclophane was introduced to the reaction.Eventually, they isolated the desired disubstituted product with a yield of 85%.We slightly modified this protocol regarding temperature and catalyst, i.e., PEPPSI-IPr was used to optimize the yield for the fourfold coupling of units 5, 6, and 7 with tetrabromo[2.2]paracyclophane(8) as depicted in Figure 2.This optimized step yielded 53%, 75% and 64% of TPCP-1, TPCP-2, and TPCP-3, respectively.Hence, the methoxy and tert-butyl substituents raised the yield of the final product by 11 or 22% absolute, respectively.
All three final products exhibited solubilities of more than 30 g L −1 in CB, which was beneficial for device processing.Yet, the unsubstituted TPCP-1 showed aggregation during storage, while the tert-butyl-and methoxy-substituted TPCP-2 and TPCP-3 remained fully dissolved in CB over several months.All three TPCPs exhibited good thermal stability with glass transition temperatures above 140 °C (Figure S1, Supporting Information).
In order to assess the economic feasibility of TPCPs as an alternative to spiro-MeOTAD, we carried out a cost estimate of TPCP-3.This estimate shows that even without further optimization of the synthesis, TPCP-3 can be produced at 25% of the cost of spiro-MeOTAD (Table S1, Supporting Information).

Optoelectronic Properties
An ideal HTL both facilitates hole transport without ohmic losses and blocks electrons from diffusing into the anode in order to prevent electrode recombination losses.Hence, HTLs are generally designed for a high hole conductivity, an IP that matches the perovskite's IP and a wider energy gap than the band gap of the perovskite.A wide optical gap (E g opt ) of the HTL also prevents parasitic light absorption.
We determined E g opt of TPCP solutions in CB by analyzing the Tauc plots of UV-vis absorbance spectra, depicted in Figure 3a.By applying linear fits around the inflection points, we obtained E g opt between 2.43 and 2.50 eV for all three TPCPs.This was ≈400 to 500 meV narrower than E g opt of the reference spiro-MeOTAD, i.e., 2.93 eV.We confirmed E g opt by studying the transitions from the ground state of the molecule, S 0 , and the lowest singlet excited state S 1 by time-dependent density functional theory (TD-DFT) calculations.Table 1 summarizes the results which also agree with the experimental findings within negligible deviations of up to 0.06 eV with a slight dependence on the size of the localized basis set.Visualizations of the transition orbitals involved in the excitations and the orbital energies are provided in Figures S2-S4 and Tables S2-S7 (Supporting Information).
Thus, TPCPs exhibit some parasitic photon absorption in the blue spectral regime.However, the strong absorption of the perovskite in the blue diminishes any parasitic absorption of blue photons in the HTL on the back side of the device. [28][31] Figure 3b depicts representative PESA measurements on TPCP-1, TPCP-2, and TPCP-3 thin films.Fitting the photoelectron yield versus the photon energy for three samples each, we found IPs of 5.42, 5.34, and 5.27 eV, respectively, which were larger than the IP of spiro-MeOTAD at 5.12 eV.The mesomeric donor effect of the methoxy group in TPCP-3 was noticeable with a lowering of the IP by 150 meV against the unsubstituted TPCP-1.The inductive donor effect of the tert-butyl group in TPCP-2 also reduced the IP, but the impact was smaller.
We also measured IPs of the TPCPs in solution with cyclic voltammetry (CV) (Table 1; and Figure S5 and Table S8, Supporting Information).We observed the same IP trend as in PESA measurements and DFT calculations, but owing to the polar solvent environment and different states of aggregation compared to a solid thin film, [32,33] the IPs obtained from CV were smaller by ≈200 meV.For comparison, the IP of the perovskite methylammonium lead iodide (MAPbI 3 ) is reported at ≈5.6 eV. [34,35]We also calculated the IPs of all three TPCPs via density functional theory (DFT) in vacuo (for the lowest energy conformer), resulting in 5.58, 5.43, or 5.26 eV, respectively.The slight differences in the calculated and the experimentally measured IPs likely stem from the treatment as an isolated molecule in the simulation while PESA is measured on solid-state thin films.However, DFT calculations and PESA measurements show the same trend of IPs between all three TPCPs as shown in Table 1.
Hence, the IPs of all investigated TPCP compounds are suitable to be used as HTLs in MAPbI 3 solar cells.Due to their slightly larger IPs compared to spiro-MeOTAD, they may also be applied to wide-bandgap perovskites that typically have larger IPs than MAPbI 3 . [36,37]

Hole Mobilities in TPCP Thin Films
HTLs must have a certain minimal thickness to ensure complete coverage of the underlying layer in a thin-film device in order to mitigate imperfections from the fabrication process.For the purpose of minimizing ohmic losses and avoiding space-charge limitation of the hole current, the HTL should possess a high hole conductivity  = q•p•μ h , which is proportional to the hole density p, the hole mobility μ h , and the elementary charge q.In order to determine the hole mobilities of the TPCPs, we employed MIS-CELIV. [40,41]Holes were injected through a molybdenum oxide (MoO x )/silver electrode into the TPCP thin films, where they accumulated at the opposite interface to an insulating lithium fluoride layer (LiF, 20 nm).Then, a linear voltage ramp was applied to extract the previously injected holes from the TPCP thin film.The charge carrier reservoir mimics an ohmic contact, creating a space-charge limited transient current, which is overlaid with the displacement current J 0 of the geometric capacitance.The time at which twice the displacement current is reached, i.e., t 2J0 , is related to μ as reported by Sandberg et al. [41] Figure 4 shows the transient curves obtained for TPCP-1, TPCP-2, TPCP-3, and spiro-MeOTAD thin films.J 0 was determined using an injection voltage of −3 V. We found that this negative injection voltage was necessary to deplete spiro-MeOTAD of holes from unintentional doping and to obtain an accurate J 0 .No unintentional doping was found in the TPCPs.A comparison of hole extraction at injection voltages of 0 and −3 V and further discussion of unintentional doping can be found in the supporting information (Figure S6, Supporting Information). [8,42]Notably, J 0 contains information about the quasi-static relative permittivity ( s ) of the semiconductor layer which influences the electric field distribution in the solar cell.By treating the insulator and the semiconductor layers as two parallel plate capacitors connected in series, we obtained  s = 3.2 ± 0.1 for the unsubstituted TPCP-1.The tert-butyl-substituted TPCP-2 exhibited a lower permittivity of  s = 2.7 ± 0.2, while the methoxysubstituted TPCP-3 had a higher permittivity of  s = 4.1 ± 0.2, Table 1.Summary of the optical and electronic properties of the TPCP compounds obtained from UV-vis spectroscopy, PESA, CV, and DFT calculations.Data of spiro-MeOTAD are provided for reference.Spiro-MeOTAD 2.93 3.11 [ 38] 5.12 ± 0.03 --a) Potential ferrocene (Fc) (E°(Fc + /Fc) = 0.400 V vs NHE). [ 39]b) Calculated with the def2-SVP basis set.c) Range for different conformers (see Supporting Information).which was very similar to the permittivity of spiro-MeOTAD ( s = 4.0 ± 0.2).At an injection voltage of 1 V, the slope of the transient current saturated in all compounds (Figure S7, Supporting Information), allowing for consistent measurements of t 2J0 and, hence, the calculation of μ h . [41]TPCP-1 achieved an excellent μ h = (3.7 ± 0.3) • 10 −5 cm 2 ⋅V −1 s −1 , outperforming both the spiro-MeOTAD reference and TPCP-3 (both which may be explained by its bulky tert-butyl units that are not part of the molecule's -electron system and hence do not contribute to charge carrier delocalization (see visualization of the HOMO in Figure S3, Supporting Information).Overall, the hole mobilities are promising but still insufficient for use as neat HTLs in high-performance PSCs.To achieve good hole conductivity, the hole mobility or the hole density have to be increased.

Electrical Doping of TPCP
Charge carrier densities and thus conductivities in semiconductors are commonly increased by electrical doping.Organic semiconductors can be p-doped by introducing strong electron acceptors, which oxidize the organic semiconductor, creating radical cations.

Energetics of Doping
To assess p-doping of TPCPs, we examined the acceptor strength of suitable dopants (i.e., the EA) and the IPs of the HTLs.The EA of F 4 TCNQ was reported between 5.08 and 5.24 eV, measured by inverse photoelectron spectroscopy, [43,44] and 5.23 to 5.33 eV according to CV measurements. [45,46]However, its actual EA depends strongly on the host material. [47]FK209 was reported to have an EA of 5.12 eV, determined by differential pulse voltammetry. [21]In relation to the IPs of the TPCPs (i.e., between 5.27 and 5.42 eV), the EAs of F 4 TCNQ and FK209 initially appear insufficient to foster charge transfer between dopants and TPCPs.However, the Coulomb binding energy of the ground state charge transfer complex and energetic disorder in organic semiconductors make the charge transfer energetically more favorable by several hundred meV. [48]Since the energy differences between the IPs of TPCPs and the EAs of F 4 TCNQ and FK209 deviate only slightly, both dopants can still be efficient.
Upon doping of the HTMs and the formation of radical cations, a change in absorption is expected due to newly available electronic transitions from and into the singly occupied molecular orbital (SOMO) of the cation.These new transitions, which are characteristic of cationic species, appear red-shifted against the original features of the absorption spectrum.Therefore, a molecule that has an E g opt in the UV spectral regime in its undoped state may absorb visible or near-infrared light after oxidation.Such observations were reported for spiro-MeOTAD cations by Fantacci et al. and Cappel et al. [38,49] Our quantum mechanical calculations identify the transition orbitals involved in the excitation of TPCP cations (Tables S5-S7, Supporting Information).The respective calculated absorption spectra are illustrated in Figure 5.The emergence of new absorption bands is clearly visible from the comparison of spectra of neutral and cationic forms of TPCPs.The electron density difference for some characteristic transitions is depicted in the insets of Figure 5.

Doping in Solution
We conducted experimental doping of the TPCPs in CB by adding F 4 TCNQ at a concentration of 10 wt.% or FK209 at a concentration of 50 wt.%(with respect to the mass of the HTM), which yielded comparable molar ratios (Table S9, Supporting Information).In order to produce strong doping signatures in optical measurements, we deliberately chose high dopant concentrations.Immediately after addition of the dopant, the color of all HTM solutions changed from completely transparent or light-yellow to black (Figure S8, Supporting Information).UV-Vis-NIR absorption measurements of diluted HTM solutions in CB demonstrate that this color change indeed originated from integer charge transfer (Figure 6; Figure S9, Supporting Information).Neutral F 4 TCNQ and the HTMs absorb in the same spectral regime (close to 3 eV) which makes it difficult to differentiate between the two. [50]We observed increased absorbance in this regime for all HTM solutions containing F 4 TCNQ, indicating the presence of F 4 TCNQ in its neutral form.However, all of these solutions also exhibited characteristic absorption features of the F 4 TCNQ − anion in the visible (3.2 eV) and near-infrared (1.5 eV) spectral regime, indicating the formation of ground state charge-transfer complexes.These absorption features were much less pronounced in the doped TPCP solutions (Figure 6a-c) than in the doped spiro-MeOTAD reference (Figure 6d).This indicated that doping of spiro-MeOTAD with F 4 TCNQ was much more efficient than doping of the TPCPs with F 4 TCNQ.Accordingly, spiro-MeOTAD doped with F 4 TCNQ also exhibited noticeable absorption features of the spiro-MeOTAD + cation in the visible and the near-infrared. [38,49]pon doping with FK209, distinct absorption bands of HTM cations appeared not only in spiro-MeOTAD, but also in the TPCPs.Specifically, new characteristic bands emerged between 2.1 and 1.5 eV as well as below 1.3 eV.Yet, in comparison to the calculated spectra (Figure 5), the absorption bands in the measured spectra (Figure 6) are red-shifted and show more broadening.In the experimental spectra of FK209-doped solutions, which contain both neutral molecules and their cations, the absorption features at ≈3 eV were less pronounced due to oxidation of the HTM molecules (Figure S9, Supporting Information).The strongest change in absorption upon addition of FK209 occurred in TPCP-3 and spiro-MeOTAD, indicating more efficient doping compared to TPCP-1 and TPCP-2.Remarkably, the main absorbance peak of TPCP-3 (at 2.98 eV) was reduced by 16%, which was similar to the reduction in absorbance found in spiro-MeOTAD (at 3.18 eV), even though the IP of TPCP-3 is 150 meV larger.This shows that FK209 is a significantly stronger dopant for TPCP than F 4 TCNQ.

Enhancing the Conductivity of TPCP Thin Films via p-Doping
In order to quantify the doping efficiency as well as the impact of electrical doping on the conductivity in solid-state samples, we conducted measurements using the same MIS-CELIV device architecture as shown in Figure 4. MIS-CELIV theory assumes no background charge carriers to be present or only charge carriers that diffuse from the electrodes into the device. [40,41]Therefore, it cannot be readily applied to electrically doped samples.[53] We observed that the conductivity of TPCP-3 and spiro-MeOTAD was improved by more than one order of magnitude upon doping, while the conductivity of TPCP-1 and TPCP-2 was improved only marginally.Furthermore, FK209doped TPCP-1 and TPCP-2 solutions in chlorobenzene showed precipitation of material (Figure S10, Supporting Information), which we attribute to poor solubility of FK209 Co(III) TFSI in comparison to FK209 Co(II) TFSI.Therefore, in the following, we focus on the investigation of TPCP-3 and benchmark its properties against spiro-MeOTAD.
F 4 TCNQ-doped samples were too leaky for the determination of doping concentrations from capacitive measurements, but the conductivity could be determined from the ohmic regime.In contrast, on FK209-doped samples, both capacitive and ohmic measurements yielded useful insights.We excluded CELIV measurements on TPCP-3 or spiro-MeOTAD samples which were doped with LiTFSI:tBP since the slope of the transient current did not saturate with increasing injection voltages.
Figure 7 shows the conductivity of TPCP-3 and spiro-MeOTAD thin films versus dopant concentration (FK209 or F 4 TCNQ), derived from the ohmic regime MIS-CELIV measurements (Figures S11 and S12, Supporting Information).In neat films, the conductivity of spiro-MeOTAD was higher than the conductivity of TPCP-3, which we attribute to the already known unintentional doping of spiro-MeOTAD (Figure S6, Supporting  Information). [8,42]Taking into account the hole mobility and the conductivity from MIS-CELIV measurements in the standard and ohmic regimes, we calculated a hole density of 2.5 • 10 16 cm −3 from unintentional doping.Upon electrical doping with F 4 TCNQ or FK209, the conductivity of spiro-MeOTAD increased with higher doping concentrations.F4TCNQ doping led to a greater increase in conductivity at the same concentration by weight.Notably, the molecular weight of FK209 is 5.4 times higher than the molecular weight of F 4 TCNQ and hence the molar concentration of FK209 is only 18% of the molar concentration of F 4 TCNQ at the same content by weight.Yet, FK209-doping increased the conductivity of TPCP-3 more than F 4 TCNQ-doping even at the same weight percentage (lower molar percentage).This suggests that the poor F 4 TCNQ-doping efficiency in solution also translated to the thin film.Remarkably, TPCP-3 doped with FK209 at a concentration of 3 wt.%achieved comparable conductivity to spiro-MeOTAD doped at the same concentration.
The primary goal of doping is the enhancement of the conductivity of the semiconductor by increasing the free charge carrier density.Yet, doping can also affect the conductivity by changing the charge carrier mobility.[56] The former leads to increased mobility at high doping concentrations, while the latter leads to decreased mobility at low doping concentrations. [57]he charge carrier concentrations from capacitive regime measurements and the conductivities from ohmic regime measurements allowed for the calculation of the charge carrier mobility.We further determined the doping efficiency from the charge carrier concentration and the number of dopant molecules per volume.In order to calculate the number of dopant molecules per volume, we first determined the mass densities of each HTL.To this end, we spin coated thin films on 25 mm × 25 mm substrates and subsequently removed the outer 5 mm edge that often contains drying artifacts, obtaining homogeneously coated 15 mm × 15 mm samples.Then, we dissolved the thin film of each sample in 1.8 mL of CB and compared the resulting solutions to reference samples with known concentrations in UV-vis measurements (Figures S13 and S14, Supporting Information).Determination of the layer thickness of equally processed samples by profilometry introduced the largest uncertainty in this procedure.Taking into account the standard deviation of the layer thickness measurements, we obtained densities of 1.29 g cm −3 ± 0.03 g cm −3 and 1.33 g cm −3 ± 0.05 g cm −3 for spiro-MeOTAD and TPCP-3, respectively.Notably, this is a significant deviation from the density of spiro-MeOTAD that was reported earlier in the literature, i.e., 1.02 g⋅cm -3 ± 0.03 g⋅cm -3 (same method) and 1.82 g cm -3 (wafer weight change). [58,59]etails on the measurement precision of our experiment are reported in Figures S13 and S14 (Supporting Information).
Based on the charge carrier densities obtained from MIS-CELIV measurements in the capacitive regime (Figure S15, Supporting Information), we calculated the doping efficiency.However, at an FK209-doping concentration of 3 wt.%, it was impossible to fit the capacitive measurements reliably because the HTLs showed metal-like behavior, approaching the saturation current of the MIS structure.Hence, we could not conclude on charge carrier concentration and mobility in those samples and focused on data from samples with lower FK209-doping concentrations of 0.5 or 1 wt.% in the following.Table 2 summarizes the electrical properties of FK209-doped thin films of TPCP-3 and spiro-MeOTAD.The corresponding measurement data and calculations are shown in Figures S11, S12, and S15 (Supporting Information).
The hole density generated in TPCP-3 was larger than in spiro-MeOTAD at the same doping concentration.This may be a result of higher mass density and doping efficiency.Yet, the standard deviation of the doping efficiency was relatively high because it combined the deviations of both the mass density and the charge carrier concentration.The hole mobility of the FK209-doped samples, which was derived from the hole concentration and the hole conductivity, was about one order of magnitude lower than in neat TPCP-3 and spiro-MeOTAD, which we attribute to Coulomb trapping of charge carriers.

Deployment of TPCP-3 in Perovskite Solar Cells
The conductivity of TPCP-3 thin films that was improved by more than one order of magnitude via doping with FK209 and a favorable IP fulfill the requirements for deployment in solar cells.Hence we incorporated TPCP-3 as HTL in PSCs utilizing the device architecture indium tin oxide (ITO)/tin oxide (SnO 2 )/4-(1′,5′-dihydro-1′-methyl-2′H- [5,6]fullereno-C 60 -I h -[1,9-c]-pyrrol-2′-yl)benzoic acid (C60-SAM)/MAPbI 3 /TPCP-3/MoO x /Ag.For reference, we replaced TPCP-3 with spiro-MeOTAD and investigated the common doping systems LiTFSI:tBP and LiTFSI:tBP:FK209.Figure 8 shows the J-V curves of solar cells in the dark and under illumination with both undoped and doped HTLs.In both cases, the undoped references (black curves) exhibit a significant s-shapes, indicating a charge carrier extraction barrier.Since we know from MIS-CELIV measurements that MoO x forms an ohmic contact to both the silver electrode and the undoped TPCP-3 or spiro-MeOTAD, we attribute this s-shape to the buildup of space charges due to the low conductivities of the undoped HTLs.These space charges partially screen the electric field across the perovskite and reduce the free energy gradient. [60]As a consequence, part of the electric field drops off across the HTL, effectively shifting the operating point of the perovskite layer toward lower voltages. [61]o cure the losses in solar cells, enhancing the conductivity of the HTLs by electrical doping is the method of choice.Following the previously discussed strategies, doping of TPCP-3 with either F 4 TCNQ or FK209 produced solar cells without s-shapes (Figure 8a).However, they exhibited roughly 100 mV lower opencircuit voltages (V OC ) than solar cells with LiTFSI:tBP-doping or LiTFSI:tBP:FK209-doping.As shown in Figure 8a, doping of TPCP-3 with the common doping system LiTFSI:tBP did not mitigate the s-shape, even if FK209 was added as a third component.This indicated inefficient doping of the thin film even though FK209 itself can oxidize TPCP-3 as previously shown in UV-Vis-NIR (Figure 6) and MIS-CELIV measurements (Table 2).The prevailing s-shape in J-V curves of devices with LiTFSI:tBP:FK209 may have been caused by dedoping, which we also observed for TPCP-3 doped via FK209 in solution over the course of a few hours when LiTFSI:tBP was present (Figure S16, Supporting Information).Lamberti et al. found dedoping of spiro-MeOTAD by tBP both in the solid state and in solution. [62]They argue that this is related to charge transfer from spiro-MeOTAD + species to tBP 0 and subsequent formation of an adduct between the charged tBP + with another spiro-MeOTAD + cation (TFSI − acts as a counter ion).We infer that a similar interaction takes place between TPCP-3 + and tBP 0 , although at a much higher reaction rate than in spiro-MeOTAD, since we did not observe discoloring of doped spiro-MeOTAD reference solutions even after 21 h (Figure S16, Supporting Information).
The s-shape of J-V curves of solar cells comprising spiro-MeOTAD was diminished by all four doping concepts, but again resulted in lower V OC s (Figure 8b).Upon doping of spiro-MeOTAD with F 4 TCNQ, the solar cells exhibited V OC = 0.90 V ± 0.02 V, while no doping or FK209-doping resulted in V OC s of 1.01 V ± 0.01 V or 1.03 V ± 0.01 V, respectively.Solar cells comprising spiro-MeOTAD doped with LiTFSI:tBP or LiTFSI:tBP:FK209 exhibited larger V OC s of 1.12 V ± 0.01 V or 1.08 V ± 0.02 V. Summaries of the key parameter statistics are shown in Figures S17-S19 (Supporting Information).
The reduced V OC s may stem from different surface recombination rates at the perovskite/HTL interface. [37,63]Zhang et al. previ-ously found that the concentrations of both the spiro-MeOTAD + cation and the dopant affect the recombination rate. [64]Furthermore, tBP is known to reduce surface recombination at interfaces. [12]Indeed, in photoluminescence (PL) experiments on spiro-MeOTAD-coated MAPbI 3 layers, we observed that neat spiro-MeOTAD quenched the PL of MAPbI 3 , and only detector noise was visible.In comparison, MAPbI 3 coated with spiro-MeOTAD, which included tBP, exhibited a PL intensity enhancement by more than one order of magnitude (Figure S20, Supporting Information).Besides surface passivation, tBP hindering hole extraction may be another origin of the increased PL.In MAPbI 3 solar cells, we quantified the impact of tBP on the opencircuit voltage by direct comparison of solar cells with pristine spiro-MeOTAD and solar cells with tBP added to spiro-MeOTAD (Figure S21, Supporting Information).Addition of tBP without any dopants led to an improvement in V OC from 1.01 V ± 0.01 V to 1.11 V ± 0.01 V, i.e., an enhancement of 100 mV.Hence, both doping-induced surface recombination and reduction of surface recombination by tBP are likely causes for the overall loss in V OC .

Thermal Stability of Solar Cells with Doped HTLs
Dopant diffusion can compromise the thermal stability of HTLs.To test for stability, we annealed the devices at temperatures that are well within the expected operation range of solar cells (70 °C, 1 h).Surprisingly, the charge carrier extraction from devices comprising TPCP-3 doped with LiTFSI:tBP vastly improved through thermal treatment (Figures 8a vs 9a; Figure S17, Supporting Information), enhancing the power conversion efficiency (PCE) from only 3.6% ± 1.1% to 16.7% ± 0.7%.Yet, we observed many electric breakdowns in the dark curves as exemplified in Figure 9a, which we attribute to pinholes in the layer stack.GI-WAXS measurements reveal that TPCP-3 forms an amorphous layer (Figure S22, Supporting Information) and its glass transition temperature of 141 °C (Figure S1, Supporting Information) is far above the annealing temperature of 70 °C.Yet, the dopants may aggregate or lead to crystallization of the entire layer by lowering the glass transition temperature. [65]Notably, the glass transition temperature of neat TPCP-3 is 15 °C higher than the glass transition temperature of spiro-MeOTAD, suggesting a better thermal stability. [65]n contrast, none of the reference devices comprising HTLs doped with LiTFSI:tBP:FK209 showed such an electric breakdown.However, the PCE of these solar cells did only improve from 3.7% ± 0.6% to 11.9% ± 1.4% after annealing whereas LiTFSI:tBP-doped devices reached 16.7% ± 0.7% after annealing.Because of the previously discussed dedoping of tBP, before thermal annealing, TPCP-3 devices doped with only FK209 (PCE = 13% ± 1.8%) or only F 4 TCNQ (PCE = 13.3% ± 1.7%) performed much better than the LiTFSI:tBP:FK209-doped devices.Moreover, they maintained the same performance before and after thermal annealing, demonstrating better thermal stability than devices with any mixture comprising LiTFSI and tBP.

Conclusion
We synthesized three novel HTMs based on TPA units linked to a tetrasubstituted PCP core via thiophene -bridges.Out of these HTLs, TPCP-1 with unsubstituted TPA showed a larger hole mobility ((3.7 ± 0.3) • 10 −5 cm 2 V −1 s −1 ) than the reference spiro-MeOTAD ((2.7 ± 0.3) • 10 −5 cm 2 V −1 s −1 ), but the conductivity of TPCP-1 was only marginally improved by doping.Yet, the methoxy-substituted derivative TPCP-3 (IP = 5.27 eV) could be readily doped by FK209 at a doping efficiency similar to spiro-MeOTAD (IP = 5.12 eV).Therefore, TPCP-3 is a very promising alternative to spiro-MeOTAD, especially in architectures where a larger IP of the HTL is required, e.g. in high-bandgap perovskite solar cells.Moreover, TPCP-3 is inexpensive compared to spiro-MeOTAD and has a higher glass transition temperature, leveraging better thermal stability.The frequently used dopant system LiTFSI:tBP is problematic for stability and hence, alternative dop-ing strategies should be considered both for spiro-MeOTAD and novel HTLs.FK209 doping increased the conductivity of TPCP-3 by more than one order of magnitude while showing promising initial device stability.The limited V OC calls for passivation strategies such as functional group engineering of the HTM itself, that render the addition of tBP, which dedopes the HTL, obsolete.

Experimental Section
Solar Cell Manufacturing-Substrate Cleaning: Glass substrates (16 × 16 mm 2 ) with structured indium tin oxide (ITO) electrodes were cleaned in an acetone ultrasonic bath, followed by rubbing with a plastic swab soaked in glass cleaner.Afterwards, the samples were rinsed with isopropanol and dried with a nitrogen gun.Before SnO 2 deposition, the samples were treated in an oxygen plasma for 2 min in order to improve surface wetting.
Perovskite: The perovskite precursor solution was prepared by dissolving 1.17 mmol lead iodide (Alfa Aesar ultra dry) and 1.17 mmol methylammonium iodide per mL of 1:1 N,N-dimethylformamide (DMF, 99.8%, anhydrous, Sigma-Aldrich): N-methylpyrrolidone (NMP, 99.5%, extra dry, Acros Organics).After stirring at room temperature for several hours, the solution was filtered with a polytetrafluoroethylene (PTFE) filter (pore size: 0.2 μm).The precursor solution was then spin coated (3000 rpm, 300 rpm s −1 , 30 s) inside a nitrogen filled glovebox and immediately transferred to a heated vacuum chamber (volume: 5 L, temperature: 40 °C), which was evacuated for 60 s using a scroll vacuum pump (12.7 m 3 h −1 , nominal final pressure: 0.007 mbar) and then refilled with nitrogen.Furthermore, samples were dried with a nitrogen gun for 60 s, followed by thermal annealing (100 °C, 30 min) on a hotplate inside the glovebox.Subsequently, all sources of solvent vapor were removed from the glovebox.
MoO x /Ag: The top electrode was processed by thermal evaporation in high vacuum through a shadow mask, forming cells with an active area of 10.5 mm 2 each through overlap with the structured ITO bottom electrode.First, molybdenum oxide was evaporated at a rate of ≈1 Å s −1 up to a thickness of 10 nm.Next, without breaking vacuum, silver was evaporated on top at a rate of 1 Å s −1 up to 10 nm and 2 Å s −1 up to the final thickness of 100 nm.
MIS-CELIV Device Manufacturing: The ITO substrates were treated with the same cleaning procedure as for solar cell fabrication.Thereafter, 20 nm of lithium fluoride were thermally evaporated at 1 Å s −1 as the insulator.The materials under investigation were dissolved in CB at 30 g L −1 .The materials were spin-coated (50 μL, 1000 rpm, 1000 rpm s −1 , 30 s), followed by a drying step (4000 rpm, 1000 rpm s −1 , 10 s).The process for the MoO x /Ag top electrode was the same as for the solar cells, except for a smaller active area of 3.5 mm 2 in order to keep the RC constant small.
UV-Vis-NIR Spectroscopy: Absorbance measurements were performed on an Agilent Cary 5000 UV-Vis-NIR spectrometer with a double-beam setup in standard 10.00 mm quartz cuvettes.For the optical energy gap determination, HTM solutions with a concentration of 30 μg mL −1 in CB were prepared.Doping of the HTMs was conducted by dissolving F 4 TCNQ and FK209 at high concentrations (40 and 200 g L −1 ) in acetonitrile and then adding these dopant solutions to HTM solutions (10 g L −1 in CB).Subsequently, the doped solutions were diluted to an HTM concentration of 30 μg mL −1 in CB.A CB reference was measured in the same holder as the solutions and subtracted from the other measurements during data analysis.
Photoelectron Spectroscopy in Air (PESA): Measurements of thin-film ionization potentials (IPs) were carried out on a RIKEN KEIKI AC-2E photoelectron spectrometer.The photoelectron yields were corrected for the quantity of light and fitted with a power number of 0.5 to obtain the IPs.
Solar Cell Characterization: Solar cells were characterized under 1 sun irradiation from a Sciencetech Lightline AX-LA200 solar simulator (AAA, ASTM E927), calibrated with a Newport 91150-KG5 silicon reference solar cell.J-V curves were recorded using a source measure unit (Keithley 2420) with sweeps at 300 mV s −1 in steps of 20 mV.First, the descending scan from 1.4 to −0.4 V was executed, and then the ascending scan from −0.4 to 1.4 V.After measurements under irradiation (descending -ascending), dark curves were captured (descending -ascending).Measurements were not corrected for spectral mismatch. [66]Spectral mismatch was estimated at 1.02 from other MAPbI 3 solar cells measured using the same setup.
MIS-CELIV: A Keysight 81150A function generator was employed to hold a constant injection voltage for a prolonged time, followed by a linear voltage ramp for charge extraction.A slope of 400 mV μs −1 was chosen for standard MIS-CELIV measurements of undoped layers and the doping-induced ohmic regime in doped layers.For measurements in the doping-induced capacitive regime, the ramp was lowered to 12 mV μs −1 .The resulting transient currents were sent through a transimpedance amplifier Femto DHPCA-100 and captured with a Keysight Infiniium DSO-S104A digital oscilloscope, averaging across 1024 to 4096 measurements for each transient curve.For ohmic regime extraction measurements, a 10 Ω resistor was added in parallel to the transimpedance amplifier input, which reduced the RC constant of the measurement setup to a sixth compared to RC constant with the usual 50 Ω load impedance.The standard deviations of parameters that were obtained from MIS-CELIV measurements were calculated from the standard deviations of the semiconductor layer thicknesses.Further information on how the charge carrier mobilities, doping concentrations and conductivities were calculated can be found in the descriptions of the graphs in the Supporting Information.It was found that it was not feasible to perform reliable transient measurements on samples doped with LiTFSI:tBP, presumably due to mobile Li + ions from LiTFSI.
Layer Thicknesses: TPCP and spiro-MeOTAD layer thicknesses were measured with a Dektak XT stylus profilometer on scratched thin films, averaging over several measurements.The standard deviations of the thickness measurements were used to calculate the standard deviations of parameters derived from MIS-CELIV measurements.
Cyclic Voltammetry: Cyclic voltammetry experiments were performed with a Gamry Interface 1010B in a three electrodes electrochemical cell.The electrochemical cell was equipped with a glassy carbon (GC) working electrode, Ag/AgNO 3 reference electrode, and a platinum counter electrode.
The experiments were performed in N 2 -saturated N,Ndimethylformamide containing 0.1 m [nBu 4 N][PF 6 ] as the electrolyte at a scan rate of 100 mV s −1 .The concentration of the investigated compounds was 3 mm.According to the IUPAC recommendation, ferrocene (Fc) was added as an internal standard after each experiment. [39]ynthesis: Details of synthetic procedures and further characterization of the compounds can be found in the Supporting Information.
Density Functional Theory Calculations: The TPCP conformers for the theoretical calculations were generated using the previously reported workflow. [6]The CREST sampling was omitted.2000 (4000 for TPCP-1) initial conformers were generated with the ETKDG (experimental-torsion distance geometry with basic knowledge) method and filtered using a root-mean-square deviation (RMSD) threshold of 2 Å (1 Å for TPCP-1) in the first step and of 1 Å in the second step. [67]Afterward, at most 10 of the remaining conformers (5 for TPCP-1, 8 for TPCP-2, and 10 for TPCP-3) were used for the DFT and TD-DFT calculations, which were performed using TURBOMOLE, version 7.4.1. [68]The conformers were optimized with the def2-SVP basis set, [69] and all calculations were run with the B3LYP functional, [70][71][72][73] using the Grimme DFT-D3 dispersion correction with Becke-Johnson damping. [74,75][78][79][80][81][82] In all cases, restricted calculations were performed for the closed-shell neutral TPCP molecules and unrestricted calculations for the TPCP cations.Starting with the optimized geometries of the lowestenergy conformer of neutral TPCPs, the respective cations were also optimized.Using the optimized structure of neutral and oxidized TPCPs, single-point calculations were performed using B3LYP/def2-TZVP, [83] from which the IPs and orbital energies were obtained.The IP was calculated as the difference between the total DFT energy for charged molecule (+1) and the total energy for the neutral molecule.[86][87][88][89] As a comparison, the excitations of the neutral TPCPs were also calculated with the def2-TZVP basis set and are shown in Figure S23 (Supporting Information).A consistent small blue-shift of 11-12 nm (0.056-0.061 eV) for the results obtained with the def2-SVP basis set was observed.Spectra were generated using Gaussian broadening with a full width at half maximum (FWHM) of 0.1 eV.An isovalue of 0.0005 a.u. was used for the visualization of electron density differences.

Figure 3 .
Figure 3. a) Tauc plots of the UV-vis absorbance spectra of the TPCP compounds in CB (30 μg mL −1 ).b) According to PESA measurements, the IPs range between 5.27 and 5.42 eV also matching the requirements of application in PSCs.Lower slopes in PESA measurements indicate lower photoelectron emission rate, e.g., due to the permeability and thickness of the layer, which are largely independent from the IP.

Figure 4 .
Figure 4. MIS-CELIV characteristics of thin films of neat TPCPs and spiro-MeOTAD.The relative permittivity  s of each compound was calculated from the respective displacement current J 0 .After injection through the MoO x /Ag top electrode by a constant injection voltage (V inj ), holes form a reservoir of charge carriers at the interface to the LiF layer.Subsequently, this charge carrier reservoir mimics an ohmic contact, producing a space-charge limited current upon extraction via a linear voltage ramp (400 mV μs −1 ) starting at t = 0.The slope of the transient current is related to the hole mobility μ h , which is largest for TPCP-1.

Figure 5 .
Figure 5. Calculated absorption spectra of the three TPCP molecules studied in their neutral and oxidized (p-doped) states using TD-B3LYP/def2-SVP level of theory in the gas phase.The electron density differences between the excited and the ground state of TPCPs, which in first place stem from some pronounced absorption bands, are depicted in visualizations of the molecules.Regions in red indicate a positive difference, i.e., electron accepting, while regions in blue indicate a negative difference, i.e., electron donating.

Figure 6 .
Figure 6.Absorbance spectra of doped (red and blue curves) and undoped (black curve) solutions of TPCP-1, TPCP-2, TPCP-3, and spiro-MeOTAD (30 μg mL −1 in CB), on logarithmic scale to better visualize the doping signatures.The addition of FK209 to the HTMs led to the emergence of absorption features of the HTM cations in the visible and near-infrared spectral regime, indicating doping.Upon addition of F 4 TCNQ, weak absorption features of the F 4 TCNQ − anion appeared.At 0.74 eV, the strong absorption of CB produced minor measurement artifacts.

Figure 7 .
Figure 7. Conductivity of TPCP-3 and spiro-MeOTAD thin films electrically doped with F 4 TCNQ or FK209 versus dopant concentration.F 4 TCNQdoping was less efficient in increasing the conductivity in TPCP-3 than in spiro-MeOTAD, but with FK209-doping the conductivities of TPCP-3 and spiro-MeOTAD converged.We note that the data points are not evenly spaced along the x-axis and dashed lines between data points serve to guide the eye.

Figure 8 .
Figure 8. a) J-V curves of solar cells comprising HTLs of TPCP-3.Devices with undoped TPCP-3 showed a significant s-shape that was mitigated by FK209-doping and F 4 TCNQ-doping, but not by LiTFSI:tBP-doping or LiTFSI:tBP:FK209-doping. b) All doping concepts diminished the s-shape in J-V curves of solar cells comprising spiro-MeOTAD.

Table 3 .
Comparison of performance parameters of solar cells comprising HTLs from TPCP-3 or spiro-MeOTAD and employing different doping strategies before and after annealing at 70 °C for 1 h.