Formamidine acetate (FAAc, 99%), cesium acetate (CsAc, 99%), methylamine (MA, AR, 30-33 wt.% in ethanol), and lead acetate trihydrate (PbAc₂·3H₂O, AR, 99.5%) were purchased from Aladdin. Hydroiodic acid (HI, 47 wt.% in water) was purchased from Macklin. PbI₂ (98%) was acquired from TCI (Shanghai) Development Co., Ltd. 2-PACz, PTAA, 2,9-Dimethyl-4,7-diphenyl-1,10-phenanthroline (BCP, 99.9%), methylammonium chloride (MACl, > 99.5%), piperazinium iodide (PI, 99%), C₆₀ (99.9%), and phenmethylammonium iodide (PMAI, 99.9%) were bought from Xi’an Yuri Solar Technology Co., Ltd. Cu (99.999%) were purchased from Fuzhou Innovation Photoelectric Technology Co. Ltd. N’, N’-Dimethylformamide (DMF, 99.8%), isopropanol (IPA, 99.5%), chlorobenzene (CB, 99.8%), dimethyl sulfoxide (DMSO, 99.9%) were obtained from J&K Scientific Co., Ltd. All chemicals were used as received, without additional purification.

The perovskite CsPbI₃, MAPbI₃, and FAPbI₃ microcrystals were synthesized using our previously established work^[7].

A Keithley 2400 source meter and a Newport solar simulator (Enlitech Solar Simulator SS-F7-3A) were employed to characterize the device performance under standard AM 1.5G conditions (100 mW cm^-2). The light intensity was precisely calibrated using a NIST-certified single-crystal Si solar cell (Newport 532 ISO1599).

An FLS1000 spectrometer (Edinburgh Instruments Ltd.) equipped with an Xe lamp was employed to measure steady-state PL spectra. TRPL decay kinetics were recorded at 787 nm with an excitation wavelength of 468 nm. The photoluminescence was excited from the perovskite side of the PVK/ITO substrate configuration. TRPL measurements were performed under two excitation geometries. The excitation laser was incident either from the perovskite side or from the HTL side, allowing comparison of carrier dynamics at the two interfaces. An ultrafast transient absorption spectrometer (Time-tech Spectra Limited Company) was used for TAS measurements. Morphological features were examined by field-emission SEM (SU8230, Hitachi).

To compare charge carrier dynamics influenced by different hole transport layers, we investigated pristine (FAPbI₃)_0.85(MAPbI₃)_0.1(CsPbI₃)_0.05 perovskite (PVK) films as a control sample (for preparation and characterization see Ref.^[7]). Additionally, we examined PVK films spin-coated on two distinct HTL materials: PTAA, a commonly used polymeric HTL (denoted as PVK/PTAA), and 2-PACz, a self-assembled monolayer-based HTL (denoted as PVK/2-PACz). These samples allow us to study how HTL materials impact charge carrier behavior, energy alignment, and interfacial properties, providing insights into optimizing HTL designs for improved perovskite optoelectronic devices. Figure 1A presents the linear absorption and PL spectra of the three investigated samples. The steady-state absorption spectra exhibit nearly identical profiles for the pristine PVK film and PVK films with HTLs, with a distinct and sharp absorption edge around 800 nm. Using the Tauc plot method (see Supplementary Figure 1), we determined that the bandgap remains nearly constant across the samples, with values of 1.552 eV, 1.554 eV, and 1.555 eV for pristine PVK, PVK/PTAA, and PVK/2-PACz films, respectively. These results confirm that the addition of HTLs does not significantly impact the intrinsic bandgap of the perovskite material, allowing for a direct comparison of their influence on charge carrier dynamics and interfacial properties. Correspondingly, Figure 1A shows that while the pristine PVK exhibits a PL peak at 792 nm, a slight blue shift is observed in the HTL-coated films. These findings indicate that while the bulk optical properties remain largely unchanged, HTLs subtly influence interfacial electronic properties and carrier dynamics. Notably, the PL intensity decreased tenfold after coating with PTAA, while it increased significantly with 2-PACz, consistent with the reported trend^[17,20].

To further assess the influence of different HTLs on interfacial carrier dynamics, TRPL measurements under 468 nm excitation were performed from both the PVK side and the ITO side, as shown in Figure 1B and C. Average carrier lifetimes were extracted through exponential fitting and compared across the three samples in Figure 1D, enabling the separation of intrinsic perovskite recombination from HTL-induced interfacial effects. When excited from the ITO side, the fitted lifetimes for the PVK, PVK/PTAA, and PVK/2-PACz samples are 3,515 ± 21 ns, 81 ± 0.8 ns, and 1,090 ± 8 ns, respectively, compared with 801 ± 6 ns, 75 ± 1.4 ns, and 591 ± 4.6 ns under PVK-side excitation. Compared with PVK-side excitation, pronounced lifetime extensions are observed for the pristine PVK and PVK/2-PACz samples, whereas the PVK/PTAA sample exhibits only a marginal increase. This distinct behavior indicates that the influence of excitation geometry strongly depends on the interfacial properties of the HTL, with 2-PACz effectively modifying interfacial carrier dynamics through defect passivation and carrier transfer, while PTAA shows only a negligible interface-dependent effect. To quantitatively evaluate this asymmetry, we introduce a difference factor $$ \beta=\frac{\tau_{H T L-s i d e}-\tau_{P V K-s i d e}}{\tau_{P V K-s i d e}} \\ $$, which reflects the relative enhancement in carrier lifetime induced by HTL-side excitation.

Notably, the pure PVK exhibits the highest β value, indicating a much longer lifetime under ITO -side excitation compared to PVK-side excitation. Given that the penetration depth at 468 nm excitation (196 nm) is smaller than the PVK thickness (~ 700 nm), the TRPL response predominantly reflects the recombination characteristics near the excitation surface. The significantly shorter lifetime observed for PVK-side excitation therefore suggests severe non-radiative recombination at the PVK surface, likely arising from a high density of surface defects. In contrast, the PVK/ITO interface shows substantially improved quality, with fewer defect-mediated recombination pathways, leading to reduced non-radiative losses and prolonged carrier lifetimes.

In the PVK/HTL system, however, the PL intensity and carrier lifetime are governed by two competing mechanisms: (i) the HTL provides an efficient pathway for extracting photoexcited holes from the perovskite, which suppresses radiative recombination; and (ii) the HTL can passivate interfacial defect states, reducing non-radiative recombination and thereby enhancing radiative emission. Compared to pristine PVK, the enhanced PL intensity observed in PVK/2-PACz indicates reduced non-radiative recombination and improved interfacial passivation by 2-PACz, consistent with previous reports^[21,22]. We note that steady-state PL intensity primarily reflects quasi-Fermi level splitting and interfacial recombination losses rather than charge-extraction efficiency, as emphasized in recent studies of perovskite optoelectronics^[18,19]. Although TRPL lifetimes alone cannot uniquely distinguish extraction from recombination processes, the excitation-geometry-dependent measurements reveal distinct interfacial carrier dynamics. Combined with TAS measurements discussed below, these observations support more efficient interfacial hole transfer in PVK/2-PACz than in PVK/PTAA. When excited from the PVK side, carriers must diffuse towards the PVK/2-PACz interface before transfer occurs, during which scattering and recombination weaken the built-in electric field and slow extraction. In contrast, ITO -side excitation enables ultrafast hole transport directly into 2-PACz, rapidly establishing an interfacial electric field that suppresses further transfer. This screening effect results in a carrier lifetime nearly twice that obtained under PVK-side excitation. For PVK/PTAA, however, the carrier lifetime shows minimal dependence on excitation direction (β ≈ 0), indicating weaker hole extraction and poorer interface passivation compared with 2-PACz. This suggests that PTAA provides limited interfacial improvement and inefficient carrier transfer in contrast to 2-PACz. We note that a rigorous assessment of charge extraction under operating conditions may additionally be obtained through bias-dependent PL measurements that monitor the evolution of emission from open-circuit to short-circuit conditions. Such measurements have recently been proposed as a direct probe of extraction efficiency in perovskite devices^[19]. While beyond the scope of the present study, our conclusions regarding carrier transfer are supported by the combined TAS and excitation-geometry-dependent TRPL results.

To understand the impact of HTLs on the photophysical properties of PSCs, we first examined the photoexcited carrier relaxation in the pristine PVK film using TAS. Figure 2A-C present the 2D pseudo-color maps of the TAS under 660 nm excitation with pump fluences ranging from 4.5 to 59.1 µJ/cm². The 660 nm pump wavelength was selected to avoid excitation of the HTL in later comparative measurements, ensuring that the observed spectral responses originate solely from the perovskite layer. In the transient absorption (TA) spectra (ΔA/A), negative signals (red) represent ground-state bleach (GSB), while positive signals (blue) correspond to excited-state absorption (ESA). At a fluence of 4.5 µJ/cm² [Figure 2A], the absorption depth of perovskite at 660 nm (~ 249.5 nm) yields a carrier density of 6.04 × 10¹⁷ cm^-3, at which thermal and Auger recombination effects can be safely neglected^[23]. The pronounced GSB at 780 nm, which is consistent with the optical bandgap of 1.555 eV (~ 797 nm) extracted from the steady-state absorption spectrum, is therefore attributed to the band filling effect^[24,25]. As shown in Supplementary Figure 2, the competition between bandgap renormalization (redshift) and the Burstein-Moss effect (B-M effect, blueshift) leads to a net redshift of the GSB peak at longer timescales. Meanwhile, the weak ESA near 800 nm vanishes within 1 ps and is subsequently overshadowed by the dominant GSB. With increasing pump fluences, as shown in Figure 2B and C, the GSB peak broadens and redshifts, indicating enhanced carrier-carrier interactions that accelerate GSB recovery and promote faster carrier thermalization at higher excitation densities. Notably, when the pump fluence exceeds 22.7 µJ/cm², a second ground-state bleach (GSB2) emerges around 815 nm, whose origin will be discussed in the following section.

We further performed global fitting of the TA spectra in Figure 2A-C to extract the corresponding decay-associated spectra (DAS), as shown in Figure 2D-F^[26]. In Figure 2D, the fitting of the TA dynamics in Figure 2A reveals two distinct decay components, indicating multiple relaxation pathways. The fast decay component, occurring within the sub-picosecond regime (blue curve), exhibits a spectral profile resembling the first derivative of the absorption band, characteristic of transient band shifts and state filling of accumulated carriers. This behavior is consistent with the instantaneous shift of the absorption band reported previously^[26-28], and is therefore assigned to hot-carrier cooling immediately after photoexcitation. In contrast, the slow decay component (orange curve) on the nanosecond timescale displays a red-shifted bleach-like spectrum, corresponding to the recovery of the ground state. We attribute this component to bimolecular of photogenerated electrons and holes^[29].

At higher fluences, as shown in Figure 2E, three decay components are required to adequately reproduce the TA dynamics. The additional intermediate component (green curve), appearing on the ~ 100 ps timescale, is attributed to Auger recombination based on the following evidence: (i) The carrier density at this fluence reaches 3.02 × 10¹⁸ cm^-3, where Auger recombination becomes significant and can no longer be neglected^[23]; (ii) As shown in Supplementary Figure 3, the ΔA^-2 kinetics at the GSB peak exhibit a linear dependence after 15 ps^[26], indicating that three-body Auger recombination dominates the carrier decay in this time window, whereas the sublinear behavior within the first ~ 15 ps will be discussed later; (iii) As illustrated in Figure 2F, the amplitude of this component decreases with increasing fluence, consistent with enhanced carrier-carrier interactions accelerating the Auger process. Meanwhile, the strengthened carrier interactions at high excitation densities also reduce the slow decay component (orange curve), with its lifetime decreasing from 1.441 ns to 1.076 ns. In contrast, the fast decay component exhibits an extended carrier lifetime at higher fluence, likely due to mechanisms such as the hot phonon bottleneck, giant polaron screening, or acoustic-optical phonon up-conversion^[1,30,31]. Notably, the spectral profile of this fast component shows a dip around 815 nm, resembling the shape of a bleach feature. This behavior suggests that the emergence of the GSB2 signal interferes with the hot-carrier cooling pathway, altering the early relaxation dynamics.

To further elucidate the influence of HTLs, we performed TAS measurements on PVK/PTAA and PVK/2-PACz films from the PVK side (see Supplementary Figures 4 and 5). In both cases, the DAS profiles remain largely unchanged across different pump fluences, indicating that incorporating HTLs does not modify the intrinsic fluence-dependent carrier relaxation pathways in PVK. Instead, the HTLs primarily influence interfacial carrier processes, specifically hole extraction from PVK into the HTL and the first-order Shockley-Reed-Hall recombination at the interface mediated by trap states.

To compare the influence of HTLs on ultrafast carrier dynamics, we plotted the normalized TA kinetics at GSB1 under a pump fluence of 4.5 µJ/cm² in Figure 3A. On longer timescales (within 7 ns), PVK/2-PACz exhibits the fastest GSB1 than both pristine PVK and PVK/PTAA. The GSB relaxation is governed by two competing processes: defect-assisted nonradiative recombination and hole transport from PVK into the HTL^[32-35]. Although PVK/PTAA shows accelerated decay due to interfacial defects, PVK/2-PACz still decays more rapidly, confirming that efficient hole extraction - not defect-driven recombination - dominates at this low carrier density. Taken together with the excitation-geometry-dependent TRPL measurements, the faster GSB recovery in PVK/2-PACz suggests more efficient interfacial carrier transfer than in PVK/PTAA. Because steady-state PL predominantly reflects interfacial passivation, the TAS results provide complementary evidence that 2-PACz not only suppresses non-radiative recombination but also facilitates carrier extraction across the PVK/HTL interface. As the excitation density increases, however, the decay kinetics of all samples gradually converge as shown in Supplementary Figure 6. This behavior arises from interfacial hole accumulation and the high carrier density within the HTL, which saturates the hole-transfer process and limits further extraction^[36], leading to overlapping decay curves.

It is worth noting that even at a high fluence of 59.1 µJ/cm², the HTLs still retain a weak degree of hole extraction. However, this does not directly imply poor device performance, because the extracted holes remain confined within the HTL rather than being collected, as would occur in a full device stack^[36]. In complete PSCs, imbalanced electron or hole transport, together with interfacial trap states, can induce charge accumulation at the PVK/HTL interface. Such accumulation reduces the open-circuit voltage and introduces pronounced current-voltage hysteresis, ultimately compromising device stability and operational durability^[37]. These observations underscore the importance of interface engineering to suppress trap-assisted barriers and maintain balanced charge extraction, both of which are crucial for pushing PSC efficiencies closer to their theoretical limits.

The appearance of a second GSB was previously reported by Zheng et al. as a new trap-related channel in PVK nanoparticles^[26]. In contrast, our observations reveal that GSB2 also emerges in thin-film PVK under high excitation. By tracking GSB2 dynamics at a fluence of 59.1 µJ/cm², we find that its temporal evolution diverges from that of GSB1 after 1,500 ps, as indicated by the green squares in Figure 3B. The GSB2 kinetic trace bends downward at longer delays, indicating that additional energy progressively flows into this channel. For clarity, the GSB2 signal in Figure 3B is scaled by a factor of 13.1.

To further probe the origin of GSB2, we examined the fluence-dependent normalized TA spectra at 2 ps [Figure 3C]. The main bleach peak blue-shifts and broadens with increasing fluence, consistent with the B-M band-filling effect^[25]. Interestingly, at intermediate fluences of 22.7 and 31.8 µJ/cm², the spectra show negligible blue-shift or broadening, deviating from the expected B-M trend. According to the B-M model, the bandgap shift scales with carrier density n as follows^[38]:

(1) $$ \Delta E_{g}^{B M}=\frac{\hbar^{2}}{2 m_{e h}^{*}} \left(3 \pi^{2} n\right)^{2 / 3} $$

where $$ \Delta E_{g}^{B M} \\ $$ is the B-M shifts, $$ \hbar \\ $$ is the Planck constant, and $$ m_{e h}^{*} $$ is the reduced effective mass of the photogenerated electron-hole pair. We therefore extracted $$ \Delta E_{g}^{B M} $$ as a function of n^2/3 from the TA spectra in Figure 3C and Supplementary Figure 7 to quantify deviations from the B-M prediction and identify the onset of GSB2-related processes.

However, as shown in Figure 3D, a plateau in $$ \Delta E_{g}^{B M} \\ $$ appears at n^2/3 ~ 2 × 10¹² cm^-2 (corresponding to a fluence of 22.7 µJ/cm²) for all three samples. Beyond this plateau, the slope changes. Since this fluence coincides with the onset of GSB2, we attribute the plateau to the emergence of a new relaxation channel associated with GSB2. By analyzing the slope before and after the plateau, we extract the variation in the reduced effective mass $$ m_{e h}^{*} $$. Interestingly, $$ m_{e h}^{*} $$ increases in the PVK/HTL systems but decreases in pristine PVK, suggesting that the HTL modifies the early-time carrier-phonon coupling environment.

To elucidate the origin of this slope change, we further analyzed ΔA^-2 dynamics for GSB1 at 22.7 µJ/cm² [Figure 3E]. All three samples exhibit a linear ΔA^-2 dependence from 15-140 ps, confirming that the ~ 110 ps component in the DAS spectra arises from Auger recombination. In contrast, the sublinear behavior within the first 15 ps indicates the involvement of a higher-order process beyond Auger recombination. At these high carrier densities, excess charges (either electrons or holes) can strongly couple to ionic vibrations, forming large polarons^[39]. Polaron formation modifies the carrier scattering landscape, increases the effective mass, and alters subsequent relaxation pathways^[40].

We therefore propose that the emergence of GSB2 at high fluence is consistent with polaron formation. Alternative explanations, including simple band filling, local heating, and carrier-carrier interactions, are unlikely to fully account for the appearance of a distinct bleach feature with markedly different temporal dynamics. The associated lattice distortion may be related to previously reported Rashba spin splitting in highly excited perovskites^[41-44], thereby opening an additional carrier-relaxation channel from the GSB1 (1.60 eV) to the lower-energy GSB2 (1.52 eV), as illustrated in Figure 3F. This mechanism explains the anomalous long-lived behavior of GSB2 in Figure 3B: once carriers accumulate in this polaron-mediated channel, the resulting screening suppresses Coulomb interactions, thereby slowing recombination and preventing GSB2 relaxation back to the ground state within our detection window. The extended persistence of carriers in this “hot” polaronic channel suggests that such states may be advantageous for long-range transport and extraction into the adjacent HTL.

In pristine PVK, polarons exhibit strong repulsive interactions when oppositely charged species remain spatially separated, and the resulting dielectric screening is pronounced. However, as the excitation density increases beyond the threshold fluence of 22.7 μJ/cm², the average polaron-polaron spacing decreases to the point where their wavefunctions begin to overlap. This overlap drives a transition from repulsive to attractive interactions^[39], which accelerates the recombination of oppositely charged polarons. Consequently, the reduced effective mass decreases above this threshold. Meanwhile, the presence of polarons is known to impede hot-carrier cooling^[44,45], consistent with our observation that the fast decay component in the DAS becomes slower with increasing excitation intensity.

When a HTL is introduced, efficient hole extraction reduces the population of positively charged polarons within the PVK layer, causing electron-induced lattice distortion to dominate the polaronic landscape. As the carrier density continues to rise, the effective special volume available to electron polarons becomes larger relative to that in pristine PVK, and therefore the system does not undergo the same repulsion-to-attraction transition. Instead, in the PVK/HTL structure, $$ m_{e h}^{*} $$ increases beyond the 22.7 μJ/cm² threshold.

This contrasting behavior originates from HTL-induced electron-hole separation at the PVK/HTL interface, which establishes a local electric field that perturbs free diffusion. The resulting spatially nonuniform carrier distribution increases the probability of carrier localization, thereby strengthening electron-lattice interactions^[46]. These interactions stabilize polaronic states and prevent their collapse even at high carrier densities. As a result, no decrease in the effective mass is observed across the accessible excitation range. Such polaron-stabilizing behavior could be further resolved using advanced nano-imaging techniques^[47].

The inverted perovskite solar cells employed in this work for comparative performance analysis had the device architecture ITO/HTL/PVK/C₆₀/BCP/Cu, as schematically represented in Figure 4A. A cross-sectional scanning electron microscopy (SEM) view of the complete device architecture is displayed in Figure 4B. The current density-voltage (J-V) characteristics of PTAA-based and 2-PACz-based PSCs are shown in Figure 4C, with devices without an HTL included for comparison. The corresponding photovoltaic parameters are summarized in Table 1. Among all the fabricated devices, the PVK/2-PACz-based PSC achieved the highest performance, which yielded a power conversion efficiency (PCE) of 26.03% with a short-circuit current (J_SC) of 25.16 mA/cm² and an open-circuit voltage (V_OC) of 1.20 V. In contrast, the PVK/PTAA-based device reaches a PCE of 23.12%. The V_OC deficits of the 2-PACz- and PTAA-based devices are 0.355 V and 0.404 V, respectively, highlighting the superior voltage management enabled by 2-PACz.

As shown in Figure 4D, MPP tracking of the champion 2-PACz-based device was performed under continuous AM 1.5G illumination with an applied bias of 1.06 V over 600 s, yielding a stabilized efficiency of 25.89%. Device reproducibility was evaluated across 40 devices [Figure 4E]. The 2-PACz-based PSCs exhibit good reproducibility, with an average PCE of 25.39% and V_OC of 1.190 V, whereas PTAA-based PSCs showed an average PCE of 22.37% and V_OC of 1.136V. These results clearly demonstrate the superior performance of 2-PACz as an HTL. The performance enhancement arises mainly from reduced non-radiative recombination losses in the PVK/2-PACz system. Compared to traditional PTAA, 2-PACz improves film uniformity and interfacial quality. As shown in Figure 1A, the PVK/2-PACz interface exhibits enhanced PL intensity due to effective defect passivation. This leads to higher fill factor (FF) and V_OC, boosting the overall device efficiency. The nearly unchanged J_SC values, together with the substantial increases in V_OC and FF, indicate that suppression of interfacial non-radiative recombination is the primary contributor to the improved device performance. TAS measurements further suggest that more efficient interfacial carrier transfer in PVK/2-PACz helps reduce charge accumulation and supports efficient device operation^[48].

To further evaluate the device performance of solar cells with different HTLs under high carrier densities, we tested concentrator solar cells using lenses to focus light from a solar simulator. Performance metrics for PTAA-based and 2-PACz-based devices under 1 sun and 5 sun illumination are summarized in Table 2. The 2-PACz-based device exhibits a V_OC of 1.27 V under 5 suns, with a corresponding PCE of 27.24%, significantly higher than the 26.03% PCE observed under 1 sun. In contrast, the PTAA-based solar cell shows an increase in PCE from 23.12% under 1 sun to 24.73% under 5 suns. However, the performance of the PTAA-based device remains slightly inferior to that of the 2-PACz-based device. As the intensity of the sun increases, the radiant temperature rises [Supplementary Figure 8]. To reduce the temperature effect on the device and further verify the influence of light irradiation, we tested the photocurrent of perovskite solar cells with varying light intensities of a 660nm laser [Supplementary Figure 9]. We found that compared to devices without HTL and those with PTAA, the device with a 2-PACz hole transport layer exhibited the highest current and the steepest slope with changing light intensity, indicating better photoresponse and fewer non-radiative recombination losses.

Although the present study focuses on ultrafast spectroscopic investigations of HTL-dependent carrier dynamics, complementary electrical transient measurements, including transient photovoltage, transient photocurrent, and space-charge-limited current analysis, could provide further insight into carrier transport and recombination under device operating conditions and will be the subject of future work.