Introduction

Continued development of organic donor and acceptor (D/A) photovoltaic materials1,2,3,4,5 has resulted in power conversion efficiencies (PCE) of organic solar cells (OSCs) based on the bulk-heterojunction (BHJ) concept6,7 approaching and exceeding 19%8,9,10,11. Both the photovoltaic external quantum efficiency (EQEPV) and the fill-factor (FF) of state-of-the-art OSCs have already been improved to over 80%12,13,14,15, approaching those of inorganic solar cells. However, the open-circuit voltage (VOC) is still low, and increasing it is critically important for further improving the performance of OSCs.

VOC is limited in OSCs due to significant voltage losses associated with non-radiative recombination of charge carriers, referred to as non-radiative voltage losses (VNR)16,17. The reported VNR for high-efficiency OSCs is as high as 0.20–0.30 V18,19, while high-efficiency inorganic solar cells have a VNR of only 0.05 V20,21. In 2007, Rau demonstrated that VNR and EQEEL are related by Eq. (1)22,

$${V}_{{NR}}=-\frac{{k}_{B}T}{q}{{{{\mathrm{ln}}}}}({EQ}{E}_{{EL}})$$
(1)

where kB is the Boltzmann constant, q is the elementary charge, and T is temperature. Therefore, a high VNR in OSCs goes hand-in-hand with a low electroluminescence quantum efficiency (EQEEL), typically in the range of 10−5 % to 10−2 %.

For OSCs, EQEEL is primarily determined by the properties of the charge transfer (CT) state formed at the D/A interfaces in the BHJ active layer, given by23,24

$${{EQE}}_{{EL}}=\frac{{k}_{r}^{{CT}}}{{k}_{r}^{{CT}}+{k}_{{nr}}^{{CT}}}=\frac{{k}_{r}^{{CT}}}{{k}_{{CT}}}$$
(2)

where \({k}_{r}^{{CT}}\) and \({k}_{{nr}}^{{CT}}\) are the radiative and non-radiative decay rate constants of the CT state, respectively, and \({k}_{{CT}}\) is the overall decay rate constant of the CT state. Due to strong vibrational coupling between the CT state and the ground state25, \({k}_{{nr}}^{{CT}}\) of OSCs is high23,25, being the main reason for the low EQEEL.

Several attempts have been made to manipulate the decay dynamics of the CT state and thereby increase the EQEEL of OSCs. Eisner et al. 26 reported that increasing the degree of hybridization of the CT state and the singlet (S1) state of the acceptor could result in increased \({k}_{r}^{{CT}}\), which in turn increased EQEEL and decreased VNR, similar to that reported by Chen et al. 27 and Qian et al. 28 However, the increase in \({k}_{r}^{{CT}}\) also leads to an increased radiative voltage loss29,30. Ullbrich et al. reported that increasing the energy of the CT state (ECT) could lead to weakened vibrational coupling between the CT state and the ground state25, giving rise to reduced \({k}_{{nr}}^{{CT}}\), increased EQEEL, and reduced VNR, without affecting the radiative voltage losses. However, increasing ECT of OSCs requires the use of larger bandgap organic absorbers with a narrower absorption spectral range, which limits the short-circuit current density (JSC) and consequently PCE.

In 2021, some of us demonstrated that increasing the spacing between D/A in the BHJ blend (DA spacing) could also lead to decreased \({k}_{{nr}}^{{CT}}\)31. In that case, EQEEL was increased for the solar cell with increased DA spacing, and VNR was reduced without reducing JSC. However, VNR remained high, over 0.3 V, and the FF was limited for the solar cell with increased DA spacing. Thus, the performance of the solar cell was considerably worse than that of state-of-the-art OSCs. Recently, Gillett et al. demonstrated that the triplet states were an additional non-radiative decay channel for the CT states32. Increasing the degree of hybridization between the donor and acceptor molecules effectively prevented the decay of CT states via the triplet states, leading to increased EQEEL. Nevertheless, despite considerable efforts that have been spent on the manipulation of the decay dynamics of CT state33,34,35,36,37,38, EQEEL is still low for the high-efficiency OSCs. This low EQEEL problem, leading to high VNR and limited photovoltaic performance, seemed inevitable for OSCs39.

In this work, we show that it is possible to increase the EQEEL of high-efficiency OSCs without compromising either JSC or FF, and without elevating radiative voltage losses. This is achieved by manipulating the dissociation dynamics of the S1 state of the acceptor (S1A) rather than the decay dynamics of the CT state. Specifically, we reevaluate the dynamic processes of excited states in OSCs and illustrate that, beyond \({k}_{r}^{{CT}}\) or \({k}_{{nr}}^{{CT}}\), the rate constant governing the dissociation of S1A (\({k}_{{DS}}\))—traditionally associated with charge carrier generation rate—likewise holds a significant role in determining the EQEEL of OSCs. Then, we demonstrate that, contrary to the prevailing belief that maximizing kDS is essential for achieving high PCE in OSCs, a decrease in kDS can actually yield enhanced solar cell performance, provided that kDS remains above the threshold required for efficient exciton dissociation. In light of this, we propose an alternative avenue for research, one that centers on the manipulation of kDS. This approach presents a promising solution to address the VNR issue and to overcome the VOC bottleneck in OSCs.

Results

Excited state dynamics in OSCs

Equation (2) was derived by assuming that the CT state was the sole decay channel for the excited states in OSCs23. However, this assumption does not hold for state-of-the-art OSCs based on non-fullerene acceptors (NFAs). In NFA OSCs, electroluminescence (EL) also includes S1A emission, resulting from the back transfer of CT state to the singlet state26,40,41. This introduces S1A as an additional decay channel for free charge carrier recombination in NFA OSCs. A schematic illustration of the excited state dynamics for NFA OSCs under EL or EQEEL measurements is presented in Fig. 1a. When singlet population upon free charge carrier recombination becomes significant, the expression for the EQEEL for NFA solar cells becomes

$${{EQE}}_{{EL}}=\frac{{n}_{{S}_{1}}{k}_{r}^{{S}_{1}}+{n}_{{CT}}{k}_{r}^{{CT}}}{{n}_{{S}_{1}}{k}_{{S}_{1}}+{n}_{{CT}}{k}_{{CT}}}$$
(3)

where nCT and nS1 are the concentrations of the CT state and S1A, respectively, kCT and kS1 are the decay rate constants of the CT state and S1A, respectively. Here, we assume that there is no decay of excited states via molecular triplet states.

Fig. 1: Determining factors for the performance of non-fullerene organic solar cells.
figure 1

a Dynamics of excited states in non-fullerene (NFA) organic solar cells (OSCs) under electroluminescence measurement. S1A, CT, and CS are abbreviations for acceptor singlet, charge transfer, and charge separated states, respectively. Note that for typical NFA OSCs, the energy of the singlet (S1) state of the donor is much higher than that of the acceptor, and thus the concentration of the excited donor S1 state is negligible at thermal equilibrium. b Power conversion efficiency (PCE) as a function of rDS and rBK, calculated using Eq. (7). \({\eta }_{{S}_{1}}\) and \({\eta }_{{CT}}\) of the solar cell are assumed to be 1% and 10−5%, respectively. The radiative limit for the open-circuit voltage, the upper limit for the short-circuit current density, and the fill factor of the solar cell are assumed to be 1.10 V, 32 mA cm−2, and 80%, respectively.

Full size image

In the case of OSCs operating as an LED in steady state, the generation rate of S1A must be in equilibrium with the rate at which S1A disappears. Then, we have

$${n}_{{CT}}{k}_{{BK}}={n}_{{S}_{1}}{k}_{{S}_{1}}+{n}_{{S}_{1}}{k}_{{DS}}$$
(4)

where kDS is the dissociation rate constant, i.e., the rate constant for the transition from S1A to the CT state, kBK is the rate constant for the transition from the CT state to S1A. Combining Eqs. (3) and (4), we derive

$${{EQE}}_{{EL}}=\frac{{\eta }_{{S}_{1}}}{1+\left(\frac{1+\frac{{k}_{{DS}}}{{k}_{{S}_{1}}}}{\frac{{k}_{{BK}}}{{k}_{{CT}}}}\right)}+\frac{{\eta }_{{CT}}}{1+\left(\frac{\frac{{k}_{{BK}}}{{k}_{{CT}}}}{1+\frac{{k}_{{DS}}}{{k}_{{S}_{1}}}}\right)}=\frac{{\eta }_{{S}_{1}}}{1+\left(\frac{1+{r}_{{DS}}}{{r}_{{BK}}}\right)}+\frac{{\eta }_{{CT}}}{1+\left(\frac{{r}_{{BK}}}{1+{r}_{{DS}}}\right)}$$
(5)

where \({\eta }_{{CT}}\) (\(=\frac{{k}_{r}^{{CT}}}{{k}_{{CT}}}\)) and \({\eta }_{S1}\) (\(=\frac{{k}_{r}^{{S}_{1}}}{{k}_{{S}_{1}}}\)) are the emission quantum efficiencies of the CT state and S1A, respectively, rDS is defined as the ratio between kDS and kS1, and rBK is the ratio between kBK and kCT. Note that \({\eta }_{{S}_{1}}\) is orders of magnitude higher than ηCT for OSCs27. Therefore, from Eq. (5), one can derive that EQEEL increases when reducing kDS, and the EL peak of the solar cell shifts from CT state emission to S1A emission. Note that when the decay of excited states via molecular triplet states is involved, Eq. (3) requires modification, resulting in a new expression for the relationship between EQEEL and the rate constants, which differs slightly from Eq. (5). Nevertheless, the conclusion that EQEEL increases with the reduction of kDS remains valid. Further details on the derivation of Eq. (5), as well as the derivation of the expression with the decay via triplets included, can be found in Supplementary Note 1.

Similar to a derivation in the literature42, by assuming that the built-in electric field is sufficiently large to dissociate CT states and extract free charge carriers, the internal quantum efficiency (IQE) of OSCs under short circuit can be expressed as (derivation of Eq. (6) is provided in the Supplementary Note 1)

$${IQE}=\frac{\frac{{k}_{{DS}}}{{k}_{{S}_{1}}}}{1+\frac{{k}_{{DS}}}{{k}_{{S}_{1}}}}=\frac{{r}_{{DS}}}{1+{r}_{{DS}}}$$
(6)

Therefore, the maximum achievable IQE decreases with reducing kDS, but remains close to its maximal value as long as \({k}_{{DS}}\gg {k}_{{S}_{1}}\). Equations (5) and (6) express that for most of the typical OSCs with high kDS, the CT state is the dominant emission channel and EQEEL is primarily determined by ηCT. Thus, EQEEL of these solar cells is low, VNR is high, and IQE is also high. On the other hand, for OSCs with kDS being lower than \({k}_{{S}_{1}}\), IQE is limited, but the dominant emission channel in these solar cells is S1A, giving rise to high EQEEL and low VNR. There is clearly a trade-off between IQE and EQEEL, and thus, between JSC and VOC in BHJ OSCs, governed by kDS.

To evaluate how kDS affects the overall performance of OSCs, we assume that \({\eta }_{S1}\) and \({\eta }_{{CT}}\) of a solar cell are 1% and 10−5% 27, respectively, the radiative limit for VOC (\({V}_{{OC}}^{{rad}}\)) of the solar cell is a constant, 1.10 V (implying that \({k}_{{S}_{1}}\) is lower than \({k}_{{CT}}\)), the upper limit for JSC(\(\, {J}_{{SC}}^{\max }\)) is 32 mA cm−2, and the FF is 80%. Then, we relate the device PCE to EQEEL and IQE, and thus, to \({r}_{{DS}}\) and \({r}_{{BK}}\), via

$${PCE}=\frac{{J}_{{SC}}\cdot {V}_{{OC}}\cdot {FF}}{{P}_{{in}}}= \frac{{J}_{{SC}}^{\max }\cdot {IQE}\cdot \left({V}_{{OC}}^{{rad}}+\frac{{kT}}{q}{{{{\mathrm{ln}}}}}\left({EQ}{E}_{{EL}}\right)\right)\cdot {FF}}{{P}_{{in}}} \\= \frac{{J}_{{SC}}^{\max }\cdot \frac{{r}_{{DS}}}{(1+{r}_{{DS}})}\cdot \left({V}_{{OC}}^{{rad}}+\frac{{kT}}{q}{{{{\mathrm{ln}}}}}\left(\frac{{\eta }_{S1}}{1+\left(\frac{1+{r}_{{DS}}}{{r}_{{BK}}}\right)}+\frac{{\eta }_{{CT}}}{1+\left(\frac{{r}_{{BK}}}{1+{r}_{{DS}}}\right)}\right)\right)\cdot {FF}}{{P}_{{in}}}$$
(7)

where Pin is the incident light intensity (100 mW cm−2). Accordingly, we calculate PCE as a function of rDS and rBK, as shown in Fig. 1b.

For typical high-efficiency OSCs, the IQE is close to 100%, and VNR is also high, indicating that rDS is high and rBK is low. This signifies that the achievable PCE for typical OSCs is depicted by the greenish region in the lower-left corner of Fig. 1b. Therefore, three important conclusions regarding the impact of decreasing kDS and rDS on the performance of OSCs can be inferred from Fig. 1b:

  1. 1.

    When rBK is greater than a certain threshold, roughly 10−3 in the example above, the PCE of the solar cell will increase from 18% to over 25% with the decrease in kDS, until kDS becomes excessively low, constraining IQE (indicated by the red arrow ① in Fig. 1b).

  2. 2.

    When rBK is lower than 10−3 and increases as kDS decreases, the PCE of the solar cell will also rise to over 25% with decreasing kDS, until kDS becomes too low (indicated by the red arrow ② in Fig. 1b).

  3. 3.

    When rBK is lower than 10−3 but does not increase as kDS decreases, the PCE of the solar cell will not increase with decreasing kDS (indicated by the red arrow ③ in Fig. 1b).

Anticipating that rBK does not necessarily reduce with a reduction in kDS, it follows that an intentional decrease in kDS within real-world OSCs could lead to an elevation in PCE.

Fine-tuning k DS in OSCs using a dual-acceptor strategy

Practically, when considering binary OSCs, reducing kDS could require a reduction in the energy difference between the CT state and S1A (ΔECT)43. Achieving this requires precise adjustments to the energy levels of the frontier molecular orbitals of either the donor or acceptor material44,45. However, the task of such precise calibration presents challenges. Additionally, kDS depends on reorganization energy and electronic coupling between the S1 and CT states, both of which are even more challenging to adjust. Consequently, there is still a lack of an effective experimental strategy for optimizing the trade-off between IQE and EQEEL, through the regulation of kDS.

In this work, we present a dual-acceptor strategy for the direct manipulation of kDS in OSCs. Differing from commonly employed ternary46,47,48,49,50 or quaternary51 strategies, which primarily focus on altering the microstructures of D/A blend films, our approach centers around modifying the energetics and dynamics of excited states within the blend films. Specifically, our active layer formulation for OSCs involves a common donor (D) combined with two different acceptor materials: a primary acceptor (A1) and a secondary acceptor (A2).

Within the D/A1/A2 ternary blend, the careful selection of A1 ensures that the energy of the CT state formed at the D/A1 interface (CT1) remains lower than the energy of the S1 state of A1 (S1A1). Consequently, the difference between the energy of S1A1 and the CT state at the D/A1 interface (ΔECT1) is anticipated to be substantial. Assuming the electronic coupling and reorganization energy of the blend of D/A1 are not significantly different from those of classic donor/acceptor blends used for organic solar cells, this high ΔECT1 is expected to lead to a high dissociation rate constant of S1A1 (\({k}_{{DS}}^{A1}\)) according to the Marcus theory52,53. Furthermore, our choice of A2 is designed to yield a CT state energy at the D/A2 interface (CT2) that closely aligns with the energy of the S1 state of A2 (S1A2). This configuration results in a low ΔECT at the D/A2 interface (ΔECT2), ideally approaching 0 eV, and consequently, a very low \({k}_{{DS}}^{A2}\). Hence, assuming a homogeneous distribution of the donor and acceptors in the D/A1/A2 ternary active layer, kDS for the ternary blend, expressed as the weighted average of \({k}_{{DS}}^{A1}\) and \({k}_{{DS}}^{A2}\), can be modulated by adjusting the A2 content, similar to what has been reported in the literature for ternary blends with two donors54: A gradual decrease of kDS is possible to achieve by progressively increasing the A2 content.

It should be noted that the influence of increased A2 content on EQEEL and the overall performance of the ternary solar cell is expected to diverge based on the selection of A2. Illustrated through the dynamic processes of excited states in the dual acceptor solar cell (Fig. 2), two distinct scenarios emerge. In both situations, kDS of the solar cell reduces with increased A2 content. In the first scenario, A2 possesses an S1 state energy lower than the ECT of the D/A1 blend. Here, increasing A2 content results in lowered ECT and subsequently elevated kCT of the ternary blend55,56, leading to a constrained increase in EQEEL in accordance with Eq. (5). In the second scenario, A2 has an S1 state energy close to or higher than the ECT of the D/A1 blend. In this case, increasing A2 content has a limited impact on ECT and kCT of the ternary blend, thus resulting in an increased EQEEL.

Fig. 2: Dynamics of excited states in the D/A1/A2 ternary solar cell.
figure 2

A1 and A2 represent the primary and secondary acceptors used in the blend, respectively. CT1 and CT2 are the CT states formed at the D/A1 interfaces and the D/A2 interfaces, respectively. Because of the high ΔECT1, \({k}_{{DS}}^{A1}\) is high, and due to the low ΔECT2, \({k}_{{DS}}^{A2}\) is low. Thus, gradually increasing the A2 content in the D/A1/A2 ternary blend leads to gradually reduced kDS of the ternary blend system. Two different cases are illustrated, depending on the selection of A2, and the impact of the increased A2 content on the electroluminescence external quantum efficiency (EQEEL) and the overall solar cell performance is anticipated to differ in these cases.

Full size image

Impact of reduced k DS on the performance of OSCs

To examine the effectiveness of the dual-acceptor strategy in modulating the kDS of organic blends, solar cells were constructed using the conventional device architecture of ITO/PEDOT:PSS/active layer/PFN-Br/Ag. Initially, we employed IT4F as the primary acceptor, Y5 as the secondary acceptor, and PM7 as the donor57,58,59. The chemical structures of the active materials are provided in Supplementary Fig. 1.

We selected PM7/IT4F as our D/A1 blend because the EL spectrum of the PM7/IT4F binary solar cell was predominantly characterized by CT state emission (Fig. 3). This observation strongly suggests a high ΔECT for the PM7/IT4F blend, which we estimate to be 0.12 eV (Supplementary Fig. 2). Therefore, kDS for the PM7/IT4F blend is expected to be high, which can also be seen from femtosecond transient absorption (fs-TA) measurements.

Fig. 3: Photophysical properties of PM7/IT4F and PM7/Y5.
figure 3

Electroluminescence (EL) spectra of the solar cells based on the binary blends of PM7/IT4F and PM7/Y5, and the ternary blends of PM7/IT4F/Y5 with varying Y5 contents, and the EL spectra of the devices based on pure IT4F and pure Y5. Source data are provided as a Source Data file.

Full size image

Specifically, a weak pump excitation at 750 nm (1.7 μJ cm−2) was selected for the fs-TA measurements to exclusively excite the S1 state of IT4F (UV-vis absorption spectra of PM7 and IT4F are provided in Supplementary Fig. 3). In this case, the only charge transfer process in the PM7/IT4F blend system during the fs-TA measurements involved hole transfer from IT4F to PM7. As a result, the negative transient absorption signals (expressed as ΔA, Supplementary Fig. 4) were detected in the sub-650 nm region and the 700–800 nm region of the fs-TA spectrum, which are attributed to the ground state bleaching (GSB) of the donor PM7 and acceptor IT4F, respectively. Because these spectral response coincide with the steady state absorption spectra of PM7 and IT4F (Supplementary Fig. 3). Consistent with the findings reported in the literature60, the GSB signals of the donor (at 630 nm) persisted for a relative extended duration, thus they can be used to represent the charge generation kinetics, i.e., the transfer of holes from the acceptor to the donor. Therefore, the time evolution of the TA signals at 630 nm is fitted using the sum of two exponential functions and an instrument response function (IRF), giving rise to a rise time constant of ≈8 ps. Importantly, the GSB signal of PM7 is observed quickly after photoexcitation, which implies an ultra-fast hole transfer from the acceptor to the donor, corresponding to a high kDS value, thanks to the high ΔECT for the PM7/IT4F blend.

We opted for Y5 as the secondary acceptor A2 due to the emission solely from the S1 state of Y5 in the EL spectrum of the PM7/Y5 binary solar cell (Fig. 3). This observation indicates a low ΔECT within the PM7/Y5 blend, and consequently, a low anticipated dissociation rate constant. We conducted fs-TA measurements for the blend film of PM7/Y5 and observed negative signals in the 650-850 nm region (Supplementary Fig. 4), assigned to the GSB of Y5 (steady-state absorption spectra of Y5 shown in Supplementary Fig. 3). However, unlike the PM7/IT4F blend, GSB signals from the donor PM7 are hardly observed in the fs-TA spectra for the PM7/Y5 blend.

In fact, the TA spectra for the PM7/Y5 system closely resemble those for the neat Y5 film, as shown in Supplementary Fig. 5. Moreover, a comparison of the GSB signals for Y5 in the neat film and those for the PM7/Y5 blend reveals very similar decay lifetimes (≈140 ps). These experimental TA features indicate that in the PM7/Y5 system, the intrinsic kinetics of Y5 dominate the photophysical process, while the contribution of the PM7 GSB signal is minimal. This could be due to either the absence of the dissociation of acceptor excitons (no transfer of holes from Y5 to PM7), or, alternatively, the holes, after transferring from Y5 to PM7, could quickly return from PM7 to Y5, owing to the extremely lower ΔECT. In both cases, the hole transfer rate is severely limited. Therefore, the kBK of the PM7/Y5 blend, with a lower ΔECT, is higher compared to that of PM7/IT4F, corresponding to a lower kDS for the PM7/Y5 blend. Similar results have been reported for organic donor/acceptor systems, where a reduced ΔECT leads to a reduction in hole transfer rate by over two orders of magnitudes53.

Since the fs-TA results indicated a lower kDS in PM7/Y5 compared to PM7/IT4F, it is anticipated that the kDS will decrease as the Y5 content increases within the ternary blend of PM7/IT4F/Y5. This is confirmed by the electric field dependent steady state PL measurements, as detailed in Supplementary Note 2. Furthermore, given that the S1 state energy of Y5 (1.43 eV, Supplementary Fig. 3) closely approximates the ECT of the PM7/IT4F blend (1.46 eV), the excited state dynamics within the PM7/IT4F/Y5 ternary blend align with the scenario depicted as case 2 in Fig. 2. Hence, the decreased kDS also prompts a shift in the primary decay pathway within ternary solar cells, transitioning from the CT state to the S1 state of Y5 as the Y5 content increases. This transition is indicated by the alteration of the predominant emission peak in the EL spectrum (Fig. 3) from the PM7/IT4F CT state (1.2 eV) to the Y5 S1 state (1.35 eV). Consequently, the device EQEEL is significantly increased, as shown in Fig. 4b: For the ternary solar cell with a Y5 content of 10%, i.e., the ratio between the weight of Y5 and the total weight of the acceptors, EQEEL is increased by over an order of magnitude compared to the PM7/IT4F binary solar cell. Thus, VNR of the ternary solar cell is reduced from 0.35 to 0.27 V (Table 1).

Fig. 4: Performance of PM7/IT4F/Y5 ternary solar cells.
figure 4

a Photovoltaic external quantum efficiency (EQEPV) of the PM7/IT4F/Y5 ternary solar cells with different Y5 contents. b Electroluminescence external quantum efficiency (EQEEL) and open-circuit voltage (VOC) of the PM7/IT4F/Y5 ternary solar cells with different Y5 contents. The EQEEL values are determined using an injection current density equal to the short-circuit current density (JSC) of the device. c Transient photovoltage (TPV) decay lifetime of the PM7/IT4F/Y5 ternary solar cells with different Y5 contents (see Supplementary Note 3 for details). d Power conversion efficiency (PCE) of the PM7/IT4F/Y5 ternary solar cells with different Y5 contents. The error bars in b and d represent the highest, lowest, and average values from multiple devices. Source data are provided as a Source Data file.

Full size image
Table 1 Photovoltaic performance of the PM7/IT4F/Y5 solar cells
Full size table

Transient photovoltage (TPV) measurements were also performed on the PM7/IT4F/Y5 ternary solar cells to evaluate the impact of increasing Y5 content on the lifetimes of charge carriers. The results are presented in Fig. 4c. Clearly, as the Y5 content increases, the photovoltage lifetime increases, indicating an extended lifespan of charge carriers. As a result, the increased EQEEL and reduced VNR contribute to an overall reduction in voltage loss and an increase in VOC, as demonstrated in Fig. 4b: The VOC of the solar cell increases from 0.82 V to 0.87 V as the Y5 content rises from 0% to 10%. It is important to note that the substantial increase in VOC underscores the distinct nature of the dual-acceptor strategy proposed herein for manipulating kDS, diverging from conventional ternary strategies. Since without accounting for the influence of kDS, the addition of a third component with an S1 state energy lower than the primary D/A system would conventionally lead to a reduction in VOC for ternary OSCs61,62,63,64,65,66,67.

It should be noted that the addition of Y5 to the binary blend of PM7/IT4F could lead to a change in the morphology of the active layer, potentially resulting in a reduction in the reorganization energy of the active layer. This, in turn, could contribute additionally to the increase in EQEEL23,24,. Therefore, we measured the EL and sensitive EQEPV spectra of the solar cell based on PM7/IT4F, PM7/IT4F/Y5, and PM7/Y5 (Supplementary Fig. 6) and determined the reorganization energy through Gaussian fitting to the tail of the EQEPV spectra. We found that the reorganization energy is approximately 0.32 eV for the PM7/IT4F binary blend, and it slightly decreases to 0.28 eV in the ternary blend with a Y5 content of 20%. This suggests that the introduction of Y5, while it may cause changes in the morphology of the active layer, does not significantly decrease the reorganization energy. Therefore, the increased EQEEL should not be ascribed to a morphological reason.

Next, we explore the implications of decreasing kDS on IQE and EQEPV of the PM7/IT4F/Y5 ternary solar cell. As illustrated in Fig. 4a, our initial observation highlights a significant difference between the EQEPV of the binary PM7/IT4F (over 80%) and PM7/Y5 solar cells (below 10%). The markedly low EQEPV for the PM7/Y5 solar cell is a foreseeable outcome, attributed to the lower ΔECT and kDS within the PM7/Y5 blend. This leads to a restricted IQE of the binary solar cell, a hypothesis that gains support from the steady state PL quenching measurements, as further elucidated in Supplementary Fig. 7. Nevertheless, the EQEPV of the PM7/IT4F solar cell remains unaffected by the introduction of a small quantity of Y5: The peak EQEPV of the ternary solar cell with a Y5 content of 10% is as high as that of the PM7/IT4F binary solar cell (≈80%). It is only when the Y5 content surpasses 10% that a distinct reduction in the peak EQEPV becomes evident. This pattern suggests that the kDS within the PM7/IT4F binary blend substantially surpasses the threshold required to achieve an IQE close to 100%. Consequently, a controlled reduction in kDS for high quantum efficiency OSCs does not necessarily lead to a direct decrease in IQE, as predicted by Eq. (6).

Since the FF of the ternary solar cell is hardly affected by the addition of a small amount of the secondary acceptor (Table 1), PCE of the solar cell could indeed be increased using the dual-acceptor strategy: The PCE of the solar cell is increased from 11.8% to 12.6%, when the Y5 content is increased from 0% to 10% (Fig. 4d). The basic photovoltaic performance parameters of the solar cells with different Y5 contents are summarized in Table 1, and the current density-voltage (J-V) curves are provided in Supplementary Fig. 8.

We have also employed the dual-acceptor strategy for solar cells based on the blend of PM7/IT4F as the primary D/A1 system, and Y5OD as A2, as well as solar cells based on PM6/fullerene as the primary D/A system, and an non-fullerene acceptor SM16 as A2. In all cases, we realized increased EQEEL, reduced overall voltage loss, and improved solar cell performance by a controlled decrease in kDS, achieved by gradually increasing the A2 content. More details are provided in Supplementary Note 4 and 5. The outcomes derived from the study of these ternary solar cells all demonstrate the potential for enhancing the performance of OSCs through the reduction of kDS. Nonetheless, it is important to underline that these material combinations were selected as model systems, because: 1) The primary D/A1 system possesses high kDS, giving rise to high IQE, while the secondary D/A2 system has low kDS. 2) The introduction of A2 content is not anticipated to significantly influence \({k}_{{S}_{1}}\) and \({\eta }_{{S}_{1}}\) within the primary D/A1 blend, because of the comparable S1 state lifetime values between A1 and A2 (the lifetime values determined from fs-TA are 40 and 140 ps for IT4F and Y5, respectively, see Supplementary Fig. 5). 3) The emission efficiencies of the S1 states of A1 and A2 also appear similar (the EQEEL of the solar cells based on pure IT4F and pure Y5 are 0.2% and 0.3%, respectively, see Supplementary Fig. 9).

The well-understood electronic properties of these material combinations allow us to have confidence in the effectiveness of the dual-acceptor strategy in modulating kDS, as well as in the deduction concerning the pivotal role of kDS in shaping the trade-off between EQEEL and IQE, and improving overall solar cell performance. However, the decrease in IQE is significant, while the increase in EQEEL is less pronounced for these solar cells when the A2 content exceeds 10%. This phenomenon could be attributed to the substantial molecular structural differences between the primary acceptor A1 and the secondary acceptor A2, leading to different stacking characteristics of the donor or the acceptor molecules in the ternary active layer with increasing A2 content (see Supplementary Fig. 10). As a result, the dual-acceptor strategy results in a limited enhancement in PCE for solar cells based on the PM7/IT4F/Y5, PM7/IT4F/Y5OD, or PM6/fullerene/SM16. This challenge will be addressed in the subsequent section of this article.

Improving the performance of high-efficiency OSCs

To further explore the advantages of the dual-acceptor strategy in enhancing the performance of high-efficiency OSCs and to establish the general applicability of the strategy, we employed D1868 and EC969 as the primary D/A1 system. For A2, we utilized SM1670, which bears a molecular structure similar to EC9. The chemical structures of these active materials are displayed in Supplementary Fig. 1.

Before constructing the ternary devices, we first studied the binary solar cells based on D18/EC9 and D18/SM16. In the D18/EC9 binary solar cell, the EL spectrum predominantly features an emission peak corresponding to the S1 state of EC9 (Fig. 5a), implying a low ΔECT for the solar cell and a reduced kDS in comparison to the PM7/IT4F binary solar cell. However, the kDS within the D18/EC9 binary solar cell remains sufficiently high to facilitate efficient exciton dissociation, which results in a high IQE and a peak EQEPV of over 80% (Fig. 5b). The EQEEL of the solar cell is constrained to around 0.02% (Fig. 5c), consequently leading to a high VNR, exceeding 0.2 V (Table 2).

Fig. 5: Performance of the solar cells based on D18/EC9/SM16.
figure 5

a Electroluminescence (EL) spectra of the solar cells based on the binary blends of D18/EC9 and D18/SM16, and the EL spectra of the devices based on pure EC9 and pure SM16, b Photovoltaic external quantum efficiency (EQEPV) of the solar cells based on D18/EC9/SM16 with different SM16 contents. c Electroluminescence external quantum efficiency (EQEEL) and open-circuit voltage (VOC) of the D18/EC9/SM16 ternary solar cells with different SM16 contents. d Power conversion efficiency (PCE) of the D18/EC9/SM16 ternary solar cells with different SM16 contents. The error bars in c and d represent the highest, lowest, and average values from multiple devices. Source data are provided as a Source Data file.

Full size image
Table 2 Photovoltaic performance of the D18/EC9/SM16 solar cells
Full size table

On the other hand, the D18/SM16 binary solar cell exhibits a predominant emission peak in the EL spectrum also stemming from the S1 state of the acceptor (Fig. 5a). This signifies a low ΔECT. However, the kDS within the D18/SM16 binary solar cell is expected to be excessively low, significantly impeding the efficiency of exciton dissociation, as validated by PL quenching measurements (Supplementary Fig. 11). Consequently, the IQE remains low, and the peak EQEPV is confined to less than 5% (Fig. 5b). Owing to the markedly low kDS, the EQEEL of the D18/SM16 binary solar cell is high, approximately 0.7% (Fig. 5c). This is paralleled by a low VNR of merely 0.12 V (Table 2).

It should also be noted that the S1 state energy of SM16 is comparable to that of EC9 (Supplementary Fig. 12). Moreover, the emission efficiencies of the S1 states for SM16 and EC9 show no significant disparities (0.9% vs. 0.4%, Supplementary Fig. 13). Therefore, in the D18/EC9/SM16 ternary blend, the values of \({k}_{{CT}}\), \({k}_{{S}_{1}}\), and \({\eta }_{{S}_{1}}\) are expected to remain relatively consistent despite variations in the SM16 content.

Then we constructed ternary solar cells based on D18/EC9/SM16, and anticipated that kDS decreases with increasing SM16 content, which then gives rise to an increase in EQEEL. This is indeed the case as shown in Fig. 5c: EQEEL of the ternary solar cell increases from 0.02% to 0.04%, with the increase of SM16 content from 0% to 10%. Since the lifetime of charge carriers also increases with SM16 content, as confirmed by TPV measurements (Supplementary Fig. 14), the overall voltage loss is reduced and the VOC is increased for the ternary solar cells (Fig. 5c). It is noted that the degree of reduction in overall voltage loss exceeds that anticipated from the increase in EQEEL, suggesting that there could be additional reasons for the increase in VOC with the addition of SM16. Nevertheless, neither the peak EQEPV (Fig. 5b) nor the FF (Table 2) of the ternary solar cell is significantly reduced by the increased SM16 content, even with a high SM16 content of 20%. This is attributed to the similar molecular structures of SM16 and EC9, leading to similar morphological properties of the active layers, regardless of the SM16 content (Supplementary Fig. 15). As a result, the PCE of the solar cell significantly increases with the addition of SM16: The PCE increases from 17.4% to 19.2% as the SM16 content increases from 0% to 10% (Fig. 5d). The basic photovoltaic performance parameters of the solar cells with different SM16 contents are summarized in Table 2, and the J-V curves are provided in Supplementary Fig. 16.

Finally, referring to Eq. (5), we can observe that the dual-acceptor strategy possesses the potential to notably enhance EQEEL, particularly when incorporating a secondary acceptor (A2) with an S1 state energy close to the ECT of the D/A2 blend but substantially higher than the S1 state energy of the other acceptor (A1). This expectation arises because using A2 in this manner elevates ECT and reduces both kCT and kDS of the ternary blend, thereby leading to a significant EQEEL increase. However, as elaborated in Supplementary Note 6, utilizing A2 with a higher S1 state energy does not positively impact the performance of D/A1/A2 solar cells due to excited state energy transfer between A2 and A1. Hence, we formulate the subsequent selection rule for A2 to modulate kDS and enhance solar cell performance: The S1 state energy of A2 should be lower than that of A1, as depicted in case 2 of Fig. 2.

Discussion

The above results affirm the general effectiveness of the dual-acceptor strategy in reducing kDS within OSCs. Furthermore, the reduced kDS can indeed result in an increase in EQEEL, a decrease in VNR, and an improvement of PCE for high-efficiency NFA OSCs, provided that the electronic properties of the active materials align with the characteristics depicted in case 2 of Fig. 2. It is crucial to emphasize that the preference for the dual-acceptor approach over dual-donor stems from the significantly lower S1 state energy of the acceptor in state-of-the-art OSCs compared to that of the donor. Consequently, the excited state concentration of the donor is negligible. For OSCs where the S1 state energy of the acceptor surpasses that of the donor, as seen in fullerene solar cells, a similar dual-donor strategy is projected to effectively decrease the kDS of the blend system, enhance EQEEL, and improve overall solar cell performance. Furthermore, Eq. (5) prompts us to acknowledge that the enhancement of EQEEL achieved by reducing kDS is highly reliant on the emission efficiency of the S1 state. Currently, the emission quantum efficiency of the most emissive acceptors in thin films for OSCs remains below 1%. Hence, in order to fully harness the benefits of the dual-acceptor strategy, there is a pressing need to develop acceptor materials with higher emission efficiency.

To conclude, this study undertook a reevaluation of the dynamic processes governing excited states in OSCs, incorporating the repopulation and redissociation of the S1 state of the acceptor. The findings reveal that the dissociation rate constant of the S1 state, denoted as kDS, traditionally associated with the generation efficiency of free charge carriers, plays a pivotal role in determining the non-radiative voltage loss within OSCs. Specifically, our results illustrate that instead of pursuing maximal kDS values for achieving high-efficiency OSCs, a deliberate reduction in kDS can lead to reduced non-radiative voltage loss and an enhanced overall device performance. In this regard, we subsequently introduced the dual-acceptor strategy, which centers on modifying the dynamic properties of excited states to effectuate a precisely controlled reduction in kDS within real-world OSCs. We successfully demonstrated that, for high-efficiency binary OSCs, incorporating a secondary acceptor with an S1 state energy closely aligned to the ECT of the blend formed by the donor and the secondary acceptor enables a decrease in kDS. This subsequently results in reduced non-radiative voltage loss and an improved solar cell performance. Consequently, the theoretical insights and experimental approach outlined in this study carry significant implications for overcoming the voltage bottleneck in OSCs, and for generating innovative concepts pertaining to the synthesis of high-efficiency photovoltaic materials, as well as the engineering of high-performance organic donor-acceptor blends.

Methods

Materials and device fabrication

The active materials, PM6, PM7, D18, IT4F, Y5, Y5OD, BTA3, and EC9 were purchased from Solarmer materials Inc (China). PC71BM was purchased from Solenne BV. SM16 were synthesized according to the literature (Adv. Funct. Mater. 32, 2107756 (2022)). PFN-Br used in this work was purchased from Sigma Aldrich. PEDOT:PSS 4083 was purchased from Heraeus. 1,8-diiodooctane (DIO), 1-chloronaphthalene (CN), chlorobenzene (CB), and chloroform (CF) were purchased from Sigma-Aldrich.

Organic solar cells with a standard architecture of ITO/PEDOT:PSS 4083/active layer/PFN-Br/Ag were fabricated in this work. The ITO substrate was cleaned with acetone, isopropanol, and ethanol solvent in sequence within an ultrasonic cleaner for a duration of 15 min. Subsequently, the ITO surface underwent a 10-minute plasma treatment. Then, PEDOT:PSS 4083 (purchased from Heraeus) solution was spin-coated (WS-650Mz-23NPPB, Laurell) on the ITO substrate at a speed of 4000 rpm, followed by annealing on a hot plate (IKA RCT digital) at 150 °C for 20 min. The thickness of PEDOT:PSS 4083 layer was approximately 30 nm. The PM7/IT4F/Y5 and PM7/IT4F/Y5OD active solutions were prepared by dissolving the materials at 20 mg mL−1 in chlorobenzene (CB) (purchased from Sigma-Aldrich) with different D/A1/A2 weight ratios. In addition, 0.5% DIO was added in the solution as the solvent additive. Subsequently, each solution was heated to 80 °C and stirred at a rate of 1000 rpm utilizing a magnetic stirrer hotplate for 8 h. The solution was allowed to stand for half an hour and then cooled to room temperature prior to use. Then, the solutions for the active layer were spin-coated (2200 rpm) on top of the PEDOT:PSS 4083 coated substrate, in a nitrogen-filled glove box (Mbraun), and the resulting active layers were annealed at 110 °C for 10 min on a hotplate. The thickness of the active layers was about 100 nm. The D18/SM16/EC9 solutions were prepared by dissolving the active materials at a concentration of 16 mg mL−1 in chloroform (CF) (from Sigma-Aldrich) with different D/A1/A2 weight ratios. Additionally, 0.5% CN (from Sigma-Aldrich) was added in the solution as an additive. All solutions were heated at 60 °C and stirred at 800 rpm on a magnetic stirrer hotplate for 4 h. The active layer solutions were spin-coated on the PEDOT:PSS 4083 coated substrates at 3000 rpm in a glove box, then the active layers were annealed at 90 °C for 5 min on a hotplate. The thickness of the active layers was about 90 nm. The PM7/BTA3/IT4F solution was prepared using CB as the solvent and the solution concentration was 20 mg mL−1, supplemented with 0.5% DIO as an additive. Prior to use, the solutions underwent heating at 80 °C and they were stirred at 1000 rpm on a magnetic stirrer hotplate for a duration of 8 h. The active layers were spin-coated onto the PEDOT:PSS 4083 coated substrates at 2500 rpm, and they were annealed at 110 °C for 10 min on a hotplate. The thickness of the active layers was about 100 nm. Following the deposition of the active layers, the PFN-Br solution (0.5 mg mL−1 in methanol, purchased from Sigma-Aldrich) was spin-coated onto the active layers at a speed of 3000 rpm. Then, a 110 nm thick Ag electrode was evaporated onto the active layers in a thermal evaporator (Mbraun) under a vacuum pressure of 3 × 10−6 mbar. The active area of the devices determined by an optical microscope was 0.04 cm2. Finally, the solar cells were encapsulated by employing glass slides in conjunction with UV-curing adhesive.

Device characterization

J-V characterization

Photovoltaic performance parameters of the organic solar cells studied in this work were characterized under simulated AM1.5 solar illumination (100 mW cm−2) from a Newport Oriel VeraSol-2™ Class AAA solar simulator. Prior to testing, the illumination intensity was calibrated by a standard silicon solar cell and a set of low pass filters. The J-V curves were measured by a Keithley 2400 source meter. The JSC values from J-V characterizations and those derived from the EQEPV spectra exhibited a marginal discrepancy (<5%).

Highly sensitive EQE PV measurements

The EQEPV spectra were measured using a halogen lamp (LSH-75, 250 W, Newport), a monochromator (CS260-RG-3-MC-A, Newport), equipped with a series of long-pass filters (600 nm, 900 nm, 1000 nm), a front-end current amplifier (SR570, Stanford), a phase-locked amplifier (SR830, Stanford Instrument), and an optical chopper (173 Hz). Calibration of the light intensity was performed using a standard Si detector (Hamamatsu s1337-1010BR). A background illumination of 100 mW cm−2 was applied during the measurements. The illuminated area was approximately 0.4 mm2, achieved through the use of a focus lens and an optical aperture.

EQE EL measurements

EQEEL measurements were conducted through a custom-built setup, which consists of a Keithley 2400 to apply electric current to the device in a dark environment. Photons emitted from the device was recorded using a Si detector (Hamamatsu s1337-1010BR) in conjunction with a Keithley 6482 Picoammeter.

Photoluminescence measurements

A Super-continuous white laser (SuperK EXU-6, NKT photonics) coupled to a narrowband filter (LLTF Contrast SR-VIS-HP8, LLTF Contrast SR-SWIR-HP8, NKT photonics) was used to excite organic thin films, and emission from the films was collected by an optical fiber (BFL200LS02, Thorlab), and directed to a fluorescence spectrometer (KYMERA-328I-B2, Andor Technology). Fluorescence signals were captured by either a Si EMCCD camera (DU970P-BVF, Andor Technology) covering the range of 400–1000 nm or an InGaAs EMCCD camera (DU491A-1.7, Andor Technology) covering the range of 1000–1500 nm. Optical losses in the fibers and the spectrometer were corrected using a standard light source (HL-3P-CAL, Ocean Optics Germany GmbH).

Electroluminescence measurements

Electroluminescence spectra were captured using a fluorescence spectrometer (KYMERA-328I-B2, Andor Technology LTD), a Si EMCCD camera (DU970P-BVF, Andor Technology), and an InGaAs EMCCD camera (DU491A-1.7, Andor Technology). Injection current was supplied to the sample device through a direct-current meter (PWS2326, Tektronix) in dark.

Transient absorption spectroscopy

The fs-TA spectra were measured by using a Helios-pump-probe setup (Ultrafast Systems), with an amplified 1030 nm fs laser (Pharos, Light Conversion, 200 fs, 200 µJ) with an effective repetition rate of 1 kHz. The probe white light was generated via non-linear phenomenon with a photonic crystal, yielding broadband probe light across the visible to NIR region from 400 to 950 nm. The pump pulse was generated with an optical parametric amplifier (Orpheus-F, Light Conversion), while the excitation intensity was tuned by density filters. The measurements were done using an excitation fluence of 1.7 μJ cm−2, Samples were spin-coated on glass substrate and measured at the ambient condition without room light.

Absorption measurements

The absorption spectra of organic thin films were recorded using a UV-Vis spectrometer (Lambda 950, PerkinElmer).

Thicknesses measurements

The thicknesses of the organic thin films were measured utilizing a surface profilometer (KLA-Tencor P-7 Stylus Profileror).

Atomic force microscope

AFM images of the active layers were recorded by MFP-3D-BIO™, Asylum Research, in a tapping-Mode.

Transmission electron microscope

TEM images were taken using a JEM 2100 transmission electron microscope with an accelerating voltage of 200 kV.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.