Incorporating semiconducting single-walled carbon nanotubes as efficient charge extractors in organic solar cells

Organic solar cells (OSCs) have emerged as a potentially inexpensive alternative to crystalline solar cells, especially for harvesting solar energy in consumer electronics. However, their efficiency and stability needs to be improved in order for them to become more widely adopted. The inherent properties of organic semiconductors, such as the formation of tightly bound exciton immediately after photo generation, which require dissociation prior to charge separation, low exciton diffusion lengths, and low charge carrier mobilities, are some of the factors limiting the performance of OSCs.¹ Significant progress has been made in some of these areas; for example, the rational design of donor-acceptor (D-A) polymers has enabled improved light harvesting,^2,3 efficient exciton dissociation,⁴ and improved the charge carrier mobilities.⁵ In contrast, it has been shown that in these efficient systems, the non-geminate recombination of electrons and holes have been the dominant loss mechanism.^6–8 Therefore, it is important to improve the extraction of photogenerated charges out of the device to suppress the non-geminate recombination and improve the device performance.

Past efforts to improve charge extraction have been focused mainly towards the optimization of the morphology of the bulk-heterojunction (BHJ), by employing techniques such as using solvent additives,⁴ thermal annealing,⁹ incorporation of block-copolymers,¹⁰ and nano-imprinting.¹¹ Most of these methods lead to the formation of purer domains of donors and acceptors, which ultimately results in improved charge generation and, consequently, power conversion efficiency (PCE). However, they do not specifically address the problem of poor charge mobility of the active-layer.

A direct method to minimize the transit time of photogenerated electron and holes from charge-separating interface to the charge-collecting electrode is to use a network of interpenetrating high mobility carbon nanotubes (CNTs) inside the photoactive layer. Multi walled carbon nanotubes (MWNTs)^12,13 and single-walled carbon nanotubes (SWNTs)^14–20 have been investigated as active layer additives in monolayer devices as acceptors^14–16 and in charge extraction limited bulk heterojunction devices.^{12,13,17–20} Some of these studies have used the pristine nanotubes,^17–19 whereas others have functionalized the nanotubes with various functional groups.^13,20,21 Most of the previous investigations have focused on randomly dispersing the CNTs within the active layer, whilst some studies have attempted building the bulk heterojunction using nanotube network as a skeleton,²² and achieving an orderly array of nanotubes that had connectivity to one of the end electrodes.²³ These studies have demonstrated success in varying degrees, but, in general, the efficiency enhancements achieved by using CNTs with^13,20 or without^18,19,22 functionalization were limited. However, remarkable progress has been made by doping the nanotubes to be either n-type or p-type designating them to behave as either electron extractors or hole extractors.^21,24

Here, we demonstrate a method that can enhance the device efficiency of OSCs significantly, with extremely low quantities of pristine SWNTs as an active layer additive on the thieno[3,4-b]thiophene/benzodithiophene (PTB7) and [6,6]-Phenyl-C71-butyric acid methyl ester (PC₇₀BM) polymer system. We hereby address two major drawbacks identified in previous work. The presence of the metallic nanotubes¹⁸ and their probable detrimental effects are eliminated by use of semiconducting (S)-SWNTs. The non-optimal dispersion of nanotubes in the active layer¹⁷ is handled by using N-Methyl-2-pyrrolidone (NMP) as the SWNT dispersant.

In most of the previous work, the same solvent that was used to make the active layer solution (chloroform,¹⁴ chlorobenzene,¹⁷ dichlorobenzene,¹³ etc., or a blend of these solvents) was used to prepare the CNT dispersion as well. However, it is difficult to disperse pristine CNTs at the individual nanotube level in the concentrations (0.5 mg/ml or above) required for use as an active layer additive. On the other hand, NMP is one of the best known dispersants of pristine CNTs. However, the active layer polymers do not dissolve well in NMP, and our preliminary efforts concluded that fabrication of NMP only device is not feasible. Next, we investigated the possibility of using NMP as a processing additive. The use of a processing additive with different evaporation rates and solubilities than the principal solvent to enhance device performance has been demonstrated for several polymer systems,^25–27 and is commonly deployed with the PTB7/PC₇₀BM system.^2,28,29 Several processing additives such as 1,8-diiodooctane,² 1-chloronaphthalene,²⁵ and 1,8-octanethiol have been identified to enhance the device efficiency. NMP as a processing additive for the PTB7/PC₇₀BM system has not been investigated. We investigated the possibility of using NMP as a processing additive to the PTB7/PC₇₀BM/1,2-dichlorobenzene system (Table I, NMP devices). NMP addition did not result in improved characteristics, in fact, devices exhibited slightly degraded performance. However, it was possible to incorporate it at up to 6% of the total solvent volume without significant performance degradation. This provides a route for introducing well dispersed SWNTs into the active layer solution.

The placement of nanotubes within the active layer is random, keeping with the solution processability of the active layer. The process of CNT dispersion and incorporation to the active layer are simple and readily scalable. Both unsorted SWNT samples containing both metallic and semiconducting nanotubes as well as sorted SWNT samples containing only semiconducting nanotubes are used in this study to investigate, (i) whether it is possible to enhance the PCE of BHJ OSCs using pristine SWNT for the particular polymer blend employed and (ii) whether S-SWNTs are better at PCE enhancement compared to unsorted (U)-SWNT, and if so to what extent.

The devices discussed here consist of an ITO coated glass substrate (transparent anode, sheet resistivity 20 Ω/sq), PEDOT:PSS hole collecting layer (40 nm), active layer (100-120 nm), and an aluminum back electrode (cathode, 100 nm). The schematic diagram of the device structure is given in Fig. 1. Active layer polymers PTB7 (Organic Nano Electronic Material inc,) and PC₇₀BM (SolenneBV) were blended in a 2:3 ratio (by weight), dissolved in 1,2-dichlorobenzene to make a solution with total polymer concentration of 35 mg/ml. Four types of devices were investigated for their performance: (i) Standard PTB7/PC₇₀BM device (“control”), (ii) PTB7/PC₇₀BM devices, NMP as the solvent additive (“NMP”), (iii) PTB7/PC₇₀BM devices, NMP + U-SWNT as additives (“U-SWNT”), and (iv) PTB7/PC₇₀BM devices, NMP + S-SWNTs as additives (“S-SWNT”) SWNT (purified 95% semiconducting and unsorted, mean diameter 1.4 nm from Nanointegris) suspensions were prepared by adding 0.2 mg/ml of nanotubes to NMP and sonicating for 8 h in a bath sonicator. According to the device type, a pre-determined volume of additive was added to the active layer solution prior to spin coating the active layer.

The cells were measured directly after the Al back electrode deposition, without further thermal annealing. The measurements were made in air at room temperature, and the devices were not encapsulated. I-V measurements were done with a computer controlled HP 4140B source meter and an Oriel solar simulator (Model 96000) was used with an AM1.5 G filter. All the measurements were carried out at the standard intensity level of 100 mW/cm², verified by a calibrated solar cell prior to measurements.

Table I summarizes the open circuit voltage (V_oc), short circuit current density (J_sc), fill factor (FF), and PCE for the devices fabricated in this work. All the results presented here are averaged over four devices of each type and the device active area is 0.08 cm².

In most of the previous studies, it was not possible to achieve an enhancement in the efficiency using pristine CNTs. Even when it was possible to achieve an enhancement, the percentage wise improvement was small (about 10%) and this was achieved with 0.5%–1% of CNT by total polymer weight.^17,19 Here, we present a similar efficiency improvement using pristine U-SWNT with efficiency enhancement of 11% with respect to the control device but with only 0.02% of SWNT (3% NMP level) in the active layer, which is two orders of magnitude less than the previously reported values (Fig. 2(a)). This is perhaps an indicator of the much better dispersion of SWNTs in the NMP, and consequently, in the photoactive layer compared to previous work. When using S-SWNT (0.02%), the efficiency enhancement is 29% with respect to the control device resulting in average device efficiency of 6.32% (peak efficiency of 6.63%). The efficiency enhancement due to SWNTs is even more prominent when the SWNT devices are compared to “NMP only” device, where the only difference between the devices is the presence and absence of the nanotubes. With respect to the “NMP only” device, the PCE enhancement in U-SWNT device is 17% and that of S-SWNT device is 36%. This also clearly indicates that S-SWNTs are much more effective in efficiency enhancement compared to U-SWNT, furthering PCE enhancement by additional 18%. When the S-SWNT concentration is increased to 0.04%, the PCE is similar to that achieved with U-SWNT at 0.04%. Though the S-SWNT is sorted, they are not 100% semiconducting with the supplier specification being 95% semiconducting by weight. Therefore, it is more probable for there to be a few residual metallic SWNT incorporated as the S-SWNT concentration increases and for the PCE to reduce.

It is evident that there is an improvement of FF in both U-SWNT and S-SWNT devices compared to “Control” and “NMP” devices, which do not have SWNTs in the active layer. This FF enhancement amounts to an improvement of 13% over the “Control” device and 11% over the “NMP” devices. Addition of any SWNT can aid the crystallinity of the active layer polymers,³⁰ consequently, improving the mobility of individual polymer phases in the BHJ and the 11% enhancement over “NMP” devices can be attributed to the mobility enhancement of the active layer due to the presence of SWNTs. It is noteworthy that FF enhancement is not sensitive to the type of SWNT present in the active layer.

However, the difference between U-SWNTs and S-SWNTs is very much apparent when considering the short circuit current density (J_sc). The J_sc enhancement in U-SWNT device is small compared to the control device and negligible compared to “NMP” device. Thus, it appears as if the small J_sc enhancement present in “U-SWNT” devices is rather due to the processing additive, rather than due to the U-SWNTs. The J_sc in S-SWNT device is significantly higher than that of control and NMP devices with 19% improvement over control device J_sc and 17% improvement over the “NMP” device. The 17% J_sc enhancement over the “NMP” devices clearly results from the presence of S-SWNTs in the active layer, and indicates more efficient charge extraction. Improved charge extraction can result from the active participation of high mobility S-SWNTs as charge extractors or from the increased mobility of the active layer itself. However, if only U-SWNTs are used then any enhancement in mobility due to enhanced crystallinity could be negated by the presence of metallic (M)-SWNTs acting as recombination centres. When only S-SWNTs are used, such recombination would be reduced making the enhanced extraction due to crystallization apparent. The use of S-SWNTs, therefore, could be enhancing the charge extraction process through the combination of the two mechanisms discussed. The improvement over the J_sc of the U-SWNT devices could be due to more S-SWNTs present in the active layer or from the suppression of detrimental effects of M-SWNTs. The optical properties of the active layer films were characterized by UV-VIS spectroscopy and no significant difference in absorption was observed after addition of CNTs.

At 6% NMP level (0.04% SWNT), the difference between NMP, U-SWNT, and S-SWNT devices diminished altogether, and efficiency falls below the control device. There is evidence from the previous work that optimal amount of processing additive is around 2%–3%.^26,27

The open circuit voltage (V_oc) of the control device is 0.75 V and this agrees well with the published results for PTB7/PC₇₀BM systems.^2,3,28 However, the addition of NMP lowered the V_oc to 0.69 V in the presence of NMP only. Therefore, as a processing additive alone, NMP is not suited for the particular system, but given its excellent SWNT dispersant properties, it is suitable to obtain an overall improved device. The presence of either U-SWNT or S-SWNT partially recovers this loss in V_oc, up to 0.72 V at 3% NMP level, but remains always below the control device. However, the enhancements in J_sc and FF are significant enough to overcome this loss in V_oc and bring the overall PCE of the SWNT added devices significantly above that of control device.

Dark J-V characteristics (Fig. 2(b)) reveal a decrease in dark current at negative bias upon the addition of NMP. The presence of NMP may change the morphology of the PTB7:PC₇₀BM bulk-heterojunction. A reduction in dark current due to the thermal coarsening of the bulk-heterojunction has been previously observed.³¹ Coarsening of donor and acceptor morphology has also been observed on addition of 1,8-diiodooctane in PTB7:PC₇₀BM and small molecule solar cells.³² Therefore, we expect similar coarsening of the donor and acceptor domains in our PTB7:PC70BM active-layer when NMP was added as co-solvent to disperse SWNT. However, in this case, the concentration of co-solvent NMP was decided based on the concentration of SWNT, and hence, a further optimization of co-solvent concentration was not done. The dark current increases after addition of the SWNT with NMP. The increase in dark current is similar to the reported mechanism of adding silver nanowires in organic solar cell,³³ where the SWNT would provide conduction pathways across the bulk-heterojunction for transport of injected charges and extraction of photo-generated charges.

To investigate the cause(s) of FF enhancement, the lumped parameter equivalent values for series and parallel device resistances were evaluated. The equivalent lumped parameter values of the devices were calculated considering the single diode model of a solar cell. The schematic of the device model is given in Fig. 3(a). Accordingly a device under illumination is described by

where J₀ is the saturation current density, J_L is the photocurrent density, R_s is the series resistance, R_sh is the shunt resistance, n is the ideality factor, q is the electron charge, k_B is the Boltzmann constant, and T is the temperature. The experimental J-V characteristics were fitted to Eq. (1) after using Lambert W function to separate the variables in Eq. (1) and then using non-linear curve fitting functions available with MATLAB™ (the method given in Ref. 34) and the parameters R_s, R_sh, J₀, n, and J_L were solved numerically. To extract the dark parameters, dark characteristics were fitted to Eq. (1) with J_L = 0 as there is no photocurrent.

Fig. 3(b) summarizes the lumped parameters extracted for the devices discussed above from the single diode model, which have been used to describe organic solar cell in the past.^34,35 Simulated parameters indicate that the addition of NMP increases the series resistance of the device, and the effect is more prominent with increasing NMP concentration. Altered drying dynamics of the active layer after adding NMP should change the morphology of the active layer, which could increase the series resistance. The decrease in dark current with the addition on NMP (Fig. 2(b)) also indicate similar conclusion. We note that the total series resistance of the device is lumped together in the model, which consists of resistance of the active layer, resistance of the charge extraction layer, the interface resistance between the layers, and finally, the resistance of the electrodes. Though, in the model, it is not possible to isolate the active layer resistance from other factors contributing to the total R_s, however, it is unlikely that the resistance of the electrodes and the resistance of charge extraction layer will change due to the addition of NMP. When we compare NMP only devices, and devices with NMP and SWNT, in most of the instances the values of R_s have reduced with the addition of SWNT indicating mobility enhancement due to the nanotubes.

Considering the shunt resistance (R_sh) following observations can be made. (i) There is a significant difference in R_sh values extracted from dark IV curve and light IV curve. This is well known not only for organic solar cells but also for many other types of solar cells, and typically, R_sh,D is much larger than R_sh,L.³⁴ (ii) Addition of SWNT has reduced both dark and light shunt resistances (R_sh,D & R_sh,L). This can be identified as a negative effect due to the addition of the nanotubes. Reduction of R_sh indicates current leakage paths, and this was present for both U-SWNTs and S-SWNTs. It indicates that nanotubes are likely to have contact with back electrode resulting in charge recombination inside the device; however, this could be suppressed by inserting a charge extraction layer between the active-layer and the cathode.

In conclusion, we report the use of purified pristine SWNTs to enhance the efficiency of PTB7/PC₇₀BM OSCs. Both U-SWNT and S-SWNT enhances the efficiency, nonetheless, the use of only S-SWNTs is far more effective in enhancing the efficiency than U-SWNT. NMP has two important functions; it acts as SWNT dispersant and also as a solvent additive modifying the morphology of bulk heterojunction. While not directly leading to efficiency enhancements when used by itself, NMP plays an important role in facilitating the dispersion of a very small quantity of CNTs (about two orders of magnitude less than previously reported values) which resulted insignificant efficiency enhancements. Given the high cost of SWNT, the reduction in the concentration of SWNT has important cost implication in using them in OCS. Therefore, the results reported showing enhanced efficiency and the reduction in the quantity of SWNT to achieve that, when combined, provides a promising path for low-cost higher-efficiency organic solar cells.

This work was supported by National Research Council, Sri Lanka (Grant No. NRC 12-082) and the International Research Centre, University of Peradeniya, Sri Lanka.