Applied Physics A

, Volume 108, Issue 3, pp 515–520

Correlation of charge extraction properties and short circuit current in various organic binary and ternary blend photovoltaic devices

Authors

    • Materials Science and EngineeringCommonwealth Scientific and Industrial Research Organisation (CSIRO)
  • Xiwen Chen
    • Materials Science and EngineeringCommonwealth Scientific and Industrial Research Organisation (CSIRO)
  • Wallace W. H. Wong
    • School of Chemistry, Bio21 InstituteUniversity of Melbourne
  • Tino Ehlig
    • Materials Science and EngineeringCommonwealth Scientific and Industrial Research Organisation (CSIRO)
  • Peter Kemppinen
    • Materials Science and EngineeringCommonwealth Scientific and Industrial Research Organisation (CSIRO)
  • Ming Chen
    • Materials Science and EngineeringCommonwealth Scientific and Industrial Research Organisation (CSIRO)
  • Scott E. Watkins
    • Materials Science and EngineeringCommonwealth Scientific and Industrial Research Organisation (CSIRO)
  • Kevin N. Winzenberg
    • Materials Science and EngineeringCommonwealth Scientific and Industrial Research Organisation (CSIRO)
  • Steven Holdcroft
    • Dept. of ChemistrySimon Fraser University
  • David J. Jones
    • School of Chemistry, Bio21 InstituteUniversity of Melbourne
  • Andrew B. Holmes
    • Materials Science and EngineeringCommonwealth Scientific and Industrial Research Organisation (CSIRO)
    • School of Chemistry, Bio21 InstituteUniversity of Melbourne
Article

DOI: 10.1007/s00339-012-7028-x

Cite this article as:
Singh, T.B., Chen, X., Wong, W.W.H. et al. Appl. Phys. A (2012) 108: 515. doi:10.1007/s00339-012-7028-x

Abstract

Charge extraction properties of various binary and ternary blends of organic photovoltaic devices covering both polymers and small molecules are studied. Due to their bipolar nature, both slow and fast carrier mobilities are identified from the extraction current transient. The equilibrium carrier concentration is also estimated for each of the blend films. The product of the slow carrier mobility and equilibrium concentration spreading two orders of magnitude can be used to estimate the short circuit current density. A good agreement between the estimated and measured short circuit current density is obtained with the accuracy reliant on the estimation of the slowest carrier mobility. This simplistic approach will be very useful to predict the short circuit current density for devices based on new materials.

1 Introduction

Solution-processed organic photovoltaic (OPV) devices offer the possibility of low-cost power generation from large area, flexible modules that can be manufactured using established printing processes. This field is underpinned by advances in small-molecule organic and polymeric semiconductors which have recently led to power conversion efficiencies (PCE) in excess of 7.4 % [1]. In these highly efficient devices, there is a common observation: When the film’s thickness increases, internal quantum efficiency (IQE) drops drastically with charge extraction and recombination becoming important determinants of efficiency [1, 2]. Often, there is a notion that charge mobility (μ) is a key efficiency indicator. In this communication, we demonstrate that μ alone cannot be used to accurately predict efficiency in solar cells.

In solar cells, one of the key determinants of PCE is the short circuit current density (Jsc). There are three techniques that can be used to estimate the Jsc for new materials without actually fabricating a solar cell; (i) optical modeling, a powerful technique that also allows estimation of Jsc [2, 3]; (ii) transient absorption spectroscopy (TAS), which allows the measurement of free charge generation efficiency (ηDISS(CT)) [4]; (iii) extraction of photogenerated charge carriers by using a linearly increasing voltage (Photo-CELIV) to measure the bipolar charge mobility (both electrons and holes), equilibrium charge carrier concentration (n) and lifetime (τ) [57]. The key difference between TAS and Photo-CELIV is that TAS optically probes all the charges including those deeply trapped within a ∼50 ns time scale, with negligible bimolecular losses prior to measurement. Photo-CELIV probes only relatively mobile charges over time scales spanning longer than ≈1 μs. Photo-CELIV studies reported to date have not fully covered a wide range of ordered and disordered small molecules and polymers [47]. Organic materials are well known to possess a high degree of disorder and short charge carrier diffusion lengths. In bulk heterojunction (BHJ) OPVs, charge transport/extraction is considered to be dominated by the drift of carriers under the built-in electric field, Ebi [8]. The role played by τ is also considered to be vital and is sensitive to many parameters including the relaxation of n toward the tail end of the density of states. Hence, measurement of bipolar (electron and hole) carrier mobility, μe,h, and n, as a function of electric field (E) is vital for any new material in order to estimate Jsc. In this work, the estimated short circuit current is referred to as Jsc,est=e×n×μslow×Ebi, where, e is the electronic charge, n is the equilibrium charge carrier concentration under illumination normalized to account for the fraction of photons absorbed at the excitation wavelength, and μslow is the slowest carrier mobility (either electrons or holes whichever is less). It is to be noted that in general, the photocurrent generation rate (G) is given by sum of the extraction rate of slow carriers (Eslow) and the recombination rate (R). However, if Eslow>R, then Jsc is determined by the mobility of the fast carrier (μfast) [9].

A major experimental challenge is how to measure μslow in optimized donor/acceptor blends. Most often only one transit time, which corresponds to one type of carrier, can be measured. This implies that μfast and μslow are either comparable or very different. In transistors, most often it is electron currents that are not observed due to the presence of unwanted electron trapping moieties at the interface. Another major scientific challenge is the accurate measurement of n from the very same device where μslow is extracted. The only technique that can combine the two measurements in one is Photo-CELIV [10]. We note that accurate measurement of μslow and n by Photo-CELIV is also affected by the offset voltage. Nevertheless, even with this uncertainty, a wealth of information is available from Photo-CELIV measurements.

In this report, we demonstrate that: (i) a general agreement is observed between independently measured, well reproduced Jsc from various optimized OPVs and Jsc,est; (ii) the product n×μslow is successfully measured for a wide range of magnitude covering both small molecules as well as polymer blends as a process to assayJsc,est; (iii) independently, μslow and n cannot be used to accurately predict Jsc,est.

2 Experimental

All of the materials reported here were prepared and characterized in various devices in our laboratories (see Electronic supplementary material, Scheme 1) [1113]. A significant amount of each polymer or small molecule, at least 0.5 g, at high purities was prepared as reported previously. For poly(9.9′-dioctylfluorene-alt-(bis-thienylene)benzothiazole) (PFOTBT), the chlorobenzene-soluble fraction was used, with an unknown but higher molecular weight than that of the chloroform-soluble fraction (Mw=17.0 kg/mol; polydispersity index(PDI)=1.71) [3]. For poly(9,9′-dihexylfluorene-alt-bithiophene) (F6T2), the batch had Mw=149.5 kg/mol and PDI=1.54 [13]. The batches of 9,9′-dioctylfluorenyl hexa-peri-hexabenzocoronene (FHBC) [12] and 7,14-bis(triisopropylsilylethynyl)dibenzo[b,def]chrysene [11] were >99 % pure as analyzed by HPLC. OPV fabrication and characterization procedures were as reported previously [11, 12]. Photo-CELIV measurements were performed on sealed OPV devices using a nitrogen pulsed laser (3 ns at 337 nm) as the excitation source with low peak power of 300 μJ/pulse. An Agilent function generator 33250A was used to supply the voltage pulse. The extraction currents were recorded using a 500 MHz Tektronix oscilloscope and a set of resistors. The delay time between the laser pulse and the voltage pulse was set using a digital delay generator, Stanford Research System DG535. Charge carriers were excited by a short light pulse and within the given delay duration (tdelay) the linearly increasing voltage pulse (increase rate A=dU/dt) of reverse direction was applied onto the sample electrodes. The pulse of current transient consisting of inter-electrode capacity current (j(0)=εεoA/d; d is thickness of layer) was analysed [10]. The possible influence of the intrinsic potential of the OPV on the amount of extracted charge was eliminated using a compensating offset voltage (Uoffset). From the time when extraction current reaches maximum (at tmax) the charge carrier mobility can be estimated according to expressions [10]:
$$ \mu = \frac{2d^{2}}{3At_{\max}^{2}} $$
(1)
The concentration of photoexcited charge carriers n, at a given tdelay, was estimated using the expression:
$$ n = \frac{1}{ed}\int_{0}^{t} j(t) - j(0)\,dt $$
(2)
where e is the electronic charge. The measured n is normalised with the absorbance of the materials at the excitation wavelength. Equation (1) assumes that n is low and not limited by charge recombination which is achieved by measuring a thin film sample under low laser intensity.

3 Results and discussion

Figure 1 shows the Incident Photon Conversion Efficiency (IPCE) spectra, of PFOTBT:PCBM, F6T2:PCBM and PFOTBT:F6T2:PCBM OPVs. The spectral response closely resembles the absorption spectra of the blends (see Electronic supplementary material, Fig. S1). The spectral responses are also additive, showing weighted contributions from the absorptions of the individual components. The inset of Fig. 1 shows the current density–voltage (JV) characteristics of the devices used for the IPCE measurements. Power Conversion Efficiencies (PCEs) of 4 % were achieved from the binary blend of PFOTBT:PCBM with an open circuit voltage, Voc of 965 mV and a Jsc of 8.85 mA/cm2. This result is an average of 6 devices with an area of 0.2 cm2 with a highest PCE of 4.3 %. These results are comparable to efficiencies reported previously [2] on a very similar device structure except for the replacement of LiF by a Ca electrode in the present study. The binary blend of F6T2:PCBM gave a PCE of 2.2 % with a Voc of 990 mV and a Jsc of 4.59 mA/cm2. The Jsc in the ternary blend, 6.18 mA/cm2, is approximately the average of the two binary blends and the ternary device has a PCE of 3.1 %. These results suggest that optimization of solar cells using ternary blends employing two polyfluorene-based donor polymers with the identical HOMO levels having different absorption band results in a device with no open circuit voltage loss and with wide spectral response. The results can be interpreted as being due to parallel-like OPVs [14]. The fill factor for all three device blends was around 0.50 and the Voc of the ternary blend is equal to that of the binary blend with the highest Voc (F6T2:PCBM, 990 mV). Table 1 summarizes the parameters extracted from the devices. From all these devices, it is clear that one of the key determinants of PCE is the Jsc. To explore this relationship with other types of donor materials, we also investigated devices based on a range of small molecules developed in our laboratories that exhibit PCEs ranging from 0.03–2.2 % [11, 12].
https://static-content.springer.com/image/art%3A10.1007%2Fs00339-012-7028-x/MediaObjects/339_2012_7028_Fig1_HTML.gif
Fig. 1

IPCE of PFOTBT:PCBM; F6T2:PCBM and PFOTBT:F6T2:PCBM OPVs; inset: their respective JV characteristics under illumination of AM1.5 simulated sunlight. The HOMO-LUMO energy band diagram of PFOTBT and F6T2 as determined by Photo Electron Spectroscopy in Air (PESA) and UVVis spectroscopy is also shown in the inset [see details as Electronic supplementary material, Fig. S2]

Table 1

Summary of parameters extracted from the optimized BHJ OPVs fabricated using the compounds and their blends with PCBM. The active layers were a 1:4 wt. ratio polymer:PCBM blend. The device area was 0.2 cm2 and the data is an average of 6 devices

Donor

Voc [mV]

Jsc [mA/cm2]

Fill factor

Efficiency [%]

PFOTBT

965

8.85

0.47

4

F6T2

990

4.59

0.49

2.2

PFT OTBT+F6T2

990

6.18

0.50

3.1

The charge extraction properties were studied by the Photo-CELIV techniques as described previously [57, 10]. Figure 2(a) shows the photo-CELIV transient of a ternary blend film of PFOTBT:F6T2:PCBM (0.5:0.5:4 wt/v) at different maximum voltage ramp, Umax, respectively. The photo-CELIV transient demonstrates a typical CELIV transient with increasing extraction current due to an increase of the applied voltage and carrier drift velocity. When n starts to decrease, due to the carrier arrival at the electrode, it forms an extraction maximum, tmax. The position of the peak markers indicate that tmax is shorter with Umax. However, estimated μslow using Eq. (1) remains as 3±1×10−3 cm2/vs. For this specific blend film, there is an absence of a second tmax meaning the measured μslow is for electrons. In this case, hole mobility in the blend is too low to be measured. This will influence the estimation of Jsc,est which will be discussed shortly. Similarly, Fig. 2(b) shows photo-CELIV transients of the same device for various delay times, tdelay. The extraction current decreases gradually with increasing tdelay due to carrier recombination. Since OPVs operate only within Ebi, we have applied the lowest possible UmaxEbi to be able to accurately estimate μ and n. From the data shown in Fig. 2(b), with Umax=1 V and electric field, E=2.5×104 V/cm, estimates of μslow=1.8×10−3 cm2/v s and n=1.35×1015/cm3 were obtained using Eqs. (1) and (2), respectively. n is normalized to the fraction of photons absorbed (1–10A, where A is the absorbance of the film at 337 nm excitation wavelength) within the least possible tdelay.
https://static-content.springer.com/image/art%3A10.1007%2Fs00339-012-7028-x/MediaObjects/339_2012_7028_Fig2_HTML.gif
Fig. 2

(a) Photo-CELIV transients are shown for devices based on a ternary blend PFOTBT:F6T2:PCBM (0.5:0.5:4 wt/v) at different applied Umax at tdelay=2 μs, (b) Photo-CELIV transients for various tdelay between the light pulse and the extraction voltage ramp. Photo-CELIV transients are shown for devices with binary blend films of: (c) PFOTBT:PCBM (1:4 wt/v) at different applied Umax, at tdelay of 2 μs and (d) HexaFHBC:PCBM (1:2 wt/v). Both (c) and (d) plots show the bipolar (fast and slow carrier) nature of charge transport of the blend films represented by two transit times (peak markers indicate the respective transit times for fast and slow carriers)

Figure 2(c) shows the photo-CELIV transients of a binary blend film of PFOTBT:PCBM (1:4 wt/v) at different Umax respectively. In the dark, at time t=15 μs, the capacitive (displacement) current step is seen in the transient due to an RC circuit response to an applied triangle-shaped voltage pulse where no charge extraction takes place. The photo-CELIV transient shows two maxima (tmax) which are indicated by the peak markers, representing μfast and μslow (bipolar), respectively. We note that the second tmax is not due to dark-CELIV as shown clearly with Umax=10 V. The observation of bipolar transport is found to be very sensitive to the stoichiometry of the blend [6]. For the PFOTBT:PCBM blend, μfast due to first tmax went up from 3.3×10−3 to 5.08×10−3 cm2/Vs with Umax=1 V→10 V. μslow due to second tmax went from 4.16×10−4 to 5×10−4 cm2/V s with Umax=8 V→10 V. The range of μ measured here are among the highest reported for OPVs to date [6, 14, 15]. The estimated ambipolar mobility values are the same order of magnitude as ambipolar FET mobility measurements performed on the same blends, which are presented in Electronic supplementary material, Fig. S3. Previous reports of space charge limited current (SCLC) measurements on hole-only devices with the same blend films yielded a μ of 8×10−5 cm2/V s [3]. If we consider our μslow as hole mobility, the values in this study are 5 times higher than the SCLC estimated μ. A plausible explanation for this difference is the role of n and polymer molecular weight on the measurement of μ. Bipolar charge extraction was also observed in a HexaFHBC: PCBM (1:2 wt/v) blend at different Umax (Fig. 2d).

Photo-CELIV transients for blend films of ten different films composing of P3HT:PCBM (1:1), PFOTBT:PCBM (1:4), F6T2:PCBM (1:4), PFOTBT:F6T2:PCBM (0.5:0.5:4), TIPS-DBC:PCBM (1:1) [11], HexaFHBC:PCBM (1:2), 2,5-FHBC:PCBM (1:2), 2,11-FHBC:PCBM (1:2), 2,11-FHBC:PCBM (1:1), and 2,11-FHBC:PCBM (1:3) were recorded. Considering the scope of this paper, only extracted μslow and n are summarized in Table 2. Detailed analyses of the Photo-CELIV data is shown in Electronic supplementary material, Fig. S4. The measured μslow for most of the blends in the present study are comparable with our bipolar FET mobilities reported elsewhere [12, 15], and for P3HT:PCBM blends reported by several groups [16].
Table 2

Summary of parameters extracted from the Photo-CELIV measurement UmaxEbi of BHJ OPV fabricated using the compounds and their blends with PCBM. Jsc,est and Jsc for respective devices are also shown

 

Composition

Ratio

Mobility

n [1/cm3]

μn [1/(cm V s)]

\(J_{\mathrm{SC\ Est}}\) [mA/cm2]

\(J_{\mathrm{SC}\mbox{-}\mathrm{meas}}\) [mA/cm2]

Polymeric donors

P3HT:PCBM

1:1

2.00E−3

1.90E+15

3.80E+12

12.16

7.60

PFOTBT :PCBM

1:4

6.50E−3

3.11E+14

2.02E+12

9.50

8.85

F6T2 :PCBM

1:4

2.91E−4

2.44E+15

7.10E+11

4.80

4.59

PFOTBT:F6T2:PCBM

0.5:0.5:4

1.80E−3

1.35E+15

2.43E+12

9.72

6.18

Small molecule donors

TIPS-DBC:PCBM*

1:1

8.50E−5

1.00E+16

8.50E+11

6.80

5.19

HexaFHBC:PCBM

1:2

8.20E−5

7.20E+14

5.90E+10

0.20

0.18

2,5-FHBC:PCBM

1:2

2.00E−3

2.50E+14

5.00E+11

2.00

2.03

2,11-FHBC:PCBM

1:2

1.53E−3

7.65E+14

1.17E+12

5.32

2.68

2,11-FHBC:PCBM

1:1

8.48E−4

1.04E+15

8.81E+11

3.66

2,11-FHBC:PCBM

1:3

9.00E−4

2.51E+14

2.26E+11

0.44

*Taken from [15]

We now turn to the observed correlation between n×μslow and Jsc for all the aforementioned devices. As seen from Table 2, there is a large variation of μslow and n among the blends. The products, n×μslow, are found to vary by orders of magnitude. Accordingly, Jsc,est is calculated as a function of n×μslow. It is particularly striking to note that μslow of blend films employed here differs by two orders of magnitude ranging from 8×10−5 cm2/V s and 6.5×10−3 cm2/V s and there is no correlation with Jsc,est. Similarly, measured n values do not directly correlate with Jsc,est. However, we observe a slightly higher n for devices with higher Jsc,est. We note that the n value measured here can be a characteristic of IQE. The magnitude of n is related to the yield of charge generation minus any fast component of the recombination kinetics and possibly trapped charges that are not extracted within the applied triangular pulse. As clearly seen, there is no direct correlation between n and μslower which means observation of larger n in some of the blends studied here is not solely due to higher μslow. n is related to ηDISS(CT) which depends on a number of parameters including temperature, electric field, initial charge separation distance, and dielectric constant. Initial charge separation distance is well known to be governed by the size of the molecule and the length of the conjugated segments. In Fig. 3, we compare the Jsc,est andJsc and show that there is a clear correlation between the two. The data presented here can be separated into two cases: (i) the blend films where we were able to measure μslow from the second tmax or both μslow and μfast and (ii) the blend films where we were unable to measure the μslow from the second tmax and, therefore, calculated μslow from the only tmax present. In the former cases, the agreement between Jsc,est and Jsc is greater than 75 %. Such examples are: PFOTBT:PCBM (1:4), F6T2:PCBM (1:4), TIPS-DBC:PCBM (1:1), HexaFHBC:PCBM (1:2), and 2,5-FHBC:PCBM (1:2). In the latter cases, due to overestimation of μslow, the independently measured Jsc is lower than Jsc,est with the accuracy falling to between 50–60 %. Such examples are P3HT:PCBM (1:1), PFOTBT:F6T2:PCBM (0.5:0.5:4), and 2,11-FHBC:PCBM (1:2). We also note that the Jsc data compared here was obtained under illumination of simulated AM 1.5 sunlight while the Jsc,est was obtained at monochromatic light irradiation at significantly lower light intensities. A mismatch factor is therefore unavoidable. The finding of a correlation between Jsc,est and Jsc also reflects that the measurement of μslow and n is not sensitive to the bandgap of the materials covered in this study. From a new material design point of view, it will be an advantage to consider design criteria such that blend films have both high μslow and n to achieve higher Jsc. The experimental measurement of both μslow and n for a given blend is therefore found to be very useful to predict the maximum attainable Jsc.
https://static-content.springer.com/image/art%3A10.1007%2Fs00339-012-7028-x/MediaObjects/339_2012_7028_Fig3_HTML.gif
Fig. 3

Plot of measured Jsc vs. estimated Jsc,est deduced from different blend materials employed in the optimized OPVs in our laboratory

4 Conclusion

In conclusion, a strong agreement between the estimated and independently measured short circuit current density covering films based on both small molecule and polymers has been presented. Charge extraction parameters such as carrier concentration and slow carrier mobility measurement were shown to be a very useful assay of short circuit current density. The product of the carrier concentration and slow carrier mobility was found to be the key parameter for current density rather than μslow or n alone. This result will be very useful to predict short circuit current density for new materials.

Supplementary material

339_2012_7028_MOESM1_ESM.doc (606 kb)
Correlation of charge extraction properties and short circuit current in various organic binary and ternary blend photovoltaic devices (DOC 606 kB)

Copyright information

© Springer-Verlag 2012