It seems that one prominent discussion in organic photovoltaics has officially ended, the one about the primary photoexcitation in disordered organic solar cells being excitons (with a binding energy clearly above 100meV) or free charges (with excitons having binding energies in the range of the thermal energy, i.e. <<100meV). Hwang, Moses and Heeger, have just published a paper on polymer:fullerene blends [Hwang 2008] where they describe the charge generation as
Mobile carriers are generated via a two-step process: initial ultrafast charge separation to an intermediate charge transfer (CT) bound state, followed by the transfer of carriers onto the bicontinuous networks.
They explicitly mention
[…] indicating ultrafast dissociation of the singlet excitons at the polymer-PCBM interface and the build-up of the initial CT state.
The paper is nice but in itself not that remarkable, except that previously, Moses and Heeger always claimed the primary photoexcitation to be free charges instead of bound excitons. Their measurements yielded exciton binding energies in the range of the thermal energy, i.e., no donor acceptor interface being necessary for charge separation. To quote an older paper [Moses 2000],
Thus, carriers are photoexcited directly and not generated via a secondary process from exciton annihilation.
Now I have to mention that in the new paper they use P3HT:PCBM, and in the old one MEH-PPV:PCBM. But as they do not mention this in the new paper, I assume that either I missed something, or they changed their point of view concerning the primary photoexcitation.
21 thoughts on “Comment on Primary Photoexcitation in Polymer:Fullerene Blends”
Hi, I tried to read the 2000 paper by Moses but I just cannot understand it. Anyway, I get the relevance of the general discussion. In a book published in 2010 by Heeger, Sariciftci and Namdas (Semiconducting and Metallic Polymers) it is mentioned that the exciton binding energy is less than 0.1 eV for PPVs and polythiophenes, so I am confused, is this discussion still going on?
Assuming the carriers could be photogenerated by thermal energy, how can the high photocurrent yield of BHJ devices compared to the ver low photocurrent yield of single layer devices be explained? Did they think/thought that somehow bimolecular recombination played a much more important role in single layer compared to BHJ? Or that maybe thermal energy is responsible for photogeneration of a few charge carriers, but the donor-acceptor interface produces a higher yield? Or does the point of view of low exciton biding enregies simply cannot explain the effectiveness of the donor-acceptor concept?
I just realized a mistake in my interpretation (the second paragraph of previous point is wrong). In single layer devices photoluminescence takes place in nanoseconds. While in BHJs back transfer recombinations is much more slow. So this explains why single layer devices would have a much more lower photocurrent even if excitons were dissociated with thermal energy. Anyway, do you know if this discussion is still going on?
I would guess that most researchers in the field agree that singlet exciton binding energies in neat polymers are much higher than 0.1eV, except for Heeger and some others. In the post above believed this discussion to be over, but it may not be completely the case;-) However, to me it seems only to be a minority discussion, despite its famous supporter. We measured the exciton binding energy of P3HT and found 0.7 eV [Deibel 2010].
Interesting topic binding energies, for me the binding energy will depend on the nature of the transition itself so there will be no hard and fast rules. Shame they are so very hard to measure. For instance it’s a shame Deibel 2010 doesn’t explain the difference in the UPS and IPES with reference 18 (nor tell us if the UPS/IPES refers to the crystalline or amorphous forms of P3HT), that the EQE spectra is better explained by exciton splitting at the aluminium electrode and a higher hole mobility in P3HT than a band gap transition that leads to free charges (such transitions are typically highly forbidden and wouldn’t compete with neutral state absorption and exciton splitting at electrode as a current source in a device such as this – remember light has to get to the aluminium electrode hence onset at 2.5 eV) and that the emission from P3HT likely originates from an excimer state and not the P3HT chain itself and so any voltage quenching results only tell us how to dissociate hole and electron on the excimer and not on the main chain (and that’s if you are happy with Onsager-Braun theory being used in this way:-)).
Shame on me?! Well, I am all for constructive criticism. However, it is not what you say, but how you say it.
Anyway, here our reply.
Ref. 18 [Feng 2005] as well as we measure regioregular P3HT. We have not determined crystallinity and have no additional information. PES Data of Ref 18 and our data are in agreement. The IPES data has lower resolution than ours and makes it difficult to determine the onset, which we used to determine the energy gap. We point out that our PES and IPES data is in very good agreement with the measurements of [Kanai 2010] who found – with a similar sample preparation – also a gap of 2.6 eV.
In our paper, we explicitly state “Synthesis residues such as Ni—typically below 100 ppm relative to P3HT—or dissociation at the electrodes can be responsible for this low-energy photocurrent, and would explain its low magnitude.” Of course, this contribution only explains the photocurrent starting at the absorption onset, but why should dissociation at the electrode lead to a strong increase of the photocurrent at 2.6eV? Your explanation “remember light has to get to the aluminum electrode” should not be relevant for a film thickness of only 70nm. Thus, we believe our explanation to be correct, in particular as the transport gap is also 2.6 eV.
Concerning “highly forbidden”, there is a finite absorption at this wavelength. Also, we do not claim that the absorption is due to direct band-to-band transition, it is rather that the photogenerated excitons are so hot that their vibrational/thermal energy drives the dissociation.
We agree that this is possible and likely. However, also the absorption by the P3HT aggregates is rather dominant, so why should this contribution be unimportant for the luminescence? Certainly, this approach only gives a lower limit of binding energy.
Not so happy, but as long as there is nothing better it is sufficient for an estimate. Also, at room temperature Braun-Onsager is probably not far off.
My bad, no shame on you Deibel, for what it’s worth I thought the paper was a good one, so good I read it twice when it was first published. I still stand by my points however, just highlighting that despite all the good work you put into this paper there is some doubt.
Ok, S, and I do value your opinion highly. So how to verify the results (or show their limits), possibly with independent measurement techniques?
Thanks for the compliment. OK, here goes. Don’t know really how to do a better job of looking for binding energies than you are doing now. The first problem is P3HT; regioregular P3HT is a two phase system (amorphous and crystalline) and we have to treat it as such. (Work on mdmoppv might be easier as mdmoppv is mostly amorphous and so we don’t have that problem. c.f. Kern 2011) Trying to separate the bits of P3HT that behave as amorphous and crystalline is difficult.
The work of Kanai et al is good as it looked at both regiorandom and regioregular P3HT so that’s a good place to start. Unfortunately, UPS/IPES on regiorandom P3HT find that the amorphous transport gap is about 2.8 eV, but this is also the optical band gap of RRa P3HT, namely 2.8 eV at peak. Does this mean that there is no binding energy? lf you use the onset energy of the RRa P3HT absorption as a measure of the optical gap (why would you want to do that, it’s not a median value..) then you have 2.8 transport gap versus 2.5eV optical gap and this leaves a binding energy of 0.3eV. Is this the right value? Seems a bit speculative (and small…)
When we go to regioregular P3HT we have to contend with an additional crystalline form of P3HT with a 2eV optical band gap, a decrease of 0.8 eV over the amorphous optical gap, but the transport gap hardly moves, only shrinking to 2.6eV. Is this a convoluted response? I am no expert, but I can’t understand easily the data that comes from UPS/IPES on P3HT (despite the excellent work of Kanai et al). Maybe IPES is too surface sensitive to give us useful bulk values anyway. I couldn’t find a lot of papers in the literature on this technique so I guess some more work in this field, discussed in the mainstream, would eventually help.
With the EQE measurements, deconvolution of competing mechanisms is also difficult. Measurements made with different thickness P3HT devices will reveal if there is a relationship between light penetration depth and EQE. Once you have examined how much a difference it makes, you can then go on to discuss (or not) wether absorption by amorphous P3HT (at .2.5 eV onset) creates photocurrent more efficiently than absorption by crystalline P3HT (hence explaining the increase in EQE). This would be hard to rule-out outright, but you could try and vary the crystalline/amorphous ratio in the devices by changing solvent and by making crystalline P3HT / amorphous P3HT bilayer devices (potentially by lift-off techniques). All this might tell you something. My guess is that absorption by amorphous P3HT plays a big role in the EQE change with a minor role played by light screening. It is easier to comment on your suggestion of charge creation by pumping crystalline P3HT with 2.5 eV light, with the high energy exciton subsequently falling into the transport gap creating polarons. If this was the case fs optical pump probe spectroscopy and microwave conductivity would be able to see such charge creation happening on a suitable timescale and with suitable yields. From my understanding of this literature there is not convincing evidence to support that.
With the PL quenching, why don’t you isolate some P3HT chains in a polystyrene matrix and then do PL quenching as a function of voltage measurements. This is a simple system and could give you the (low end) binding energy of an exciton on a single P3HT chain. It will also help clarify if the OB theory is suitable for exciton splitting under voltage. I don’t think OB theory is the right choice here. For separation of already individual charges, such as CT states in organic solar cells , I think the theory is appropriate (and why wouldn’t it be?). For separation of excitons however, because hole and electron are correlated completely, i.e., it acts neither like a hole nor an electron but as something else, I don’t think it is appropriate. For instance, OB theory in Deibel 2010 needs the sum of hole and electron mobility as an input, but I don’t see how the mobility of free holes and electron will help in understanding exciton dissociation. What is the mobility of a hole with an electron right next to it anyway? What is the mobility of a hole in a exciton shaped relaxation well rather than a polaron one? Because electron and hole are one, individual hole and electron mobilities are meaningless. This may explain why the value for the exciton diffusion length (Ld) derived from individual hole electron mobilities is incorrect (?), and always would be I guess, because exciton movement is mostly by Forster energy transfer and not a sequential arrangement of hole and electron hops. At high fields, where the electron and hole in an exciton are screened from each other, then the concept of an exciton may be incorrect, and so OB theory could start to become correct. In this case you might then be able to assess the binding energy using OB. To be more accurate however, we would also have to consider the role of electric field on the absorption transition (does absorption in an electric field create a more distorted exciton in the first place) and how does it change the oscillator strength of the fluorescence with changes to kf (some of this may be evident in Fig.1 of Kern 2011). My thoughts are that OB however is not good here, and that a field dependent Marcus electron transfer type theory would work better instead. An electron transfer out of the exciton would effectively split the exciton and quench the emission. From Marcus you could get an energy barrier that would represent the same as ‘binding energy’ does here and a rate for the exciton splitting. You could then go out and use time resolved microwave conductivity or optical measurements to confirm that rate and the model. Finally, the description of electron-hole separation (rs) could be correlated to the idea of dipole moments of excitons by electro-absorption measurements.
What do you think?
Thanks! Here my thoughts before leaving for a few days of vacation;-)
You are right, to investigate the exciton binding energy really in detail, you need to distinguish amorphous and crystalline regimes. Not sure if this is possible, though. In the mix you could determine the lower boundary. Maybe easier to consider regiorandom and regioregular cases, or indeed an amorphous polymer such as a PPV derivative. Concerning peak vs onset, we compared two onset values (transport gap vs absorption), so that is consistent. I do not have absorption data of regiorandom P3HT at hand, but is the absorption onset (once again) really as high as 2.5eV? If so, indeed 0.3eV sound too small to me as well. By the way, in the next paragraph you write that the amorphoug/regiorandom optical gap is at 2.8 eV (i.e. 0.8eV larger than for regioregular P3HT), is that a typo?
Concerning EQE, I agree, the way to go is to test different thicknesses, and ideally consider IQE. Concerning the combination os fs photophysics with transient microwave conductivity (TRMC): the latter has limited time window (due to the cavity) of usually at least 10 to 100ns. I think that this combination has not been used much in literature, but if you give me a hint to suitable paper, I’d be happy to read a bit more about it. Concerning IPES, not many groups can do it (properly), therefore you do not see it much in literature. For isntance, you need to very carefully make sure that you are not charging your sample with the incident electrons, which will lead to energy level shift.
Field induced PL quenching in a polystyrene matrix will lead to charging… will reduce electric field. We tried blocking layers. Comment concerning OB for singlets might be right. For criticism concerning the application of OB to CT, see e.g. work by M. Tachiya. Also, our Monte Carlo data from the 2009 paper could not be completely described by OB. An alternative model with activation energy, even simpler than using Marcus theory, was recently presented [Ojala 2011], although only for bilayers. Concerning Electroabsorption, I know only the basics of the principle, so how can it complement the other measurements?
At least the EQE we can do in the near future, for the other measurements I am still not sure of the best combination.
Although we have not had time yet to look into some of these issues experimentally, one small addition: the mobility in Braun-Onsager comes from the Langevin term and is included due to detailed balance (see Braun’s derivation in his 1984 paper). Therefore, it is the sum of electron and hole mobility and not any “CT mobility”.
Hi everyone. Just a few more imputs and no big answers on this quite interesting topic:
– Concerning the role of regiorandom P3HT, I checked a bit where the absorption’s onset is: we find it around 2.2-2.3eV in films. This corresponds pretty well to the absorption onset of isolated (non aglomerated) chains of regio-regular P3HT in solution (see [Clark 2007], fig. 1a and 1b), and so rather likely to the onset of the amorphous domains’ absorption in regio regular P3HT films.
– Concerning the role of screening in the EQE results, I agree that this effect most likely exists, and I also agree that the effect is probably less important than other effects (be it over-transport-gap excitation or amorphous regions excitation). Indeed, excitation close to let’s say 2 eV, where the absorption —and thus the screening— is more or less the same as at 2.8eV, results in a photocurrent generation indeed higher than at the aborption maximum of P3HT (2.2-2.4 eV) but only slightly, and clearly lower than the photocurrent generated by an incident light of energy around 2.8 eV. Well, this is clearly not quite the proof that a thickness-dependent study would be, but it’s already a hint I would say.
– To come back to regio random P3HT again, it is expected that intrachain exciton (such as those encountered in regio random P3HT) have lower binding energy than interchain ones (such as those encountered in regio regular P3HT) (see for example [Brazovskii 2010] section 5, p 2459). I would not go that far as to numerically evaluate this difference, so
I can not tell if a binding of 0.3 eV for RRa-P3HT compared to 0.7 eV for RR-P3HT is correct. But at least the tendency is the expected one.
– Concerning the generation of charges in pure P3HT upon high energy light excitation: femtosecond pump probe spectroscopy indeed sees it: Piris et al. claimed a charge generation efficiency of up to 15% in pure P3HT in that time scale ([Piris 2009]). Unfortunately, due to efficient Langevin-type recombination, those charges do not survive much over the ns time scale and are not visible at room temperature in a timescale over 10ns. However, they can be seen at lower temperatures (by transient absorption), although it’s a difficult trade-off to not decrease the temperature too much and increase the lifetime of other excited species (triplets? interchain excitons? polaron pairs?) and thus disturb your measurements (see [Gorenflot 2014]).
Generally, I agree that there is a lot more to learn, for example as you propose, with the help of a polystyrene matrix. Or also by trying other excitation wavelengths for field induced PL quenching. Note that excitation wavelength-dependent measurements exist in literature on MeLPPP [Hertel 2002, figure 3 (I think)]. The results seemed to be that the absolute quenching value at a given field depended on the excitation wavelength, but the field-dependence of the quenching was always the same (more visually: the higher exitation energy resulted in kind of an vertical offset on a log-log Quenching vs Field graph). But of course, the material is more simple than regio-regular P3HT.
Hope I brought more help than confusion… Not quite sure :/
Oh, right, I forgot the pb of chargeing in polystyrene matrix :/
Hi, I keep visiting this blog and learning from it, thanks a lot for your good explanations. Regarding this interesting topic, I just read a paper in which a different idea is described (Adv Funct Mater 2012 from Heeger group). Authors propose that before charge thermalization and exciton formation the excited state is less localized than the exciton, and hence the electron transfer from donor to acceptor may happen before exciton formation (before thermalization). They mention that the thermalization of the excited state to the exciton state is on the picosecond regime, while the ultrafast electron transfer from donor to acceptor occurs faster.
The way I understood this is that at least for photons with energies above the bandgap, charge transfer states or free charges will be formed without exciton formation. Their main argument I think is that exciton diffusion is slow, so that in the femtosecond regime the exciton only gets to diffuse about 0.1 nm, and hence this picture cannot explain ultrafast electron transfer experimental data.
In the paper it is even mentioned that exciton diffusion and previous discussion on binding energy are not that important if their proposed charge photogeneration mechanism is correct. What do you think about this idea? If true, do you think that the excess thermal energy available from the dissociation of this delocalized states plays an important role in free charge carrier formation?
I am a new in this field, so I am not even sure if this is a new idea or if I misunderstood something. I hope I got the general idea from the paper correct…
Hi Jose, on vacation, answer will take a while… Best, Carsten
No-one really knows what goes on at the kind of timescales you are talking about (tens of femtoseconds or less) or what the nature of excited states are before atomic nucleii start to adjust to the excitation and ‘confine’ it. You have to remember that most of this can only be tested via pump-probe spectroscopy methods and although lasers and techniques are getting faster they have only recently achieved a few femtosecond timescale resolution and then only with specialist equipment. Given the lack of information all ideas are possible. The difference between the ideas of Heeger and colleagues and the current mainstream is that if electron transfer is very quick then charge spitting can occur before ‘confinment’ and exciton binding energy doesn’t need to be accounted for in calculations. (I too hope I paraphrased this correctly.) Wether they are correct or not, who knows?
Hi, it is true it is impossible to know what happens at those timescales. The experiments done for which they made the conclusions mentioned on my previous post were fluorescence up conversion spectroscopy studies. These studies involve a very short laser pulse to excite electrons in the polymer, after that they analyze fluorescence as a function of time. Their fastest fluorescence spectra is at 0.5 ps after the excitation. The fluorescence spectra shows evidence that the excited states finish relaxing after 200 ps. However, at 0.5 ps they observe half of the full Stokes shift, implying that half of the excitations relax before this time.
What I do not understand yet is why do you need complete exciton dissociation (all or almost all generated excitons to dissociate) to explain ultrafast charge transfer experiments? This is their main argument to indicate that maybe the photogeneration mechanism is different.
I am not very familiar with experiments proving fast charge transfer, but I remember seeing that high resolution transient absorption studies show how the periodic relaxations observed in the transient absorption of for pure polymer disappear after a few tenths of femtoseconds when adding an electron acceptor. Do you need charge transfer from a high percentage of the photoexcitations to observe that results? Maybe the effects introduced by a few excitons that dissociate are enough. I do not know…. Anyway, has anyone study the effect of the excess photon energy on charge photogeneration efficiency? Assuming the new idea is correct, do you think this is important?
hmmm, these are some good questions and I don’t have an easy answer. I like the FUC work of Heeger et al and think the experiments were brilliant but I still have doubts about the interpretation of what is happening before the resolution of their equipment (i.e. <500 fs). Before then it remains guesswork although educated. There is also an issue that although the exciton is still reconfiguring bond lengths at 200 ps, this does not necessarily have any bearing on the strength of the exciton binding. You could argue that exciton binding depends on how closely overlapped HOMO and LUMO orbitals are in space. A very long exciton but with good overlap between HOMO and LUMO orbitals in space would have a high binding energy while even a small exciton but with HOMO and LUMO on very different parts of the molecule would have a weaker binding. I must concede however that a longer exciton makes it more physically possible to have a larger separation between hole and electron in exciton, it's just not guaranteed.
Transient absorption measurements are more-or-less linearly correlated to concentration, so it measures what the bulk of the population are doing. A small subset of transient species could not be followed unless they have a specific probe wavelength where they dominate the signal.
Yes, I have briefly, studied the effect of wavelength on charge photo-generation efficiency but it would I guess depend on polymer type, purity, morphology, presence of acceptors, and photo-excitation density. Until you have a top-notch grasp of polymer photophysics this kind of stuff is hard to understand. I am more than ten years into it and it's hard. My advice would be to read everything recent and form a consensus and then pick holes in it for experiments. As a general rule, anything written before 2008-ish be wary of (as well as anything that Carsten has ever written on the subject…lol).
Thanks for the reponse and nice explanations. Regarding the effect of the wavelength in charge photogeneration… I was thinking in an experiment: What if IQE is obtained by varying the light intensity at each wavelength such that the mobility and recombination conditions are the same in the whole range of wavelenghths of the experiment? I think under this conditions the IQE might represent charge photogeneration efficiency. However, I do not know if this would be an easy experiment. For example, does the mobility depend on charge density or on the density of occupied states above the LUMO level? I am curious about this things but unfortunately I haven’t had the time this last days to read more.
This experiment sounds plausible theoretically but due to trap s in organic systems mobility can depend on density if your charge density is below the trap limit (traps in system slow down charges). Above the trap-limit, mobility should be constant (I guess, Carsten does a lot of stuff on trap measurements and mobility as do Imperial College, look for the work of O’Regan and Shuttle). Also in this experiment it would be hard to distinguish geminate and second order recombination depending on wavelength.
This experiment could work but only at very very low light intensities in a system were there has been proven to be no geminate recombination. At very low light intensities there would be no second order recombination, plus minimal trap filling so mobility would not depend on concentration (i.e. all traps are available to each charge), and then you would finally be able to extract directly (free) charge generation efficiency from incident photons (after accounting for absorption including constructive/destructive interference.) The current might be so small though that you could not measure in these conditions.
Thinking big, you could do this on a single photon level, firing a single photon in and then somehow trying to measure that single hole or electron. You could do this by trying to get that single electron into a multiplier so it can be measured, or by introducing a fluorescent molecule and getting the recombination of hole and electron to happen there so that you can then detect that single photon (or not).
However you do the experiment, I think it actually worth doing.
Charge photogeneration was studied on neat polymers [Hertel 2002] and also blends, including the IQE [Lee 2010]. Indeed the trick is to adjust the light intensity to very low values in order to avoid nongeminate recombination (direct and trap-assisted). Lee found that a sub(polymer)gap excitation into CT states lead to the same high IQE than abovegap excitation for P3HT:PCBM. I like this work a lot, although it would be nice to see it for other polymer systems as well. However, the absorption cross section of the former is much lower, so that most charges by far are generated from above gap illumination in an operating solar cell, of course. Field dependence of charge photogeneration was, e.g., studied by the Neher group [Kniepert 2011], showing that at least in the working regime of the solar cell the photogeneration is field independent for annealed P3HT:PCBM. We showed recently [Mingebach 2012] this this holds true for lower temperatures as well. The Neher group also looked at other materials, showing field dependent photogeneration with a nicely optimised measurement setup [Albrecht 2012]. All of these TDCF measurements where all taken at low light intensities. Still does not tell you much about the detailed photophysics at ultrafast timescales, though;-)