Overcoming the water oxidative limit for ultra-high-workfunction hole-doped polymers

It is widely thought that the water-oxidation reaction limits the maximum work function to about 5.25 eV for hole-doped semiconductors exposed to the ambient, constrained by the oxidation potential of air-saturated water. Here, we show that polymer organic semiconductors, when hole-doped, can show work functions up to 5.9 eV, and yet remain stable in the ambient. We further show that de-doping of the polymer is not determined by the oxidation of bulk water, as previously thought, due to its general absence, but by the counter-balancing anion and its ubiquitously hydrated complexes. The effective donor levels of these species, representing the edge of the ‘chemical’ density of states, can be depressed to about 6.0 eV below vacuum level. This can be achieved by raising the oxidation potential for hydronium generation, using large super-acid anions that are themselves also stable against oxidation. In this way, we demonstrate that poly(fluorene-alt-triarylamine) derivatives with tethered perfluoroalkyl-sulfonylimidosulfonyl anions can provide ambient solution-processability directly in the ultrahigh-workfunction hole-doped state to give films with good thermal stability. These results lay the path for design of soft materials for battery, bio-electronic and thermoelectric applications.

I enjoyed reading the manuscript and I believe the work is very relevant, both in advancing understanding of hole-doped polymer layers, as well as the associated technological advances made. Although high work function materials are already available, air stability in combination with solution processability is currently lacking. I therefore believe that this manuscript would be suitable for publication in Nature Communications.
I have one comment that the authors should address: Fig. 5 shows convincingly that hole injection from the hole-doped polymer is equally efficient as hole injection from the MoO3/Ag electrode, which, on itself, is a respectable achievement for an airprocessed hole-injection layer. However, it is not convincingly demonstrated that hole injection from these electrodes is Ohmic. I would like to encourage the authors to investigate a wider PFOP layer thickness range, and report the thickness dependence of the injected current density, which may reveal if an injection barrier is still present.
In this manuscript, the authors report their findings of stable ultrahigh-workfunction states beyond the bulk-water limit with thermal and ambient stability. DFT calculations were performed and doped thin films were characterized with UV-vis, FTIR, UPS and XPS. The Injection efficacy as well as temporal and thermal stability were also examined. Controlling the workfunction in ambient conditions is important for those in electronics and energy transformation and storage. The manuscript may be suitable for publication in Nat. Comm. after the following modifications/clarifications: 1. In the manuscript there are extensive discussions about the hydration of anions with water. The experimental verification, however, is with the dopant C2F5SIS, a perfluoroalkylsulfonylimidosulfonyl derivative immiscible with water. Can there be a simpler explanation to the observed high workfunction and stability as simply due to the hydrophobicity of the dopant that keeps water away from the surface? 2. In Theoretical considerations, the first water molecule is taken to be bonded to the ion cluster and the subsequent ones hydrogen bonded. Is the nature of the first bonding ionic or of some other type? 3. In Computational methodology, separability and additivity of the dominant interactions are assumed. The authors should justify the assumption. 4. The data presented in Fig. 2 indicate that for weak acids the donor energy decreases as the cluster size increases, while the reversed trend is observed for stronger acids. The authors should clarify if it is expected and what is the consequence. 5. Although UPS provides accurate measurements of the workfunction, it should be cautioned that complications may occur that affect the interpretation of the result. One possible complication is the photovoltaic effect if there's any surface band bending. Other factors include charging and dehydration in vacuum. Such possibilities in the present case should be discussed. Kelvin probe measurements may by-pass these problems. 6. Comparison of the measurements and calculations. While C2F5SIS is used in most of the experiments, there's no DFT calculation to compare with. It's peculiar that the authors did quite extensive calculations for potential dopants other than the one they actually used for the testing. If there's a reason behind it, they should explain it. In any case, the relevance of the calculation and the implications to the measured changes of the workfunction presented in Fig. 7 should be addressed. 7. The XPS in Fig. S5 is for mTFF-SO3. The authors should provide their rational on doing so instead of on the material systems that they studied to demonstrate the high workfunction. It should also be specified if the binding energy is defined from the vacuum level as in Fig. 7. 8. The doping level is deduced from UV-Vis measurements. The authors should check if such determination is consistent with the concentration analysis of the XPS core levels. More detailed analysis may even provide information on the hydration by detailed decomposition of the core levels, for example the O 1s. 9. The bias for the UPS measurements and pressure of the chamber should be provided. 10. The abbreviation pTFF is undefined. The molecular structures of the materials measured should also be provided in the Supplementary materials.

Reviewer 1
We thank this Reviewer for positive comments. The Reviewer has raised an important query, which we have now addressed with more experiments and a detailed note below.
1. "… it is not convincingly demonstrated that hole injection from these electrodes is Ohmic.
I would like to encourage the authors to investigate a wider PFOP layer thickness range, and report the thickness dependence of the injected current density, which may reveal if an injection barrier is still present."

Response:
We have studied this problem in some detail over recent few years. The existence of traps complicates simple analysis, but the conclusion of Ohmic contact attained is robust. See the new data included in new Supplementary Fig S12 now. For thin films (< 100 nm) of the benchmark semiconductor PFOP (ionization energy, 5.8 eV) with mTFF-C 2 F 5 SIS as hole injection layer, and Al as hole-exit contact, the current density shows the expected V 2 scaling for current densities larger than about 500 mA cm -2 that is characteristic of spacecharge-limited conduction (SCLC) of the Mott-Gurney law (Fig S12c, old data from 24 months ago). For larger film thicknesses and/or lower current densities, the Mott-Gurney slope drifts above 2. This is characteristic of trap limited SCLC, which points to the existence of traps in polyfluorene, the greater the deviation, the deeper the traps, as is well known in the literature [Paul Blom and co-workers, Nature Materials 2019; Martijn Kemerink and co-workers, Nature Materials 2019]. To be very sure of this and address any remaining scepticism, we have made new experiments with the device configuration pTFF-C 2 F 5 SIS/ PFOP/ MoO 3 , where the top evaporated MoO 3 acts as hole exit contact or hole injection contact, as desired, to provide the reference for the bottom polymer hole contact. The fitting of the PFOP film with two strong hole contacts at both its interfaces also better help to fill traps. We found indeed that the Mott-Gurney slopes for the thinner films and larger current densities (> 200 mA/cm 2 ) are very close to 2, while the slopes for the thicker films and lower current densities also become uniformly closer to 2 (Fig S12a and b, new data from last month). The thickness dependence also scales as expected, around d -3 to d -4 . Most compellingly, the hole current injected from the spin-on polymer contact is actually (slightly) larger than that injected from the evaporated MoO 3 contact. Since MoO 3 is well known to be practically Ohmic, this observation together with that of the correct Mott-Gurney index allows us to be sure that the polymer hole contact is Ohmic.

Reviewer 3
We thank this Reviewer for attention to many details. We overlooked to report some of these in the original manuscript -sorry. These are now compiled in various new Supplementary Figures, including Figure S5 for the chemical structures of the materials. Our line-by-line responses are as follows.
1. "Can there be a simpler explanation to the observed high workfunction and stability as simply due to the hydrophobicity of the dopant that keeps water away from the surface?"

Response:
This is the natural starting assumption, and was indeed our first working hypothesis. Detailed thermogravimetry however shows that the perfluoroalkylsulfonylimidosulfonyl anions are still fairly hygroscopic, binding about half as many water molecules as sulfonate anions, although they no longer confer water solubility to the polymers. See . This amount is sufficient to 'kill off' all the holes even if one sixth of the water can act as chemical trap, since two anions are present for each hole, one of them counterbalanced by a spectator cation for solubility. FTIR measurements show the tethered perfluoroalkylsulfonylimidosulfonyl-anion films in fact sorb moisture very rapidly from the ambient, Supplementary Figure S10. This is also made clear by Figure 6, which shows water sorption on two time scales, but de-doping occurs on the slower timescale.
The 'hydrophobicity' of the dopant anions is not sufficient to keep water away from the system.
In separate studies, we found that polyelectrolyte systems with both large 'hydrophobic' cations, like tetramethylammonium and tetraphenylphosphonium, and large 'hydrophobic' anions, like the ones here, are still hygroscopic, binding a significant amount of water in the as-deposited films. We traced this to the coulomb binding of water in molecular cavities in these materials.
Thus the reason why water can surprisingly be compatible with ultrahigh work-function states is due to energetics. The most vulnerable water associated with large anions can have their HOMO level sufficiently depressed to avoid becoming a chemical hole trap.
Were this not the case, it would have been impossible to achieve air stability for any solution processable ultrahigh-workfunction polymer, since elimination of water is not really possible in 'soft' matter containing ions. 2. "Is the nature of the first bonding ionic or of some other type?"

Response:
We assume the Reviewer is referring to the bonding between the various water species and the ion cluster. The bonding of water to the anion is by hydrogen-bonding. This is clear from the DFT calculations. [Yes, DFT/CAM-B3LYP/6-31++G(d,p) appears to reproduce Hbonding fairly well, based on benchmark calculations.] When the water is ionized, a hydronium ion forms upon geometric relaxation, as one would expect. The binding to the ion cluster comprises both coulomb and H-bonding. We have allowed geometric relaxation for the water molecules in the vicinity of the ion cluster. The results suggest the natural formation of H bonding network amongst the water molecules and also with the attached anion. This is now briefly discussed in the Method section.
3. "… separability and additivity of the dominant interactions are assumed. The authors should justify the assumption."

Response:
This is a necessary assumption in order to factorize the gargantuan degrees of freedom in the configuration space to smaller subsets that can be handled by calculations, in the same spirit as the Born-Oppenheimer approximation. The essential idea is:  (1) Quantum mechanical interactions, including H-bonding, are of very short range.
Therefore the hydrated anion complex X -(H 2 O) p is treated as a quantum mechanical entity.
(2) These energies are then shifted by the 'background' Coulomb potential set up at X -(H 2 O) p by the other surrounding ions. The ion-cluster configurations are sampled by MM2/PM3 and their energies computed at the less costly PM3 level since these configurations are primarily determined by Coulomb interactions between the constituent ions subjected to steric and tethering constraints. (3) Longer-range polarization of the matrix is evaluated by a classical polarizable continuum model.

"
The data presented in Fig. 2 indicate that for weak acids the donor energy decreases as the cluster size increases, while the reversed trend is observed for stronger acids. The authors should clarify if it is expected and what is the consequence."

Response:
We thank the Reviewer for pointing out that we did not explain so well in the earlier manuscript draft. Sorry. We have now improved the discussion with one new paragraph in the main text. Basically the differences have a really intuitive explanation. For weak acid anions, the product of chemical hole trapping in the hydrated anion complex is a radical, not a hydronium. Calculations show the radical is primarily derived from the anion. In this case, the initial anion charge is stabilized by water molecules. So as the hydrating water cluster size increases, the hydrated anion donor level becomes deeper, i.e., more difficult to detach its electron. For strong acid anions, the product is a hydronium. In this case, the final hydronium charge is stabilized by hydration. So as the water cluster size increases, the anion…water donor level becomes shallower.
The dependence on the ion cluster size, i.e., number of ion pairs in the local cluster, is also in opposite directions for these two cases. For the weak acid anions, increasing ion cluster size increases stabilization of the initial anion charge (Madelung effect). This lowers the donor level. For the strong acid anions, increasing ion cluster size also increases stabilization of the final hydronium charge, but this raises the donor level, as the hydronium is now easier to form.
Since hydration effects level off beyond 5 H 2 O per anion in the ambient, the donor energies in Fig 2 provide an estimate of the limit of stability of ultrahigh work functions for different hydration number and ion-cluster size. We have expanded three paragraphs to clarify these now.
5. "Although UPS provides accurate measurements of the workfunction, …. One possible complication is the photovoltaic effect if there's any surface band bending. Other factors include charging and dehydration in vacuum. Such possibilities in the present case should be discussed." Response: Yes, we have included a simple discussion now. Below is our detailed response.
Surface band bending. The materials are in the strong doping regime, doping level ≳ 10% of repeat units, corresponding to a few 10 20 cm -3 . At such high carrier density, surface depletion and band bending effects are not present. We know this is the case also because the measured work function is independent of film thickness from 5 to 20 nm. So there is no significant surface photo voltage effect.
Charging. The doped films have an electrical conductivity of ≳ 10 −3 S cm -1 , depending on semiconductor core. UPS and XPS were performed on films usually less than 20-nm thick. The photoemission current is typically 1 nA mm -2 . The voltage drop is negligible, much less than μV. We know this is the case also because there is no shift in XPS core level between the doped and undoped films.
Dehydration. In extensive work, we have established the work function is set by the electronic structure of the polymer and its doping level, modified by Madelung potential of the counter-ion and other spectator ions in the vicinity of the carrier. During UPS/XPS measurement, the film retains tightly-bound water at the level of 1-1.5 H 2 O per anion. This water is not ionized nor oriented, and so does not significantly affect work function. We know this is the case because the effective work function measured in the diodes by electroabsorption spectroscopy (where more water is present) agrees with the UPS vacuum work function to within 0.  Fig. 7 should be addressed"

Response:
We have fixed this gap now. We have added one more model anion for C 2 F 5 SIS, (CH 3 SO 2 )(CF 3 SO 2 )Nto Figure 2, Supplementary Table S3. The attachment of alkyl chain to the methyl end does not shift energetics much. We have also improved the discussions on anion effects. Please note that ions in Figure 2 are only model ions. They are not directly useful as counter-ions for hole-doped polymer organic semiconductors, because of 'dopant migration'. To overcome this, we need to immobilize them, for example by tethering to alkyl chains. Tethering through the alkyl moiety does not change the energetics of the hydrated anion. Indeed, experimentally measured stability tracks the predicted stability limits remarkably well for a simple zero-free-parameter model. Re Figure 7, we have included one more experimental set now, in new Figure 7b. This experimental set confirms that the measured recovery in work function in Figure 7 (now Figure 7a) is a feature of self-compensated, hole-doped polymers, due to the serendipitous formation of a reversible 'protection' surface dipole in the presence of liquid water. The presence of dipole is now confirmed by the rigid VB spectra shift observed in the selfcompensated hole-doped polymer, but not in the one that is conventionally doped with SbF 6 -.
7. "The XPS in Fig. S5 is for mTFF-SO3. The authors should provide their rationale on doing so instead of on the material systems that they studied to demonstrate the high workfunction. It should also be specified if the binding energy is defined from the vacuum level as in Fig. 7"

Response:
Yes, all binding energies are referenced to vacuum level. We have now made clear in the figure captions. The reason why we presented XPS of mTFF-SO 3 in Fig S5 is to point out the excessive loss of Na and conversion of SO 3to SO 3 H in the widely used sulfonate systems that somehow has escaped the other investigators. Following Reviewer's suggestion, we have now added a new XPS measurement on the more desirable system pTFF-C 2 F 5 SIS to show that this is indeed free of such problems, in new Supplementary Figure S9.

"
The doping level is deduced from UV-Vis measurements. The authors should check if such determination is consistent with the concentration analysis of the XPS core levels. More detailed analysis may even provide information on the hydration by detailed decomposition of the core levels, for example the O 1s."

Response:
For the triarylaminium polymer cores, hole doping leads to the emergence of P 1 and P 2 subgap polaron absorptions at ca. 1.0 and 2.5 eV respectively, similar to molecular triarylaminium salts. For these salts and polymers, it is well known that the polaron absorption band intensity scales linearly with the doping level, as also is the loss of the π*←π absorption band intensity. Therefore we used the fractional bleaching of the π*←π absorption band to estimate the hole doping level, calibrated using the undoped spectrum (0 h + / per repeat unit) and fully-doped spectrum (1 h + / per repeat unit).
Such an evaluation has been firmly established by XPS quantification for a model system through ratio of core intensity of N 1s + to total N 1s , and SbF 6 − to total N 1s [Png et al, Nat Commun 7:11948 (2016)]. The method however cannot be applied to R F SIS-compensated films, because R F SIS N 1s overlaps with triarylamine N 1s in the undoped state, and with the triarylaminium N 1s in the doped state.
We have studied the O1s intensities now to infer that H 2 O is present at 1.0-1.5 H 2 O per anion. We have mentioned this in the text now.