In Situ Generation of n‐Type Dopants by Thermal Decarboxylation

Molecular doping is a powerful and increasingly popular approach toward enhancing electronic properties of organic semiconductors (OSCs) past their intrinsic limits. The development of n‐type dopants has been hampered, however, by their poor stability and high air‐reactivity, a consequence of their generally electron rich nature. Here, the use of air‐stable carboxylated dopant precursors is reported to overcome this challenge. Active dopants are readily generated in solution by thermal decarboxylation and applied in n‐type organic field‐effect transistors (OFETs). Both 1,3‐dimethylimidazolium‐2‐carboxylate (CO2‐DMI) and novel dopant 1,3‐dimethylbenzimidazolium‐2‐carboxylate (CO2‐DMBI) are applied to n‐type OFETs employing well‐known organic semiconductors (OSCs) P(NDI2OD‐T2), PCBM, and O‐IDTBR. Successful improvement of performance in all devices demonstrates the versatility of the dopants across a variety of OSCs. Experimental and computational studies indicate that electron transfer from the dopant to the host OSC is preceded by decarboxylation of the precursor, followed by dimerization to form the active dopant species. Transistor studies highlight CO2‐DMBI as the most effective dopant, improving electron mobility by up to one order of magnitude, while CO2‐DMI holds the advantage of commercial availability.


Introduction
In the past few decades, organic semiconductors (OSCs) have developed from lab curiosities to commercial applications. They have found use in technologies such as organic field-effect which may be added to the OSC layer via thermal deposition, through admixing, or via post-deposition techniques. [14][15][16][17][18] Although n-dopants have also been investigated for many years, their transition from fundamental research to common practice has progressed more slowly than that for p-dopants. [5,11] One of the main obstacles has been the air sensitivity of n-type dopants, a consequence of their typically electron rich nature. This is particularly the case for dopants that work by the electron transfer mechanisms, in which an electron is transferred from the highest occupied molecular orbital (HOMO) of the dopant to the lowest unoccupied molecular orbital (LUMO) of the host. In order to make this energetically favorable, the dopant should have a high lying HOMO (i.e., small ionization potential), but this can lead to difficulties with synthesis, storage, and handling. [19][20][21] One approach to solving the air-sensitivity issues has been the development of n-dopants that work by alternative mechanisms. For example, one of the most popular n-type dopants to date, (4-(1,3-dimethyl-2,3-dihydro-1H-benzoimidazol-2-yl) phenyl)dimethylamine (N-DMBI), belongs to a class of dopants that act through a doping mechanism involving hydride transfer, rather than electron transfer. [22][23][24] Such species are typically air stable, but can suffer from the drawback of undesired side products as a result of the hydride transfer mechanism, as well as poor predictability regarding their suitability with different OSCs. [23] An alternative strategy to address stability concerns is the design of so-called precursor-type dopants: air-stable dopant precursors that can be activated in the presence of the OSC. Examples of such dopants include dimeric species where electron transfer is coupled with the cleavage of weak CC bonds -a mechanism that may be initiated thermally or through photoactivation. [25][26][27][28][29][30] In our search for new precursor type dopants, we drew inspiration from the field of so-called super-electron donors, in which organometallic reductants have been replaced by fully organic reducing agents. [31][32][33] These are often based on highly electron rich amine substituted alkenes, which are formed by dimerization of carbene precursors. Facing similar issues with air-stability, Broggi and co-workers have pioneered the use of carboxylate protected carbene precursors, which are bench stable, but can be thermally decarboxylated in situ to form carbenes. Their subsequent dimerization generates the active reducing agent. Such species include 1,3-dimethylimidazolium-2-carboxylate (CO 2 -DMI) and 1,3-dimethylbenzimidazolium-2-carboxylate (CO 2 -DMBI), displaying chemical stability under ambient atmosphere over several months. [34,35] This concept has also recently been demonstrated in the context of doping by Pei and co-workers, who utilized CO 2 -DMI via a post-deposition doping approach in thermoelectric devices. [36] In this work we explore the use of air-stable carboxylate adducts as dopants for the active layer of OFET devices. We show that CO 2 -DMI and the novel dopant CO 2 -DMBI can be used to dope n-type OSCs through in situ thermal activation in the solution formulation, prior to coating. Furthermore, the versatility of these dopants is demonstrated through successful implementation in a variety of n-type OSCs -P(NDI2OD-T2) (also known as N2200), PCBM, and O-IDTBR (full chemical names provided in the Supporting Information) -encompassing both polymeric and small molecule species with varying energetic characteristics (Figure 1). We investigate the origin of the doping effect and compare electronic and morphological effects of the two dopant precursors. Our findings show that CO 2 -DMBI is the more effective dopant throughout, improving electron mobility by up to one order of magnitude compared to the pristine semiconductor.

Optoelectronic Properties of Doped Systems
While doping mechanisms may vary, and sometimes emphasize different aspects of device physics, they generally include an alteration of the electronic properties of the host material upon charge transfer between the host and the dopant. [5,37] Several studies on doping mechanisms have established integer charge transfer, i.e., electron transfer between the dopant and the host material, as one of the main doping mechanisms, although there are variations between different doping systems. [38][39][40][41][42] As electron transfer in a doping context holds many similarities with redox chemistry, investigating the optoelectronic and electrochemical properties of dopant/OSC blends is a simple screen for the possibility of device doping and helps to establish the doping mechanism.
As an initial indication of relevant electron transfer processes, absorption spectra were collected for pristine and doped OSC solutions in chlorobenzene. Studying the normalized absorption spectra of P(NDI2OD-T2) solutions doped with CO 2 -DMI and CO 2 -DMBI shows a clear suppression and slight blue-shift of the peak ≈700 nm, with the emergence of a new peak ≈485 nm upon doping (Figure 2a; Figure S1, Supporting Information). This change is more prominent upon doping with a very high dopant concentration (10 wt.%), and is very similar to absorption trends observed upon electrochemical reduction of P(NDI2OD-T2), indicating the formation of a radical anionic species. [43] Similar effects have also been observed in previous studies of doping of P(NDI2OD-T2). [44][45][46] The effect was much more pronounced with CO 2 -DMI compared to CO 2 -DMBI, indicating a more reductive effect upon thermal activation of CO 2 -DMI.
Normalized absorption spectra of doped PCBM solutions show a slight increase in intensity of the broad peak ≈500 nm and the sharp peak ≈430 nm, along with a lesser drop in intensity ≈315 nm (Figure 2a; Figure S2, Supporting Information). These changes are more pronounced with CO 2 -DMI, which also elevates the baseline in the region above 700 nm. Similar trends have been observed in previous studies investigating n-doping of PCBM and its derivatives, once again indicating the formation of radical anions. [40,47,48] Meanwhile, differences in absorption spectra of pure and doped O-IDTBR solutions are more subtle, with high dopant concentrations (10 wt.%) leading to a very slight blue-shift of the main peak ≈645 nm and a slight enhancement of the peak ≈405 nm. The blue shift is much more pronounced in thin films ( Figure S3, Supporting Information), which is possibly an effect of changes in morphology rather than energetic properties. Once again, CO 2 -DMI has a stronger effect on the absorption spectra than CO 2 -DMBI. In the case of thin-film absorption spectra of P(NDI2OD-T2) and PCBM, trends are less pronounced, and more difficult to interpret, possibly as an effect of competing morphological and energetic effects on the absorption characteristics (Figures S1, S2, Supporting Information).
To confirm the electron transfer from dopants to host materials and the formation of radical ionic species in doped thinfilms, electron paramagnetic resonance (EPR) spectroscopy was conducted. [49] Figure 2b shows the normalized spectra of films of each OSC material, doped with 1 wt.% of each dopant. Sharp peaks in the region ≈3375 G (g-factor 2.000-2.005, Figure S4, Supporting Information) confirm the presence of unpaired electrons in all doped systems. [50] In contrast, EPR peaks are completely absent in pristine films in all cases, as would be expected for the neutral species. These trends align well with those typically seen in doping processes involving integer charge transfer. [49]

Dopant Activation
Previous studies on carboxylate precursors have provided insights into the activation mechanism of such species. [34,35] It is clear that the carboxylate group -that acts as a stabilizing   group against spontaneous oxidation in air or other environments -is removed as carbon dioxide gas upon heating, leaving behind an N-heterocyclic carbene (NHC). In the case of CO 2 -DMBI, the addition of small amounts of water, or UV radiation, has also been demonstrated to accelerate the activation. [35] NHCs are known to reversibly dimerize into their corresponding tetraazafulvalenes, in a process known as the Wanzlick equilibrium, with the position of the equilibrium determined by a variety of factors, such as N-substituents and the singlet-triplet energy gap. [51][52][53][54][55] It is typically these neutral dimeric species that act as electron donors, driven by aromatic stabilization of the oxidized products (Figure 3a). [33] The dimeric form of NHCs has been thoroughly studied in the context of chemical reduction, including the investigation of numerous diimidazolinylidenes and dibenzimidazolinylidene derivatives. [31][32][33] Di-1,3-dimethylbenzimidazolin-2-ylidene (DMBI 2 ) has been exemplified in literature as a thermodynamically stable dimer, benefitting from a singlet-triplet energy gap lower than that of DMI 2 and little disruption of any aromatic properties after dimerization, as its aromatic character is located primarily on the fused benzene ring and may not fully extend to the 5-membered ring. [53,56] Likewise, mechanistic studies of CO 2 -DMBI confirmed the formation of DMBI 2 upon decarboxylation: as the carboxylate adduct was heated in presence of hexachloroethane (C 2 Cl 6 ), a dichloride salt of the dimer formed, confirming the formation of highly reductive DMBI 2 . [35] Notably, the DMBI 2 dimer has a structure similar to previously reported dopant DMBI-BDZC, and to some extent tetrakis(dimethylamino)ethylene (TDAE). [57,58] Imidazolin-2-ylidene, on the other hand, exists primarily as a monomeric carbene, with a near thermoneutral dimerization,  insufficient of overcoming the entropic loss and kinetic barriers thereof. [53,56,59] Although DMI 2 has been successfully isolated through reduction of its oxidized form, it is extremely unstable, and decomposes into the monomeric carbene within minutes. [60] Furthermore, mechanistic studies of the decarboxylation of CO 2 -DMI in dimethylformamide (DMF) has suggested the formation of a formamide-derived electron donor following deprotonation of the solvent by the monomeric carbene. [34] Meanwhile, it is less certain what species would act as the active dopant in an aprotic environment in the absence of DMF. [36] To investigate this matter, we heated CO 2 -DMI and C 2 Cl 6 in anhydrous chlorobenzene at 80 °C for 1 h. Mass spectrometry and 1 H NMR spectroscopy of the resulting solids confirmed the presence of DMI 2 Cl 2 , suggesting that DMI 2 may be the reducing/doping species upon thermal activation of CO 2 -DMI in chlorobenzene ( Figures S5, S6, Supporting Information). [60] We also note that the strong nucleophilic and electron-donating ability of carbene has been observed to result in doping by nucleophilic addition to the organic semiconductor in certain cases, and we cannot rule out a similar mechanism occurring here [36,61] To further confirm that the dimers are capable of doping OSCs we prepared chemically pure DMBI 2 under inert conditions and added as a solution in chlorobenzene to a solution of P(NDI2OD-T2). Accordingly, the resulting absorption spectrum ( Figure S7, Supporting Information) followed the same trends as previously described for P(NDI2OD-T2) doped with CO 2 -DMBI, with a suppression of the peak ≈700 nm and an increase in absorption ≈485 nm. Interestingly, this was only observed on heating, with little change observed at room temperature. This implies a dual role of the thermal activation step, which is necessary not only for the decarboxylation of dopant precursors, but also to overcome the activation energy related to the electron transfer. This is reminiscent of how reduction reactions with similar carbene dimers are typically conducted under elevated temperature to facilitate the electron transfer. [62] HOMO levels of the active dopant species are of great importance, since they may reveal the limitations of the dopants with respect to host OSCs. Cyclic voltammetry (CV) was performed to achieve an estimation of energy levels, in this case by measuring the reduction potential of the corresponding cationic salt, due to the high reactivity of the neutral dimer species. The reduction potential of the dihexafluorophosphate salt DMI 2 (PF 6 ) 2 dissolved in DMF has previously been reported as a two-electron reduction at −1.84 V against the ferrocene/ferrocenium couple (Fc/Fc + ), corresponding to an estimated DMI 2 2+ LUMO at −2.96 eV. [63] Measuring the CV of DMBI 2 (PF 6 ) 2 under equal conditions ( Figure S8, Supporting Information) reveals a two-electron reduction peak at −1.33 V against Fc/Fc + , indicating a lower DMBI 2 2+ LUMO at −3.47 eV. As these molecular orbitals correspond to the HOMOs of the neutral dimeric species, this may explain the lesser reducing effects of the CO 2 -DMBI dopant compared to CO 2 -DMI, as observed in the absorption spectra of doped OSCs. Either way, energy levels of both dopant dimers lie comfortably above the LUMOs of the doped OSCs, which range from −3.99 to −3.60 eV (Figure 3b), and thus should allow for efficient electron transfer. [64][65][66][67] The energy levels obtained from CV data are in good agreement with density functional theory (DFT) calculations performed on the B3LYP/DZVP level. Calculations of DMI 2 (PF 6 ) 2 and DMBI 2 (PF 6 ) 2 complexes (whereby the HOMOs of the DMI 2 and DMBI 2 moieties become unoccupied as their electrons are transferred to the PF 6 counterparts) predict LUMO energy levels of the cationic dimers at −3.10 and −3.54 eV respectively ( Figure S9a,b, Supporting Information), which lies within 0.15 eV of values obtained through CV. Calculations on isolated neutral dimers (in absence of PF 6 − ) resulted in slightly deeper HOMO levels at −3.46 and −3.90 eV for DMI 2 and DMBI 2 respectively (Figure 3c; Figure S9c,d, Supporting Information). This reduces the energy offset with respect to OSC LUMOs, with the DMBI 2 HOMO equal to or even slightly below the LUMOs of the studied OSCs. We note, however, that doping may still occur, if there is sufficient interaction between the dopant molecule and a proximal OSC unit. [29,37] As we will discuss shortly, this is indeed the case for the dopants and OSCs of interest in this work. Meanwhile, the calculated HOMO levels of both carboxylates and carbenes are very deep (−5.72 eV and lower, Figure S10, Supporting Information), ruling out spontaneous electron transfer from these species to the OSCs.
Further closed-shell DFT calculations on complexes between active dopant dimers and host OSCs show that DMI 2 interacts strongly with each OSC (so that the frontier HOMO orbitals of dopant-host complexes have contributions from both moieties), resulting in partial charge transfer from the dimer to the P(NDI2OD-T2), PCBM, or O-IDTBR, corresponding to 0.67, 0.30, and 0.61 electrons respectively (Figure 3d; Figure S11, Supporting Information). Interactions between DMBI 2 and P(NDI2OD-T2) resulted in a slightly lower partial charge transfer of 0.47 electrons. The unequivocal tendency of DMI 2 and DMBI 2 to donate electrons to the host OSCs has been verified with calculations employing a number of other basis sets, as detailed in Table S1 (Supporting Information). These findings confirm that both dimerized molecules act as n-type dopants for the respective OSC systems. Open-shell DFT calculations provide further evidence for electron transfer, showing that integer charge transfer of one electron from DMI 2 to either of the OSCs is energetically favorable over partial charge transfer by 0.05 to 0.40 eV. Furthermore, spin density plots obtained with constrained DFT calculations ( Figure S12, Supporting Information) show that the resulting unpaired electrons are delocalized on the dopant and host moieties.
In addition, the species formed through hybridization with DMI 2 are predicted to have HOMO values of −3.99, −3.79, and −3.89 eV for P(NDI2OD-T2), PCBM, and O-IDTBR respectively. As these values lie near the LUMO values of the pristine OSCs (Figure 3b), there is a possibility that the doping process may involve an additional charge transfer step between such charge transfer complexes and surrounding OSCs; this is particularly the case for DMI 2 , as the HOMO of the DMBI 2 /P(NDI2OD-T2) complex is predicted at a deeper −4.31 eV.

Organic Field-Effect Transistors
To evaluate the ability of the dopants to improve device performance past the limits of pristine OSCs, bottom-contact top-gate (BC-TG) OFETs were fabricated. P(NDI2OD-T2), PCBM, and O-IDTBR were selected as the semiconductor layers to explore the versatility of the proposed dopants. The selected materials cover a range of well-known n-type materials, including polymers, small molecules, and fullerene and non-fullerene species, with a variety of LUMO levels. Dopants were activated prior to spin-coating by heating the pre-mixed solution in chlorobenzene to 80 °C for 1 h. Figure 4 outlines representative transfer characteristics of devices made with each OSC and doped with various concentrations of either dopant as well as average mobility values at the respective doping levels. Furthermore, their output characteristics are plotted in Figures S13-S15 (Supporting Information) and all device parameters are outlined in Table S2 (Supporting Information). P(NDI2OD-T2) devices doped with CO 2 -DMI and CO 2 -DMBI follow a similar pattern, where optimal doping concentrations were found at 0.07 and 0.05 wt.% respectively, followed by a decline in electron mobility at higher concentrations. At their optimal configuration, average electron mobility increased from 0.17 to 0.26 and 0.31 cm 2 V −1 s −1 respectively, with maximum values at 0.30 and 0.40 cm 2 V −1 s −1 . As such, CO 2 -DMBI was proven as a more efficient device dopant for P(NDI2OD-T2), approximately doubling electron mobility at optimal conditions. Adv. Funct. Mater. 2023, 33, 2212305 The increase in device performance is also associated with a reduction of threshold voltage of ≈3.5 V in both cases, which is related to trap filling and increased electron density by the dopant addition. Upon 0.1 wt.% doping with either dopant, mobilities dropped below that of devices made with pristine P(NDI2OD-T2).
In PCBM devices, the two dopants seem to behave quite differently, most likely due to the ambipolarity of PCBM. Figure 4b shows the transfer characteristics from pristine and doped OFETs. In the case of CO 2 -DMI it can be seen that the p-type character is present even at high dopant loadings, whereas in the case of CO 2 -DMBI even small dopant addition results in suppression of the p-type character, allowing for higher n-type charge transport. At low dopant concentrations, CO 2 -DMI outperforms its counterpart, with an average electron mobility value of 0.010 cm 2 V −1 s −1 , almost three times as high as the pristine device at 0.0036 cm 2 V −1 s −1 . Meanwhile, for CO 2 -DMBI doped OFETs, mobility enhancement is achieved at higher dopant loadings. At 0.1 wt.%, the average device mobility was improved by almost one order of magnitude, reaching 0.033 cm 2 V −1 s −1 , with a maximum value of 0.045 cm 2 V −1 s −1 , and thus overall outperforming the CO 2 -DMI. Further increasing the amount of dopant resulted in a reduction of performance.
Finally, devices made with O-IDTBR also demonstrated an increase in electron mobility upon doping. Doping with CO 2 -DMI resulted in a subtle upswing in mobility upon 0.1 wt.% doping, and remained at a similar level also at 0.3 and 0.5 wt.%, reaching 0.42 cm 2 V −1 s −1 , compared to the original 0.35 cm 2 V −1 s −1 . Once again, CO 2 -DMBI resulted in a greater improvement, reaching an average mobility of 0.47 cm 2 V −1 s −1 at 0.05 wt.%, but with a gradual decrease at higher concentrations. In both cases a reduction in threshold voltage was observed when doping with the optimum amount (Table S2, Supporting Information). Both CO 2 -DMI and CO 2 -DMBI have been demonstrated to successfully improve electron charge carrier mobility of OFETs upon optimization of dopant concentration.

Thin Film Morphology
The transistor results suggest that the electron affinity of the dopant is not the only factor influencing device properties, and changes in the active layer morphology may also play a role. Recent results have suggested that even at low doping concentrations, as in the current study, doping can fundamentally impact crystallization and orientation of the semiconductor. [5,57,68] To investigate the effect of doping on morphology, atomic force microscopy (AFM) was used to map the topography of spun-cast thin-films of each dopant/OSC system (Figure 5a; Figures S16-S18, Supporting Information). Here, we expanded our range of dopant concentrations, from low to very high (0, 0.05, 0.1, 1, and 10 wt.%).
Pristine P(NDI2OD-T2) exhibits a clearly fibrillar surface structure, with a root mean square (RMS) roughness of 2.12 nm, similar to what has previously been reported. [69,70] Within dopant concentrations that are relevant in a device context, i.e., 0.1 wt.% and below, no major changes could be seen to this pattern, with RMS values remaining similar to the pristine film. However, at 1 wt.%, the fibrillar structure appears to disintegrate; this is more prominent in the case of CO 2 -DMI, while CO 2 -DMBI retains a partly fibrillar surface structure. This is also reflected by the RMS, reported at 1.50 nm when doping with 1 wt.% CO 2 -DMI, indicating a considerable flattening of the surface, compared to the lesser flattening of the CO 2 -DMBI doped film with an RMS of 1.90 nm. At 10 wt.% dopant concentration, the fibrillar surface structure completely disintegrated in both films.
While AFM images of PCBM films did not exhibit any obvious surface patterns -showing featureless and smooth surfaces throughout -the surface topography of pristine O-IDTBR films appeared crystalline, with clear boundaries between adjacent sheet-like crystallites. This surface structure was maintained also after doping with up to 0.1 wt.% CO 2 -DMI. At higher dopant concentrations significant change occurred, with the large platelets being replaced with a relatively featureless, smooth film (RMS = 0.295 nm, cf. ≥ 5.78 nm). In the case of CO 2 -DMBI doping, the crystalline surface morphology was maintained up to 1 wt.% doping, followed by a similar morphological transition upon addition of 10 wt.% dopant ( Figure S18, Supporting Information).
Another powerful tool to assess film morphology is X-ray diffraction (XRD). [71] To gain further insights into the morphological properties of doped OSCs, XRD measurements were conducted on films drop-cast onto silicon wafers (Figure 5b; Figure S19, Supporting Information). The diffraction pattern of pristine P(NDI2OD-T2) clearly displays the (200) and (300) lamellar peaks as well as the broad (010) π-stacking peak, in good agreement with literature, while the (100) peak is out of range. [72] The overall trend upon doping with CO 2 -DMI is that the (200) d-spacing gradually decreases past 0.05 wt.% dopant concentration from ≈12.4 to ≈11.8 Å at 1 wt.%, with a significant drop to ≈10.3 Å at 10 wt.%. Meanwhile, lamellar spacing does not change upon doping with CO 2 -DMBI until 10 wt.% concentration, resulting in a lesser drop from ≈12.4 to ≈11.1 Å. Furthermore, the lamellar peaks become more pronounced upon 10 wt.% doping, particularly in the case of CO 2 -DMI. No clear trends are observed in the (010) d-spacing, which remains at ≈3.9 Å up until 1 wt.% dopant. At 10 wt.%, however, there is no longer any indication of ordered π-stacking as peaks disappear in favor of a broad, seemingly amorphous, peak. This is unsurprising, considering the complete disruption of fibrillar surface morphology at such concentrations, as observed with AFM. As such, we concluded that the dopant disturbs the morphology, albeit at loadings beyond the optimized OFET concentrations.
In the case of PCBM films, the pristine pattern has many features in common with crystalline PCBM as reported in literature, although slightly broadened and less defined than a fully crystalline species. [73] This indicates the presence of nanoscale crystallites in an amorphous environment. This pattern is largely maintained upon doping with 0.05 wt.% of either dopant, but starts to fade upon doping at 0.1 wt.%, particularly with CO 2 -DMI. Upon doping with 1 wt.% CO 2 -DMI, only two broad peaks remain, indicating an amorphous structure, while some crystalline features still remain upon doping with 1 wt.% CO 2 -DMBI. [74] These features finally disappear at 10 wt.%. Likewise, pristine O-IDTBR exhibits crystalline features, again in agreement with previously reported diffractograms. [75] The crystalline structure is seemingly unaffected by addition of CO 2 -DMI up to 0.1 wt.%, and of CO 2 -DMBI up to 1 wt.%. At higher concentrations, all peaks disappear, indicating a complete disruption of the crystalline structure, in agreement with the AFM results.

Conclusion
In summary, CO 2 -DMI and CO 2 -DMBI have been demonstrated as effective and versatile dopants of n-type OSCs. While the dopant precursors remain stable and unreactive at room temperature -even under ambient conditions -UV-vis and EPR measurements of doped films and solutions indicate efficient electron transfer from the dopant to the host OSC upon in situ thermal activation in solution prior to film formation. Experimental and computational studies suggest the formation of dimeric species upon decarboxylation of the dopant precursors, which act as the active dopant species. Device data of doped OFETs employing P(NDI2OD-T2), PCBM, and O-IDTBR as the semiconductor layer shows an improvement of electron mobility for all materials with both dopants. Throughout, devices doped with CO 2 -DMBI perform better than the corresponding devices doped with CO 2 -DMI, with CO 2 -DMBI doped PCBM devices exhibiting a 10-fold increase in electron mobility. AFM and XRD studies indicate that CO 2 -DMBI has a lesser impact on the thin-film morphology of the investigated OSC systems, although the extent of these effects is not obvious at device-level dopant concentrations. Overall, this work highlights carboxylation as an effective design tool toward air-stable dopant precursors, which can be thermally activated in situ through thermal decarboxylation to enhance electron mobility of a wide range of n-type OSCs.

Experimental Section
A full description of experimental procedures is provided in the Supporting Information.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.