Letter reSeArCH
which in turn depends on ionic interactions, but also on statistics—that
is, on entropy. Here, working under the assumption that the energetic
gain via anion exchange is established to some extent by the entropy
factor, we assessed the role of these ionic interactions by comparing
combinations of eight cations, X+, and six anions, Y−, in conjunction
with systematic variations in the effective ion radius, Reff, that sepa-
rates the charges. The correlation between the sizes of the molecular
ions having unique shapes and the associated ionic interactions can be
determined from spatial maps of the electrostatic potentials based on
van der Waals surfaces (Fig. 2 ; see Supplementary Information section
1.4). The results of density functional theory (DFT) calculations show
that smaller ions have higher electrostatic potentials on their surfaces.
We discuss here the roles of ionic interactions on the basis of experi-
mental observations, which demonstrate that the anion-exchange effi-
ciency (F4TCNQ•− → Y−) and doping concentration both correlate
with the strength of the ionic interaction. Specifically, the highest dop-
ing concentration and most efficient anion exchange are realized when
using a salt composed of small cations and large anions, which possess
high and low electrostatic surface potentials, respectively.
Initially, we assessed the manner in which the electrostatic poten-
tial of the ionic liquid anion Y− affects the anion-exchange doping
by considering ionic liquids with four different anions. These anions
were tetrafluoroborate (BF 4 −), hexafluorophosphate (PF 6 −), tris(pen-
tafluoroethyl)trifluorophosphate (FAP−) and TFSI−. In each case, the
associated cation was 1-butyl-3-methylimidazolium (BMIM+). Anion-
exchange doping was performed by immersing PBTTT thin films in a
solution of F4TCNQ dissolved in each ionic liquid for 10 min at 60 °C.
Figure 3a presents the optical absorption spectra obtained from PBTTT
thin films doped via anion exchange with the four different anions.
The F4TCNQ•− doublet at 775 nm and 881 nm is observed only for
Y− = BF 4 − and PF 6 −, and this result is indicative of inefficient anionexchange. Interestingly, although smaller anions (such as BF 4 − and
PF 6 −) should be more mobile than larger anions (FAP− and TFSI−), the
anion-exchange efficiency shows the opposite trend. This clearly sug-
gests that the present anion exchange proceeds only when the exchange
lowers the free energy of the system, where the gain of Gibbs free energy
is defined as Δex. The results of DFT calculations also show that anion
exchange with TFSI− is more energetically favourable than that with
BF 4 − (see Supplementary Information section 1.5). This binding pref-
erence can be understood by considering both the size and the shape
of the ions, because delocalized anions such as TFSI− will preferentially
couple with the delocalized charges on the PBTTT.
We further assessed the mechanism associated with the anion-
exchange doping here, which is based on the energetic gain Δex, by
investigating the effect of the cation. These trials used three different
cations: Li+, EMIM+ and methyltributylphosphonium (MtBPho+). In
each case, Y− was fixed as TFSI−. The cation does not directly interact
with the PBTTT, but rather can be considered as a spectator ion in
the anion-exchange process that nevertheless affects the Δex value by
forming ion pairs. Assessing the disappearance of the F4TCNQ•− dou-
blet shows that only MtBPho+ did not permit efficient anion exchange
(Fig. 3b; see Extended Data Fig. 5 for additional data). One possible
explanation for the beneficial effects of a small cation is that the initial
ion pair must be composed of ions with poor affinity for one another
(for example, a small cation and large anion) so as to produce a large
Δex. This explanation is in good agreement with the empirical hard and
soft acid and base (HSAB) theory, which is often invoked to explain
ion-exchange tendencies^13.
The optimization of X + and Y− is of great importance, not only
because it affects the anion-exchange efficiency, but also because of
its effect on the doping level, as verified by conductivity measure-
ments. During these analyses, we doped PBTTT thin films via anion0.40.30.20.10.0Absorbance500 1,000 1,50 0 2,000 2,500
Wavelength (nm)Pristine
F4TCNQ doped
Anion exchange
with EMIM-TFSIAbsorbance (a.u.)2,050 2,100 2,15 0 2,20 0 2,250
Wavenumber (cm–1)PristineF4TCNQ dopedCŁN stretchingAnion exchange with EMIM-TFSIPBTTT + F4TCNQ + X+ + Y– ĺ [PBTTT•+ F4TCNQ•–] + X+ + Y–
[PBTTT•+ F4TCNQ•–]^ + X+ + Y– ĺ [PBTTT•+ Y–] + X+ + F4TCNQ•–i
ii, iiiiii iiiPBTTT + F4TCNQ ĺ [PBTTT•+ F4TCNQ•–]Conventional molecular dopingAnion-exchange dopingX+ Y–Dopant-dissolved ionic liquid
F4TCNQHoleHole
PBTTT lm Charge transferF4TCNQ•–F4TCNQ X+ Y–HoleHole
PBTTT lmF4TCNQ•–Dopant-dissolved solution
F4TCNQHoleHole
PBTTT lm Charge transferF4TCNQ•–Anion exchangeF4TCNQ X+HoleHole
PBTTT lmF4TCNQ•–Y–***PBTTT F4TCNQa
b
cd
F FF FNNNNSC 14 H 29SSSC 14 H 29nFig. 1 | Summary of anion-exchange
doping. a, b, Diagrams showing conventional
molecular doping (a) and anion-exchange
doping (b). The inset at the top right shows
the chemical structures of PBTTT and
F4TCNQ. In anion-exchange doping, a thin
film of PBTTT is doped with the initiator
molecule F4TCNQ (i) via a charge-transfer
interaction. ii, The F4TCNQ radical anions
are replaced by the Y− anions (where Y is
TFSI) of the ionic liquid. iii, This anion
exchange (F4TCNQ•− → TFSI−) results in
the formation of a solid-state donor–acceptor
complex ([PBTTT•+ TFSI−]). c, d, Optical
absorbance (c) and FTIR spectra (d) of
pristine PBTTT (black), F4TCNQ-doped
PBTTT (orange), and PBTTT doped via
anion exchange (blue). The centre value of the
peak marked with a single asterisk is 553 nm.
The centre values of the doublet peak (double
asterisk) are 775 nm and 881 nm.29 AUGUSt 2019 | VOL 572 | NAtUre | 635