1. Introduction
Thin-film organic materials have become increasingly important as the active electronic medium in a wide range of large-area imaging and display devices. An important class of such organic materials are molecularly doped polymers (MDPs) which comprise an electronically inert polymer binder into which an electrically active dopant molecule is dispersed. Careful chemical selection of the polymer-dopant combination, and optimisation of the dopant concentration, permits the benchmark mobility-lifetime (μτ) product to be sensitively controlled for the majority electronic carrier according to the target application. The μτ product has historically been recognized as a significant controlling parameter for the achievable efficiency of thin-film inorganic solar cells [
1,
2], and more recently in the development of optimized thin-film organic solar cells [
3] and large-volume crystalline detectors for ionizing radiation [
4]. The long-term stability of such μτ magnitudes in MDPs may be seriously compromised, however, via unwanted internal chemical reactions. Such reactions are generally induced by internal photo-excitation processes and are significantly more damaging for the dopant molecules which are subsequently rendered to become electronically inactive.
Mitigation of photo-degradation processes in MDP thin-films consequently demands that a comprehensive understanding of the underlying processes and parameters that are dominant in driving these internal processes are fully identified and numerated. Previous work in this area has focused upon a specific photo-cyclic degradation process that is known to occur for a small hydrazone molecule p-diethylaminobenzaldehyde -1,1′-diphenylhydrazone (DEH) in MDP thin films [
5,
6,
7,
8]. The photo-cyclic conversion of DEH to the electronically inactive indazole (IND) product requires that the DEH molecule be first excited into a (reactive) state by appropriate absorption of UV radiation, and that excited DEH molecules have a supply of absorbed oxygen to then complete the reactive transition to IND. Modelling of the conversion dynamics of DEH to IND has previously been undertaken using rate-equation [
7] and cellular-automata (CA) approaches [
8]. From these studies it was found that the conversion dynamics of the hydrazone molecular dopant to indazole photo-product (
P) could be analytically described as a function of photo-exposure time (
t) by a stretched-exponential function such that:
In Equation (1) the fitting parameters
Γ and
γ are expected to be close to unity [
8] provided the concentration of dopant (c
M) is kept sufficiently low (<20% by weight relative to the polymer). The other parameter
τ in Equation (1) represents a characteristic time constant given by
τ = c
M/c
Oβ where c
O represents the concentration of soluble oxygen, and β gives the molecular relaxation rate of reactive DEH back to the (un-reactive) ground state. It is noted that short degradation time’s
τ may therefore be facilitated by high levels of soluble oxygen c
O within the MDP film. Polymer morphologies which facilitate high soluble oxygen levels, such as those that contain micro-porous channels, should accordingly be avoided when identifying potential dopant-polymer pairs for MDP applications.
A consequence of the dynamical increase in IND photo-product with exposure time according to Equation (1) is that the MDP films are expected to photo-bleach if, relative to the absorption strength of the DEH molecule, the IND molecule is relatively transparent at the selected UV excitation wavelength [
6]. A simple experimental methodology to investigate the conversion dynamics is therefore presented by selecting an appropriate UV exposure wavelength which may serve the dual function of initiating the photo-excitation process and providing a reference transmission signal through the film to determine the optical absorbance. As noted from previous cellular-automata simulations of the absorbance dynamics [
8] the overall change in absorbance will then be determined by the changing proportion of DEH to IND as photo-exposure progress, and will be sensitive to the relative magnitude of the DEH and IND molecular absorption coefficients for the film thickness used. The proposed absorption approach may consequently offer a simple, affordable approach to investigate the porosity of thin polymer films which may be inaccessible using traditional bulk material methods [
9,
10,
11] or which require the use of free-standing films [
10,
11,
12]. Experimental absorption data to test these model predictions have now been acquired and are presented in the present work. In particular the potential use of the technique to evaluate and identify thin-film polymers which may contain micro-porous regions, and would thus be unsuitable for MDP consideration, is evaluated.
4. Discussion
The significant red-shift observed in
Figure 2 for the unexposed PE MDP spectra is consistent with the DEH excitation (i) and relaxation (ii) processes depicted in
Figure 1. In the unexposed state these MDP films are initially strongly absorbing at the UV wavelength and virtually none of this incident radiation is therefore transmitted to the spectrometer. Sufficiently short exposure times of typically <10s are adequate to permit the spectra to be experimentally recorded, yet ensure that the majority of the excited DEH molecules subsequently relax to their original ground state via process (ii) in preference to reacting with soluble oxygen to form IND (processes (iii) and (iv) in
Figure 1). The relaxation process (ii) will consequently emit a photon but this is expected to be at a lower energy as the excited electrons thermalise to the lowest energy states within the DEH LUMO band. For MDP materials the LUMO and HOMO Gaussian bandwidths are expected to increase with the dopant concentration due to random dipole disorder [
8,
14]. For the DEH molecule which possesses a dipole moment ~2.1 debye this results in a bandwidth of around 0.11 eV for c
M = 50% [
14]. For energetically symmetric LUMO and HOMO bands the net reduction in emitted photon energy during process (ii) is therefore expected to be around 0.22 eV with an associated red shift of about 30 nm relative to the incident UV peak. This is quantitatively consistent with the data in
Figure 2 where the peak for the c
M = 50% occurs at about 419 nm. The spectral red shift is furthermore observed to become progressively smaller as c
M is reduced which is expected due to the associated dipole disorder narrowing of the HOMO and LUMO bandwidths.
For fully exposed films the spectral response is anticipated to be significantly different due to the conversion of DEH to IND. This is confirmed for the lightly doped PE MDP specimens in
Figure 3 where spectra for the fully exposed and unexposed films are compared. In contrast to the red-shifted spectra recorded at the start of the exposure period, the fully exposed films are now observed to have a spectral signature that matches the incident UV profile. This confirms that the films are now entirely transparent to the incident UV radiation and have therefore become UV bleached due to the replacement of strongly absorbing DEH molecules with the non-absorbing IND photo product.
The spectral phenomena observed in
Figure 2 and
Figure 3 suggest that it should therefore be possible to dynamically monitor the DEH to IND photo-conversion process by either recording the intensity of the emitted red-shifted radiation or the amount of transmitted UV LED light. Consideration of the spectrometer’s signal gain requirements when operated in a kinetic acquisition mode indicated that monitoring of the transmitted UV signal was more reliable, however, and permitted a greater range of film absorbance to be monitored across the maximum exposure period. The absorbance data presented in
Figure 4 and
Figure 5 have consequently been determined using this reference UV signal approach.
As already noted for the
Figure 4 data, the recorded change in absorbance was found to be highly reproducible across randomly selected regions of the film. The overall reduction in absorbance for this specimen corresponds to an increase by a factor of around 160 in received UV light at the spectrometer once exposure is completed. Such a large increase in signal proved challenging to process using the spectrometer’s fixed-gain setting in the kinetics mode as the gain required at the start of the exposure period to reliably establish a reference baseline would sometimes result in detector saturation before full-exposure had been completed. To experimentally mitigate this issue specimens were latterly prepared with a reduced target thickness of around 1 μm so that the starting reference signals could be obtained using a minimal spectrometer gain. It is consequently suggested that a bespoke experimental setup that comprises a UV-sensitive photodiode operating with a trans-impedance amplifier would provide an optimised setup to record this type of dynamical absorbance data. Evaluation of such a proto-type system using a micro-controller and digital potentiometers in the feedback loop to optimise the signal gain has provided similar absorbance results to those obtained from the spectrometer unit. The overall cost of such a bespoke system is presently estimated to be around 5–10% of the spectrometer’s listed purchase price.
Whilst the absolute absorbance change over the exposure period presents many interesting experimental challenges, the associated timescale and temporal signature for the change contains potentially important information about the underlying polymer morphology. The overall change in the film absorbance (
δA) with time may be functionally expressed as
δA(t) = −
log[I(
t)/I
0] where I
0 and I(
t) are, respectively, the intensities of UV light received at the spectrometer at the start (
t = 0) of the exposure period and after an exposure period t has elapsed. Assuming a simple de Beer’s law dependence for UV absorption through the film thickness permits
δA(
t) to then be expressed in terms of the film absorption coefficient (
α) as:
In Equation (2)
L is again the film thickness, and
α0 and
α(
t), respectively, represent the corresponding film absorption coefficients at the start of the exposure period, and following an interval t of UV exposure. The overall absorption coefficient for the film is entirely controlled by specific UV absorption events by DEH and IND molecules, however, and may consequently be expressed in terms of the associated absorption coefficients for these respective molecular species (
αDEH and
αIND). If these molecular absorption events occur independently then the overall film absorption coefficient may be estimated from the weighted proportion of DEH to IND that exists in the film at any time during exposure. From Equation (1) it follows that
α(
t) = [1 −
P(
t)]
αDEH +
P(
t)
αIND and using this in Equation (2) finally yields the following expression for the absorbance dynamics:
It is noted from Equation (3) that since P(t) increases with exposure time, then the film absorbance is expected to decrease (UV bleaching) provided αDEH > αIND. The maximum relative change in absorbance, which is acquired after full exposure when P(t) → Γ, will then be dictated by the film thickness L. The selection of an appropriate film thickness is therefore expected to be key in achieving an appropriate response range for δA but whilst it is tempting to use thicker films this may have experimental consequences as already noted in the preceding discussion concerning the spectrometer gain.
The ability of Equation (3) to fit experimental data is demonstrated in
Figure 4 for the 2.5 μm thick polyester film where c
M = 10%. The optimum fit as shown by the open symbols returns values of
Γ = 1.0 and
γ = 0.9 in accordance with the low-doping expectations previously noted for the
P(
t) function [
8]. The characteristic time constant is then
τ = 2 × 10
4 s, whilst the overall change in the absorbance for the nominal film thickness requires that
αIND/
αDEH = 0.1 using a value of
αDEH = 3 × 10
4 cm
−1 determined from optical transmission measurements. Whilst there is a reasonable level of confidence in the fitted parameters (
Γ,
γ and
τ) that control the temporal shape of the absorbance change, it is evident that uncertainties associated with the film thickness (ΔL in
Table 1) and the appropriate magnitude for
αDEH may result in unacceptably large values for
αIND being returned to account for the net change in absorbance. A constraint that
αIND/
αDEH should remain sufficiently low (≤0.1) during the fitting procedure was consequently applied to comply with the previously noted observation [
6] that the IND molecule is optically “transparent” at the incident UV wavelength. This additional constraint on the fitting procedure was not, however, found to significantly affect the emergent
P(
t) parameters (
Γ,
γ and
τ). The further application of Equation (3) to fit experimental data in a more heavily doped polyester film (c
M = 20%) is shown if
Figure 5 where comparable parameters are found for
Γ and
γ and a similar value of
τ = 2 × 10
4 s is extracted. The similarity of these time constants does not initially appear to be consistent with the CA expression
τ = c
M/c
Oβ where it is anticipated that under identical experimental conditions the required photo-degradation time will scale linearly with the amount of DEH dopant provided c
M < 40% [
8]. As the film thicknesses are significantly different for the PE data presented in
Figure 4 and
Figure 5, the simple ×2 scaling factor predicted from c
M considerations alone is therefore not observed (there is simply more DEH present in the thicker c
M = 10% film which requires a longer time to degrade under equivalent UV exposure conditions). The expected time constant scaling with c
M is, however, confirmed for the PE films in
Table 1 which have comparable thicknesses. Here, the time constant values for the c
M = 10% and c
M = 20% PE films have been entered for convenience under the column headed
τs for reasons that are described below.
It is evident that for the other data presented in
Figure 5 for the polystyrene and polycarbonate films Equation (3) will no longer provide an appropriate description of the absorbance dynamics as these films appear to exhibit a 2-stage reduction in
δA that occur at distinctive times. Compared to the single-stage polyester response, the initial fall of
δA for these films occur at significantly shorter times which suggests that fast conversion to IND occurs at a proportion of the DEH sites (stage 1) before a slower conversion of the remaining DEH sites is detected at later times (stage 2). An obvious modification to Equation (3) to permit this 2-stage process to be analytically modelled is to simply replace the
P(
t) element from Equation (1) with an equivalent weighted expression for the fast and slow DEH components. Assuming that the degradation processes within these DEH components are independent the modified replacement for
P(
t) would then simply be:
In Equation (4) the parameter
φf represents the proportion of DEH sites which may undergo fast degradation, and
τf and
τs denote the associated time constants for the fast and slow reaction sites, respectively. The ability of Equation (4) to fit the c
M = 20% polystyrene data is illustrated in
Figure 5 where the proportion of DEH sites that undergo fast degradation is found to be around 80%. Similar fitting procedures applied to the c
M = 10% polystyrene film, and equivalently doped polycarbonate specimens, return broadly similar proportions of fast to slow DEH sites as evidenced in
Table 1. The associated time constants for the fast reaction sites is furthermore found to be about two orders of magnitude lower than for the slow sites for both of these polymer binders. The single time constant extracted for the polyester films falls between these magnitudes although closer to the slow values and these parameters have consequently been entered under the
τs column in
Table 1 for convenience with
φf consequently set to zero for this particular binder.
It is interesting to consider the physical origin of the emergence of the significantly faster degradation process in the polystyrene and polycarbonate films which is absent in the polyester samples. As noted in the introduction, the signature time-constant
τ = c
M/c
Oβ so that for equivalently doped films the shorter time-constant implies that either the concentration c
O of soluble oxygen is enhanced at a fraction (
φf) of the DEH sites and/or the relaxation rates β for this subset of excited DEH sites is increased [
8]. Similar spectral responses to those presented for the polyester films in
Figure 2 and
Figure 3 are also found for the polystyrene and polycarbonate samples, however, which suggests that the relaxation of excited DEH sites is principally controlled by direct transitions between the molecular HOMO and LUMO states. The influence of the polymer in the relaxation process is therefore likely to be minimal (other than to permit limited thermalisation within the HOMO) and the β parameter is consequently likely to be a specific characteristic of the DEH molecule and polymer independent. Any influence of the polymer binder upon
τ is thus inferred to arise mainly through the available concentration of soluble oxygen to complete the photo-cyclic conversion to IND [
5].
The presence of two distinctive degradation stages in the polystyrene and polycarbonate samples would then appear to indicate that the concentration of soluble oxygen throughout the polymer is non-uniform for these samples with regions of higher c
O resulting in the faster degradation process. Film morphology properties that may explain the presence of locally enhanced oxygen levels include surface topography characteristics (roughness and associated porosity) [
15,
16], and the presence of micro-pores within the bulk of the polymer binder [
9,
17]. However, the relatively large value of 0.8 deduced for the
φf parameter is inconsistent with enhanced degradation being entirely confined to a small fraction of the film thickness in the vicinity of the film surface (assuming uniform dispersal of the DEH dopant throughout the film bulk). It is furthermore noted that, for the thinner ~1 μm films in
Table 1, the film roughness ΔL is around 20% of the film thickness for all of the polymer binders. A more likely explanation for the PS and PC behaviour is that these particular polymers contain a significant network of micro-pores throughout the film bulk that are able to maintain an enhanced concentration of oxygen across the majority (~80%) of the DEH sites. A minority (~20%) of DEH sites in these films are presumably more isolated from the micro-pores and thus degrade at a significantly slower rate. The PE samples must by contrast possess a more uniform morphology throughout the entire film bulk which yields the single degradation rate response. The
τ magnitude found for the PE films may accordingly indicate that the underlying bulk morphology for this binder is entirely devoid of micro-pores.
Whilst the preceding arguments offer some insight into how the observed degradation dynamics may be controlled via the various MDP binder morphologies it is still important to confirm that these processes proceed independently and that Equation (4) may therefore be applied with confidence. Further cellular-automata modelling of the degradation dynamics has accordingly been undertaken in which the original CA rule-set [
8] was modified to now permit a pre-defined fraction (
φf) of excited DEH cells to convert to IND cells with an enhanced probability. Such CA cells therefore constitute fast reaction sites within the MDP bulk that have an enhanced oxygen concentration supplied by micro-pores. It is found that the output generated from this modified CA approach may indeed be functionally described by Equation (4) and that the parameters returned (
φf,
τf and
τs) are consistent with those employed by the CA rule-set (
φf, c
O and β).