- About this Journal ·
- Abstracting and Indexing ·
- Aims and Scope ·
- Article Processing Charges ·
- Articles in Press ·
- Author Guidelines ·
- Bibliographic Information ·
- Citations to this Journal ·
- Contact Information ·
- Editorial Board ·
- Editorial Workflow ·
- Free eTOC Alerts ·
- Publication Ethics ·
- Reviewers Acknowledgment ·
- Submit a Manuscript ·
- Subscription Information ·
- Table of Contents

Journal of Spectroscopy

Volume 2013 (2013), Article ID 346826, 16 pages

http://dx.doi.org/10.1155/2013/346826

## The Kinetics of Joined Action of Triplet-Triplet Annihilation and First-Order Decay of Molecules in T_{1} State in the Case of Nondominant First-Order Process: The Kinetic Model in the Case of Spatially Periodic Excitation

^{1}Spectroscopy and Photochemical Kinetics Department, Max Planck Institute for Biophysical Chemistry, Am Fassberg 11, 37077 Göttingen, Germany^{2}Department of Photochemistry and Spectroscopy, Institute of Physical Chemistry, Polish Academy of Sciences, Kasprzaka 44/52, 01-224 Warsaw, Poland^{3}Department of Characterisation of Nanoelectronic Structures, Institute of Electron Technology, Al. Lotników 32/46, 02-668 Warsaw, Poland

Received 24 June 2012; Accepted 23 August 2012

Academic Editor: Gianfranco Giubileo

Copyright © 2013 Paweł Borowicz and Bernhard Nickel. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

#### Abstract

In this paper the model developed for estimation of the diffusion coefficient of the molecules in the triplet state is presented. The model is based on the intuitive modification of the Smoluchowski equation for the time-dependent rate parameter. Since the sample is irradiated with the spatially periodic pattern nonexponential effects can be expected in the areas of the constructive interference of the exciting laser beams. This nonexponential effects introduce changes in the observed kinetics of the diffusion-controlled triplet-triplet annihilation. Due to irradiation with so-called long excitation pulse these non-exponential effects are very weak, so they can be described with introducing very simple correction to the kinetic model described in the first paper of this series. The values of diffusion coefficient of anthracene are used to calculate the annihilation radius from the data for spatially homogeneous excitation.

#### 1. Introduction

The technique of measuring the diffusion coefficient using spatially periodic excitation was first developed by Avakian and Merrifield [1], Ern et al. [2], and Ern [3]. In those papers the authors used a pattern that consists of parallel slits to create the spatially inhomogeneous distribution of the measured species, namely, triplet excitons in anthracene crystals. The technique was modified by Nickel [4], instead of the pattern with parallel slits the interference of two laser beams was used to create spatially periodic distribution of the density of the molecules in the T_{1} state. In this way the drawback of original Avakian’s method, the diffraction of the light on the slits was avoided. This method was applied in measurements of the diffusion coefficient of different organic molecules [5, 6].

The main aim of this work is to take into account the effect of the joined action of the first-order decay of the molecules in the triplet state and diffusion-controlled triplet-triplet annihilation (TTA) in the kinetic model that takes into account the effects caused by spatially periodic excitation. The starting points are the intuitive treatment of the spatially homogeneous excitation [7] and the Nickel’s model of TTA with spatially periodic excitation [6]. The main aim of the present approach is the introducing of the joined action of TTA and first-order decay to the term describing the kinetics of the TTA for inhomogeneous excitation—in the *diffusional relaxation* term. The derived model was applied for the evaluation of the diffusion coefficient of anthracene in the T_{1} state. The values obtained with the introduced model are compared with the data obtained from the Nickel’s model developed in the framework of standard Smoluchowski approach.

The temperature and viscosity dependence of diffusion coefficient is another problem discussed in this paper. The hydrodynamic theory of diffusion predicts the linear dependence between the diffusion coefficient and the ratio —Stokes-Einstein (SE) model [8]. At the other side the hole theory of diffusion [9] predicts that viscosity and diffusion constant vary exponentially with the temperature. In this model diffusion is described as a process that has an activation barrier. To allow the molecule to diffuse a free space (hole) must be created in the solute. The energy that is necessary for the creation of this hole is described in terms of the activation barrier. In the hole model the diffusion coefficient tends to the limited value when temperature tends to infinity, whereas in the SE model the diffusion coefficient tends to infinity with the increase of temperature. The Stokes theory of diffusion was modified by Gierer and Wirtz [8, 10]. Instead of the continuous medium they took into account the finite dimensions of the solvent and solute molecules. They introduced to the SE model a correction which is a function of the radii of solvent and solute species.

In this work we postulate the equation for the diffusion coefficient as dependent on the which combines the finite limit of the diffusion constant for large values of as in the hole theory of diffusion and the linear dependence of the diffusion coefficient for small as predicted by SE equation.

#### 2. Description of the Kinetic Model

##### 2.1. Temporal Behavior of Delayed Fluorescence after Spatially Periodic Excitation

The kinetic equation derived for spatially periodic excitation within framework of standard Smoluchowski model is the start position of the discussion. In the case of strongly dominant first-order decay: and for the spatial period the equation for the delayed fluorescence intensity can be written in the following way [6]:
where is the constant, is annihilation radius, is mutual diffusion coefficient, is the first-order rate constant, is the density of the molecules in T_{1} state at the time , and is the normalized parameter specifying the degree of interference. Its value varies from 0—no interference to 1—complete interference. The parameter (*diffusional relaxation constant*) describes the diffusional relaxation resulting from inhomogeneous (here spatially periodic) distribution of the molecules in the T_{1} state at and is a function of the spatial period of excitation and the absolute diffusion coefficient (in the case of TTA ):
Equation (1) can be divided into two parts:

The first one describes the decay of the molecules in the T_{1} state after spatially homogeneous excitation in standard Smoluchowski approach:
where the convention of the indications () is the same as in the paper [11] it means is the time-dependent rate parameters in the Smoluchowski original model. The modification of the “homogeneous” kinetics was described previously [7] for the case of strongly dominant first-order decay and extended to nondominant first-order decay in the first paper of this series [12].

The other term introduces the changes in the kinetics caused by spatially periodic excitation, namely, describes the diffusional relaxation of the inhomogeneous initial distribution of the molecules in the triplet state:
This relaxation modifies the kinetics of homogeneous distribution of the molecules in the triplet state. Since TTA do not contribute directly to this term the changes in time-dependent intensity of the delayed fluorescence are caused only by the average decay of the concentration of the molecules in T_{1} state in the sample. The changes in the local concentration involve the changes in the efficiency of the TTA in different areas of the sample. The lack of the direct contribution of the TTA to the diffusional distribution manifest itself by the absolute not relative diffusion coefficient appearing in the term .

The diffusional relaxation will be influenced by the first-order decay which decrease homogeneously the concentration of molecules in the triplet state. The diffusional relaxation constant depends on the gradient in the distribution of molecules in the T_{1} state between areas of the highest and the lowest concentration of the reacting species. This gradient is built by spatially periodic excitation, so one of the factors influencing the parameter is the spatial period . The mobility of the molecules is represented by absolute diffusion coefficient . However, due to homogeneous decay of the molecules in the T_{1} state via first-order decay the initially built gradient will change with the time by joined action of the redistribution and the first-order decay. We assume that the first-order decay can be treated as dominant and the diffusional relaxation is the function of the mobility of molecules. That means that homogeneous decay do not influence the redistribution process in other way as changing the gradient via homogeneous decrease of the concentration of the molecules in the T_{1} state. In such a case the parameter introduced as diffusional relaxation constant should be corrected by time-dependent factor. Taking into account the independence of first-order decay and the redistribution of the molecules one can represent the diffusional relaxation in terms of joined action of the first-order decay and the diffusional redistribution by introducing the time-dependent parameter instead of in the periodic term:
So intuitively the modified equation for the time-dependent intensity of delayed fluorescence should have the form in the case of dominant first-order decay:

There is also one more point in the model that should be discussed. Since, as mentioned before, in different places in the sample one has different intensity of the exciting laser light it cannot be excluded that in the areas with higher exciting intensity small components of nonexponential behavior of the sample exist. Due to irradiation with so-called long excitation pulse, where the intensity is so small that by homogeneous excitation the nonexponential behavior of the kinetics can be generally avoided, we expect to have the nonexponential behavior only in the part directly related to the TTA, it means in homogeneous term of . The most sensitive part in in the context of nonexponential behavior is , because in the changes of contribute linearly to the time-dependent rate parameter. So the most simple way to test the nonexponential behavior of the spatially periodic decay is to use the following modification of the model:(1)the time-dependent rate parameter has the form as in the case of the intuitively modified model with first-order decay to be dominant:
(2)the density of the molecules the T_{1} state has the nonexponential form:
where donate ,(3)the periodic part of is the same as in (5).

The density described in point 2 has the form as in (4) in the first paper of this series [12], but with as time-dependent rate parameter defined by (7).

To sum up: the time-dependent intensity of the delayed fluorescence for nonexponential decay with spatially periodic excitation should have the form:

In the temperature above 150 K the contribution of the short time effect to the intensity of the delayed fluorescence decreases very fast. If the spatial period of the irradiation is enough large the diffusion relaxation resulting from inhomogeneous excitation takes place mainly for so large values of the delay that the contribution of the nonstationary part can be neglected. Since it takes place generally at the temperatures, where the diffusion coefficient is large the nonexponential approximation equation (9) seems to be the proper choice in this case. However due to negligible contribution of the short-time effect the equation can be simplified by neglecting the contribution of the nonstationary term. Neglecting the ratio: one obtains the simplified dependence of for spatially periodic excitation in the case of nonnegligible contribution of nonexponential effect to the kinetics:

##### 2.2. Temperature and Viscosity Dependence of Diffusion Coefficient

The basic model of the temperature and viscosity dependence of the diffusion constant, SE model, assumes that the particle moves in continuous medium. The diffusion coefficient is a linear function of the ratio temperature to viscosity : where is the Boltzmann constant and is the radius of the particle. Viscosity is a function of temperature and it decreases with the increase of . In SE model if and if . The lower limit of is reasonable: with the viscosity , so the mobility of the molecules should be very small. The other limit: seems to be artificial. This limit suggests that at very high temperatures the process of diffusion should be infinitively fast, it means the infinitively large mass should diffuse through finite area within the finite time period. At the other side the hole model of the diffusion postulate the diffusion coefficient to tend to finite limit and to be exponentially dependent on the temperature: In this model the process of diffusion of solute molecule is described as series of jumps between equilibrium positions of the solute molecule. The equilibrium positions are separated by barriers. This equilibriums result from the finite (nonzero) dimensions of the solvent and solute molecules. The origin of the barrier is to create of the “hole” among the solvent molecules. This “hole” is the free space where the solute molecule can migrate. The energy is the free energy of the barrier which corresponds to the process of creating a hole in the solvent. Since the energy depends on viscosity, the diffusion coefficient also depends on even if it is not written explicit in (12). The hole model of diffusion takes into account the corpuscular nature of the solvent.

Since measured values of diffusion coefficient show nonlinear dependence as a function of and the course of the functions seems to tend to finite limit (see Section 4) it would be sensible to postulate an equation for satisfying following conditions:(1) for ;(2)for small the new equation should be approximated by SE model;(3) for .

We postulate the equation in the form:
The model defined by (13) will be named as *combined model* because it has as an argument of the function as in the SE model and has the exponential dependence of the diffusion coefficient from the independent variable and finite limit of the diffusion constant for high temperatures as in the hole model.

Since the difference equals to 0 for , ; for and the expansion of (13) in Taylor series around gives From (11) and (14) one can obtain The finite limit of the diffusion coefficient when the value of the argument of function tending to infinity is the great advantage of the combined one. This kind of the dependence of the diffusion coefficient shows that the mass transport tends to saturation at very high temperatures. This kind of behavior, namely, the saturation of the efficiency of the process for conditions tending to the limit is observed in the case of other transport processes. The example of the setup where the saturation is observed is charge transport in structures like Junction Field-Effect Transistors (JFET) or Metal Semiconductor Field-Effect Transistors (MESFET) [13]. The saturation effect in the electronic structures is observed for drain current for JEFT and MESFET structures [13] and as well for accumulation-mode (AM) -channel silicon-on-insulator (SOI) metal-oxide-semiconductor-fiels-effect-transistor (MOSFET) or enhancement-mode (EM) -channel SOI MOSFET [14]. The saturation effect in the case of electronic devices is so important that the models of charge transport developed for structures like SOI -type MOSFET [15] or high electron mobility transistors (HEMTs) based on AlGaN/GaN [16] structures have this effect as an necessary condition of correctness of the model. The sublinear behavior of charge transfer with the tendency to saturation is also observed for organic crystals that is for perylene or deuterated naphtalene [17].

#### 3. Experimental

##### 3.1. Materials

The purification of the materials: anthracene and cis- and transisomers of 2-methylcyclohexane (DMCH), as well the preparation of the sample was described in previous articles [7, 11, 18].

##### 3.2. Apparatus

The apparatus was described in details also in previous papers [7, 11, 18] where so-called side excitation was used. Here we will concentrate on the parts of the setup that make possible to realize two way of excitation: the side one and the bottom one. In this paper we will use the names: illumination/excitation from the side (side excitation) or from the bottom (bottom excitation). Special attention will be focused on the elements necessary to excite the sample in spatially periodic way.

In order to create the spatially periodic distribution of the molecules in the T_{1} state in the sample in wide range of viscosity two options of the excitation were used. The details of the part of the apparatus generating spatially inhomogeneous distribution of the molecules in the T_{1} state are presented in Figure 1. Starting from side illumination: The laser beam prepared by the fast chopper [7, 11] was reflected in the vertical direction towards mirror M8 (see scheme of the setup in [7, 11, 12]). M8 was used to reflect the laser beam horizontally to generate the homogeneous or spatially periodic distribution of the molecules in the T_{1} state. One of the wall of the sample cell was partly covered with the dielectric mirror reflecting back the laser beam from Ar^{+} laser. Since the coherence length of the laser beam was 50 mm and the pathway in the cuvette was equal to 10 mm the reflection of the 30 *μ*s long pulse generated the standing wave in the sample. In the case of homogeneous excitation the cell was moved down and the excitation beam was passing the sample over the top edge of the dielectric mirror [7, 11, 18]. This option of spatially periodic excitation is presented in Figure 1(a). The other possibility of inhomogeneous excitation: the illumination from bottom is presented in Figure 1(b). The mirror signed as M7 in [7, 11] was replaced by M7′. Mirror M7′ is adjusted in such a way that it directs the laser beam to the mirror M8′ and then to the beamsplitter BS instead of reflecting it in the vertical direction to the M8. Then the laser beam was divided by beam splitter BS. Both beams were reflected by mirror M0 and entered the cell through the prism placed on the bottom wall of the cuvette. The prism angle was equal to 60°. Due to refraction of both beams on the prism walls the directions of both laser beams changed in such a way that the beams crossed in the sample and generate the spatially periodic distribution of the exciting light. In order to measure the decay in the case of homogeneous excitation from the bottom one of the crossing beam was cut-off. In the case of interference of two laser beams having equal intensities the light intensity in the maximum should be four times larger than the intensity of each interfering beam. To have the same intensity of the exciting light for homogeneous excitation and in places of positive interference of laser beams different energy of laser pulses were used for homogeneous and spatially periodic excitation. The energy of the pulse in the case of homogeneous excitation was 4 times larger in comparison with the energy of the laser pulses used for spatially periodic excitation.

The spatial period of the distribution of the exciting light can be calculated from the following equation [4]: where is the wavelength of the exciting laser light, the refraction coefficient of the solvent for the wavelength , and () is the angle between two crossing beams in the solvent. The angle corresponds to the standing wave (the side illumination). The idea of spatially periodic excitation is presented in Figure 2.

##### 3.3. Evaluation Procedure

The data were fitted using the following procedure. First the decies obtained for spatially homogeneous excitation were fitted with intuitively modified model [7]: where is a dark current, is the amplitude, will be called Smoluchowski parameter (, and is the first-order decay time (). Obtained from the fits of homogeneously excited decies the parameters and were used in the fitting of decies measured for spatially periodic excited samples. To fit the periodically excited decies the following functions were used: where describes the degree of the interference, is the diffusion relaxation time (), and plays the role of the indicator of nonexponential behavior and equals The function defined by (18) describes the model developed by Nickel [4–6] within standard Smoluchowski approach. It was used here as a reference. The function defined by (19) can be treated as intuitively modified (18) and describes the condition of strongly dominant first-order decay. The equation (20) introduces to (19) the nonexponential approximation on the lowest possible level (see Section 2).

The functions given by (19) and (20) take into account both components of the decay in intuitively modified model: the short time part and the stationary part. The diffusional relaxation has to play important role in the kinetics on different time scale in comparison with short time effect. Otherwise it will be impossible to differentiate between these two components of the model in the case of analysis of the data measured with spatially periodic excitation of the sample.

The measurements of the diffusion coefficient were performed in the temperature range from 135 K to 155 K. The viscosity of the mixture solvent changes over four orders of magnitude. In the vicinity of the ends of the temperature range the differences in time scales of the short time effect and diffusional relaxation may be so large that the simplified equation for the may offer more stable fit than (19) or (20). Taking the diffusion coefficients used in previous papers [7, 11] and the approximate value of annihilation radius equal to 1 nm and the spatial period of excitation as used in the experiment one can estimate the contributions of the short time term and the diffusional relaxation term to the whole decay of the delayed fluorescence. The measurements were done for two values of spatial radius: 119 nm and 305 nm. In the next paragraph the estimation of the short time term and the diffusional relaxation term will be presented.

In the case of low viscosity where the diffusion is fast the short time effect decay also very fast. In order to have reasonable time for the diffusional relaxation one should excite the sample with enough large spatial period. For example, in the case of the decay measured at K after the delay equal to about 2 ms the term describing short time effect equals to about 0.06 whereas the diffusional relaxation term: equals to ~0.60 if the spatial period of excitation equals to about 305 nm. After the delay equal to 20 ms the short time term equals to 0.02 and the diffusional relaxation equals to about 0.28. So, if the spatial period of excitation is enough large the diffusional relaxation mainly takes place in the delay range where flow can be approximated with stationary part of the intuitively modified Smoluchowski equation. Since the bottom excitation gives small signal, because only small part of the illuminated part of sample contributes to the measured signal the number of the fitted parameter should be limited in order to make the fitting procedure as stable as possible. In such a case (20) should be simplified:

At the other side in high viscosities the diffusional relaxation can also be important for longer delay than the short time effect. For example, in the case of the decay measured at 135 K and after the delay equal to about 3 ms the short time term decreases to the value equal about 0.09 whereas the diffusional relaxation term equals to about 0.95 if the spatial period of excitation equals to 119 nm. After 20 ms the short time term decreases to 0.06 and diffusional relaxation term to about 0.80. So, as in the case of high temperature the main part of diffusional relaxation takes place when the process can be described with stationary part of the intuitively modified Smoluchowski equation. Here the intensity of the delayed fluorescence is low due to low mobility of the molecules. In such conditions the fit of the function with minimum number of important parameters can give more stable results than the fit of the whole function defined by (19). The simplified equation (19) has the following form:

The other approximation used here is the same as in previous papers [7, 11]: it means that modified initial condition introduced by Nickel et al. [19] have significant influence on the initial part of the decay. In those papers the decay was divided into *anti-Smoluchowski* and *Smoluchowski* time ranges. Within the *anti-Smoluchowski* time range the modified initial conditions introduce to the temporal behavior of the decay significant changes in comparison with standard Smoluchowski behavior. In the *Smoluchowski* time range the course of the decay of the delayed fluorescence is the same for both types of initial conditions. The points are systematically cut off from the beginning of the decay of spatially homogeneous excited samples. In this way the part of the decay where the modified initial condition introduce the important changes to the kinetics is removed. The parameters and were obtained from the evaluation of the decies measured for the spatially homogeneous excited samples. They are kept constant in the case of the fitting of the spatially periodic excited samples. The values of the parameters , and in the case of nonexponential behavior are obtained from *Smoluchowski* time range of spatially periodic excited decay. The *anti-Smoluchowski* and *Smoluchowski* time ranges are assumed to be the same in the case of homogeneously and spatially periodic excited samples. The reason to avoid the *anti-Smoluchowski* time range in the fitting procedure is the same as in previous papers:(i)the modified initial conditions do not affect the general idea of the joined action of two processes: the first-order decay and TTA;(ii)the mathematical description of the kinetics is much simpler for the standard Smoluchowski initial conditions.The *Smoluchowski*, *anti-Smoluchowski* time range are presented in Figure 3 together with the delay range of particular interest.

The values of the parameters , , and eventually were obtained as follows. The functions defined by (18) ÷ (20) were fitted to the measured decies with the systematic cutting off the points from the beginning. Using this procedure one obtains the parameters as functions of the delay, . In the case of (19) for exponential behavior or (20) in the case of nonexponential behavior one should obtain the constant values (within the accuracy of the measurement and the fitting procedure) of parameters , , and eventually in the time range of particular interest (see Figure 3). In the case of the decies measured near the limits of the temperature range the simplified functions: equation for exponential behavior and for nonexponential case were fitted and the results were compared with the fits performed for the whole functions (see (19) and (20)). The option giving more stable fit and precise results were taken to the further analysis. The values of parameters and obtained from the fitting of (18) were taken as reference. The values of the parameters were averaged over the range of particular interest. The mutual diffusion coefficient was calculated from the parameter :
where the spatial period was calculated for the wavelength of the Ar^{+} laser nm and the refractive index of *cis*/*trans*-DMCH. The refractive index of the mixture solvent was measure and published previously [20].

The temperature and viscosity dependence of the diffusion coefficient was examined with the following procedure. First the linear functions was fitted to the values of the absolute diffusion coefficient calculated from (22):

Since in the SE model the coefficients and obtained from fits of (23a) and (23b) should be equal within experimental error. The coefficient can be treated as an additional component of the standard deviation of . The contribution of this component should be important in high viscosities. Also the other model of temperature/viscosity dependence of the diffusion constant was fitted to the experimental points: The values of diffusion coefficient were calculated from both models: SE model (see (23a), (23b), and combined model (24)).

The annihilation radius was calculated from (Smoluchowski) parameter obtained from the fits of the decies of homogeneously excited sample: using values of calculated from the both models of temperature/viscosity dependence of the diffusion coefficient. The values of calculated for the experimental data resulting from measurements for short and middle excitation pulses are compared with those obtained for the long excitation pulse [7, 11]. Also the values of the annihilation radius calculated with the diffusion coefficient obtained from SE and combined models are compared with each other.

#### 4. Results and Discussion

##### 4.1. Comparison of the Nickel’s and Modified Approaches, Exponential Decay

In Figure 4 there is presented the comparison of the fit of two models: original Nickel’s approach (18) and intuitively modified one (19). Since the parameters and were taken from the fit of homogeneously excited decay they are not presented here. The comparison is reduced to the diffusional relaxation time , —Figure 4(a) and the degree of interference —Figure 4(b). Both parameters were obtained from the fitting procedure performed for the decay measured at 142 K. The parameters and are presented as dependent on the delay, . The course of the parameters are similar for both models. In the case of (Figure 4(a)) after the decrease for *μ*s the values reach shallow minimum which can be treated constant within the accuracy of the measurement and fitting procedure. For *μ*s the diffusional relaxation time has clear tendency to grow up with the increase of the delay. The values of obtained from the intuitive model are about 500 *μ*s smaller than those from Nickel’s original approach in the range of , where the plateau is reached. The course of the parameter (Figure 4(b)) as a function of the delay corresponds to the course of . For *μ*s the values of the degree of interference increase with the increase of the delay. In the range 500 *μ*s < < 2500 *μ*s there is a shallow maximum where the values can be treated as constant. For *μ*s the values of decrease with the increase of the delay. The insert in Figure 4(b) presents as a function of . For the delay larger than 500 *μ*s the values of the can be treated as constant and equal to about 1.2 for both models.

In Figure 5 there are presented the values of diffusion coefficient (average values) calculated from original Nickel’s and intuitively modified model as a function of in the range below 0.12 K/m Pa s (what corresponds to the temperature below 149 K). The experimental data are presented together with the fit of SE model. The measurements performed in the range K/m Pa s (the temperature below 145 K) were done for the side excitation. In the case of higher temperatures the larger spatial period of excitation was necessary. These measurements were done for bottom excitation. From the course of the experimental values can be deduced that original Nickel’s approach gives the values of the diffusion coefficient that have a tendency to deviate from the linear SE model. This tendency can be corrected introducing the intuitive modification to the Nickel’s formula for the temporal dependence of the intensity of delayed fluorescence. The correction gives the results consistent with SE model in the range 0.055 K/m Pa s < < 0.12 K/m Pa s. The difference between experimental diffusion coefficient and SE model in the range ~0.015 K/m Pa s < < ~0.055 K/m Pa s is not fully corrected with the intuitive modification of the Nickel’s equation. The additional difference between the experimental values and SE model results from nonexponential decay of the spatially periodic excited sample. This is due to the side excitation of the sample within this range of where the energy of the exciting pulse was significantly larger than in the case of bottom excitation (see Section 3).

##### 4.2. Nonexponential Contribution to the Modified Model in the Case of Spatially Periodic Excitation

The parameters and as a function of the delay are presented in Figure 6, where the values obtained from exponential and nonexponential models (see *Description of the model*, part *Temporal behavior of delayed fluorescence after spatially periodic excitation*) are compared. In Figure 6(a) there is presented diffusional relaxation time as a function of the delay. For exponential model the values of decrease with the increase of the delay for *μ*s. For ~400 *μ*s < < ~1300 *μ*s the values obtained from exponential model show shallow minimum. The values within this range of can be treated as constant within experimental error. For *μ*s the rapid increase of with the delay is observed. In the case of nonexponential the course of is similar to the exponential case: for *μ*s the decrease of with the increase of the delay is observed. For ~400 *μ*s < < ~1400 *μ*s the values of can be treated as scattered around the constant. For *μ*s the parameter has tendency to decrease with the increase of . Also the increase of the standard deviation is observed for *μ*s. The behavior of degree of interference—Figure 6(b) corresponds to that of diffusional relaxation time. For *μ*s the values of obtained from both types of modified model (exponential and nonexponential) increase with the increase of the delay. In the case of exponential fit parameter reaches the shallow maximum for ~400 *μ*s < ~600 *μ*s and for larger the decrease of the value of with the increase of the delay is observed. The values of resulting from nonexponential fit can be treated as constant within the accuracy of the measurement and fitting procedure for ~400 *μ*s < < ~1000 *μ*s. For *μ*s the values of the parameter are scattered and the standard deviation is much larger than for the delay below 1 ms. The insert in Figure 6(b) presents the indicator of nonexponential behavior—. Although the values are very small (not larger than 0.003) they can be treated as constant within the presented accuracy for the delay between ~400 *μ*s and a few milliseconds.

In Figure 7 there is presented diffusion coefficient obtained from kinetic model with taking into account the nonexponential character of the decay in the range ~0.015 K/m Pa s < < ~0.055 K/m Pa s. The presented values were calculated from exponential: K/m Pa s and nonexponential: ~0.015 K/m Pa s < < ~0.055 K/m Pa s fits. The “experimental” values of the diffusion constant are presented together with the SE model. The range of from 0.003 to 0.052 K/m Pa s corresponds to the temperature range 135 K ÷ 145 K. The difference between SE model and the result of the fitting of exponential modified model of the periodically excited delayed fluorescence increases with the increase of . The nonexponential model of the decay seems to reproduce better the SE dependence of the diffusion coefficient than the exponential approximation of the decay. The effect is small but the tendency seems to be unequivocal.

##### 4.3. Negligible Short Time Effect

As mentioned in Evaluation procedure in some cases the short time effect can be neglected and the simplified functions of can be used. In Figure 8 there is presented an example of evaluation of the decay measured at 155 K. Here the results of the fitting procedure obtained from whole and simplified kinetic models are compared. The course of the parameters obtained with or without the contribution of the short term are almost the same. The diffusional relaxation time (Figure 8(a)) shows the increase of the value of the parameter for *μ*s. For the delay from the range ~250 *μ*s to ~1 ms the values are scattered around the constant. For ms the increase of the standard deviation and the value of is observed. The increase of reflects the decrease of the contribution of the diffusional relaxation to the total decay. The values obtained for the parameter (Figure 8(b)) show the decrease as a function of the delay in the range: *μ*s. For ~250 *μ*s < < ~700 *μ*s the values of can be treated as scattered around the constant. The dispersion of the values obtained with the neglecting of the short time effect is significantly smaller in comparison with the data resulting from the fit of (20). For *μ*s the the values of decrease with the increase of the delay. The parameter tends asymptotically to zero. The values of are presented in the insert in Figure 8(b). The values of are constant and equal to about 1.05 for both version of the kinetic model.

In Figure 9 there is presented similar comparison as in Figure 8. Here however, the evaluation was performed for the decay measured at 136 K. The complete function of the kinetic model was defined by (19), the simplified one by . In Figure 9(a) the diffusional relaxation time is presented as dependent on the delay. In the case of the fit with short time term included only first two points are presented in the plot. The other values obtained from the fit of (19) are much larger than the scale and their standard deviation is larger than the value of the parameter. So they are omitted in the plot. The results of the fit of show the decrease of with the increase of the delay for *μ*s. For ~800 *μ*s < ~8000 *μ*s the values of the diffusional relaxation time can be treated as scattered around the constant. The standard deviation grow up with the delay. This behavior reflects the decrease of the contribution of the diffusional relaxation to the whole decay with the increase of the delay. For the decrease of the with the increase of the delay is observed. The degree of interference shows similar behavior—Figure 9(b). The data obtained from fit of (19) (with short-time term included) are not presented due to the standard deviation overcoming 100% of the value of the parameter. In the case of the analysis neglecting short time effect the values of are scattered around the constant for < 8 ms. The exception is the first point which standard deviations is larger than the value. The increase of the standard deviation is observed as in the case of the diffusional relaxation time. This increase is very clear for the delay above ~5.5 ms. For ms the increase of the with the increase of the delay is observed. Also the increase of for the delay above 8 ms is much more significant. From the course of and the ratio one can deduce that the contribution of the diffusional relaxation to the whole decay decreases significantly within the delay range ~5.5 ms < < 8 ms. For the delay larger than 8 ms this contribution seems to be negligible. The insert in the Figure 9(b) presents the as dependent on the delay. The value obtained for the fit of (19) are spread between 1 and 2.4, whereas the values of resulting from fit of are constant and equal to 1 for *μ*s.

It seems that in the case of periodically excited samples in some cases the simplified functions with reduced number of parameters give better results than the complete one. This is due to the complicated temporal dependence of the delayed fluoresce which is described with function having up to 7 parameters. In some cases simplification of fitted function results in avoiding the parameters that values are negligible. This gives in turn more stable fit and more reasonable results.

##### 4.4. Viscosity and Temperature Dependence of the Diffusion Coefficient

In Figure 10 there is presented the diffusion coefficient as dependent on . Together with experimental points fitted models are presented: SE model and a combined model. The course of experimental points shows that the both models give the similar values of the diffusion coefficient up to K/m Pa s. For larger values of the diffusion coefficient shows the deviation from SE model towards lower values. The other model seems to reproduce better than the SE model the course of the experimental points. The values of extrapolated to the room temperature with different models are as follows,(1)SE model: nm^{2} s^{−1} = cm^{2} s^{−1},(2)combined model: nm^{2} s^{−1} = cm^{2} s^{−1}.Let us compare the diffusion coefficient measured for organic molecules (having the size similar to anthracene) in organic solvents at room temperature with the data extrapolated with both models. The collection of the anthracene absolute diffusion coefficients () in different solvents can be found in [6]. The values at 25°C are in hexane cm^{2} s^{−1}, in octane cm^{2} s^{−1}, in hexadecane cm^{2} s^{−1}, in perfluororohexane cm^{2} s^{−1}, and in methylcyclo-hexane cm^{2} s^{−1}. The diffusion constant of pyrene and 9,10-diphenylanthracene in hexane equal to cm^{2} s^{−1} and cm^{2} s^{−1}, respectively. Taking into account that the cited data can be summed up: the absolute diffusion coefficient at room temperature in the range between 10^{−5} and 10^{−4} cm^{2} s^{−1}. The value of extrapolated with the combined model ~ cm^{2} s^{−1} can be treated as reasonable, whereas the values calculated with SE model overestimate the experimental day by a few orders of magnitude.

The expansion in the Taylor series of the combined model gives the product of the coefficients and (see (24)) equal to nm^{2} K (m Pa s^{2})^{−1}, where the standard deviation was calculated from
The slope obtained from the fit of SE model equals to nm^{2} K (m Pa s^{2})^{−1}. Taking into account standard deviations of the above discussed quantities both presented above values can be treated as equal.

##### 4.5. Annihilation Radii

Using the diffusion coefficients obtained from both discussed above models the annihilation radii were calculated. In Figure 11 there are presented annihilation radii obtained for the measurements with short excitation pulse and spatially homogeneous excitation. The data presented in Figure 11 consist of values calculated from exponential and nonexponential approaches of the temporal dependence of the delayed fluorescence intensity. In Figure 11(a) the values of annihilation radii calculated with from SE model are presented. For the temperature range 132 K ÷ 146 K the values of are calculated from parameter obtained from exponential fit of (squares). In the case of the temperature range 142 K ÷ 150 K the values are taken from nonexponential fit (circles). The values calculated from exponential model of the kinetics of TTA can be treated as constant for the temperature range 138 K ÷ 144 K. For the temperatures below 138 K the the values of decrease with the decrease of the temperature. The increase of the standard deviation with the decrease of the temperature in the range 132 K ÷ 138 K reflects the growing up uncertainty of the diffusion coefficient with the decrease of the temperature. For the temperatures 145 K and 146 K the obtained from exponential kinetics has the tendency to grow up with the temperature. In the case of the temperature above 146 K the values of the annihilation radii were not average, because there was no range of the delay (for ms) where parameter can be treated as constant in the case of exponential model. In the case of nonexponential fit the values of were calculated for the decies measured in the temperature range 142 K ÷ 150 K. The values obtained from nonexponential fit are constant within this temperature range. For the temperature between 142 K and 144 K the values of resulting from exponential and nonexponential models of the decay are equal. In Figure 11(b) there is presented annihilation radius calculated from the combined model. The values of are presented as dependent on the temperature. The behavior of as a function of the temperature is practically the same in the case of Figure 11(b) as described above for Figure 11(a). The only difference between Figures 11(a) and 11(b) is the behavior of standard deviation in the temperature range 132 K ÷ 138 K. In the case of Figure 11(b) it is constant what shows that the standard deviation of does not increase with the drop of the temperature as in the case of SE model.

In Figure 12 there is presented the annihilation radius as a function of temperature for the short excitation pulse and the model based on the non-Fickian treatment of TTA [11]. The partition of the temperature range between exponential and nonexponential part is the same as in Figure 11. Parts (a) and (b) present the data calculated for the same model of the viscosity and/or temperature dependence of as in the case of Figure 11:(i)part (a): calculated from SE model;(ii)part (b): calculated from combined model.In the case of Figures 12(a) and 12(b) the course of as a function of the temperature is similar to that presented in Figures 11(a) and 11(b). The are two differences between the behavior presented in Figures 12(a) and 12(b), and Figures 11(a) and 11(b). First the annihilation radii calculated from the non-Fickian treatment are larger in comparison with that from the intuitive modification of Smoluchowski equation. The average value for the “intuitive” equals to about 0.6 nm, whereas the non-Fickian approach gives the value of annihilation radius equal to about 1 nm. This discrepancy was already discussed in previous paper [11]. It results from the screening effect of the first coordination shell which is reproduced in the non-Fickian approach. The other difference is the scattering of the values of the annihilation radius in the range 132 K ÷ 134 K which is not observed in the case of intuitively modified model. At the other side, the values of calculated for parameter obtained for intuitively modified kinetic model has the tendency to decrease with the decrease of the temperature in the range 132 K ÷ 134 K. In the case of non-Fickian kinetic model this tendency is not observed.

In Figure 13 there is presented the annihilation radius as dependent on the temperature. Here the values were calculated for middle excitation pulse and intuitively modified model of temporal behavior of the intensity of delayed fluorescence. The convention is the same as in the case of Figures 11 and 12. It means: in Figure 13(a) the values of are calculated using the diffusion coefficients from SE model and in Figure 13(b)—from combined model. The course of the annihilation radius as a function of temperature presented in Figure 13 is very similar to that shown in Figure 11. The changes of the standard deviation shows the same behavior as a function of temperature for short and middle excitation pulses. It means that the main factor contributing to the accuracy of is the standard deviation of the mutual diffusion coefficient—. Also the average value of in the temperature range where annihilation radius can be treated as constant is for both kind of excitation pulse equal within the accuracy of the measurement and the accuracy of the evaluation procedure.

In Figure 14 there is presented the temperature dependence of in the case of middle excitation pulse and the non-Fickian model of the temporal behavior of the intensity of delayed fluorescence. The convention is the same as in Figures 11, 12, and 13: part (a) presents values of calculated with from SE model and part (b)—with from combined model. The behavior presented in Figure 14 is similar to that one presented in Figure 12 for short excitation pulse. There is only one difference: the average value of in the temperature range where the annihilation radius can be treated as constant equals to about 0.7 nm for middle excitation pulse. This value is slightly smaller than average for short excitation pulse and the same model of temporal behavior of .

#### 5. Concluding Remarks

The introduction of the nonexponential modification to the model describing the temporal behavior of the intensity of delayed fluorescence allowed to enlarge the range of the applicability of the model. In the case exponential model to obtain a good fit one has to perform the experiment in such condition that the measurement of single decay takes very long time (several hours). The measurement of the kinetics of TTA taking long time is extremely difficult because the temperature of the sample must be stabilized within the range of ±0.1 K. Otherwise the fluctuations of viscosity are too large to get reasonable values of parameters. In the case of application of exponential model the mathematical condition: must be satisfied. In the case of nonexponential approach one can obtain good fit for the same parameter equal up to about 0.06. The values of the annihilation radius obtained from nonexponential model are in a good agreement with that from literature and from modified exponential models [7, 11].

It seems that the combined model reproduces quite well the dependence between the diffusion coefficient and the ratio temperature to viscosity over several orders of magnitude. This treatment also removes the artificial upper limit of the diffusion coefficient in SE model: when . In the case of very high viscosities application combined or SE model to calculate from intuitively modified Smoluchowski equation results in the decrease of annihilation radius with the decrease of temperature. This kind of decrease of is not observed for the data calculated from non-Fickian kinetic approach. In the case of SE model the increase of with the decrease of the temperature for K is observed. The reason for this behavior is increasing contribution of the uncertainty concerning with the constant factor from (23b).

There are two limits of the discussed model of the kinetics of TTA. First, the nonexponential modifications are based on time-dependent rate parameter from Smoluchowski equation. So in the matter of fact it contains the traces of the time-dependent rate parameter applied for the case of dominant first-order decay [7, 11, 18]. The other limit is concerned with the case of spatially periodic excitation. The evaluation of spatially periodic excited samples is performed as for the spherically symmetric problem, although this irradiation introduces the cylindrical symmetry to the sample which axis is perpendicular to direction of the changes of the intensity of exciting light.

#### Acknowledgments

The authors thank professor Jürgen Troe (Max-Planck Institute, Göttingen, Germany) for his generous support of the research project. The discussion with professor K. Rotkiewicz (Institute of Physical Chemistry PAS, Warsaw, Poland) was very fruitful for preparation of this paper. The support of stay in Göttingen (Germany) of one of the authors (P. Borowicz) by Max-Planck-Gesellschaft (from October 1, 2000 to Septamber 30, 2001) and by Alexander von Humboldt-Foundation (from October 1, 2001 to December 31, 2002) is gratefully acknowledged. Dr B. Nickel passed away 27/01/2002.

#### References

- P. Avakian and R. E. Merrifield, “Experimental determination of the diffusion length of triplet excitons in anthracene crystals,”
*Physical Review Letters*, vol. 13, no. 18, pp. 541–543, 1964. View at Publisher · View at Google Scholar · View at Scopus - V. Ern, P. Avakian, and R. E. Merrifield, “Diffusion of triplet excitons in anthracene crystals,”
*Physical Review*, vol. 148, no. 2, pp. 862–867, 1966. View at Publisher · View at Google Scholar · View at Scopus - V. Ern, “Anisotropy of triplet exciton diffusion in anthracene,”
*Physical Review Letters*, vol. 22, no. 8, pp. 343–345, 1969. View at Publisher · View at Google Scholar · View at Scopus - B. Nickel, “A modification of the avakian-merrifield method for the determination of the diffusion constant of triplet states,”
*Berichte der Bunsengesellschaft für Physikalische Chemie*, vol. 76, pp. 582–584, 1972. View at Google Scholar - B. Nickel and U. Nickel, “The diffusion constant of pyrene molecules in the triplet state in glycerol from −17°C to +15°C,”
*Berichte der Bunsengesellschaft für Physikalische Chemie*, vol. 76, pp. 584–589, 1972. View at Google Scholar - E. G. Meyer and B. Nickel, “Diffusion coefficients of aromatic hydrocarbons in their lowest triplet state: anthracene in hexane, octane, hexadecane, perfluorohexane, and methycyclohexane, pyrene and 9, 10-diphenylanthracene in hexane,”
*Verlag der Zeitschrift für Naturforschung*, vol. 35, pp. 503–520, 1980. View at Google Scholar - P. Borowicz and B. Nickel, “Application of high-accuracy time-resolved laser spectroscopy to the study of diffusion-controlled triplet-triplet annihilation,”
*Opto-electronics Review*, vol. 12, no. 3, pp. 325–332, 2004. View at Google Scholar · View at Scopus - J. B. Birks, Ed.,
*Organic Molecular Photophysics*, vol. 1, Wiley-Interscience, London, UK, 1973. - S. Glasstone, K. J. Laidler, and H. Eyring, Eds.,
*The Theory of Rate Processes*, McGraw-Hill, New York, NY, USA, 1941. - K. Wirtz, “Kinetische theorie der thermoosmose,”
*Verlag der Zeitschrift für Naturforschung*, vol. 3, pp. 380–386, 1948. View at Google Scholar - P. Borowicz and B. Nickel, “Triplet-triplet annihilation in viscous solutions as an example of non-fickian diffusion,”
*Journal of the Optical Society of America B*, vol. 22, no. 2, pp. 315–322, 2005. View at Publisher · View at Google Scholar · View at Scopus - P. Borowicz and B. Nickel, “The kinetics of joined action of triplet-triplet annihilation and first-order decay of molecules in the T1 state in the case of nondominant first-order process,”
*ISRN Spectroscopy*, vol. 2012, Article ID 316037, 13 pages, 2012. View at Publisher · View at Google Scholar - M. Balucani, V. N. Dobrovolsky, A. V. Osipov, and A. Ferrari, “Model of the drain current saturation in long-gate JFETs and MESFETs,”
*Solid-State Electronics*, vol. 49, no. 8, pp. 1251–1254, 2005. View at Publisher · View at Google Scholar · View at Scopus - M. Bellodi and J. A. Martino, “Study of the leakage drain current carriers in silicon-on-insulator MOSFETs at high temperatures,”
*Solid-State Electronics*, vol. 45, no. 5, pp. 683–688, 2001. View at Publisher · View at Google Scholar · View at Scopus - C. J. Sheu and S. L. Jang, “Modeling of electron gate current and post-stress drain current of p-type silicon-on-insulator MOSFETs,”
*Solid-State Electronics*, vol. 47, no. 4, pp. 705–711, 2003. View at Publisher · View at Google Scholar · View at Scopus - A. Christou, “Charge transport in low-dimensional nitride semiconductor heterostructures,”
*Physica B*, vol. 296, no. 1–3, pp. 264–270, 2001. View at Publisher · View at Google Scholar · View at Scopus - N. Karl, “Charge carrier transport in organic semiconductors,”
*Synthetic Metals*, vol. 133-134, pp. 649–657, 2003. View at Publisher · View at Google Scholar · View at Scopus - B. Nickel, P. Borowicz, A. A. Ruth, and J. Troe, “Application of Smoluchowski's generalized theory to the kinetics of triplet-triplet annihilation of anthracene in viscous solution after long-pulse excitation,”
*Physical Chemistry Chemical Physics*, vol. 6, no. 13, pp. 3350–3363, 2004. View at Publisher · View at Google Scholar · View at Scopus - B. Nickel, H. E. Wilhelm, and C. P. Jaensch, “Effect of the Förster energy transfers ${S}_{1}+{S}_{3}\to {S}_{0}+{S}_{n}$ and ${S}_{1}+{T}_{1}\to {S}_{0}+{T}_{n}$ on the time dependence of the delayed fluorescence from aromatic compounds: anti-Smoluchowski and Smoluchowski temporal behavior,”
*Optika i Spektroskopiya*, vol. 83, no. 4, pp. 541–556, 1997. View at Google Scholar · View at Scopus - J. Jasny, B. Nickel, and P. Borowicz, “Wavelength- and temperature-dependent measurement of refractive indices,”
*Journal of the Optical Society of America B*, vol. 21, no. 4, pp. 729–738, 2004. View at Publisher · View at Google Scholar · View at Scopus