High Performance Recycling of Polymers by Means of Their Fluorescence Lifetimes *

Technical polymers could be identified by means of their remarkably strong auto fluorescence. The time constants of this fluorescence proved to be characteristic for the individual polymers and can be economically determined by integrating procedures. The thus obtained unequivocal identification is presented for their sorting for recycling. Furthermore, polymeric materials were doped with fluorescent dyes allowing a fine-classification of special batches.


Introduction
The recycling of organic polymers obtains an increasing interest both in research and technology.There is a necessity for the development of efficient processes because of increasing environmental pollution by polymers ("plastic planet").Moreover, their recycling may open an economic source for organic materials.The majority of technical polymers are thermoplasts and melt and moulding again is attractive for their easy re-use.However, the immiscibility and incompatibility of organic polymers are therefore the main obstacles because lacking uniformity as low as 5% lowers the value of polymers appreciably and an even higher uniformity is required for high performance materials.Pure polymers for recycling may be collected in polymer-processing manufactories, however, the majority of collected material forms mixtures where an efficient sorting is required before processing.The machine-based recognition of polymers is a prerequisite for such processes where methods using the density or electrostatic properties were described [1]- [3].Optical methods are more attractive because of simple, stable and efficient technology where fluorescence is advantageous [4]-[8] because of unproblematic light path and detection.The doping of polymers with fluorescent markers [9] and their re-identification by the spectral resolution of their fluorescence in combination with a binary coding was described in preceding papers [10] [11].This demonstrated the efficiency of the application of fluorescence.However, there are two topics for a fundamental improvement: 1) Only doped material can be recycled where the recycling has to be already targeted in the production of final products; undefined wastes cannot be recycled in this way; 2) the spectral resolution for every flake for recycling costs appreciable efforts for detection and signal processing.Optical processes for the sorting of undoped material would bring about an appreciable progress and would even allow working up deposited material.

Auto Fluorescence of Polymers
The identification of polymers was concentrated to the technical high performance products Luran ® , Delrin ® and Ultramid ® .We found an appreciable strong auto fluorescence of these technical materials with standard optical exciting at 365 nm where mercury lamps may be applied as a light source; see Figure 1.Slight variations of the wavelengths of excitation do not alter the spectra.
The investigated polymers exhibit individual shapes of their auto fluorescence spectra; see Figure 1.We preferred a fluorescence excitation at 365 nm where intense light sources are available.A slight variation of the wavelength of excitation does not influence the fluorescence.The spectra may be used for the identification and sorting of polymers by means of methods of pattern search.Thus, even undoped material can be sorted, however, this requires still an appreciable effort of calculation.As an alternative, we investigated the fluorescence lifetimes of the auto fluorescence and found remarkable differences for various polymers; see Table 1, lines 1 to 3. Fluorescence decay essentially proceeds first order in time with the time constant τ.Minor, less important bi-exponential components (τ bi ) could be detected, however, a mono-exponential interpretation is sufficient for the identification by far.The decay curves can be easily splitted into two branches representing each of the single components of the fluorescence lifetime.Factors about two are between the decay times τ for Delrin ® , Ultramid ® and Luran ® allowing an unambiguous identification of the polymers.
The decay curves of the auto fluorescence of the polymers are reported in Figures 2(a)-(c) and are clearly indicating their pronounced differences.These can be even more easily seen in the fitted function in Figure 2(d).
A simple logarithmic representation of the right branch of the decay curves is by far sufficient for the determination of the differences in lifetimes; right scales in Figure 2.

Fluorescent Labels
An additional labelling of the polymers by means of fluorescent dyes was taken into account for not only identifying the basic polymeric material but also special technological batches.We applied the perylene ester 1 (PTIE), the perylene carboxylic bisimide 2 (S-13) and the terrylene carboxylic bisimide 3 (S-13TBI) because of their light fastness and high fluorescence quantum yields.The fluorescence of these dyes proceeds in different spectral regions forming three channels for detection as can be seen from their fluorescence spectra in Figure 3.The spectra in various polymeric materials differ only slightly from the spectra in solution because solvatochromism of the dyes is weak.As a consequence, the three channels of fluorescence can be taken to be invariant with respect to the tested material.The labelling of polymers can proceed with a binary coding where the first or the second dye or both were applied and so on resulting in 2 n − 1 possibilities for labelling with n as the number of applied fluorescent dyes: Thus, seven individual batches may be labelled for each polymeric material with the application on dyes 1 to 3. The fluorescence spectra may be applied for the identification of the labelling with the individual dyes.The formation of the second derivative of the spectra improves [10] [11] the security of detection.
Furthermore, we found that the time constants for fluorescence decay vary both with the applied dye and the applied polymer; see Table 1.Such combinations can be taken as an additional pattern for the recognition of the  entire batch of a polymer for further improvement for the identification of polymers.Moreover, the determination of decay times needs no calibration concerning the fluorescence intensities (such a calibration may by applied with the auto fluorescence of polymers as internal standards), because the exponential decay remains similar independent from the starting intensity and some dead time before acquisition; this may be advantageous, even for very inhomogeneous flakes for recycling concerning size and shape.We tested the reproducibility of the determined time constant of fluorescence decay and found standard deviation only for the second decimal; see Table 2 for examples.As a consequence, the reproducibility is good enough by far for the unequivocally discrimination between the individual samples; on the other hand, even an

Time-Resolved Detection
The first order exponential decay curves need not be completely registered and fitted because there are well established mathematical procedures [12]- [14] for the determination of the time constant by means of the measurements of two points of the decay curve or even more appropriate by two integrated regions, preferment before and behind the half time (t 1/2 ).This is schematically shown with two Gaussian-shaped samplings in Figure 4 where the integrating measurements improve the signal to noise ratio.One up to two ns time for integration time seem to be appropriate concerning a decay time of about 5 ns for the majority of fluorescent structures.The fluorescence is induced by the periodically pulsed light for excitation where one can expect a sufficiently complete fluorescence decay of 70 ns for the case of an unfavourable lifetime of 10 ns.As a consequence, an unproblematic frequency of about 15 MHz results for repetition.The two regions of integration may be selected by means of two phase sensitive detectors (PSD) and a phase shift between the two analyzing signals for sampling.These need not be applied for each pulse, but may be distributed, for example, between two consecutive pulses.
A further improvement of the signal to noise ratio may be obtained by the accumulation of the signals of detection.One can roughly calculate the upper limit of the detection for sorting with industrial flakes for recycling of dimension of max. 10 mm.Minimal 20 mm space between the individual flakes seem to be realistic and a transport of maximal 500 m/s where 200 pulses for excitation at 15 MHz of repetition should be obtainable for a single flake; this should be more than sufficient for a good signal to noise ratio for an unequivocal sorting.An average mass of about 25 mg was found for standard industrial recycling flakes resulting in a sorting capacity of 1.5 tons of material per hour.This has to be taken as an upper technological limit for a permanent sorting of polymers covered by the described method.The bottleneck for such capacities seems to be more the mechanics than the methodology of detection.Both electronics and mechanics become much more simple for lower demand.

Conclusion
The unequivocal identification of technical polymers by means of their time constants of auto fluorescence decay is a promising method for their sorting for recycling.Time constants can be economically determined by phase-shifted integration of the fluorescence response of pulsed optical excitation.The auto fluorescence of polymers can be applied for the identification of the basic material where a doping with fluorescent dyes allows the further fine-classification of special batches.A binary coding of the doping with n fluorescent dyes results in 2 n − 1 possibilities for the labelling of batches.

Figure 2 .
Figure 2. Fluorescence decay of polymers at 573 nm in linear (left) and logarithmic scales (right) and the characteristic of the light pulse of excitation at 365 nm as dotted lines.Mono-exponentially fitted functions of decay as solid lines.(a) Fluorescence decay of Luran ® ; (b) fluorescence decay of Delrin ® ; (c) fluorescence decay of Ultramid ® ; (d) comparison of the fitted functions for Delrin ® : Solid line, Ultramid ® : Dotted line and Luran ® : Dashed line.

I
absolute determination of the time constant is not necessary as long as the complete setup produces sufficiently reproductive values.

Figure 4 .
Figure 4. Schematic first order decay (solid line) with a time constant of τ = 14.4ns corresponding to a half life t 1/2 of 10 ns.Gaussian-shaped samplings at t = 5 ns (dashed curve) and at t = 15 ns (dotted curve).

Table 1 .
Fluorescence lifetimes of genuine polymers, the fluorescence labels in chloroform solution and doped polymers.
a) Fluorescence lifetime; b) Additional biexponential component; c) Wavelength of excitation in nm; d) Wavelength of detection in nm.

Table 2 .
Test of reproducibility of the time constant of fluorescence decay including the applied method; measurements with individually prepared and re-oriented samples of labelled granulates.
a) Time constant of fluorescence decay; b) Mean value, standard deviation s; c) Wavelengths of excitation in nm; d) Wavelengths of detection in nm.