Electrospinning of recycled PET to generate tough mesomorphic fibre membranes for smoke filtration

Tough fibrous membranes for smoke filtration have been developed from recycled polyethylene terephthalate (PET) bottles by solution electrospinning. The fibre thicknesses were controlled from 0.4 to 4.3 mm by adjustment of the spinning conditions. The highest fibre strength and toughness were obtained for fibres with an average diameter of 1.0 mm, 62.5 MPa and 65.8 MJ m , respectively. The Xray diffraction (XRD) patterns of the fibres showed a skewed amorphous halo, whereas the differential scanning calorimetry (DSC) results revealed an apparent crystallinity of 6–8% for the 0.4 and 1 mm fibres and 0.2% crystallinity for the 4.3 mm fibres. Heat shrinkage experiments were conducted by exposing the fibres to a temperature above their glass transition temperature (Tg). The test revealed a remarkable capability of the thinnest fibres to shrink by 50%, which was in contrast to the 4.3 mm fibres, which displayed only 4% shrinkage. These thinner fibres also showed a significantly higher glass transition temperature (+15 C) than that of the 4.3 mm fibres. The results suggested an internal morphology with a high degree of molecular orientation in the amorphous segments along the thinner fibres, consistent with a constrained mesomorphic phase formed during their rapid solidification in the electric field. Air filtration was demonstrated with cigarette smoke as a model substance passed through the fibre mats. The 0.4 mm fibres showed the most effective smoke filtration and a capacity to absorb 43 its own weight in smoke residuals, whereas the 1 mm fibres showed the best combination of filtration capacity (32 ) and mechanical robustness. The use of recycled PET in the form of nanofibres is a novel way of turning waste into higher-value products.


Introduction
Useful recycling of plastic polymer waste is a growing concern over the entire globe.In the United States in 2012, 32 million tons of plastic waste were generated, amounting to 12.7% by weight of the total municipal solid waste, 1 but about twice that if considering the volume. 2Only 9% of this plastic waste was recovered for recycling, 1 ca.30% of which was related to the collection of used poly(ethylene terephthalate) (PET) bottles. 3hese post-consumer PET bottles are nowadays highly valued in manufacturing since they are inexpensive compared to virgin PET, which would have an equivalent market price of ca. 150 million USD if no PET would be recycled in the USA. 4 The current market for recycled PET ranges from engineering plastics, automobiles, packaged foods, containers, eece fabric, and different kinds of lm. 5,6However, virgin PET also has additional use in applications such as protective clothing, membranes, vascular gras, tissue scaffolding, and ltration. 6,7hus, recycled PET could be an ideal cost-effective choice in a variety of applications.
Bottle-grade PET exists as a semi-crystalline thermoplastic with high impact and tensile strength, chemical resistance, and a reasonable thermal stability, 8,9 but since recycled PET has been in contact with a variety of substances and environments, applications in a biological setting are not suitable.This fact, in addition to the growing concerns on environmental air pollution, 10 means that ltration is one of the most promising application for recycled ultrathin PET bres.Nonwoven fabrics of PET could here play a momentous role in dust ltration because of their porous structure and low cost of manufacturing in combination with its unique mechanical properties. 10Utilization of recycled PET bottles for the manufacture of protable ltration products would help to offset the cost of recycling and encourage the collection of post-consumer PET bottles.The PET polyester is also particularly marketable for recycling since it is one of the most easily identiable thermoplastics since almost all beverage bottles are made of PET.
An increasingly popular technique for the production of nonwoven membranes is electrospinning, which is one of the most rapidly growing industrial polymer processing methods for the production of ultrathin bres.Studies on the electrospinning of PET exist, but only one published study concerns the electrospinning of recycled PET, 9 where electrospinning from melt was applied to prepare ca.30 mm thick bres.Polymer melt spinning is a useful technique to handle thermoplastics when traces of other plastics are possibly present as long as the polymers show melt characteristics that enable them to mix properly at a given process temperature.The drawback is the considerable amounts of heat (energy) required to process the material, and the inferior mechanical properties that comes from repeated heating cycles. 11,12Another limitation of the electrospinning of polymer melts is the relatively high viscosity of the spun material, making it difficult to produce very thin uniform bres. 13A possible alternative advantageous method would be to carry out the electrospinning of recycled PET from solvents, since it allows for unprecedented control of bre morphology.The solvents could be recycled by process-coupled condensation units, with the additional cost for this offset by the thermal energy savings and the benets of separation of llers aer precipitation and sedimentation.
In this present study, solvent-based electrospinning was used to produce recycled PET micro-and nanobres for application-oriented research.The possibility of producing large quantities of continuous ultrane bres with a thickness ca. 2 orders of magnitude smaller than previously reported is shown. 14Fibre morphology and mean diameter were studied in relation to the ow rate and concentration of recycled PET.The electrospun bre properties were determined by cold-eldemission scanning electron microscopy (FE-SEM), infrared spectroscopy (IR), tensile testing, X-ray diffraction (XRD), and differential scanning calorimetry (DSC).The prepared bres were collected as aligned bre mats for mechanical testing in a micromechanical stage, and as randomly oriented bres in isotropic bre mats.The bre mats were demonstrated as lters for airborne substances such as polycyclic aromatic hydrocarbons (PAH), persistent organic pollutant (POP) and ne particulate matter (nanoparticles) since the effect of these in our environment is a current concern. 15In this study cigarette smoke was chosen as a model substance, as it contains several thousands of components, which includes PAH, POP and ca.5% particulate matter. 16,17The results pave the way for supplementary studies on the optimization of these ultrane membranes for use in nanoltration and other applications in consumer and industrial products.

Preparation of polymer solution for electrospinning
Recycled polyethylene terephthalate (PET) was obtained from Coca Cola bottles (500 mL) collected from the same production batch (Coca-Cola Enterprises Sweden AB).The PET bottles were produced from at least 99.8 wt% PET without plasticizers, as determined from the data-sheet from the manufacturer.The molecular weight of this type of PET, which is commonly used during the production of PET bottles, typically lies in the range of 30-80 kDa. 8All the bottles were cleaned and rinsed with pure ethanol prior to drying, followed by shredding into 5 Â 5 mm 2 small pieces.Triuoroacetic acid (TFA, Reagent Plus, 99%) was purchased from Sigma-Aldrich and dichloromethane (DCM, ACS Grade Stabilized, 99.5%) was purchased from VWR International.A mixture of TFA and DCM in a ratio 70/30 by weight was prepared.The recycled PET was then added to the solvent mixture at concentrations of 10, 15, and 20 wt%.Solutions were mixed for 24 h to ensure complete dissolution of the PET at a temperature of 20 C.

Electrospinning of unaligned bres
The polymer solutions were fed at different rates (20, 10 and 5 mL min À1 ) from a 5 mL solvent resistant syringe through a polytetrauoroethylene tube attached to a steel needle (inner diameter 0.60 mm) mounted 25 cm above the collection plate.An electric potential between 7 and 12 kV m À1 (depending on the concentration of PET and ow rate) was applied by adjusting the voltage from the needle to the collection plate.The faster the ow rate, the greater was the voltage required for spinning. 18ll samples were spun for 20 min to ensure that a substantial amount of bres could be obtained for FE-SEM and IR-spectroscopy analysis.

Electrospinning of aligned bres
The preparation of the aligned bres was identical with that of the unaligned bres except that solutions were fed at a xed ow rate and voltage.The 10 and 15 wt% PET concentrations were spun with a ow rate of 5 mL min À1 , whereas the 20 wt% concentration was spun with a ow rate of 10 mL min À1 , as these conditions gave the most uniform bres.The bres were collected on a rotating alignment drum (diameter ¼ 50 mm) rotating at 2000 rpm at the same spinning distance (25 cm) as the unaligned bres.The total spin time was varied depending on the PET concentration of the sample (10 wt% spun for 4 h, 15 wt% for 2 h, and 20 wt% for 1 h) in order to produce bre mats of similar thicknesses.

Tensile testing
Fig. 1 shows the micromechanical tensile testing stage with a mounted PET bre mat.The tensile tests were performed according to a previously reported Template Transfer Method (TTM). 19,20our pieces of uorinated ethylene propylene release lms (thickness ¼ 76 mm) were rst secured along the centre of the alignment drum.Aluminium foil templates (thickness ¼ 30 mm) with a cut out window (length ¼ 5 mm, width ¼ 10 mm) were then attached over the release lms, which were attached to the drum with copper tape (thickness ¼ 66 mm).The cut-out window allowed bre testing on a Deben Microtest micromechanical stage (Fig. 1).Before removal from the alignment drum, bres were xed to both sides of the window using alkoxy-ethylcyanoacrylate (Loctite 460, Hennkel AG & Co. KGaA, Germany).
Once secured in the tensile stage, the side panels of the window were cut (while leaving the bres intact).The length, width, and thickness of a piece of PET bottle was measured and weighed and the density of the PET was determined to be 1.37 g cm À3 .This value was in agreement with the previously reported density of bottle-grade PET (which is also close to that of amorphous PET) 21 and was used to calculate the thickness of the bre mats aer electrospinning, i.e. by weighing a known area of the bre mat taken adjacent to the template on the collection drum.The tensile strength was taken as the highest stress supported during the test, the Young's modulus was taken as the initial linear slope of the stress-strain curve and the toughness was calculated as the total area under the stress-strain curve.

Determination of bre size and morphology
Using detailed scanning electron micrographs, 200 bres were measured from each spun sample of aligned and unaligned bres.The angular deviation of aligned bres was calculated from a minimum of 50 bres from each spun sample.For these measurements, a Hitachi S-4800 cold-eld-emission scanning electron microscope (FE-SEM) was used in conjunction with a thin Pt/Pd 60/40 coating, (10 s at 80 mA) in a Cressington 208HR high-resolution sputter.
The molecular draw ratio of each aligned bre mat (20 Â 20 mm 2 cut-outs) was evaluated from their shrinkage upon heating to 150 C, over a temperature-controlled aluminium surface.

Infra-red spectroscopy (IR)
All IR spectroscopy measurements were performed on a Perkin-Elmer Spectrum 2000 using a 1 cm À1 scan step and a single reection attenuated total reectance stage (ATR) MKII Golden Gate unit.

X-ray diffraction (XRD)
The electrospun samples were prepared as thick sample discs to obtain the highest possible diffraction intensity, i.e. by compressing the bres in a Specac circular dye (diameter 10 mm) for 1 min and 100 kN at room temperature.X-ray diffractograms on samples were taken at room temperature using a PANalytical X'pert Pro MPD diffractometer with a Cu-Ka source (wavelength 1.5418 Å) at a step size of 1 arcmin (2q) and scan step time of 51 s.

Differential scanning calorimetry (DSC)
Thermograms of bres were obtained with a temperature-and energy-calibrated Mettler-Toledo DSC1.Each sample was weighed at 2.0 AE 0.1 mg, enclosed in a 100 mL standard aluminium crucible with one hole in the cover.The samples were heated from 30 C to 300 C at a rate of 10 C min À1 under a nitrogen atmosphere at a gas ow rate of 50 mL min À1 , and allowed to rest for 5 min before subsequent cooling at 10 C min À1 .The degree of crystallinity (X c,DSC ) was calculated, under the assumption that the morphology of amorphous regions before cold crystallisation and aer melting are the same, from the equation: 22 where DH f is the enthalpy of fusion, DH c is the enthalpy of cold crystallization, assessed directly from DSC thermograms, and DH 0 f (136 J g À1 ) is the enthalpy of fusion of 100% crystalline PET. 23

Smoke ltration
To determine the smoke ltering properties of the bre mats with different bre diameters, the mats were evaluated by securing them over a glass tube (9 mm diameter) tted within a tubular system for smoke transportation and testing (see Fig. 7c).A cigarette (Marlboro Gold, Philip Morris International) with its lter removed was tted into the other end of the glass tube at a distance of 25 mm from the bre mat.The specic surface area of each lter was calculated from the measured average diameter (from SEM) of the bres.IR-spectroscopy was employed to determine what sort of components were absorbed, small amounts of smoke were used for this purpose, i.e. 2 mL of smoke was passed through each mat for each mg of bre mat present, at a rate of ca. 100 mL s À1 .The ltration capacity was measured gravimetrically as the total mass uptake aer exposure to an excess of cigarette smoke (1000 mg of tobacco for each mg of lter), a comparison to the commercial lter tip used in the original cigarette was made using the same proportions of tobacco to lter.

Results and discussion
Morphology and composition of electrospun recycled PET-bres Fig. 2a-c shows the result of increasing the concentration of PET in the solvent carrier.By adjusting the concentration, the average bre diameter could be precisely controlled, while the bre morphologies were almost unaffected and showed continuous long bres with smooth surfaces.The insets in the Fig. 1 Photograph of the micromechanical tensile stage with mounted PET fibres (at ca.100% strain) according to the Template Transfer Method (TTM). 19icrographs show the size distribution of the bres, obtained by measuring the diameters of at least 200 bres for each sample.The average bre diameters at these ow rates increased from 0.41 mm (AE0.12 mm, i.e.AE30%) at 10 wt%, to 1.0 mm (AE0.19 mm, i.e.AE19%) at 15 wt%, to 4.3 mm (AE0.34 mm, i.e.AE8%) at 20 wt% PET.The relative standard deviation was largest for the thinner bres and decreased with increasing bre thickness, possibly due to the lower viscosity and high surface tension of the electrospun solutions for these samples, as previously reported. 24Fig. 2d shows the relationship between the bre diameter and the PET concentration at different ow rates.A ow rate three times as high (at a given concentration) had only a marginal effect on the thickness of the bres.For all the evaluated formulations and ow rates, the bres with greatest uniformity were always prepared at the lowest spinnable ow rate (5 mL min À1 for 10 and 15 wt% PET, 10 mL min À1 for 20 wt% PET), Fig. 2a-c.For ease of reference, the above bres are hereaer referred to by their approximate diameters: 0.4, 1.0 and 4.3 mm.1240 and 1092 cm À1 , respectively, and the C-H wagging vibrations from the aromatic structures at 722 cm À1 .Triuoroacetic acid would, if present, be visible within the 1250-1180 cm À1 region due to the C-F stretching modes (1244, 1240 and 1199 cm À1 ), 25 whereas the dichloromethane C-Cl symmetric and asymmetric stretching should appear at 874 and 771 cm À1 .Among these peaks, only those of 1199, 874 and 771 cm À1 could be used to assess trace solvents remaining in the prepared bre mats, due to the overlapping nature of the remaining peaks.However, none of these peaks showed any signicant difference from those of the spectra for the PET samples cut from the original bottle.This PET bottle displayed a minor peak at 1340 cm À1 and a slightly smaller intensity of the peaks in the 3100-2800 cm À1 region.
In summary, the electrospinning from solution allowed the regeneration of the PET polymer as 0.4, 1.0 and 4.3 mm thick continuous bres without any trapped solvent phases or apparent degradation of the original material.

Tensile testing
Tensile testing was conducted on all the prepared bres as aligned bre mats, shown in Fig. 4. The bres had an average angular deviation of 12.4 , as concluded from FE-SEM analysis.As the bre diameter decreased, an increase in strength and in Young's modulus was observed.This could be explained by a higher degree of molecular alignment when thinner bres were produced in the electrical eld, due to the increased bre stretching. 26The strain at break is usually reduced when the strength and modulus increase, but, the results contradicted this trade-off with a positive correlation between strength and strain at break, as shown in Fig. 4, i.e. the strongest bres also had the highest strain at break.Of the different bres tested, the 1.0 mm bres showed the best combination of mean tensile strength (62.5 MPa), modulus (1.39 GPa) and toughness (65.5 MJ m À3 ), Fig. 4. The 0.4 mm bres had a similar strength and a higher modulus, but a signicantly lower elongation at break.The bres with a diameter of 4.3 mm had a much lower tensile strength (À87%), lower modulus (À89%) and lower toughness (À97%), than the 1.0 mm bres, Fig. 5.
A comparison between the 1.0 mm bres (i.e.low density bre mats) and that of bottle-grade PET (tensile strength of ca.79 MPa, Young's modulus of ca.2.8 GPa and elongation at break of ca.70%) 27 showed that the 1.0 mm bres had only marginally lower strength and modulus but a signicantly higher strain at break, i.e. a higher toughness.The toughness of the bre mats was comparable to that of amorphous biaxially oriented lms of PET, pre-drawn to a draw ratio of ca.3-4. 28eleirinho et al. 6 measured the mechanical properties along only lightly oriented electrospun bre mats (spun virgin PET and by using the same solvent ratios as in the present study) with bre diameters of ca.0.6 mm.The stiffness, strength and extensibility (strain at break) were, respectively, 50-60 MPa, 1-2 MPa and 20-40%.These properties were signicantly lower than those of the present bre mats, showing the importance of bre alignment when evaluating the bre properties.Commercial single melt-spun PET laments without additional post-drawing have a stiffness, strength and extensibility on the order of 2-14 GPa, 170-550 MPa and 45-300%. 29,30Post-drawn PET have been reported to have a stiffness and strength of 15 GPa and 1100 MPa, but at the expense of much inferior extensibility when compared to the present bres. 31Hence the conclusion from this comparison is that electrospinning PET at the present conditions does not create the same amount of high molecular orientation as in the strongest commercially post-  drawn bres.Instead the mechanical properties of the present as-spun bre are comparable with as-spun single PET bre laments.
Crystal structure and molecular morphology (XRD and DSC) Fig. 6a shows X-ray diffractograms of the PET bottle and the electrospun bre mats.3][34] As seen in Fig. 6a, the diffractogram of the PET bottle showed only one relatively wide single diffraction peak at 25.5 , corresponding to the (100) crystal plane, together with an amorphous halo at slightly lower angles.Nunes et al. 34 ascribed this particular semi-crystalline structure of PET bottles to the biaxial molecular orientation achieved in the bottle wall during blow moulding.The electrospun bres displayed broader halos with centres at 22-23 and the absence of the peak at 25.5 , indicating that they consisted mainly of amorphous material.Hadjizadeh et al. 33 reported a similar lack of crystallinity in electrospun samples and attributed this to the rapid solidication of stretched PET chains.The solidication rate at the conditions used in the present study was ca.1-10 milliseconds, which is much shorter than the time frame of ca.50 s required for the crystallization of PET at a cooling rate of 100 C min À1 . 35However, it is also known that PET can form a mesomorphic phase (oriented amorphous phase) which does not reveal itself clearly in the diffractograms. 36The formation of an oriented mesomorphic phase inside electrospun PET bres has also been proposed by Wang et al. 24 Keum et al. showed that the presence of this oriented amorphous phase reveals itself as a skew in the amorphous halo. 37This sort of skewed (unsymmetrical) amorphous halo was most noticeable in the case of the 0.4 and the 1.0 mm bres, as can be seen in Fig. 6a.
The existence of an oriented molecular morphology inside the bres was veried by a shrinkage test, where the different bre mats were heated to a temperature of 150 C. The shrinkage in the bre direction was 54% for the 0.4 mm and 43% for the 1.0 mm bres, whereas the shrinkage for the 4.3 mm bres was only ca.4% (Table 1).This conrmed that there was a molecular orientation, since shrinkage of 54% in the bre direction would require an initial molecular draw ratio of 2.2 or more, and complete relaxation could not be expected due to molecular hysteresis.Similar draw ratios have previously been observed in other electrospun polymers, measured by infra-red dichroism spectroscopy. 38he existence of a mesomorphic phase in the two thinnest bres was further veried by the DSC analysis.Fig. 6b shows the heating thermograms of the PET bottle and the electrospun bres with diameters from 0.4 mm to 4.3 mm.No signicant difference in crystallisation temperature during heating (122 AE 2 C) or melting point (247 AE 3 C) could be observed between the bre samples and the original PET bottle.The glass transition temperature (T g ) and degree of apparent crystallinity (X c,DSC ) as calculated by eqn (1), did however show signicant differences for the different samples, see Table 1.
The bre mats consisting of 0.4 mm and 1.0 mm diameter bres showed the highest T g (80.5-81.0C, Table 1).This higher T g was consistent with a more constrained/stretched amorphous phase, and the much greater shrinkage observed when these bres were heated. 39On the other hand, the 4.3 mm bres showed a T g of only 66.4 C, which was consistent with almost no shrinking (3.6%) of these bres.The thermogram of the PET bottle (Fig. 6b) showed no clear glass transition due to the smaller amount of amorphous material in the original PET bottle which had a 21.2% crystalline content.The 4.3 mm bres showed no signicant crystallinity (0.2%), whereas the 0.4 and 1.0 mm bres showed an apparent degree of crystallinity of 6-8%, Table 1.The inconsistency with the X-ray diffractograms in Fig. 6a stems from the fact that eqn (1) assumes that the amorphous morphology before crystallisation and aer melting are the same.This was not the case for the 0.4 and 1.0 mm bres, since they initially contained a highly oriented amorphous morphology, i.e. a mesomorphic phase, and the change in enthalpy to the crystalline phase was thus less than that of an un-oriented amorphous material (e.g. the 4.3 mm bres).On the macroscopic scale, this can be interpreted as meaning that there were local regions, smaller than the detection limit of XRD (ca. 3 nm), [39][40][41][42] with a greater ordering of PET molecules than in the completely amorphous counterpart.None of the bres contained any conventional PET crystals and they were considered to be completely amorphous.However, in the case of the two thinnest bres, a large degree of molecular orientation existed, which was higher than the molecular orientation in the commercial semi-crystalline PET bottle.

Inuence of molecular morphology on mechanical properties
The higher modulus and tensile strength of the two thinnest bres (0.4 and 1.0 mm), than of the 4.3 mm bres was attributed to the greater amount of local regions with intensied intermolecular interactions due to the oriented mesomorphic phase, diverting a larger portion of the stress to the stronger covalent backbone of the polymer.This explanation was consistent with the largest shrinkage for the 0.4 mm bres upon heating (54%, Table 1), and the high Young's modulus which exceeded that of all the other bres (2.1 GPa, Fig. 5).The more extensive initial molecular orientation due to the larger draw ratio for the 0.4 mm bres also explained the reduced strain at break of these bres compared with that of the 1.0 mm bres.Extensive molecular stretching is known to lower the strain at break.The most prominent example of this is probably the drawing and preparation of ultra-high-molecular-weight polyethylene bres (Spectra® bres), which show only a fraction of the strain at break of the un-stretched material. 43oke ltration Fig. 7a and b show one of the electrospun bre mats before and aer smoke ltration.The ltration capabilities of the mats were tested using a cigarette as the smoke source.The bre mat/ membrane was secured inside the tubular system according to Fig. 7c.The air/smoke ow was controlled via a ow rate valve (not shown) and was adjusted to expose the bre mats to the same amount of smoke (2 mL of smoke per mg of bre mat, at 100 mL min À1 ).The bre mats turned brown immediately when subjected to the smoke (see ESI †), and although the effectiveness of the ltration was obvious to the naked eye (Fig. 7b), additional IR spectroscopy measurements were made, see Fig. 7d.All the spectra were normalized to the absorbance of the characteristic PET bands (Fig. 3).A substantial increase in the IR-absorbance of the mat aer the smoke ltration occurred in the 3000-2850 cm À1 range, corresponding to C-H stretching.This C-H stretching can be related to the many carcinogenic hydrocarbons that exist in tobacco smoke. 16The additional hump shown in the 3500-3200 cm À1 region corresponds to the O-H stretching (possibly also N-H stretching) characteristic of alcohols, hundreds of which are identied in tobacco smoke and possibly contribute in tumorigenesis. 16Overall the absorption of these hydrocarbons/alcohols increased as the bre diameter decreased, i.e. when the specic surface area (SSA) increased.The SSA for each membrane was calculated by assuming smooth and long cylindrical bres, shown in Table 2.
The ltration capacities of the electrospun PET membranes were further evaluated by measuring the total mass uptake when a large excess of cigarette smoke was passed through each lter.The results veried that the thinner bres (0.4 mm) contained the highest amount of absorbed smoke components (43.7 times its own weight, Table 2), the bre mat weightincrease was ca. 2 times higher than that seen for the 4.3 mm bres (absorbed 26.4 times its own weight).Small amounts of volatile components were absorbed in all the lters (e.g.water), which revealed themselves when drying the membranes in vacuum for >3 h.The amount of evaporated volatile components in the thickest bres (4.3 mm), was 12.6 wt% of the total absorbed mass, whereas the absorbed substances in the 0.4 mm bres contained only 4.9 wt% volatiles.
The original cellulose acetate cigarette lter tip absorbed only 2.7 times its own weight during the same treatment.In addition to the inferior ltration capacity, these much thicker bres absorbed a larger portion of volatiles, 25.2 wt% of the absorbed matter.This can be interpreted as if the bre mats with thick bres and thus larger pores, had a much lower capability of trapping small solid/non-volatile particles.

Conclusions
Solution-electrospinning has been used to prepare the thinnest ever reported PET bres obtained from recycled PET.The thermal (DSC) and the IR characteristics of the polymer were the same before and aer electrospinning, neither showed any evidence of degradation in the electrospun bres.The average bre diameter was varied by over one order of magnitude (from 0.4 mm to 4.3 mm) by adjusting the concentration of recycled PET in the electrospinning solutions.The uniformity of the bres depended on the feed rate of the solutions, the most uniform bres being formed at the slowest rates.A template transfer method (TTM) was applied to carry out micromechanical tensile testing on aligned bre mats, and it was found that the strength, strain at break, and modulus increased as the bre diameter decreased.This resulted in a large increase in toughness of the bres, making them useful for ltration applications.The 1.0 mm thick PET bres displayed high strength, a high modulus and high toughness (62.5 MPa, 1.39 GPa and 65.5 MJ m À3 ).The improvement in the mechanical properties originated from the formation of a completely amorphous mesomorphic phase with highly oriented PET molecules, as a result of the extensive bre stretching and the rapid solidication of the bres during the spinning.The molecular draw ratio was greater than 2.2 for the thinnest bres, which was higher than that in the original PET bottle.The bre mats were evaluated as ltration devices for air ltration of condensing hydrocarbons and particulate matter.Tobacco smoke was chosen as a model substance as it contains thousands of different particulate and vapour phase substances.IR-analysis conrmed an increased absorption efficiency of hydrocarbons and alcohols as the bre diameter decreased.Gravimetrical measurements also showed an increased ltration capacity (especially of non-volatile particles) with decreased bre diameter.A ltration capacity of more than 43 times the lters own weight was seen for lters with an average bre diameter of 0.4 mm, which is signicantly higher than shown for the original cigarette lter tip.
The combination of the large-scale availability of recycled PET, the electrospinning-induced formation of a toughnessenhancing internal morphology yielding ca. 30 times tougher bre mats, and the high affinity of these PET bre mats for airborne hydrocarbons, also open the way for the applied use of recycled PET in a range of industrial lters.In the future we foresee that these recyclable electrospun non-woven lters develop into biodegradable materials that not only show specic absorption characteristics but also with a tailored lifespan for the intended application, e.g.materials based on biopolyesters.

Fig. 3
shows the acquired bre IR-spectra for the electrospun bres together with the original bottle PET.The analysis was made to also ensure complete TFA/DCM evaporation in all samples (only 1.0 mm bres are shown).The four major peaks associated with the inherent structure of the polyethylene terephthalate were the terephthalic acid ester C]O group at 1714 cm À1 , the asymmetric C-C-O and the O-C-C stretching at

Fig. 2
Fig. 2 Electron micrographs show the un-aligned electrospun fibres spun from solutions with different concentrations of PET (a): 10, (b): 15, (c): 20 wt%, with the same flow rates as used for tensile testing (a: 5, b: 5, c: 10 mL min À1 ), insets show the fibre diameter distributions, (d) shows the average fibre diameter spun at different flow rates (mL min À1 ) as a function of PET concentration in the electrospinning solutions.

Fig. 3
Fig.3IR-spectra of the 1.0 mm thick electrospun PET fibres and of the original PET bottle.The fibres were analysed 30-60 min after electrospinning to ensure that no solvent remained after this time period, i.e. prior to performing other measurements.All electrospun fibres showed the same IR-spectra.

Fig. 5
Fig. 5 The mechanical properties of the 0.4, 1.0 and 4.3 mm fibre mats as a function of average fibre diameter.

Fig. 6 X
Fig. 6 X-ray diffractograms (a) and DSC thermograms (b) for the original PET bottle and fibres spun to different diameters.

Fig. 7
Fig. 7 Photographs of fibre mats (a) before and (b) after smoke filtration testing (1.0 mm diameter), conducted according to the scheme shown in (c).IR-spectroscopy (d) of a clean fibre mat compared to that of smoke-exposed fibre mats with average fibre diameters of 0.4, 1.0 and 4.3 mm.

Table 1
Thermal properties and shrinkage upon heating for the PET a