Release of Carbon Nanotubes from Polymer Nanocomposites

: Carbon nanotube (CNT)/polymer nanocomposites have superior properties compared to the neat polymer matrix. They now are widely used in industry, but questions have been raised about the risks of such materials since CNTs can be toxic when inhaled. For a risk assessment of CNT nanocomposites, it is crucial to know whether CNTs from nanocomposites can be released into the environment or if they remain embedded in the matrix. This review article summarizes the studies that investigated the release of CNTs from nanocomposites during the service life. Three scenarios are reviewed, the release of particles due to mechanical impact, the release due to weathering processes, and the release due to ﬁre. A release during composite production and disposal is not incorporated.


Introduction
Carbon nanotubes (CNTs) exhibit unique mechanical, electrical, and optical properties, therefore they have been considered as a nanofiller for composites.CNT nanocomposites have superior or additional properties compared to their neat matrix materials.These include mechanical properties such as tensile strength and Young's modulus [1,2], energy absorption [3], improved scratch and wear resistance [4], electrical and thermal conductivity [5,6], fire resistance [7], and optical properties [8].Nowadays, CNT nanocomposites are widely used for a variety of applications in different industries, e.g., for automotive, aerospace, defense, electronics, energy, and sporting goods [9].
Because of the extensive usage, the question has been raised whether these nanocomposites pose a threat to the human health and the environment since it is already known, that CNTs can cause health problems.They can cause oxidative stress, genotoxicity, inflammation, and also fibrosis [10][11][12][13][14][15][16].
During the life cycle of a CNT nanocomposite, different scenarios can be identified where CNTs might be released into the environment and pose a threat.Those scenarios have already been developed in several publications [9,[17][18][19][20][21][22][23].In general, the life cycle can be divided into three stages where exposure to humans can occur, production and processing, service life, and disposal.The purpose of this paper is to review the scientific work up to date that investigated the potential of release of CNTs into the environment.It concentrates on the possible release during the service life where untrained humans are in contact with nanocomposites.Three possible pathways are considered, particle release due to mechanical impact, degradation of the matrix due to weathering processes, and release due to fire incidents.
A summary of all used abbreviations in this review is presented in Table 1.

Release of Particles Due to Mechanical Impact
From nanocomposites, particles can be released due to several mechanical actions.Those involve scratching, sanding, sawing, and drilling.So far, no standard method has been established to simulate these actions and also no model has been established to compare them.To measure the release of particles during the lifespan of a consumer product, the Taber Abraser can be used.It is a widely used device to simulate sanding processes and to study abrasion resistances of materials and coatings with its own international standards (e.g., ISO 7784-2:1997; ISO 9352:1995; ISO 5470-1:1999; ASTM G195-08).It provides a continuous abrasion process under defined conditions that can be repeated by any other laboratory.Several studies have already used the Taber Abraser to generate particles from nanocomposites, to measure the particle size distribution, and to search for released nanoparticles [24][25][26][27][28].
Other studies simulate sanding process directly with sanding machines and other workshop tools.Those measurements are closer to the reality for processes with high shear forces but it is more complicated to measure the particle size distributions because it can happen that the sanding device is producing nanoparticles itself [29,30], or that big differences in the particle concentrations can appear either because the particle concentration depends on the applied pressure, or due to burst events [28].To achieve repeatable results for real sanding studies, self-made measurement devices have to be established as it has been done e.g., by Göhler et al. [31].
For the characterization of the released particles, several measurement devices are available.A review on that subject has been done already by Kuhlbusch et al. [32].Generally, it is expected that an abrasion experiment releases particles in the size range from a few nanometers to several micrometers.A determination for the full size range has been done e.g., by Wohlleben et al. [28].They used Fraunhofer diffraction and measured particles with diameters up to 200 µm for a PU nanocomposite.To measure the release of individual free standing CNTs, measurement devices that can detect and collect particles in the nanosize are preferred.
The reviewed papers of this section are summarized in Table 2.
The first study that investigated the release of nanofibers from a nanocomposite was done by Mazzuckelli et al. [33].They measured the release of particles and fibers from a carbon nanofiber (CNF) composite during different stages of the production process including wet sawing.Particle concentration measurements during weighing and mixing of CNFs in an unventilated area and also during wet-saw cutting of the composite revealed a release of airborne particles.Further, filter sampling according to NIOSH Manual of Analytical Methods (NMAM) method 5040 showed that also CNFs have been released during those treatments.
Bello et al. [34][35][36] have published several studies with applied wet and dry drilling, band-sawing under dry conditions, and sawing with a rotary cutting wheel under wet conditions on nanocomposites.The investigated samples consisted either of a carbon fiber/epoxy composite with aligned CNTs placed at the center ply interface or a woven alumina fiber cloth with aligned CNTs grown on the surface of the fibers, impregnated with an epoxy.During the processing of the composites, the aerosol particle concentration and particle size distribution in the nano-and microrange was measured.Further, particles were collected with an asbestos sampling cassette for fiber analysis and also TEM samples have been produced.The aerosol measurements showed that drilling releases more particles with a broader particle size distribution than sawing, this is partially caused by smoke that is generated during drilling.TEM analysis of collected particles show that respirable fibers are released during all processes, but single CNTs have not been found.In contrast to sawing, drilling caused a release of CNT agglomerates [36].
Cena and Peters [37] investigated the particle release during weighing of CNTs and sanding of a 2 wt% CNT/epoxy composite.Only few airborne nanoparticles were measured during the two actions.
Images from TEM imaging show micrometer sized particles with protruding CNTs.The authors conclude that as long as the toxicity of such particles is unknown, precautions should be applied to avoid exposures to workmen.Further, they found out that a biological safety cabinet was more effective than a custom fume hood to avoid exposures.
A POM/CNT nanocomposite was investigated by Wohlleben et al. [25].Abrasion was applied by the Taber Abraser or by a sanding machine.The particle size distribution of abraded particles was measured by AUC (sanding machine) and by SMPS (Taber Abraser, only for airborne particles).The AUC data shows that particles >2.7 µm represent more than 99 wt% of the released particles.According to the SMPS data, 80% of the measured particles in the nanorange are smaller than 100 nm, but no influence of the nanofiller on the particle size distribution could be detected neither by AUC nor by SMPS.Further, no release of CNTs due to abrasion was measured.The authors also investigated the toxicology of abraded particles by in-vivo instillation in rats.No difference could be observed between powders from the neat matrix material and the nanocomposites.
With the same methods, Wohlleben et al. [28] also investigated a PU/CNT nanocomposite.Measurements by SMPS show that only few particles in the nanorange are generated and a release of CNTs was not detected.Most of the released particles were bigger than 10 µm and no protruding CNTs are visible on TEM images.The authors explain the missing protruding CNTs with the possibility that the elasticity of the PU allows the polymer to reflow around the CNTs during fragmentation.Cytotoxicity tests with abraded particles showed no difference between the PU and the CNT/PU samples, no acute toxicity was measured for both samples.
Methner et al. [38] investigated the release of CNFs from an epoxy composite during wet sawing, surface grinding, and belt sanding.Particles for TEM measurements were collected in the process area and at the personal breathing zone.Except a sample from belt sanding where the ventilation was turned on, all TEM samples contained free standing CNF fibers and sometimes also a larger amount of loosely agglomerated CNFs was found.
Ogura et al. [39] used a microgrinder to release particles from a 5 wt% of SWCNT/PP nanocomposite.The released particles were characterized by SMPS and SEM.By the usage of a thermodenuder with a working temperature of 190 • C, they were able to show that more than 99.9 % of the detected particles, in the range from 10 nm to 1000 nm, are volatile particles that have been released by the friction heat of the grinding process.The SEM images from collected particles show particles with protruding CNTs but no free standing CNTs.
Golanski et al. [40] researched the release of CNTs from CNT/PC and CNT/PA6 composites with up to 4 wt% of CNTs.They simulated low abrasion with a rake mounted on a Taber Linear Abraser and high abrasion with a metallic brush.With the rake, only few particles were emitted (< 100 p/cm 3 ) and no release of CNTs was detected.With the metallic brush, two kinds of samples were investigated, one with well dispersed CNTs and one sample with poorly dispersed CNTs.No release of CNTs could be measured for the samples with well dispersed CNTs, however for the samples with poorly distributed CNTs, individual free standing CNTs were observed on TEM grids.The authors concluded that if there are CNT agglomerates present in the material due to a bad dispersion, these CNTs have a greater probability to be aerosolized.Schlagenhauf et al. [27] investigated the release of CNTs from an epoxy/CNT nanocomposite with 0.1 wt% and 1 wt% CNTs.Particle generation was conducted with a Taber Abraser and all abraded particles were collected and characterized.The particle size distributions show for all samples four size modes, one in the nanorange at about 400 nm, and three in the microrange at about 0.7 µm, 1.2 µm, and 2.4 µm while the sample with 1 wt% CNTs showed slightly larger particles for all modes.TEM sampling of released particles revealed that particles with protruding CNTs, agglomerates of CNTs, and also free standing CNTs were emitted during the abrasion process (see Figure 1).The free standing CNTs are clearly shorter than the incorporated CNTs (between 50 nm and 1 µm) and thus have been chopped during the abrasion process.Despite that a release of the nanofiller was observed, it was not possible to detect it with aerosol instruments.The authors explain this with the fact that both the low filler content of the CNTs in the composite and the broad size distribution of the free standing CNTs does not allow a detection by SMPS.The size mode, generated by the CNTs seemed to be within the measurement error.Hellmann et al. [41] measured the particle release from a CNT/epoxy composite during sanding by a sanding machine.SEM measurements show particles with protruding CNTs, but no free standing CNTs have been found.
Huang et al. [42] investigated self-made and commercial available CNT/epoxy nanocomposites with up to 4 wt% CNTs.Abrasion was simulated with a sanding machine using different grit sizes and sanding speeds.The measured particle size distributions show two size modes, the first one with particles smaller than 100 nm contains the major amount of particles, the second mode contains particles from 500 nm to 5 µm.The measured particle concentrations increased with higher CNT loads.No free CNTs were observed except for the sample with 4 wt% of CNTs.

Release Due to Weathering Processes
Weathering can cause a weakening or even a degradation of a nanocomposite matrix and thus expose the nanofiller to the environment.It can be categorized based of three sources of weathering, UV-light, elevated temperatures, water absorption, or a combination thereof.
Degradation by UV-light causes mainly damage on the sample surface [43].Combined with the presence of oxygen, UV-light can cause oxidation of the polymer and chain scission can form molecular products that are either volatile or lie at the surface.A common instrument to characterize the UV degradation process is the ATR-FTIR [44].It allows to follow different degradation products and to define a polymer specific degradation index [43,45,46].
Elevated temperatures can cause thermooxidation with similar degradation products as UV degradation [44].Thermooxidation first affects the surface layer, this is important for coatings and adhesives.Depending on the oxygen diffusion kinetics, also bulk properties of composites will be affected [47].
Degradation by water uptake can influence the bulk properties of polymers and their composites.The affected properties among others are the stiffness, the interfacial strength, the damping ratio, weakening of the fiber-matrix interface in composites, plasticization, swelling, and softening [48].The maximum water uptake depends on the relative humidity but not on the temperature [49].Further e.g., for epoxy, different species of water are observed in the polymer, it can be free or bound to the matrix [50].Different models for the water uptake of nanocomposites have been developed and tested by Liu et al. [51,52].
The reviewed papers of this section are summarized in Table 3. Nguyen et al. and Petersen et al. [53][54][55] investigated the degradation of a CNT/epoxy nanocomposite under intensive UV-light [63].They measured the weight loss of the samples and the degradation of the matrix by ATR-FTIR.The results show that the composite degrades slower than the neat epoxy.In contrary to a SiO 2 /epoxy that the same research group investigated [46], a release of the nanofiller could not be measured.Due to degradation of the matrix material, CNTs formed a network on the surface of the samples but an analysis of released particles did not show free standing CNTs.Further, the scratch resistance of the CNT layer was tested.It could be shown that CNT network on the weathered epoxy surface is more mechanically resistant to scratching than the neat epoxy.The authors conclude that this finding indicates that it is unlikely that the CNTs are readily released into the environment.
The degradation by UV-light of a CNT/PP and a CNT/PE composite has been investigated by Bocchini et al. [56].The degradation of the matrix was measured by ATR-FTIR and the results show that the two matrix materials behave differently.The CNT/PP composite shows a similar behavior as the analyzed CNT/epoxy nanocomposite of Nguyen et al. [53,54], the CNTs are adsorbing the UV-light and thus, less degradation compared to the neat PP can be measured.For the CNT/PE composite, almost no difference between polymer and composite can be measured.The authors propose that besides the photooxidation by the UV light, also thermooxdiation plays a role during the degradation process.Since the CNTs are converting light into heat, they conclude that the composites have a higher temperature during the degradation experiment and thus have a higher thermooxidation.For the PE composite, the positive effect of the CNTs seems to be balanced by the negative effects of the higher temperature.
Besides the abrasion processes, Wohlleben et al. [25,28] and Hirth et al. [57] investigated the UV-degradation of CNT/POM, CNT/epoxy and CNT/PU nanocomposites.During the experiment, the samples were protected by covers to ensure that no loose particles are blown or washed away.Afterwards, either the loose particles on the sample surface, or particles that have been released into water during sonication have been analyzed by TEM.For the CNT/epoxy composite, the TEM images show particles with embedded or protruding CNTs.By application of high shear forces, also free standing CNTs were observed.
Asmatulu et al. [58] investigated the resistance against weathering of neat epoxy and CNT/epoxy coatings.The coatings were exposed to UV-light only or to a combination of UV-light and salt fog (according to ASTM B117).The coatings with incorporated CNTs showed a slower weight loss and smaller cracks were generated on the surface compared to the neat epoxy.
Orlov et al. [59] degraded different nanocomposites with CNTs and amino functionalized CNTs with a combination of UV light and moisture.The authors observed degradation of the matrix by FTIR and the formation of cracks by microscope.SEM imaging of samples revealed that loose CNTs can be found on the samples surfaces as well as within the cracks.Even though the amino functionalized CNTs were better dispersed, the polymer matrix was much less homogeneous.This might be explained with a lower degree of polymer cross linking compared to the composite with neat CNTs.Vilar et al. [60] investigated a CNT/PA6 composite.They weathered the samples under UV light in combination with moisture.After the weathering process, SEM images showed CNTs that are exposed to the environment on the sample surface.The authors also calcinated the samples in order to regain the nanofillers.In contrary to the unexposed samples, the calcination of aged samples caused the release of free standing CNTs.
Busquets-Fitè et al. [61] weathered CNT/PP, CNT/EVA, and CNT/PA6 nanocomposites under UV light combined with moisture.They gathered the aging water and collected the released particles.No released CNTs have been found.
The degradation of a CNT/epoxy nanocomposite with neat and amino functionalized CNTs by UV light and elevated humidity has been investigated by Ging et al. [62].Images by SEM revealed that several forms of CNTs can be found on the weathered surface, free standing agglomerates of CNTs, partially exposed CNTs due to crack formation, and encapsulated CNTs (see Figure 2).The authors conclude that there is a possibility that CNTs can be released into the environment after the matrix is degraded, but it has to be considered that this process takes time and thus the releasing rate would be rather low.Further, the authors milled the exposed nanocomposite films and tested the toxicity of the particles with Drosophila larvae.In contrary to the neat CNTs, the embedded CNTs were not toxic.Changes in the electrical conductivity of a CNT nanocomposite can be used to monitor its degradation.Therefore Barkoula et al. [48] made an attempt to correlate the water uptake of a CNT/CRP/epoxy composite with the change of the electrical resistance.They also compared the water uptake of a CNT/epoxy composite with the neat epoxy.The neat epoxy system exhibited a slightly lower water uptake than the modified systems.These results are in good agreement with the results of Starkova et al. [49] where the water uptake of a CNT/epoxy composite was measured under different relative humidities and different temperatures.At lower temperatures, it could be shown that the water uptake rate for the neat epoxy is faster in comparison with the nanocomposites but the equilibrium weight gain was not influenced by the nanofiller.Further, the addition of CNTs to the epoxy also did not influence the swelling of the samples.

Release Due to Fire
Due to improper disposal by fire or by accidental burning of CNT nanocomposites, the nanofiller can be released into the environment since the decomposition temperature of CNTs is much higher than of the polymer matrix, they can be stable up to 600 • C [64].So far, no detailed study has been published that investigates the release of CNTs from nanocomposites due to an event of fire.However, in the field of fire inhibition, CNT/polymer nanocomposites have been already investigated and the analysis of the produced char can give an indication of the possible release of CNTs into the environment.
The reviewed papers of this section are summarized in Table 4. Kashiwagi et al. [7] studied the mechanism of fire inhibition by CNTs with a PP nanocomposite.The experiments showed that the present CNTs modified the thermal and oxidation properties of the PP matrix.Cone calorimetry measurements showed that the heat release of the PP nanocomposite is significantly reduced compared to the neat PP.The analysis of the sample residues revealed that partially oxidized agglomerated CNTs are present in the char (see Figure 3).Further, the color of the char was red because the iron catalyst in the nanotubes was oxidized.The authors concluded that the formed CNT network increased the mechanical integrity of a protective layer which could act as a thermal insulation layer and also as a barrier for evolved degradation products to the gas phase.
Further experiments by the same group with PP, PS, and PMMA nanocomposites confirmed these results [65][66][67].For all samples, a protection layer made of CNTs was formed during cone calorimeter experiments.The network layer showed physical integrity and had about the same mass as the initial mass of the CNTs in the nanocomposite.
Schartel et al. [68] investigated the fire resistance of a CNT/PA6 nanocomposite.They showed that the CNT network remained in the composite, influenced the melt viscosity, and thus prevented dripping and flowing of the burning sample.The authors concluded that CNTs may be a highly interesting filling material for flame retardation, but only for distinct systems and scenarios.Since dripping is inhibited by the CNT network, the conversion of the material is increased, this might be negative for passing flammability tests.
Verdejo et al. [69] used CNTs to change the fire properties of silicone foams.Also for this study, a CNT network was observed in the residues of the burned foams.Kim et al. [70] investigated a CNT/PEN nanocomposite.The thermal stability could be enhanced by adding CNTs to PEN, but the thermal decomposition kinetics depended strongly on the CNT content.SEM analysis of the residues showed also for this nanocomposite the formation of a CNT network during the decomposition process.
Fu et al. [71] created a CNT/wood/PE nanocomposite that contained either neat CNTs or hydroxylated CNTs (CNT-OH).The hydroxylated CNTs showed a slightly better flame retardancy due to their improved interfacial compatibilization between CNT-OH and wood flour as well as the polymer matrix.SEM images of the char residues of the composite with untreated CNTs showed free CNTs on the char surface.For the composite with functionalized CNTs, no free CNTs could be detected.Zammarano et al. [72] produced a PU foam with incorporated CNFs und tested the flammability of the composite by a cone calorimeter.Due to the incorporated CNFs, flame spread was reduced by preventing the heat transfer of burning material to adjacent surfaces.A thermal stable entangled fiber network was formed that also prevented the collapse of the foam.Smoke and char of this burned CNF/PU foam was then analyzed by Nyden et al. [73] and Uddin et al. [74].They burned the composite and collected samples from the released smoke on filters.Further, they also suspended the char in water in order to collect particles for electron microscopy analysis.The analysis of the samples showed that free standing CNFs are present in the char but they were not airborne.The authors conclude that any airborne CNF has been destroyed in the flames.Further, collected char was shaked to see whether CNFs can be released into air.The measurements showed that the particle concentration was increased by an order of magnitude.The released particles further were trapped in water and by measuring their concentration, the rate of aerosolization of the CNFs was approximately determined to 1.4 mg/hour.Dittrich et al. [75] compared the flame retardance properties of different carbonaceous nanofillers.All filler materials enhanced the flame retardancy of the PP matrix by the formation of a protection layer.Functionalized graphene showed the highest potential as flame retardant material.
Bouillard et al. [76] was the first study to investigate the formation of airborne CNTs during the combustion of a CNT nanocomposite.They used a CNT/ABS composite with 3 wt% of CNTs, combusted the sample in a furnace and collected released particles on TEM grids.Analysis by TEM revealed that free standing CNTs and also agglomerates of CNTs were released into the air.The authors report that the numbers were quite significant posing a possible sanitary risk in the case of accidental scenarios.

Conclusions
A summary for all investigated release studies is given in Table 5, it shows that there are still some gaps that have to be filled.Further, the CNT release from other polymers that have been considered as matrix material for CNT nanocomposites have not been investigated at all, e.g., polyvinyl acetate (PVA) [77], polyimide (PI) [78], polytrimethylene terephthalate (PTT) [79], or vinylester/polyester [80].The studies that investigated the release due to mechanical impact do not give a coherent picture.It stands out that both studies that investigated the release of CNFs from composites find airborne CNFs in the breathing area, not only under dry conditions but also for wet sawing.The number of studies however is limited and no universal conclusions can be drawn.
For CNT nanocomposites, it can be concluded that the expected release scenarios include free standing CNTs, agglomerated CNTs, and particles with-and without protruding CNTs.Depending on the applied shear force during the abrasion process, it is also expected that released CNTs will be shorter than the processed CNTs in the composite.Normally during abrasion processes, the particle concentration in air is too low for particle agglomeration.This means that the finding of agglomerates can indicate a poor distribution of the CNTs in the investigated nanocomposite.
As already discussed by Hirth et al. [57], the release abilities of CNT/polymer composites can be divided into two categories.The first category includes the ductile matrix materials.A release of CNTs has not been observed for those materials.It seems that the ductile matrix is able to reflow around the CNTs during fragmentation [28], for those composites, no or only few protruding CNTs from abraded particles are expected.The second category includes brittle materials, here protruding CNTs from abraded particles are observed for all tested composites.For this category, also a release of CNTs has been observed but only for few tested materials or under special circumstances as a high filler loading [42] or the presence of agglomerated CNTs in the material [40].If other factors, such as the influence of the filler-matrix interface, play a role in the CNT release abilities of brittle composites still has to be evaluated.An approach to reduce or even to prevent the release of CNTs due to abrasion can be the functionalization of CNTs.E.g., for epoxy composites, functionalized CNTs can be better dispersed in the epoxy resin [81] and the interaction between CNT and matrix can be improved [82].
For the weathering studies, no differences among the investigated CNT/polymer composites can be found.All of them expose CNTs to the environment when the matrix is degraded by UV-light.The CNTs form a network and are not easily to detach from the samples.Even though already several studies have been published with confirming results, it is not possible to draw a conclusion for a risk assessment of the considered scenarios since only the worst case has been investigated where the composite has not been protected by additives.Further, the investigation of a CNT release due to a combination of weathering and abrasion processes would be interesting since it is closer to the real life (e.g., usage of outdoor materials or polishing of a dulled surface).Also toxicity studies on this subject are desired.The formed CNT layer on the surface of weathered composites could be a source of a high quantity of released free standing CNTs and thus maybe pose a health risk.For the weathering by elevated temperatures and water, no conclusions can be drawn since their effect on the CNT release has not been investigated yet.
In contrary to incineration where under high temperatures CNTs are destroyed [83], a fire does not degrade all CNTs in composites, they even can be used as flame retardants.All the fire release studies show that during a fire incident, a network of CNTs is formed in the char.It hinders a dripping of molten composite and influences the degradation properties.For future studies, despite the positive flame retardant properties, also exposure to CNTs has to be regarded.For CNFs, it has been already shown that they can be released from the char into the air.For CNTs, such measurements are still missing.Further, the toxicity of the inhalable fraction of the released particles would be important.As a fire is a single incident, only short term exposure would have to be considered.

Figure 1 .
Figure 1.TEM images of abraded particles from an epoxy/CNT nanocomposite.(a) A free standing individual CNT; (b) an agglomerate of CNTs with a couple of individual CNTs scattered nearby; (c) protruding CNTs from an abraded particle [27].

Figure 2 .
Figure 2. Impact of UV light on a CNT/epoxy nanocomposite: the initial flat surface (left) gets cracks and unprotected CNTs are forming a layer (right).Reprinted from Science of the Total Environment, 473474, J. Ging et al., Development of a conceptual framework for evaluation of nanomaterials release from nanocomposites: Environmental and toxicological implications, p. 9-19, Copyright © 2014, with permission from Elsevier [62].

Table 2 .
Carbon Nanotube (CNT) release studies -Release of particles due to mechanical impact.

Table 5 .
Release studies for all investigated polymers.