Chloroform desorption from poly(lactic acid) nanocomposites: a thermal desorption spectroscopy study

Polylactic acid (PLA) is a biopolymer with high optical transparency and good workability properties which is currently investigated for the replacement of commercial petroleum-based polymers in packaging technology [1]. The addition of bio-based nanoparticles such as nano-cellulose [2] has demonstrated a reliable way to improve the PLA mechanical properties, thermal stability [3]. Moreover cellulose nanofibers have been investigated as additive particle in biopolymer films due to their ability to improve the barrier properties against water [4], water vapor [5], air and oxygen [6] and CO₂ [7]. Cellulose nanofibrils (CNF) are elongated nanoparticles extracted from cellulose by acid hydrolysis having diameter ~ few nm and length in the 100 nm range with both amorphous and crystalline regions: CNF are of great interest given their gas-impermeable nature, high optical transparency and excellent mechanical properties [2]. A reliable process for the production of PLA-CNF nanocomposite films is the solvent casting method [8], [9]. This process involves the dissolution of polymer pellets in a solvent which is then loaded with the proper amount of filler nanoparticles, previously surface functionalized to favor their dispersion in the nanocomposites. The resulting solution is then poured onto a glass plate where it is then dried to form a film. Residual solvent tend to remain in the casted film after drying [10].

In this communication we present an experimental study on the desorption kinetics of residual solvent molecules (chloroform, CHCl₃) from biopolymer nanocomposites consisting of lauryl-functionalized cellulose nano-fibrils (CNF) particles dispersed in a poly(lactic acid). The solvent desorption kinetics was studied in structurally characterized nanocomposites by Thermal Desorption spectroscopy (TDS) using a Quadrupole Mass Spectrometer (QMS). The obtained TDS spectra were then analyzed fitting experimental data by curves obtained through a model describing the transport of the CHCl₃ molecules hosted in the PLA matrix to the sample surface where their desorption occurs. We obtained information on the physical-chemical state of the residual solvent molecules in the PLA layers and a description of the microscopic mechanism controlling their release. Information on the residual solvent desorption kinetics are important for applications in particular in fields such as food and electronic packaging, as well as for the development of innovative medical material [11], [12], [13]. The obtained information are also of basic interest because the role of residual solvent in the dispersion of the filler particles in the casted polymer films, which determines different properties of the nanocomposite, is not well investigated [14], [15], [16], [17].

PLA nanocomposites were prepared by solution mixing and casting according to the following procedure. A first solution was prepared dissolving poly(lactic acid) (PLA 4032D, Nature works LLC, density of 1.24 g/cm³ and 160°C melting point) in CHCl₃ (1 g PLA/25 ml CHCl₃) under magnetic stirring at 40°C. Cellulose nano-fibrils were obtained from cotton through acid hydrolysis and were functionalized by grafting hydrophobic lauryl chain via esterification. A second solution was then prepared by dissolving the laury-functionalized cellulose nano-fibrils [7] in CHCl₃ (0.1 g CNF/10 ml CHCl₃) using ultra-sonicator at 200 W in an ice bath. The PLA-CHCl₃ and CNF–CHCl₃ solutions were then mixed together by magnetic stirrer for 3 h. Nanocomposite films having different additive content were obtained casting mixed solutions with proper PLA-CNF weight ratio in a petri dish after solvent evaporation at room temperature for 24 h and then in a ventilated oven at 40°C for 4 h. Pure PLA films 49±1 μm thick and nanocomposite films with 3 wt.% and 5 wt.% LNC content having thickness of 53±1 and 52±1 μm, respectively, have been examined.

Information on the nanocomposite structure and on the dispersion degree of the filler particles were obtained by Scanning Electron Microscopy (SEM) analyzing the LN₂ fractured cross-sections of nanocomposite films (JEOL high resolution Field Emission Scanning Electron Microscope at 5 kV accelerating voltage). The SEM micrograph in Fig. 1a shows the cross-section of the pure PLA film which presents a smooth and uniform morphology without specific structures. Micrographs in Fig. 1b,c show the cross-section of the nanocomposite sample with 3 and 5 wt.% additive content, respectively: here cavities with not uniform size and shape are present and dispersed filler particles can be distinguished in the cavity walls (see dashed circles). Occasionally, in the nanocomposite with 5 wt.% filler content, CNF aggregates with sub-micrometer size and irregular shape can be observed. SEM micrographs, presented in a previous paper, show that at filler content larger than ~6 wt.% a relevant fraction of the CNF filler particles form CNF aggregates with size in the few micrometer range [7].

Fig. 1:

Cross-sectional SEM micrographs of the pure PLA sample (a), PLA + 3 wt.% CNF (b) and PLA + 5 wt.% CNF (c) nanocomposites.

Figure 2a shows the DSC spectra of the pure and nanocomposite samples (Mettler Toledo TC15 calorimeter, 10 mg samples, N₂ flow of 150 ml/min at 10°C/min). The addition of CNF filler particles does not influence the glass transition temperature value T_g=330 K, nor the melting temperature value T_m=438 K. The most relevant change shown by the DSC spectra is the decrease of the PLA crystalline fraction from 2.7% for the pure PLA to 1.4% for the PLA nanocomposite with 5 wt.% CNF [7]. We also note that the crystallization peak at ~400 K shown by the pure PLA disappears introducing the CNF additive. Figure 2b shows the thermogravimetric curves obtained in the 300–700 K temperature interval at a heating rate of 10 K/min, under a nitrogen flow of 150 ml/min (TGA Q5000 from TA instrument). The TGA curve of the pure PLA sample shows same features as the nanocomposites: (i) a first mass loss starting at ~350 K and (ii) a single step decomposition process starting at ~600 K with maximum decomposition rate at ~640 K.

Fig. 2:

DSC spectra of pure PLA and of the PLA nanocomposites with 3 and 5 wt.% CNF content (a). Thermo-gravimetric curves of pure PLA and of the PLA nanocomposites with 3 and 5 wt.% CNF content (b).

The kinetics of the residual CHCl₃ solvent desorption process was studied by Thermal Desorption Spectroscopy (TDS) using film samples with 2×5 mm size and thickness values reported above. Experiments were carried out monitoring the CHCl₃ mass signal during sample heating in a continuously pumped UHV chamber from T₀=300 K to T_f=500 K, which is below the sample decomposition temperature, with a linear temperature ramp of pendency α=0.12 K/s. The solvent desorption signal was monitored by a Quadrupole Mass Spectrometer (QMS) equipped with grid-type ion source (operating at 90 V ionization potential) and Secondary Electron Multiplier acquiring the QMS I_m/e ion currents pertinent to the CHCl₃ fragmentation products: m/e=83, 85 and 47 [18]. The I_m/e QMS signal is proportional to the CHCl₃ partial pressure PCHCl3: under continuous pumping conditions PCHCl3 is a measure of the CHCl₃ desorption rate jdes(t)=SpA PCHCl3 where S_p is the pumping speed of the vacuum system and A the “effective” surface area of the film sample [19]. Relative indeterminations in the CHCl₃ desorption signal are evaluated by the relation δIm/ebackIm/e where δIm/eback is the mean value of the fluctuations of the Im/eback background signal [20]. Before gas transport tests, the nanocomposite sample was kept in UHV conditions at 300 K inside the experimental apparatus for at least 10 h.

The normalized TDS spectrum of the pure PLA sample is shown in Fig. 3: experimental data are presented as open symbols while the experimental indeterminations are inside the size of the symbol. Solid lines are fitting of the experimental data, see text below. The spectrum exhibits a single, narrow desorption peak centered at temperature T_P=393±1 K with FWHM ~10 K: this peak presents an exponentially increasing leading edge at T≅340 K and a sudden high-temperature cut-off at T≅400 K which are features typical of zero-order desorption kinetics [21]. Note that the desorption process occurs in the same temperature range where TGA analysis evidences the first mass loss, see right panel of Fig. 2; note also that solvent release occurs in the T_g<T<T_m temperature interval, see left panel of Fig. 2.

Fig. 3:

Open symbols: CHCl₃ TDS spectrum of PLA. Lines: TDS curves numerically obtained assuming different solvent transport processes. Dashed and dotted lines: diffusion- controlled transport mechanism with E_diff=0.5 eV and D₀=8×10⁻³ cm²/s (see text). Thin solid line: de-trapping model with E_des=1.17 eV, p₀=1.0×10¹³ s⁻¹ and ε_D=0. Thick solid line: de-trapping model with E_des=1.11 eV, p₀=1.0×10¹³ s⁻¹ and ε_D=−0.07 eV.

The normalized TDS spectra pertinent to the PLA nanocomposites with 3 and 5 wt.% filler content are presented in Figs. 4 and 5, respectively. These spectra appear more structured than that of the pure sample. The spectrum of the PLA + 3 wt.% CNF nanocomposite presents, in fact, a shoulder in the high temperature side of the peak at T_P=388±1 K. The spectrum of the PLA + 5 wt.% CNF nanocomposite consists of two partially overlapping narrow peaks with similar intensity centered at T_P,a=377±1 K and T_P,b=400±1 K.

Fig. 4:

Open symbols: CHCl₃ TDS spectrum of the PLA + 3 CNF wt.% nanocomposite. Dashed line: de-trapping model with E_des=1.11 eV, p₀=1.0×10¹³ s⁻¹ and ε_D=−0.08 eV. Dotted line: de-trapping model with E_des=1.21 eV, p₀=1.0×10¹³ s⁻¹ and ε_D=0. Solid line: convolution of dotted and dashed lines.

Fig. 5:

Open symbols: CHCl₃ TDS spectrum of the PLA + 5 CNF wt.% nanocomposite. Dashed line: de-trapping model with E_des=1.07 eV, p₀=1.0×10¹³ s⁻¹ and ε_D=−0.07 eV. Dotted line: de-trapping model with E_des=1.20 eV, p₀=1.0×10¹³ s⁻¹ and ε_D=0. Solid line: convolution of dotted and dashed lines.

The desorption of molecules dispersed in a host matrix is composed by three processes: (i) the release (de-trapping) of the particles from matrix sites (traps) where they are preferentially hosted, (ii) their diffusion to the matrix surface and (iii) their desorption from surface sites. According to this model, the concentration of CHCl₃ molecules c(x, t) in free volume sites of the host matrix changes with time according to the following continuity equations:

In eq. 1(a) N₀ is the density of trapping sites per unit volume in the x=[0, h] domain, x=0 and x=h are the film opposite surfaces and x the direction along which net solvent transport occurs. Because x≪L edge effects can be neglected and the transport problem can be studied in a 1-dimensional geometry. θ(x, t) is the fractional occupancy at time t of the traps sites located at depth 0<x<h: trap saturation occurs when θ(x, t)=1 while θ(x, t)=0 indicates empty sites. This model assumes two simplifying hypotheses: (i) freely diffusing CHCl₃ molecules released from trap sites cannot by trapped again and (ii) trap sites are isolated and there is no path for the transfer of molecules between traps. The previous equation is numerically solved with the following boundary conditions:

which hold as the sample surface layers are exposed to the vacuum in the pumped UHV chamber; the solvent desorption flux j(t) was then calculated by the relation:

The molecular diffusivity D which control the solvent transport to the surface sites and the de-trapping parameter p which controls the molecular release from trap sites, are fitting parameters of this model. During the TDS run, D changes with temperature T(t)=T₀+αt according to the Arrhenius relation D(T)=D0exp(−EdiffkB  T(t)) where E_diff is the activation energy for CHCl₃ diffusion in PLA and D₀ the pre-exponential factor. The p=p(T) parameter describes the probability that a trapped molecules is released from the trap site: it is proportional to the oscillation frequency of the molecule in the trapping site p₀ and the probability that after an oscillation the molecule escapes from the trap which depends on the trapping energy E_des. This parameter also changes during the TDS ramp according to the Arrhenius relation p(T)=p0exp(−Edes(θ)kB  T(t)). To describe possible interactions between CHCl₃ molecules in trapping sites we introduce a θ-dependent activation energy for desorption:

where Edes0 is the activation energy for desorption of solvent molecules from trapping sites at low fractional occupancy while the ε_D term describes attractive (ε_D<0) or repulsive (ε_D>0) interactions between trapped molecules. Note that the ε_D term plays in the present system the role of the “lateral” interaction term in the Elovich equation, thus describing interaction between adsorbed neighboring molecules in surface processes [22], [23].

As a first attempt we assumed that solvent molecules are hosted only in free volume sites of the host matrix, θ(x, t=0), and modeled the solvent desorption as diffusion-controlled process [24]. We have not found information on the CHCl₃ diffusivity in PLA and thus reasonable values of the E_diff and D₀ parameters have to be guessed. In the simulations we have assumed a lower limit of 0.35 eV for E_diff which correspond to the activation energy of diffusion for CO₂ diffusion in PLA [25] and different c(x, t=0) profiles as initial solvent concentration in the PLA layers. Simulated TDS curves cannot reproduce the experimental TDS spectra. Each (E_diff, D₀) parameters couple that produces TDS curves with the correct peak position at T_P=393±1 K, gives rise to peak having shape different to the observed one: simulated peaks, in fact, present much larger FWHM than experimentally observed. The dashed line in Fig. 3, for example, was obtained with narrow Gaussian c(x, t=0) profile with x₀=25 μm and σ=1 μm: this curve presents a TDS peak with FHWM ~50 K. Even larger FWHM are observed with TDS curves obtained with uniform c(x, t=0)=c₀ profiles, see dotted line in Fig. 3.

We then modeled the solvent desorption assuming that the transport of solvent molecules to the sample surface is a fast process: the release of solvent molecules from their trapping sites controls the desorption of solvent CHCl₃ molecules from the nanocomposite samples. The CHCl₃ desorption flux j_des is thus proportional to the solvent release rate r:

As an attempt we assumed ε_D=0 and we solved Eq. 5 with the initial condition θ (t=0)=θ₀=1. To avoid an arbitrary choice of the pre-exponential factor we fixed it at p0=kB Th~1.0×1013 s−1 which is a common chosen value in TDS data analysis [26].

The experimental TDS peak position was reproduced with E_des=1.17 eV, see thin solid line in Fig. 3: here also, similarly to that observed with the diffusion model, the obtained spectrum fails in reproducing the shape and FWHM of the observed one. The TDS peak presents, in fact, a symmetrical shape and much broader FWHM than experimentally observed.

The experimental TDS spectrum of the pure PLA sample was accurately reproduced considering the presence of lateral interactions between the trapped molecules: the thick solid line in Fig. 3 fitting the experimental TDS spectrum was, in fact, obtained with Edes0=1.11 eV, p0=1.0×1013s−1 and ε_D=−0.08 eV. Note that the obtained TDS curves exhibits the typical shape of zero-order desorption processes with the same FWHM as the observed TDS spectrum.

Figure 6 shows, as a function of temperature T, quantities relevant to describe the present desorption process, namely the fractional occupancy term θ and the simulated desorption curve in the upper panel, the p parameter in lower panel when Edes(θ)=Edes0−εD θ and, for sake of comparison, when Edes(θ)=Edes0. This figure describes the correlation between these quantities and illustrates the apparent zero-order kinetics of the desorption process. Specifically, comparing the θ and p curves, we observe that the ε_Dθ term in E_des(θ) forces the desorption constant p keeping low value of as long as θ~1. When trapping sites start to release trapped molecules, that is θ diminishes, then the ε_Dθ term reduces and a fast increase of the de-trapping constant p occurs: this fast increase triggers a self- sustaining “explosive” desorption process having as fingerprint the very narrow TDS peak and the apparent zero-order kinetics.

Fig. 6:

Upper panel: TDS curve (j_des) and fractional occupancy θ as a function of temperature (simulated curves). Lower panel: desorption constant p obtained with Edes(θ)=Edes0−εDθ (solid line) and with Edes(θ)=Edes0 (dashed line) as a function of temperature. Curves were obtained with Edes0=1.11 eV, p₀=1.0×10¹³ s⁻¹ and ε_D=−0.08 eV.

Before going on with the discussion it is worthy to mention that extrended simulation works considering the complete diffusion-detrapping model, as described by eqs. 1(a,b), has revealed that the shape and position of the simulated TDS curves do not change changing diffusion-related parameters such as sample size h or the initial distribution profiles θ(x, t=0) of the trapped molecules, as long as the activation energy for diffusion E_diff is lower than E_des. The fast diffusion process of solvent molecules is clearly connected to the fact that: (i) the PLA matrix is in its rubbery phase and (ii) kinetic processes occur at temperatures T~T_m (see DSC in Fig. 2).

Thermal desorption processes of molecules adsorbed at the solid surfaces exhibit apparent zero-order kinetics when a fraction of adsorbed molecules form aggregates acting, as long as they are present, as a reservoir to maintain constant the surface coverage: the “lateral” energy term is negative describing attractive interactions between aggregated molecules [21], [26], [27].

The obtained ε_D value of –0.08 eV indicates that residual solvent molecules attractively interact forming aggregates in open volumes (cavities) reasonably located in the boundary layers between the amorphous and crystalline regions of the present semi-crystalline polymer [28], [29]. Molecules hosted in filled trapping sites (θ~1), due to their attractive interaction, have to overcome a potential energy barrier which is ε_D~0.08 eV larger than that in empty trapping sites (θ~0).

It is worthy to mention that the obtained ε_D value well compare with values found in different surface studies. Nagai, for example, analyzing TDS spectra of the Xe/CO/W(110) system evaluated an interaction energy term of 0.88 kcal/mol≅0.04 eV [30]. Varma and Dowben analyzing N₂ desorption spectra from Ni surfaces obtained a value of 5 kJ/mol≅0.05 eV which increased when N₂ molecules formed ordered over-layers [31]. Hood et al. finally, obtained a values of 0.45 kcal/mol≅0.02 eV in N₂ desorption processes from Ru(001) surfaces [32]. Note also that the present values are of the same order as the minimum of the Lennard-Jones potential energy curve between like molecules, Xe–Xe: 0.0197 eV, CO₂–CO₂: 0.0163 eV, CH₄–CH₄: 0.0118 eV [33] and suggests that the trapped molecule interacts with the next-neighbor ones [30], [31], [32].

Let us consider now the TDS spectra of the PLA nanocomposites in Figs. 4 and 5. These spectra can be reproduced by the overlapping two TDS peaks both obtained assuming the CHCl₃ de-trapping process as rate limiting step in the desorption kinetics. The first peak, see dashed lines in Figs. 4 and 5, shows FWHM~10 K, is compatible with a zero-order desorption kinetics and is due to solvent de-trapping from aggregates in the PLA matrix. In fact, it can be accurately reproduced with p0=1.0×1013s−1, ε_D=−0.07 eV and activation energies similar to those of the pure PLA film: Edes0=1.1 eV in the nanocomposite with 3 wt.% filler content and 1.07 eV in the nanocomposite with 5 wt. filler content. The second peak, see dotted lines in Figs. 4 and 5, shows a larger FWHM, ~60 K, and its shape is compatible with a first-order desorption kinetics. This peak was reproduced with p0=1.0×1013 s−1 but: (i) larger activation energy for desorption Edes0= 1.21 eV in the nanocomposite with 3 wt.% filler content and Edes0= 1.20 eV in the nanocomposite with 5 wt.% filler content and (ii) ε_D=0. SEM analysis indicate that in both samples CNF particles are well dispersed in the PLA matrix layers, see Fig. 1, without changing its crystalline fraction and T_g value as indicated by the DSC tests. Given the impermeable nature of nano-cellulose, the second TDS peak can be tentatively attributed to the release of dispersed CHCl₃ molecules trapped in interface sites between PLA matrix and surface-functionalized additive. The comparison between the integrated areas of the two TDS peaks indicates that this second contribution increases with the filler content as it accounts for the release of ~46% and ~75% solvent molecules in the nanocomposite with 3 and 5 wt.% filler content, respectively.

In conclusion, PLA-based biopolymer nanocomposites were prepared by solvent casting method with filler contents up to 5 wt.%. TDS spectra reveal that the release of residual CHCl₃ solvent molecules occurs when the PLA matrix is in its rubbery phase from two kind of trapping sites. The solvent release from the PLA matrix follows an apparent zero-order desorption kinetics and the obtained TDS spectra can be reproduced assuming that solvent release is controlled by the thermally activated de-trapping of aggregated CHCl₃ molecules attractively interacting. The addition of filler particles introduces supplementary trapping sites where dispersed CHCl₃ molecules are hosted.