Functional Barriers in PET Recycled Bottles. Part ILDiffusion of Pollutants During Processing

P. Y. Pennarun, Y. Ngono, P. Dole, A. Feigenbaum

INRA (ADEME, ECO EMBALLAGES), CPCB Moulin de la Housse, 51687 Reims Cedex 2, France

Received 30 June 2003; accepted 22 October 2003

ABSTRACT: Recycled plastics may be polluted by variouschemicals available to consumera. When such matériels areused to manufacture packaging materials intended to be indirect contact with food, thèse pollutants may rr.igrate intofood. The use of multilayer structures with a virgin polymerlayer as a functional barrier may prevent such migration.This work deals with thé possible diffusion of pollutantsinto and through thé virgin layer during processing at hightempératures. A test is designed to measure diffusion coef-ficients in elastomeric and glassy polymers, in their moltenstate; 2.5-dimethoxyacetophenone was used as a surrogatepollutant, and its concentration profiles are monitored byUV microscopy. Based on diffusion coefficients and on ac-tivation énergies, glassy polymers appear to be much betterbarriers than polyolefins, even in their molten state. We thenfocus on poly(ethylene terephthalate) (PET) bottle preforms,

processed by injection molding. In a first approach, a worstcase situation was simulated by numerical analysis, usingboth overestimated diffusion coefficient values and very lowvalues of activation energy. It appears that migration intofoodstuffs through functional barriers cari only be observedwith unrealistic high diffusion coefficients and low activa-tion énergies. Thèse results were confirmed by expérimentaldétermination of thé concentration gradients of model pol-lutants in multilayer structures. It appears that little diffu-sion occurs, despite at a very high température. The effectsof température and thickness of thé functional barrier arediscussed. © 2004 Wiley Periodicals, Inc. J Appl Polym Sci 92:2859-2870, 2004

Key words: PET; diffusion; functional barrier; modeling;food packaging

INTRODUCTION

The possibility of recycling plastics into food-contactmaterials has been largely studied.1'2 Many technolo-gies are now available. However, because recoveredplastics may be polluted by various chemicals, whenfood packaging materials are produced, each processmust be carefully optimized in terms of plastics puri-fication (removal of low-molecular chemical contami-nants) for preventing thé migration of thé pollutantsinto food. A possible approach consists of makingarticles where thé recycled plastic is incorporated intomultilayer structures, and thus, separated from théfood by a layer of new (virgin) polymer, which is thendeemed to behave as a barrier layer (called functionalbarrier) to migration.

The functional barrier may play its rôle as long as itis not itself polluted by thé contaminants, a situationwhich may already occur when thé contaminatedlayer and thé functional barrier are assembled at hightempérature. There has been considérable work inliterature focusing on migration from such multilayer

Correspondance to: A. Feigenbaum.Contract grant sponsor: ECO-EMBALLAGES.Contract grant sponsor: ADEME.Contract grant sponsor: Région Champagne-Ardenne.

Journal of Applied Polymer Science, Vol. 92, 2859-2870 (2004)© 2004 Wiley Periodicals, Inc.

structures, where thé layer (called hère inner layer)located beyond thé food contact layer (thé functionalbarrier) has been deliberately polluted with modelcontaminants (surrogates). Some authors hâve mod-eled thé migration behavior from such structures,emphasizing that thé governing factors are (1) thédiffusion coefficient of thé pollutant in migrationconditions (at ambient température), and (2) théinitial pollution of thé barrier layer after its process-ing (e.g., coextrusion or coinjection). To simulatediffusion during processing, Piringer3'4 uses a con-stant (and very high) diffusion coefficient, whereasPérou et al.5'6 uses a diffusivity gradient, describedby time- and space-dependent température gradi-ents during processing. The problem is (1) to de-scribe correctly thé température gradient (i.e., toknow ail thé détails of thé process conditions, par-ticularly cooling conditions), and (2.) to use correctvalues of diffusion constants.

The current article focused on diffusion between thédifférent layers during poly(ethylene terephthalate)(PET) preform injection. The two following methodswill be used to demonstrate that thé diffusion duringprocessing leads to a negligible pollution of thé bar-rier: (1) numerical simulations, using diffusion coeffi-cients at molten state, measured by UV microscopy;and (2) expérimental concentration depth profilingfrom artificially polluted multilayer bottles. In addi-

2860 PENNARUN ET AL.

It Rn « Susiif ra.,i«ii{Rot* = Qvâèe neSas

Figure 1 Représentation of a preform: polar coordinates.

tion, a method is designed to détermine diffusioncoefficients in molten polymers.

NUMERICAL SIMULATION OF DIFFUSION INPREFORMS DURING COINJECTION

Diffusion in a plane sheet

The mathematical treatment of diffusion kinetics ofchemical substances and of heat is based on concen-tration or température gradients between two zonesseparated by a homogeneous médium. The flux of achemical or température across a plane perp>endicularto thé diffusion direction is proportional to thé con-centration or to thé température gradient. The math-ematical treatment for concentration and températureare thé same.

Thèse gradients follow thé Fick's law for matter andthé Fourier law for heat:

ôC ô / dC\Second Fick's law: — = — D—

dt dx \ dx/ (1)

Boundary conditions

A convection parameter, hc, and thé concentration inthé external médium are introduced to take into ac-count mass transfer at interface. The first Fick's lawleads to

«(i x=0 — Cex() (4)

where hc is a convection parameter at thé surface incontact with thé médium in contact, and Cext is théconcentration in this médium.

The same équation is obtained for heat diffusion(replacing C by T, D by a, and hc by hT).

Diffusion in a cylinder with symmetry axis (fig. 1)

Because thé preform is a cylinder with thick walls, itcan be assumed as an infinité cylinder, in thé directionof symmetry axis OZ. As a resuit, eq. (1) can be writtenas:

Fourier's law:_ d I dT~dx A (2)

where C is thé diffusant concentration, x is thé thick-ness of thé section considered, t is thé time, D is thédiffusion coefficient (diffusivity), T is thé température(K), p is thé matériel density, Cp is thé heat capacity ata constant pressure, and A is thé thermal conductivity.

If pCp is constant, eq. (2) can be simplified by intro-ducing a, thé thermal diffusivity:

dC _ ddt ~ dx

.andx (3)

where a = A/pCp. Mathematical resolution of thèseéquations is quite difficult because thé températureand concentration dependence with x and t is notknown. Numerical analysis is used, with Taylor de-velopments7'8; thé thickness is segmented into 100elementary volumes.

dC _ d I dC~dt ~ or \ ~dr

D dC(5)

where r is thé radial distance perpendicular to théz-axis (Fig. 1).

Application to multilayer preform processing(fig. 2)

When preforms are manufactured, PET injection intothé mold takes place in three steps. First, virgin PET isinjected into thé mold, thé walls being efficientlycooled (a, b) (Fig. 2). Zones in contact with thé moldbegin irnmediately cooling (c). Température and vis-cosity gradients appear in thé material.

In a second step, while thé virgin PET is still moltenin thé middle of thé mold (c), recycled PET is injected(d). Virgin molten PET is pushed and continues to fillthé mold until it reaches thé top of thé mold. Newvirgin PET zones in contact with thé mold are cooling

FUNCTIONAL BARRIERS IN PET RECYCLING. II 2861

a) b) «) *) e) I)Air Virgin malien PET KacjdedPET Vifgin CfloW PET

B B

Figure 2 Différent steps of multilayer preform injections.See text for a, b, c, d, e, and f States.

(e). [Pictures (d) and (e) represent continuous injection(thé real situation is an average of each).]

Finally, virgin resin is injected to ensure filling ofthé mold and to compensate for thermal retraction ofthé preform (f).

The following expérimental parameters, obtainedfrom thé manufacturer, are used for thé modeling ofthé process. The inside and outside radiuses are 0.8895and 1.3210 cm, respectively. Time 0 stands for thébeginning of thé injection. Virgin PET is injected up tot = 4.3 s. Recycled PET (25% of thé preform) is injectedfor 1.5 s in thé middle of thé preform. Then, thépreform is cooled.

Assumptions for heat transfer modeling

The mold (core and walls) is kept at constant tempér-ature equal to 8°C (information given by thé manu-facturer).

Two cases of heat convection between PET and thémold wall are considered: hT = +=° (perfect convec-tion) and hT = 0.0074 (typical value representing apoor convection), for a température of 110°C at thésurface of PET, 30 s after injection (information givenby manufacturer).

The PET does not crystallize, even for thé lowestvalue of convection factor HT.

The température at time t = 0 is homogeneous ailover thé preform and equal at 280°C.

Thermal diffusivity a(T) is linear in two différenttempérature intervais:9

for T>503 K a(T) =

for T<503 K

-0.02 X 10~3 X T + 11.46 X 10~3

(6)

a(T) = -4.565 X 10"6 X T + 3.546

where T is thé température (K).

Assumptions on surrogate diffusion in PET

The diffusivity D follows an Arrhenius-type behavior:D = D0 e\p(-EA/RT), where EA is thé activationenergy of diffusion (J mol"1), T is thé température (K),R is thé perfect gas constant (8.315 J moP1 K"1), andthé preexponential factor D0 is a constant (defined bythé values of EA and D at a référence température).

Surrogates (Table I) or pollutants are homoge-

TABLE ISurrogates Used and Their Classification

Set

AAAAAABB

BB

BCCCCCC

Surrogates

1,1,1-TrichloroethaneDimethyl sulfoxyde (DMSO)Methyl palmitateBenzophenonePhenylcyclohexaneEthyl hydrodnnamatePhénol2,6-Di-terf-butyl-4-

methylphenol (BHT)Chlorobenzene2,5-Thiophenediylbis(5-feri-

butyl-l,3-benzoxazole)(Uvitex)

1-Chlorooctane2,4-PentanedioneAzobenzeneNonaneDibutyl phthalate (DBP)Phenyl benzoateToluène

Properties

V, MF, BP = 75°CNV, P, BP = 189°CNV, NPNV, MP, BP = 305°CNV, NP, BP = 240°CMV, NP, BP = 247°CNV, P, BP = 182°CNV, MP, BP = 265°C

V, P, BP = 131°CFluorescent dye

MV, NP, BP = 183°CMV, P, BP = 133°CDye,NV, NP, BF = 151°CNV, MF, BP = 340°CNV, MP, B]3 = 299°CV, NP, BP -•= 110°C

M(g/mol)

M = 13378

27018216017894

220

113431

14910018212827819892

Concentration (mg/kgSolubility in water impregnated PET bottle)

<1000 ppm (20°C)10

Very soluble11

Not soluble11

<1000 ppm (20°C)12

Not soluble"Not soluble11

93 g/L (25C'C)10

<0.01 g/L (20°C)13

500 ppm (20°C)10

Nd

Nd>10 g/L12

NdNot soluble<400 mg/L (25°C)10

Nd515 ppm (20°C)10

26901363704

29101285587

2616872

1324570

1552785921624533810704

V: volatile; NV: not volatile; MV: médium volatile; P: polar; NP: not polar; MP: médium polar; Nd: not determined.

2862 PENNARUN ET AL.

TABLE IIDiffusion Coefficients of 2,5-Dimethoxyacetophenone (DMA) in Polymers Around thé Melt Température

Diffusion coefficient (XlO6 cm2/s)

Température EVA

100120125 5135140142142145 4150155160162162165 6170170170180182182182185190190200202210210222225230245250260280290

HDPE LDPE IIDPE EP PP PS PA EVOH PVDC PAN PVC PET

0.0030.008

0.050.05

6 11.5 15

0.0940.007

0.185.3 0.26

17 1120

0.370.12

0.030.5

4.7 ; 0.530 40 12

0.67.2 1.56.3 0.33 0.3 0.09

6.950

8.812 1 0.25

550.6

20 1.251.5

22132.5

neously distributed in recycled PET and their concen-trations are taken equal to 100 (arbitrary units).

Transfer of surrogates to thé outside of thé preformis impossible because of thé mold walls (hc = 0).

EXPERIMENTAL DETERMINATION OFDIFFUSION COEFFICIENTS AT

MOLTEN STATE

Materials

A set of 13 polymers used in food packaging waschosen for this study. Thèse polymers are presented inTable IL Polymers were polluted with a low molecularweight UV probe, 2,5-dimethoxyacetophenone (DMA),and with a high molecular weight red dye (acid orange)(both obtained from Aldrich, Strasbourg, France).

IPolluted polymers

Granules of thé polymer under this study wereground in liquid nitrogen to facilitate homogenous

mixing with thé UV probe. The powder obtained wasthen immersed into a methanol solution containing atleast 3000 ppm of thé UV probe compound, 1500 ppmof acid orange, and 1000 ppm of antioxidant to avoidthermal dégradation during thé diffusion experiment.After blending, this mixture was exposed to air flowovernight to remove thé methanol. Drying was thenachieved in a ventilated oven, at 65°C (thé methanolboiling température), for 4 h. The dried and contami-nated powder was then extruded to produce diffusionplate spécimens (100-500 /wn thick).

Methods

Two plates of thé same polymer (100-500 /xm thick,depending on thé polymer intrinsic UV absorption),one polluted and one virgin, were placed, in an alumi-num mold facing each other on their smallest crosssection (Fig. 3). The mold was sandwiched betweentwo irori plates and heated under a slight pressure up

FUNCTIONAL BARRIERS IN PET RECYCLING. II 2863

Polluted pofywier

I Virgin polywerVirgin polymer of ter af ime *t" cf dfffustors

Figure 3 Principle of thé diffusion test around melt tem-pératures.

to thé testing température for 10 min. After this heat-ing session, thé sample was then removed and ana-lyzed with a UV microspectrophotometer (Cari ZeissUV-visible spectrophotometer equipped with a Xénonlamp and a microscope unit for analysis of small ar-eas). The concentration gradient of DMA along théx-axis is recorded (maximum absorption at 330 nm);thé diffusion coefficient is then calculated from thé fitof thé expérimental and calculated concentration pro-files, assuming Fickian diffusion and a constant diffu-sion coefficient (Fig. 4).

The red dye allows us to locate thé initial interfacebetween thé two plates; thé analysis can be done onlyon thé initially virgin side (uncolored).

The main difficulty of thé test is how to avoid blend-ing of thé two plates during their melting. This wouldlead to concentration gradients of DMA not connectedto its diffusion. Blending of thé plates is easily de-tected by a red color gradient, because thé high mo-lecular weight acid orange can be considered as im-mobile on thé test time scale; therefore, any présenceof a gradient is due to blending of thé molten poly-mers.

The following précautions were used to avoidblending: thé virgin and thé polluted plates rnust hâvethé same thickness; sample thickness must be very

close to and lower than that of thé mold; and thé areaof both plates must be slightly lower than thé moldarea (to compensate for thé thermal expansion duringmelting).

Despite thèse précautions, at least 50% of thé pre-pared samples cannot be used for analysis, due toblending.

EXPERIMENTAL EVALUATION OF THEPOLLUTION OF BARRIER LAYER IN

PREFORMS

List of surrogates

Model substances (surrogates) were selected to cover abroad range of différent chemical structures, func-tional groups, and molecular weight (Table I). Themodel substances were subdivided into three groups(A, B, and C) to minimize thé reactivity between themand to avoid overloading in thé PET (Table I). Thèsesurrogates were incorporated into PET flakes by usingdichloromethane as carrying solvent, as reported else-where.14'15 The concentrations of thé surrogates in PETafter processing bottles were adjusted as far as possi-ble in thé range of 500-1500 ppm of each in thé recy-cled layer (Table I) to hâve measurable amounts fordiffusion and migration experiments.

Manufacture of preform

Impregnated PET was dried at 150°C for thé first 3 h ofthé cycle to remove dichloromethane and for a secondcycle to remove water just before injection took place.Trilayer preforms were manufactured by Amcor PETPackaging Recycling France (Dunkerque, France), théimpregnated PET being used instead of thé recycledPET in thé usual process (Fig. 2). The impregnated

54000,5 |

6400

Distance from interface with polluted area (Mm)

7400 8400 9400 10400 11400

• •

Figure 4 Diffusion profiles of DMA in HDPE at différent températures (contact time = 10 min).

2864 PENNARUN ET AL.

Prdbmi s

uttîng Im

Figure 5 Top view (parallel to thé direction of thé mic-rotome) of slices (thickness Ar) sectioned in thé preform. O isthé center of thé cylindrical part of thé preform; other lettersrepresent intersections between linear cutting lines and cir-cles delimiting rings (thickness Ar).

PET représenta 25% of thé preform (excluding théneck and thé bottom).

Three différent preform batches were processed,each containing a différent group of surrogates (setsA, B, or C, Table I). Because groups B and C of surro-gates contained a dye, thé impregnated layer could belocated visually.

Microtome sectioning of preforms and analysisof slices

The bottom cylindrical part of thé preform was se-lected for analyses, as it is thé closest to thé injectionpoint and thé hottest part containing surrogates. Itwas eut into 1-cm-high rings (internai and externaldiameters were 0.8895 and 1.3210 cm, respectively).The thickness of thé ring was then eut into longitudi-nal slices (every Ar = 50 ju-m), in a direction parallel tothé symmetry axis of thé preform (Fig. 5), using amicrotome (LEICA RM 2165 microtome, 1.5 cm/min).

Slices were weighed on a microbalance (SARTO-RIUS M2P), and their weight was compared with thétheoretical values to check their regularity. Each slicewas then extracted separately for 12 h in a 100-/^L vialwith 50 |U,L dichloromethane containing ait internaistandard.

Quantification of surrogates in extracts

Surrogates in extracts are quantified with 2 ju,L injectedin GC-flame ionization détecter (FID) with splitlessinjection technique (Fisons Instrument GC 8160). In-jector température is 250°C. Splitless time was 15 s andflow was 20 mL/min. The column was a DB5-MS J&WScientific (15 m X 0.32 mm X 1 /xm). Carrier gas (He)

flow was 2 mL/min at 40°C. FID température was320°C; H2 and air flows were 25 and 250 mL/min,respectively.

Oven program for surrogates in group A was 40°Cfor 4 min, ramp 15°C/min to 132°C, isotherm for 6min, heating 15°C/min until 270°C, and isotherm for 3min.

Oven program for surrogates group B was 40°C for4 min, ramp 10°C/min to 145°C, ramp 15°C/min to200°C, 30 to 320°C, and isotherm for 13 min.

Oven program for surrogates group C was 40°C for8 min, ramp 15°C/min to 170°C, ramp 2°C/min to180°C, ramp 15 to 240°C, and isotherm for 2 min.

Mathematical treattnent of extracts

Because of thé cylindrical geometry of preforms, andbecause of thé geometry of thé sections (rectangular),thé concentration profile of thé surrogates in cylindri-cal layers must be reconstructed from data in rectan-gular samples. The procédure is schematized by Fig-ure 5.

The weight of each slice, and thé amount of surro-gates, had to be mathematically corrected. Because théthickness of thé ring is relatively constant, thé polymerweight is then correlated with thé sectioned surface.The surface areas of slices (A) were calculated by thédifférence of arc segments by thé following équation,which is obvious from Figure 5 (example of thé arcsegment AQG):

MAQG) = -0- (OQ)2 Koc - sin«AOG) (7)

where (OQ) is thé radius, a is thé angle (radians), andAAQG) is thé corresponding area.

The concentration in thé first slice is directly asso-ciated with thé surface (IPQ). In thé second slice, théamount of surrogates is associated with thé surface(MLPI). The concentration in thé surface (CKN) is thendeduced by subtracting thé total amount of surrogatesin slice 2 from those in slices (MCNI) and (KLPN),which hâve thé same concentration as in thé surface(IPQ).

In thé third slice, thé amount of surrogates is relatedto thé surface (HSLM). The concentration in thé sur-face (REJ) is obtained by subtracting from thé totalamount of surrogates in slice 3 from those of thésurfaces, where (BRJC) and (EFKJ) are thé concentra-tion of thé second surface (CKN), and (HBCM) and(FSLK) are thé concentration of thé first surface (IPQ).

The concentrations are calculated step by step, byrecursive treatment from thé outside to thé inside ofthé ring with thé incrément Ar.

FUNCTIONAL BARRIERS IN PET RECYCLING. II 2865

-7M 7

-15

-17

-19 -

-21 -

-23 -

-25 J

0.0019 O.OOZ1 0.0023 0,0025

35 KJ / mol

0.0027

Figure 6 Arrhenius plot of DMA diffusion coefficients in différent polymers at températures above thé melt.

RESULTS AND DISCUSSION

Expérimental measurement of diffusioncoefficients in usual packaging polymers at moltenstate (fig. 6)

To measure diffusion coefficients around thé melttempérature, a spécial test was designed: two rectan-gular plates were put in contact along their smallestcross section. Both plates were made of thé same poly-mer, but one contained in addition DMA, as a UVabsorbing surrogate, and acid orange (to check forpossible blending of surrogates). Concentration pro-files of DMA into thé virgin plate, monitored by UVmicroscopy, allowed calculating diffusion coefficients(Table II). Diffusion coefficients obtained for polyole-fins are very high and close to thé values in liquids.16

Diffusion coefficients are much lower for more polarpolymers. Low values obtained for polyamide (PA-6)and for EVOH, especially, which can be linked to théformation of polar interactions, such as hydrogenbonds between thé polymer containing NH or OHgroups and thé carbonyl groups of DMA, which low-ers thé mobility.17 However, intramolecular interac-tions between macromolecules themselves would leadto thé same effect (decrease of thé network mobility,leading to a decrease of thé probe mobility). Whateverthé effect is, thèse polymers appear to keep barrierproperties even in their molten state.

Effect of température

The range of températures used in thé industry forprocessing polymers may be very large. Several tem-pératures were therefore tested for each polymer: thélowest température is thé melting température for semi-crystalline polymers, or thé glass transition for amor-phous polymers. Higher températures were chosen tocover thé limits of classical processing conditions.

A low-temperature effect on thé diffusion coeffi-cient of DMA is observed in polyolefins. The mostimportant variation is observed for high-density poly-ethylene (HDPE) and linear low-density polyethylene(LLDPE) (60 kj/mol, assuming an Arrhenius-type be-havior). The activation energy is lower for thé otherpolyolefins. Therefore, an average value of 35 KJ/molcan be given for thé tested polyolefin panel in thétempérature range used in this study.

Effect of polymer polarity

For polar polymers, thé activation energy is higher:110 and 95 kj/mol, respectively, for PAN and PVDC,respectively. For thèse polymers, diffusion in thé mol-ten state is a more thermally activated process.

It appears that thé higher thé diffusion coefficient ofDMA, thé lower its activation energy is. This trendwas also observed for thé diffusion of différent molé-cules in polypropylene.18

2866 PENNARUN ET AL.

-Rextemat wall

T*m|Mntur*{K)

• 540-560• 520 - 540• SOO - 520

• 480 -500• 4» -480

• 440 -480

• 420-440

• 400 - 420

• 380 -400

• 360 -380P 340 -360d 320 - 340• 300 - 320

• 280-300

™Rinsîde wali25

Figure 7 Simulation of températures profiles across PET trilayer preform, as a function of time and thickness, with hT = +°

In summary, thé diffusion coefficient at molten statestrongly dépends on thé polymer polarity, with poly-mer-polymer and/or polymer-probe (surrogate) in-teractions, and is more activated when thé diffusioncoefficient is lower. As a conséquence, thé materialsare good barriers because of a low D and a high EA

(diffusion quickly slowed by cooling).In thé further simulations in this article, we will use

as référence unrealistic (worst case) couples of (EA, D0)values corresponding to higher diffusion coefficientsthan thé values obtained for DMA in PET (Le., at280°C: D = 3 X 1CT6 cm2/s and EA = 110 KJ/mol) andthé values obtained for DMA in any melted polymer(i.e., generally D less than 10~5 cm2/s and EA morethan 35 KJ/mol around melt température).

thé preform remains at 280°C for a few seconds afterthé end of injection and then decreases rapidly.

For hT = +°°, after 16 s, thé température is below100°C, even in thé bulk.

Obviously if heat convection at thé surface is rate-limiting, thé preform cooling will be very slow; thistempérature profile can be considered as a real casesituation, in contact with a mold: thé hT = 0.0074 valuewas calculated from thé measurements of thé preformsurface températures (by an infrared caméra), imme-diately after removal from thé mold, and after différ-ent times of cooling.

An asymmetry of température profiles is found, dueto thé cylindrical geometry, as thé polymer quantity ishigher near thé external surface.

Numerical simulation of diffusion profiles; in PETpreforms during processing

This section describes thé simulations of funcitional bar-rier contamination during processing. The mold is keptat 8°C, and PET is injected at 280°C. The températuregradient in thé thickness is first calculated. Next, thédiffusion of surrogates is calculated taking into accountthé time- and space-dependent température gradient.

Simulation of température gradient

The simulation of preform cooling is presented inFigures 7 and 8 for hT = +°° and hT = 0.0074, respec-tively. At t = 4.3 s, impregnated PET is injected for1.5 s and then cooled. The température decreasesquickly near thé surfaces of thé mold. The middle of

Simulation of surrogate diffusion in exaggeratedconditions

Figure 9 shows thé behavior of pollutants defined bythé following unrealistic set of constants:

EA = 5 KJ mor1

D0 = 5 X 10~4 cm2/s (which is équivalent to D = 1.7

X IQ-W/s at 280°C)

The concentration of surrogates is 0 ail over thé moldfor f < 4.3 s. After injection of polluted PET, thédiffusion of surrogates starts and continues until théend of thé cooling. In this example, when D is veryhigh, thé surfaces are highly contaminated at thé end

FUNCTIONAL BARRIERS IN PET RECYCLING. II 2867

InitialRecyeled part

Extemal wail

Température {KJ

1IIU 'U 'MUa 4»n 480a 470

— a 4?,og u 4-a*ora

•&•Mta«j

J£ !st-

4-.C4 V4^"410400300380370

•wrt3403303»

a sic

-lO.tuj30520

-5)0* 500- 4*^ï-480470

-460-4M-440

430-420•410-4M-390

~ ̂ ^t-3SO-340-330-320- 110

'00,'90

10 15 20Time (s)

Internai wall

Figure 8 Simulation of température profiles across PET trilayer preform, as a function of time and thickness, with hT+ 0.0074.

of thé simulation. Moreover, as thé activation energyis very low, diffusion still occurs after 30 s.

As could be expected from thé asymmetric tempér-ature profiles, asymmetric concentration profiles areobtained. This asymmetry is more and more promi-nent as time increases, as shown by surface (externaland internai) kinetics (Fig. 10).

The set of unrealistic values was chosen to illustratethé phenomenon. However, for more realistic D0 andEA sets of values, thé contamination of thé barrier

layer becomes very low. This is illustrated by Figure11. The pollution of thé internai surface is simulated30 s after injection, as a function of D0 and EA. Thetime t = 30 s was chosen, because when EA > 15KJ/mol, diffusion reaches a pseudoequilibrium before30 s. Lower values of EA are meaningless for PET.

A sharp concentration increase is observed, show-ing that some couples of values lead to a close to zérocontamination, and some couples lead to a high con-tamination level. Assuming an arbitrary limit of 10~7

Concentration(arbitrary unils)__

B 90m K

wTS70«8«06$5046403SM252015105-

- 1OT-85-90-85-ao-79-TO-es-60-SS-g»-«-40-35-30-25•20-IS10

-3

Figure 9 Simulation of concentration profiles during diffusion of a surrogate in thé thickness of a preform, as a function oftime and of thickness (calculated with hT = 0.0074, EA = 5 KJ mol"1, and D0 = 5 X 10~4 cm2/s).

2868 PENNARUN ET AL.

Figure 10 Concentrations at surfaces of a preform obtained by simulation with hT = 0.0074, EA — 5 KJ/mol, and D0 = 5X 10~4 cm2/s. Comparison between inside (•) and outside (A) concentrations in a cylindrical preform, and on surface planesheet ( * ) in thé same conditions.

for thé surface contamination (very low; this has to becomparée! to C0 = 100), we can look in thé plane (Cext

= 10~7) for which (D0, EA) couples thé surface iscontaminated. Isodiffusion coefficient straight Unesare represented in thé plane. The (D280T

= 10~5

cm2/s) straight line has no intersection (in thé studiedunrealistically large area) with thé contamination sur-face map, showing that no (D0, EA) couple is compat-ible with a 10~7 surface contamination. Even such alow contamination level cannot be reached.

Obviously thé straight lines corresponding to lowerD280oC values give no intersection with thé contamina-tion surface map. It can be also concluded that ailsituations with D28o°c < 10~5 cm2/s lead to a negligi-ble contamination of thé barrier layer.

Diffusion coefficients of surrogates and of con-taminants are expected to be lower than 10"5 cm2/s,as is thé case for data in this work (measured forPET or for those overestimated from polyolefin behav-ior) (Table II).

Concentrationsat internai

surface

D(280'C)= ICTcinVs(In thé plane (Éa., Log 08))

20 000

Ea (J/mol)10 000

Figure 11 Concentrations at thé surface of a preform after 30 s in thé mold for hT = 0.0074 (injection of virgin PET 4.3 s,injection of recycled 1.5 s, cooling 30 s) as a function of EA and Log(D0 cm2/s). Values of EA and Log(D0) corresponding toD = 10~5 cm2/s are represented in thé plane (£„, Log D0) by a straight line.

FUNCTIONAL BAKRIERS IN PET RECYCLING. II 2869

11+01

-2,9

* Hinflnite

* 0,1

• 0,05

- 0,02

» Hmouid=O.OQ74

-*- 0,005

-*- 0,002

-*™HaIr=O.OQ712

-2,7 -2,5 -2,3

Figure 12 Concentration at thé inside surface after 30 s in thé mold for différent hT, as a function of Log(D0 cm2/s). EA =10 KJ/mol.

Effect of h,

The previous simulations were made by using an /ZT

value, supposed to correspond to thé real processstudied in this work (calculated from expérimentaltempérature measurements). As thé efficiency of heatexchange can vary with thé types of mold, we inves-tigate in this section thé influence of thé variable hT.

Figure 12 shows thé variation of thé concentration atthé surface, as a function of D0, for différent values ofhT (from hT infinité, instantaneous surface heat ex-change, to hr = 0.00712, which corresponds to ex-changes with static air, i.e., worst case). A pessimisticunrealistic value is taken for EA (EA — 10 KJ/mol).

If we arbitrarily take a surface pollution limit of C= C0/109 = 10~7, we can deduce thé highest accept-able D0 values, corresponding to each hT value:

for hT = 7.12 X 10~4, thé highest acceptable D0

= 1.12 X 10"4 cm2/sfor hT -» +°c, thé highest acceptable D0 = 3.98

X 10~4 cm2/s.

This corresponds to thé following range for thé diffu-sion coefficient at 280°C: 1.3 X 10~5 to 4.5 X 10"5

cm2/s. Thèse values are again higher than thé référ-ence values measured is this work.

Expérimental détermination of concentrationprofiles of surrogates in thé PET preform

Figure 13 shows thé concentration profiles obtainedwith surrogate set B. The other sets (A and C) led tothé same type of results, for instance,

1. AU molécules hâve thé same behavior, even ifthey may hâve very différent chemical andphysical characteristics. For example, UvitexOB has thé highest molecular weight, then thélowest diffusion coefficient. However, its con-centration profile in thé thickness is thé same asthat of other substances. This shows that nodétectable diffusion occurs.

2. The profiles are almost vertical between pol-luted and virgin layers, which shows thatblending and diffusion effects are limited.

3. The recycled part of thé preform is not in thécenter of thé preform total thickness (4.32 mm).According to thé manufacturer, this off-center isinhérent to thé coinjection process and cannot beimproved. As a conséquence, thé thickness of théfunctional barrier in thé preform is around 0.82mm, corresponding roughly to 20% of thé totalthickness. This corresponds, in thé final bottle, toa 55- to 60-jum-thick functional barrier. An obser-vation under microscopy of bottle wall slices ob-tained from thé preforms containing dyes (UvitexOB and azobenzene) showed that thé functionalbarrier thickness is around 60 to 70 jam.

The (not optimal) dimension of thé functional bar-rier is rather due to a dissymmetry of thé inner layerinjection, and a polymer layer blending effect, thandue to diffusion of pollutants during processing.

CONCLUSION

Functional barriers must behave as barriers even inmolten polymers. A double approach was used to

2870 PENNARUN ET AL.

1,0

0,8

0,6

0,4

0,2

-0,2

- Chlorobenzcine

-Phénol

- Chlorooctane

-BHT

-Uvitex

Distance to thé external surface

500 1 000 1 500 2 000 2 500 3 000 3 500 4 000 4 500

Figure 13 Expérimental surrogate concentration profiles (with surrogate set B) in thé thickness of a preform.

demonstrate that pollutants introduced in a preforminner layer do not diffuse into thé functional barrierduring processing.

In thé first approach (simulation), thé functionalbarrier pollution was simulated by numerical anal-ysis. Unrealistic sets of diffusion parameters arenecessary to obtain a sensible pollution of thé virginlayer.

In thé second approach (expérimental), thé diffu-sion of real pollutants is overestimated by thé use oflow molecular weight surrogates including nonpolarspecies which do not interact with PET. The radialconcentration profiles of thé surrogates show that nodiffusion occurred for any surrogate during preformmanufacturing.

Our positive conclusion with PET preforrns cannotbe generalized to any PET-based material. In fact,diffusion is negligible compared to thé thickness ofPET preforms. For thinner materials (generally thécase for coextrusion), thé conclusions could be lessdefinite.

Useful surrogate diffusion data, determined in var-ious polymers at molten state by a mode! test, areavailable in this article and can be used for numericalsimulations.

The authors thank ECO-EMBALLAGES, ADEME and Ré-gion Champagne-Ardenne for grants for P.Y.P. (E-A, A,RCA) and for Y.N.G. (RCA) and for financial support. Partof this work was also supported by thé EU DG Research(PAIR 984318 research program).

Références1. Bayer, F. L. Food Addit Contam 1997, 14, 661-670.2. Franz., R.; Huber, M.; Piringer, O.; Damant, A. P.; Jickells, S. M.;

Castle, L. J Agric Food Chem 1996, 44, 892-897.3. Piringer, O.; Franz, E.; Huber, M.; Begley, T. H.; McNeal, T. P. J

Agric Food Chem 1998, 46, 1532-1538.4. Franz, R.; Huber, M.; Piringer, O. Food Addit Contam 1997, 14,

627-640.5. Pérou, A. L.; Vergnaud, J. M. Plast Rubber Compos 1999, 28 (2),

74-79.6. Pérou, A. L.; Laoubi, S.; Vergnaud, J. M. J Appl Polym Sci 1999,

73 (10), 1939-1948.7. Reynier, A.; Dole, P.; Feigenbaum, A. Food Addit Contam 2002,

19 (1), 89-102.8. Vergnaud, J. M. Liquid Transport Processes in Polymeric Ma-

terials. Modeling and Industrial Applications; Prentice Hall:Englewood Cliffs, NJ, 1991.

9. Morikawa, J.; Hashimoto, T. Polymer 1997, 38, 5397-5400.10. Montgomery, J. H.; Welkom, L. M. Groundwater Chemicals

Desk Référence; Lewis Publishers: Boca Raton, FL, 1991.11. Weast, R. C; Astle, M. J.; Beyer, W. H. CRC Handbook of

Chemistry and Physics; CRC Press: Boca Raton, FL, 1986-1987.12. Budavari, S.; O'neil, M. J.; Smith, A.; Heckelman, P. E.; Kin-

neary, J. F. The Merck Index, 12th éd.; Merck Research Labora-tories, Division of Merck & Co.: Whitehouse, NJ, 1996.

13. CIBA, Technical da:a of BHT; Raschig AG, Sparte Chemie,Mundenheimer Str. 100, D-67061 Ludwigshafen.

14. Pennarun, P. Y.; Dole, P.; Feigenbaum, A.; Saillard, P. FoodAddit Contam to appear.

15. Pennarun, P. Y.; Dole, P.; Feigenbaum, A. J Appl Polym Sci toappear.

16. Brandrup, J.; Immergut, E. H. Polymer Handbook, 3rd éd.;Wiley-Interscience: New York, 1989.

17. Feigenbaum, A.; Ducruet, V.; Riquet, A. M.; Scholler, D. J ChemEduc 1993, 883-886.

18. Reynier, A.; Dole, P.; Feigenbaum, A. J Appl Polym Sci 2001, 82(10), 2434-2443.