Flame retarding poly(methyl methacrylate) with phosphorus-containing compounds: comparison of an additive with

a reactive approach

Dennis Pricea,*, Kelly Pyraha, T. Richard Hulla, G. John Milnesa, John R. Ebdonb,Barry J. Huntb, Paul Josephb, Christopher S. Konkelb,1

aInstitute of Materials Research, Cockroft Building, University of Salford, Salford M5 4WT, UKbThe Polymer Centre, Department of Chemistry, University of Sheffield, Sheffield S3 7HF, UK

Abstract

The flame retardance and thermal stability of a methyl methacrylate (MMA) copolymer reactively modified by copolymerisation

of the MMA with diethyl (methacryloyloxymethyl) phosphonate (DEMMP) have been compared with those of poly(methylmethacrylate) (PMMA) containing equivalent amounts of the additive diethyl ethyl phosphonate (DEEP). DEEP can be regardedas having a structure similar to that of a DEMMP comonomer unit and therefore the two compounds might be expected to confer

about the same levels of flame retardance to PMMA when used at similar concentrations. The incorporation of 3.5 wt.% phos-phorus in both cases raises the limiting oxygen index of PMMA from 17.2 to over 22. However, cone calorimetry shows that theMMA/DEMMP copolymer is inherently more flame retardant than PMMA containing DEEP: the former has a significantly lower

peak rate of heat release than the latter (449 and 583 kW m�2, respectively) and gives rise to a greater amount of char. Thermo-gravimetric analysis (TGA) of the polymers indicates also that the MMA/DEMMP copolymer is more thermally stable thanPMMA whilst PMMA containing DEEP is less thermally stable. Dynamic mechanical thermal analysis (DMTA) shows that theMMA/DEMMP copolymer has physical and mechanical properties similar to those of PMMA, whilst the low molecular weight

DEEP plasticises PMMA, resulting in a significantly reduced glass transition temperature, Tg. A condensed phase mechanism offlame retardance in MMA/DEMMP has been identified. # 2001 Elsevier Science Ltd. All rights reserved.

Keywords: Poly(methyl methacrylate); Phosphorus; Flame retardance; Diethyl ethyl phosphonate; Diethyl (methacryloyloxymethyl) phosphonate

1. Introduction

Chain reaction polymers, such as poly(methyl metha-crylate) (PMMA), are traditionally rendered flame retar-dant by the physical incorporation of additives, and theuse of phosphorus-containing compounds as such is wellestablished [1]. The incorporation of additives howeverhas several disadvantages. The additive is often requiredin high loadings to be effective (typically 10–40 wt.%),which may result in adverse changes to the physical andmechanical properties to the polymer. This can oftenresult in the polymer being no longer suitable for a parti-cular end-use. Additives can also be leached from thepolymer through normal service and ageing, which poses

an environmental threat and, as a result, reduction in anyflame-retardant effect. An alternative approach is thechemical incorporation of the flame-retardant species viacopolymerisation or some other type of chemical mod-ification. The relatively low loadings required to achievesufficient flame retardance, and careful selection of thecomonomer, may keep detrimental changes to the physi-cal and mechanical properties of the polymer at anacceptable level. Furthermore, the flame retardant is thennot easily lost from the polymer, eliminating one of themajor problems associated with the additive approach.

Chemical incorporation of a phosphorus-containingflame retardant into a polymer may promote cross-linkingand char formation during combustion, i.e. a condensedphase mechanism of flame retardance. A condensed-phasemechanism in which phosphorus is retained in the charmay be preferred to a vapour-phase mechanism so as notto unnecessarily increase the toxicity of the combustionproducts [1].

0141-3910/01/$ - see front matter # 2001 Elsevier Science Ltd. All rights reserved.

PI I : S0141-3910(01 )00184-7

Polymer Degradation and Stability 74 (2001) 441–447

www.elsevier.com/locate/polydegstab

* Corresponding author.

E-mail address: d.price@Salford.ac.uk (D. Price).1 Present address: ICI STG, Wilton Centre, PO Box 90, Mid-

dlesbrough, Cleveland TS90 8JE, UK.

Phosphorus-containing components have alreadybeen used in the synthesis of several flame-retardant,step-reaction polymers, e.g. polyesters [2–5], poly-urethanes [6–8], and epoxy resins [9–15]. However, theiruse in the synthesis of chain reaction polymers is muchless well developed, although there has been consider-able pioneering work in this area by Allen and co-workers who have demonstrated flame retardance inseveral chain reaction polymers containing vinyl phos-phazenes as comonomers [16]. As part of an ongoingprogramme to examine the efficacy of phosphorus-con-taining monomers as flame-retardant components inchain reaction polymers, especially styrenics and acryl-ics [17–19], we report here some studies of the thermaldegradation and combustion of poly(methyl methacry-late-stat-diethyl(methacryloyloxymethyl) phosphonate)(MMA/DEMMP copolymer). DEMMP (I) and analo-gous phosphonate monomers have been described in theliterature as potential components of adhesives andflame retardant plastics [20–22]. However, to the best ofour knowledge, no detailed examinations of the proper-ties of acrylic polymers based upon DEMMP have beenreported, although we have already published some datarelating to the thermal stability and flame retardance ofMMA-DEMMP copolymers [23,24]. Here we report onthe thermal degradation and combustion behaviours ofPMMA containing the additive diethyl ethyl phospho-nate (DEEP, II) and compare these with the corre-sponding behaviours of PMMA and MMA/DEMMPcopolymer, allowing a better assessment of the relativemerits of the reactive and additive approaches to flameretarding acrylic plastics. We regard DEEP as having astructure more closely similar to that of the DEMMPcomonomer unit than the flame retardant additives trie-thyl phosphate and diethyl hydroxymethyl phosphonate,on which we have previously reported [24].

2. Experimental

2.1. Diethyl (methacryloyloxymethyl) phosphonate(DEMMP)

DEMMP was synthesized by the condensation ofmethacryloyl chloride with diethyl hydroxymethyl phos-phonate following, with minor adaptations, the method ofLiepins et al. [25] We have previously reported in detailthe preparation and purification of this monomer and sodo not repeat it here [24].

2.2. Synthesis of polymer plaques

Plaques of PMMA suitable for cone calorimetricexperiments were prepared in the laboratory as follows.A mixture of methyl methacrylate (MMA) (ca. 350 ml),benzoyl peroxide (3 g l�1) and lauryl peroxide (1.5 g l�1)was purged with nitrogen for 10 min and then heatedunder nitrogen, with stirring to 70 �C. When the mixturebecame sufficiently viscous (typically 1 h) it was pouredinto a cell made from two thick, borosilicate glass plates,separated around their edges by a silicone rubber gas-ket. The assemblies were placed in an air oven and theplaques cured at 40 �C for 24 h, 60 �C for 8 h and 80 �Cfor 24 h. Optimum curing of the plaques was deter-mined by trial and error. The resulting plaques mea-sured ca. 120�120�4 mm and were cut to size for conecalorimetry experiments.

Plaques of PMMA containing DEEP were preparedas above with a minor adaptation. DEEP was added tothe viscous prepolymer mixture (prior to pouringbetween the glass plates) and left to stir for ca. 5 min,allowing the mixture to become homogeneous. Theamount of DEEP added was calculated to yield a pla-que containing 3.5 wt.% phosphorus. Curing was car-ried out in a manner identical to that for PMMA.

Plaques of MMA/DEMMP copolymer containing 10mol% DEMMP (corresponding also to 3.5 wt.% phos-phorus) were prepared in the laboratory by heating de-oxygenated mixtures of MMA, DEMMP and azoisobu-tyronitrile (2 g l�1) between glass plates separated aroundtheir edges with a silicone rubber gasket. Heating to effectcure was carried out in an air oven, first at 80 �C for 2 h,then at 100 �C for 2 h and finally at 120 �C for 2 h.

2.3. Limiting oxygen indices

Limiting oxygen indices (LOI-ASTM-D-2863) weremeasured on a Stanton-Redcroft FTA flammability uniton off-cuts of polymer plaque measuring 100�6�4 mm.

2.4. Thermogravimetric analysis

Thermogravimetric analyses (TGA) were carried outon ca. 10 mg samples contained in silica crucibles on a

442 D. Price et al. / Polymer Degradation and Stability 74 (2001) 441–447

Mettler TG 50 thermogravimetric analyser under nitro-gen and in air at a heating rate of 10 �C min�1. The samplewas prepared by removing fine scrapings of material fromthe edge of a polymer plaque with a sharp blade.

2.5. Cone calorimetry

The combustion behaviour under ventilated condi-tions was measured using a Fire Testing Technologycone calorimeter, in conformance with ISO DIS 5660[26]. Testing was performed on 100�100�4 mm plaquesat a heat flux of 35 kW m�2. Ideally, the specimenshould be thermally thick for cone calorimetry experi-ments. However, the procedure for preparing the pla-ques is in the early stages of development and the depth ofthe rubber gasket determines sample thickness. In additionto this, synthesis of phosphorus-containing comonomers isvery time consuming and preparing thicker samples wasnot within the scope of this investigation. Experimentswere carried out at a heat flux of 35 kW m�2 which is therecommended irradiance for exploratory testing [27]. Theplaques were placed in a sample holder with retainerframe, resulting in 88 cm2 of the sample surface beingexposed to the radiating cone heater. Each test was carriedout twice to ensure the results obtained were reproducible.Optimally, the test should be repeated at least three times,but limited quantities of DEMMP allowed for only twoplaques of the MMA/DEMMP copolymer to be pre-pared and tested.

2.6. Dynamic mechanical thermal analysis (DMTA)

DMTA was carried out on a Polymer LaboratoriesDMTA Mark II instrument at a fixed frequency of 1 Hzand at a heating rate of 5 �C min�1 over the temperatureranges 35–180 �C on small pieces of polymer plaquesheld in a single-point cantilever clamp. The principal

data recorded were the temperatures corresponding tothe maximum values of tan d, which approximate to thezero frequency glass transition temperatures, Tg.

3. Results and discussion

3.1. Combustion behaviour

Limiting oxygen indices of the PMMA, MMA/DEMMP copolymer and of the PMMA containingDEEP are given in Table 1.

It can be seen that the incorporation of just 3.5 wt.%phosphorus increases the LOI of PMMA from 17.2 toabove 22 vol.%, indicating a reduction in flammability.The LOIs obtained for the PMMA with the DEEP addi-tive and the DEMMP reactive are comparable at 22.4 and22.8 vol.%, respectively. However, whilst PMMA burnswith a clean, non-smoky flame, both phosphorus-con-taining polymers produce smoke during combustion.

Whilst measurement of LOI is a useful, small-scaletest that highlights and ranks flame retardance in poly-mers, it is not a reliable indicator of how a material willperform in a real fire. The method of choice for this is thecone calorimer, in which a polymer plaque is irradiated ata pre-determined heat flux, simulating the conditions ofan advancing flame front. Fire parameters including timeto ignition, heat release rate, carbon monoxide, carbondioxide yields, and smoke yield are obtained from a singleexperiment. Some important parameters obtained forplaques of PMMA and of the modified polymers aregiven in Table 2; some of the data are also presented asa function of time in Fig. 1.

Both the PMMA containing DEEP and the MMA/DEMMP copolymer have longer times to sustainedignition than PMMA, although flashing is observed forthe PMMA containing the additive DEEP. The resultsshow that the modified polymers have reduced peak ratesof heat release (RHR) as compared to the homopolymer,PMMA. However, the modified polymers produce farmore smoke and carbon monoxide during combustionthan does PMMA. This is due to incomplete combustionand is indicative of a substantial vapour phase compo-nent to the mechanism of flame retardance. It is clearfrom the results that whilst both phosphorus-containing

Table 1

Limiting oxygen indices

Sample Wt.% P LOI (vol.% O2)

PMMA 0 17.2

MMA/DEMMP 3.5 22.8

PMMA+DEEP 3.5 22.4

Table 2

Cone calorimetric data measured with an irradiance of 35 kW m�2

Parameter PMMA MMA/DEMMP PMMA+DEEP

Time to ignition (s) 50�3 60�2 63�3

Peak heat release rate (kW m�2) 641�5% 449�5% 583�5%

Peak carbon monoxide yield (kg kg�1) 0.016�0.0015 0.171�0.002 0.109�0.0015

Peak carbon dioxide yield (kg kg�1) 3.90�0.4 2.67�0.1 2.85�0.2

Peak specific extinction area (m2 kg�1) 179�38 736�40 389�46

Residue (wt.%) 1.3�0.5 6.7�0.5 2.7�0.8

D. Price et al. / Polymer Degradation and Stability 74 (2001) 441–447 443

polymers are inherently more flame retarded than thehomopolymer, there are significant differences in theflame-retardant characteristics of the MMA/DEMMPcopolymer and the PMMA containing DEEP.

The peak rate of heat release of a polymer is consideredto be one of the most important parameters in assessingpotential behaviour in a real fire. The peak rate of heatrelease for PMMA is high at 641 kW m�2. The PMMAcontaining the additive, DEEP, has a slightly lower peakrate of heat release at 583 kW m�2, which cannot be con-sidered as a significant reduction. The MMA/DEMMPcopolymer however, has a significantly reduced rate ofheat release as compared to the PMMA containingDEEP at 449 kW m�2. As MMA/DEMMP and PMMAplus DEEP contain 3.5 wt.% phosphorus in similarchemical environments, the reduction in the rate of heatrelease observed in the former must be attributed to thereactive incorporation of the phosphonate units. Fur-ther evidence for increased flame retardant efficiency inthe MMA/DEMMP copolymer as compared to PMMAplus DEEP comes from the residual masses obtainedafter combustion. Both PMMA and the PMMA con-taining DEEP combust to leave little residue. The MMA/DEMMP copolymer, on the other hand, retains over 6.5wt.% of its mass as a carbonaceous residue or char. Thisindicates also a condensed-phase component to themechanism of flame retardance in MMA/DEMMP inwhich fewer volatile products of degradation fuel theflame and more of the polymer is retained as char. Theprobable condensed-phase mechanism of flame retar-dance in MMA/DEMMP has been described in detail

elsewhere [24]. In summary, ethene is eliminated fromthe DEMMP comonomer units leaving bound phos-phonic acid species which transesterify with MMA unitsto give methacrylic acid (MA) units. The MA units thendehydrate and probably react with other MMA units,via alcoholysis, to form anhydride links, disrupting thechain unzipping of PMMA. The anhydride sequencessubsequently decarboxylate, resulting in unsaturated charprecursors. This is in contrast to the cross-linkingmechanism leading to char formation proposed by Loma-kin et al. [28] from studies of the thermal degradation oftwo (trimethylolpropane triacrylate and trimethylolpro-pane trimethacrylate) network copolymers of methylmethacrylate.

3.2. Thermal degradation behaviour

The thermal degradation behaviours of the PMMAcontaining DEEP and of the MMA/DEMMP copoly-mer have been assessed by dynamic thermogravimetric(TG) analysis under nitrogen and in air at a heating rateof 10 �C min�1. The TG curves and derivative TGcurves are given in Figs. 2 and 3. Clearly, the thermaldegradation behaviour of the phosphorus-containingpolymers is significantly different to that of the homo-polymer, PMMA.

The thermal degradation mechanism of PMMA is sim-ply a reverse of the polymerisation, which occurs afterchain scission [29]. The depolymerisation of polymer toevolve monomer is termed ‘‘unzipping’’. The degrada-tion of PMMA is however influenced by the type of

Fig. 1. (a) Rate of heat release (RHR), (b) rate of carbon monoxide production, (c) mass loss and (d) rate of smoke production. All measured at an

irradiance of 35 kW m�2. Key: PMMA ( ), MMA/DEMMP ( ), PMMA+DEEP ( ).

444 D. Price et al. / Polymer Degradation and Stability 74 (2001) 441–447

polymerisation, i.e. free radical or ionic [29,30] and alsoby the presence of comonomers. Current TGA experi-ments have shown that thermal degradation (pyrolysis)of the PMMA under nitrogen occurs after 145 �C at arelatively slow rate. At temperatures above 200 �C, therate of degradation increases slightly. The rate of pyr-olysis then rapidly increases after 300 �C, and by 400 �C,the pyrolysis is complete, resulting in 100% mass loss.In air, significant degradation of the PMMA is notobserved until 230 �C. At temperatures above 230 �C,the degradation of PMMA is rapidly increased and100% mass loss occurs below 380 �C.

The MMA/DEMMP copolymer starts to pyrolyseabove 150 �C, but at a much slower rate than thehomopolymer, PMMA. The DTG curve for pyrolysis ofthe MMA/DEMMP copolymer contains two peaks,indicating that the degradation mechanism is differentto that of PMMA, for which there is only a single peak

in the DTG curve. TGA/DSC/MS runs of the DEMMPcopolymer has shown that the first peak on the DTGcurve corresponds to loss of MMA and C2 hydro-carbons. The second peak corresponds to loss of COand water, which is consistent with the proposedmechanism outlined above. A relatively stable carbo-naceous residue or char is formed during the pyrolysisequal to 5.0 wt.% at 500 �C. TGA of the DEMMPcopolymer in air shows that degradation occurs after200 �C, which is 50 �C higher than for the degradationunder nitrogen. In air, the TG curve has a third broadpeak that is not observed during the degradation undernitrogen. The peak ranges from approximately 450–570 �C and is probably due to char oxidation.

The TG curves obtained for the pyrolysis of thePMMA containing DEEP are again different to those ofboth PMMA and the MMA/DEMMP copolymer.Under nitrogen, weight loss from PMMA plus DEEP

Fig. 2. (a) TG and (b) DTG curves recorded for pyrolyses under nitrogen at 10 �C min�1. Key: PMMA ( ), MMA/DEMMP ( ),

PMMA+DEEP ( ).

D. Price et al. / Polymer Degradation and Stability 74 (2001) 441–447 445

begins at ca. 95 �C. This is significantly lower than thetemperature at which PMMA begins to degrade and isalmost certainly a consequence of the volatilisation ofDEEP. At 310 �C, with only 63 wt.% of the samplemass remaining, the rate of degradation increasesresulting in 100% mass loss by 400 �C.

3.3. Mechanical properties

The physical and chemical modification of PMMAshould not affect the mechanical properties of the poly-mer to such an extent that it would be unsuitable for aparticular end-use. One method to assess the effects ofthe phosphorus-containing compounds on the mechan-ical properties is to determine the glass transition tem-perature (Tg) of the modified polymers. This was achievedby dynamic mechanical thermal analysis (DMTA and theresults are given in Table 3.

The incorporation of DEEP significantly plasticisesPMMA as indicated by the marked reduction in Tg

relative to that of PMMA. The chemical incorporationof the flame-retardant moiety in the form of the como-nomer, DEMMP, however, only slightly reduces the Tg

from 124 to 117 �C. Thus, these Tg values indicate thatthe reactive approach to flame retardance would appearto only slightly affect the mechanical properties ofPMMA in contrast to the additive approach, which hasa significant effect. Further measurements of other

Fig. 3. (a) TG and (b) DTG curves recorded for pyrolyses in air at 10 �C min�1. Key: PMMA ( ), MMA/DEMMP ( ).

Table 3

Tg determined by DMTA

Sample Tg (�C)

PMMA 124

MMA/DEMMP 117

PMMA+DEEP Ca. 70

446 D. Price et al. / Polymer Degradation and Stability 74 (2001) 441–447

parameters of mechanical strength are required to con-firm this.

4. Conclusions

The LOI data indicate that the incorporation of just 3.5wt.% phosphorus in PMMA either in the form of theadditive diethyl ethyl phosphonate, DEEP, or in the formof the comonomer, diethyl (methacryloyloxymethyl)phosphonate, DEMMP, reduces flammability. However,cone calorimetry indicates that the MMA/DEMMPcopolymer is more flame retardant than the PMMA con-taining DEEP in terms both of reduced rate of heatrelease and of increased char formation. Flame retardanceof the MMA/DEMMP copolymers occurs by both con-densed (char-forming) and vapour-phase mechanisms.The condensed phase mechanism is believed to arise bythe formation of anhydride links which disrupt the ther-mal unzipping and hence reduce the rate of fuel evolu-tion. At present, the mechanism of the vapour-phaseprocess is not known, but is likely to be due to the parti-cipation of phosphorus-containing free radicals in radicalpropagation processes leading to an overall reduction inrates of propagation and hence in rates and efficiencies ofcombustion. Flame retardance of the PMMA containingDEEP occurs overwhelmingly in the vapour phase as verylittle char is produced during combustion. Again, themechanism is unknown but is most likely to be due to theparticipation of phosphorus-containing radicals in gas-phase chain reactions. In addition to improving flameretardance in PMMA, the chemical incorporation ofDEMMP has little effect on the physical and mechanicalproperties. The physical incorporation of the additiveDEEP, however, significantly degrades the properties ofPMMA as shown by the reduced glass transition tem-perature and lower thermal stability, whilst only slightlyreducing the flammability of PMMA as compared toDEMMP.

Acknowledgements

We thank the Engineering and Physical SciencesResearch Council (EPSRC), the Ministry of Defence(MoD) and ICI for financial support. (Grant Nos. GR/L85879 and GR/L85886), Dr. N.R. Bosley of LGC

(North West), Runcorn, Cheshire, for supplying theTGA/DSC/MS data.

References

[1] For a review of this area, see, for example: Ebdon JR, Jones MS. In:

Salamone JC, editor. Polymeric materials encyclopedia. Boca

Raton: CRC Press, 1996. p. 2301–411 (and references cited therein).

[2] Delaviz Y, Gungor A, McGrath JE, Gibson HW. Polymer 1992;

33:5346.

[3] Ma ZL, Zhao WG, Liu YF, Shi JR. J Appl Polym Sci 1997;

63:151.

[4] Wang CS, Shieh JY, Sun YM. J Appl Polym Sci 1998;70:1959.

[5] Wang CS, Lin CH, Chen CY. J Polym Sci, Part A, Polym Chem

1998;36:3 051 .

[6] Lee FT, Nicholson P, Green J. J Fire Retard Chem 1982;9:194.

[7] Sivriev C, Zabski L. Eur Polym J 1994;30:509.

[8] Liu YL, Hsiue GH, Lan CW, Chiu YS. J Polym Sci, Part A,

Polym Chem 1997;35:1769.

[9] Derouet D, Morvan F, Brosse JC. J Appl Polym Sci 1996;

62:1855.

[10] Liu YL, Hsiue GH, Chiu YS. J Polym Sci, Part A, Polym Chem

1997;35:565.

[11] Cho CS, Chen LW, Chiu YS. Polym Bull 1998;41:45.

[12] Wang CS, Shieh JY. Polymer 1998;39:5819.

[13] Buckingham MR, Lindsay AJ, Stevenson DE, Muller G, Morel

E, Coates B, Henry Y. Polym Degrad Stab 1996;54:311.

[14] Levchik SV, Camino G, Luda MP, Costa L, Muller G, Costes B,

Henry Y. Polym Adv Technol 1996;7:823.

[15] Cho CS, Fu SC, Chen LW, Wu TR. Polym Int 1998;47:203.

[16] Allen CW, Bahadur M. Phosphorus Sulfur 1993;76:463.

[17] Banks M, Ebdon JR, Johnson M. Polymer 1993;34:4547.

[18] Banks M, Ebdon JR, Johnson M. Polymer 1994;35:3470.

[19] Ebdon JR, Joseph P, Hunt BJ, Price D, Milnes GJ, Gao F. In:

Lewin M, editor. Recent advances in flame retardancy of poly-

meric materials, vol VIII. Norwalk: BCC, 1997.

[20] O’Brien JL, Park E, Lane CA. US Patent 2,934,555, 1960.

[21] O’Brien JL, Park E, Lane CA. US Patent 3.030,347, 1962.

[22] Moszner N, Zeuner F, Fischer UK, Rheinberger V. Macromol

Chem Phys 1999;200:1062.

[23] Price D, Pyrah K, Hull TR, Milnes GJ, Woolley WD, Ebdon JR,

Hunt BJ, Konkel CS. Polym Int 2000;49:1164.

[24] Ebdon JR, Hunt BJ, Joseph P, Konkel CS, Price D, Pyrah K, et

al. Polym Degrad Stab 2000;70:425.

[25] Liepens R, Surles JR, Morosoff N, Stannett V, Duffy JJ, Day

FH. J Polym Sci 1978;22:2403.

[26] ISO DIS 5660: Fire test—reaction to fire—rate of heat release

from building products, 1990.

[27] STD.BSI DD 246-ENGL 1999: Recommendations for the use of

the cone calorimeter.

[28] Lomakin SM, Brown JE, Breese RS, Nyden MR. Polym Degrad

Stab 1993;41:229–43.

[29] McNeill IC. Comprehensive polym. sci., vol. 6/15. Pergamon

Press, 1989. p. 458–62.

[30] Kashiwagi T. Macromolecules 1986;19:2160.

D. Price et al. / Polymer Degradation and Stability 74 (2001) 441–447 447