You are on page 1of 8

Rheologica Acta

Rheol Acta 32:57-64 (1993)

A note on estimating mastercurves
J. Honerkamp and J. Weese
Universit~t Freiburg, Germany

Abstract: There are several quantities in rheology which show a scaling behavior.
One well known example is the time-temperature superposition principle of material functions characterizing the linear viscoelastic properties of polymer melts.
We propose a mathematical shift procedure for the calculation of mastercurves
and the corresponding scaling factors from experimental data which show such
a scaling behavior. In order to demonstrate the applicability of the shift procedure mastercurves and scaling factors are determined for material functions of
several polystyrene melts and for the specific viscosity of polyisobutylene in
cyclohexane.

Key words: Scaling behavior - mastercurve - time-temperature superposition
principle - specific viscosity

1. Introduction

It is well known that material functions characterizing the linear viscoelastic properties of polymer melts
or solutions in the terminal and plateau region show
a scaling behavior: by scaling the experimental data
for a material function measured at different temperatures a unique mastercurve referring to an arbitrary
reference temperature To can be obtained (Ferry,
1980). For the storage modulus G~r(co) this so called
time-temperature superposition principle can be expressed by the equation

G'r(o) ) : CTo( T) G'ro (aro ( T)co ) .

(1)

Thereby the scaling factors ar0(T) and CTo(T) have
been introduced.
These material functions are not the only quantities
showing such a scaling behavior. The specific viscosity r/~t(c) of polymer solutions in dependence on the
concentration c is another example. With appropriate
scaling factors aMo(M) and CMo(M) the specific
viscosity t/~t(c) is independent of the molar mass M
of the polymer sample (Utracki and Simha, 1963;
Baloch 1988a, 1988b):

r/~(c) = CMo(M) rl~o(aMo(M) c)

(2)

Thereby, M0 denotes an arbitrary reference molar
mass.
In practice there are two reasons for the determination of scaling factors from experimental data obtained at different temperatures or for different molar
masses. First, the measuring range of an experimental
device can be extended by the calculation of mastercurves. Furthermore, the scaling factors provide information about physical properties of the system.
Unfortunately, there is no well-known mathematical method for the calculation of scaling factors and
graphical methods are used to estimate them. For that
reason we propose and discuss a mathematical shifting procedure in this publication.
In the following section the shift procedure is introduced. In Sect. 3 this procedure is used to calculate
the scaling factors for material functions of several
polystyrene samples. As another application example
the scaling behavior of the specific viscosity of a
polymer solution is studied in Sect. 4.

2. Shift procedure

The first part of this section includes a mathematical formulation of the problem which must be
solved in order to calculate shift factors from experi-

. . = g2j(tj. polynomials with . 7~o(2m) and For other choices the quantity Z 2 has no local minimum. (5) X (Tj. 1 (1993) mental data. i+r&(2j))+Y~o(2j))) 2 j = O i = 1 •j. the determination of the shift factors yao(21) . The shift procedure itself is presented in the second part. These problems can be avoided by the use of a polynomial. . A m in the following text. . the data are affected by absolute errors of size a. .0 e 8) with respect to y~0(21). these data are considered as realizations of the random variables Ga. aj. For every parameter 2j there is therefore a data set [g~. however.~o(2m) and r~0(2i) . . i + 72). ..g.. .o(t) is. . Because of the experimental'noise. .i = 1. ra0(2 ) and 7~0(2) correspond with the logarithm of the scaling factors ar0(T) (aMo(M)) and CTo(T ) (CMo(M)) and satisfy the equations rzo(2o)=O Z 2= ~ -Tf-(g~. . (If e. standard normally distributed.i-(gao(tj. % : in addition to the shift factors 7 z o ( 2 0 . i = 1 .. . aj.i (6) with respect to Yao(21) .~o(21).i. . . . ). . .y~o(2m) and rao(21) .. .} for the function gaj(t) at the points {tj i. (1) or (2)) on a double logarithmic scale. If.58 Rheologica Acta. random variables and the aj. . 1 . ~. not known. i) the data are affected by relative errors of the size a. . r{0(2m) for the shift factors would be obtained by minimizing 2. however. . .. . If e. .In practice. . z 2 =j =2O i2= l (4b) In practice. rao(2m). 1989) are used the quantity ~ 2 h a s discontinuities which lead to serious problems in practice. ~'). 1986) could be used and estimates y{0(21).T For the approximation of the function g~o(t) spline-functions seem to be more appropriate. Vol. 7~o(2p) and rzo(21) . We propose therefore to approximate the function g~o(t) by a polynomial np g~o(t) -~ ~ ckt k and to define the estimates y~o(21) . The function g. . 1"20(21) . . . . .i = 1 . . t represents the logarithm of the frequency co (concentration c) and 2 denotes the temperature T (molar mass M). y. In the last part of this section simulated data are used to test the shift procedure. . Concerning the shift procedure described so far some remarks are necessary: - In this equation the ej. . . .. . Z'X0(2m) and c0. .2 m . i specify the error of each data point up to an usually unknown scaling factor a. (9) j=O 2. On this scale the scaling becomes a shift and the equivalent to the scaling relation can be written as gx (t) = g~0 (t + rXo(2) ) + Y*0(2) . nj} available. i = l. .2"20(2m) by t h e m i n i m u m o f (4a) m ~ yao(20) = 0 . 7ao (Am) and rZo(21) . the function g~(t) is measured for several parameters 2 denoted by 20 . j. . igj. . .1 Formulation o f the p r o b l e m For the definition of the shift procedure it is convenient to consider the scaling relation (Eqs. linear or cubic splines (Stoer. "l'~to(2m)must be calculated. j = O. . . . . . (3) In this notation ga (t) represents the logarithm of the storage modulus G~(co) (specific viscosity t/~t(c)). i g j... r~o(2m) the function gzo(t) itself has to be fitted to the data. . h e range in which the degree np of the polynomial can be chosen is limited by m l<np< ~] n j . . i are independent.nj. 32. i) + ff ffj. . . . . . . i -- . . nj. .0(2m) would be no problem: a simple least-squares method (Press et al. . .2 Estimating mastereurves by optimization If the function g~o(t) would be known. No. . .'20(2m). i= g2j(tj. .')~o(2m) and G ¢7 ~')~0(21)' " . . . . . . i .. m (7) k=0 . . . . .g. . .) From these data the shift factors y. T~0(21) . .

642 + 0.023 1.019 0.0 I z i I z 36. 2.. ...036 1. Shift factors r~(j) and ?~(j) calculated from the simulated data (Fig.024 1.023 0.100 ..032 0. The solid line marks the reference curve go(t) Table 1...049 + 0. I I -2"0-00 ~...003 + 0.666 + 0. tj.5 3. .0 0.068 0.648 _ 0.1.677 + 0.055 1.0 10.499 _ 0.062 _+0.. the estimates show o n l y a slight d e p e n d e n c e o n 10.069 + 0.940 _+0.034 0. 3..075 _ 0.... 2... A note on estimating mastercurves m 59 2..5 t t Fig. 1974).667 _+0.3). Shift factors r~(j) and y~(j) calculated from the simulated data (Fig.020 1.0 2.929 + 0..5 3.1 .054 + 0.667 . 1) in dependence of the degree np of the polynomial data. E a c h d a t a set was obtained by e v a l u a t i n g the shifted function g 0 ( t + r 0 ( J ) ) + 7 0 ( J ) (Fig.. I n addit i o n .o .o- led to good results. . .020 - 1..5 S 5.035 0.019 - exact 0.0 I zIzIzzzzIIIIzzIzIIzzIz~Izz~z~=Iz=zI=~I~=' - 5. Figure 2 shows the estimates for the shift factors in d e p e n d e n c e of the degree o f the p o l y n o m i a l . 1) for different degrees r/p of the polynomial..Honerkamp and Weese.0 Z l ] ~zzSO np Fig.667 1. I n this r a n g e the estimates for the shift factors were n e a r l y i n d e p e n d e n t of the degree o f the p o l y n o m i a l (see Sect..5 7.041 1.041 1.031 0.037 0.662 + 0. Simulated data sets used for the test of the shift procedure...(2) 1.677 _ 0...035 0.049 + 0.943 _+0.0 I I I I i 4().724 ___0.023 1.0 l l l 25. T h e n a G a u s s i a n r a n d o m n u m b e r corr e s p o n d i n g with a n absolute error of 0.5 °°0.0 l l l l 15.(2) I l l l 5J.020 0.022 0.0- m nj/20<_np<_ ~ nj/4 j=o (10) j=o "r.042 1.40.0 1. 1.676 _+0.924 + 0.072 +_0.024 1. 3~o 2.023 0.933 + 0. The last row contains the exact values used to simulate the data np z~(1) 2 3 4 5 10 20 30 40 0...961 + 0. Mastercurve obtained for the simulated data with np = 10.119 1. solid line) at the points tj.659 + 0.033 0.(1) 010- _. F u r t h e r m o r e . 1.0 2.229 _+0.0.035 .672 ___0..3 Test of the method with simulated data F o r the test of the shift p r o c e d u r e three d a t a sets have been simulated (Fig.. The solid line marks the reference curve go(t) Fig.043 1..O l l l l l l l 1(3.5 2.641 + 0. T a b l e 1 includes some n u m e r i c a l values for the estimates together with the c o r r e s p o n d i n g errors (Bard..934 _+0. ... The resulting mastercurves (Fig.0 i i I I I 35.042 1. 1).653 + 0.5 10 1. F o r 4_<np_<45 the estimates are i n good a g r e e m e n t with the values used to simulate the ~" / 1-.030 1.025 1...670 + 0.90 73(1) r~(2) y~(2) 1.022 0.633 + 0..0 7. 3) are therefore a good a p p r o x i m a t i o n of the f u n c t i o n go(t).Q l l l I I I l 2()...05 was added.

No.3 )~lw/MN PS 1 PS 2 PS 3 PS 4 PS 5 PS 6 39 70 128 275 770 3000 1. The results for the mastercurves and the corresponding scaling factors aro(T) and CTo(T) are shown in Figs.m. The vertical scaling factors CT0(T) (Fig.To~ Table 2. 32. Molar mass and polydispersity index of the nearly monodisperse PS samples (Schausberger et al. It should be mentioned that c(To)(To. Application to material functions @ lO~ In this section the shift procedure is used to calculate mastercurves for material functions of six nearly monodisperse PS samples and a commercial PS sample. 10 -6 I i 10 -s i. 4 and 5 and Table 3. +PS3.To~) = ( ..m.05 1o' o o ~ +-~o a #oo O~o × e a i 10-~ b iiiiIiq 10-~ i LIIIHr I 10-s i iiiiiii I 10-~ i i HL]II I 10-3 I I IIIIII I I I I]][11 10-~ 10- I I IIIlll I I I IIIILI I 100 ld I 11111111 102 I IIIIII1~ ~03 ~ (s-') Fig.2 Results for a commercial PS sample For the commercial PS sample data for the dynamic moduli and the relaxation modulus were available... The data have been measured at temperatures between 122 ° and 250°C.07 1. 3. x PS2.60 Rheologica Acta. As reference temperature To = 150°C was chosen. solid line) by the WLF-relation (Ferry. 6 and 7 and Table 4. I i.103 (Table 2) and were measured at temperatures between 150 ° and 270 °C. ~..33)°C. 1985) Sample M" 10.05 1.To) (11) T . 4). For polynomials with a higher degree the estimates could not be calculated because of numerical instabilities. Vol. Mastercurves for the dynamic moduli G'r0(co) and G}n(co) (b) of the six nearly monodisperse PS samples (<>'PSI. The results for the mastercurves and the corresponding scaling factors azo(T) and CTo(T) are shown in Figs. + o o ~o~. [] PS6) The appropriate WLF-parameters have been determined by a nonlinear least-squares fit (Press et al. In every example the shift procedure should lead to the same result for the scaling factors. A PS4. In practice this limitation is no disadvantage because the estimates are not expected to become better..uj 1¢ 4 l lumlt 10-3 i luml 10-2 i turn11 10- i in. 1986) and are given by c(To)(To. 'i'~ . o + i. (1985).80_6. ~o~ ~o~ 5 u ~10 ~ 3 k @ D k @ u ® ~ + k @ 2 ~ 2 u 3. The agreement of the results can therefore be considered as a test of the shift procedure.7 6 3 _ 7 8 ) ° C and T= = (44.05 1. .07 1.T=) has been calculated instead of c(T o) because this quantity is independent of the reference temperature To. 4. © PS5. Only the calculation time increases with an increasing degree of the polynomial. As reference temperature To = 180°C was chosen. 'IC13 ~o% 3. For all samples the shift procedure led to smooth mastercurves (Fig. 1980) log aro(T) - c(To)(T. The horizontal scaling factors aro(T) are well represented (Fig.u. I 1d 10 ° i J . 5a) obtained for the six samples are in good agreement with each other. 1 (1993) the degree of the polynomial in this range. 5 b) are within the error given by 1. 5a..06 1. They characterize the dynamic moduli of six nearly monodisperse PS samples with molar masses between 39-103 and 3000.1 Results for six nearly monodisperse PS samples The data used in this section were published by Schausberger et al. The values for the horizontal scaling factors aro(T) (Fig.~ a 2 I i in.

986 + 0.528 _+0.012 1.405 +_0.004 + 0.3 10 -3 10 -4 Table 4.101)" 10 -2 c ~_(160) c br~(180) 1.013 1. + P S 3 .999 _+0. ~ P S 4 . (122) c~r°.10 -1 (6.003 0.005 0.023)" 10 +4 (0.943 -+0.027 PS 4 PS 5 PS 6 a~n (160 ) a~ro(180 ) a~.007 0.017 o 6 0.0 20'0.~(240) a}~(270) (2. 10.0o~ Ii i ~ 7._(122) abr~(133) abr~(143) abr~(153) abr~(164) a~r°.314 +_0. 10 -2 (1.482 -+ 0.010 1. © P S 5 .012 10-1 10 -2 10.3 (1.010).990 _+0.992 _+0.0 1. Scaling factors a}o(T ) and c}n(T ) calculated from the data for the dynamic moduli and the relaxation modulus of the commercial PS sample lo0 G"(o)) and G"(co) G(t) (1.010 0.Honerkamp and Weese.10 +1 (6. ld" 10-4 1400 16'00 18'00 20'0.231 _+0.(200) abT~(220) a-~ (250) 8.007 0.(133) cbr~(143) c~r~(153) c ~° (164) c~r~(200) c ~r~(220) c~(250) 16'0.147 ± 0.014 - 0.583 ± 0.002 0. 109).108 + 0.037).003 0.0 T (oc) 2.320 _+O.068)" 10 -2 id' a~.593 ± 0.3 (1. Scaling factors PS samples a~-o(T) and C~.548 _+0.776_+ 0.428 ± 0.0 26i0.250)" (1.104-+0.567+_0. 10 +4 (1.063 -+ 0.021)'10 +3 (1.833+0.009 0.052)" 10 -2 (4.009 _+0.o(T) calculated 61 from the data for the dynamic moduli of the nearly monodisperse PS 1 PS 2 PS 3 a)n (160 ) a~(180) (2.054).275 ± 0.014 ld 1.026)" 10-1 (2.858 ± 0.919 + O.0 c~.991 +-0.009 0.~ - (2.710 ± 0.993 ± 0. 10 -z (2.286 -+ 0.007 0.0 2210.988 _+0..014 1.386 -+0.10 +2 (2. c ~-n(160) c ~r~(180) c ~_(200) c ~ (220) c ~n(240) c ~-0(270) 0.536_+0.579 ± 0.031). Scaling factors a~.530 +_0.095).090)" 10(2.040).076)" (5.524_+0.003 0. 10-2 . 10.099 -+ 0.007 1.5 0.462 _+0.003 0.003 1.047). 10 -4 - (1.399 + 0.264 -+O.013 1.0 2810.120_+0.0 T (°C) 24'0.0 2210.656 _+0.088)" 10-1 (2.176___ 0.0 26'00 28100 Fig.033).385 _+0.027 _+0.008 0.285 + 0. The solid line (a) marks the temperature dependence of the scaling factor ar0(T) given by the W L F relation .(200) a ~ (220) a~.591 ± 0.058)" (2.002 1.195 ± 0.527_+0. 10 -3 (4.095)" 10 -2 (4.572 ± 0.958_+0.003 0. x P S 2 .877 ± 0.038).062)" 10.001 0.997 -+ 0.008 - 1.890 -+0.027 - 0.029 _+0.026).017)" 10 + 1 (5.008)" 10 +2 (2.o 1.091)" (1.0 2410.0 !8'0.038 _+0.594 _+0. 126).002 0.10-1 (2.026)" 10 +o (1.029 + 0.007 0. [] PS6). 5. A note on estimating mastercurves Table 3.5 8.990 -+0. 10 -2 (1.987 -+ 0.522_+0.002 0.009)' 10 + 3 (1.961 +_0.202 _+0.061)" 10 + 0 (1.012)" 10-1 (5. 10 -3 - (2.003 0. 184).760±0.017:t:0.970 _ 0. 8 o b 1400 0..011).020 +__0.n(T ) and c}n(T) calculated from the data for the dynamic rfioduli of the~nearly monodisperse PS samples ( o PS1.

.~10'~ ..105 and 31.10 s for concentrations between 0.75 ___3. 8 and 9... 1 (1993) For the dynamic moduli as well as for the relaxation modulus good results for the mastercurves (Fig.0 Fig. They vary between 0.6.110-2 .4 g/cm 3 and refer to a temperature of 25 °C..0 24Jo.\ \ '\ lo' 10-5' b ''"'"10-"' '"'"ld z' '"'"~10.~' ''""10- ' ''"'"~10 ° '''"'10' ' '"j'~2 t (4 Fig.. 7b).o 18'o.4. 7a). As reference the sample with a molar mass of M= 1.06) °C were obtained.4 and 1.0 o 107~ ~vo J o"* .. No.. 6) have been obtained. 8) is obtained.. 6. Scaling factors a~. the shift procedure led to horizontal scaling factors ar0(T) which are in good agreement with each other (Fig. 10"1 ' ' ' '"i'~ ' .0 2610...o 16~o. Somewhat amazing is the temperature dependence of the vertical scaling factors cr0(T ) (Fig. As in the examples of Sect. 7. The results for the mastercurve and the corresponding scaling factors aMo(M) and CMo(M) are shown in Figs.4.001 g/cm 3 and 0..10~l . .2 without a systematic temperature dependence. ~0~ ~o~ ~. Vol.(T) (b) calculated from the data for the dy°namic moduli ~nd the relaxation modulus of the commercial PS sample..0 2410.o 22'0. This fact may be attributed to inaccuracies in the preparation of the samples.. These values are in good agreement with the values in Sect. o o J 14100 a aB 1.. For the WLF-parameters the values c(To)(To-T~)= ( . .. ~0~ ° ~0' ~oo 1~' :: ~0-~ 12o.$'00 16'0....q105 (s-b 120..o 26100 1.0 T (oc) 2. - & 105: '".0 T (°C) 22~0...0 18'00 20'0. These data characterize the specific viscosity of five polyisobutylene samples with molar masses between 1.62 Rheologica Acta.738 __+_39) °C and Too = (44.. l0 s was chosen..... The solid line (a) marks the temperature dependence of "the scaling factor aro(T) given by the WLF relation 1if= 4.. 3 the determination of the scaling factors with the shift procedure from experimental data is no problem and a smooth mastercurve (Fig. 3.o 2o'o. In addition.~(T)(a) and c~.1 and lead to a good approximation of the temperature dependence of the horizontal scaling factors aro(T ). 10 -~'''' ' ' '"'"~100 ..0 e e b o o ~cf 10-z .. . Application to the specific viscosity >.o- 0. Mastercurves for the dynamic moduli G'T0(og) and G~0(co) (a) and the relaxation modulus Gro(T) of the commercial PS sample The data for the specific viscosity in dependence of the concentration for polyisobutylene in cyclohexane were published by Baloch (1988).. 32.. The horizontal scaling fac- .5- • o 0.

for supplying us with the experimental data of the commercial polystyrene sample.. This conclusion and the assumption that for small concentrations c the specific viscosity t/~t(c ) is given by rl~(c) = [~llMC (12) . Conclusions A shift procedure for the calculation of mastercurves and the corresponding scaling factors from data showing a scaling behavior has been proposed and discussed. Mastercurve for the specific viscosity r/~q0(c) of polyisobutylene in cyclohexan A o tors aMo(M) (Fig. Against this the vertical scaling factors Cuo(M) (Fig. . . However. the results become more and more accurate with decreasing data errors. . it should be kept in mind that large data errors lead to inexact results. 9a) show a molar mass dependence.Honerkamp and Weese. 5.. . . 1@ b . It was also possible to calculate mastercurves and scaling factors for the specific viscosity o f polyisobutylene in cyclohexane. Acknowledgement J. This result is in agreement with the relations several authors (Baloch. . New Jersey. The method has been tested with experimental data for material functions of several polystyrene melts and led to consistent results for the scaling factors. ~ 1@ L ~ i t . 9. it can also be applied to experimental data which are affected by large statistical errors or by systematic errors. Though the shift procedure has been applied to experimental data with small statistical errors. Scaling factors a~n(M) (a) and C~o(M) (b) calculated from the data for the specific vascosity of polyisobutylene in cyclohexan leads to the relation aMo(M) = [nlM/[nlM0 • (13) The intrinsic viscosity [r/] M and the scaling factor aMo(M) have therefore the same molar mass dependence. Weese gratefully acknowledges financial support by Rheometrics Inc. 1988) propose for the horizontal scaling factors. A note on estimating mastercurves 63 ld ~ ld I@' o o I I o o [] 1@ $ oo o io~ [] t~ ~. We thank Rheometrics Inc. r7 lo~ M Fig. 8. 9b) seem to be independent of the molar mass. On the other hand.~10 6 I [] [] o# [] 10~ [] 103 lOo 1@ [] I@ /@ a 1@ M ng ° id io° ld q~o oo 1o ~ 10-2 1o-~ 10 ~ c (g/cm 10 o 10' ~) Fig.

1 (1993) (Submitted on June 8. Academic Press. Vetterling WT (1986) Numerical recipes.Phys B 27(2):151 Bard Y (1974) Nonlinear parameter estimation. Wiley & Sons. Schindlauer G. Teukolsky SA. Germany . J. 1992. Vol. No. Br.64 Rheologica Acta. Springer Verlag. Flannery BP. in revised form on Nov 16.Chem A 25(4):363 Baloch MK (1988) J Macromol Sci . 3 W-7800 Freiburg i. Janeschitz-Kriegl H (1985) Rheol Acta 24:220 Stoer J (1989) Numerische Mathematik 1. J. Cambridge University Press Schausberger A. New York Ferry JD (1980) Viscoelastic properties of polymers. Berlin Utracki L. New York Press WH. Honerkarnp Fakult~tt for Physik Universit~tt Freiburg Hermann Herder Str. 1992) References Baloch MK (1988) J Macromol Sci . 32. Simha R (1963) J Polym Sci A 1:1089 Correspondence to: Prof.