Interlaboratory tests on polymers by differential scanning calorimetry (DSC): determination and...

Polymer Testing 22 (2003) 419–428www.elsevier.com/locate/polytest

Test Method Comparison

Interlaboratory tests on polymers by differential scanningcalorimetry (DSC): determination and comparison of

oxidation induction time (OIT) and oxidation inductiontemperature (OIT∗)

Manfred Schmida,∗, Samuel Affolterb

a EMPA (Swiss Federal Laboratories for Materials Testing and Research), Lerchenfeldstrasse 5, CH-9014 St. Gallen, Switzerlandb Interstaatliche Fachhochschule fur Technik Buchs (NTB), Werdenbergstrasse 4, CH-9470 Buchs (SG), Switzerland

Received 29 July 2002; accepted 4 September 2002

Abstract

This report presents results obtained from two interlaboratory tests: 14 and 16 participants, respectively, mainly fromindustry and research institutes, measured oxidation induction times (OIT) and oxidation induction temperatures (OIT∗)on four different grades of polyethylene by differential scanning calorimetry (DSC). The measured data were collectedby EMPA and evaluated using a robust statistical method. Repeatability and reproducibility data were of special interest.

The results of the OIT tests indicate that the determination of OIT involves a high degree of uncertainty in respectof the measured data, particularly for low OIT values. This would seem to show that the significance of OIT measure-ments for quality control purposes or lifetime predictions of polyolefin parts has to be viewed critically. DeterminingOIT∗ could therefore be a valuable alternative for less stabilised polyolefins (low OIT values) in particular. However,OIT∗ measurement clearly indicates that the ability to distinguish between different samples decreases drastically asOIT∗ data increase.

One interlaboratory test participant uses chemiluminescence instead of DSC to determine OIT values. In the contextof this study it was shown that chemiluminescence could also provide OIT results within the range of DSC interlabora-tory test data.

Overall, the evaluated data can be used to estimate the significance of in-house and external OIT/OIT∗ measurementsand their compatibility. This report should therefore support day-to-day work in analytical laboratories where oxidationinduction values are measured by DSC. 2002 Elsevier Science Ltd. All rights reserved.

Keywords: Interlaboratory test; Round Robin; DSC; Thermal analysis; Oxidation induction time; Oxidation induction temperature

1. Interlaboratory tests

1.1. Introduction

EMPA St. Gallen organises interlaboratory tests onpolymeric materials on a biennial basis. The participants

∗ Corresponding author. Tel.:+41-71-274 76 74; fax:+41-71-274 77 88.

E-mail address: [email protected] (M. Schmid).

0142-9418/03/$ - see front matter 2002 Elsevier Science Ltd. All rights reserved.doi:10.1016/S0142-9418(02)00122-8

are usually industrial laboratories and laboratories atinstitutes that test, research and develop polymericmaterials. The present interlaboratory tests took place in1998 and 2000.

ISO 5725 (1994) describes the procedures for interlab-oratory tests. The requirements governing the parti-cipants’ organisation, equipment and instrumentation, thestatistical evaluation and the calculation of the relevantinterlaboratory data are precisely defined. The followingfeatures depict the philosophy of the interlaboratory tests:

420 M. Schmid, S. Affolter / Polymer Testing 22 (2003) 419–428

� Interlaboratory tests produce data on the repeatabilityand reproducibility of the measurement results. Thesedata represent valuable information that can be usedto determine the uncertainty of a measuring pro-cedure. The uncertainty of a measurement is mostlyunknown for new test methods. Available calculationmethods do not usually supply the representativeuncertainty values under rough industrial conditions.

� Older standardised methods will be replaced with thedevelopment of new analytical methods. Interlabora-tory tests make a contribution to the comparison ofthe validity of traditional and newly developed testmethods.

� Interlaboratory tests should also make a contributionto the mutual acceptance of measurements. The dataare produced under routine conditions in industrialand institutional laboratories in order to create realreproducibility values. Day-to-day experience showsthat two laboratories produce different results despiteusing the same method. These findings often lead todisputes between partners (i.e. supplier andcustomer). In such cases, data from interlaboratorytests can help in interpreting and assessing the resultsand changing practice in the laboratory, if required.

� The interlaboratory data often lead to dramaticimprovements as a result of adjusting or even chang-ing processing parameters of measuring methods.

� The participants have an opportunity to test thereliability of their results and therefore also test theprofessionalism of their laboratory (proficiencytesting).

The statistical analysis of the present interlaboratory testswas performed according to the robust statistical methoddescribed by Lischer [1].

1.2. Definition of statistical terms

The most important factors that produce deviationsbetween individual measured results are: (a) the operator,(b) the equipment and the analytical instruments, (c) thecalibration of the equipment and the instruments and (d)the environmental effects during the test procedure, i.e.influence of temperature, humidity, light, pollution andso on.

The factors mentioned are mathematically describedby calculating the main characteristic interlaboratoryresults, the robust standard deviation of repeatability sr

(factors (a), (b), (c) and (d) are identical; repeatabilityconditions) and the robust standard deviation of repro-ducibility sR (factors (a), (b), (c) and (d) are varied;reproducibility conditions). The source and the signifi-cance of these and other statistical terms used in interlab-oratory tests are defined in Fig. 1.

Besides the robust standard deviations, the corre-

sponding limits were also calculated. These values aredefined as follows:

r, repeatability limit (r � 2.8 × sr): the value less thanor equal to which the absolute difference between twotest results obtained under repeatability conditions maybe expected to be with a probability of 95%;R, reproducibility limit (R � 2.8 × sR): the value lessthan or equal to which the absolute difference betweentwo test results obtained under reproducibility conditionsmay be expected to be with a probability of 95%.

2. Determination of oxidation induction time andtemperature

2.1. Introduction

Depending on the specific application, plasticmaterials have to be stabilised to a greater or lesserextent against oxidation and environmental influences. Asimple method to check the efficiency of the stabilisersor stabilising systems used is to determine the oxidationinduction time (OIT) or oxidation induction temperature(OIT∗) of the molten material. Especially for polyole-fins, OIT and/or OIT∗ measurements are well estab-lished for quality control purposes as a quick screeningmethod to check the activity of the stabilisation systemused. The OIT measurement is most popular in this case.Many publications deal with this in detail [2]. The effectsof different measuring parameters on the precision of theresults have been investigated in detail [3,4], as well asthe derivation of lifetime prediction for polyolefintubes [5].

In the present report, not only is the well-establishedand popular OIT measurement discussed, but thedynamic OIT∗ measurement is introduced as well. Afew standards are available especially for OIT measure-ment (see Table 1). Although OIT∗ measurement is notvery popular, it has already been the subject of an inter-laboratory test [6] and is described in one standard pro-duced in Finland [7].

Below, both methods will be described in detail andcompared with each other on the basis of four differentpolyethylene types (two Interlaboratory test campaignsin 1998 and 2000). EMPA supplied detailed instructionsto all participants on how to perform the tests. Altogether14 (1998) and 16 (2000) laboratories participated in theprogramme. After collecting all the results, EMPAdeveloped the statistics and revealed the detailed resultsto the participants in the form of survey presentations.An example of such a comparison sheet can be seen inFig. 2. From such plots, the participants can estimatetheir position in the whole field of results and assess theirreliability (e.g. finding a systematic error during themeasurement).

421M. Schmid, S. Affolter / Polymer Testing 22 (2003) 419–428

Fig. 1. Correlation between statistical terms (sr, sL and sR).

Table 1Actual standards for determination of oxidation induction times (OIT)

Standard Title

EN 728 (1997) Plastics piping and ducting systems—Polyolefin pipes and fittings—Determination ofoxidation induction time

ISO/CD 11357-6.3 (1999) Plastics—Differential Scanning Calorimetry (DSC)—Part 6: Oxidation Induction TimeASTM D 3895 (1995) Test Method for Oxidation Induction Time of Polyolefins by Differential Scanning

Calorimetry

Fig. 2. Example for the presentation of the results of interlaboratory tests.

2.2. Principle of measurement

2.2.1. Oxidation induction time (OIT)The sequence of a standardised OIT measurement

using the DSC method according to EN 728 is outlined

in Fig. 3. A sample of the polymer (approximately 15mg) is placed in a clean aluminium pan. After pos-itioning the uncovered sample pan together with anempty reference pan in a calibrated DSC oven, a nitrogenatmosphere is established in the measuring cell. Then,


Fig. 3. Schematic diagram for the determination of OIT as the time period between t1 and t2 (onset of decomposition).

the sample and the reference are heated rapidly (at least�20 K/min) to the temperature at which the OIT valueis to be determined. When the required temperature isreached for the first time an isothermal step of 3 minfollows. After reaching this point (indicated as t1 in Fig.3) the atmosphere is switched to oxygen and the DSCoven is held at the same temperature until an exothermalsignal (oxidation) can be recognised. The onset of thisoxidation signal corresponds to a time t2. The OIT valuecan now be determined as the time between t1 and t2, asdescribed in Fig. 3. According to EN 728, the flow ofnitrogen and oxygen should be adjusted to 50ml/min±10% during the entire measuring procedure.

Frequently however, the oxidation signal is less pro-nounced than indicated in Fig. 3, making the determi-nation of a clearly defined onset temperature difficult(Fig. 4 shows two real OIT curves with exact (curve b)and problematic (curve a) determination of t2). In thiscase the evaluation of t2 can still take place because adeviation from the base line, defined in advance, is takenas the time t2. It should be pointed out, however, thatthe two methods yield different OIT values of t2. Thus,

Fig. 4. Realistic OIT curves with good (curve b) and problematic (curve a) determination of t2.

comparative series of measurements must always be per-formed with the same evaluation method in order toensure comparability of results.

Finding a suitable measuring temperature for the iso-thermal phase often causes further difficulties with OITmeasurements. If the temperature is too low there is asubstantial increase in the duration of the measurement.On the other hand, if the temperatures are too high, oxi-dation takes place immediately after the introduction ofoxygen. The onset temperature of the decomposition sig-nal (t2) can no longer be determined. Acceptable OITtimes (range from 30 to 60 min) are frequently obtainedwith polyolefins at 200 or 210 °C. Usually, however,extensive preliminary tests are necessary to optimise thetest conditions for OIT measurements of unknownsamples. In this connection, the dynamic OIT∗ measure-ment described below requires substantially less effort.

2.2.2. Oxidation induction temperature (OIT∗)The oxidation induction temperature (OIT ∗) is evalu-

ated in accordance with Fig. 5. The sample is heated upcontinuously (i.e. 10 °C/min) under a pure oxygen (or


Fig. 5. Schematic diagram for the determination of OIT∗ as the onset temperature of decomposition.

air) gas flow. A change of gases at a defined time, asstated under OIT measurement, is not necessary. OIT∗is determined as that point in the thermogram where theonset of the decomposition signal results. OIT∗ is usu-ally more clearly pronounced as the onset time t2 in OITmeasurements (t2 is necessary for the determination ofthe OIT values).

Finally, to give an overview, the key points of bothmethods (determination of OIT versus OIT∗) are sum-marised in Table 2. Comparing the points in Table 2, itis obvious that the OIT∗ method needs less effort in set-ting up the measurements and in the majority of cases itdelivers clearly defined onset points.

2.3. Results

2.3.1. Oxidation induction time (OIT)Four commercial grades of polyethylene with different

levels of stabilisation were selected as test materials forthe OIT test (see Table 3 for exact description of thematerials). Altogether 14 (1998) and 16 (2000) labora-tories took part in the interlaboratory tests. The test pro-gramme was accurately prescribed for all the participat-

Table 2Correlation and comparison of the two oxidation induction measuring principles

Oxidation induction time (OIT) Oxidation induction temperature (OIT∗)

Standardised yes no (except in Finland)Measuring principle static dynamicPreliminary tests yes (evaluation of capable measuring no

temperature)Gas change yes (possible source of error) noOnset of oxidation signal partially difficult to recognise and to mostly pronounced very well and clear to analyse

analyze

ing laboratories. The temperature of the isothermaloxidation phase was defined as 210 °C.

The interlaboratory test data of the OIT values aresummarised in Table 4. From this, it is possible to estab-lish that the repeatability and reproducibility of the OITvalues depends on the order of magnitude of the medianvalues of the samples. The lower the OIT median values,the greater the relative values (sr relative, sR relative) ofthe standard deviations. This correlation becomes clearerif the data are compared with data from other interlabora-tory tests (see Table 5).

From the data in Table 5, it is possible to see, as atrend, what influence the OIT median values have on thesr and sR results. The lower the OIT values, the greaterthe relative standard deviation becomes. This could beseen with the relative reproducibility standard deviation(sR relative) in particular. Even a borderline within therange of 10–20 min seems to be present, below whichthe sR values rise significantly (relative reproducibilitystandard deviations greater than 60%!). Evaluations ofOIT values of samples with very low stabilisatoramounts (OIT lower than 15 min) must therefore beregarded critically. One possibility to improve the pre-


Table 3Materials used in the interlaboratory test

Sample Material Producer Type/visual characterisation

Sample 1 PE-HD BASF AG Lupolen 4261 A/white powderSample 2 PE-LD BASF AG Lupolen 1852 H schw. 412/black granulesSample 3 PE-HD DOW Plastics BG 10050/black granulesSample 4 PE-HD Hoechst Hostalen CRP 100/black granules

Table 4OIT median values, standard deviations (sr, sR), relative standard deviations (sr relative, sR relativea), repeatability and reproducibilitylimit (r, R) of the four PE ring test materials

Value Unit Sample 1 (PE-HD) Sample 2 (PE-LD) Sample 3 (PE-HD) Sample 4 (PE-HD)

Median value [min] 3.4 18.9 36.9 62.4sr [min] 0.6 1.2 2.1 1.7sr relativea [%] 17.8 6.1 5.8 2.7sR [min] 2.1 2.0 6.5 9.5sR relativea [%] 62.1 10.8 17.6 15.3r [min] 1.7 3.2 5.9 4.8R [min] 6.0 5.7 18.2 26.6

a sr,R relative [%] � (sr,R × 100%)/median value.

cision of OIT in that case would be to lower the iso-thermal temperature in order to increase t2 to a timeregion above 20 min.

2.3.2. Oxidation induction temperature (OIT∗)The key data of the interlaboratory test obtained from

the evaluation of the OIT∗ values are summarised inTable 6. The standard deviations of the temperaturevalues in Table 6 seem to be reasonable because theyhave a similar order of magnitude as those found in otherinterlaboratory tests, where temperatures were evaluatedby means of DSC measurements [8,9].

In the case of temperature measurements, it is quiteclear that calculated relative values (sr relative and sR

relative) strongly depend on the temperature scale used.For instance, if absolute temperatures (Kelvin scale) areused in our interlaboratory tests, the relative standarddeviation would become smaller. Another proposal couldbe to define the melting point of every material as thezero point of the respective OIT∗ measurement. In thiscase the relative standard deviations would increase sub-stantially. In our interlaboratory test, we decided to usetemperatures corresponding to the centigrade scale,because it is quite common in DSC measurements.Therefore, all further discussions and comparisons ofrelative temperature data refer to °C.

2.3.3. Comparison of OIT and OIT∗ dataTable 7 summarises the median values, standard devi-

ations (sr, sR) and relative standard deviations (sr relative,sR relative) of the four PE samples and the two oxidationinduction methods. By comparing the relative standarddeviations (sr relative and sR relative) of OIT and OIT∗data it can be seen that the dynamic measurement(OIT∗) is associated with significantly smaller relativestandard deviations than the static OIT method (see Fig.6), both in the case of the repeatability standard deviationand in the case of the reproducibility standard deviation.This could lead to the conclusion that OIT∗ measure-ments are always more reproducible than OIT tests andshould be preferred in all cases. Additionally, the set upof OIT∗ measurements is even more easy.

However, if one compares the absolute values, then itis noticeable that in the case of OIT∗ measurements thedifferential between individual samples decreases sig-nificantly with rising temperature. This correlation isshown clearly in Fig. 7. The four PE interlaboratory testsamples are plotted here in an OIT/OIT∗ diagram withthe median values as x, y values (symbol �) and thereproducibility standard deviations (sR) as uncertaintybars. It can be clearly seen that due to the exponentialcharacter of the oxidative decomposition of plastics withrespect to temperature (dotted line in Fig. 7), OIT∗


Table 5Comparison of relative standard deviations (sr relative and sR relativea) of OIT data from own (EMPA) and foreign interlaboratorytests (described in the mentioned standards)

Data from own and Material Number of Interlaboratory test key dataforeign interlaboratory participantstests

OIT median value sr relativea (%) SR relativea (%)(min)

EMPA 1998b PE-HD 14 3.4 15.4 64.7OKI 2000c PE-HD 13 3.7 33.7 62.2OKI 2000c PE-HD 12 9.3 11.8 64.1EMPA 1998b PE-LD 14 18.9 6.9 11.1ASTM D 3895d PE-LD 11 23.4 13.2 20.1ISO/CD 11357-6e PE-LD –e 24.0 11.7 17.9EMPA 2000b PE-HD 16 36.9 5.8 17.6EMPA 2000b PE-HD 16 62.4 2.7 15.3ASTM D 3895d PE-LD 11 79.9 11.4 18.9ISO/CD 11357-6e PE-LD –e 83.4 11.0 20.9ASTM D 3895d PE-LLD 11 119 6.6 14.0ISO/CD 11357-6e PE-LLD –e 120 6.5 12.2ISO/CD 11357-6e PE-HD –e 163 5.1 13.3ASTM D 3895d PE-HD 11 166 4.9 14.5

a sr,R relative [%] � (sr,R × 100%)/median value.b Interlaboratory test performed by EMPA in 1998 and 2000; evaluation of data with robust statistics.c Interlaboratory test performed by the Austrian Plastics Institute; evaluation of data according ISO 5725-2 with consideration

of outliers.d Interlaboratory test performed by ASTM in 1991; evaluation of data with ASTM E 691.e Not specified regarding evaluation method and number of participants.

Table 6OIT∗ median values, standard deviations (sr, sR), relative standard deviations (sr relativea, sR relativea), repeatability and reproducibilitylimit (r, R) of the four PE interlaboratory test materials

Value Unit Sample 1 (PE-HD) Sample 2 (PE-LD) Sample 3 (PE-HD) Sample 4 (PE-HD)

OIT∗ median value [°C] 217 242 248 254sr [°C] 2.4 0.7 0.9 1.5sr relativea [%] 1.1 0.3 0.4 0.6sR [°C] 4.0 2.2 2.8 4.1sR relativea %] 1.8 0.9 1.2 1.6r [°C] 6.7 1.9 2.5 4.2R [°C] 11.1 6.1 7.8 11.5


values above approx. 240 °C lead to a cumulativeaccumulation of the measured OIT∗ values. This finallyleads to an overlay of the error bars concerning abscissa(overlapping region in Fig. 7). Thus, the differentiationbetween samples no longer exists in the case of the OIT∗values despite very small relative standard deviations(e.g. sample 3 and sample 4 may give the same OIT∗result although they contain a different amount of oxi-dation stabilisers). In contrast to that, it is possible to

distinguish the samples by way of the OIT values(ordinate axis) despite high relative standard deviations.

2.3.4. ChemiluminescenceOne interlaboratory test participant determined the

OIT values not by DSC measurement, but by using arecently developed measuring instrument for the deter-mination of chemiluminescence (CL). Here, the stabilityof plastics during thermal oxidation is observed by light


Table 7OIT and OIT∗ median values, standard deviations (sr, sR) and relative standard deviations (sr relativea, sR relativea) of the four PEinterlaboratory test samples

Method Median values sr sr relativea (%) sR sR relativea (%)

Sample 1 (PE-HD) OIT 3.0 min 0.5 min 15.4 2.2 min 64.8OIT∗ 217 °C 2.4 °C 1.1 4.0 °C 1.8

Sample 2 (PE-LD) OIT 18.9 min 1.2 min 6.2 2.1 min 11.0OIT∗ 242 °C 0.7 °C 0.3 2.2 °C 0.9




Fig. 6. Comparison of the relative standard deviations of OIT and OIT∗ method.

Fig. 7. X,Y-pair of OIT/OIT∗ values (symbol �) of the four PE ring test samples with the corresponding reproducibility standarddeviations (sR) as uncertainty bars.


Fig. 8. Determination of OIT data with chemiluminescence according two different methods.

emission (see Fig. 8 for an example of such a chemilumi-nescence measurement). The data determined with thismethod are summarised and compared with the datafrom DSC in Table 8. Every sample is evaluated usingtwo different methods. Firstly by determination of themaximum slope of the CL curve and secondly by deter-mination of the “onset” on the CL curve (see Fig. 8).From the comparison of the data in Table 8 it can beseen that the CL data, particularly evaluated accordingto CL method 2 (determination of the “onset” of the CLcurve), agree closely with the data of the DSC measure-ments. Thus, this “new” method seems to be quite suit-able for the determination of OIT values.

Table 8Comparison of the OIT data of the interlaboratory test and the data received from chemiluminescence (CL) measurements (medianvalues and standard deviations); CL measurements were evaluated according to two different methods (described in Fig. 8)

Sample 3 Sample 4

Median value (min) sr resp. si (min) Median value (min) sr resp. si (min)

Interlaboratory test data (DSC) 36.9 2.1 62.4 1.7CL method 1a 49.9 4.5 67.1 1.9CL method 2b 38.0 3.2 62.9 2.3

a Evaluation by determination of the maximum slope of the CL curves.b Evaluation by determination of the “onset“ of the CL curves.

3. Conclusion

The present report indicates what repeatability andreproducibility standard deviations and what correspond-ing limits must be taken into account when oxidationinduction time (OIT) or temperature (OIT∗) measure-ments are performed using DSC. At a glance, the reportshould therefore provide support for day-to-day work inanalytical laboratories where OIT data are measured.

The data from this OIT interlaboratory test demon-strate that the determination of the OIT shows a substan-tial variation in the measured values, particularly for verylow OIT values. Here, only a reduction of the tempera-


ture of the isothermal phase (T�210 °C) or a reductionof the oxygen content in the measuring chamber couldincrease the differentiation of similar samples. The highvalues for repeatability and reproducibility standarddeviations also show that the significance of OITmeasurements, e.g. with regard to quality control or life-time predictions of polyolefin parts, has to be viewedrather critically.

Particularly for very low OIT values (polyolefins withlow stabilisation), the dynamic procedure for the deter-mination of the OIT∗ seems to be a good alternative.However, the OIT∗ data also show clearly that differen-tiation between individual samples decreases rapidly asOIT∗ values increase.

Beyond this, it was shown within the scope of theinterlaboratory test that chemiluminescence measure-ments also provide comparable OIT results.

Acknowledgements

The authors wish to thank Dr M. Weber who handledthe statistical software and evaluated the statistical data,Dr A. Ritter who supported the project by a lot of differ-ent activities and ideas, and all the participants in theinterlaboratory test campaigns (1998 and 2000) who pro-duced the results presented in this report. Also we thank

all our colleagues at the Institute who made it possiblefor the interlaboratory tests to be carried out and thecompanies and institutes which supported the projectfinancially or by supplying test materials free of charge.

References

[1] P. Lischer, Robust statistical methods in interlaboratoryanalytical studies, in: H. Rieder (Ed.), Robust Statistics,Data Analysis, and Computer Intensive Methods, LectureNotes in Statistics (109), Springer Verlag, New York, 1996,pp. 251–265.

[2] J.R. Pauquet, R.V. Todesco, W.O. Drake WO, Limitationsand applications of oxidation induction time (OIT) to qual-ity control of polyolefins, 42nd International Wire & CableSymposium, 1993.

[3] S. Wallius, Angew. Makromol. Chem. 212 (1993) 103.[4] A.T. Riga, G.H. Patterson, Oxidative Behavior of Materials

by Thermal Analytical Techniques, ASTM STP 1326,1997.

[5] H. Gebler, Kunstst. 79 (1989) 9.[6] L. Koski, K. Saarela, J. Thermal Anal. 25 (1982) 167.[7] SFS 3449, Plastic pipes—Thermoanalytic method for the

determination of oxidation stability (dynamic method), Fin-land, 1982.

[8] S. Affolter, M. Schmid, B. Wampfler, Kautsch. GummiKunstst. 52 (1999) 519.

[9] M. Schmid, S. Affolter, A. Ritter, Macromol. Mater. Eng.286 (10) (2001) 605.

Interlaboratory tests on polymers by differential scanning calorimetry (DSC): determination and...

Documents

Transcript of Interlaboratory tests on polymers by differential scanning calorimetry (DSC): determination and...