Accred Qual Assur (2012) 17:439–444 DOI 10.1007/s00769-012-0900-8

PRACTITIONER’S REPORT

Determination of brominated flame retardants: a proficiency test Fernando Cordeiro • Piotr Robouch • Thomas Linsinger • Beatriz de la Calle

Received: 29 February 2012 / Accepted: 8 May 2012 / Published online: 25 May 2012  Springer-Verlag 2012

Abstract This manuscript presents the results of a proficiency test for the determination of total bromine and several polybrominated biphenyls and diphenyl ethers (PBB and PBDE) in plastic. The test material used is a poly(ethylene terephthalate) (PET) granulate fortified with a mixture of PBB and PBDE. Up to twenty laboratories from 15 countries registered for the exercise and reported results. Homogeneity and stability were investigated to asses the adequacy of the selected test material. Laboratory results were rated with z and f scores according to ISO 13528. The standard deviation for proficiency assessment was set to 25 % of the assigned values. This exercise highlights the difficulties laboratories have to provide consistent values for the investigated measurands. Many participants reported underestimated measurand values; satisfactory z scores ranged from 61 to 88 %. The critical experimental parameters are identified and discussed. Keywords Proficiency test  External quality assessment  Flame retardants  Polymers  Scoring

Introduction Brominated flame retardants (BFR) are a group of organic bromine compounds producing an inhibitory

Presented at the Eurachem PT Workshop, October 2011, Istanbul, Turkey. F. Cordeiro (&)  P. Robouch  T. Linsinger  B. de la Calle European Commission, Joint Research Centre, Institute for Reference Materials and Measurements (EC-JRC-IRMM), Retieseweg 111, 2440 Geel, Belgium e-mail: [email protected]

effect on the ignition of combustible organic materials such as textiles, paper and plastics. They are extensively used in electronics, clothes and furniture. Not strongly bound to the host polymer (plastic), BFRs are able to ‘‘bleed’’ from it and become an environmental contaminant. The European Commission Directive 2002/95/EC on the restriction of the use of certain hazardous substances in electrical and electronic equipment (RoHS) [1] bans the use of certain polybrominated flame retardants in electric and electronic devices from 1 July 2006 unless no technical substitutes exist. Moreover, the European Regulation 850/2004 [2] implementing the Stockholm Convention on persistent organic pollutants (POPs) and the Commission Decision 2005/618/EC [3] set the ‘‘maximum limits for the total sum of polybrominated biphenyls (PBB) and polybrominated diphenyl ethers (PBDE) to 0.1 g/100 g’’ (expressed in mass fraction). These restrictions were further confirmed by the EU Regulation 756/2010 [4]. In 2007, a proficiency test (PT) was organized to test the measurement capabilities of laboratories to determine brominated flame retardants in plastic [5]. The outcome of that PT indicated ‘‘a clear need for a learning process among the laboratories involved’’. This conclusion was based on the fact that the observed between-laboratory relative standard deviations ranged from 22 to 61 % for the different BFRs congeners. A significant improvement was achieved in the certification exercise of two polymer-based test materials where the observed between-laboratory variability ranged from 3 to 12% [6, 7]. The present work reports the outcome of the second proficiency test organized by the Institute for Reference Materials and Measurements (IMEP-26) aimed to estimate analytical performance of laboratories when determining polybrominated flame retardants in plastic.

123

440

Invitation, registration and sample distribution IMEP-26 was advertised on the IRMM website [8] and announced to the European Co-operation for Accreditation (EA) and to the Asian and Pacific Laboratory Accreditation Cooperation (APLAC). Both organizations were invited to nominate laboratories for participation. Each participant received an information letter related to measurands, sample storage conditions and the envisaged time frame. Laboratories were requested to use the analytical method they normally apply and to perform two or three independent measurements. They were asked to report the mean of experimental replicates, the associated expanded measurement uncertainty together with the coverage factor, and to describe the analytical technique used. The results were to be reported with the same number of significant digits they usually report to customers. The measurands were defined as ‘‘total mass fraction of bromine, sum of polybrominated biphenyls (PBBs) and sum of polybrominated diphenyl ethers (PBDEs) as well as the mass fraction of several specific brominated diphenyl ethers (BDE-47, BDE-99, BDE-183 and BDE-209) and decabrominated biphenyl (BB-209) in plastic’’.

Accred Qual Assur (2012) 17:439–444

sufficiently homogeneous for the purpose of this PT for all the investigated measurands. The stability data acquired in the frame of the stability monitoring program of the IRMM Reference Material Unit clearly indicate that the former IRMM-310 is stable; no further study was performed.

Reference values and their uncertainties The following three expert laboratories, later referred as Certifiers, were contacted to analyse the test material and to confirm whether the IRMM-310 informative values could be used as assigned values (Xref): • • •

Belgian Nuclear Research Centre (SCK/CEN, Mol, Belgium) ¨ V Rheinland Taiwan Ltd (Taiwan); and TU Vlaamse Instelling Voor Technologisch Onderzoek (VITO, Belgium)

The means reported by the Certifiers (Xcert) together with the associated expanded uncertainties (Ucert) are compared to the IRMM-310 informative values (Table 1). Total bromine

Test material Preparation The IMEP-26 test material is a polyethylene terephthalate (PET) spiked with PBDEs and PBBs concentrations matching the maximum levels set in the RoHS Directive [1, 3]. This material was first produced for the 2007 PT [5] and was available as a quality control material (IRMM310). It was finally removed from the IRMM catalogue when two new certified reference materials for BFRs in polymers (ERM-EC590 and ERM-EC591) were marketed [7]. The former IRMM-310 material was re-labelled to avoid any identification.

The IRMM-310 informative value for total mass fraction of Br was determined in 2007 by an expert laboratory having used solid sampling neutron activation analysis with k0 internal standardization (k0NAA) and having demonstrated its measurement capability in the frame of the certification campaign of other polymer materials [7]. While Certifier 3 used k0NAA and confirmed the original total mass fraction of Br values, the values reported by the other two Certifiers—using high-resolution inductively coupled plasma

Table 1 Results and associated expanded uncertainties reported by the certifying laboratories (Xcert ± Ucert) compared to the IRMM-310 informative values (X ± U, k = 2) Analyte

Certifier 1 Xcert ± Ucert

Certifier 2 Xcert ± Ucert

Certifier 3 Xcert ± Ucert

Total Br

1407 ± 225

1600.1 ± 245

2245 ± 110

Sum PBDEs

1507 ± 452

806.9 ± 7.2

Homogeneity and stability The experimental design used for the assessment of the homogeneity of the test sample complies with the requirements set by the ISO 13528 [9]. These tests compare the between-bottle standard deviation (ubb) with the standard deviation for PT assessment (^ r). ubb was estimated using the SoftCRM software [10, 11] and ranged from 0.4 to 1.1 % for all BFR congeners and 2.1 % for the total bromine mass fraction (expressed as a relative standard ^ was set to 25 % of the assigned values of each deviation). r measurand. The tests indicated that the test material was

123

IRMM-310 (X ± U) 2300 ± 120 1800 ± 135

Sum PBBs

825 ± 297

323.0 ± 7.2

700 ± 110

BDE-47

226 ± 17

130.3 ± 3.4

227 ± 25

BDE-99

320 ± 45

168.0 ± 6.2

307 ± 31

BDE-183

94 ± 18

40.2 ± 1.4

150 ± 17

BDE-209

728 ± 160

339.6 ± 9.2

689 ± 128

BB-209

781 ± 281

323.0 ± 7.2

700 ± 110

All values are expressed in mg kg-1 The numbering of certifiers does not correspond to the listing order in ‘‘Reference values and their uncertainties’’

Accred Qual Assur (2012) 17:439–444

mass spectrometry or X-ray fluorescence spectrometry— were significantly lower. This discrepancy was attributed to the loss of analyte during sample treatment. Thus, the IRMM-310 informative value for the total mass fraction of Br was used as assigned value in IMEP-26.

441 Table 2 Assigned values (Xref), associated expanded uncertainties r) for IMEP(Uref, k = 2) and standard deviation for PT assessment (^ ^ were set as 0.25 Xref 26. r Analyte

Xref

Uref

^ r

Total Br

2300

136

575

Sum PBDEs

1800

136

450

110 25

175 56.8

BFRs congeners

Sum PBBs BDE-47

700 227

The IEC 62321:2008 [12] standard stressed the importance of an efficient extraction for the determination of total mass fraction of PBBs and PBDEs in polymers. It explicitly recommended (1) cryogenic grinding of the sample to increase the extraction efficiency and (2) instrumentally assisted extraction technique (i.e. Soxhlet extraction) prior to the gas chromatography mass spectrometry (GC–MS) measurements. These recommendations were proven to be critical experimental parameters and led to satisfactory results for PBBs and PBDEs congeners in the 2007 PT [5]. Certifier 1 cryogenically ground all the samples, carried out Soxhlet extraction and performed high-resolution GC– MS analysis using isotopically labelled internal standards. Certifier 2 used a simplified sample treatment where toluene was added to the samples cut into small pieces, and the mixture was shaken manually. No grinding, no instrumental extraction (i.e. Soxhlet) was applied, and no isotopically labelled internal standards were used for quantification. His reported results were systematically the lowest (Table 1). Therefore, the values reported by Certifier 2 were excluded from statistical evaluation. Despite this significant difference between their reported results, both certifiers used a CRM to assess their measurement trueness. All the IRMM-310 informative values, confirmed by Certifier 1 and Certifier 3 (for total Br), were set as assigned values (Xref) for IMEP-26. No reference value could be assigned to the mass fraction of BDE-183, due to the highly discrepant results reported by the expert laboratories and the IRMM-310 informative value. No scoring could therefore be provided for this measurand. Table 2 presents all the assigned values (Xref), the standard deviation for the proficiency test assessment (^ r) and the reference expanded uncertainties (Uref) calculated as follows: qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Uref ¼ 2 u2310 þ u2bb ð1Þ

BDE-99

307

31

BDE-209

689

128

172

700

110

175

where u310 is the IRMM-310 reference standard uncertainty and ubb is the standard uncertainty estimated from homogeneity studies Scores and evaluation criteria Individual laboratory performance is expressed in terms of z and f scores in accordance with ISO 13528 [9].

BB-209

76.8

-1

Values expressed in mg kg

xlab  Xref ^ r xlab  Xref ffi f ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi u2ref þ u2lab

z¼

ð2Þ ð3Þ

where xlab is the measurement result reported by a participant, Xref is the assigned value, uref is the standard uncertainty of the reference value, ulab is the standard uncertainty reported by a ^ is the standard deviation for proficiency test participant and r assessment These scores are interpreted as: • • •

Satisfactory result for | z or f | B2, Questionable result for 2 \ | z or f | B3, Unsatisfactory result for | z or f | [3.

The z score compares the participant’s deviation from the reference value with the target standard deviation for the proficiency assessment set to 25 % of the assigned value of each measurand. The f score indicates whether the range reported by the laboratory is in agreement with the assigned one. Unsatisfactory f score can be due to an inaccurate estimate of either the mass fraction or its measurement uncertainty (or both). Generally, the standard uncertainty of the laboratory (ulab) was calculated dividing the reported expanded uncertainty by the reported coverage factor (k). When k was not specified, the reported expanded uncertainty was considered as the half-width of a rectangular distribution; ulab was then calculated dividing this half-width by H3, as recommended by EURACHEM/CITAC guide [13]. When no uncertainty was reported, ulab was set to 0.

Evaluation of results From the 25 laboratories having registered, only 23 submitted results and completed the questionnaire. Eight of them reported results for total mass fraction of bromine, while 21 laboratories reported results for the BFRs

123

442

Accred Qual Assur (2012) 17:439–444

congeners. ‘‘Less than’’ values were not included in the evaluation. The between-laboratory relative standard deviations obtained in this exercise ranges from 48 to 68 %. This range is wider than the one observed during the certification campaign (3–12 % [6]), but is similar to the one of the 2007 PT using the same test material (from 22 to 61 % [5]). This exercise may indicate that the laboratory performances for the determination of BFRs in plastics did not improve since the 2007 PT exercise. The results reported by participants and the consequent assessment of performances are presented in the IMEP-26 report [14]. It is surprising to see that even when requested, only 60 % of the participants reported measurement uncertainties, when the majority (75 %) claimed providing uncertainty statement

to their customers and being accredited according to ISO 17025. Figure 1 presents the z and f scores obtained in this exercise. The percentage of satisfactory z scores ranged from 61 to 88 %, while the share of satisfactory f scores ranged between 33 % and 50 %. Only 40 % of the participants obtained satisfactory z and f scores for the various BFR congeners investigated. The majority of the participants underestimated their measurement uncertainty [14].

Bimodal distribution of the results A bimodal distribution was observed for most the BFRs congeners, except for the total mass fraction of BDE-47

z scores

a

100%

0 12.5

10 22

24 0

score (%)

75%

14

18

29 14

24

18

22

10

61

62

Sum PBB

BB-209

50%

87.5 76

72

67

25%

64

0% Total Br

BDE-47

Sum PBDE

Satisfactory

b

BDE-209

BFRs congeners Questionable Unsatisfactory

ζ scores 100%

75%

62.5

score (%)

BDE-99

57

57

56

62

50%

52

0 0

5

10

11

38

33

33

33

Total Br

BDE-47

Sum PBDE

Sum PBB

BB-209

5

5

25%

37.5

45

48

50

BDE-99

BDE-209

0%

BFRs congeners Satisfactory Questionable Unsatisfactory Fig. 1 Overview of scores: a z scores and b f scores for the seven analytes

123

Accred Qual Assur (2012) 17:439–444

443

a

Kernel Density Plot BDE-209

| z| < 2

0.0014

X Ref

0.0012 0.001

0.0008 0.0006

0.0004 0.0002

0 -1000

-500

0

500

1000

1500

2000

| z| > 2

b

Fig. 2 Partial least squares regression model for BDE-209 data, presenting the score plot (a) and the loading plot (b). Only the two first PLS components (PC1 and PC2) are shown. L and Q denote the

laboratory and the question codes. zBDE-209 represents the projection of the z score under investigation on the loading plot. In the kernel density plot the x-axis reads in mg kg-1

and the sum of PBBs. The higher mode shown in the kernel density plot [15] (Fig. 2) corresponds to the assigned value (Xref), while the other mode is a factor of two lower. A holistic approach was applied to identify the reasons for such discrepancies exploiting the information of the questionnaire filled by the participants. Multivariate statistical models, such as partial least squares regression models (PLS-R), were applied for the BFR congeners using The Unscrambler X 10.1 software (CAMO Software AS, Norway) to identify and quantify the multivariate relationship, for each measurand, between the answers to the questionnaire (categorical variables) and the estimated laboratory z score. The PLS-R models were cross-validated and captured most of the structured information with low model errors, thus explaining from 79 to 92 % of the variances observed. Figure 2 illustrates the PLS-R model for BDE-209. The score plot (Fig. 2a) provides the sample patterns allowing the identification of any clustering among the calculated z scores, while the loading plot (Fig. 2b) illustrates the interrelationship among all the categorical variables and their relationship with the corresponding z score. It enables

the identification of which answers to the questionnaire have the biggest influence in the PLS-R model, thus providing reasons for the observed bimodality of the distributions. The following positively influencing experimental parameters were used by laboratories having reported satisfactory results, as indicated in the answers to the questions indicated between parenthesis: • • • • •

Grinding the sample (Q15); Soxhlet extraction (Q12a); Participating regularly in ILCs for this type of analysis (Q20); Using analytical GC–MS (as opposed to GC-ECD or HPLC–UV-FLD); and Carrying out this type of analysis regularly (Q23).

Similarly, some negatively influencing parameters, resulting in unsatisfactory results, were identified. Among them we wish to stress the following: the use of static or— ultrasonic extraction (Q12d or Q12c) and—the use of long (30–50 m) GC columns (Q07). These observations confirm once again the recommendations provided in the

123

444

certification report of ERM-EC 590 and 591 [6] and in the IEC 62321 standard [12]. The use of a short chromatographic column (10–15 cm) was recommended to reduce any potential degradation of higher brominated congeners. It is worth mentioning the IEC 62321 standard inserts the determination of PBBs and PBDEs congeners in polymers by GC–MS as an ‘‘informative annex’’ due to the large scatter results reported in the corresponding ring trial. The standard provides a list of commercially available BFR calibrants without any minimum required purity. The use of matrix-matched certified reference materials should be recommended and should be considered in next revision of the standard [7].

Conclusion Compared to the first ILC exercise in 2007, no significant improvement in analytical performances is identified by this proficiency testing round related to the determination of BFR congeners in plastics. Up to 33 % of the participants underestimated the sum of PBDEs mass fractions, for a test material containing 1.8 g kg-1. Hence, they would have failed in supporting the legal requirement setting a maximum of 1 g kg-1. The multivariate data analysis identified four critical experimental parameters enabling reliable results: sample grinding, instrumental extraction technique (i.e. Soxhlet), effective shaking (i.e. ultrasonic) and the use of short GC columns to avoid congener degradation. Additionally, the use of matrix-matched certified reference materials should be considered to improve the state of the practice in this field. Acknowledgments Authors wish to thank J. Charoud-Got, I. Verbist and F. Ulberth for the contribution to this project and for the constructive discussions. Following laboratories are acknowledged for their participation in IMEP-26: National Measurements Institute, Australia; Umweltbundesamt Austria, Austria; SGS Belgium NV, Belgium; SGS do Brasil Ltd, Brasil; Intertek Testing Services Wuxi Ltd, China; Intertek Testing Services Guangzhou Ltd, China; Intertek Testing Services Shenzhen Ltd, China; SGS Hong Kong Ltd, China; CMA Industrial Development Foundation Ltd, China; Electronics Testing Center, Taiwan, China; PICA Pru¨finstitut Chemische Analytik GmbH, Germany; Intertek Consumer Goods GmbH, Germany; Fraunhofer Institut fu¨r Verlahrenstechnik und Verpackung IVV, Germany; Ostthu¨ringische Materialpru¨fgesellschaft fu¨r Textil und Kunststoffe GmbH; Germany; Pruefinstitute Hansecontrol, Germany; Eurofins GfA GmbH, Germany; IISG Istituto Italiano Sicurezza dei Giocattoli, Italy; Israel Chemicals—Industrial Products, Israel; Institute for Environmental Studies (IVM), Netherlands; KOMAG

123

Accred Qual Assur (2012) 17:439–444 Institute of Mining Technology, Poland; SGS Philippines, Inc., Philippines; Swiss Quality Testing Services SQTS, Switzerland; Intertek Testing Services (Thailand) Ltd, Thailand; St. Louis Testing Laboratories, United States; Intertek Consumer Goods, UK Ltd, United Kingdom.

References 1. European Directive 2002/95/EC on the restriction of the use of hazardous substances in electrical and electronic equipment (RoHS) 2. Commission Regulation (EC) 850/2004 on persistent organic pollutants (POPs) 3. Commission Decision 2005/618/EC amending Directive 2002/95/EC for the purpose of establishing the maximum concentration values for certain hazardous substances in electrical and electronic equipment 4. Commission Regulation (EC) 756/2010 amending Regulation 850/2004 5. Linsinger TP, Voorspoels S, Liebich A (2008) Results of an interlaboratory comparison on the determination of polybrominated flame retardants in poly(ethyleneterephthalate). Anal Bioanal Chem 390:399–409 6. Linsinger TP, Birgersson-Liebich A, Lamberty A, Pellizzato F, Venelinov T, Voorspoels S (2009) Development of the first certified reference materials for several brominated flame retardants in polymers. Anal Chem 81(10):3792–3800 7. http://www.erm-crm.org/ERM_products/search/reports/EC590591.pdf 8. IMEP Web page: http://irmm.jrc.ec.europa.eu/interlaboratory_ comparisons/imep/Imep-26/Pages/IMEP-26.aspx 9. ISO 13528 (2005) Statistical Methods for Use in Proficiency Testing by Interlaboratory Comparisons. Issued by ISO Geneva (CH), International Organization for Standardization 10. Van der Veen AMH, Linsinger TP, Pauwels J (2001) Uncertainty calculations in the certification of reference materials. 2 homogeneity study. Accredit Qual Assur 6:26–30 11. Bonas G, Zervou M, Papaeoannou T, Lees M (2003) SoftCRM: a new software for the certification of reference materials. Accredit Qual Assur 8:101–107 12. IEC 62321 Ed.1 (2008) Electrotechnical products–Determination of levels of six regulated substances (lead, mercury, cadmium, hexavalent chromium, polybrominated biphenyls, polybrominated diphenyl ethers). Issued by International Electrotechnical Commission (IEC), 2008 13. EURACHEM/CITAC (2000), Quantifying Uncertainty in Analytical Measurement. http://www.eurachem.org. Accessed 14 April 2012 14. Cordeiro F, Verbist I, Robouch P, Linsinger TP, De la Calle MB (2011) IMEP-26: Determination of brominated flame retardants in plastic. EUR 24874, Publication Office of the European Union, ISBN 978-92-79-20657-3 15. Representing data distributions with Kernel density estimates (2006), Issued by the Statistical Subcommittee of the Analytical Methods Committee (AMC) of the Royal Society of Chemistry (RSC), AMC Technical Brief No 4