Alkylphenol retention indices

Journal of Chromatography A, 1123 (2006) 98–105

Alkylphenol retention indices Svein A. Mjøs a , Sonnich Meier b , Stepan Boitsov b,∗ a b

Norwegian Institute of Fisheries and Aquaculture Research, Bergen, Norway Institute of Marine Research, P.O. Box 1870 Nordnes, N-5817 Bergen, Norway

Received 15 February 2006; received in revised form 28 April 2006; accepted 1 May 2006 Available online 15 May 2006

Abstract A novel type of retention indices for alkylphenols and related compounds are proposed. The alkylphenol retention indices (APRI) use parasubstituted n-alkylphenols as reference series. APRI for para-n-alkylphenols are per definition equal to the number of carbon atoms in the alkyl substituent; the value for phenol is zero. Application of the APRI system with different types of derivatisation of the phenolic hydroxy group showed that the derivatisation has limited influence on these indices. Especially para-substituted alkylphenols gave APRI values that could be transferred with high accuracy from one type of derivative to another. By comparing results obtained with different gradients in temperature-programmed GC, it was also shown that APRI is less affected by chromatographic conditions than retention indices based on n-alkanes. © 2006 Elsevier B.V. All rights reserved. Keywords: Gas chromatography; Alkylphenols; Retention indices; APRI

1. Introduction Alkylphenols, and especially long-chained branched parasubstituted isomers, are known to have oestrogenic activity and therefore represent an environmental problem. There are several sources of alkylphenols in the environment. Because of the antioxidant properties nonylphenols are used as plastic additives. Nonylphenol ethoxylates are applied as surfactants and emulsifying agents in a large range of industrial products [1,2]. Alkylphenols are present in raw oil and those of shorter alkyl chain may enter the environment via produced water from offshore oil installations [3]. Produced water is defined as the water that comes up with oil and gas from sea bed reservoirs, separated on the platform from the oil and discharged into the sea. The alkylphenols in produced water show an extreme diversity in molecular structure, both with respect to the alkyl substituent and the position of the substituents in the phenyl ring, and several hundred alkylphenols may be present in produced water. The many isomers represent a challenge in chromatographic analysis of produced water, and accurate characterisation of the chromatographic properties of the compounds would facil-

∗

Corresponding author. Tel.: +47 55236394; fax: +47 55238584. E-mail address: [email protected] (S. Boitsov).

0021-9673/$ – see front matter © 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.chroma.2006.05.002

itate identification. It is well known that retention indices are more reproducible than retention times, and they are therefore more suitable for characterisation of compounds analysed by gas chromatography. On a retention index scale, the retention of a compound is described in relation to the retention of a series of homologues. The most common system is the Kovats indices [4], where n-alkanes are applied as reference compounds. At isothermal conditions there is a linear relationship between log t R and the number of carbon atoms in homologous series, and the Kovats index, I, for a compound, x, can be calculated by the following equation: Ix = 100n

  log tR(x) − log tR(z)

  log tR(z+n) − log tR(z)

+ 100z

(1)

where t R is adjusted retention times of the compound of interest and two n-alkanes eluting on each side of the compound. z represents the number of carbon atoms in the n-alkane eluting before x, and n is the difference in carbon atoms between the two n-alkane references. For maximal accuracy, it is recommended that n is one. Kovats indices acquired at isothermal conditions are usually assumed to be invariant to differences in column dimensions and carrier gas flow, but are highly dependent on the stationary phase and also influenced by the oven temperature. Thus, I acquired at a certain stationary phase at a certain temperature is a characteristic property for a compound that can

S.A. Mjøs et al. / J. Chromatogr. A 1123 (2006) 98–105

be used for identification purposes. In many practical situations, the effect of temperature is small enough to be neglected. The use of retention indices has been extended to temperature-programmed gas chromatography where there exists an approximately linear relationship between retention times and the number of carbon atoms in a homologous series. In temperature-programmed GC, I is usually calculated by the Van den Dool and Kratz equation [5]: Ix = 100n

tR(x) − tR(z) + 100z tR(z+n) − tR(z)

(2)

The parameters n, x and z are the same as in Eq. (1). Eq. (2) gives the same results whether applied with neat retention times or adjusted retention times. As for Eq. (1), it is recommended that n is one. In addition to application of the equations above, various approaches based on higher order regressions have also been applied [6–10]. Although isothermal retention indices are independent of carrier gas flow, column dimensions and phase ratios, temperatureprogrammed indices are not. All these factors will influence the elution temperature of a compound; increasing the temperature gradient, column length or phase ratio, or decreased carrier gas flow rate, will move the I in the same direction as increased temperature in isothermal chromatography. Thus, I acquired under temperature-programmed conditions are generally less reproducible than I acquired under isothermal conditions [11]. While Kovats indices are the dominating general-purpose retention index system, a large number of alternative series with other calibration compounds than n-alkanes have been applied for special purposes, the most successful approach may be the use of equivalent chain lengths [12] for fatty acid analysis. Alternative retention indices for various purposes have been extensively reviewed elsewhere [13,14]. The motivation for using other calibration standards than alkanes are basically the following: (1) n-Alkanes cannot be analysed by several common detection methods such as negative ion chemical ionisation mass spectrometry, electron capture detectors and element specific detectors. (2) Retention indices based on molecules with the same functional groups as the analytes of interest are usually more reproducible, and vary less with chromatographic conditions than I. (3) A second calibration standard of n-alkanes is not necessary if the retention index scale is defined by some of the analytes of interest. (4) n-Alkanes have poor chromatographic properties on highly polar stationary phases. Point 4 is not an issue in alkylphenol analysis, since phases with low to medium polarity are usually preferred. However, the other arguments, especially 1 and 2, are of importance. We therefore propose an alkylphenol retention index (APRI) system based on the homologous series of the para-substituted n-alkylphenols. The index for a para-substituted n-alkylphenol is by definition equal to the number of carbon atoms in the alkyl

99

substituent. Thus, by the principle devised by Van den Dool and Kratz [5], APRI can be calculated by the following equation: APRIx = n

tR(x) − tR(z) +z tR(z+n) − tR(z)

(3)

where tR is retention time of the compound of interest, x, and two para-substituted n-alkylphenols eluting on each side of the compound. z represents the number of carbon atoms in the alkyl chains of the para-alkylphenols eluting before x, and n is the difference in the number of carbon atoms between the two references. z is zero if the first reference compound is phenol. It is shown that APRI is more robust than I towards changes in chromatographic properties (temperature gradient) and that APRI to a certain degree can be transferred between different derivatives of the hydroxyl group that is common in alkylphenol analyses. The proposed system is currently applied for characterisation of alkylphenols in water produced from offshore oil installations in the North Sea. 2. Experimental 2.1. Reference compounds 4-(1,1-Dimethylbutyl)phenol; 4-(1-methyl-2,2-dimethylpropyl)phenol and 4-(1,1-dimethylpentyl)phenol were synthesized from the corresponding tertiary alcohols as described elsewhere [15]. Crystalline products were recrystallised from hexane–dichloromethane 3:1 mixture and hexane several times. Liquid products were separated from solvents by rotavapour. The structures and purity of the synthesized products were confirmed by 400 MHz H1 NMR and GC–MS. Technical nonylphenol was acquired from Sigma–Aldrich (Oslo, Norway). Nonylphenol isomers were identified from the gas chromatographic elution patterns and identifications in [2,16,17]. Other alkylphenols were pure (>97%) reference compounds purchased from Sigma–Aldrich. An n-alkane reference mixture containing every n-alkane from C11 to C26 (except C16) was applied for calculation of I. The n-alkanes were acquired from Kebo-Lab (Oslo, Norway) 2.2. Derivatisations The phenol hydroxy group was derivatised using three different derivatisation methods. Formate esters of the alkylphenols were prepared by derivatisation with methyl chloroformate (MCF) as described by Grahl-Nielsen [18]. Pentafluorobenzyl ethers were prepared by derivatisation with pentafluorobenzyl bromide (PFBB) as described by Nakamura et al. [19]. Pentafluorobenzoyl esters were prepared from pentafluorobenzoyl chloride (PFBC) according to Boitsov et al. [3]. The structure of the different derivatives can be seen in Fig. 1. 2.3. Gas chromatography All analyses were performed on an Agilent 6890 GC system connected to an Agilent 5973 mass selective detector with electron impact ionization, and scanning from m/z 50 to 500. A

100

S.A. Mjøs et al. / J. Chromatogr. A 1123 (2006) 98–105

3. Results and discussion 3.1. Definitions According to IUPAC conventions, Kovats indices (I) are by definition acquired at isothermal conditions [20]. However, to emphasize the difference between the various retention indices, the term I is in this work used for retention index scales for linear temperature programs where the index for an n-alkane by definition is equal to the number of carbon atoms multiplied by 100. The term APRI is used for the novel retention indices where the index for a para-substituted n-alkylphenol by definition is equal to the number of carbon atoms in the alkyl substituent. The term RI means any retention index. Fig. 1. Structure of derivatised and free alkylphenols and relationship between retention time and number of carbons in the alkyl chain for para-n-alkylphenols. The alkylphenol retention index (APRI) is per definition equal to the number of carbons in the alkyl chain in para-n-alkylphenols.

Varian FactorFour VF-5ms (Varian, Lake Forest, CA, USA), L = 50 m, I.D. = 0.25 mm, df = 0.25 ␮m, was used as analytical column. Helium (99.9999%) was used as carrier gas in 22 cm/s constant flow mode. The samples (1 ␮l hexane solution) were injected in splitless mode and the split valve was opened after 2 min. Three different temperature programs were applied. The temperature was 60 ◦ C at injection. After 2 min the temperature was increased to 100 ◦ C by a temperature programming rate of 30 ◦ C/min, which was followed by a second rate of 1 ◦ C/min (Program A), 2 ◦ C/min (Program B) or 4 ◦ C/min (Program C). 2.4. Calculations Retention indices were calculated by an in-house written program, ‘Q (10-05)’, programmed in Matlab 6.5 (Mathworks, Natick, MA, USA). This program calculates the relationship between retention indices and retention time (at the peak apex) by stepwise second order regressions according to ref. [9]. This calculation method gives more smooth regressions than stepwise linear regressions (Eqs. (2) and (3)) in areas with non-linear relationships between retention times and retention indices. The calculated regression lines for the relationships between alkylphenol retention indices and retention times for Program B is shown in Fig. 1. The root mean squared deviation (RMSD) is used to compare retention indices calculated at different conditions (variation in derivative or temperature gradients). The RMSD for comparison of two variables, A and B, is calculated according to the equation below:
 N 1 RMSD =  (4) (An − Bn )2 N n=1

Both poor correlation and systematic difference (bias) between the two variables will give increased RMSD.

3.2. APRI-indices for different derivatives In a complex molecule with several functional groups, the retention index can be divided into contributions from each functional group and interaction effects between the different groups [19], and can be expressed as in Eq. (5):  (5) RI RI = RIderivative + RIphenol + RIalkyl + interactions

It should be emphasised that the different factors cannot always be accurately determined, and that that there are alternative ways to separate the factors. The phenol oxygen atom may for instance be considered a part of a derivative that is connected to a phenyl ring. Alkylphenols may also have several alkyl groups connected to the phenyl ring. In a retention index system where the reference series have the same functional groups as the analytes of interest, the two first factors in Eq. (5) may be excluded and the APRI retention index may be expressed as in Eq. (6):  APRI = APRIalkyl + APRIinteractions (6) Interaction effects in Eq. (6) include both interactions between the alkyl substituents and the derivative, and interactions between different alkyl groups in the molecule. Thus, if the interaction effects between the derivative and the alkyl groups are negligible, the APRI should be independent of hydroxy group derivatisation. The largest interaction effects for alkylphenols are the so-called “ortho effects” [21–24]. These effects occur in molecules with large substituents in ortho-position, which may effectively shield for interactions between the stationary phase and the polar functional group. APRI for 45 alkylphenols acquired with free phenols and with the three derivatives are listed in Table 1. Statistics for comparison of the derivatives are given in Table 2. The best correlation was found between PFBC and MCF derivatives; the worst correlation was between PFBC and free phenols. APRI for PFBC is plotted against MCF in Fig. 2a and against free phenols in Fig. 2b. The largest differences between the two derivatives were seen for 2,3,6-trimethylphenol (16); 2,6-dimethylphenol (3) and 2,4,6-trimethylphenol (12) where APRI for PFBC was approximately 0.3 units higher than the index for MCF. These three compounds are the only compounds in the data set with

S.A. Mjøs et al. / J. Chromatogr. A 1123 (2006) 98–105

101

Table 1 APRI for alkylphenols analysed with different derivatives No.

Compound

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45

2-Methylphenol 3-Methylphenol 2,6-Dimethylphenol 2-Ethylphenol 2,4-Dimethylphenol 2,5-Dimethylphenol 3-Ethylphenol 3,5-Dimethylphenol 2,3-Dimethylphenol 3,4-Dimethylphenol 2-Isopropylphenol 2,4,6-Trimethylphenol 2-n-Propylphenol 4-Isopropylphenol 3-Isopropylphenol 2,3,6-Trimethylphenol 3-n-Propylphenol 3-Ethyl-5-methylphenol 2-tert-Butylphenol 2,3,5-Trimethylphenol 3-tert-Butylphenol 4-tert-Butylphenol 4-sec-Butylphenol 2-Methyl-5-isopropylphenol 3-Methyl-4-isopropylphenol 2-tert-Butyl-4-methylphenol 2-Methyl-4-tert-butylphenol 2,6-Di-isopropylphenol 2-tert-Butyl-5-methylphenol 4(1,1-Dimethylpropyl)phenol 2-tert-Butyl-4-ethylphenol 2,5-Di-isopropylphenol 4(1,1-Dimethylbutyl)phenol 4(1,2,2-Trimethylpropyl)phenol 4(1-Ethyl-1-methylpropyl)-2-metylphenol 4(1,1-Dimethylpentyl)phenol 4(1,1-Dimethylhexyl)phenol 2-Methyl-4(1,1-dimethylhexyl)phenol 4(1,3-Dimethyl-1-propylbutyl)phenol 4(1,1,3-Trimethylhexyl)phenol 4(1-Ethyl-1,4-dimethylpentyl)phenol 4(1,1,4-Trimethylhexyl)phenol 4(1-Ethyl-1,3-dimethylpetyl)phenol 4(1-Methyl-2-ethylhexyl)phenol 4(1,1,2-Trimethylhexyl)phenol

a b c d e f

Note

a a b

a

b b

a a a b a a b b b b b,c,d,e b,c b,d,f b,c b,d b,f b,c,d

Free phenol

MCF

PFBB

PFBC

0.7866 1.0000 1.4318 1.6743 1.8154 1.8401 2.0038 2.0372 2.1441 2.3166 2.3461 2.4977 2.5997 2.6580 2.6614 2.8106 3.0000 3.0072 3.1241 3.1619 3.3388 3.3528 3.5541 3.5600 3.7385 3.9199 3.9819 4.0084 4.0429 4.3823 4.6403 4.7275 5.1490 5.3420 5.8761 6.0371 6.4168 6.7943 7.2605 7.3569 7.4359 7.5218 7.5405 7.7442 7.8169

0.5849 0.9349 1.1660 1.3056 1.6486 1.5912 1.8500 1.9824 1.9192 2.3649 1.7625 2.2517 2.1230 2.6572 2.4214 2.5407 2.8131 2.8921 2.5327 2.9669 3.0478 3.3649 3.5350 3.3585 3.7668 3.4568 3.8159 3.3440 3.4568 4.3715 4.1552 3.9852 5.1216 5.3225 5.6228 5.9764 6.3177 6.4881 7.1089 7.2192 7.3119 7.4150 7.3998 7.6392 7.7089

0.6288 0.9004 1.4177 1.3071 1.6405 1.4857 1.8104 1.7899 1.9538 2.2590 1.6968 2.4240 2.0425 2.6100 2.2840 2.7266 2.7266 2.6429 2.5716 2.7694 2.8458 3.2962 3.4840 3.0456 3.5623 3.4108 3.6969 2.9869 3.3742 4.3192 4.0729 3.6263 5.0179 5.2832 5.5311 5.8532 6.2124 6.3648 6.9368 7.1284 7.1501 7.2656 7.2805 7.4753 7.6112

0.6721 0.8973 1.4729 1.3177 1.7283 1.6356 1.8628 1.7787 1.9989 2.3246 1.6346 2.5461 2.0673 2.6389 2.2991 2.8637 2.7300 2.7322 2.4619 2.9890 2.8637 3.3391 3.5283 3.1571 3.6858 3.3531 3.8519 3.1807 3.3361 4.3760 4.0479 3.7367 5.1130 5.3500 5.7205 5.9460 6.3436 6.6067 7.0772 7.2104 7.2767 7.3771 7.4064 7.6219 7.6942

Large (>C2) substituent in ortho-position (see Fig. 2). Para-substituted only (see Fig. 2). Tentatively identified according to [2]. Tentatively identified according to [16]. Tentatively identified as 4(1,3-dimethyl-1-isopropylbutyl)phenol in [17]. Tentatively identified according to [17].

methyl-substitutes in both ortho-positions. However, the effect causing the deviations is not the typical ortho-effect caused by shielding of the functional group. All compounds with large groups in the ortho-position (open circles in Fig. 2) had higher values for MCF than PFBC derivatives. This effect is more pronounced when PFBC is compared to un-derivatised phenols in Fig. 2b. The eight compounds with larger groups than ethyl in ortho positions are the eight compounds with the highest deviations and all these compounds have larger values for free phe-

nols than PFBC. The two largest deviations were observed for 2,5-diisopropylphenol (32) and 2,6-diisopropylphenol (28) with 0.99 and 0.87 units, respectively. Large deviations are also found for several compounds with large substituents in meta-position, such as 3-tert-butylphenol (21), 2-methyl-5-isopropylphenol (24) and 3-isopropylphenol (15). The alkylphenols that are only para-substituted show high correlation between all derivatives with r2 > 0.999. The RMSD for all comparisons were <0.2 and as low as 0.023 when MCF

102

S.A. Mjøs et al. / J. Chromatogr. A 1123 (2006) 98–105

Table 2 Statistics based on data in Table 1 for APRI for alkylphenols analysed as different derivatives Derivative

All

aa

ba

Correl.

Free OH Free OH Free OH MCF MCF PFBB

MCF PFBB PFBC PFBB PFBC PFBC

0.9913 0.9864 0.9852 0.9969 0.9966 0.9989

Para-substituted only (r2 )b

Biasc

RMSD

Correl. (r2 )b

Biasc

RMSD

−0.219 −0.323 −0.209 −0.132 −0.018 0.115

0.292 0.389 0.344 0.150 0.121 0.105

0.9998 0.9994 0.9997 0.9997 0.9999 0.9999

−0.073 −0.176 −0.085 −0.103 −0.011 0.092

0.092 0.200 0.105 0.112 0.023 0.099

a Free OH: free phenols; MCF: formate esters; PFBB: pentafluorobenzyl ethers; PFBC: pentafluorobenzoyl esters. The structures of derivatised compounds (as para-substituted phenols) are illustrated in Fig. 1. b Coefficient of determination (r2 ) for linear regression between APRI acquired with the different derivatives. c Bias: mean b − mean a.

was compared with PFBC. For comparison, the peak width at half height measured on PFBC derivatives of 4-n-alkylphenols varied from 0.011 to 0.016 APRI units. Thus, it can be concluded that APRI for para-substituted alkylphenols can be transferred between different derivatives with high accuracy. Correct identification of alkylphenol structures in complex mixtures containing a large number of isomers, such as produced water from oil installations, is a difficult task. In these cases, it is important for ecotoxicological purposes to determine whether a specific compound is a para-substituted alkylphenol with no other alkyl groups, or if large groups are present in ortho-position. The para-substituted alkylphenols have been shown to have high oestrogenic activity, while ortho-substituted alkylphenols do not interact with the oestrogen receptor [25]. Comparison of APRI of different derivatives may therefore help in evaluating the oestrogenic activity of a given compound. An alternative may be to analyse the same derivatives on columns with different polarities. Different patterns may be found on other stationary phases; phases with high affinities for free OHgroups may for example be expected to give lower APRI for ortho-substituted compounds when underivatised phenols are compared to derivatised phenols. 3.3. Comparison of APRI and I

Fig. 2. APRI for MCF derivatives (a) and free phenols (b) plotted against PFBC derivatives. Values for para-n-alkylphenols are given by definition. The line y = x marks the ideal line with identical APRI values for the derivatives in the plots. The numbers on single observations with large deviations from y = x refer to Table 1.

An important issue is the robustness of retention indices towards differences in chromatographic conditions. Although differences in RI caused by different chromatographic conditions may be utilized for identification purposes [8], indices that are stable towards differences in temperature programs and column dimensions are usually preferred. The stability of APRI and I were investigated by applying temperature gradients of 1, 2 and 4 ◦ C/min, using GC programs A–C, respectively (see Section 2.3). Both free phenols and PFBC derivates were analysed. Retention indices and statistics describing the differences for RI acquired by the 2 and 4 ◦ C/min gradients relative to the RI acquired by 1 ◦ C/min are given in Table 3. Because the differences between two homologues by definition is 100 in the Kovats system and 1 in the APRI system, Bias and RMSD for APRI are multiplied by 100 to get error estimates that are directly comparable. The terms “RMSD100 ” and “Bias100 ” are used for these estimates.

Table 3 APRI and I for free phenols and PFBC derivatives analysed with three programs with different temperature gradients No.

Compounds

Free phenols, APRI Program

2-Methylphenol 3-Methylphenol 2,6-Dimethylphenol 2-Ethylphenol 2,4-Dimethylphenol 2,5-Dimethylphenol 3-Ethylphenol 3,5-Dimethylphenol 2,3-Dimethylphenol 3,4-Dimethylphenol 2-Isopropylphenol 2,4,6-Trimethylphenol 2-n-Propylphenol 4-Isopropylphenol 3-Isopropylphenol 2,3,6-Trimethylphenol 3-n-Propylphenol 2-tert-Butylphenol 2,3,5-Trimethylphenol 3-tert-Butylphenol 4-tert-Butylphenol 4-sec-Butylphenol 2-Methyl-5-isopropylphenol 3-Methyl-4-isopropylphenol 2-tert-Butyl-4-methylphenol 2,6-Di-isopropylphenol 2-tert-Butyl-5-methylphenol 4(1,1-Dimethylpropyl)phenol 4(1,1-Dimethylbutyl)phenol 4(1,2,2-Trimethylpropyl)phenol 4(1,1-Dimethylpentyl)phenol 4(1,1-Dimethylhexyl)phenol 2-tert-Butyl-6-methylphenol Deviations Correl. (r2 )b RMSD Biasc RMSD100 d Bias100 c

Program

Free phenols, I Ba

Program

Ca

Program

Aa

PFBC, APRI

Program

Ba

Program

1.4252 1.6810 1.8217 1.8431

1.4260 1.6699 1.8180 1.8363

1.4403 1.6670 1.8125 1.8347

1114.2 1136.2 1148.9 1150.9

1114.9 1136.5 1150.1 1151.8

1117.2 1137.9 1151.4 1153.5

2.0328 2.1310 2.3056 2.3431 2.4832 2.5990

2.0278 2.1357 2.3083 2.3399 2.4932 2.5950

2.0243 2.1426 2.3140 2.3354 2.5051 2.5925

1169.5 1178.9 1195.4 1198.9 1211.0 1221.3

1170.0 1180.1 1196.3 1199.2 1212.9 1222.2

1171.3 1182.2 1198.2 1200.2 1215.8 1223.9

2.6644 2.7927

2.6560 2.8041

2.6491 2.8224

1227.3 1239.3

1227.9 1241.9

1229.2 1245.7

3.1244 3.1532 3.3463 3.3463 3.5514 3.5644 3.7384 3.9257 4.0016 4.0436 4.3671

3.1206 3.1569 3.3365 3.3454 3.5483 3.5541 3.7346 3.9162 4.0027 4.0388 4.3777

3.1173 3.1647 3.3355 3.3474 3.5490 3.5405 3.7355 3.9077 4.0051 4.0336 4.3903

1273.2 1276.1 1294.7 1294.7 1313.7 1315.0 1332.2 1352.2 1361.1 1365.4 1397.2

1273.5 1277.2 1294.8 1295.6 1315.3 1315.8 1334.0 1353.1 1362.7 1366.4 1400.1

1274.9 1279.7 1296.7 1297.8 1317.9 1317.0 1336.9 1355.0 1365.6 1368.5 1404.4

6.3888 3.4627

6.4145 3.4742

6.4432 3.4878

1602.3 1305.3

1608.2 1308.0

1615.5 1311.7

0.9999 0.0086 −0.0002 0.86 −0.02

0.9998 0.0179 0.0024 1.79 0.24

0.9999 1.79 1.35 1.79 1.35

Ca

Program 0.6647 0.8982 1.4590 1.3135 1.7252 1.6367 1.8762 1.7850 1.9972 2.3186 1.6431 2.5363 2.0660 2.6386 2.3093 2.8487 2.7358 2.4535 2.9831 2.8757 3.3347 3.5225 3.1795 3.6855 3.3526 3.1976 3.3347 4.3605 5.1089 5.3218 5.9404 6.3097

0.9997 4.47 3.72 4.47 3.72

PFBC, I Program

Ba

Program

0.6712 0.8975 1.4716 1.3176 1.7286 1.6329 1.8639 1.7789 1.9990 2.3226 1.6370 2.5435 2.0650 2.6381 2.2993 2.8643 2.7268 2.4638 2.9825 2.8643 3.3365 3.5252 3.1533 3.6816 3.3545 3.1768 3.3365 4.3748 5.1127 5.3496 5.9456 6.3436

0.6748 0.8902 1.4878 1.3152 1.7332 1.6341 1.8499 1.7732 2.0280 2.3263 1.6282 2.5547 2.0623 2.6328 2.2829 2.8891 2.7175 2.4725 2.9850 2.8490 3.3408 3.5286 3.1229 3.6767 3.3487 3.1517 3.3329 4.3906 5.1239 5.3784 5.9482 6.3756

0.9999 0.0117 0.0014 1.17 0.14

0.9997 0.0255 0.0029 2.55 0.29

Ca

Program Aa

Program Ba

Program Ca

1552.6 1576.4 1630.7 1616.8 1655.6 1647.4 1669.3 1661.0 1680.0 1708.0 1648.0 1728.0 1685.9 1737.5 1707.1 1756.9 1746.5 1720.3 1769.2 1759.4 1804.8 1823.6 1789.2 1840.1 1806.6 1791.0 1804.8 1909.3 1984.4 2006.1 2069.5 2107.6

1555.1 1578.2 1633.7 1619.1 1657.8 1648.9 1670.2 1662.5 1682.3 1710.9 1649.3 1731.3 1688.0 1740.1 1708.8 1761.1 1748.3 1723.9 1771.8 1761.1 1807.7 1826.7 1789.2 1842.6 1809.5 1791.6 1807.7 1913.7 1987.8 2011.9 2073.1 2114.2

1558.1 1580.1 1638.0 1621.5 1661.0 1651.8 1671.7 1664.7 1687.7 1714.4 1651.2 1735.6 1690.7 1742.9 1710.4 1766.7 1750.8 1727.9 1775.5 1763.0 1811.6 1830.5 1789.6 1845.6 1812.4 1792.5 1810.8 1918.7 1992.6 2018.7 2077.3 2121.3

0.9999 2.92 2.65 2.92 2.65

0.9997 6.52 6.00 6.52 6.00

103

Deviation statistics are for Program B and Program C relative to Prg. A. a Temperature gradients: Program A: 1 ◦ C/min, Program B: 2 ◦ C/min, Program C: 4 ◦ C/min, see methods section for further details. b Coefficient of determination (r2 ) for linear regression between Prg.1 and other programs. c Mean of Program A subtracted from mean of Program B or Program C. d RMSD and Bias for APRI multiplied by 100, RMSD and bias for I multiplied by 1. See text for details.

Aa

S.A. Mjøs et al. / J. Chromatogr. A 1123 (2006) 98–105

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 19 20 21 22 23 24 25 26 28 29 30 33 34 36 37 46

Aa

104

S.A. Mjøs et al. / J. Chromatogr. A 1123 (2006) 98–105

Table 4 Kovats indices for para-n-nonylphenols analysed as free phenols and PFBC derivatives. Free phenols Program Phenol 4-Methylphenol 4-Ethylphenol 4-n-Propylphenol 4-n-Butylphenol 4-n-Pentylphenol 4-n-Hexylphenol 4-n-Heptylphenol 4-n-Octylphenol 4-n-Nonylphenol a b

1004.1b 1077.3 1165.8 1259.9 1360.6 1460.0 1561.6 1665.5 1768.1 1871.8

Aa

PFBC derivatives Program 995.5b 1076.0 1167.2 1260.4 1361.8 1462.5 1565.0 1669.0 1772.3 1876.7

Ba

Program

Ca

986.8b 1075.3 1168.5 1261.7 1364.4 1465.9 1568.8 1673.8 1778.2 1883.4

Program Aa

Program Ba

Program Ca

1480.9 1586.2 1680.0 1770.2 1872.9 1973.0 2075.2 2178.4 2281.1 2383.9

1482.4 1588.2 1682.3 1773.2 1875.9 1976.3 2078.6 2181.8 2284.5 2387.1

1484.7 1590.9 1685.1 1776.6 1879.3 1979.7 2082.4 2185.2 2287.9 2391.3

Temperature gradients: Program A: 1 ◦ C/min, Program B: 2 ◦ C/min, Program C: 4 ◦ C/min, see Section 2 for further details. Tailing peak, I may be inaccurate.

There are minimal differences in the coefficients of determination (r2 ) from linear regressions between retention indices acquired with the different programs; the differences are basically found in the bias and RMSD. As can be expected, the differences between 1 and 4 ◦ C temperature gradients are larger than the differences between 1 and 2 ◦ C. While the APRI indices are nearly unbiased (Bias100 < 0.3), the bias for I varies between 1.4 and 6.0. The differences in bias can be explained by the differences between Eqs. (5) and (6). It is known that I of compounds with aromatic rings show a relatively high temperature dependence on stationary phases with low polarity [26–28]. Thus, the different temperature gradients will have large effects on RIderivative and RIphenol in Eq. (5). The I bias for PFBC derivatives, which have two aromatic rings, is 1.6–2.0 times larger than the bias of the free phenols, which have one aromatic ring. In the APRI system the effects of the phenol and the derivative are eliminated (Eq. (6)). The differences in bias affect the overall error (RMSD100 ), which is 2.1–2.6 times larger for the I system than for the APRI system. Thus, it can be concluded that APRI gives more robust indices than I for alkylphenols. I for the 4-n-alkylphenols are given in Table 4. With the exception of phenol and 4-methylphenol, I increase with the temperature gradient, i.e. I at 4 ◦ C is higher than I at 1 ◦ C. The effect of the temperature also increases with the chain length of the alkyl group, i.e. the temperature gradient has the largest effects on the compounds with the largest alkyl groups. The addition of methylene units in homologous series are usually expected to give increments of approximately 100 I per added unit, but exceptions for the first members of the homologous series are frequently reported [22,24,29–32]. The addition of a para-methyl group to the free phenol increases I by approximately 80, the difference between 4-ethylphenol (C2) and 4-methylphenol (C1) is approximately 91; and the increments approach 104 (average of four last increments) in an asymptotic manner. A similar pattern has previously been reported for alkylbenzenes [29,30]. In contrast, the PFBC derivatives follow a surprisingly complex pattern. The increments are 106 from C0 to C1, but only 94 from C1 to C2 and 91 from C2 to C3. The effect of methy-

lene addition then rises sharply to 103 (C3–C4) and falls to 100 (C4–C5) before the effect stabilizes around 103 I units per added methylene unit. The pattern is reproducible and almost identical for all three temperature gradients. The fact that the I increments per methylene unit approach values slightly above 100 has been explained by influence from adjacent electronegative groups [21]. However, another explanation is that contributions from the aromatic groups, RIderivative and RIphenol (Eq. (5)), are affected by the higher elution temperature of the higher homologues and contribute to a stronger retention. 4. Conclusions Alkylphenol retention indices were calculated with paran-nonylphenols as references. Derivatisation of the phenolic hydroxy group has limited influence on these indices, and especially APRI values for para-substituted alkylphenols can be transferred with high accuracy from one type of derivative to another. By comparing results obtained with different gradients in temperature-programmed GC, it was also shown that APRI is less affected by chromatographic conditions than Kovats indices. Acknowledgements We are grateful to Norwegian Research Council for financial support within project 164401/S40 and to Professor Jon Songstad at University of Bergen for help with the synthesis of reference compounds. References [1] A.C. Nimrod, W.H. Benson, Crit. Rev. Toxicol. 26 (1996) 335. [2] Y.S. Kim, T. Katase, S. Sekine, T. Inoue, M. Makino, T. Uchiyama, Y. Fujimoto, N. Yamashita, Chemosphere 54 (2004) 1127. [3] S. Boitsov, S. Meier, J. Klungsøyr, A. Svardal, J. Chromatogr. A 1059 (2004) 131. [4] E. Kov´ats, Helv. Chim. Acta 41 (1958) 1915. [5] H. Van den Dool, P.D. Kratz, J. Chromatogr. 11 (1963) 463. [6] F.J. Heeg, R. Zinburg, H.J. Neu, K. Ballschmiter, Chromatographia 12 (1979) 458.

S.A. Mjøs et al. / J. Chromatogr. A 1123 (2006) 98–105 [7] R. Lebr´on-Aguilar, J.E. Quintanilla-L´opez, J.A. Garc´ıa-Dom´ınguez, J. Chromatogr. A 945 (2002) 185. [8] S.A. Mjøs, J. Chromatogr. A 1015 (2003) 151. [9] S.A. Mjøs, J. Chromatogr. A 1061 (2004) 201. [10] F.R. Gonzales, J.L. Alessandrini, A.M. Nardillo, J. Chromatogr. A 810 (1998) 105. [11] F.R. Gonzalez, A.M. Nardillo, J. Chromatogr. A 842 (1999) 29. [12] T.K. Miwa, K.L. Mikolajczak, F.R. Earle, I.A. Wolff, Anal. Chem. 32 (1960) 1739. [13] G. Castello, J. Chromatogr. A 842 (1999) 51. [14] M.B. Evans, J. Chromatogr. 472 (1989) 93. [15] R.C. Huston, T.Y. Hsieh, J. Am. Chem. Soc. 58 (1936) 439. [16] B. Thiele, V. Heinke, E. Kleist, K. Guenther, Environ. Sci. Technol. 38 (2004) 3405. [17] T.F. Wheeler, J.R. Heim, M.R. LaTorre, B. Janes, J. Chromatogr. Sci. 35 (1997) 19. [18] O. Grahl-Nielsen, Sarsia 72 (1987) 375.

[19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29] [30] [31] [32]

105

S. Nakamura, T.H. Sian, S. Daishima, J. Chromatogr. A 919 (2001) 275. L.S. Ettre, Pure Appl. Chem. 65 (1993) 819. C.T. Peng, J. Chromatogr. A 678 (1994) 189. C.T. Peng, S.F. Ding, R.L. Hua, Z.C. Yang, J. Chromatogr. 436 (1988) 137. C.T. Peng, Z.C. Yang, S.F. Ding, J. Chromatogr. 586 (1991) 85. W. Engewald, U. Billing, I. Topalova, N. Petsev, J. Chromatogr. 446 (1988) 71. E.J. Routledge, J.P. Sumpter, J. Biol. Chem. 272 (1997) 3280. Analytical Methods Committee, Analyst 122 (1997) 1167. J. Curvers, J. Rijks, C. Cramers, K. Knauss, P. Larsson, J. High Resolut. Chromatogr. Chromatogr. Commun. 8 (1985) 611. I.O. Korhonen, J. Chromatogr. 363 (1986) 277. R.V. Golovnya, J. Chromatogr. 364 (1986) 193. C.T. Peng, J. Chromatogr. A 903 (2000) 117. J.R. Ashes, J.K. Haken, J. Chromatogr. 101 (1974) 103. J.R. Ashes, J.K. Haken, J. Chromatogr. 111 (1975) 171.