Journal of Chromatography A, 1327 (2014) 141–148
Contents lists available at ScienceDirect
Journal of Chromatography A journal homepage: www.elsevier.com/locate/chroma
Ultra high efficiency/low pressure supercritical fluid chromatography with superficially porous particles for triglyceride separation夽 E. Lesellier a,∗ , A. Latos a , A. Lopes de Oliveira b a
Institut de Chimie Organique et Analytique (ICOA), Université d’Orléans, CNRS UMR 7311, B.P. 6759, rue de Chartres, 45067 Orléans cedex 2, France Faculdade de Zootecnia e Engenharia de Alimentos, Universidade de São Paulo, FZEA/USP Caixa Postal 23, CEP: 13635-900, Pirassununga, São Paulo, France b
a r t i c l e
i n f o
Article history: Received 26 September 2013 Received in revised form 12 December 2013 Accepted 16 December 2013 Available online 25 December 2013 Keywords: Column coupling Derringer function Method development UHE/LP-SFC Superficially porous particle Triglycerides
a b s t r a c t This paper reports the development of the separation of vegetable oil triglycerides (TG) in supercritical chromatography (SFC), using superficially porous particles (SPPs). The SPP, having a small diameter (2–3 m), provide a higher theoretical plate number (N), which allows to improve separation of critical pairs of compounds. However, compared to fully porous particles of larger diameter (5 m), the pressure drop is also increased. Fortunately, supercritical fluids have a low viscosity, which allows coupling several columns to achieve high N values, while maintaining flow rate above 1 ml/min, ensuring a ultra high efficiency (UHE) at low pressure (LP) (below 40 MPa), with regards to the one reached with liquid and sub-two micron particles (around 100 MPa). The use of two detector systems (UV and ELSD) connected in series to the UHE–LP-SFC system provides complementary responses, due to their specific detection principles. Working in a first part with three coupled Kinetex C18 columns (45 cm total length), the effect of modifier nature and percentage were studied with two reference oils, argan and rapeseed, chosen for their different and well-known TG composition. The analytical method was developed from previous studies performed with fully porous particles (FPP). Optimized conditions with three Kinetex were as follows: 17 ◦ C, 12% of ACN/MeOH (90/10; v/v). With these conditions, and by using an increased length of Kinetex C18 column (60 cm), another additional column was selected from ten different commercial SPP C18 bonded phases, by applying a Derringer function on varied parameters: theoretical plate number (TPN), separation index (SI) for critical pairs of peaks (the peaks of compounds difficult to separate due to subtle structural differences), the analysis duration, and the total peak number. This function normalizes the values of any parameters, between 0 and 1, from the worst value to the better, allowing to take account of various parameters in the final choice. Finally, by using four Kinetex C18 plus one Accucore C18 (75 cm total column length), a high-performance separation of triglycerides was achieved, with reasonable analysis duration and isocratic conditions. These conditions can be applied to varied vegetable oils. Identification of the numerous separated peaks of rapeseed oil was achieved by using published data and chromatographic retention behaviour. © 2013 Elsevier B.V. All rights reserved.
1. Introduction Triglycerides (TG) are the main constituents of vegetable oils. They are composed by three fatty acid chains, having varied length, from 14 to 24 carbons, and double bonds, from 0 to 4 (Table 1). Due to their low volatility, they are generally not analyzed in gaseous phase chromatography (GC), but either by high-performance liquid
夽 Presented at the 39th International Symposium on High-Performance LiquidPhase Separations and Related Techniques, Amsterdam, Netherlands, 16–20 June 2013. ∗ Corresponding author. Tel.: +33 2 28 49 45 88; fax: +33 2 38 41 72 81. E-mail address: [email protected] (E. Lesellier). 0021-9673/$ – see front matter © 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.chroma.2013.12.046
chromatography (HPLC) or by supercritical fluid chromatography (SFC). Two HPLC methods can be used, reverse phase liquid chromatography (RPLC) [1–5], or silver-ion mode [6–8]. With the former, using non-polar octadecylsiloxane-bonded silica (ODS) stationary phases, the retention of TG increases with the partition number (PN). PN is calculated from the total carbon number of the fatty acid chain, minus twice the total double bond number. For instance, this partition number is equal to 36 for LnLnLn (Ln = C18:3), this value is obtained from this calculation: PN = 3 × 18 − 2 × 9 = 36; 18 being the carbon number of each Ln chain and 9 the total double bond number. Moreover, due to the low polarity of the studied compounds (log P > 5), the use of water is generally avoided in the mobile phase, leading to a non-aqueous mode, called NARP. Elution gradients are also often used [1,3–5] to
142
E. Lesellier et al. / J. Chromatogr. A 1327 (2014) 141–148
Table 1 List of name, abbreviation, carbon number (CN), double-bond number (DB) and partition number (PN) for main fatty acids of vegetable oils. Trivial name
Abbreviation
CN
DB
PN
Myristic Palmitic Palmitoleic Margaric Margaroleic Stearic Oleic Linoleic Linolenic Arachidic Gadoleic Behenic Lignoceric
M P Po Ma Mo S O L Ln A G B Li
14 16 16 17 17 18 18 18 18 20 20 22 24
0 0 1 0 1 0 1 2 3 0 1 0 0
14 16 14 17 15 18 16 14 12 20 18 22 24
reduce the retention time which could be equal to 300 min in isocratic mode with two coupled columns [2] against 100–140 min in gradient mode [3,5]. However, the separation of several critical pairs remains difficult by NARP-LC [5]. First, separation of compounds composed by three C18 chains, having the same total carbon and double-bond number (identical PN), but with a different repartition onto the three chains, for instance: LLLn and OLnLn, LLL and OLLn, OLL and OOLn or SLL and SOLn [9,10]. These couples can be separated by HPLC using two-coupled C18 column, and an elution gradient with water in the first part, thereby increasing the analysis duration from 100 to 140 min [3]. Another separation seems impossible whatever the system used: the separation between couples SLL/POL, SLLn/POLn and SOL/POO having the same partition number with one saturated chain, either S (C18) or P (C16) [3,11]. The same difficulties of separation are encountered for the couples P0 OO/OOL or P0 PO/POL [5]. Separations of regio-isomers (sn-2; sn-3) are achieved by NARPLC, when using polymeric bonded ODS phases at low temperature [12], whereas LC–MS/MS systems allow to quantify these isomers on the basis of ion abundance [9]. The use of low temperature in NARP-LC was also reported in other studies but the retention times were prohibitive [13,14]. For this kind of separation, temperatures of 25–40 ◦ C are preferred for silver-ion chromatography to favour good peak shapes [15]. Silver-ion chromatography displays a different retention order in regards of the one in NARP-LC, as larger double-bond number induces longer retention. It also allows the separation of regioisomers, having the same chains but with a different position onto the glycerol moiety (as OLL and LOL) [8]. Unfortunately, silverion chromatography suffers from poor reproducibility and peak shapes, requires long equilibration times and freshly prepared mobile phases [16]. However, reproducibility seems improved by the addition of 2-propanol to the hexane–acetonitrile mobile phase, favouring the miscibility between the two previous solvents [15]. Increase in theoretical plate number can improve overall separation performance. This can be achieved by using superficially porous particles (SPPs) increasing the total column length. Despite the lower viscosities of non-aqueous mobile phases in regards to aqueous ones, the column length, and/or the flow rate, is limited when using pumps allowing for classical inlet pressure conditions [17,18]. Recently the use of SPP, also called core-shell or fused-core particles, has allowed to couple columns in HPLC, for the analysis of a very complex oil sample [197]. Four columns were serially coupled, and the inlet pressure was around 900 bars at the end of the elution gradient. 137 compounds from menhaden (fish) oil were identified with mass spectrometry, with total analysis duration equal to 190 min.
Two-dimensional studies using both liquid modes were achieved [19,20]. In this case, silver-ion mode is performed first, then NARP-LC. However, despite the short column used in the second dimension, the flow rate in the first one should be in the 10–20 l/min range, to ensure the separation of all peaks in the second dimension. This low flow rate dramatically increases the analysis duration. SFC has also been used for the separation of TG, with ODS stationary phases [21–26]. Due to the low viscosity of the CO2 -based mobile phase, high flow rates and long columns together provide shorter retention time and greater separation efficiencies in regards to HPLC. By using C18-bonded phases, the general retention rules are identical to those observed in RPLC, i.e. they are based on the partition number. Besides, due to the nature of the mobile phase, selectivity between TG having P chains are somewhat modified due to the increase in the P chain solubility in SFC, leading to improved separation. For instance the retention order in NARP-LC is LLL/OLL/PLL, whereas in SFC it is LLL/PLL/OLL. The improved solubility of the P chains in regards to the O one provides a retention inversion of the TG differing only by one P or one O chain. Moreover, the selectivity between the three (LLL, PLL, OLL) is strongly improved in SFC due, in part to the greater peak capacity and to the larger retention difference between PLL and OLL. Recently, SPP were also used in SFC to achieve 120,000 theoretical plates, by coupling four 15 cm columns [27]. Due to their reduced particle size, and to their great particle size homogeneity, these SPP phases provide higher efficiencies, with a reduced inlet pressure, compared to the ones observed using sub-2 m fully porous particles. This property, in addition of the low viscosity of super(sub)critical fluids would allow the use of very long columns to improve the separation efficiency for TG, without the need for ultra-high pressure chromatographic systems. The use of such high column length leads to get ultra high efficiency/low pressure SFC (UHE/LP-SFC). This paper will describe the method development used in UHE/LP-SFC, based on a previous separation obtained by SFC using fully porous particles (FPP) (Hypersil ODS). The composition of the mobile phase, the flow rate, temperature, backpressure and the choice of the stationary phases will be studied to ensure the best TG separation. Finally, the coupling of five columns filled with superficially porous particles, allows improving the separation performance of TG of vegetable oils with regard to previous works performed in the last century [23].
2. Materials and methods Chromatographic separations were carried out using equipment manufactured by Jasco (Tokyo, Japan). One pump model 2080-CO2 Plus was used for carbon dioxide and a second one model 2080 Plus for the modifier. Control of the mobile phase composition was performed by the CO2 pump. When the two solvents (modifier and CO2 ) were mixed, the fluid was introduced into a dynamic mixing chamber PU 4046 (Pye Unicam, Cambridge, UK) connected to a pulsation damper (Sedere, Orleans, France). The injector valve was supplied with a 20 L loop (model 7125 Rheodyne, Cotati, CA). The columns were thermostated by an oven (Jetstream 2 Plus, Hewlett-Packard, Palo Alto, CA). The detector was a Gilson UV 151 detector (provided by Waters, Milford, MA) equipped with a pressure-resistant cell. The detection wavelength was 210 nm. After the detector, the outlet column pressure was controlled by a Jasco BP-2080 Plus back pressure regulator (BPR). The outlet regulator tube (internal diameter 0.25 mm) was heated to 60 ◦ C to avoid ice formation during the CO2 depressurization. The ELSD model
E. Lesellier et al. / J. Chromatogr. A 1327 (2014) 141–148 mV
143
3. Results and discussion
400
OLL
3.1. Optimization of operating conditions
350 OOLn OOL
LLL
300
OLLn
250
POL
200
SLL
POO SOL PSL
50
OOO
POLn
LnLnLn
100
LnLnL PLnLn LLLn
150
PLLn
OLnLn
PLL
SOO
0 0
5
10
15
20
25
30
35
40
Time (minutes)
Fig. 1. Chromatograms of argan (full red line) and rapeseed (blue dashed line) oils by SFC. Three Kinetex C18 in series (3 × 150 mm × 4.6 mm i.d., 2.6 m). UV detection 210 nm. T = 15 ◦ C; P out = 10 MPa. Mobile phase: CO2 –AcN 90:10 (v/v). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of the article.)
Sedex 85 (SEDERE, Alfortville, France) was connected to the outlet capillary. The fittings of the nebulizer tubing were slightly different with regard to the ones in HPLC, despite the use of the standard capillary diameter equal to 150 m [28]. The conditions for ELSD were 3 bar for the nebulizer gas (N2 ) and the drift temperature set at 40 ◦ C. UV and ELSD detectors were used in series to provide gains in sensitivity or uniformity of response based on the physical properties of the analyte. For UV, the response is related to the molar absorption coefficient (ε), which favours the response of TG having double bonds. For ELSD, the response is related to the droplet size after the nebulizer, which should be uniform since the conditions of the experimental conditions (mobile phase composition and flow rate) are constant. Consequently, the UV provides a higher sensitivity and identity for the unsaturated TG whereas the ELSD is better suited for a uniform response which regards to the amount of each TG present. The oils studied were argan and rapeseed, provided by Olvea (St Leonard, France). The oils were dissolved in a methanol–methylene chloride 50:50 (v/v) mixture. The injection volume was 1 l. Identification of peaks was done by: (i) the retention order which depends on the partition, number, i.e. increasing partition number causes increased retention; (ii) the published composition of these common vegetable oil and the relative amount of major and minor peaks; (iii) the comparison of the retention time of the peaks of the two selected oils and the knowledge of their identical (LLL, PLL, OLL, POL, OOL, POO, OOO, SOL, SOO) and specific (LnLnLn, PlnLn, LLLn, OLnLn, OLLn, POLn, OOLn, PPL, PPO, PSL) TG. Moreover, for each analytical condition studied, the chromatograms of the two oils were overlaid in the goal to ensure the reproducibility of the experiments; i.e. the reproducibility of the retention time for the identified TG (see in Fig. 1). The column dimensions were all 150 mm × 4.6 mm: Kinetex C18 (2.6 m) (Phenomenex, Le Pecq, France); Kinetex XB C18 (2.6 m) (Phenomenex, Le Pecq, France); Aeris peptide XB (3.6 m) (Phenomenex, Le Pecq, France); Aeris Wide Pore (3.6 m) (Phenomenex, Le Pecq, France); Accucore C18 (2.6 mm) (Thermoelectron, Les Ulis, France); Poroshell C18 (2.7 m) (Agilent, Les Ulis, France); Nucleoshell C18 (2.7 m) (Nacherey-Nagel, Hoert, France); Halo C18 (2.7 m) (AMT, Wilmington, USA); Halo Peptide ES (2.7 m) (AMT, Wilmington, USA); Ascentis Express C18 (2.7 m) (Supelco, St Quentin Fallavier, France). Chromatograms were recorded using the Azur software (Datalys, St Martin d’Hyeres, France).
The method development was performed in two parts. Firstly, three identical Kinetex C18 columns (45 cm) were used, with the goal to provide sufficient separation performance in the separation of critical pairs of compounds, thanks to the high number of theoretical plates (90,000 plates), without producing too long analysis duration (30–40 min). The studied parameters were the nature and the percentage of modifier, the column temperature and the drift evaporative tube temperature (ELSD detection). Analyses were performed with a constant flow rate of 2 ml/min, and a back-pressure of 10 MPa, by analysing two reference oils, argan and rapeseed. These two oils were chosen because of their well-known composition and their significant differences (Fig. 1). The argan oil is composed mainly by linoleic (C18:1)(L), oleic (C18:2)(O), stearic (C18:0)(S) and palmitic (C16:0)(P) chains. It allows the study of the change in selectivity for varied pairs of compounds differing either by one double-bond (LLL/OLL) or by one chain (LLL/PLL; PLL/OLL). The rapeseed oil contains the same fatty acids but also the linolenic one (C18:3)(Ln), which leads to numerous additional TG. Except the TG Po (C16:1) and some TG with long chains (C22, C24), most of the common fatty acids are present in these two oils. Consequently, the analytical conditions suited for the separation of TG in these two oils should be efficient for the separation of the TG in all other vegetable oils. Obviously, for other vegetable oil, such as peanut oil, containing higher alkyl chain lengths (C20, C22, C24), co-elution of some TG are reported, for instance between SOO (C18:0/C18:1/C18:1) and PGO (C16:0/C20:1/C18:1) [28]. These two TG display the same partition number (50), the same carbon number (54) and the same double bond number (2), they only differ by the relative chain length. In these cases, mass spectrometry could provide additional identification based on fragmentation patterns. Some couples present in rapeseed oil only differ by the location of one double-bond onto the chains: LLLn/OLnLn; LLL/OLLn; OLL/OLnLn. These pairs of TG are called critical pairs, because the structural difference between these compounds (the location of a double bond in the alkyl chain) is very slight, and these compounds are often difficult to separate. Previous studies have shown that the most suited modifier for separation of these couples in SFC was acetonitrile (AcN) [23–25]. Fig. 1 displays the chromatogram of the argan and rapeseed oil with 10% acetonitrile in carbon dioxide at 15 ◦ C. For argan oil (red chromatogram), these conditions would be sufficient to ensure the separation of the TG, because no critical pairs are present in this oil. Peaks were identified based on previous works in SFC with C18-bonded stationary phases [25]. For rapeseed, there is no full baseline resolution between the peaks of critical pairs, but for each pair, the two compounds can be clearly seen. The increase from 10 to 30% acetonitrile (Fig. 2) leads to a strong increase in retention. This behaviour was previously reported for hydrophobic compounds in SFC with C18-bonded stationary phases. It is related to the decrease in the solubility of TG when the proportion of polar modifier increases [23–25]. The acetonitrile increase also favours the selectivity between the critical pairs (see arrows), but degrades the separation between other compounds, because of the decrease in the total peak number. However, the TG having two identical chains and differing by one P or one O chain (PLL/LLL; POLn/OLL) then have identical retention. Two temperatures (15 ◦ C and 25 ◦ C) were studied with 10% acetonitrile as modifier. The temperature effects on selectivity were previously studied on fully porous particles between 5 and 35 ◦ C [23–25]. Temperature decrease induces increased selectivity for most peak pairs, mainly for the critical pairs, but also causes increased
144
E. Lesellier et al. / J. Chromatogr. A 1327 (2014) 141–148
mV 350
mV
500
300 400
250 300
200 150
200
100 100
50 0
0 0
10
20
30
40
50
60
-50
Time (minutes)
Fig. 2. UV chromatograms of rapeseed oil with 10% AcN (black full line), and with 30% AcN (red dashed line). Arrows show PLL/LLL and POLn/OLL). Other conditions as in Fig. 1. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of the article.)
retention time by reducing the TG solubility into the mobile phase. The results obtained with superficially porous particles are in accordance with the previous studies. To balance the selectivity and the analysis duration increase, optimal temperatures are set between 15 and 20 ◦ C. We also observed that the ELSD response was dramatically reduced by the temperature decrease with acetonitrile as modifier (Fig. 3). This low response in ELSD-SFC with this detector using acetonitrile as modifier was previously observed with another compound (caffeine) [29]. This behaviour is probably due to limited interactions between the acetonitrile molecules during the nebulization of the mobile phase, which does not provide droplets of sufficient size to ensure light scattering. Fortunately, the addition of methanol (10%) to acetonitrile provides a dramatically improved ELSD response, due to the strong hydrogen bonds allowing for a better detection of numerous compounds (Fig. 3). This addition induces other changes: the decrease of the retention time, and the decrease of the separation between the critical pairs (OLL/OLLn). The increase of the methanol percentage in acetonitrile, from 10 to 20%, does not further improve the response, and leads to slightly reduced selectivity between the critical pairs, but also for others, such as POO and PSL. The partial conclusion of these preliminary studies is the need for a small proportion of methanol in the modifier to improve the ELSD response (10% methanol in the acetonitrile modifier), but this addition is detrimental to the separation of critical pairs. This reduction can be compensated by the increase of the acetonitrile percentage, but the final amount of modifier should not be excessive to avoid co-elution of TG having P or O chains. Fig. 4 shows the changes in selectivity performances due to (i) the addition of mV 700 600 500 OLL/OOLn
400 300 200 100 0 15 -100
17
19
21
23
25
27
29
31
0
70
33
35
Time (minutes)
Fig. 3. ELSD chromatograms of rapeseed oil with 10% AcN (black dashed line) and with 10% of an AcN–MeOH 90:10 mixture (red full line). Other conditions as in Fig. 1. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of the article.)
-100
5
10
15
20
25
30
35
40
Time (minutes)
Fig. 4. UV chromatograms of rapeseed oil with 10% AcN (black full line), 10% of an AcN–MeOH 90:10 mixture (red dashed line), and 12% of an AcN–MeOH 90:10 mixture (dotted blue line). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of the article.)
10% methanol in acetonitrile (total amount of modifier 10%), which reduces both the analysis duration and slightly the critical pair selectivity, and (ii) the increase of the total modifier percentage, from 10 to 12% (acetonitrile–methanol 90:10), which again reduces the analysis duration but does not reduce the selectivity for critical pairs. The latter conditions were then selected for the second part of the method development. For temperature and backpressure, the final conditions selected were 17 ◦ C and 10 MPa. The selected temperature of 17 ◦ C, which allows satisfactory selectivities, depends on the modifier percentage and nature of the modifier, and of the stationary phases (SP) used. In previous studies [23–25], the selected SP was Hypersil ODS and the total modifier percentage was 6% with a temperature equal to 16 ◦ C. As expressed previously in this paper, due to the analytical conditions, 12% of modifier and 17 ◦ C, the density of the fluid is high, and obviously its state is not supercritical. At this density, above 1, the pressure changes do not significantly modify the eluent strength of the mobile phase, i.e. the analysis duration. The value of 10 MPa for back pressure was selected to limit the inlet pressure below 40 MPa when working with a long column length. 3.2. Optimization of the set of columns with Derringer function This second part was related to the increase of the column length by using four Kinetex C18 columns (total length 60 cm). This increase to 60 cm by adding one more Kinetex C18 does not induce a significant change in selectivity. In regard to the inlet pressure achieved with 60 cm of column length, one more column (to reach 75 cm total column length) can be added to provide a higher theoretical plate number. This additional column was selected from the 10 columns listed in the experimental part. In this part, the optimized operating conditions were selected: outlet pressure is equal to 10 MPa, temperature is set at 17 ◦ C, and the mobile phase contains 12% of a 90:10 acetonitrile–methanol mixture. The flow rate was set at 1.6 ml/min to limit the inlet pressure at 360 bars (below the maximum pressure allowed by the pumping system). To select the additional column from the 10 possibilities, we used a Derringer function, which allows comparison of varied parameters (Tables 2 and 3). The study was carried out with the two vegetable oils, and the final conclusion for the choice of the additional column was the same. Only results for rapeseed, containing the greater number of TG and the critical pairs of peaks are presented in Tables 2 and 3. The selected optimization parameters were: (i) the number of theoretical plates measured on several major peaks (OLnLn, OLLn, OOLn, POL, OOL, OOO), from which an average value was calculated,
E. Lesellier et al. / J. Chromatogr. A 1327 (2014) 141–148
145
Table 2 Values of the studied parameters for 10 different columns coupled to four Kinetex C18 (rapeseed oil). In bold the lowest and the highest values. Rapeseed oil KinC18
AerPep
AerWid
Theoretical plates number: TPN 114,566 118,790 118,174 OLnLn 101,344 110,542 111,998 OLLn 90,774 102,873 100,025 OOLn POL 108,830 115,767 107,481 OOL 78,091 93,466 94,841 72,088 94,673 91,431 OOO 94,282 106,019 103,992 Average Separation index: SI (calculation described in Fig. 5) 0.008 0.069 0.075 OLnLn/LLLn OLLn/LLL 0.081 0.077 0.138 OOLn/OLL 0.062 0.053 0.081 Analyse time: tR OOO 47.13 54.85 40.70 tR (OOO) Selectivity: ˛ POO/peak 1.007 1.000 1.023 ˛ Numbers of peaks at chromatograms 24 22 23 N◦
HalC18
HalPep
AscExp
PorShe
Nucloe
AccC18
KinXB
113,640 107,227 95,957 109,837 92,741 90,153 101,592
114,146 105,839 97,707 100,957 84,427 78,602 96,946
101,233 94,085 84,444 92,212 82,986 78,329 88,882
93,581 86,540 79,453 79,726 73,384 72,539 80,871
116,740 108,274 97,642 107,613 86,913 83,619 100,133
128,024 119,390 108,607 118,121 100,009 97,667 111,970
131,008 121,280 112,718 118,556 97,795 90,573 111,988
0.000 0.025 0.028
0.041 0.071 0.049
0.065 0.098 0.066
0.045 0.110 0.069
0.047 0.108 0.069
0.029 0.056 0.044
0.024 0.070 0.053
53.96
50.60
54.61
60.61
50.92
46.23
50.27
1.018
1.025
1.018
1.013
1.017
1.026
1.024
24
22
23
23
24
25
23
and OOL with a response equal to 400 mV, and the limit was set at 4 mV above the base line. Due to the higher sensitivity of UV detection, these measurements were achieved based on chromatograms recorded with UV detection. The separation index (SI) was chosen instead of the separation factor. Obviously, the separation factor (˛) was identical for the three critical pairs, as their structural differences are the same, but despite this identical selectivity value, the baseline return could be different for these three couples due to co-elution of minor TG, which could occur with some studied columns (it is not the case of the five coupled Kinetex C18 presented in Fig. 5). In the same way, resolution was not selected because of the lack of accuracy of the values for couples of peak composed by a minor and a major one, mainly for partial separations. Table 2 shows that: Fig. 5. UV chromatogram of rapeseed oil with five 15-cm Kinetex columns (2.6 m). Analytical conditions are described in the text. The couples of peaks used for the calculation of the SI parameter are indicated, as well as the six peaks used for the Theoretical plate number (TBN) calculation, the OOO peak used for the analysis duration, and the couple for the separation factor (˛ POO). The way of calculating the separation index (SI) is also described.
(ii) the separation index (SI) measured as described in Fig. 5, for three selected critical pairs (OLnLn/LLLn; OLLn/LLL and OOLn/OLL), (iii) the retention time for OOO, the last significant peak, (iv) the separation factor for POO and an unidentified peak, and finally (v) the total number of peaks on the chromatogram. This last measurement took into account the peaks with intensity equal at least to 1% of the major peak. For instance in Fig. 5, the major peaks are OOLn
(i) The average theoretical plate number (TPN) ranges from 80,871 (with Poroshell 120 C18) to 111,988 (with Kinetex XB C18). (ii) The best separation indices are obtained for Halo C18 (lowest SI values), and the worst for Aeris Wide pore. (iii) This last column also displays the lowest retention time for OOO (40.7 min), as these two parameters (SI and tR OOO) are somewhat correlated, due to the low specific surface area for ˚ On the contrary, for tR OOO, the this large pore silica (300 A). Poroshell 120 C18 displays the largest value, due to a strong bonded-phase hydrophobicity. The general retention for OOO of these 10 C18 fused-core phases is rather well described by the hydrophobicity obtained from the carotenoid test [30].
Table 3 Values of the Derringer function for 10 different columns coupled to four Kinetex C18 (rapeseed oil). Column
Rapeseed oil KinC18 AerPep AerWid HalC18 HalPep AscExp PorShe Nucloe ACCC18 KinXB
TPN average
0.431 0.808 0.743 0.666 0.517 0.257 0.000 0.619 0.999 1.000
Separation index OLnLn/LLLn
OLLn/LLL
OOLn/OLL
0.893 0.080 0.000 1.000 0.453 0.133 0.400 0.373 0.613 0.680
0.413 0.442 0.000 1.000 0.486 0.290 0.203 0.217 0.594 0.493
0.235 0.346 0.000 1.000 0.395 0.185 0.148 0.148 0.457 0.346
tR OOO
˛
N◦ of peak
Results
0.677 0.289 1.000 0.334 0.503 0.301 0.000 0.487 0.925 0.519
0.269 0.000 0.900 0.676 0.953 0.693 0.507 0.671 1.000 0.939
0.667 0.000 0.333 0.667 0.000 0.333 0.333 0.667 1.000 0.333
0.005 0.000 0.000 0.100 0.000 0.001 0.000 0.002 0.154 0.019
146
E. Lesellier et al. / J. Chromatogr. A 1327 (2014) 141–148
Derringer funcon values Accucore C18
1,100
1,0
Halo Pepde ES C18 Halo C18
0,900
Kinetex XB C18
0,700
Kinetex C18 Aeris Pepde C18
0,500
Aeris Wide pore C18 Poroshell C18 Nucleoshell C18
0,300
Ascens Express C18
0,100
0,0 -0,10080000
85000
90000
95000
100000
105000
110000
115000
TBN Average Fig. 6. Plot of the Derringer function values from the values of average theoretical plate number (TBN) for the 10 different columns studied.
(iv) The highest value for the separation factor between POO and the unidentified peak is obtained for the Accucore C18 phase. (v) The largest total number of peaks is also obtained on Accucore C18. Table 3 shows the values of the Derringer function calculated from the raw data. Each function was defined as a linear function varying from the minimum to maximum value observed among the 10 columns from 0 to 1. For example, for the average theoretical plate number (column 1 Table 3), the Derringer function value is equal to 1 for Kinetex XB C18 and 0 for Poroshell 120 C18. The values for the other columns are proportional, for instance, for Halo Peptide, the average plate number is 96,946, almost the medium value between the two extreme values (80,871 and 111,988), and the Derringer value is 0.517. Fig. 6 shows the correspondence between the TPN average values and the Derringer function values. Obviously, the larger is the TPN average value, the larger is the Derringer function value. Another example is that of the peak number: the values observed are 22, 23, 24 and 25 peaks, leading to a Derringer function varying from 0 for 22 peaks, 0.333 for 23, 0.667 for 24 to 1 for 25 peaks. The final result, last column in Table 3, is obtained by multiplying the seven Derringer values for all 10 columns. Of course, if one of these seven values is equal to 0, the final result is equal to 0, as is observed for four columns (Aeris Peptide, Aeris Wide pore, Halo Peptide and Poroshell). The first three have a low specific surface area due to large pore size, as discussed previously. For Poroshell 120 C18, the efficiency measured with TGs is poor and the retention great. It probably shows that the bonding density of this phase is high, that reduces the diffusion into the stationary phase for the TG, which are rather bigger than most classical chemical compounds, explaining the lower efficiency of this phase. Finally, the best value (0.154) is obtained for Accucore C18, and the second best is for Halo C18 (0.100). Fig. 7 shows one chromatogram obtained with these three chromatographic systems: (A) 5 Kinetex C18, (B) 4 Kinetex + 1 Halo C18 and (C) 4 Kinetex + 1 Accucore C18. As expected from Table 2, the analysis duration increases from system C, to A and B; the separation indices (SI) are higher for systems B and C, but the separation factor between POO and the unidentified peak is low with systems A and B. Consequently, Accucore C18 was retained as the fifth column coupled to four Kinetex C18. The separation developed with superficially porous particles allows to couple five columns (75 cm total length), ensuring both a lower analysis duration and an improved separation quality in SFC, in regards of the ones reported in NARP-LC [3–5,8]. The separation obtained with the SPP looks very close to the one previously
Fig. 7. UV chromatograms for three chromatographic systems with five columns (four 15 cm Kinetex C18 with: one 15-cm Accucore C18 in blue; one 15-cm Halo C18 in red; one 15-cm Kinetex C18 in green). Mobile phase: CO2 –AcN–MeOH (88:10.8:1.2, v/v/v), outlet pressure 10 MPa, temperature 17 ◦ C, flow rate 1.6 ml/min, detection wavelength 210 nm. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of the article.)
obtained with FPP [25]. However, with the FPP the column length was equal to 120 cm, the retention time for OOO was around 55 min, at a flow rate of 3 ml/min, whereas the retention of OOO with the SPP is equal to 42 min, with a flow rate equal to 1.6 ml/min, that means a reduction in the solvent consummation by a factor 2.5. Moreover, the quality of separation is strongly improved, with baseline resolution observed for most successive peaks. The increase of the theoretical plate number with 75 cm of SPP with regards to 120 cm of FPP also leads to get thinner peaks, that both allows the improvement of sensitivity and the increase of the peak capacity. These effects together provide the separation and detection of numerous additional peaks. Fig. 8 shows the responses obtained by UV and ELSD detector, and the attempts of identification for most of the peaks. As expected, the response is higher with ELSD, mainly for the major peaks OLLn, OLL, OOLn, OOL POO and OOO. For LnLnLn, LLnLn, PLL and some other peaks at the beginning of the chromatogram (not identified), the UV response is higher, due to the great number of double bonds of the TG. A new attempt of identification was done for these additional peaks, marked with an asterisk (Table 4, Figs. 8 and 9). This identification was done first from previous SFC studies for the main peaks [25], then from studies performed by HPLC/MS by Holcapek
Fig. 8. UV (blue) and ELSD (red) chromatograms for rapeseed oil with the optimal selected chromatographic system (60 cm Kinetex C18 + 15 cm Accucore C18). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of the article.)
E. Lesellier et al. / J. Chromatogr. A 1327 (2014) 141–148 Table 4 Attempt of peak identification for rapeseed oil separated by the UHE/LP-SFC system with SPP particles. n◦ 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 a b
TG
CN
DB
PN
tR min
Log k
% in Ref. [3]
LnLnLn LLnLn PLnLn LLLn OLnLn
54 54
9 8
36 38
20.46 22.74
3.316 3.797
0.521 0.579
Contents lists available at ScienceDirect
Journal of Chromatography A journal homepage: www.elsevier.com/locate/chroma
Ultra high efficiency/low pressure supercritical fluid chromatography with superficially porous particles for triglyceride separation夽 E. Lesellier a,∗ , A. Latos a , A. Lopes de Oliveira b a
Institut de Chimie Organique et Analytique (ICOA), Université d’Orléans, CNRS UMR 7311, B.P. 6759, rue de Chartres, 45067 Orléans cedex 2, France Faculdade de Zootecnia e Engenharia de Alimentos, Universidade de São Paulo, FZEA/USP Caixa Postal 23, CEP: 13635-900, Pirassununga, São Paulo, France b
a r t i c l e
i n f o
Article history: Received 26 September 2013 Received in revised form 12 December 2013 Accepted 16 December 2013 Available online 25 December 2013 Keywords: Column coupling Derringer function Method development UHE/LP-SFC Superficially porous particle Triglycerides
a b s t r a c t This paper reports the development of the separation of vegetable oil triglycerides (TG) in supercritical chromatography (SFC), using superficially porous particles (SPPs). The SPP, having a small diameter (2–3 m), provide a higher theoretical plate number (N), which allows to improve separation of critical pairs of compounds. However, compared to fully porous particles of larger diameter (5 m), the pressure drop is also increased. Fortunately, supercritical fluids have a low viscosity, which allows coupling several columns to achieve high N values, while maintaining flow rate above 1 ml/min, ensuring a ultra high efficiency (UHE) at low pressure (LP) (below 40 MPa), with regards to the one reached with liquid and sub-two micron particles (around 100 MPa). The use of two detector systems (UV and ELSD) connected in series to the UHE–LP-SFC system provides complementary responses, due to their specific detection principles. Working in a first part with three coupled Kinetex C18 columns (45 cm total length), the effect of modifier nature and percentage were studied with two reference oils, argan and rapeseed, chosen for their different and well-known TG composition. The analytical method was developed from previous studies performed with fully porous particles (FPP). Optimized conditions with three Kinetex were as follows: 17 ◦ C, 12% of ACN/MeOH (90/10; v/v). With these conditions, and by using an increased length of Kinetex C18 column (60 cm), another additional column was selected from ten different commercial SPP C18 bonded phases, by applying a Derringer function on varied parameters: theoretical plate number (TPN), separation index (SI) for critical pairs of peaks (the peaks of compounds difficult to separate due to subtle structural differences), the analysis duration, and the total peak number. This function normalizes the values of any parameters, between 0 and 1, from the worst value to the better, allowing to take account of various parameters in the final choice. Finally, by using four Kinetex C18 plus one Accucore C18 (75 cm total column length), a high-performance separation of triglycerides was achieved, with reasonable analysis duration and isocratic conditions. These conditions can be applied to varied vegetable oils. Identification of the numerous separated peaks of rapeseed oil was achieved by using published data and chromatographic retention behaviour. © 2013 Elsevier B.V. All rights reserved.
1. Introduction Triglycerides (TG) are the main constituents of vegetable oils. They are composed by three fatty acid chains, having varied length, from 14 to 24 carbons, and double bonds, from 0 to 4 (Table 1). Due to their low volatility, they are generally not analyzed in gaseous phase chromatography (GC), but either by high-performance liquid
夽 Presented at the 39th International Symposium on High-Performance LiquidPhase Separations and Related Techniques, Amsterdam, Netherlands, 16–20 June 2013. ∗ Corresponding author. Tel.: +33 2 28 49 45 88; fax: +33 2 38 41 72 81. E-mail address: [email protected] (E. Lesellier). 0021-9673/$ – see front matter © 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.chroma.2013.12.046
chromatography (HPLC) or by supercritical fluid chromatography (SFC). Two HPLC methods can be used, reverse phase liquid chromatography (RPLC) [1–5], or silver-ion mode [6–8]. With the former, using non-polar octadecylsiloxane-bonded silica (ODS) stationary phases, the retention of TG increases with the partition number (PN). PN is calculated from the total carbon number of the fatty acid chain, minus twice the total double bond number. For instance, this partition number is equal to 36 for LnLnLn (Ln = C18:3), this value is obtained from this calculation: PN = 3 × 18 − 2 × 9 = 36; 18 being the carbon number of each Ln chain and 9 the total double bond number. Moreover, due to the low polarity of the studied compounds (log P > 5), the use of water is generally avoided in the mobile phase, leading to a non-aqueous mode, called NARP. Elution gradients are also often used [1,3–5] to
142
E. Lesellier et al. / J. Chromatogr. A 1327 (2014) 141–148
Table 1 List of name, abbreviation, carbon number (CN), double-bond number (DB) and partition number (PN) for main fatty acids of vegetable oils. Trivial name
Abbreviation
CN
DB
PN
Myristic Palmitic Palmitoleic Margaric Margaroleic Stearic Oleic Linoleic Linolenic Arachidic Gadoleic Behenic Lignoceric
M P Po Ma Mo S O L Ln A G B Li
14 16 16 17 17 18 18 18 18 20 20 22 24
0 0 1 0 1 0 1 2 3 0 1 0 0
14 16 14 17 15 18 16 14 12 20 18 22 24
reduce the retention time which could be equal to 300 min in isocratic mode with two coupled columns [2] against 100–140 min in gradient mode [3,5]. However, the separation of several critical pairs remains difficult by NARP-LC [5]. First, separation of compounds composed by three C18 chains, having the same total carbon and double-bond number (identical PN), but with a different repartition onto the three chains, for instance: LLLn and OLnLn, LLL and OLLn, OLL and OOLn or SLL and SOLn [9,10]. These couples can be separated by HPLC using two-coupled C18 column, and an elution gradient with water in the first part, thereby increasing the analysis duration from 100 to 140 min [3]. Another separation seems impossible whatever the system used: the separation between couples SLL/POL, SLLn/POLn and SOL/POO having the same partition number with one saturated chain, either S (C18) or P (C16) [3,11]. The same difficulties of separation are encountered for the couples P0 OO/OOL or P0 PO/POL [5]. Separations of regio-isomers (sn-2; sn-3) are achieved by NARPLC, when using polymeric bonded ODS phases at low temperature [12], whereas LC–MS/MS systems allow to quantify these isomers on the basis of ion abundance [9]. The use of low temperature in NARP-LC was also reported in other studies but the retention times were prohibitive [13,14]. For this kind of separation, temperatures of 25–40 ◦ C are preferred for silver-ion chromatography to favour good peak shapes [15]. Silver-ion chromatography displays a different retention order in regards of the one in NARP-LC, as larger double-bond number induces longer retention. It also allows the separation of regioisomers, having the same chains but with a different position onto the glycerol moiety (as OLL and LOL) [8]. Unfortunately, silverion chromatography suffers from poor reproducibility and peak shapes, requires long equilibration times and freshly prepared mobile phases [16]. However, reproducibility seems improved by the addition of 2-propanol to the hexane–acetonitrile mobile phase, favouring the miscibility between the two previous solvents [15]. Increase in theoretical plate number can improve overall separation performance. This can be achieved by using superficially porous particles (SPPs) increasing the total column length. Despite the lower viscosities of non-aqueous mobile phases in regards to aqueous ones, the column length, and/or the flow rate, is limited when using pumps allowing for classical inlet pressure conditions [17,18]. Recently the use of SPP, also called core-shell or fused-core particles, has allowed to couple columns in HPLC, for the analysis of a very complex oil sample [197]. Four columns were serially coupled, and the inlet pressure was around 900 bars at the end of the elution gradient. 137 compounds from menhaden (fish) oil were identified with mass spectrometry, with total analysis duration equal to 190 min.
Two-dimensional studies using both liquid modes were achieved [19,20]. In this case, silver-ion mode is performed first, then NARP-LC. However, despite the short column used in the second dimension, the flow rate in the first one should be in the 10–20 l/min range, to ensure the separation of all peaks in the second dimension. This low flow rate dramatically increases the analysis duration. SFC has also been used for the separation of TG, with ODS stationary phases [21–26]. Due to the low viscosity of the CO2 -based mobile phase, high flow rates and long columns together provide shorter retention time and greater separation efficiencies in regards to HPLC. By using C18-bonded phases, the general retention rules are identical to those observed in RPLC, i.e. they are based on the partition number. Besides, due to the nature of the mobile phase, selectivity between TG having P chains are somewhat modified due to the increase in the P chain solubility in SFC, leading to improved separation. For instance the retention order in NARP-LC is LLL/OLL/PLL, whereas in SFC it is LLL/PLL/OLL. The improved solubility of the P chains in regards to the O one provides a retention inversion of the TG differing only by one P or one O chain. Moreover, the selectivity between the three (LLL, PLL, OLL) is strongly improved in SFC due, in part to the greater peak capacity and to the larger retention difference between PLL and OLL. Recently, SPP were also used in SFC to achieve 120,000 theoretical plates, by coupling four 15 cm columns [27]. Due to their reduced particle size, and to their great particle size homogeneity, these SPP phases provide higher efficiencies, with a reduced inlet pressure, compared to the ones observed using sub-2 m fully porous particles. This property, in addition of the low viscosity of super(sub)critical fluids would allow the use of very long columns to improve the separation efficiency for TG, without the need for ultra-high pressure chromatographic systems. The use of such high column length leads to get ultra high efficiency/low pressure SFC (UHE/LP-SFC). This paper will describe the method development used in UHE/LP-SFC, based on a previous separation obtained by SFC using fully porous particles (FPP) (Hypersil ODS). The composition of the mobile phase, the flow rate, temperature, backpressure and the choice of the stationary phases will be studied to ensure the best TG separation. Finally, the coupling of five columns filled with superficially porous particles, allows improving the separation performance of TG of vegetable oils with regard to previous works performed in the last century [23].
2. Materials and methods Chromatographic separations were carried out using equipment manufactured by Jasco (Tokyo, Japan). One pump model 2080-CO2 Plus was used for carbon dioxide and a second one model 2080 Plus for the modifier. Control of the mobile phase composition was performed by the CO2 pump. When the two solvents (modifier and CO2 ) were mixed, the fluid was introduced into a dynamic mixing chamber PU 4046 (Pye Unicam, Cambridge, UK) connected to a pulsation damper (Sedere, Orleans, France). The injector valve was supplied with a 20 L loop (model 7125 Rheodyne, Cotati, CA). The columns were thermostated by an oven (Jetstream 2 Plus, Hewlett-Packard, Palo Alto, CA). The detector was a Gilson UV 151 detector (provided by Waters, Milford, MA) equipped with a pressure-resistant cell. The detection wavelength was 210 nm. After the detector, the outlet column pressure was controlled by a Jasco BP-2080 Plus back pressure regulator (BPR). The outlet regulator tube (internal diameter 0.25 mm) was heated to 60 ◦ C to avoid ice formation during the CO2 depressurization. The ELSD model
E. Lesellier et al. / J. Chromatogr. A 1327 (2014) 141–148 mV
143
3. Results and discussion
400
OLL
3.1. Optimization of operating conditions
350 OOLn OOL
LLL
300
OLLn
250
POL
200
SLL
POO SOL PSL
50
OOO
POLn
LnLnLn
100
LnLnL PLnLn LLLn
150
PLLn
OLnLn
PLL
SOO
0 0
5
10
15
20
25
30
35
40
Time (minutes)
Fig. 1. Chromatograms of argan (full red line) and rapeseed (blue dashed line) oils by SFC. Three Kinetex C18 in series (3 × 150 mm × 4.6 mm i.d., 2.6 m). UV detection 210 nm. T = 15 ◦ C; P out = 10 MPa. Mobile phase: CO2 –AcN 90:10 (v/v). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of the article.)
Sedex 85 (SEDERE, Alfortville, France) was connected to the outlet capillary. The fittings of the nebulizer tubing were slightly different with regard to the ones in HPLC, despite the use of the standard capillary diameter equal to 150 m [28]. The conditions for ELSD were 3 bar for the nebulizer gas (N2 ) and the drift temperature set at 40 ◦ C. UV and ELSD detectors were used in series to provide gains in sensitivity or uniformity of response based on the physical properties of the analyte. For UV, the response is related to the molar absorption coefficient (ε), which favours the response of TG having double bonds. For ELSD, the response is related to the droplet size after the nebulizer, which should be uniform since the conditions of the experimental conditions (mobile phase composition and flow rate) are constant. Consequently, the UV provides a higher sensitivity and identity for the unsaturated TG whereas the ELSD is better suited for a uniform response which regards to the amount of each TG present. The oils studied were argan and rapeseed, provided by Olvea (St Leonard, France). The oils were dissolved in a methanol–methylene chloride 50:50 (v/v) mixture. The injection volume was 1 l. Identification of peaks was done by: (i) the retention order which depends on the partition, number, i.e. increasing partition number causes increased retention; (ii) the published composition of these common vegetable oil and the relative amount of major and minor peaks; (iii) the comparison of the retention time of the peaks of the two selected oils and the knowledge of their identical (LLL, PLL, OLL, POL, OOL, POO, OOO, SOL, SOO) and specific (LnLnLn, PlnLn, LLLn, OLnLn, OLLn, POLn, OOLn, PPL, PPO, PSL) TG. Moreover, for each analytical condition studied, the chromatograms of the two oils were overlaid in the goal to ensure the reproducibility of the experiments; i.e. the reproducibility of the retention time for the identified TG (see in Fig. 1). The column dimensions were all 150 mm × 4.6 mm: Kinetex C18 (2.6 m) (Phenomenex, Le Pecq, France); Kinetex XB C18 (2.6 m) (Phenomenex, Le Pecq, France); Aeris peptide XB (3.6 m) (Phenomenex, Le Pecq, France); Aeris Wide Pore (3.6 m) (Phenomenex, Le Pecq, France); Accucore C18 (2.6 mm) (Thermoelectron, Les Ulis, France); Poroshell C18 (2.7 m) (Agilent, Les Ulis, France); Nucleoshell C18 (2.7 m) (Nacherey-Nagel, Hoert, France); Halo C18 (2.7 m) (AMT, Wilmington, USA); Halo Peptide ES (2.7 m) (AMT, Wilmington, USA); Ascentis Express C18 (2.7 m) (Supelco, St Quentin Fallavier, France). Chromatograms were recorded using the Azur software (Datalys, St Martin d’Hyeres, France).
The method development was performed in two parts. Firstly, three identical Kinetex C18 columns (45 cm) were used, with the goal to provide sufficient separation performance in the separation of critical pairs of compounds, thanks to the high number of theoretical plates (90,000 plates), without producing too long analysis duration (30–40 min). The studied parameters were the nature and the percentage of modifier, the column temperature and the drift evaporative tube temperature (ELSD detection). Analyses were performed with a constant flow rate of 2 ml/min, and a back-pressure of 10 MPa, by analysing two reference oils, argan and rapeseed. These two oils were chosen because of their well-known composition and their significant differences (Fig. 1). The argan oil is composed mainly by linoleic (C18:1)(L), oleic (C18:2)(O), stearic (C18:0)(S) and palmitic (C16:0)(P) chains. It allows the study of the change in selectivity for varied pairs of compounds differing either by one double-bond (LLL/OLL) or by one chain (LLL/PLL; PLL/OLL). The rapeseed oil contains the same fatty acids but also the linolenic one (C18:3)(Ln), which leads to numerous additional TG. Except the TG Po (C16:1) and some TG with long chains (C22, C24), most of the common fatty acids are present in these two oils. Consequently, the analytical conditions suited for the separation of TG in these two oils should be efficient for the separation of the TG in all other vegetable oils. Obviously, for other vegetable oil, such as peanut oil, containing higher alkyl chain lengths (C20, C22, C24), co-elution of some TG are reported, for instance between SOO (C18:0/C18:1/C18:1) and PGO (C16:0/C20:1/C18:1) [28]. These two TG display the same partition number (50), the same carbon number (54) and the same double bond number (2), they only differ by the relative chain length. In these cases, mass spectrometry could provide additional identification based on fragmentation patterns. Some couples present in rapeseed oil only differ by the location of one double-bond onto the chains: LLLn/OLnLn; LLL/OLLn; OLL/OLnLn. These pairs of TG are called critical pairs, because the structural difference between these compounds (the location of a double bond in the alkyl chain) is very slight, and these compounds are often difficult to separate. Previous studies have shown that the most suited modifier for separation of these couples in SFC was acetonitrile (AcN) [23–25]. Fig. 1 displays the chromatogram of the argan and rapeseed oil with 10% acetonitrile in carbon dioxide at 15 ◦ C. For argan oil (red chromatogram), these conditions would be sufficient to ensure the separation of the TG, because no critical pairs are present in this oil. Peaks were identified based on previous works in SFC with C18-bonded stationary phases [25]. For rapeseed, there is no full baseline resolution between the peaks of critical pairs, but for each pair, the two compounds can be clearly seen. The increase from 10 to 30% acetonitrile (Fig. 2) leads to a strong increase in retention. This behaviour was previously reported for hydrophobic compounds in SFC with C18-bonded stationary phases. It is related to the decrease in the solubility of TG when the proportion of polar modifier increases [23–25]. The acetonitrile increase also favours the selectivity between the critical pairs (see arrows), but degrades the separation between other compounds, because of the decrease in the total peak number. However, the TG having two identical chains and differing by one P or one O chain (PLL/LLL; POLn/OLL) then have identical retention. Two temperatures (15 ◦ C and 25 ◦ C) were studied with 10% acetonitrile as modifier. The temperature effects on selectivity were previously studied on fully porous particles between 5 and 35 ◦ C [23–25]. Temperature decrease induces increased selectivity for most peak pairs, mainly for the critical pairs, but also causes increased
144
E. Lesellier et al. / J. Chromatogr. A 1327 (2014) 141–148
mV 350
mV
500
300 400
250 300
200 150
200
100 100
50 0
0 0
10
20
30
40
50
60
-50
Time (minutes)
Fig. 2. UV chromatograms of rapeseed oil with 10% AcN (black full line), and with 30% AcN (red dashed line). Arrows show PLL/LLL and POLn/OLL). Other conditions as in Fig. 1. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of the article.)
retention time by reducing the TG solubility into the mobile phase. The results obtained with superficially porous particles are in accordance with the previous studies. To balance the selectivity and the analysis duration increase, optimal temperatures are set between 15 and 20 ◦ C. We also observed that the ELSD response was dramatically reduced by the temperature decrease with acetonitrile as modifier (Fig. 3). This low response in ELSD-SFC with this detector using acetonitrile as modifier was previously observed with another compound (caffeine) [29]. This behaviour is probably due to limited interactions between the acetonitrile molecules during the nebulization of the mobile phase, which does not provide droplets of sufficient size to ensure light scattering. Fortunately, the addition of methanol (10%) to acetonitrile provides a dramatically improved ELSD response, due to the strong hydrogen bonds allowing for a better detection of numerous compounds (Fig. 3). This addition induces other changes: the decrease of the retention time, and the decrease of the separation between the critical pairs (OLL/OLLn). The increase of the methanol percentage in acetonitrile, from 10 to 20%, does not further improve the response, and leads to slightly reduced selectivity between the critical pairs, but also for others, such as POO and PSL. The partial conclusion of these preliminary studies is the need for a small proportion of methanol in the modifier to improve the ELSD response (10% methanol in the acetonitrile modifier), but this addition is detrimental to the separation of critical pairs. This reduction can be compensated by the increase of the acetonitrile percentage, but the final amount of modifier should not be excessive to avoid co-elution of TG having P or O chains. Fig. 4 shows the changes in selectivity performances due to (i) the addition of mV 700 600 500 OLL/OOLn
400 300 200 100 0 15 -100
17
19
21
23
25
27
29
31
0
70
33
35
Time (minutes)
Fig. 3. ELSD chromatograms of rapeseed oil with 10% AcN (black dashed line) and with 10% of an AcN–MeOH 90:10 mixture (red full line). Other conditions as in Fig. 1. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of the article.)
-100
5
10
15
20
25
30
35
40
Time (minutes)
Fig. 4. UV chromatograms of rapeseed oil with 10% AcN (black full line), 10% of an AcN–MeOH 90:10 mixture (red dashed line), and 12% of an AcN–MeOH 90:10 mixture (dotted blue line). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of the article.)
10% methanol in acetonitrile (total amount of modifier 10%), which reduces both the analysis duration and slightly the critical pair selectivity, and (ii) the increase of the total modifier percentage, from 10 to 12% (acetonitrile–methanol 90:10), which again reduces the analysis duration but does not reduce the selectivity for critical pairs. The latter conditions were then selected for the second part of the method development. For temperature and backpressure, the final conditions selected were 17 ◦ C and 10 MPa. The selected temperature of 17 ◦ C, which allows satisfactory selectivities, depends on the modifier percentage and nature of the modifier, and of the stationary phases (SP) used. In previous studies [23–25], the selected SP was Hypersil ODS and the total modifier percentage was 6% with a temperature equal to 16 ◦ C. As expressed previously in this paper, due to the analytical conditions, 12% of modifier and 17 ◦ C, the density of the fluid is high, and obviously its state is not supercritical. At this density, above 1, the pressure changes do not significantly modify the eluent strength of the mobile phase, i.e. the analysis duration. The value of 10 MPa for back pressure was selected to limit the inlet pressure below 40 MPa when working with a long column length. 3.2. Optimization of the set of columns with Derringer function This second part was related to the increase of the column length by using four Kinetex C18 columns (total length 60 cm). This increase to 60 cm by adding one more Kinetex C18 does not induce a significant change in selectivity. In regard to the inlet pressure achieved with 60 cm of column length, one more column (to reach 75 cm total column length) can be added to provide a higher theoretical plate number. This additional column was selected from the 10 columns listed in the experimental part. In this part, the optimized operating conditions were selected: outlet pressure is equal to 10 MPa, temperature is set at 17 ◦ C, and the mobile phase contains 12% of a 90:10 acetonitrile–methanol mixture. The flow rate was set at 1.6 ml/min to limit the inlet pressure at 360 bars (below the maximum pressure allowed by the pumping system). To select the additional column from the 10 possibilities, we used a Derringer function, which allows comparison of varied parameters (Tables 2 and 3). The study was carried out with the two vegetable oils, and the final conclusion for the choice of the additional column was the same. Only results for rapeseed, containing the greater number of TG and the critical pairs of peaks are presented in Tables 2 and 3. The selected optimization parameters were: (i) the number of theoretical plates measured on several major peaks (OLnLn, OLLn, OOLn, POL, OOL, OOO), from which an average value was calculated,
E. Lesellier et al. / J. Chromatogr. A 1327 (2014) 141–148
145
Table 2 Values of the studied parameters for 10 different columns coupled to four Kinetex C18 (rapeseed oil). In bold the lowest and the highest values. Rapeseed oil KinC18
AerPep
AerWid
Theoretical plates number: TPN 114,566 118,790 118,174 OLnLn 101,344 110,542 111,998 OLLn 90,774 102,873 100,025 OOLn POL 108,830 115,767 107,481 OOL 78,091 93,466 94,841 72,088 94,673 91,431 OOO 94,282 106,019 103,992 Average Separation index: SI (calculation described in Fig. 5) 0.008 0.069 0.075 OLnLn/LLLn OLLn/LLL 0.081 0.077 0.138 OOLn/OLL 0.062 0.053 0.081 Analyse time: tR OOO 47.13 54.85 40.70 tR (OOO) Selectivity: ˛ POO/peak 1.007 1.000 1.023 ˛ Numbers of peaks at chromatograms 24 22 23 N◦
HalC18
HalPep
AscExp
PorShe
Nucloe
AccC18
KinXB
113,640 107,227 95,957 109,837 92,741 90,153 101,592
114,146 105,839 97,707 100,957 84,427 78,602 96,946
101,233 94,085 84,444 92,212 82,986 78,329 88,882
93,581 86,540 79,453 79,726 73,384 72,539 80,871
116,740 108,274 97,642 107,613 86,913 83,619 100,133
128,024 119,390 108,607 118,121 100,009 97,667 111,970
131,008 121,280 112,718 118,556 97,795 90,573 111,988
0.000 0.025 0.028
0.041 0.071 0.049
0.065 0.098 0.066
0.045 0.110 0.069
0.047 0.108 0.069
0.029 0.056 0.044
0.024 0.070 0.053
53.96
50.60
54.61
60.61
50.92
46.23
50.27
1.018
1.025
1.018
1.013
1.017
1.026
1.024
24
22
23
23
24
25
23
and OOL with a response equal to 400 mV, and the limit was set at 4 mV above the base line. Due to the higher sensitivity of UV detection, these measurements were achieved based on chromatograms recorded with UV detection. The separation index (SI) was chosen instead of the separation factor. Obviously, the separation factor (˛) was identical for the three critical pairs, as their structural differences are the same, but despite this identical selectivity value, the baseline return could be different for these three couples due to co-elution of minor TG, which could occur with some studied columns (it is not the case of the five coupled Kinetex C18 presented in Fig. 5). In the same way, resolution was not selected because of the lack of accuracy of the values for couples of peak composed by a minor and a major one, mainly for partial separations. Table 2 shows that: Fig. 5. UV chromatogram of rapeseed oil with five 15-cm Kinetex columns (2.6 m). Analytical conditions are described in the text. The couples of peaks used for the calculation of the SI parameter are indicated, as well as the six peaks used for the Theoretical plate number (TBN) calculation, the OOO peak used for the analysis duration, and the couple for the separation factor (˛ POO). The way of calculating the separation index (SI) is also described.
(ii) the separation index (SI) measured as described in Fig. 5, for three selected critical pairs (OLnLn/LLLn; OLLn/LLL and OOLn/OLL), (iii) the retention time for OOO, the last significant peak, (iv) the separation factor for POO and an unidentified peak, and finally (v) the total number of peaks on the chromatogram. This last measurement took into account the peaks with intensity equal at least to 1% of the major peak. For instance in Fig. 5, the major peaks are OOLn
(i) The average theoretical plate number (TPN) ranges from 80,871 (with Poroshell 120 C18) to 111,988 (with Kinetex XB C18). (ii) The best separation indices are obtained for Halo C18 (lowest SI values), and the worst for Aeris Wide pore. (iii) This last column also displays the lowest retention time for OOO (40.7 min), as these two parameters (SI and tR OOO) are somewhat correlated, due to the low specific surface area for ˚ On the contrary, for tR OOO, the this large pore silica (300 A). Poroshell 120 C18 displays the largest value, due to a strong bonded-phase hydrophobicity. The general retention for OOO of these 10 C18 fused-core phases is rather well described by the hydrophobicity obtained from the carotenoid test [30].
Table 3 Values of the Derringer function for 10 different columns coupled to four Kinetex C18 (rapeseed oil). Column
Rapeseed oil KinC18 AerPep AerWid HalC18 HalPep AscExp PorShe Nucloe ACCC18 KinXB
TPN average
0.431 0.808 0.743 0.666 0.517 0.257 0.000 0.619 0.999 1.000
Separation index OLnLn/LLLn
OLLn/LLL
OOLn/OLL
0.893 0.080 0.000 1.000 0.453 0.133 0.400 0.373 0.613 0.680
0.413 0.442 0.000 1.000 0.486 0.290 0.203 0.217 0.594 0.493
0.235 0.346 0.000 1.000 0.395 0.185 0.148 0.148 0.457 0.346
tR OOO
˛
N◦ of peak
Results
0.677 0.289 1.000 0.334 0.503 0.301 0.000 0.487 0.925 0.519
0.269 0.000 0.900 0.676 0.953 0.693 0.507 0.671 1.000 0.939
0.667 0.000 0.333 0.667 0.000 0.333 0.333 0.667 1.000 0.333
0.005 0.000 0.000 0.100 0.000 0.001 0.000 0.002 0.154 0.019
146
E. Lesellier et al. / J. Chromatogr. A 1327 (2014) 141–148
Derringer funcon values Accucore C18
1,100
1,0
Halo Pepde ES C18 Halo C18
0,900
Kinetex XB C18
0,700
Kinetex C18 Aeris Pepde C18
0,500
Aeris Wide pore C18 Poroshell C18 Nucleoshell C18
0,300
Ascens Express C18
0,100
0,0 -0,10080000
85000
90000
95000
100000
105000
110000
115000
TBN Average Fig. 6. Plot of the Derringer function values from the values of average theoretical plate number (TBN) for the 10 different columns studied.
(iv) The highest value for the separation factor between POO and the unidentified peak is obtained for the Accucore C18 phase. (v) The largest total number of peaks is also obtained on Accucore C18. Table 3 shows the values of the Derringer function calculated from the raw data. Each function was defined as a linear function varying from the minimum to maximum value observed among the 10 columns from 0 to 1. For example, for the average theoretical plate number (column 1 Table 3), the Derringer function value is equal to 1 for Kinetex XB C18 and 0 for Poroshell 120 C18. The values for the other columns are proportional, for instance, for Halo Peptide, the average plate number is 96,946, almost the medium value between the two extreme values (80,871 and 111,988), and the Derringer value is 0.517. Fig. 6 shows the correspondence between the TPN average values and the Derringer function values. Obviously, the larger is the TPN average value, the larger is the Derringer function value. Another example is that of the peak number: the values observed are 22, 23, 24 and 25 peaks, leading to a Derringer function varying from 0 for 22 peaks, 0.333 for 23, 0.667 for 24 to 1 for 25 peaks. The final result, last column in Table 3, is obtained by multiplying the seven Derringer values for all 10 columns. Of course, if one of these seven values is equal to 0, the final result is equal to 0, as is observed for four columns (Aeris Peptide, Aeris Wide pore, Halo Peptide and Poroshell). The first three have a low specific surface area due to large pore size, as discussed previously. For Poroshell 120 C18, the efficiency measured with TGs is poor and the retention great. It probably shows that the bonding density of this phase is high, that reduces the diffusion into the stationary phase for the TG, which are rather bigger than most classical chemical compounds, explaining the lower efficiency of this phase. Finally, the best value (0.154) is obtained for Accucore C18, and the second best is for Halo C18 (0.100). Fig. 7 shows one chromatogram obtained with these three chromatographic systems: (A) 5 Kinetex C18, (B) 4 Kinetex + 1 Halo C18 and (C) 4 Kinetex + 1 Accucore C18. As expected from Table 2, the analysis duration increases from system C, to A and B; the separation indices (SI) are higher for systems B and C, but the separation factor between POO and the unidentified peak is low with systems A and B. Consequently, Accucore C18 was retained as the fifth column coupled to four Kinetex C18. The separation developed with superficially porous particles allows to couple five columns (75 cm total length), ensuring both a lower analysis duration and an improved separation quality in SFC, in regards of the ones reported in NARP-LC [3–5,8]. The separation obtained with the SPP looks very close to the one previously
Fig. 7. UV chromatograms for three chromatographic systems with five columns (four 15 cm Kinetex C18 with: one 15-cm Accucore C18 in blue; one 15-cm Halo C18 in red; one 15-cm Kinetex C18 in green). Mobile phase: CO2 –AcN–MeOH (88:10.8:1.2, v/v/v), outlet pressure 10 MPa, temperature 17 ◦ C, flow rate 1.6 ml/min, detection wavelength 210 nm. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of the article.)
obtained with FPP [25]. However, with the FPP the column length was equal to 120 cm, the retention time for OOO was around 55 min, at a flow rate of 3 ml/min, whereas the retention of OOO with the SPP is equal to 42 min, with a flow rate equal to 1.6 ml/min, that means a reduction in the solvent consummation by a factor 2.5. Moreover, the quality of separation is strongly improved, with baseline resolution observed for most successive peaks. The increase of the theoretical plate number with 75 cm of SPP with regards to 120 cm of FPP also leads to get thinner peaks, that both allows the improvement of sensitivity and the increase of the peak capacity. These effects together provide the separation and detection of numerous additional peaks. Fig. 8 shows the responses obtained by UV and ELSD detector, and the attempts of identification for most of the peaks. As expected, the response is higher with ELSD, mainly for the major peaks OLLn, OLL, OOLn, OOL POO and OOO. For LnLnLn, LLnLn, PLL and some other peaks at the beginning of the chromatogram (not identified), the UV response is higher, due to the great number of double bonds of the TG. A new attempt of identification was done for these additional peaks, marked with an asterisk (Table 4, Figs. 8 and 9). This identification was done first from previous SFC studies for the main peaks [25], then from studies performed by HPLC/MS by Holcapek
Fig. 8. UV (blue) and ELSD (red) chromatograms for rapeseed oil with the optimal selected chromatographic system (60 cm Kinetex C18 + 15 cm Accucore C18). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of the article.)
E. Lesellier et al. / J. Chromatogr. A 1327 (2014) 141–148 Table 4 Attempt of peak identification for rapeseed oil separated by the UHE/LP-SFC system with SPP particles. n◦ 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 a b
TG
CN
DB
PN
tR min
Log k
% in Ref. [3]
LnLnLn LLnLn PLnLn LLLn OLnLn
54 54
9 8
36 38
20.46 22.74
3.316 3.797
0.521 0.579
