A multi-location peak parking approach for calculation of second dimensional retention indices for improved volatile compound identification with cryogen-free comprehensive heart-cut two-dimensional gas chromatography

Abstract

Comprehensive heart-cut multidimensional gas chromatography (CH/C MDGC) without a cryogenic trapping device was developed with an established approach for calculation of first and second dimensional retention indices (¹I and ²I) for improved compound identification. A first dimensional (¹D) DB-1MS column (60 m) and a second dimensional (²D) DB-WAX column (60 m) were applied with a Deans switch (DS) using a constant H/C window of 0.2 min and a periodic multiple heartcut strategy comprising 225H/C throughout the CH/C. ¹I was calculated based on comparison of the middle of the heartcut time with the alkane retention times on the ¹D column. A multi-location peak parking approach using sixteen sets of automated injections of alkane references was also established with the least square curve fitting method for construction of the alkane isovolatility curves which were applied for ²I calculation. The untargeted compound analysis of a perfume sample was then performed according to comparison with the libraries of mass spectra, ¹I and ²I. The CH/C MDGC system with a 25 h analysis time showed a peak capacity (n_c) of 9198 and 128 separated peaks with 71 compounds successfully identified according to MS, ¹I and ²I library match under the established error approximation criteria. Furthermore, relationship between the analysis time and number of separated peaks was proposed based on the set of 84 identifiable compounds. With the compensation of lower separation performance and greater I errors, the analysis time could be reduced by applying a 2.5 min H/C window with a total analysis time of 2 h and n_c of 1134.

---

1. Introduction

Comprehensive two dimensional gas chromatography (GC × GC) offers separation with high resolution and enhanced peak capacity improving the analysis of complex samples such as petrochemicals, essential oils, and other food or biological samples.^1–4 This technique solves the co-elution problem, enhances peak capacity and presents more reliable results than those obtained with conventional one dimensional GC.^5,6 GC × GC hyphenated with mass spectrometry (GC × GC-MS) also provides untargeted compound analysis with higher confidence.^7–9 This technique relies on an effective modulation process (using a cryogenic, valve based or flow modulator) which transfers all components obtained from a ¹D to a ²D column with different selectivity.^10,11 This can be generally performed by using a short ²D column (≤5 m) which enables fast ²D separation (<20 s) completing comprehensive analysis within a single injection.^12,13

Alternatively, multiple injection techniques can be applied to allow the use of a longer ²D column and longer analysis time improving separation and peak capacity. Such operation has been recognized as comprehensive heart-cut multidimensional GC (CH/C MDGC),¹⁴ and can be performed by using a heartcut device, Deans switch (DS) or switching valve, and a modulator.^2,15,16 The modulator can also be replaced with a cryogenic trapping device with a lower cost.^17,18 To this end, the DS heartcuts a selected pulse^19,20 from a ¹D column to be trapped at the cryogenic device located close to a ²D column inlet. This is followed by the pulse release within a narrow band of a sample undergoing ²D separation, which can be performed using an independent flow and temperature program for improved ²D separation. However, a cryogenic approach still involves additional devices (e.g. a trapping channel and a switching valve for automated on/off operation), cryogen consumption and system maintenance. With the compensation of lower separation efficiency, a system employing a DS without using a cryogenic trapping device (cryogen-free CH/C MDGC) has been reported.¹⁶ This approach applied a single DS with a H/C window of 0.2 min in order to transfer a narrow band of the sample from a ¹D onto a ²D column without trapping, which improved the separation performance. A challenge is to develop a data analysis approach for improved compound identification with a cryogen-free CH/C MDGC-MS system.

Traditional methods used to identify untargeted compounds with GC × GC-MS are matched with MS spectra and ¹I libraries.³ Calculation of ¹I for a peak of interest is performed by additional injection of n-alkanes followed by comparison of the peak ¹D retention time (¹t_R) with that of the alkanes according to van den Dool and Kratz relationship.²¹ Cryogen-free CH/C MDGC can also apply the same method for ¹I calculation albeit the middle of the H/C is used instead of ¹t_R.¹⁶ Moreover, ²I can be taken into account to improve confidence in compound identification. For conventional GC × GC, ²I calculation can be performed by multiple injections of n-alkanes under the same temperature programs used for sample analysis or the temperature program with steps of different isothermal temperatures.^3,9,22 This is performed to generate isovolatility curves (¹t_Rvs.²t_R plots) for alkanes on a ²D column, and the curves are used to calculate ²I for a peak. ²I calculation has also been reported with H/C MDGC using cryogenic trapping devices where all the trapped compounds simultaneously started eluting onto a ²D column under the same flow and temperature programs.³ This approach involves a single injection of alkanes onto the applied system. All the alkanes elute through a ¹D column and are trapped at the inlet of the ²D column. The oven is then cooled down prior to the alkane release onto a ²D column under an identical ²D flow and temperature program to that applied for sample separation. This allows direct ²I calculation using the same approach as that applied for ¹I calculation. However, this approach cannot be applicable for CH/C MDGC with different compounds eluting under different ²D conditions. In addition, an approach for ²I calculation with cryogen-free CH/C MDGC has not been reported.

In this study, a cryogen-free CH/C MDGC-MS system with a narrow H/C window of 0.2 min was applied, and relevant data analysis and experimental approaches for ¹I and ²I calculation were established. Isovolatility curves were constructed according to a peak parking approach allowing the alkane peaks to park at different axial positions along the ¹D column prior to elution onto the ²D column at different temperatures. This was adapted from the concept of multi-location peak parking (MLPP) which was reported in liquid chromatography and differently applied to investigate peak broadening after parking (stopped flow) for different periods prior to the peak elution and detection in order to determine the longitudinal diffusion coefficient of the analyte.²³ An example application was demonstrated for perfume analysis where compounds were identified according to comparison with MS spectra, ¹I and ²I from the libraries. The possible compounds observed in perfumes included aldehydes, alcohols, lactones, esters and terpene derivatives.^24,25

2. Experimental

2.1 Materials and chemicals

Perfume was purchased from a local supermarket, Thailand, and kept in a refrigerator at 4 °C prior to use. n-alkane standards (C6–C22) and all the standard compounds were purchased from Sigma-Aldrich (St. Louis, MO).

2.2 Instrumental

2.2.1 Comprehensive H/C MDGC. An Agilent 7890A gas chromatograph coupled with an Agilent 5975C QMS (Agilent technologies Inc., US) was applied in this study. The perfume sample was diluted in ethanol (10% v/v). This diluted sample or n-alkane mixture (100 mg L⁻¹ in hexane) was injected (1 μL) at 250 °C with a split ratio of 20 : 1. A ¹D nonpolar DB-1MS column (60 m × 0.25 mm I.D. × 0.25 μm; J&W Scientific, US) and a ²D polar DB-Wax column (60 m × 0.25 mm × 0.25 μm; J&W Scientific, US) were used for separation. The restrictor column (1.5 m × 0.1 mm; Agilent technologies Inc.) was applied to balance the flow with the ²D column. The outlets of the restrictor and the ²D columns were connected to a flame ionization detector (FID) and MS as reported,¹⁶ respectively. A Dean switch device (DS, Agilent technologies Inc.) is an interface connecting the ¹D, ²D and restrictor columns. In this instrument, the DS was operated in off and on modes corresponding to analyte pulse transfer to the FID and MS, respectively. The flow rates of the carrier gas (99.999% He) in ¹D and ²D separation were 2.0 and 4.0 mL min⁻¹, respectively. The GC oven temperature program was set at 60 °C, increased to 250 °C with a rate of 4 °C min⁻¹ and held at this temperature for 12.5 min. The FID temperature was set at 270 °C using flow rates of 40, 400 and 30 mL min⁻¹ for hydrogen, air and N₂ (makeup gas), respectively. For MS conditions, the ion source temperature was set at 250 °C, the electron ionization voltage was −70 eV and a mass range of 28–550 m/z was applied. The multiple H/C approach with a constant window of 0.2 min was applied. This required 25 injections in order to comprehensively sample all the peaks obtained from ¹D separation. The H/C events are provided in Table S-1 (ESI).†

2.2.2 Multi-location peak parking and construction of isovolatility curves. 16 sets of experiments were conducted each consisting of 4 sequences (two performed for identical parking and the others performed for the elution with the DS off and on). Details of the experimental approach are summarized in Table S-2 (ESI).†

2.3 Data analysis

A contour plot for the CH/C MDGC-MS analysis covering 10–60 min of ¹D separation was obtained using a combination of all the elution profiles (time vs. intensity profiles each of which belonged to each H/C event within every 5 min) into a single matrix (containing ¹t_Rvs.²t_Rvs. intensity coordinates). The ¹t_R coordinate is the H/C center time (t_H/Cmid) of the corresponding H/C event. ²t_R coordinates of each profile were obtained from the time coordinates observed with MS (t_observed) subtracted with t_H/Cmid of each H/C event. The matrix was then converted into a contour plot by using Fortner Transform 3.3 (Fortner, Inc., Savoy, IL). ChemStation software was used to process and identify separated peaks, and Excel 2016 was used for chromatographic parameter calculation and data visualisation. Peaks of interest were tentatively identified by comparison of their mass spectra with those from the NIST 14 library with MS match scores of >700, and further confirmed according to comparison of their experimental ¹I and ²I values with those from the NIST 14 within the differences of ±37 and ±44 units, respectively; see the discussion of the errors associated with the I calculation in Section 3.4. Note that when there was no literature ¹I available for a peak of interest (e.g. with a signal to noise ratio of >3), the peak was herein defined as a detected peak which was not identifiable.

2.3.1 Calculation of ¹t_R, ²t_R and ¹I. For ¹I calculation, the n-alkane standard mixture was injected under the same oven program applied for the perfume sample with the DS off. The van den Dool and Kratz equation was used to calculate ¹I for a peak with the relationship²⁶¹t_R(i) is the retention time of the peak on the ¹D column. n and n + 1 are the number of carbons of alkanes with the elution times bracketing ¹t_R(i). In this study, ¹t_R is calculated from t_H/Cmid ast_H/Cmid, t_H/Cstart and t_H/Cend are the H/C center time, and start time and end time of the H/C, respectively. ²t_R is retention time of the peak on the ²D column which can be calculated as

²t_R = t_observed − ¹t_R (3)

t_observed is the peak time observed with the MS detector (after elution through ¹D and ²D columns). In the case of modulated (or split) peaks into different H/C fractions, calculation of ¹t_R and ²t_R of a compound was performed using the weighted average of each sub-peak with the corresponding peak area.

2.3.2 Isovolatility curve and ²I calculation. A regression model used to calculate the isovolatility curves for all the alkanes is shown in eqn (4).²⁶

²t_R,cal = exp(exp(p₁ × ²T_e + p₂) × N + exp(p₃ × ²T_e + p₄) + p₅) (4)

²t_R,cal is calculated ²t_R. ²T_e is the elution temperature onto the ²D column which can be calculated as ²T_e = γ¹t_R + T₀; ²T_e ≤ final temperature with γ = temperature increase rate and T₀ = starting oven temperature. p_1–5 values are constants obtained from the least square curve fitting (using Solver in Microsoft Excel) between the experimental ²t_R and the ²t_R,cal data with the same ¹t_R along the isovolatility curves of alkanes with the carbon numbers of N (N = 8, 9, 10, …, 18, 19 and 20).

After calculation of ¹t_R and ²t_R for a peak of interest, eqn (4) was applied to approximate ²t_R of the n-alkane references eluting at ¹t_R of the peak under the same experimental conditions applied for the perfume sample. Kovatś relationship was then used to calculate ²I of this peak with the relationship:²²

²t_R(i), ²t_R(n) and ²t_R(n+1) are ²t_R of the peak, the alkane eluting before the peak and alkane eluting after the peak, respectively.

3. Results and discussion

3.1 Analysis with the cryogenic free comprehensive H/C MDGC approach

All peaks of the perfume sample in ¹D separation were H/C with a constant window of 0.2 min prior to ²D separation. In order to shorten the total analysis time, a cyclic multiple H/C strategy was applied²⁷ applying a ²D separation time of 5 min throughout the whole comprehensive analysis. This required 25 injections with a total analysis time of 25 h. Some of the H/C results are provided in Fig. 1.

Fig. 1 Results of 1st, 2nd and 3rd injections obtained from the CH/C MDGC analysis of the perfume sample using a H/C window of 0.2 min: (A–C, respectively) FID chromatograms and (D–F, respectively) total ion chromatograms. A–C represent ¹D separation results, and D–F represent ¹D + ²D separation.

Fig. 2A shows the contour plot obtained from the overall comprehensive analysis with the compound profile shown in Table 1. From the contour plot, several peaks coeluting in the ¹D separation were well separated onto the ²D polar column. For example, peaks 10, 26 and 28 co-eluting in ¹D separation were clearly separated onto the ²D polar column. The enhanced separation was obtained according to the long ²D column (60 m) with a thick film of the stationary phase (0.25 μm) which increases separation performance and peak capacity.¹⁸ With this approach, several compounds underwent wraparound during the separation as indicated by the yellow circles in Fig. 2A, which allowed effective use of ²D separation space. In addition, ²I calculation could be more effective with a smaller error under wraparound conditions. Note that due to the non-linearly decreasing nature of the isovolatility curves (e.g. see Fig. 4), the isovolatility curves of different alkanes get closer at higher ¹t_R and lower ²t_R leading to a high ²I calculation error.

Fig. 2 (A) The contour plot of the perfume sample obtained from the cryogen-free CH/C MDGC approach with the peaks of compounds undergoing wraparound circled in yellow and the split peaks of the same compound without wraparound circled in blue, and (B) the ¹t_Rvs.²t_R plot of alkane C19 under the same experimental conditions with the data points of two adjacent split peaks highlighted along the isovolatility curve undergoing no wraparound (the points where actual ²t_R of alkane C19 of <5 min), 1st wraparound (the actual ²t_R between 5 and 10 min) and 2nd wraparound (the actual ²t_R between 10 and 15 min) shown by the , , and , respectively.

Table 1 Tentative identification of compounds in the perfume sample. Compounds 1–71 are confirmed from the MS match score of >700, and ¹I and ²I differences of within ±37 and ±44 units, respectively, from literature values. Compounds 72–81 and 82–84 are confirmed by only the MS and ¹I match with “²I not available in the literature” and “their ²t_R values above the investigated range of the isovolatility curves”, respectively

No.	^1 t R (min)	^2 t R (min)	Compound	%Area	Match	Rank in MS library match	Literature		^1 I difference	^2 I difference
							^1 I	^2 I
a literature values obtained from the NIST 14 library (RI polar column). b literature values obtained from http://webbook.nist.gov/ (normal alkane RI, polar column, and temperature ramp).
1	10.40	4.92	2-Methyl-2-buten-1-ol	0.01	822	2	760	1320^a	−5	39
2	12.95	4.53	1-Hexanol	0.03	916	1	854	1348	−5	21
3	14.60	3.54	3-Methyl-2-butenyl acetate	0.69	948	1	902	1225	−1	17
4	15.84	5.34	Benzaldehyde	0.02	919	2	933	1520	1	1
5	16.00	2.73	α-Pinene	0.09	920	1	933	1033	5	−32
6	17.60	3.04	β-Myrcene	0.01	898	1	983	1169	−2	−9
7	17.62	2.90	β-Pinene	0.08	922	1	973	1128	8	−23
8	17.84	3.51	cis-3-Hexenyl acetate	0.04	934	1	986	1317	1	−16
9	18.45	4.46	1-Methyl-4-methoxybenzene	5.45	939	1	1003	1461	0	21
10	18.62	10.28	Benzenemethanol	0.42	921	1	1012	1864	−5	18
11	19.41	3.14	Limonene	0.33	924	2	1023	1216	3	10
12	19.49	3.15	Eucalyptol	0.01	805	1	1022	1227	6	5
13	19.80	8.41	α-Methylbenzyl alcohol	0.01	795	1	1037	1801	−1	32
14	19.94	3.09	trans-β-Ocimene	0.01	848	1	1038	1251	1	−31
15	19.54	10.29	Dipropylene glycol	5.36	907	1	1056	1892^b	−26	18
16	20.60	3.09	1,4-Cyclohexadiene, 1-methyl-4-(1-methylethyl)-	0.02	894	1	1050	1259	6	−27
17	20.74	3.92	2,6-Dimethyl-7-octen-2-ol	0.50	929	1	1062	1473	−3	−27
18	21.40	5.44	Methyl benzoate	4.32	956	1	1072	1654	3	13
19	21.31	4.00	trans-Linalool oxide	0.06	885	1	1074	1461	−1	14
20	21.90	4.28	Linalool	9.84	972	1	1086	1541	2	−5
21	22.20	4.14	3,5-Dimethylanisole	0.02	874	4	1105	1533^b	−10	−14
22	22.40	3.45	trans-Rose oxide	0.03	903	1	1121	1383	−21	−4
23	22.10	8.82	Phenylethyl alcohol	6.56	957	1	1088	1895	5	26
24	22.80	4.57	Fenchol	0.02	855	1	1105	1580	5	22
25	23.20	4.00	1,2-Dihydrolinalool	0.04	864	1	1121	1520	−1	−2
26	23.60	6.42	β-Phenethyl formate	0.41	908	2	1149	1771^a	−19	43
27	23.80	4.75	β-Terpineol	0.05	895	1	1137	1627^a	−3	22
28	23.97	5.65	Benzyl ethanoate	12.41	975	1	1139	1762	0	−6
29	24.35	3.75	cis-p-Menthan-3-one	0.02	911	1	1148	1528	0	−32
30	24.51	5.01	Ethyl benzoate	0.05	901	1	1150	1698	2	4
31	25.00	7.45	2,6-Dimethyl-3,7-octadiene-2,6-diol	0.02	885	1	1185	1945^a	−21	−16
32	25.01	4.60	1-Methyl-4-(1-methylethyl) cyclohexanol	0.92	933	2	1156	1650	8	10
33	25.20	4.56	Isomenthol	0.02	755	4	1186	1667	−17	−8
34	25.20	5.00	α-Methylbenzyl acetate	0.23	938	1	1166	1687^a	3	31
35	25.60	5.83	Methyl salicylate	0.25	952	1	1174	1795	5	24
36	25.70	4.40	Dihydrocitronellol	0.03	783	1	1182	1680	−1	−33
37	25.64	4.90	α-Terpineol	0.38	924	2	1175	1705	5	13
38	26.00	4.78	γ-Terpineol	0.14	897	1	1178	1684^a	11	27
39	26.93	4.89	Citronellol	5.48	953	1	1211	1761	1	−11
40	27.01	5.25	cis-Geraniol	0.36	928	1	1213	1815	1	−17
41	27.29	5.20	Isogeraniol	0.10	826	1	1237	1812	−16	−12
42	27.60	5.50	β-Phenethyl acetate	0.09	928	2	1228	1822	1	19
43	27.96	5.32	trans-Geraniol	4.93	953	1	1237	1854	1	−22
44	28.07	3.62	Linalyl acetate	2.08	955	1	1241	1563	−1	−12
45	28.30	4.63	α-Citral	0.03	899	1	1249	1732^a	−3	18
46	28.80	6.02	Hydroxycitronellal	1.03	935	1	1266	1929^a	−7	1
47	29.20	4.93	Anethole	0.20	936	2	1264	1817^a	5	2
48	29.40	4.76	Benzyl 2-methylpropanoate	0.01	881	1	1273	1784^a	1	17
49	31.00	8.95	Methyl anthranilate	0.22	962	1	1326	2194	−11	17
50	31.40	6.44	Dihydro-5-pentyl-2(3H)-furanone	0.22	942	1	1324	2024^b	2	23
51	31.56	3.66	Citronellol acetate	0.70	894	1	1335	1668	−5	−20
52	31.80	7.47	Eugenol	0.84	947	1	1335	2168	2	−25
53	32.71	3.92	Geranyl acetate	1.70	956	1	1361	1764	−1	−25
54	33.00	4.15	1-(2,6,6-Trimethyl-1-cyclohexen-1-yl)-2-buten-1-one	0.03	680	3	1395	1830^a	−27	−32
55	33.20	5.29	3-Methyl-2-(cis-2-penten-1-yl)-2-cyclopenten-1-one	0.02	833	1	1368	1964	5	19
56	34.79	6.23	3-Phenyl-2-propenyl acetate	2.90	951	1	1418	2176	−2	−39
57	34.80	4.00	Nopyl acetate	0.05	809	1	1413	1777^a	3	40
58	35.40	6.19	5-Hexyldihydro-2(3H)-furanone	0.04	847	1	1428	2119	5	35
59	36.00	3.57	β-Chamigrene	0.04	725	26	1475	1737	−26	1
60	36.28	4.67	Pentyl benzoate	0.07	828	1	1456	1987	1	1
61	36.80	3.53	α-Selinene	0.02	677	33	1491	1751	−19	−3
62	36.90	4.47	β-Ionone	0.77	922	1	1491	1971^a	−17	0
63	37.00	4.07	α-Cetone	1.33	933	1	1484	1877	−7	19
64	38.20	4.14	1-(2,6,6-Trimethylcyclohex-2-en-1-yl)-1-pentene-3-one	0.45	902	3	1503	1933	8	11
65	38.70	4.70	Isoamyl salicylate	0.12	884	1	1515	2033^a	10	35
66	38.90	4.17	β-Methyl ionone	1.86	736	2	1557	1988^b	−26	−15
67	39.30	5.23	α-(Trichloromethyl)benzyl acetate	0.35	856	1	1538	2197^a	4	−28
68	39.60	4.24	3,7,11-Trimethyl-1,6,10-dodecatrien-3-ol	0.03	730	1	1551	2020	0	−9
69	39.90	6.46	Diethyl phthalate	9.14	946	1	1564	2378^b	−5	−41
70	42.20	5.19	Methyl (2-pentyl-3-oxocyclopentyl) acetate	0.61	896	1	1649	2264^a	−23	−1
71	43.23	5.08	cis-3-Hexenyl salicylate	0.28	898	1	1644	2280^a	13	3
72	23.80	8.50	3,7-Dimethyl-1-octen-3-ol	0.01	761	1	1122	Not available	12	—
73	32.76	3.76	4-tert-Butylcyclohexyl acetate	0.83	932	1	1346	Not available	16	—
74	35.20	5.17	3,4-Dihydro-β-ionone	0.25	729	7	1460	Not available	−33	—
75	35.65	4.91	3-(4-Isopropylphenyl)-2-methylpropionaldehyde	0.25	921	1	1424	Not available	16	—
76	38.19	4.73	Lilial	6.94	909	1	1500	Not available	10	—
77	39.00	7.50	α-Methyl-3,4-methylenedioxy-hydrocinnamic aldehyde	0.61	940	1	1542	Not available	−8	—
78	39.00	9.06	Methyl 2-formamidobenzoate	0.03	878	1	1564	Not available	−30	—
79	39.05	4.23	6-Methyl-β-ionone	0.43	763	2	1566	Not available	−31	—
80	45.00	3.99	Caryophyllene acetate	0.08	792	1	1720	Not available	−10	—
81	47.45	4.72	9-Acetyl-2,6,6,8-tetramethyltricyclo (5.3.1.01,5) undec-8-ene	2.38	952	1	1780	Not available	5	—
82	41.20	7.95	1-Acetonaphthone	0.39	926	3	1592	2471^b	5	—
83	46.17	7.48	Benzyl benzoate	0.97	935	1	1732	2272	14	—
84	49.84	9.65	Benzyl salicylate	1.87	956	1	1836	2784^a	20	—

---

It should be noted for the CH/C approach that separate injections (runs) were performed resulting in several independent data sets prior to the combination into the CH/C analysis. Thus, the wraparound only occurred within the individual set having a time interval of 5n min (n is times of wraparounds = 1, 2, 3, …) between each H/C event. This is isolated from the adjacent H/C events performed in a different run.

In other words, wraparound occurs within the next adjacent modulation event(s) in conventional GC × GC using single injection whilst the wraparound in the CH/C analysis here solely affects the H/C event(s) within the next 5n min (not the adjacent event). Thus, the apparent ¹t_R value of a peak with wraparound in Fig. 2A should be subtracted with 5n min with the ²t_R added with 5n before reporting the actual ¹t_R and ²t_R data in Table 1. As a result, the possible ¹I values for a peak with vs. without wraparound are different with the 5n min intervals. This facilitates identification of n leading to actual ¹t_R and ²t_R for the wraparound peaks with the CH/C analysis according to the ¹I information for each peak observed with MS.

Take the example for a peak with a t_observed (= ¹t_R + ²t_R) of 28.90 min (see compound 10 in Table 1, ¹t_R = 18.62 min, ²t_R = t_observed − ¹t_R = 10.28 min on the contour plot). This compound was identified by comparison with the MS library as benzenemethanol (the match score of 921) and the literature I on the ¹D column of 1012. According to the 5n min concept, the possible H/C interval before this compound could be either 18.5–18.7 or 23.5–23.7. The former was assigned as the correct interval since this resulted in an experimental ¹I of 1009 being closer to the database value of 1012. The actual ¹t_R and ²t_R of this peak identified according to the weighted average approach were 18.70 and 10.20 min, respectively. In the case that the MS library provided incorrect compound identity (e.g. not matched with the ²I data), a lower rank identity with the next highest MS library match score and available I information was proposed. The calculation process to obtain the actual ¹t_R and ²t_R of this peak was then repeated until the compound identity showed both ¹I and ²I matched with the library data. It should be noted that compound wraparound can also be confirmed by performing a single H/C experiment solely focusing on a peak of interest.

In addition, several split peaks of the same compounds were observed; see most of the circled peaks in the contour plot (Fig. 2A). These split peaks (sub-peaks) resulted from a peak of a compound in the ¹D separation sampled into different H/C windows. With the high ²D separation efficiency and the great difference between ²D separation temperature of any two adjacent H/C events, these sub-peaks showed significantly different ²t_R (e.g. baseline separation) leading to the split peaks of the same compounds in the contour plot. ¹t_Rvs.²t_R data along the isovolatility curve of alkane C19 experimentally obtained using the peak parking approach were provided as an example to confirm such a peak splitting mechanism (Fig. 2B). With the focus on three sets of two adjacent sub-peaks undergoing no wraparound, 1st wraparound and 2nd wraparound, the sub-peaks of the alkane were separated better at the higher order of wraparound (more clearly split peaks at the lower temperature). Note that the presence of these split peaks is not problematic in data analysis since the individual sub peaks could be identified as the same compounds, and the overall peak areas were obtained by the summation of all the sub-peak areas. Also note that ¹t_R and ²t_R profiles were obtained based on the weighted averages of all the sub-peak times and their corresponding peak areas. After the determination of actual ¹t_R and ²t_R based on comparison with the MS and ¹I library, the compound identity was further confirmed from ²I data. Calculation of ²I was performed by using isovolatility curves.

3.2 Multi-location peak parking and construction of isovolatility curves

Generation of n-alkane isovolatility curves in this study is based on multiple injections and variation of an alkane elution temperature (or time) on the 2nd column. This was adapted from the multi-location peak parking (MLPP) technique applying a stopped flow and multiple injections within a single analysis. This technique was previously reported for investigation of longitudinal diffusion at different axial positions along the columns in liquid chromatography.²³ Application of this technique here provided the advantage of isovolatility curve construction without the use of a cryogenic trap. However, the approach has been adapted where a compound was injected under a continuous flow with the increasing oven temperature program. The program was stopped at different times with a subsequent temperature drop to 35 °C. This parked a peak of an alkane at different axial positions along the ¹D column. To this end, n-alkane standards were injected under the temperature program starting from 35 °C which was increased with a rate of 20 °C min⁻¹ and stopped at 100, 120, …, 220, 240 °C and 240 °C held for different times as the parking process for the alkanes (see Fig. 3 and Section 2.2.2 with details for the 16 sets of the parking approach provided in Table S-2†). There are 4 runs applied within each parking-eluting set. For example, the first set employed 1st and 3rd runs (for the peak parking) with the temperature program from 35 °C to 100 °C whilst the temperature program applied in 2nd and 4th runs (for the peak eluting) was the same as that applied in the perfume sample separation (see also Set 1 in Table S-2†). This corresponds to generation of one data point along each isovolatility curve for all the alkanes. The second parking-eluting set employed 1st and 3rd (parking) runs with the temperature program from 35 °C to 120 °C using the same temperature program applied in 2nd and 4th (eluting) runs as that applied in the perfume sample separation. Note that different alkanes were parked at different positions along the capillary (as a result of an alkane undergoing different temperature programs), and the smaller alkanes were not parked (eluting to the detector during the parking) when the elution was stopped at higher oven temperature (e.g. alkane C8 not parked inside the ¹D column with the parking process reaching 240 °C). With the additional blank injection, the parked peaks were then eluted through either the restrictor to the FID or the ²D column to the MS under the same temperature program used for the sample separation; see the example for alkane C17 provided in Fig. 4A. Different data sets of ¹t_R and ²t_R along the isovolatility curve of the alkane can be calculated from the alkane peaks at the FID and MS (Fig. 3). A regression model used for calculating the isovolatility curves for all the alkanes is shown in eqn (4) with the values of p₁, p₂, p₃, p₄ and p₅ being −0.0039, −0.6181, −26.4775, 5.5559 and −3.9524, respectively. The curve fitting results are provided in Table S-3 (ESI†). All the alkane isovolatility curves are plotted in Fig. 4B, with each alkane showing abrupt decrease in ²t_R at higher ¹t_R. This corresponded to the parking steps with the increasing final temperature from 100 °C to 240 °C. To validate the parking approach, the constructed isovolatility curves were applied to calculate ²I (²I_Cal) for 10 standard compounds with different functionalities. The ²I_Cal values were compared with I from the NIST 14 library on the polar columns (²I_Lit) showing differences (²I_Lit − ²I_Cal) ranging from −35 to 41. This is within the expected value of ±44 units indicating the reliability of the method. The related data are provided in Table S-4 (ESI†). These curves were further used to calculate ²I for the volatile compounds in the perfume sample.

Fig. 3 Diagram illustrating the peak parking and eluting approach for different alkanes (one parking-eluting event consisting of 4 runs).

Fig. 4 (A) The example of raw FID and MS chromatograms (with the DS off and on, respectively) to obtain the data points along the isovolatility curve of alkane C17, and (B) the generated isovolatility curves of all the n-alkanes obtained from 16 sets of 64 automated runs. Sets 1, 11 and 13 represent the 1st, 11th and 13th parking-eluting events provided in Table S-2.†

3.3 Tentative identification of compounds in perfume samples

The CH/C analysis of the perfume sample revealed 128 separated compounds which were successfully identified solely using the MS library with a match score of >700. ¹I and ²I were then applied to confirm the compounds obtained from the MS library. The results are summarized in Table 1 showing 84 identifiable compounds which can be divided into 3 groups according to the identification criteria. The first group is confirmed from the MS match scores of >700, and ¹I and ²I differences of within ±37 and ±44 units, respectively, from literature values. The second group is confirmed by only the MS and ¹I matches with “²I not available in the literature”. The third group is confirmed by only the MS and ¹I matches with “their ²t_R values above the investigated range of the isovolatility curves (where ²I could not be calculated)”. There were 71 compounds in the first group with five major compounds of benzyl ethanoate (12.41% peak area), linalool (9.84%), diethyl phthalate (9.14%), phenylethyl alcohol (6.56%) and citronellol (5.48%). In the second group, there were 10 compounds with three major compounds of Lilial (6.94%), 9-acetyl-2,6,6,8-tetramethyltricyclo(5.3.1.01,5)undec-8-ene (2.38%) and α-methyl-3,4-methylenedioxy-hydrocinnamic aldehyde (0.83%) whilst 1-acetonaphthone (0.39%), benzyl benzoate (0.97%) and benzyl salicylate (1.87%) were identified for the last group. For high confidence identification of compounds in the first group, most of the compounds with the highest MS match scores were correctly identified based on the confirmation from ¹I and ²I. However, 18 compounds were incorrectly identified solely by the MS match such as α-selinene (compound 61 in Table 1) with the 33rd rank of the MS match. This compound was correctly identified according to both ¹I and ²I match.

3.4 Effects of the H/C window and analysis time on the number of identified peaks and retention index calculation errors in comprehensive H/C analysis

For a given set of 84 identified compounds (Table 1), the number of successfully separated peaks (which could be identified according to comparison with the ¹I, ²I and/or MS database) using different H/C windows was approximated by using a previously established method.² In this study, simulation of ¹t_R and ²t_R is not required since they could be readily obtained from the comprehensive H/C analysis experiment above with a 0.2 min H/C window and a total analysis time of 25 h. The same compounds sampled with different H/C windows are expected to show almost the same ¹t_R + ²t_R (= t_observed detected by MS in this study) since compound elution has not been modulated or stopped by a cryogenic trap. Note that the small variation of t_observed could be caused by different H/C positions along the peak as well as the shift in peak time caused by intensity change according to the peak splitting using different H/C windows. Thus, ¹t_R and ²t_R can be recalculated into ¹t_R,recal and ²t_R,recal from t_observed of each compound for a given H/C window by using eqn (2) and (3). Also note in this section that ¹t_R,recal and ²t_R,recal of each compound were approximated from the sub-peak with the highest area (not the weighted average approach as that performed in the above section). The ¹t_R,recal and ²t_R,recal of the 84 identified compounds were then converted into grid-scale coordinates² (containing the data points of ¹t_{R,recal,grid-scale} and ²t_{R,recal,grid-scale} of all the peaks obtained from their ¹t_R,recal and ²t_R,recal divided by the ¹D and ²D widths, respectively). These coordinates represent the separated peaks with a resolution of ≥1 (with ‘1’ defined as baseline separation) in ¹D or ²D separation using average ¹D and ²D widths (¹w_b,ave and ²w_b,ave) of 0.2–2.5 min and 0.07–0.10 min, respectively, depending on the applied H/C window. The number of separated compounds could then be approximated by counting the number of grid-scale coordinates of the separated peaks. The related data of ¹t_{R,recal,grid-scale} and ²t_{R,recal,grid-scale} and number of separated peaks in the CH/C analysis with different H/C windows and analysis times are provided in ESI Microsoft Excel.†Fig. 5A shows the number of identified peaks slightly decreased (by 4%) with the increasing H/C window from 0.2 to 2.5 min. However, the ¹D peak capacity (¹n_c, approximated according to the approach proposed in (ref. 16) dropped more significantly (by 83%, see () in Fig. 5B) with the increasing H/C window from 0.2 to 2.5 min. The corresponding ²D peak capacity (²n_c) only decreased by 26% (see () in Fig. 5B). The better correlation between ²n_c and the number of identified peaks suggests that the separation performance was mainly governed by ²D separation using the long ²D column (60 m) with a high phase ratio (0.25 μm stationary phase film thickness). In addition, the number of identified peaks slightly increased with the total comprehensive analysis time until reaching the upper limit (around 84 peaks, see Fig. 5C). This indicates that there were no more compounds in this perfume sample or the selectivity of the selected column set reached the maximum limit.² As a result, application of a 2.5 min H/C window with a total analysis time of 2 h could result in a total peak capacity of 1134 with 81 identified peaks.

Fig. 5 Effect of the H/C window with the CH/C MDGC analysis on the (A) number of identified peaks and (B) ¹n_c () and ²n_c (), with the corresponding plots of (C) total separation time vs. number of identified peaks and (D) additional errors in ¹I () and ²I () approximated from the average ¹t_R and ²t_R values (in Table 1).

However, additional errors (to the conventional limit of ±30) in ¹I and ²I calculation were significantly increased with a wider H/C window (Fig. 5D), due to the ¹t_R and ²t_R errors calculated using eqn (1) and (2) discussed above. To this end, application of the 2.5 min H/C window led to errors of ±91 and ±318 in ¹I and ²I calculation, respectively, compared with the ±7 and ±14 obtained with a 0.2 min H/C window. This can be compared with the cryogenic H/C MDGC system which is expected to show the same ¹I additional errors whilst there is a much smaller additional ²I error (e.g. caused by the compound that could not be effectively trapped/released). This is due to the fact that all the H/C compounds were trapped to start eluting on the ²D column at the same time.

4. Conclusion

A simple peak parking approach was established and successfully applied for construction of alkane isovolatility curves and further ²I calculation. This was demonstrated with one of the most difficult MDGC cases applying comprehensive H/C MDGC without a cryogenic device. This approach applied a narrow H/C window of 0.2 min which required 25 injections to complete the overall comprehensive analysis. According to the developed peak parking technique, ²I values were calculated for 71 compounds in the perfume sample where identities of 15 compounds can be corrected based on the ²I information. The compromise between analysis confidence (error) and analysis time was also proposed depending on the application goal. To this end, the developed approach is expected to be useful for high confidence untargeted analysis of complex samples and applicable with any CH/C MDGC operation in the future.

Conflicts of interest

There are no conflicts to declare.

Acknowledgements

The authors gratefully acknowledge the financial support from the Chulalongkorn University's 100th Anniversary Chulalongkorn University Fund for Doctoral Scholarship and the 90th Anniversary of Chulalongkorn University, Ratchadaphiseksomphot Fund, and the research facility from the center for advanced analytical technology, Dow chemical Thailand Ltd, Map Ta Phut Industrial Estate, Rayong. CK acknowledges the Second Century Fund (C2F), Chulalongkorn University.

Notes and references

1. P. J. Marriott, S.-T. Chin, B. Maikhunthod, H.-G. Schmarr and S. Bieri, TrAC, Trends Anal. Chem., 2012, 34, 1–21.
2. C. Kulsing, Y. Nolvachai, P. Rawson, D. J. Evans and P. J. Marriott, Anal. Chem., 2016, 88, 3529–3538.
3. Y. Nolvachai, C. Kulsing and P. J. Marriott, TrAC, Trends Anal. Chem., 2017, 96, 124–137.
4. C. Mallet, LC·GC Eur., 2017, 13, 28–33.
5. P. Q. Tranchida, D. Sciarrone, P. Dugo and L. Mondello, Anal. Chim. Acta, 2012, 716, 66–75.
6. D. Barden and L. M. Gregor, LC·GC Eur., 2017, 13, 14–20.
7. B. J. Peter, J. Anal. Bioanal. Tech., 2016, 7, 2.
8. V. Ponsin, T. E. Buscheck and D. Hunkeler, J. Chromatogr. A, 2017, 1492, 117–128.
9. M. Jiang, C. Kulsing and P. J. Marriott, Anal. Bioanal. Chem., 2018, 410, 3185–3196.
10. M. Adahchour, J. Beens, R. J. J. Vreuls and U. A. Th. Brinkman, TrAC, Trends Anal. Chem., 2006, 25, 540–553.
11. M. Camenzuli, LC·GC Eur., 2017, 30, 346–351.
12. S. E. Prebihalo, K. L. Berrier, C. E. Freye, H. D. Bahaghighat, N. R. Moore, D. K. Pinkerton and R. E. Synovec, Anal. Chem., 2018, 90, 505–532.
13. Z. Liu and J. B. Phillips, J. Chromatogr. Sci., 1991, 29, 227–231.
14. C. Kulsing, Y. Nolvachai and P. J. Marriott, TrAC, Trends Anal. Chem., 2020, 130, 115995.
15. L. F. de Alencastro, D. Grandjean and J. Tarradellas, Chimia, 2003, 57, 499–504.
16. N. Thongdorn, T. Nhujak, P. Janta, A. Rueangthaweep, N. Hinchiranan and C. Kulsing, Anal. Methods, 2020, 12, 5160–5167.
17. C. Kulsing, Y. Nolvachai, Y. F. Wong, M. I. Glouzman and P. J. Marriott, J. Chromatogr. A, 2018, 1546, 97–105.
18. H. D. Waktola, C. Kulsing, Y. Nolvachai and P. J. Marriott, J. Chromatogr. A, 2018, 1549, 77–84.
19. D. R. Deans, J. Chromatogr. A, 1965, 18, 477–481.
20. P. Boeker, J. Leppert, B. Mysliwietz and P. S. Lammers, Anal. Chem., 2013, 85, 9021–9030.
21. B. Girard, J. Chromatogr. A, 1996, 721, 279–288.
22. M. Jiang, C. Kulsing, Y. Nolvachai and P. J. Marriott, Anal. Chem., 2015, 87, 5753–5761.
23. F. Gritti and G. Guiochon, J. Chromatogr. A, 2011, 1218, 896–906.
24. A. Amenduni, F. Massari, J. Palmisani, G. de Gennaro, M. Brattoli and M. Tutino, Environ. Eng. Manag. J., 2016, 15, 1963–1969.
25. A. Chisvert, L.-N. Marina, P. Miralles and A. Salvador, in Analysis of cosmetic products, 2nd edn, 2018, Elsevier, Boston. pp. 225–248.
26. B. Wang, H. Shen, A. Fang, D.-S. Huang, C. Jiang, J. Zhang and P. Chen, J. Chromatogr. A, 2016, 1451, 127–134.
27. B. Mitrevski, R. L. Webster, P. Rawson, D. J. Evans, H.-K. Choi and P. J. Marriott, J. Chromatogr. A, 2012, 1224, 89–96.

---

Footnote

† Electronic supplementary information (ESI) available: Tables S-1–S-4 and supplementary information Microsoft Excel. See DOI: 10.1039/d0ay01976c

---