*2.1. Comparison of 1DGC and GC*×*GC*

In typical 1DGC analysis, it is often difficult to achieve pure mass spectra for compounds in a co-elution peak, thus leading to unreliable results. With improved separation power and enhanced sensitivity, the GC×GC technique is able to resolve and detect more volatile aroma compounds in a complex sample compared to conventional one-dimensional GC-MS [30]. A clear illustration demonstrating the employment of GC×GC is presented in Figure 1. Both the chromatogram obtained by GC×GC–TOF/MS usinga4s modulation period and the total ion chromatography by 1DGC are shown. As can be seen in the partial chromatograms obtained by 1DGC and GC×GC-QTOFMS, linalool (Peak 1, <sup>1</sup>*t*<sup>R</sup> = 19.783 min, <sup>2</sup>*t*<sup>R</sup> = 1.405 s) and 2-nonen-1-ol (Peak 2, <sup>1</sup>*t*<sup>R</sup> = 19.849 min, <sup>2</sup>*t*<sup>R</sup> = 1.447 s) were responsible for the two peaks detected between retention times of 19.380 min and 19.850 min on the HP-5 MS column. However, three other minor compounds in addition to these two peaks were further separated as they exhibited different polarities on the DB-17 MS column; these were linalool oxide (Peak 3, <sup>1</sup>*t*<sup>R</sup> = 19.383 min, <sup>2</sup>*t*<sup>R</sup> = 1.467 s), *p*-cymenene (Peak 4, <sup>1</sup>*t*<sup>R</sup> = 19.450 min, <sup>2</sup>*t*<sup>R</sup> = 1.627 s), and benzoic acid, methyl ester (Peak 5, <sup>1</sup>*t*<sup>R</sup> = 19.716 min, <sup>2</sup>*t*<sup>R</sup> = 2.017 s). The co-eluted compounds in the peak region were interfered with by dominating compounds and would usually be ignored due to their low concentrations. In summary, GC×GC successfully resolved a total of 129 compounds, while only 45 compounds were separated in 1DGC (Table S1). The results revealed the great advantages of GC×GC analysis, which is suitable for the investigation of volatile compounds in complex samples.

**Figure 1.** Chromatographic analysis of *Rhododendron* by GC-quadrupole time-of-flight mass spectrometry (QTOFMS) and a comprehensive two-dimensional gas chromatography–mass spectrometry (GC×GC)-QTOFMS color diagram (1: linalool; 2: 2-nonen-1-ol; 3: linalool oxide; 4: *p*-cymenene; 5: benzoic acid, methyl ester).

#### *2.2. Identification of Common Volatile Components*

GC×GC–QTOFMS was used to characterize the detailed chemical composition of all the samples. Several hundred peaks were generated in the GC×GC contour plot with a peak detection threshold of S/N > 3. In total, 129 volatile compounds were tentatively identified in four *Rhododendron* samples based on (1) spectral similarity (both match and reverse match scores of >750), (2) comparison with molecular ions (within 5 ppm), if they existed, and (3) retention index (RI, ±35). Table S1 lists the complete information of the 129 volatile constituents detected by GC×GC-QTOFMS.

Figure 2 introduces the identification process of two examples (1,2-dimethoxybenzene and lilac aldehyde D). First, the National Insititute of Standards and Technology (NIST) library search for Peak 162 and Peak 189 resulted in 7 and 5 possible compounds, respectively, with match factor >750. Then, only exact mass analyses within a mass accuracy of <5 ppm were considered. For Peak 162, the measured accurate mass was 138.0676, which corresponds to a formula of C8H10O2. The accurate mass reduced the number of possible compounds to two isomers (1,2-dimethoxybenzene and 1,4-dimethoxybenzene). Last, their retention indices were reviewed for further confirmation. The GC×GC analysis provided an experimental RI value of 1151 for this peak, which matched 1,2-dimethoxybenzene (literature RI value of 1151) rather than 1,4-dimethoxybenzene (RIlit = 1168). Therefore 1,2-dimethoxybenzene was the final identified compound for Peak 162.

Taking Peak 189 as another example: The NIST library search provided several possible compound matches. Among them, seven possible compounds were screened out according to their relatively high match scores. Subsequently, the mass spectrum provided a measured mass of 168.1148, corresponding to a chemical formula of C10H16O2. This indicated that 2-methyl-2-(2-oxopropyl) cyclohexanone, lilac aldehyde D, and 2-hydroxy-4,4,6,6-tetramethyl-2- cyclohexen-1-one were three

possible compounds with the theoretical molecular ion mass of 168.1145. Lastly, the experimental RI value of the peak (RIexp = 1190) confirmed that lilac aldehyde D with RIlit of 1169 was the final identified compound for Peak 189, while the other two candidates, 2-methyl-2-(2-oxopropyl) cyclohexanone (RIlit = 1360) and 2-hydroxy-4,4,6,6-tetramethyl-2-cyclohexen-1-one (RIlit = 1272), were screened out. The results emphasis the importance of applying further confirmation on the base of the spectral library match, since the compound with the highest match factor might be mistaken for the identity of the component [31]. In conclusion, with complementary identification processes, GC×GC coupled with high-resolution QTOFMS produces more precise compound identification results.

**Figure 2.** Diagram illustrating the process of compound confirmation in GC×GC-QTOFMS.
