instrumental analysis of volatile (flavour) compounds in

HAL Id: hal-00929514https://hal.archives-ouvertes.fr/hal-00929514

Submitted on 1 Jan 1997

HAL is a multi-disciplinary open accessarchive for the deposit and dissemination of sci-entific research documents, whether they are pub-lished or not. The documents may come fromteaching and research institutions in France orabroad, or from public or private research centers.

L’archive ouverte pluridisciplinaire HAL, estdestinée au dépôt et à la diffusion de documentsscientifiques de niveau recherche, publiés ou non,émanant des établissements d’enseignement et derecherche français ou étrangers, des laboratoirespublics ou privés.

Instrumental analysis of volatile (flavour) compounds inmilk and dairy products

R Mariaca, Jo Bosset

To cite this version:R Mariaca, Jo Bosset. Instrumental analysis of volatile (flavour) compounds in milk and dairy prod-ucts. Le Lait, INRA Editions, 1997, 77 (1), pp.13-40. �hal-00929514�

https://hal.archives-ouvertes.fr/hal-00929514

https://hal.archives-ouvertes.fr

Lait (1997) 77, 13-40© ElsevierllNRA

13

Review

Instrumental analysis of volatile (flavour)compounds in milk and dairy products

R Mariaca, JO Bosset *

Federal Dairy Research Institute/FAM, CH-3097 Liebefeld-Bern, Switzerland

Summary - The present article reviews the most commonly used methods, techniques and equip-ments for instrumental analysis of volatile (flavour) compounds in milk and dairy products. After list-ing sorne previous important review articles, several methods commonly used for samp1e treatmentare described, as weil as the following techniques for extraction and concentration prior to gas chro-matographic (GC) analysis: static and dynamic headspace, steam distillation, high-vacuum distilla-tion, molecular distillation, direct extraction (liquidlliquid or liquidlsolid), supercritical fluid extrac-tion (SFE), simultaneous (steam) distillation extraction (SDE), dialysis, solid-phase extraction (SPE)and solid-phase microextraction (SPME). Two c1assical injection deviees are also described: on-column injection and the so-called 'purge and trap' system. The main advantages and disadvan-tages of CUITentcommercially available types of fused silica capillary columns are briefly considered.The newly developed 'chiral' phases are also described. The article reviews sorne of the numerousdetection systems used for qualitative and/or quantitative analyses such as FID or MS detection,FTIR detection, SCD, FPD and NPD detectors used for sulfur- and nitrogen-containing compo-nents, AED detection and the 'sniffing deviee'. Sorne usefullibrary search systems such as PBM,INCaS ™ and SISCaM (ie, MassLib®) are mentioned to complete the overview of this topic.Finally, this paper briefly points out sorne other methods (ie, photometrie), capable of determiningvarious specifie chemical functions responsible for flavour (carbonyl compounds, etc), as weil aspromising techniques involving new electronic noses.

milk / dairy product / volatile compound / flavour / analytical method

Résumé - L'analyse instrumentale des composés volatils (de l'arôme) du lait et des produitslaitiers, Le présent article passe en revue les méthodes, techniques et équipements les plus utilisés

Oral communication at the lDF Symposium 'Ripening and Quality of Cheeses', Besançon, France, February26-28, 1996.* Correspondence and reprints

14 R Mariaca, JO Bosset

pour l'analyse instrumentale des composés volatils (parfois responsables de la flaveur du lait et desproduits laitiers). Après avoir cité quelques articles de revue importants, certaines méthodes com-munément employées pour le traitement des échantillons sont décrites, de même que les techniquespermettant l'extraction, la concentration et l'injection des composés de l'arôme en vue de leur chro-matographie en phase gazeuse (CPG) : analyse statique et dynamique d'espace de tête ou d'effluves(vheadspace»), distillation à la vapeur, distillation sous vide poussé, distillation moléculaire, extrac-tion directe liquide/liquide ou liquide/solide, extraction supercritique (SFE), extraction-distillationsimultanée (SDE), extraction en phase solide (SPE) et microextraction en phase solide (SPME).Les deux techniques d'injection les plus classiques sont également présentées: l'injection surcolonne et le système appelé «purge and trap» (par entraînement et piègeage). Les principaux avan-tages et inconvénients des colonnes capillaires modernes en silice fondue sont brièvement cités.Les phases chirales récemment développées sont également décrites. Cet article cite quelques-uns desdétecteurs utilisés pour l'analyse qualitative et/ou quantitative des composés de l'arôme: FlD,MSD, FfIR, SCD, FPD et NPD (ces trois derniers détecteurs étant utilisés spécifiquement pourl'analyse des composés soufrés et azotés), AED ainsi que la détection par «sniffing», Les systèmesde recherche par bibliothèques de spectres tels que le PBM, INCOSTM et SISCOM (par exemple,Masslib®) sont mentionnés. Cet article fait un rappel de quelques autres méthodes (par exemple, pho-tométriques) à même de déterminer des fonctions chimiques responsables de la flaveur (composéscarbonylés, par exemple), ainsi que quelques techniques des plus prometteuses incluant les récentsnez électroniques.

lait / produit laitier / composé volatil / arôme / méthode d'analyse

INTRODUCTION

Our civilisation is characterised by 'hedonism',which partially explains why tlavours play anincreasing role in human nutrition and why somuch research has been carried out on this topic.Involved in the difference between gastronomyand nutrition, flavours have an impact on oursenses of smell and tas te, whereas food colourshave an impact on our vision. F1avours are notreally nutrients. They can stimulate our appetiteor warn us of a lack of freshness, or of the pres-ence of anti-nutrients, possibly preventing foodpoisoning.

There has been constant progress over thelast 25 years in the development of analyticalmethods for detecting volatile, odiferous corn-ponents such as tlavours and perfumes. Thisresearch has been supported by the food indus-try on the one hand, and the perfume and cos-metics industry on the other. For the former, the

goal has often been to replace labile compoundsor to concentrate interesting components, whilefor the latter, the aim has been to synthetise rareand expensive natural essences. Research on'off-flaveur compounds' in foods has coveredail steps from raw materials to final products,including the manufacture, transformation, pack-aging, transportation and storage of foods.

The development of efficient analytical tools,such as gas chromatography (GC) and relatedtechniques, has greatly contributed to progress inthis area. Main accessories include 'on-column'or 'purge and trap' inje ctors, fused silica capil-lary columns with stabilised (crosslinked) sta-tionary phases, highly efficient detectors, andfullYautomatic PC-analysers with software suchas Massl.ibv, Systat®, Sysgraph'ê, etc for datacollection and manipulation.

Sorne excellent review articles have beenpublished on instrumental and analytical assess-ment of volatile (tlavour) compounds (Day,

Instrumental analysis of volatile compounds

1967; Forss et al, 1967; Schwartz et al, 1968;Wong and Parks, 1968; Forss, 1969, 1971, 1972;Weurrnan, 1969; Paillard et al, 1970; Teranishiet al, 1971; Adda and Dumont, 1972; Dumontand Adda, 1972; Evans, 1972; Moinas et al,1973; Ney, 1973; Lacrampe et al, 1975; Jen-nings and Filsoof, 1977; Bemelmans, 1979;Nursten, 1979; Benkler and Reineccius, 1980;Lamparsky and Klimes, 1981; Maarse and Belz,1981; Cronin, 1982; Manning and Priee, 1983;Jennings and Rapp, 1983; Leahy and Reinec-cius, 1984; Nüüez and Bemelmans, 1984;Reineccius and Anandaraman, 1984; Sugisawaet al, 1984; Schreier and Idstein, 1985; Grosch,1990; Klein et al, 1990; Vandeweghe andReineccius, 1990; Careri et al, 1994; Xan-thopoulos et al, 1994; Urbach, 1996) as weil asa reference book (Marsili, 1996). The aim ofthis article is to complete and update existingliterature on instrumental analysis of volatile(flavour) compounds from milk and dairy prod-ucts.

PREPARATION AND TREATMENTOF THE SAMPLE, TECHNIQUESOF EXTRAC'!ION, CONCENTRATIONAND INJECTION OF THE VOLATILES

Sample preparation and treatment

General principles

Various procedures have been applied to the iso-lation of volatiles from complex matrices such asmilk and dairy products. Since most volatilesare present in both the sample and vapour phasein small amounts or even traces « 10 ug/kg),different steps for the extraction, concentrationand injection of volatiles into the GC arerequired.

Due to the sensitivity of sorne compounds toheat and/or oxygen, precautions have to be takenduring the preparation of the sample and the iso-lation of the volatiles to ensure that they remainunchanged and to minimise their los ses. More-over, the formation of new compounds (arte-

15

facts) should be prevented. Possible contamina-tion from the atmosphere and laboratory per-sonnel or their activities (eg, by smoking, cos-metics or solvent i~ the vicinity) must be avoidedin trace analytical determinations by workingunder clean-room conditions.

Only materials and chemicals of high puritymust be used. They should always be checkedfor contamination before use, and if neeessarycleaned and purified. Distilled and boiled ultra-pure water should be used, especially for 'purgeand trap' techniques. Blank test values shouldbe established on a regular basis to ensure theabsence of contamination or, if not possible,to subtract the blank from the sample chro-matogram.

Homogenisation of solidand semi-solid samples

Volatile (aroma) compounds in dairy productsare generally distributed in a heterogeneous way.Trace analysis requires careful homogenisationof the sam pie prim to analysis.

A commonly used, convenient and mild tech-nique for chee se is, for ex ample, freezing fol-lowed by grating of the sam pIe at low tempera-ture. If necessary, the powder obtained can thenbe dispersed in water using a high speedhomogeniser to obtain representative sampi esprior to taking an aliquot.

When using headspace techniques, thehomogenate obtained will often be adjusted withalkali to pH 7.5 prior to headspace analysis tokeep the high concentration of volatile acids dis-solved in the aqueous solution as salts. The frac-tion to be analysed then mai nIy con tains tracesof the numerous neutral and alkaline componentspresent (Imhof and Bosset, 1991). Making ahomogenate alkaline can destroy sorne essentialcomponents, such as lactones (Urbach et al, 1972).

Separation into acidic, alkalineand neutral fractions

In general, f1avour from dairy products is madeup of a large number of volatile compounds


(Nursten, 1977; Maarse, 1983; Bosset et al,1995; Molimard and Spinnler, 1996; Urbach,1996). These may include significant amounts offree carboxylic acids, sulfur compounds, alkalinenitrogen-containing substances such as aminesand substituted pyrazines (Nursten and Sheen,1974) and pyridines and many neutral com-pounds such as carbonyl compounds (mostlymethylketones and aldehydes), primary and sec-ondary alcohols, esters, lactones, ethers, aliphaticand aromatic hydrocarbons (sorne ofthem poly-cyclic: Dafflon et al, 1995), as weil as multi-functional components. In consequence, a pre-liminary separation into acidic, alkaline andneutral fractions can often help if it is includedbefore their instrumental and sensory analysis(Cronin, 1982). Figure 1 highlights a commonlyused scheme for such a separation.

Extraction and concentration techniques

In general, volatile (aroma) compounds are fairlyIipophilic. They are consequently dissolved inthe original fatty phase or bound to the proteins.

1) make alkaline

2) extract wilh elherpH 11.5

organic layer

The first step in each procedure consists of theirextraction from such a matrix and usually resultsin a dilute aqueous solution. Prior to GC-anal-ysis, these volatiles must be separated from waterand usually concentrated by using an organicsolvent or solvent mixture. However, some mod-em techniques now allow capture of these com-pounds by direct adsorption onto a sol id phase,so avoiding tedious and time-consuming steps.

Numerous techniques have been proposedfor the extraction, concentration and injectionof volatile food product components, especiallyof milk products. An update of the methods mostfrequently used may be summarised as follows:

Headspace techniques

Regarding headspace techniques, the sample tobe analysed is a gas phase aliquot containingvolatiles released from the condensed phase. Intheir classical static form (Gomer et al, 1968;Hild, 1979; Kahlhofer, 1981; Priee and Man-ning, 1983; Degorce-Dumas et al, 1986; McNealand Hollifield, 1990; Ulberth, 1991; Gaafar,1992; Linssen et al, 1995), these techniques are

1) adjusl la pH 2.0organic layer(free acid fraclion)2) extract wilh ether

1) adjusl la pH 4.5

2) extract with ether

aqueous layer(alkaline fraction)

1) adjust la pH 7.52) ether

free alkalinefraction

Fig I.Technique used to isolate the acidic, neutral and alkaline volatile (flavour) compounds.Technique utilisée pour séparer les composés volatils acides, neutres et alcalins (de l'arôme).


restricted to the most volatile components. Thedetection of flavour compounds greatly dependson their concentration and vapour pressure, asweIl as on the temperature and matrix of thefood product. A very interesting application isthe analysis of volatile acids of cheese, basedon esterification in the gas phase prior to analysis(Badertscher et al, 1993). To obtain more con-centrated extracts, dynamic headspace methodshave been developed (Oria et al, 1987; Urbach,1987; Horwood, 1989; Imhofand Bosset, 1991;Wijesundera and Urbach, 1993; Contarini andLeardi, 1994; O'Hare and Nursten, 1994; Woodet al, 1994; Yang and Min, 1994), in which thevolatile components of the gas phase are con-tinuously removed and concentrated in a coldtrap or adsorbed onto an inert support, and finallyrecovered either by thermal desorption or byelution with a suitable solvent (Westendorf,1985).

Steam distillation techniques

Regarding ste am distillation techniques, theaqueous distillate is concentrated by Iiquid/liq-uid partitioning or by cryoconcentration.Although widely used, the se very popular tech-niques present several problems. For example,highly volatile components may have low recov-eries and/or may be masked by the chromato-graphie peak of the solvent. The solvent maycontaminate the sample. In addition, unstablecompounds may decompose thermally if thesteam distillation and solvent extraction are notcarried out under reduced pressure. This tech-nique has been largely applied to cheese vari-eties such as Emmental (Vâmos-Vigyazô andKiss-Kutz, 1974), Limburger (Parliment et al,1982), Fontina (Ney and Wirotama, 1978) andCheddar (Vandeweghe and Reineccius, 1990), asweIl as to milk (Jeon et al, 1978; Alm, 1982)and other milk products such as butter (Stark etal, 1978), ghee (Jain and Singhal, 1969; Wadhwaet al, 1979), butter oil (Urbach et al, 1972) andmilk powder (Ferretti and Flanagan, 1972).Steam distillation has been described recently(Xanthopoulos et al, 1994) as an advantageousmethod for isolating volatile substances like

17

acetaldehyde, ethanol, diacetyl and acetoin indairy products.

High-vacuum distillation techniques

High-vacuum distillation techniques producesmall volumes of conceritrated aqueous extracts,which are in turn extracted with organic sol-vents such as methylene chloride and diethylether. These techniques prevent thermal degra-dation by working at ambient or even sub-ambi-ent temperature, but usually require large sam-pie amounts and the operations are verytime-consuming. High-vacuum techniques havebeen applied to solid fat after ultracentrifuga-tion at 25 000 g (McGugan et al, 1968; Man-ning, 1974) or liquid fat (Liebich et al, 1970)from a homogeneous suspension of fat in water(Bradley and Stine, 1968; Parliment et al, 1982),or even from fat/water/isopentane mixtures(Sloot and Harkes, 1975). Sour cream, as anaqueous suspension, can also be distilled underhigh vacuum (Mick et al, 1982). Moinas (1973)used a 'gas stripping' technique combined withhigh vacuum distillation by purging the volatileswith inert gas at 10w temperatures (Moinas etal, 1973; Groux and Moinas, 1974; Kowalewskaet al, 1985). The alkaline volatile componentsof Swiss Gruyère cheese have been isolated byapplying high vacuum gas stripping to gratedcheese samples at room temperature, followed byether extraction of the distillates after they weremade alkaline (Liardon et al, 1982). The volatileodorous constituents of different French bluecheeses were isolated by high-vacuum distilla-tion (Gallois and Langlois, 1990). Volatile com-pounds of different French Appellation d'Orig-ine Contrôlée (AOC) cheeses were extracted byvacuum distillation (Guichard, 1994).

Molecular distillation techniques

Molecular distillation techniques are very similarto high-vacuum distillation and involve the directtransfer of volatile compounds from the matrix toa cold condenser. These techniques require veryshort distances between the food sample and con-denser, as weil as the use ofhigh-vacuum pump-ing systems « 10-3 Pa). Such techniques have

taneous extraction ofmono-, di- and triglyceridesand the volatile compounds, the application ofthese techniques is limited to samples with a lowfat content. Currently in the case of chee se sam-pies, SFE-SFC (supercritical fluid chrornatog-raphy) coupling (Gmür et al, 1987a,b,c) seemsto be a possible way of making use of this promis-ing technique; Gmür et al (1986) have alreadyused this on a preparative scale. Commerciallyavailable SFE equipment (ie, Hewlett Packardtype 7680Aff SFE module or Applied Separa-tions type Spe-de SFE) now allows a release ofpressure (after the restrictor) onto a cartridge con-taining a reverse phase material (similar to that ofreverse-phase high performance liquid chro-matography, RP-HPLC columns), which can beeluted stepwise by eluents of different polarities.This improvement in SFE systems now permitsanalysis of trace components in the presence ofhigh amounts of glycerides without any interfer-ence or other difficulty.


been used to isolate volatiles from fats and oils,and necessitate water-free samples. Libbey et al(1963) called this 'cold-finger molecular distil-lation'. Day and Libbey (1964) used this tech-nique to analyse cheese flavour (Cheddar), Starket al (1976) for butter oil, and Urbach (1982) formilk fat. Dumont et al (1974) described this tech-nique as 'distillation sous vide poussé' (trans-lated as 'high-vacuum distillation'). Moleculardistillation has also been carried out by Guichardet al (1987) for Comté cheese, and by Guichard(1995) for Parmigiano, Emmental, Comté andBeaufort cheeses. Volatile compounds from oldCheddar cheese were isolated by molecular dis-tillation (Christensen and Reineccius, 1995).

Direct extraction techniques

Regarding direct extraction techniques, an extractis obtained by liquid/liquid or liquid/solid par-titioning. While the se techniques are generallyrapid and efficient, they should only be appliedto samples with a very low fat content, to keepsimultaneous fat extraction low. Because of itslow solvent power for triglycerides, acetonitrilecan be used to extract cheese flavours. How-ever, the relatively high boiling point ofthis sol-vent (82 "C) causes loss of volatile componentsduring the concentration step. Several peaks inthe chromatogram will be masked by this sol-vent (Wong and Parks, 1968; Lamparsky andKlimes, 1981). Preininger et al (1994) treatedgrated Swiss cheese directly with diethyl ether,filtered off the insoluble material and finally,under high vacuum, distilled the volatile com-pounds from the non-volatile material (ie, fat),which had dissolved in this solvent. In order toseparate the neutral and alkaline fractions fromthe acidic fraction the distillate obtained, whichsmelled intensively of Emmentaler cheese, wastreated successively with aqueous sodium bicar-bonate and aqueous hydrochloric acid (fig 1).

SupercriticaI fluid extraction methods (SFE)

Supercritical fluid extraction methods (SFE) usingcarbon dioxide as solvent, avoid the problems ofconcentrating the extracts. Because of the simul-

SimuItaneous (steam) distillation extractionmethods (SDE)

Simultaneous (steam) distillation extractionmethods (SDE), which were first developed byNickerson and Likens (1966), use only very low-boiling sol vents such as pentane for the con-centration of the aroma volatiles. In a continuousprocess, the condensing water vapour is extractedby the condensing solvent vapour, yielding ahigh extraction rate. Both the water and the sol-vent are recirculated. A micro-scale modifica-tion, proposed by Godefroot et al (1981), andde Frutos et al (1988) has recently been appliedto cheese volatiles. Maignial et al (1992) haveproposed a system working under reduced pres-sure and at low temperature (20--40 "C) to pre-vent thermal generation of artifacts. Blanch etal (1993) have proposed a micro SDE apparatusworking at normal pressure. Careri et al (1994)used this method for the extraction of volatiles inParmesan cheese.

Dialysis techniques

Dialysis techniques, which separate volatilesaccording to their ability to diffuse through a


membrane, result in a concentration gradient.Although a high degree of concentration wasobtained, the sample preparation and dialysiswere very time-consuming. This technique wasdeveloped by Benkler and Reineccius (1980),modified by Chang and Reineccius (1985), andcompared with other techniques by Vandewegheand Reineccius (1990). A good review on theuse of membranes in sample preparation is givenby Majors (1995). Application of pervaporation(partial vaporisation through a non-porous perms-elective membrane) in food processing has beenreported by Karlsson and Trâgârdh (1996).

Solid-phase extraction methods (SPE)

Solid-phase extraction methods (SPE) have beenused for the selective separation and concentrationof analytes from Iiquid sampi es. Extraction of theanalytes is based on the distribution of dissolvedsubstances (ie, volatiles) between a solid-phase sur-face and the Iiquid sampie. Separation may be aresult of differences in polarity, molecular size, oreven differences with respect to ion-exchange capac-ity. Solid-phase extraction or even micro-solid-phase extraction (SPME) techniques (Woolley andMani, 1994) are relatively new methods for isolat-ing volatiles. Takacs (1989) reported that flavourcomponents in UHT processed milk could beextracted with CI8 Sep-Pak materials and subse-quently eluted with methylene chloride. A rapidand sensitive solid phase method was developedby Coulibaly and Jeon (1992) for the extraction oflactones at ppb levels. Analysis of volatiles in var-ious wines using solid phase extraction is describedby Sedlâckovâ et al (1995). The use of an on-IineSPE-GC-MSD instrument has been recentlydescribed by Brinkman and Vreuls (1995).

Ali aqueous extracts may be concentratedusing the freeze-concentration technique, weildescribed by Maarse and Belz (1981), andquoted from the Iiterature by Urbach (1996).Commercial equipment had been proposed (eg,Virtis Co, model 3-100, ser 1049), but lessexpensive kitchen deviees used for domestic icecream preparation also seem to be convenientfor such a purpose.

19

Injection techniques

Two types of injection are currently used: on-column injection and 'purge and trap' injection.One of the most popular and least expensivepieces of equipment used for many years forloading the volatile (f1avour) components offoodstuffs onto a capillary column is the so-called 'cold' injection technique or on-columninjection, where the liquid extract is injecteddirectly onto the column. This prevents the lossof substances as weil as the discrimination whichoccurs with the use of a split/splitless inje ctor.Band broadening can be suppressed by station-ary phase focussing via a retention gap consist-ing of a length of uncoated column material(Grob et al, 1985).

The other very common injection techniqueused is the so-called 'purge and trap' methodalready mentioned above (see 'dynamicheadspace technique'), which is based on theadsorption of compounds onto a trap and/or con-densation of the volatiles by low temperature('cryofocussing') before injection (Imhof andBosset, 1991).

CHROMATOGRAPHIC SEPARATION,DETECTION, IDENTIFICATIONAND QUANTIFICATION

One-dimensional gas chromatography

The qualitative and quantitative analyses of com-plex mixtures of volatile (aroma) componentsrequire extremely efficient capillary columns(packed columns are no longer used). The GCseparation should be performed exploiting boththe volatility and the polarity of the analytes. Itmust be able to coyer a wide spectrum of con-stituents from non-polar to very polar ones inline with the great variety of the chemical func-tions present in the mixture. Major progress inthis area has been achieved by the introductionof capillary columns based on flexible fused sil-ica tubing material, weil known under the abbre-

20 R Mariaea, JO Bosset

viation FSaT (fused silica open tubularcolumns),

Most capillary columns used today contain astationary-phase layer that has been cross-Iinkedwith the fused silica after coating. Today ail tra-ditional high molecular weight phases are avail-able as 'bonded' or 'stabilised stationary phase'under various trade names. These modem cross-linked phases have polarities and separation per-formances at least comparable to or even betterthan those of the silicone gum or Carbowax-type phases they replaced. Capillary columnswith internai diameters ranging from 0.10 mm(small bore) up to 0.53 mm (wide bore) are com-monly used for the analysis of volatile (flavour)compounds.

Polyethyleneglycols (PEG, Carbowax) ofvarious types have been the most popular highlypolar phases and the best stationary phases forthe separation of food flavour components formore th an 25 years (Mick et al, 1982). How-ever, the performance of polydimethylsiloxaneas phases with a lower polarity came to be testedrecently, especially in dynamic headspace anal-ysis (ie, using 'purge and trap' equipment).These Jess polar phases prevent the haphazardshifts in retenti on time observed for sorne com-pounds (due to moisture or the high water con-tent of aqueous samples) when PEG-coated cap-illary columns were used. Moreover,polydimethylsiloxane phases allow use of amuch wider temperature range (from -30 up to300 "C) th an PEG phases (from + 45 up to250 "C) and are less sensitive to oxygen at hightemperature (improved longevity).

The best performance for the GC analysis ofsam pies of ripe Swiss Emmental cheese wasobtained with a 30 m x 0.32 mm id capillarycolumn coated with a 4 um layer of 100% poly-dimethyl siloxane (SPB-I-SULFUR) (Imhofand Bosset, 1994a). The film thickness of 4 umensures better resolution and higher loadingcapacity of the column with samples contain-ing many compounds in different concentrations(fig 2). This column made possible the identifi-cation of the highest number of compounds (ie,58) in the cheese sample (table 1). Furthermore,

highly volatile compounds such as ethanal couldbe easily detected. Sanz et al (1992) used amixed-phase (43% FFAP: polar phase and 57%av -1: non polar) and found it to be better forthe analysis of volatile fractions from foods.

aüm x O.32mm x 1pm

UlliJI, ..............,. .l~ 25 35 4S

Ti me (min.)

Fig 2. Influence of eapillary eolumn dimensions andlayer thiekness on the ehromatograms obtained froma sample of ripe Swiss Emmental eheese using a polydi-methylsiloxane stationary phase (SPB-l-SULFUR).The same sample was analysed with the same dynamieheadspaee method (purge and trap) using identiealsample preparation but three different eapillaryeolumns. The temperature programme (45 "C for 13min; 5 "Czmin to 220 "C; 220 "C for JO min) as wellas the seale (abundanee and retention lime) used wereidentieal for the three ehromatograms. For further ana-lytieal details, see Imhof and Bosset (1994a).Influence des dimensions de la colonne capillaire et del'épaisseur du film sur les chromatogrammes d'unéchantillon d'emmental suisse affiné, en utilisant unephase stationnaire en polydiméthylsiloxane (SPB-I-SULFUR). Le même échantillon a été analysé avecla même méthode (espace de tête dynamique par«purge and trap»} en recourant à la même préparationde l'échantillon, mais à trois colonnes capillaires dif-férentes. Le programme de température ( 13 min à

45 "C; 5 "Cànin jusqu'à 220 "C; 10 min à 220 oC), demême que les échelles utilisées (abondance et temps derétention), étaient identiques pour les trois chro-matogrammes. Pour d'autres détails analytiques, voirlmhof et Bosset (1994a).

Instrumental analysis of volatile compounds 21

Table I. Influence of the characteristics of the separation column on the resolution (number of peaks separated*).Influence des caractéristiques de la colonne de séparation sur la résolution (nombre de pics séparés*).

No of Stationary Dimensions Polarity Composition Suppliercompounds phase [m x mm x pm)identifled

48 DB-wax 60 x 0.25 x 0.25 Polar 100% Polyethyleneglycol J+W33 Supelco wax 60 x 0.32 x 0.5 Polar 100% Polyethyleneglycol Supelco51 av 1701 50 x 0.32 x 0.5 Intermediate 88% Methylsiloxane Macherey & Nagel

7% Phenylsiloxane5% Cyanopropylsiloxane

51 Ultra-2 50 x 0.32 x 0.5 Non polar 95% Dimethylpolysiloxane Hewlett-Packard5% Diphenylpolysiloxane

54 SE54 50 x 0.32 x 0.5 Non polar 94% Methy Isiloxane Macherey & Nagel5% Phenylsiloxane1% Vinylsiloxane

58 SPB-I** 30 x 0.32 x 4.0 Non polar 100% Dimethylpolysiloxane Supelco47 SPB-l 30 x 0.32 x 1.0 Non polar 100% Dimethylpolysiloxane Supelco57 SPB-I 60 x 0.25 x 1.0 Non polar 100% Dimethylpolysiloxane Supelco

* Swiss Emmental cheese sam pIe (Imhof and Bosset, 1994a); ** SPB-I SULFUR.* Échantillon d'emmental suisse (lmhof el Bosset, 1994a); ** SPB-/ SULFUR.

Table II Iists sorne chromatographie columnsused for the analysis of volatile components inmilk and dairy products.

Multidimensional gas chromatography(MDGC)

The recent development of chiral stationaryphases such as cyclodextrins (Bicchi et al, 1992)brought about new highly efficient separationtechniques, such as enantioselective multidi-mensional gas chromatography (MDGC). Theuse of chiral MDGC in the food industry hasdramatically increased in the past few years. Ithas been applied to identify for example theadulteration of essential oils, fruit juices andmany flavoured beverages (Woolley and Mani,1994). An enantioselective fingerprint can beused to differentiate natural from synthetic rawmaterials and such a fingerprint is much more

difficult to imitate in an adulterated product(Pierce, 1995). The other method describedrecently by Casabianca et al (1995) deals withthe combination of gas chromatography withisotope-ratio mass spectrometry (GC-IRMS).Signifieant differences were found between theratio of stable isotopes (eg, l3C/l2C, 2H/IH and180/(60) of natural and synthetic flavours infood products. Scope and limitations of enan-tioselectivity and isotope discrimination arereported by Mosandl (1995).

Detection systems

Many detectors have already been proposed forthe qualitative and quantitative analysis ofvolatile (flavour) constituents of foods. Althoughthe following list is not exhaustive, it does indi-cate those that have been most frequently quotedin the literature:


Table Il. Overview of sorne separation columns used for the analysis of volatile components of milk and dairyproducts.Tableau synoptique de quelques colonnes de séparation utilisées pour l'analyse des composés volatils du lait eldes produits laitiers.

Stationary Dimensions Polarity Compounds/product Rejphase (m x mm x um} investigated

FFAP 60 x 0.3 x? Polar Swiss Gruyère cheese Liardon et al (1982)DB-5 30 x 0.32 x 1.0 Non polar Gruyère de Comté Guichard et al (1987)DB-l 30 x 0.53 x 1.5 Non polar VOC inmilk McNeal and Hollifield (1990)SE-30 22 xO.3 xO.1 Non polar Artisanal cheese varieties de Frutos et al (1991)SP 1000 25 x 0.2 x 0.4 PolarDB-wax 60 x 0.25 x 0.25 Polar Parmigiano-Reggiano Bosset and Gauch (1993)

Mahon, Comté, Beaufort,Appenzeller

SE-54 30 x 0.32 x 0.3 Non polar Key odorants Preininger et al (1994)OV-1701 30 x 0.32 x 0.3 Intermediate in Emmentaler cheeseDB-FFAP ? ? ? PolarDB 1701 60 x 0.32 x? lntermediate AOC chee ses Guichard (1994)DB-5 30 x 0.25 x 1.0 Non polar Cheddar cheese Yang and Min (1994)DB-wax 30 x 0.25 x 0.25 Polar Parmesan cheese Careri et al (1994)SE-54 25 x 0.32 x 1.0 Non polar Flavour defects in milk Ulberth and Roubicek (1995)

powderDB-wax 30 x 0.25 x 0.25 Polar Cheddar cheese Christensen and Reineccius(1995)CPSil5 CB 50 x 0.32 x 5.0 Non polar Gorgonzola cheese Contarini and Toppino (1995)

'1: Not specified in the original publication. VOC: volatile organic compounds; AOC: appellation d'origine contrôlée.? : No" spécifié dans la publication d'origine .. VOC: composés organiques volatils: AOC: appellation d'origine contrôlée.

The f1ame ionisation detector (FID)

The flame ionisation detector (FIO) is certainlythe most popular detector used for GC analysis.Ils main advantages are its low cost, universalapplicability (except for carbon dioxide andformic acid), extremely wide linear range ofconcentration for quantitative determinations,and long-term stability (the response factors foranalytes remain very constant over long peri-ods). As a consequence of its universality, thistype of detector is neither specifie, nor selec-tive. GC-FIO systems have been widely appliedto dairy products (Barbieri et al, 1994; Careri etal, 1994; Preininger et al, 1994). Grigoryeva etal (1994) used it for the analysis of Dutch cheeseand Monnet et al (1994) to investigate volatilecompounds in fermented milk by static

headspace techniques. Georgala et al (1995) alsoused this detector for the assay of flavour com-pounds in ewe's milk and yoghurt, Ulberth andRoubicek (1995) for detecting flavour defectscaused by lipid oxidation, and Contarini andToppino (1995) for volatile organic compounds(VOC) in Gorgonzola cheese during ripening.

The mass selective detector (MSD)

The mass selective detector (MSO) is much moreexpensive than the FIO, but it allows both qual-itative and quantitative determinations. Twomain types are now commercially available: thequadrupole and the ion trap system. Both sys-tems may be used in the full-scan mode (eg,total ion CUITent,TIC) in order to identify com-pounds via their fragmentation pattern, and in


the selected ion storage (SIS) or selected ionmonitoring (SIM) mode to quantify target corn-pounds with high selectivity and sensitivity. Inthe scan mode, the identification is mostly per-formed by comparison with a library of refer-ence spectra (eg, EPA/NIH library, Wileylibrary, TNO library). In the SIS or SIM mode,the voltages on the ion trap or quadrupole rodsare adjusted stepwise to detect only a small num-ber of selected ions (eg, molecular ion, heavyions, etc) highly specifie for the substances tobe quantified. Applications of the MSD/SCANmode to volatile compounds of Emmental cheesehave recently been reported by Bosset et al(1995). The MSD/SIM mode was applied forthe detection of sulfur-containing compoundsin milk after various heat treatments (Bosset etal, 1994, 1996). The GC-MSD technique is themost popular technique used to identify volatilecompounds in dairy products: neutral volatilecompounds in fresh bovine, ovine, caprine andwater buffalo raw milks (Moio et al, 1993a);off-flaveur compounds in spray-dried skim milkpowder (Shiratsuchi et al, 1994); the volatilecomponents of Cheddar and Swiss chee se (Yangand Min, 1994); the volatile components of full-fat, reduced-fat and low-fat Cheddar cheese(Delahunty et al, 1994); the aroma fraction ofParmesan cheese (Barbieri et al, 1994), andvolatile compounds in goat milk chee se (Vidal-Aragon et al, 1994). Dynamic headspace analysiswith MS detection was also used by Laye et al(1995) to identify 33 volatile compounds fromcommercial whey prote in concentrate (WPC).

Ion-trap instruments where ionisation and rnassanalysis both occur in the ion-trap produce non-standard mass spectra which cannot he comparedwith commercial data bases, In the latest ion-trapinstruments, ionisation and mass analysis havebeen separated and it is now c1aimed that theseinstruments do produce standard spectra.

The Fourier transform infrared spectroscopy(FTIR) deteetor

The Fourier transform infrared spectroscopydetection (FTIR) is very expensive and less sen-

23

sitive than the MS detection due to its muchhigher dead volume and the secondary infraredemission from the heated cell (light-pipe), Sim-i1arly to MS, the recorded spectra can be com-pared with a library ofFrIR reference spectra toaid in compound identification, but the com-mercially available databases are much smallerthan those of mass spectra. The FTIR spectraare useful complements to mass spectra, espe-cially for the differentiation of isomers for whichMS is not usually helpful (Jackson et al, 1993).Terpenes are an example of substances that mayhave many isomers due to the position of doublebonds. In GC/IR, however, ail of these sub-stances show specifie absorption bands at dif-ferent wave numbers (Kasalinsky and McDon-ald, 1983). The principal capability of the useof high-resolution gas chromatography(HRGC)-FTIR in the analysis of food tlavourshas been demonstrated by Schreier and Idstein(1985). Industrial applications of GCIFTIR havebeen described by Namba (1990) with sorneinteresting examples relevant to foods andfragrances.

The flame photometrie deteetor (FPD)

The tlame photometrie detector (FPD) is highlyselective for sulphur-containing componentswhen a suitable filter is used. Although reliableand sensitive, this detector also suffers from anumber of major drawbacks, such as: the weilknown but unavoidable quenching effects (inter-ferences, eg, from hydrocarbons and carbonmonoxide); the non-linear (but sigmoidal)response of the output signal vs the concentrationof sulphur species (Burdge and Farwell, 1994).

Prior to the introduction of the SCD (see thefollowing detector), the dominant sulphur-selec-tive detector used in GC was the FPD. Thisdetector configured in parallel with an FID wasused by Manning and Moore (1979) to de ter-mine hydrogen sulphide and methanethiol byheadspace analysis of hard cheeses. Aston andDouglas (1981) described the detection of car-bonyl sulphide in Cheddar chee se also using aheadspace technique, whereas Parliment et al(1982) examined the VOC of Limburger cheese.


The sulphur chemiluminescence(SCD) detecter

The sulphur chemiluminescence detector (SCO)is an universal sulphur detector. It has very highsensitivity and selectivity, a linear response toconcentration of sulphur species, a large dynamicrange, and is free from quenching effects over awide range. A detailed description of the char-acteristics of this type of detector is given byBurdge et al (1994). Tuan et al (1995) recentlypublished an interesting evaluation of the per-formances of FPO, MSO (working in SCANand SIM mode) and SCO for the determinationof sulph ur in natural gas. They c1aimed that theSCO represented the best choice. The detectionof sulphur f1avour volatiles in cooked milk wasreported by Steely (1994). Other applicationsusing static headspace techniques are reportedeg, by Nedjma and Maujean (1995) in the anal-ysis of sulphur volatiles in water-alcohol solu-tions and brandies.

The atomic emission detector (AED)

The atomic emission detector (AEO) is a newdetector derived from the well-known atomicemission spectroscopy, used for the analysis ofmetals and metalloids. lt can be tuned to detectany element in any compound that may be elutedby Ge. The evaluation of the various detectionsystems for the determination of volatile sul-phur compounds in foods was reviewed by Mis-try et al (1994). The study of dynamic range,minimum detectable level and selectivity for aseries of sulphur compounds showed that: theupper limit of the linear dynamic range for theAEO is six to eight times greater than for theFPO or the SCO (ie, 1200-1550 ng of the com-pounds injected for the AEO vs 200 ng for theFPO and SCO); the minimum detectable level ofthe sulphur compounds with the AEO is as lowas 1 pg (vs 200 pg with the SCO and FPO); theselectivity of the AEO can be enhanced usingthe 'snapshot' option. A snapshot is a selectedsegment of the emission spectrum showing spe-cific elemental emission wavelengths.

The nitrogen phosphorus detecter (NPD)

The nitrogen phosphorus detector (NPO) ishighly selective for nitrogen-containing com-pounds (amines, pyrazines, pyridines, etc). Ithas similar operating conditions to the FlO detec-tor, and is also of low cost. GC-NPO systemshave been used in sorne investigations of flavourcomponents. Nitrogen-containing volatiles ofSwiss Gruyère chee se have been determined byLiardon et al (1982). Flavour defects in milkand dairy products due to pyrazines have beenreported by Morgan (1976). Lund (1994) alsoused this detector for the determination of2-methoxy-3-alkylpyrazines, eg, in carrots.

The 'sniffing device'

The 'sniffing deviee' uses the human nose as adetector. lt plays a key role as a unique inter-face with sensory analyses due to its specificityfor f1avour and off-f1avour compounds. Sub-stances are sniffed individually as they are elutedfrom the GC column. Oescriptors may beattributed to each retention time correspondingto an odour-active component (Sâvenhed et al,1985). Quantitative approaches were developedby Aeree et al (1984) with the so-called CHARM(combined hedonic aroma response measure-ment). The analysis is based on sniffing of GCruns of seriai dilutions of an odour essence, inorder to determine the presence of odour-activezones. The sniffer points out the start and theend of each particular odour perception, and des-ignates it with sensory descriptors. A CHARMvalue is calculated according to the formulac = d /1-1 where Il represents the number of coin-cident responses and d the seriai dilution factor.UlIrich and Grosch (1987) worked with a simi-lar technique, aroma extract dilution analysis(AEOA), based on the dilution factor (OF),which represents the highest dilution where anodour-active component may still be detected.The resulting values are proportional to the odournumber first defined by Rothe and Thomas(1963) as the 'aroma value' (ratio of the con-centration of an odour-active compound in thematrix to the detection threshold of this com-


pound in the same matrix). Miranda-Lopez etal (1992) have developed a quantitative tech-nique called gas chromatography-olfactometry(Osme or Ge-a), based on the quantitative mea-surement of the perceived odour intensity ofextracted components after separation on a GCcolumn. Compared to the CHARM or AEDAtests (Etiévant et al, 1994), this method is notbased on odour detection thresholds but on odourintensity. Guichard et al (1995) compared thislatter technique with both former ones and foundsimilar results. Moio et al (1994) used it for raw,pasteurised and UHT bovine milk. The power-fui odorants in bovine, ovine, caprine and waterbuffalo milk had been previously determinedby Moio et al (l993b) using GC-O. Guichard(1994) studied the key flavour compounds inAppellation d'Origine Contrôlée (AOC) cheeseby a GC-sniffing technique.

Table III. Multiple detection of volatile compounds.Détection multiple des composés volatils.

25

Multiple detection of volatile (aroma)compounds

Multiple detection of volatile (aroma) corn-pounds has also been proposed by sorne authors(table III). By using several detectors sirnulta-neously, identification is made more reliable.Detectors may be configured in series or in par-aile!. Common destructive detectors such as AD,NPD, FPD and MSD must be mounted in par-allel, or only as the final detector when config-ured in series. The FTIR detector, being non-destructive, is coupled in series with MSD insorne commercial systems (GC-FTIR-MSD).The software then allows simultaneous record-ing of the infrared and mass spectra of the elut-ing compounds. The future in this respectdefinitely belongs to the spectroscopie detec-tors that allow selective recognition of the

Simultaneous detection Investigated product Re!

FIDINPD Nitrogen volatiles in Swiss Gruyère cheese Liardon et al (1982)

Idem Sulfur volatiles in Limburger cheese Parliment et al (1982)

FIDIMSD VOC in office environment Mogl et al (1995)

Idem Fatty acid methyl esters Traitler and Horman (1990)

FIDfMSD Sulfur compounds in milk Bosset et al (1996)FPDfMSD

FTIRfMSD Flavour compounds in wine Buchbauer et al (1994)

FIDfsniffing Aroma of aged Cheddar cheese Christensen and Reineccius(1995)

Idem Sulfur volatiles in Cucumis melo Wyllie et al (1994)

MSDfsniffing Cheddar cheese Arora et al (1995)

Idem Blue crab meat volatiles Chung and Cadwallader (1994)

FID/SCDfMSD Flavour compounds in whiskey MacNamara et al (1995)

FIO: flame ionisation detector; NPO: nitrogen phosphorous detector; MSO: mass selective detector; FPO: flarne photometriedetector; FrIR: Fourier transform infrared spectroscopy detector; SCO: sulphur chemiluminescence detector; VOC: volatileorganic compounds.


chromatographed compounds. Today, hyphen-ated techniques such as GC-MSD, GC-FTIRand GC-AED are the most powerful techniquesavailable. Simultaneous detection allows theidentification and quantification of almost ailaroma compounds present in sufficient amountsafter column effluent splitting.

Quantitative analysis

Two main routes have been described for thequantification of volatile compounds. The firsttechnique is based on spiking the matrix withlabelled compounds whose extraction rate andloss during sample treatment, response factorduring detection, etc, are equal to those of the(unlabelled) substances to be quantified.Schieberle and Grosch (1987) described thismethod, called 'isotope dilution assay', for theanalysis of aroma compounds. The quantificationof a compound in the original matrix is basedon the peak height-ratio (or area-ratio) of theunlabelled to the labelled compound (as an inter-nai standard) using the MS detector in the SIMor SIS mode. Thirteen compounds were quanti-fied by this method in Emmentaler cheese(Grosch et al, 1994; Preininger and Grosch,1994; Preininger et al, 1996).

The second route uses a corn mon standardaddition method with increasing quantities ofthe constituents to be quantified (Imhof and Bos-set, 1994b). Thirty-three VOC were quanti ta-tively determined in this way in pasteurised milkand yoghurts in conjunction with dynamicheadspace analysis and MS detection.

DATA COLLECTION AND PROCESSING,STA TISTICAL ANAL YSIS

Having acquired the mass spectra using elec-tron-impact ionisation (El), the interpretation ofthe data stored in the computer can begin. In thefull scan mode, the mass spectra contain peakintensities on a mass to charge scale (mlz). Chro-matograms from selected ion recordings (SIM or

SIS mode according to the type of equipmentused) and data from multichannel analysers canalso be handled. Various sophisticated softwarepackages exist for the manipulation of data, dis-play of spectra and chemical structure plots asweil as for peak integration to obtain quantitativeinformation. The main goal is the elucidation ofthe chemical structure of unknown compoundsfrom their mass spectra. The fastest approachfor extracting structural information from spec-tra is library search combined with the spectro-scopist's expertise.

Dedicated software for the coding and search-ing of the relevant spectra exists in most modemGC-MS systems: a matching procedure com-pares unknown mass spectra and their intensities(mlz values) with those of the reference library.For ail useful matches, sorne matching factorsare calculated and a ranked Iist of the 'n' bestmatches ('hit list') is presented, sometimes evenas spectra together with their correspondingchemical structures (fig 3).

The following MS libraries are frequentlyused (nonexhaustive list): commerciallibrariesof electron-impact ionisation spectra Iike Wileywith over 275 000 spectra, National Institute ofStandards and Technology (NIST) with '" 63 000spectra, the Central Institute for Nutrition andFood Research, TNO, Zeist (The Netherlands)library with '" 1 600 spectra (especially fromfood volatiles), and the Chemical Concept (Ger-many) library with '" 35 000 spectra; userlibraries of certified spectra; user libraries ofunknown or uncertified spectra.

The quality of sorne of the data contained insorne databases is questionable: spectra are oftenstrongly curtailed as to the number of peaks andhence are incomplete, and incorrect spectra takenfrom old collections may be reincluded in newcompilations. Furthermore, threshold spectrafrequently miss small but significant peaks.

Among the most powerful search algorithmscurrently in use, SISCOM (Search for Identicaland Similar COMpounds; Damen et al, 1978;Domokos et al, 1983), PBM (Probability BasedMatching System; Pesyna et al, 1976; Atwater etal, 1985) and INCOSTM type searches could be


1 1,1 1'/1 ~I, Xl,

it! ! [ 1 1 !

1 ~I, >s,l,II

....

l, l,l, !! ! r! ! 1 136 Hl1

1 1" 1111 Il Il 1 Il 1.111ni, , , î 136 H2

1 1" UI II Il ' Il 1,11, III, ! ! ~ 136 H3

1 l, UI ,! 1 ,t r Il,, J r, I~ 136 H41

1 1" l,II 1 Il Il. 1.11, III l,~I 136 H5

1 ,50 100

Fig 3. Reference spectra of an unknown compound (x), identified as œ-pinene (H 1). The other spectra (H2-H5)correspond to other structures (best matches) found in different libraries.Spectres de référence d'un composé inconnu (x) identifié comme étant le a-pinène (HI). Les autres spectres(H2-H5) correspondent allX structures les plus voisines trouvées dans les différentes banques de réference.

mentioned. The SISCOM approach, featuringmultiple rankings and also a neutralloss search,often allows the identification of the structuralclass of compounds, even if no reference spec-trum is available. Both spectra and structuresmay be searched within MassLib®, the softwarefeaturing SISCOM. It has become possible toidentify not only similar spectra, but also toretrieve spectra of similar structures, revealingmuch of the complex relationship between spec-tra and structures (Henneberg et al, 1993). Thestructure search procedure also finds poor qual-ity spectra: this is often a considerable advan-tage for the expert trying to interpret the spec-trum of an unknown compound.

The use of statistics is also helpful for theidentification of structural information. Besides

structure statistics on 'hit lists', Massl.ibê offersprincipal compone nt analysis or PCA (Varmuzaet al, 1989). For a group of spectra - forinstance the members of a hit Iist - a number ofIl spectral features is chosen to represent thespectra as points in an n-dirnensional featurespace. The task is then to investigate the rela-tive position of the objects in this n-dimensionalspace, eg, to find clusters of spectra which lieclose together or to determine which spectra arethe nearest neighbours of an unknown spectrum.If such clusters of references can be found bythe proper selection of features, the position of anobject, most often the unknown spectrum, can beused (after a PCA has substantially reduced thenumber of dimensions) for predictions about theclass to which the object belongs. The pointers


obtained using the various methods finally makepossible a stepwise approach to structural ele-ments or partial structures, hopefully even theunknown structure itself (figs 3, 4).

Massl.ib'ê also offers a second tool for theautomatic identification of compounds in rou-tine work: the retention index option. Userlibraries may additionally contain the retenti on

2

o oo

indices for an apolar, polar and an user-definedcolumn. A calibration file containing index val-ues and their corresponding retenti on timesmakes possible the calibration of chro-matograms. The two-dimensional approach over-cornes the problems associated with substanceshaving very similar mass spectra, eg, for ter-penes (fig 5).

o

o

o

1

Fig 4. Principal compone nt analysis showing the position of library spectra (0) and thus their relative distancesto the unknown (+). The black lozenge corresponds to the unknown compound. The libraries may contain morethan one spectrum of a particular compound. Each square (with or without a line segment) represents a possibleche mical structure within a cluster. The distance between the black lozenge and a given square is a measure ofthe similarity between the se structures. For better readability of this figure, only the squares with a line seg-ment show the corresponding chemical structure.Analyse en composantes principales montrant la position des spectres de masse des banques de référence (0) etleur distance relative au spectre inconnu (.). Le losange noir correspond au composé inconnu. Les banques deréférence peuvent comporter plusieurs spectres d'un composé donné. Chaque carré (segmenté ou non) représenteune structure chimique possible à l'intérieur d'un «cluster». La distance entre le losange noir et un carré donnéest une mesure de la similitude existant entre ces structures. Pour améliorer la lisibilité de la figure, seuls les car-rés segmentés indiquent la structure chimique correspondante.


'I"' 'l'" '1" "1" , , 1" " l' '" l' " , l' " 'l' " , 1" " 1" " 1" " l' '" l' " 'l' "'1 "'1"" 1" "1" " 1, , 35~0'0' , , 36~00' , , 37~O'O' '" 38~0'O' , , 39~00' , , 401:0'0

...cg0z

*10 =:N.,.,V

llll"lljlll

, 34 ~00

29

5696

5918

Fig 5. Combined search procedure (total ion current, TIC, and retention index) to identify terpenes.The presence of several possible names for a given mass spectrum and a given retention index indicates that it couldbe useful to use another stationary phase to identify the unknown compound (abscissa: retention time, min).Procédure de recherche combinée (courant ionique total et indice de rétention) pour l'identification de ter-pènes. L'existence de plusieurs noms possibles pour un spectre de masse et un indice de rétention donnés indiquequ'il pourrait être utile de recourir à d'autres phases stationnaires pour identifier le composé inconnu (abscisse .-temps de rétention, min).

ELECTRONIC OR ARTIFICIAL NOSES

The idea of an instrument for the analysis ofvolatile (aroma) compounds that mimics thehuman nose is not new. The goal has been toobtain a relatively quick response, as compara-ble as possible to that of the human nose. How-ever, technological improvements have onlyrecently provided 'sniffing' deviees called 'elec-tronic' or 'artificial' noses which c1aim to satisfysuch requirements.

The principle of the functioning of such anose is c1aimed to he similar to that of the humannose: several reactions take place between one ormore volatile (flavour) components and an arrayof sensors (from four up to 32 according to thetype of electronic nose), Their relatively rapidreaction involves the adsorptionldesorption ofvolatiles on the surface of the sensors. Thesereactions are rneasured, and their resulls areexpressed numericaIly. With the help of dedi-cated software, the numbers are then submittedto data analysis or signal processing and pre-sented graphicaIly.

Several types of gas sensors can be used forthis task, Because of the specificity, selectivityand sensitivity of these sensors (more than 60different polymers are available today), a finalpattern ('fingerprint') of the sample can beobtained, The main types of sensors which havebeen developed for such analyses are thefollowing:

- conducting organic polymers, such as polypyr-role and polyaniline, working at relatively lowtempe ratures (usually < 80 "C), They requirevery little energy power. Il is possible to pre-pare polymers which are highly specifie for par-ticular classes of volatiles;

- semiconductors such as metal oxides (Sn02,ZnO, W03) working at high temperatures (upto 650 "C), which avoids interference from waterand improves the response and regenerationtime, In both cases, it is possible to measure thechange in the electrical resistance (or conduc-tance) during the reactions (fig 6) of the sensorswith the volatile substances (Zannoni, 1995);


800-

Ê:2. 60.0-CIloc1'1 40.0-(/)

iiiCIl

CI: 20.0-

TETE DE MOINE

o.o-·~<::;:~-~..,--,--+Lç-'--r---..--.----io 20 40 60 80 1llJ 20 40 60 80 100

Time (s)

Fi~ 6. Change in resistance using an electronic nose(FOX 3000, Alpha MOS) with 12 sensors by analysingvapours (volatiles) of Tête de Moine and SwissGruyère cheese. The vapours are injected over the 12-sensor array for 60 s before switching off the auto-matie injection valve. The response of six sensors isplotted: absorption for 60 s, desorption for 60 s.Variation de la résistance d'un nez électronique (FOX3000, Alpha MOS) à 12 capteurs lors de l'analyse deseffluves defromages de type Tête de Moine et Gruyèresuisse. Les effluves sont injectées dans 1II1 réseau de 12capteurs pendant 60 s avant la fermeture automatiquede la valve d'injection. La réponse de six capteurs estreprésentée graphiquement: 60 s d'adsorption, 60 s dedésorption.

- quartz resonators, which consist of a piezo-electric quartz oscillator coated with a sensingmembrane. The adsorption of volatile (flavour)components results in a mass change that canbe measured as a shi ft of the initial frequency(Zannoni, 1995);

- surface acoustic wave (SA W), amperometricand metal oxide semiconductor field effecl tran-sistor (MOSFET) deviees can also be used asgas sensors for this purpose (Zannoni, 1995).

With conducting organic polymers, thechange in conductance can be caused by twodifferent mechanisms: direct interaction betweenthe gas phase molecules (volatiles) and thepolarons, which have n cationic sites, and bipo-larons with 2n cationic sites in the pol ymer;interaction between the vapour molecules andthe anionic counterions (primary dopants) andsolvent matrix.

It is likely that both processes occur simul-taneously in the pol ymer. The interactionbetween the anal yte and the polymer (secondarydoping) is rapid and reversible, The analytemolecules only interact with the surface layers ofthe polymer, which explains the relatively shortrecovery time after an exposure.

ln porous metal oxide sensors (MOS),chemisorbed oxygen irreversibly oxidises volatile(flavour) compounds, a process which producesone or more conducting electrons. The free elec-Irons tend to reduce the potential barrier betweenoxide grains, increasing the electron mobilityand overall electrical conductivity of the material.This reaction depends on various physical andchemical parameters such as the microstructureof the sensor film and the operating temperatureof the oxide (typically 250-650 "C), The mate-rial shows a sensitivity at the sub-ppm level forreducing combustible gases (Pearce et al, 1993;Tan et al, 1995). The mechanism can be sum-marised by the following reaction: [R] + [0-]-> [RO] + lel (Tan et al, 1995).

After data generation and collection, the finalstage of data processing uses pattern recogni-tion algorithms, such as discriminant functions,K-nearest neighbour, template matching, c1usteranalysis, and partial least squares, in an artifi-cial neural network (ANN) to distinguishbetween similar volatiles, The system is self-training; the more data are presented, the morediscriminating the system becomes. These datasets are processed in a manner similar to that ofthe brain. The ANN's processing elements, ornodes, are akin to the neurones in the brain.Learning is accomplished by varying the empha-sis placed on the output of one sensor in rela-tion to another. It is derived from the mathe-matical or 'Euclidean' distance between datasets, Pattern recognition involves training andprediction. In the training phase, patterns fromthe measurements of standard reference sam-pies are used to develop mathematical rules forassigning them to their respective classes (fig 7).In the prediction phase, the rules evolved areapplied to assign unknown data sets to respectiveclasses.


+12 r----,----,.-----,---,APPENZELL --+~

Q~~""~");ru.srr

CD ~GJx; TETE DE MOINE r

GRUYERE

+6

-6 -

-12 '-----:':- -'- -:':- J

-to +'0a

Fig 7. Clusters obtained by discriminant factorial anal-ysis of five Swiss cheese varieties by using a 12-sen-sor electronic nose (FOX 3000, Alpha MaS). Refer tofigure 6 for operating conditions.Clusters obtenus par analyse factorielle discriminantede cinq sortes de fromages suisses en utilisant 1111 «nezélectronique» à 12 capteurs (FOX 3000, Alpha MaS).Voir figure 6 pour les conditions opératoires.

OTHER INSTRUMENTAL METHODS

In addition to the methods already described inthis article for qualitative and quantitative anal-ysis of volatile (tlavour) compounds, photo me-try should be also mentioned. This techniquemay be applied to sorne classes of compoundsbearing certain functional groups, such as car-bonyls, carboxyls and alcohols, which can bederivatised by introducing specifie chromophoreswith absorption in the UV or visible region(Imhof and Bosset, 1989). These methods havebeen used especially for quantifying carbonylcompounds (methyl ketones and aldehydes) withreagents such as 2,4 dinitrophenylhydrazine(DNPH) to form the corresponding colouredphenylhydrazones. Today, however, photomet-rie methods are less frequently used, modernGC methods being much more sensitive, spe-cific, reproducible and therefore very trustwor-thy. The former methods, however, can be use-fui in the exploratory stages and again later asconfirmatory tests. Furthermore, photometriemethods require frequent calibration and theelimination of background noise using a blank.In addition, the absorption bands used corre-spond more to chemical 'functions' (monocar-

bonyl, ketoglycerides, etc) than to particularsubstances. As a result, the photometricallydetermined absorbance corresponds to theabsorbance of a mixture of compounds, eventhough each component has in fact its own spe-cific molecular absorbance.

Several other methods have also been pro-posed for the analysis of volatile (tlavour) com-pounds, su ch as high pressure liquid chro-matography (HPLC) (Kubeczka, 1981; Daftlonet al, 1995), or NMR (Formâcek and Kubeczka,1982) which cannot be treated in the presentarticle as they would be out of context.

CONCLUSION

This review updates the possibilities and limi-tations of modern instrumental analysis ofvolatile (tlavour) compounds in milk and dairyproducts and describes sorne of the difficultiesencountered in carrying it out. The harder it is toovercome an analytical challenge, the more cre-ative the solution must be.

In general, GC is the method universallyapplied with the help of several highly efficientmodes of extraction, (pre)concentration, injec-tion, separation, and qualitative or quantitativedetection of the volatile (tlavour) compounds.Ali these techniques, however, have a corn monmajor drawback: the volatiles analysed are notentrapped or embedded in their original matrix,but have been converted into an extract. Con-sequently, they are present at concentrations thatdiverge greatly from those in the original foodsample. Furthermore, the rheological propertiesof the matrix itself, which play a key role intlavour perception, are completely lost by thisstrategy.

In spite of the remarkable progress and devel-opment in this field, there is no ideal universalmethod for the simultaneous analysis of ailvolatile (flavour) compounds. Each methodbriefly described in this review presents bothadvantages and disadvantages (eg, table IV).This probably explains the numerous methodscurrently available. They may be easy and time-

32R

Mariaca,

JOB

osset


saving, or complicated and tedious. Sorne ofthem are specifie for certain compounds, andsuitable for routine assays, while others are moreappropriate for research. The numerous meth-ods and techniques proposed are complemen-tary, but no one in particular seems to predom-inate today.

The other instrumental approach recentlyproposed is the so-called 'electronic' or 'artifi-cial' nose, which allows one to investigatevolatile (flavour) compounds directly in theiroriginal food matrix. Consequently, the elec-tronic nose represents real progress. Such ananalysis is very rapid compared to, for exam-pIe, a GC analysis with its series of steps. How-ever, in the perception of volatile flavours, anelectronic nose, equipped with 4, 5, 6,12, 18 or32 sensors, cannot compete with the thousandsof sensitive nerve endings active in the humannose and retronasal system.

Finally, in both strategies, ie, the sequentialapproach (GC analysis) and the direct globalapproach (assays with an electronic nose), highlyefficient software is essential for the statisticaltreatment of data such as univariate and multi-variate analyses.

To conclude, on the one hand, even the mostefficient hardware and software currently avail-able cannot compare with the extraordinarycapacities of the human brain to analyse flavourstimuli with the help of memory in a context oflong-term cultural and social education. On theother hand, it is weil known that odour-detectingcapacities vary greatly among individuals. Theresult of smelling can be influenced by the sniff-ing technique, the type of odorant mixtures, theadaptation to the environ ment, and the age andthe psychophysical conditions of the subject.Many volatile molecules are barely detectableby the human nose. In addition, the response ofour senses are difficult to quantify, even whenusing standard calibration procedures. For ailthe se reasons, sensory analysis can be muchmore difficult and costly to apply in compari-son with instrumental analysis, but, as far asflavour is concerned, the human nose is found tobe the ultimate arbiter.

33

ACKNOWLEDGMENTS

The authors are grateful to T Berger (FAM,Liebefeld-Bern, Switzerland), F Friedli (MSPFriedli & Co, Kôniz, Switzerland), S Gregory(Sensory Science, NZ Dairy Research Institute,Palmerston North, New Zealand), L Hunt (Pala-tine, IL, 60067, USA), H Nursten (Dept FoodSei Technol, University of Reading, Reading,UK), V Raverdino (Hewlett-Packard SA,Meyrin, Switzerland), N Skinner (Centre suissed'électronique et de microtechnique, Neuchâ-tel, Switzerland), G Urbach and C Wijesundera(CSIRO Div, Food Sei Technol, Melbourne Lab-oratory, Highett, VIC, Australia) for carefulreviewing of this article and valuable linguisticassistance, as weil as to R Gauch (FAM, Liebe-feld-Bern, Switzerland) for technical assistance.

REFERENCES

Aeree TE, Barnard J, Cunningham DG (1984) A pro-cedure for the sensory analysis of gas chromato-graphie effluents. Food Chem 14,273-286

Adda J, Dumont JP (1972) Etude des différentes mé-thodes d'extraction d'arôme de fromage. /nd Ali-ment Agric 89, 143-145

Alm L (1982) Effect of fermentation on volatile acidsand ethanol in Swedish dairy products. J DairySei 65, 186-190

Arora G, Cormier F, Lee B (1995) Analysis of odor-active volatiles in Cheddar chee se headspace bymultidimensional GC/MS/sniffing. J Agric FoodChem 43, 748-752

Aston JW, Douglas K (1981) Detection of carbonylsulphide in Cheddar cheese headspace. J DairyRes 48, 473-478

Atwater BL, Stauffer DB, McLafferty FW, PetersonDW (1985) Reliability ranking and scalingimprovements to the probability based matchingsystem for unknown mass spectra. Anal Chem57,899-903

Badertscher R, Liniger A, Steiger G (1993) Bestim-mung der flüchtigen Fettsâuren in Kâse aus demWasserdampfdestillat mit 'Headspace-GCIFID'.FAM-/nfo, June 1993, 272w, 1-9

Barbieri G, Bolzoni L, Careri M, Mangia A, ParolariG, Spagnoli S, Virgili R (1994) Study of the


volatile fraction of Parmesan cheese. J Agric FoodChem 42,1170-1176

Bemelmans JMH (1979) Review of isolation and con-centration techniques. In: Progress in FlavourResearclr(Land DG, Nursten HE, eds) Appl SciPubl, London, 79-98

Benkler KF, Reineccius,GA (1980) Flavor isolationfrom fatty foods via solvent extraction and mem-brane dialysis, J Food Sei 45, 1084-1085

Bicchi C, Artuffo G, D'Amato A, Galli A, Galli M(1992) Cyclodextrin derivatives in the GC sepa-ration of racemic mixtures of volatile compounds.Part IV. J High Resol Chromatogr 15, 655-658

Blanch GP, Tabera J, Herraiz M, Reglero G (1993)Preconcentration of volatile components of foods:optimization of the steam distillation-solventextraction at normal pressure. J Chromatogr 628,261-268

Bosset JO, Gauch R (1993) Comparison of the volatilellavour compounds of six European 'AOC' cheesesby using a new dynamic headspace GC-MSmethod. lnt Dairy J 3,359-377

Bosset JO, Bühler-Moor U, Eberhard P, Gauch R,Lavanchy P, Sieber R (1994) Dimethylsulfid- undDimethyldisulfidgehait sowie Aroma ais Indika-toren für die thermische Behandlung von Milch.Schwei: Milchwirtsch Forsch 23, 67-71

Bosset JO, Gauch R, Mariaca R, Klein B (1995) Com-parison of various sample treatments for the anal-ysis of volatile compounds by GC-MS: applicationto the Swiss Emmental cheese. Mitt Geb Lebens-mittelunters Hyg 86, 672-698

Bosset JO, Eberhard P, Gallmann P, Gauch R, Rat-tray W, Sieber R (1996) Occurrence and behaviourof volatile sulfur-containing compounds in milkby heat treatment. In: Heat Treatments and A lter-native Methods. Proc IDt Symp, Vienna, Austria,6-8 Sept 1995, 409-421

Bradley RL Jr, Stine CM (1968) Agas chromato-graphie study of the volatile flavors in natural andspray-dried Cheddar cheese. J Gas Chromatogr6,344-348

Brinkman UAT, Vreuls RH (1995) Solid-phase extrac-tion for on-line sample treatment in capillary gaschromatography. LC-GC lnt 8, 694-698

Buchbauer G, Jirovetz L, Wasicky M, Herlitschka A,Nikiforov A (1994) Aroma von Weissweinblüten:Korrelation sensorischer Daten mit Headspace-Inhaltstoffen. ZLebensm Unters Forsch 199, 1-4

Burdge JR, Farwell sa (1994) Comparison of com-bustion sources for sulfur chemiluminescencedetection. J High Resol Chromatogr 17,22-24

Burdge JR, Cai Z, Papillon J, Farwell S (1994) A novelfocusing injection technique for chemilumines-cent detection of volatile sulfur compounds sepa-rated by HRGe. J High Resol Chromatogr 17,695-699

Careri M, Manini P, Spagnoli S, Barbieri G, BolzoniL (1994) Simultaneous distillation-extraction anddynamic headspace methods in the gas chromato-graphie analysis of Parmesan cheese volatiles,Chromatographia 38, 386-394

Casabianca H, Graff JB, Jarne P, Perrucchietti C (1995)Application of hyphenated techniques to the chro-matographic authentication of flavors in food prod-ucts and perfumes. J High Resol Chromatogr 18,279-285

Chang YI, Reineccius GA (1985) Isolation of llavorcompounds in model systems by countercurrentcontinuous dialysis. J Agric Food Chem 33,1168-1173

Christensen KR, Reineccius GA (1995) Aroma extractdilution analysis of aged Cheddar cheese. J FoodSei 60, 218-220

Chung HY, Cadwallader KR (1994) Aroma extractdilution analysis of blue crab meat volatiles.J Agric Food Chem 42,2867-2870

Contarini G, Leardi R (1994) Use of experimentaldesign to optimize the analysis of volatile com-pounds by dynamic headspace extraction followedby cold trapping and capillary Ge. J High ResolChromatogr 17, 91-94

Contarini G, Toppino PM (1995) Lipolysis in Gor-gonzola cheese during ripening. Int Dairy J 5,141-155

Coulibaly K, Jeon IJ (1992) Sol id-phase extraction ofless volatile flavor compounds from uitra-hightemperature processed milk. J Agric Food Chem40,612-616

Cronin DA (1982) Techniques of analysis of llavours.ln: Food Flavours. Part A. Introduction (MortonID, MacLeod Al, eds) Elsevier, Amsterdam, 15-48

Dafflon 0, Gobet H, Koch H, Bosset JO (1995) Ledosage des hydrocarbures aromatiques polycy-cliques dans le poisson, les produits carnés et lefromage par chromatographie liquide à haute per-formance. Trav Chim Aliment Hyg 86, 534-555

Damen H, Henneberg D, Weimann B (1978) Siscom- a new library search system for mass spectra.Anal Chim Acta 103,289-302

Day EA (1967) Cheese flavor. In: Symposium onFoods: The Chemistry and Physiology of Flavors(Schultz HW, Day EA, Libbey LM, eds) Avi Publ,Westport, CT, 331-361


Day EA, Libbey LM (1964) Cheddar chee se flavor:gas chromatographie mass spectral analyses of theneutral components of the aroma fraction. J FoodSci 29, 583-589

de Frutos M, Sanz J, Martinez-Castro 1 (1988) Simul-taneous distillation--extraction (SDE) method inthe qualitative and quantitative GC analysis ofcheese volatile components. Chromatographia 25,861-864

de Frutos M, Sanz J, Martinez-Castro 1 (1991) Char-acterization of artisanal cheeses by GC and GClMSanalysis of their medium volatility (SDE) fraction.J Agric Food Chem 39, 524-530

Degorce-Dumas R, Goursaud J, Leveau JY (1986)Analyse de composés volatils du yaourt par chro-matographie en phase gazeuse--espace de tête(headspace). Ind Aliment Agric 103,805-808

Delahunty CM, Piggott JR, Conner JM, Paterson A(1994) Low-fat Cheddar cheese flavour: flavourrelease in the mouth. In: Trends in FlavourResearch. (Maarse H, van der Heij DG, eds) Else-vier, Amsterdam, 47-52

Domokos L, Henneberg D, Weimann B (1983) Opti-mization of search algorithms for a mass spectralibrary. Anal ChimActa 150,37-44

Dumont JP, Adda J (1972) Isolement des constituantsde l'arôme des fromages : comparaison de mé-thodes. Lait 52, 311-323

Dumont JP, Roger S, Adda J (1974) Composés volatilsdu fromage entier et du fromage râpé: exempledu Parmesan. Lait 54,386-396

Etiévant PX, Moio L, Guichard E, Langlois D,Leschaeve l, Schlich P, Chambellant E (1994)Aroma extract dilution analysis (AEDA) and therepresentativeness of the odour of food extracts.ln: Trends in Flavour Research (Maarse H, vander Heij DG, eds) Elsevier, Amsterdam, 179-190

Evans EW (1972) Instrumental and analytical assess-ment of flavour. J Soc Dairy Technol25, 125-135

Ferretti A, Flanagan VP (1972) Steam volatile con-stituents of stale nonfat dry milk. The role of theMaillard reaction in staling. J Agric Food Chem20,695-698

Forrnâcek V, Kubeczka KH (1982) Essential ailsAnalysis by Capillary Gas Chromatography andCarbon-/3NMR Spectroscopy. John Wiley andSons, Bristol

Forss DA (1969) Flavors of dairy products: a reviewof recent advances. J Dairy Sei 52, 832-840

Forss DA (1971) The flavors of dairy fats - a review.J Am ail Chem Soc 48, 702-710

35

Forss DA (1972) Odor and flavor compounds fromlipids, In: Progress in the Chemistry of Fats andOther Lipids, vol XIII, part 4 (Holman RT, ed)Perga mon Press, Oxford, 177-258

Forss DA, Jacobsen VM, Ramshaw EH (1967) Con-centration of volatile compounds from dilute aque-ous solutions. J Agric Food Chem 15, 1104-1107

Gaafar AM (1992) Volatile flavour compounds ofyoghurt.lnt J Food Sei Technol Tl . 87-91

Gallois A, Langlois D (1990) New results in thevolatile odorous compounds of French cheeses.Lait 70, 89-106

Georgala AIK, Tsakalidou E, Kandarakis l, Kalant-zopoulos G (1995) Flavour production in ewesmilk and ewe's milk yoghurt, by single strains andcombinations of Streptococcus thermophilus andLactobacillus delbrueckii subsp bulgaricus, iso-lated from traditional Greek yoghurt. Lait 75,271-283

Gmür W, Bosset JO, Plattner E (1986) Lëslichkeitvon einigen Kâseinhaltsstoffen in überkritischemKohlendioxid. Lebensm Wiss Technol 19,419-425

Gmür W, Bosset JO, Plattner E (l987a) Beitrag zurdirekten Kopplung Fluidextraktion-Kapillarfluid-chromatographie, 1. Theoretische Optimierungeiniger wichtiger apparativer Parameter. J Chro-matogr388,143-150

Gmür W, Bosset JO, Plattner E (l987b) Beitrag zurdirekten Kopplung Fluidextraktion-Kapillarfluid-chromatographie, II. Aufbau eines Prototyps undAnwendungsbeispiele. J Chromatogr 388,335-349

Gmür W, Bosset JO, Plattner E (l987c) Beitrag zurdirekten Kopplung Fluidextraktion-Kapillarfluid-chromatographie, III. Experimentelle Optimierungdes Druck- und Temperaturprogramms in der Flu-idchromatographie im Hinblick auf die Analysevon Milchprodukten. Mitt Geb LebensmitteluntersHyg 78, 21-35

Godefroot M, Sandra P, Verzele M (1981) Newmethod for quantitative essential oil analysis.J Chromatogr 203,325-335

Gomer F, Palo V, Bertan M (1968) Verânderungendes Gehaltes der flüchtigen Stoffe wâhrend derJoghurtreifung. Milchwissenschaft 23, 94-100

Grigoryeva DN, Golovnya RV, Syomina LA, Vasi-Iyev AV, Samusenko AL (1994) Correlationbetween sensory and gas chromatographie char-acteristics of Dutch cheese. In: Trends in FlavourResearch. (Maarse H, van der Heij DG, eds) Else-vier, Amsterdam, 283-286


Grob K, Karrer G, Riekkola ML (1985) On-columninjection of large sample volumes using the reten-tion gap technique in capillary gas chromatography.J Chromatogr 334, 129-155

Grosch W (1990) Analyse von Aromastoffen. ChemUnserer Zeit 24 (2), 82-89

Grosch W, Preininger M, Warmke R, Belitz HD(1994) Studies on the f1avour of Swiss cheese(Emmentaler).ln: Aroma Perception, Formation,Evaluation (Rothe M, Kruse HP, eds) DeutschesInst Ernâhrungsforsch, Postdam-Rehbrücke,425-439

Groux M, Moinas M (1974) La flaveur des fromages.II. Etude comparative de la fraction volatile neutrede divers fromages. Lait 54, 44-52

Guichard E (1994) Study of key f1avor compounds inAOC cheese. In: Flair Sens Projeet (MatforskFlair Sens Project, Management Team, eds)Âs-Trykk, As, Norway, 87-89

Guichard E, Berdagué JL, Grappin R, Fournier N(1987) Affinage et qualité du Gruyère de Comté. V.Influence de l'affinage sur la teneur en composésvolatils. Lait 67,319-338

Guichard H (1995) Contribution au développementlie méthodes instrumentales pour l'étude des frac-tions volatile et azotée soluble dans l'eau de fro-mages à pâte pressée cuite. Relation avec l'analysesensorielle. Thèse de doctorat, Univ Bourgogne,France

Guichard H, Guichard E, Langlois D, Issanchou S,Abboll N (1995) GC sniffing analysis: olfactiveintensity measurement by two methods. Z LebensmUnters Forseh 201, 344-350

Henneberg D, Weimann B, Zalfen U (1993) Com-puter-aided interpretation of mass spectra usingdatabases with spectra and structures. 1. Structuresearches. Org Mass Speetrom 28, 198-206

Hild J (1979) Quantitative Bestimmung einigerwichtiger Aromakomponenten in Milcherzeug-nissen rnittels Head-Space- Technik. Milehwis-sensehaft 34, 281-283

Horwood JF (1989) Headspace analysis of cheese.Aust J Dairy Teehnol44, 91-96

Imhof R, Bosset JO (1989) Einfache quantitative pho-tometrische Bestimmung von 'Gesamt' -Car-bonylkomponenten in biologischen Medien. MittGeb Lebensmittelunters Hyg 88, 409-419

Irnhof R, Bosset JO (1991) Comparison of two systemsfor sample preparation and injection by dynamicheadspace GC analysis. J High Resol Chromatogr14,621-625

Imhof R, Bosset JO (1994a) The performance of var-ious capillary columns for the analysis of volatilef1avor compounds in dairy products by dynamicheadspace gas chromatography. J High ResolChromatogr 17, 25-30

Imhof R, Bosset JO (I994b) Quantitative GC-MSanalysis of volatile f1avour compounds in pas-teurized milk and fermented milk products apply-ing a standard addition method. Lebensm WissTeehno127,265c269

Jackson P, Dent G, Carter D, Schofield DJ, ChalmersJM, Visser T, Vredenbregt M (1993) Investiga-tion of high sensitivity GC-FTIR as an analyticaltool for structural identification. J High Resol Chro-matogr 16, 515-521

Jain MK, Singhal OP (1969) Chemistry of gheef1avours. II. Variations in carbonylic tlavour com-pounds in ghee made from desi buller. Indian JDairy Sei 22, 103-108

Jennings WG, Filsoof M (1977) Comparison of sam-pIe preparation techniques for gas chromatographieanalysis. J Agrie Food Chem 25, 440-445

Jennings WG, Rapp A (1983) Sample PreparationforGas Chromatographie Analysis (Bertsch W, Jen-nings WG, Kaiser RE, eds) A Hütig Verlag,Heidelberg, 1-104

Jeon II, Thomas EL, Reineccius GA (1978) Productionof volatile f1avor compounds in ultra-high tem-perature processed milk during aseptic storage.J Agrie Food Chem 26,1183-1188

Kahlhofer H (1981) Aromakontrolle mit Hilfe derHead Space- Technik. Milehw Beriehte 68,197-202

Karlsson HOE, Trâgârdh G (1996) Application of per-vaporation in food processing. Trends Food SeiTeehnoI7,78-83

Kasalinsky VF, McDonald JT Jr (1983) GC/FTIRspectra of terpenes. J Chromatogr Sei 21, 193-200

Klein B, Gauch R, Bosset JO (1990) Comparison offour extraction, concentration and injection tech-niques for volatile compounds analysis by GC-MS:an application to the study of the volatile tlavour ofSwiss Emmentaler chee se. In: Flavour Scienceand Technology (Bessière Y, Thomas AF, eds)John Wiley and Sons, Chichester, 205-208

Kowalewska J, Zelazowska H, Babuchowski A, Ham-mond EG, G1atz BA, Ross F (1985) Isolation ofaroma-bearing material from Lactobacillus hel-vetieus culture and cheese. J Dairy Sei 68,2165-2171

Kubeczka KH (1981) Application of HPLC for theseparation of tlavour compounds. In: Flavour '81,


3rd Weurman Symp Proc Int Conf Munich, April28-30, 1981 (Schreier P, ed) W de Gruyter, Berlin,345-359

Lacrampe JL, Dherbomez M, Weber F (1975) Prob-lèmes d'extraction et d'analyse de certains com-posés aromatiques des produits laitiers. 1. Tech-niques de concentration et d'extraction. Lait 55,517-529

Lamparsky D, Klimes I (1981) Cheddar cheese flavour- its formation in the light of new analytical results.ln: Flavour '81. 3rd Weumum Symp Proc lnt ConfMunich. April 28-30. 1981 (Schreier P, ed) W deGruyter, Berlin. 557-577

Laye I, Karleskind D, Morr CV (1995) Dynamicheadspace analysis of aceelerated storage com-mercial whey protein coneentrate using four dif-ferent adsorbent traps. Milchwissenschaft 50,268-272

Leahy MM, Reineccius GA (1984) Comparison ofmethods for the isolation of volatile compoundsfrom aqueous model systems. In: Analysis ofVolatiles (Schreier P, ed) W de Gruyter, Berlin,19-48

Liardon R, Bosset JO, Blanc B (1982) The aromacomposition of Swiss Gruyère cheese. 1. The alka-line volatile components. Lebensm Wiss Technol15,143-147

Libbey LM, Bills DD, Day EA (1963) A techniquefor the study of lipid-soluble food flavor volatiles.J Food Sei 28,329-333

Liebich HM, Douglas DR, Bayer E, Zlatkis A (1970)The volatile flavor components of Cheddar cheese.J Chromatogr Sei 8, 355-359

Linssen J, Reitsma H, Cozijnsen J (1995) Staticheadspace gas chromatography of acetaldehyde inaqueous foods and polythene terephthalate.Z Lebensm Unters Forsch 201, 253-255

Lund ED (1994) Determination of 2-methoxy-3-alkylpyrazines in carrot products by gas chro-matography/nitrogen-phosphorus detection. JAOAC Int 77, 416-420

Maarse H (1983) Volatile Compounds in Food. Ser1: Qualitative Data. Ser 2: Quantitative Data. DivNutrition and Food Research TNO Inst CIVO -Analysis TNO, Zeist

Maarse H, Belz R (1981) Isolation, separation andidentification of volatile compounds in aromaresearch. In: Handbuch der Aromaforschung(Rothe M, ed) Akademie Verlag, Berlin, 1-290

MacNamara K, Brunerie P, Squarcia F, RozenblumA (1995) Investigation of flavour compounds inwhiskey spent lees.ln: Food Flavors: Generation.

37

Analysis and Process Influence. Proc 8th Int Fla-vor Conf Cos, Greeee, 6-8 July 1994 (Charalam-bous G, ed) Elsevier, Amsterdam, 1753-1766

Maignial L, Pibarot P, Bonetti G, Chaintreau A, Mar-ion JP (1992) Simultaneous distillation--extractionunder static vacuum: isolation of volatile com-pounds at room temperature. J Chromatogr 606,87-94

Majors RE (1995) Sample preparation perspectives.The use of membranes in sample preparation.LC-GC Int 8, 375-380

Manning DJ (1974) Sulphur compounds in relationto Cheddar cheese flavour. J Dairy Res 41, 81-87

Manning DJ, Moore C (1979) Headspace analysis ofhard cheeses. J Dairy Res 46, 539-545

Manning DJ, Priee JC (1983) Instrumental methodsin determining flavour. J Soc Dairy Technol 36,33-35

Marsili R (1996) Techniques for analyzing food aroma.ln: Food Sei Technol Serf19 (Marsili R, ed) Mar-cei Dekker Inc, New York and Basel, 1-400

McGugan WA, Howsam SG, Ellion JA, Emmons DB,Reiter B, Sharpe ME (1968) Neutral volatiles inCheddar cheese made aseptically with and withoutstarter culture. J Dairy Res 35,237-245

McNeal TP, Hollifield HC (1990) Quantitative mul-tiresidue analyses for volatile organics in waterand milk, using a fused silica open-tubular wide-bore capillary column and automated headspacegas chromatogrography. J Assoc Off Anal Chem73,328-331

Mick S, Mick W, Schreier P (1982) The compositionof neutral volatile constituents of sour cream but-ter. Milchwissenschaft 37,661-665

Miranda-Lopez R, Libbey LM, Watson BT, McDanielMR (1992) Odor analysis of Pinot Noir wines fromgrapes of different maturities by agas chro-matography-olfactometry technique (Os me).J Food Sei 57, 985-1019

Mistry BS, Reineccius GA, Jasper BL (1994) Com-parison of gas chromatographie detectors for theanalysis of volatile sulfur compounds in foods.ln: ACS Symp Ser 564. Sulfur Compounds in Foods(Mussinan CJ, Keelan ME, eds) Am Chem Soc,Washington DC, 8-21

Mogl S, Haas J, Knutti R (1995) Bestimmung vonVOC in Bürorâumen, Analytischer Erfahrungs-bericht über eine europâische Studie an 56 Büro-gebâuden. Mitt Geb Lebensmittelunters Hyg 86,657-671


Moinas M (1973) Arôme volatil de produits laitiers- extraction et identification. Trav Chim AlimentHyg 64, 59-65

Moinas M, Groux M, Horman 1 (1973) La Ilaveur desfromages. I. Une méthodologie nouvelle d'isole-ment de constituants volatils. Application auRoquefort et au Camembert. Lait 53, 601-609

Moio L, Dekimpe J, Etievant P, Addeo F (1993a) Neu-tral volatile compounds in the raw milks from dif-ferent species. J Dairy Res 60,199-213

Moio L, Langlois D, Etievant P, Addeo F (1993b)Powerful odorants in bovine, ovine, caprine andwater buffalo milk determined by means of gaschromatography-olfactometry. J Dairy Res 60,215-222

Moio L, Etievant P, Langlois D, Dekimpe J, AddeoF (1994) Detection ofpowerful odorants in heatedmilk by use of extract dilution sniffing analysis.J Dairy Res 61,385-394

Molimard P, Spinnler HE (1996) Review: compoundsinvolved in the Ilavor of surface mould-ripenedcheeses: origins and properties. J Dairy Sei 79,169-184

Monnet C, Schmitt P, Divies C (1994) Method forassaying volatile compounds by headspace gaschromatography and application to growing startercultures. J Dairy Sci 77, 1809-1815

Morgan ME (1976) The chemistry of sorne micro-bially-induced Ilavor defects in milk and dairyfoods. Biotechnol Bioeng 18, 953-965

Mosandl A (1995) Enantioselective capillary gas chro-matography and stable isotope ratio mass spec-trometry in the authenticity control of Ilavors andessential oils. Food Rev lnt 11,597-664

Namba R (1990) Industrial applications of GC/FT-IR.ln: Practical Fourier Transform lnfrared Spec-troscopy (Ferraro JR, Krishnan K, eds) AcademiePress, New York, 469-518

Nedjma M, Maujean A (1995) Improved chromato-graphie analysis of volatile sulfur compounds bythe static headspace technique on water-alcoholsolutions and brandies with chemiluminescencedetection. J Chromatogr A 704, 495-502

Ney KH (1973) Technik der Aromauntersuchung.Gordian 73, 380-387

Ney KH, Wirotama IPG (1978) Untersuchung derAromabestandteile von Fontina, einem italie-nischen Kâse, Feue Seifen Anstrichm 80, 249-251

Nickerson GB, Likens ST (1966) Gas chromatographieevidence for the occurrence of hop oil componentsin heer. J Chromatogr 21, 1-5

Nüüez AJ, Bemelmans JMH (1984) Isolation meth-ods for the volatile components of grapefruit juice.Distillation and solvent extraction methods.Chromatographia 18, 153-158

Nursten HE (1977) The important volatile compo-nents of foods. ln: Sensory Properties of Foods(Birch GG, Brennan JG, Parker KJ, eds) Appl SciPubl, London, 151-166

Nursten HE (1979) Why Ilavour research? How farhave we come since 1975 and where now? ln:Progress in Flavour Research (Land DG, NurstenHE, eds) Appl Sei Publ, London, 337-355

Nursten HE, Sheen MR (1974) Volatile Ilavour com-ponents of cooked potato. J Sei Food Agric 25,643-663

O'Hare WT, Nursten HE (1994) Prediction of late gasformation in commercially produced Cheddarcheese by headspacc measurements at 8 weeks.ln: Trends in Flavour Research. (Maarse H, vander Hcij DG, eds) Elsevier, Amsterdam, 151-155

Oria R, Condon S, Sala TF (1987) Headspace profilessampling and recording method for cheese.Milchwissenschaft 42,713-716

Paillard N, Pitoulis S, Mattei A (1970) Techniques depréparation et analyse de l'arôme de quelquesfruits. Lebensm Wiss TecJlI1ol3, 107-114

Parliment TH, Kolor MG, Rizzo DJ (1982) Volatilecomponents of Limburger cheese. J Agric FoodChem 30, 1006-1008

Pearce TC, Gardner JW, Friel S, Bartlett PN, Blair N(1993) An electronic nose for monitoring theIlavour of beers. Analyst 118,371-377

Pesyna GM, Venkataraghavan R, Dayringer HE,McLafferty FW (1976) Probability based matchingsystem using a large collection of reference rnassspectra. Anal Chem 48, 1362-1368

Pierce K (1995) Reflections on chirality. Anal EurApril. 31-32

Preininger M, Grosch W (1994) Evaluation of keyodorants of the neutral volatiles of Emmentalercheese by the calculation of odour activity values.Lebensm Wiss Technol27, 237-244

Preininger M, Rychlik M, Grosch W (1994) Pote ntodorants of the neutral volatile fraction of Swisscheese (Emmentaler). ln: Trends in FlavourResearch. (Maarse H, van der Heij DG, eds) Else-vier, Amsterdam, 267-270

Preininger M, Warmke R, Grosch W (1996) Identifi-cation of the character impact Ilavour compoundsof Swiss cheese by sensory studies of models.Z Lebensm Unters Forsch 202, 30-34


Priee JC. Manning Dl (1983) A new technique for theheadspace analysis of ha rd cheese. J Dairy Res50.381-385

Reineccius GA. Anandaraman S (1984) Analysis ofvolatile flavors. In: Food Constituents and FoodResidues (Lawrence JF. ed) M Dekker, Basel,195-293

Rothe M. Thomas B (1963) Aromastoffe des Brotes.Z Lebensm Unters Forsch 119.302-310

Sanz J, de Frutos M. Martinez-Castro 1 (1992) Designof a mixed-phase capillary column for the gas chro-matographie separation of the volatile componentsofcheese. Chromatographia 33. 213-217

Sâvenhed R. Borén H. Grimwall A (1985) Strippinganalysis and chromatographie sniffing for thesource identification of odorous compounds indrinking water. J Chromatogr 328. 219-231

Schieberle p. Grosch W (1987) Quantitative analysisof aroma compounds in wheat and rye bread crustsusing a stable isotope dilution assay. J Agric FoodChem 35. 252-257

Schreier p. Idstein H (1985) High-resolution gas chro-matography - Fourier transform infrared spec-troscopy in f1avour analysis: limits and perspec-tives. Z Lebensm Vnters Forsch 180. 183-188

Schwartz DP. Shamey J. Brewington CR. Parks OW(1968) Methods for the isolation and character-ization of constituents of natural products. X. Newand improved methods for the analysis of carbonyl2,4-dinitrophenylhydrazones and 2,4-dinitro-phenylosazones. Microchem J 13.407-417

Sedlûckovâ B. Kakalfkovâ L. Matisovà E. Farkas P.Dobos A (1995) Analysis of volatile compounds invarious wines. Anal Methods Instrum 2.106-108

Shiratsuchi H. Shimoda M. Imayoshi K. Noda K.Osajima Y (1994) Volatile flavor compounds inspray-dried skim milk powder. J Agric Food Chem42.984-988

Sloot D. Harkes PD (1975) Volatile trace componentsin Gouda cheese. J Agric Food Chem 23. 356-357

Stark W. Urbach G. Hamilton JS (1976) Volatile com-pounds in butter oil. V. The quantitative estimationof phenol. o-methoxyphcnol, m- and p-cresol.indole and skatole by cold-finger molecular dis-tillation. J Dairy Res 43. 479-489

Stark W. Urbach G. Cook U. Ashes JR (1978) Theeffect of diet on the y- and o-Iactone and methylketone potentials of caprine butterfat. J Dairy Res45.209-221

Steely JS (1994) Chemiluminescence detection of sul-fur compounds in cooked milk. In: ACS Symp Ser

39

564. Sulfur Compounds in Foods (Mussinan CJ.Keelan ME. eds) Am Chem Soc. Washington DC.22-35

Sugisawa H. Chen C. Nabeta K (1984) Simultaneousdistillation adsorption and its application. In: Anal-ysis of Volatiles (Schreier p. ed) W de Gruyter,Berlin. 357-369

Takacs F Jr (1989) Solid-phase extraction of stale fla-vor components from ultra-high-ternperature pro-cessed milk. MS thesis, Kansas State Univ, Man-hattan. KS

Tan T. Lucas Q. Moy L. Gardner J. Bartlett P (1995)The electronic nose - a new instrument for sensingvapours. LC GC Int 8. 218-225

Teranishi R. Issenberg p. Homstein 1. Wick EL (1971)Flavor Research, Principles and Techniques.M Dekker Inc, New York

Traitler H. Horman 1 (1990) GC-MS; Simultaneousquantification and identification by column efflu-ent splitting. J High Resol Chromatogr 13.197-198

Tuan HPT. Janssen HGM. Cramers CA. Kuiper-vanLoo EM. Vlap H (1995) Evaluation of the perfor-mance of various universal and selective detectorsfor sulfur determination in natural gas. J HighResol Chromatogr 18. 333-342

Ulberth F (1991) Headspace gas chromatographie esti-mation of sorne yogurt volatiles. J Assoc Off AnalChem 74. 630-634

Ulberth F. Roubicek D (1995) Monitoring of oxidativedeterioration of milk powder by headspace gaschromatogrography. Int Dairy J 5.523-531

Ullrich F. Grosch W (1987) Identification of the mostintense volatile f1avour compounds formed dur-ing autoxidation of linoleic acid. Z Lebensm UntersForsch 184.277-282

Urbach G (1982) The effect of different feeds on thelactone and methyl ketone precursors of milk fat-a review. Lebensm Wiss Technol t S, 62-67

Urbach G (1987) Dynamic headspace gas chro-matography of volatile compounds in milk.J Chromatogr 404. 163-174

Urbach G (1996) The chemical and biochemical basisof chee se and milk aroma. In: Microbiology andBiochemistry ofCheese and Fermented Milk (LawBA. ed) Blackie Acad & Prof; imprint of Chap-man & Hall. London. 253-298

Urbach G. Stark W. Forss DA (1972) Volatile com-pounds in butter oil. II. Flavour and f1avour thresh-olds of lactones, fatty acids, phenols. indole and


skatole in deodorized synthetic butter. J Dairy Res39,35-47

Vârnos-Vigyazé L, Kiss-Kutz N (1974) Estimationof volatile acids in Emmental cheese by gas-liquidchromatography. Acta Aliment 3,309-325

Vandeweghe P, Reineccius GA (1990) Comparisonof flavor isolation techniques applied to Cheddarcheese. J Agric Food Chem 38, 1549-1552

Varmuza K, Werther W, Lohninger H (1989) Anwen-dung chemometrischer Methoden zur Un ter-suchung von Massenspektren-Struktur-Beziehun-gen. In: Software-Entwicklung in der Chemie 3(Gauglitz G, ed) Springer Verlag, Berlin, 267-278

Vidal-Aragon MC, Sabio E, Gonzâlez J, Mas M (1994)Contribution to the study of the volatile compoundsin goat milk chee ses from Extremadura: effect ofthe season of manufacture. Alimentaria 31 (258),25-29

Wadhwa B, Bindal MP, Jain MK (1979) Isolation,fractionation and characterization of lactonic com-ponents of cow ghee. Milchwissenschaft 34,481-483

Westendorf RG (1985) An instrument for thermaldesorption of adsorbent traps. Am Lab December,72-76

Weurman C (1969) Isolation and concentration ofvolatiles in food odor research. J Agric Food Chem17,370-384

Wijesundera C, Urbach G (1993) Flavour of Cheddarcheese. Final Rep, Dairy Res Dev Corp, ProjectCst66, CSIRO Div Food Science and Technology,Melbourne Laboratory, Highett Vic, Australia

Wong NP, Parks OW (1968) Simple technique forextracting flavor compounds from fatty foods.J DairySci 51,1768-1769

Wood AF, Aston JW, Douglas GK (1994) A cold-trap method for the collection and determination ofheadspace compounds from cheese. Aust J DairyTechnoI49,42-47

Woolley C, Mani V (1994) Schnelle und einfacheBestimmung von Aromastoffen mit Festphasen-mikroextraktion und Kapillar Ge. SupelcoReporter 13 (4), 9-10

Wyllie SG, Leach DN, Wang Y, Shewfelt RL (1994)Sulfur volatiles in Cucumis melo cv Makdimon(muskmelon) aroma. In: ACS Symp Ser 564. Sul-fur Compounds in Foods (Mussinan CJ, KeelanME, eds) Am Chem Soc, Washington DC, 36-48

Xanthopoulos V, Picque D, Bassit N, Boquien CY,Corrieu G (1994) Methods for the determination ofaroma compounds in dairy products: a comparativestudy. J Dairy Res 61, 289-297

Yang WT, Min DB (1994) Dynamic headspace anal-yses of volatile compounds of Cheddar and Swisscheese during ripening. J Food Sei 59,1309-1312

Zannoni M (1995) Preliminary results of employ ofan artificial nose for the evaluation of cheese. SeiTec Latt-Casearia 46,277-289

instrumental analysis of volatile (flavour) compounds in

Documents