A Historical Survey of Coffee Aroma Research 59 (1844) assamar but “got stuck’ without success in an unbelievable series of analyses of caramels of various shades of brown and difficult to identify. In 1858, Vlaanderen and Mulder claimed to have isolated from Jara coffees ‘Kaffeesiiure’ (C)Hy0: and acids from its transformations (one “Kaffeansdure’ (CigHs0s) and two ‘Coerulinstiuren’ (C\sHsOs, CiuH)Os], and *Katfeéisiure’ (Ci4HyO)2). Since these substances were relatively impure, no subsequent study proved possible on the basis of their results. Hlasiwetz (1867) published a study on the tannic acids of coffee and reviewed the previous studies. Bernheimer (1880a,b) observed with sober realism that, apart from non-volatile components such as caffeine, the famous ‘assamar’, oleic and palmitic acids, ‘chloroginic’ acid and its dccomposition products, as well as various phenolic compounds, all that had really been identified as a volatile and aromatic component of roast coffee was acetic acid, He embarked on the first genuine study of the roasting products of coffee by constructing special apparatus and applying thermal treatment to S0kg of green coffee which unfortunately, no doubt for reasons of economy, he describes as damaged. He first observed that roasting caused a 25% loss in the weight of the mass treated. The release of carbon dioxide and formation of pyrrole are clearly observed. He isolated $ liters of liquid condensate and 680g of solid extract. For the first time he isolated an ethereal extract with a very pleasant odor but which, on distillation, gave only water, acetic acid and a liquid whose odor was strongly reminiscent of acetone. By progressive heating he obtained, around 190-200°C, a relatively characteristic liquid fraction, but at 300°C a fatty mass started to distil having the consistency of butter and an odor of rancid fat. Without further ado he successively treated part of the distillate—a brownish oil insoluble in hot water, alcohol or ether—with caustic soda and then sulfuric acid, thus irreparably destroying the particularly delicate components. From the more or less pyrolized mass he nevertheless sueceeded in extracting, crystallizing and identifying hydroquinone. Since at this stage there remained only 100mg of extract, he gave up all hope of any subsequent elementary analysis, He continued the analysis of the fraction of distillate collected between 195 and 197°C, which he called ‘caffeol’, in the hope either of identifying it simply as a component or of isolating a known substance from it. By elementary analysis he characterized this extract as a compound with the formula CxHy)Q2. which recalls creosol or veratrol, and for which he proposed the structure of a methyl derivative of saligenin, either 2- methoxybenzyl alcohol, or 2-(methoxymethyljphenol. Having isolated only 20g of this substance from the initial SOkg of coffee, he decided 10 forsake the analysis, He does however indicate that the baich of coffee yielded 140g of caffeine, a proportion of 0.28% of the roasted beans. To sum up, the results of this analysis, though modest. indicate that the main components of coffee are palmitic acid, caffeine, ‘caffeo!’ (an as yet undefined substance), acetic acid and carbonic acid. His most original discoveries are rather the identification of hydroquinone, methylamine, pyrrole and perhaps acetone. Cech (1880) published a short and unsubstantial article on the preparation of a roasted coffee oil. He obtained yields between 8 and 13% from the roasted beans, mentioned the possibility of using it as a liquor ingredient and simply observed the slow formation of fine caffeine needles In 1887, Padé described a piece of apparatus that enabled researchers to determine precisely the absolute densities of coffees and to detect coffee frauds involving manipulation of green coffees and ‘watering of roast coffees. In this connection he mentions and comments on a fraudulent method whereby products released during roasting are condensed and reinserted into the coffee beans. After a summary analysis he deduces that this concentrated liquid contains caffeine, ammonia, amines, ammonia salt acrid bases near pyridinie compounds and fatty acids. By watering down the still-hot coffee with th mixture, 14 (0 15% of the lost weight can be fraudulently restored to it. In 1895, Monari and Scoccianti demonstrated the formation of pyridine by roasting coffee; they could not confirm the presence of methylamine and trimethylamine in the volatile components 0 Coffee Flavor Chemistry In 1898 Jacckle resumed study of the ‘caffeol” to which Bernheimer (18802) had assigned the formula HET: Pac, At thre ies he caine aianire at reenic conmmounds produced by the decomposition oflarger molecules during roasting was still considered to be a single constituent called ‘eaffeol’. He analyzed the volatile components supplied to him by a company whose roaster blow-off pipes were provided with a snow-cooled condensation system. The extract thus obtained appeared representative. with a good aromatic odor, but sour and bitter in taste. Using vapor distillation, Jaeckle isolated an oil which was fairly representative but which had a relatively pronounced acetic odor. During this analys he lengthened the list of identified components by the addition of furfurol (in fact furfural), trimethy mine, formic acid and resorcinol, but did not find Bernheimer’s ‘caffeo!’ Nineteenth century research into coffee aroma ends with the laborious and somewhat doubtful identification of some ten volatile components: carbonic, formic and acetic acids, methylamine, tri- methylamine, acetone, furfural, furfurylic alcohol, pyrrole and pyridine, as well as a few ill-defined phenolic components. ‘Ata time when organic nomenclature was neither codified nor scrupulously respected, and still highly empirical. risks of confusion abounded in the naming of aromatic extracts, their more or less purified components, and synthetic products. Thus, the aromatic distillate isolated by Bernheimer in 1880. and which he called “caffeol’, is changed to ‘kaffeol’ by Erdmann (1902a). This researcher takes over Bernheimer’s assumptions regarding the presence of methylated derivatives of saligenin (2-hydroxyben- zyl alcohol) and denies their presence in this ‘caffeol’, This is because 2-methoxybenzyl alcohol, already synthesized by Cannizzaro and Kérner (1872). possesses markedly different physical und organoleptic characteristics; this incompatibility of structure had. indeed, already been mentioned by Bétsch (1880), who supported the second assumption, of the methyl ether of saligenin [2-(methoxymethyl}phenol}. Unfortunately. this product, prepared subsequently by Thiele and Dimroth (1899), presented no coffee odor whatsoever. The contents of the article by Erdmann (1902a) exemplified the new proach to the search for, and confirmation of, volatile aromatic components; through developments in organic synthe- sis, reference products were used systematically to check the validity of an identification, Erdmann thus invalidated the elucidation of « principle allegedly olfactively characteristic of roast coffee. In the course of his research he prepared and carefully studied an extract, which he now called Kaffeedl, obtained by hydrodistillation of 225 ke powdered roast cofTee. By extraction with ether, he obtained @ yield of only 228 mg/kg, which he felt portended difficulties in conducting an effective study with the analytical resources then available. He undertook distillation under atmospheric pressure and obtained a clear acid distillate between 150 and 190°C and a few brownish drops up to 230°C. The residue took the form ‘of a thick, brownish mass. He also performed « careful distillation of the neutral fraction under reduced pressure (12-[3mm and 9.5mm), He observed that a major portion of this extract. ie. 431% of the total with a boiling point of between 70 and 80°C, chiefly consisted of furfurylic alcohol. Erdmann (1902b) confirmed the nature of this product by treating it with diphenylurea chloride and preparing a erystalline carbamate, He also drew attention to its formation by degradation of sugars and pyrolysis of lignous fibers. At the cost of laborious extractions, fractionations and derivatizations of the aromatic oily extract, he isolated methyl ethyl acctic acid, the structure of which was at first contested by Reichstein and Staudinger (1926b) (in 1955, they refer to Erdmann and quote ‘methyl ethyl acetic acid’). In the less volatile fractions of the distillation under reduced pressure (b.p.: 93-150°C/12-13mm), he freed the phenol components that possess the characteristic odor of coffee. Thus, in the acid fraction he demonstrated a pronounced flavor of guaiacol and creosol, and suggested that several other phenol components were present. This discovery thus supports observations mentioned earlier on coffee’s antiseptic properties and its action on certain lower organisms: they also echo the empiric observations of Weiss (1832) seventy years earlier. Erdmann also made an interesting observation: roasting equal weights of cane sugar, green coffee extract prepared by Rochleder’s (1844, 1846) technigue—and caffeine A Historical Survey of Coffee Aroma Research 61 produced, after percolation, a beverage with the flavor aud aroma of coffee, whereas binary mixtures of these components produce by no means the same flavor. From these observations we may deduce that ten years before Maillard (1912), Erdmann had discovered empirically the non-enzymatic browning reaction resulting from the heating of sugars with amino acids, peptides and other nitrogenous components. ‘The publications of Gorter (1908a,b) review and take stock of work on the non-volatile components of coffee over the previous 70 years. These studies show inier alia that the chlorogenie acid isolated byPayen ts indeed split by alkaline hydrolysis into Caticic acid (Cohig4) and quinic acid (U7 ht 25). bey also recall the isolation of quinic acid by Payen (1846a) and by Zwenger and Siebert (1861). In 1912. Grafe takes over the confused nomenclature of previous studies in regarding “Kaffeol’, “Katleon’, ‘Caféone’ and "Kaffeedl’ as the group of aromatic substances obtained with a yield of 0.3 0.45% by hydrodistillation of roast coffee powder and ether extraction. He contributes no notable advance in mentioning that 38% of the exiract consists of acetic and valerianie acid, that 50% is furfurylic aleohol and other furans, and that the rest is a mixture of phenols with the odor of creosote and a pyridinie derivative that could be responsible for coffee’s specific aroma. His article includes a ‘comparative analysis of the components of normal coffee and decaffeinated cofTee, the method for which had recently been patented. In 1913, Bertrand and Weisweiller, unaware of the previous discovery by Monari and Scoccianti, claimed to have identified pyridine in coffee for the first time. They considered that the proportion of this, volatile base is larger than the total of all the other bodies thus far reported in coffee. In several freshly roasted commercial samples they found quantities of 200-250 mg/kg, and proposed investigating whether pyridine plays an appreciable role in the physiological action of coffee infusion, In a bizarre owist, Sethness, in 1924, unaware of Erdmann’s (1902a) previous invalidation of Bernhei- mer’s (1880a.b) results and having himself conducted no tests on the product he isolates, again asserts that the methyl cther of saligenin is the major component of the aromatic oil of coffee. He cites the work of Lehmann and Wilhelm (1898), whom he wrongly quotes ‘as having also found caffeol, or chemically speaking the methyl ether of suligenin to be the principal ingredient of the fraction possessing the powerful coffee aroma’ (Table 4.2) Inthe years 1921-30, Reichstcin and Staudinger started exhaustive research on the aroma of coffee. The work was sponsored by the Swiss Confederate Foundation for the Promotion of National Economics through Scientific Research and, later, by Internationale Nahrungs und Genussmittel AG, Schaffhausen, Table 4.2 The aroma constituents of roasted coffee identified before 1902 Discoverers: ‘Compounds suspected or (tentatively) identified Lampadius (1832) Acetic acid Robiquet and Boutron (1837) Sulfur-containing compounds, gallic acid Rochleder (1844) Palmitie acid Rochlader (1846) Oleic acid Payen (1846) Quinic acid Payen (1849) Volatile hydrocarbons Bernheimer (1880) Methyiamine, pyrrole, acetone, carbonic acid, hydroquinone Padé (1887) Ammonia, amines Monari and Seoecianti (1895) Pyridine Jaeckle (1898) Furfurol (furfuraldehyde), trimethylamine, formic acid, resorcinol Erdmann (1902) Mcthylethylacetic acid (or valerianic acid), furfury! aleohol, phenolic compounds @ Coffee Flavor Chemistry Switzerland. In fact this research was economically motivated since, during and after the First World ‘War, the effects of coffee shortage were keenly felt in Europe. With this grant, Reichstein and Staudinger conducted a particularly brilliant study and, using classic derivatizing techniques and measurements of physical constants, succeeded in identifying ca 30 new components. Since their results are only to be found in patents, they have often escaped the scientific community's notice. Particularly worthy of mention is British Patent 246454 (Reichstein and Staudinger, 19261), describing the method of isolating the ‘aromatic principle’ of roast coffee, and British Patent 260960 (Reichstein and Staudinger, 1926b), describing the components identified and ‘a new or improved method of producing artificial coffee oil. ‘These results were to be published again only in the years 1950-SS (Reichstein and Staudinger, 1950b, 1955a,b), when the authors wished to gain acceptance for the priority of their discovery of particularlyoriginal components. It was then universally admitted that not one of the substances Isolated from rousted coffee possessed in itself the characteristic flavor aimed at. Nevertheless, there was little aware- ness that specific compounds were of paramount importance. This applied particularly to furfuryl mercaptan, 2 chemical unknown in the twenties when the authors detected and synthesized it for purposes of comparison. Pure furfuryl mereaptan has a persistent, obtrusive odor. Highly thinned, however, it exhales « pleasant note indicative of coffee. Marketing of this component under the name ‘coffee-captan’, announced in 1950 as an innovation by Cargill Scientific, Inc, New York in 1950, prompted Reichstein and Staudinger to redirect attention to their werk (1950b) and the priority of all their studies between 192] and 1930, as well as their patents filed between 1925 and 1929 (see References) They emphasize that furfuryl mercaptan does not in itself possess the aromatic qualities of coffee but is one of its most characteristic components. This product was, moreover, already contained in a more claborate formula marketed under the name ‘Cofarom’ for many years by the company Haarmann & Reimer (Holzminden). Their article, designed to set the record straight, likewise includes a list of the many patenis attesting to the priority and validity of their work. In a more detailed publication (Reich- stein and Staudinger, 195Sa), they reviewed a century of research into the aromatic components of coffee. They pointed out the difficultics and failures— inevitable for most of the time of thei predecessors, To recover the aromatic substances, large quantities of freshly roasted and ground coffee were first heated to 100°C under a good vacuum, at about 3 torr. The rising gases were collected after gradual cooling down to —180°C. Thus it became possible to concentrate strongly all the aromatics present, and to maintain their structure unchanged, avoiding the formation of undesirable by-products. By distillation a highly unstable yellow-colored oil was recovered and stored at 80°C. Through exerting separation procedures, over 70 characteristic substances could be isolated from this source ‘and each of them identified beyond doubt by the presence of crystalline derivatives, and by chemical analysis “The authors observed that an exhaustive list of all the chemicals present in coffee flavor had not yet been compiled, but they believed they had identified the components that are present at the higher ratio of weight, and those which principally control the odor note, Most of the substances identified were well- known compounds present in other roasted products us well, for instance in caramel sugar, cocoa, baked bread and—partially—even in wood tar. However, some of the chemicals detected were new and, obviously, characteristic of roasted coffee. Traces of methyl mercaptan, which was already known at that time and which smells even worse, were also detected in coffee aroma. Commenting on this observation, Reichstein and Staudinger note that it is generally known that many popular raw materials and synthetic perfume compounds owe their characteristic note, which is extremely pleasant. to the olfactory sense, to their content of small quantities in additives which carry a rather unpleasant odor in themselves but prove very attractive in thinned solutions and in admixture with other oils. The authors tried to reconstitute coffee aroma, and only by combining over 40 of the substances extracted from coffee Al orical Survey of Coffee Aroma Research 63 oil did it become possible to develop compound mixtures which, upon adequate thinning. held a typical coffee aroma, though this was not up to par in quality with the natural product. Reichstein and Staudinger (1955a) also compared the most important components of raw and roasted coffee. ‘They observed that sugars and tanni¢ acid are dramatically reduced in quantity while the caffeine content remains constant. Finally, they obscrved that the fats do not undergo degradation but play 2 role tuough their capacity to adsorb and protect the volatile aromas of evaporation, oxidation and deterior- ation. In conclusion, the authors identified numerous new constituents mentioned in the Table 4.3. Schmalfuss and Barthmeyer (1929), still unaware of the results of the analytical work of Reichstein and Staudinger, studied the presence of diacetyl in foods. They noticed the caramel and buttery note of this compound that results from the oxidation of acetylmethylcarbinol (3-hydroxy-2-butanone) and quanti- fied its presence in butter, tobacco, coffee, cocoa, beer and honey. Under the title “The staling of coffee’, Prescott et al. (1937a,b) drew attention to the fragility of the volatile components of roasted coffee to preservation and to the modification of flavor known variously as “flatness, staleness or rancidity’, It was generally assumed that oxidation or saturation of fats wasthe principal cause of (his TaNcidlly. Eres cot Ce a. ee) eno noe eee eS believing that the oxidation theory was inadequate to explain all the changes that were brought about. Deterioration is also due to hydrolysis, alteration and volatilization of the aromat principles. After Table 4,3 Volatile compounds identified by Reichstein and Staudinger (1926) Hydrogen sulfide Methylmereaptan Furfurylmercaptan Higher mercaptans Dimethyl sulfide Higher sulfides Acetaldehyde Methylethylacctaldehyde (2-methylbutanal) Furfurol Methylfurfurot 2-Acetylfuran Acetone Higher aliphatic aldehydes and ketones Ketones of the furan series Diacetyl Acetylpropiony! Onysulfides and merc With the above men Methyl alcoho! Higher aliphatic alcohols, Acctol Furfury! alcohol Acetic acid ans of carbonyl compounds joned mercaptans sovaletic acid Higher fatty acids Palmitic acid Esters of the former two acids with the above mentioned aleohols Methyl acetate Methyl 3-methylbutanoate Phenol Catechol Guaiaco! p-Vinylguaiacol Dihydroxyacetophenone Forther phenols and phenol ethers of a higher valency Maltol Pyridine Pyrazine Methylpyrazine 2.5-Dimethylpyravine Dimethylpyrazine Higher homologues of pyrazine N-Methylpyrrole N-Purfurylpyrrole Further higher pyrroles Higher derivatives of furan Furfuryl acetate Purfury! 3-anethylbutanoate ‘Naphthalene 4 Coffee Flavor Chemistry summarizing previous work on the volatile constituents of coffee, the authors performed different extractions of volatile constituents. Having observed that the percentage of volatile oil in roasted coffee is always very small, since Erdmann (1902a) obtained only 83.3 of volatile oil from 150 kg of roasted coffee (0.05%), four different methods of isolation were used. The authors were only able to identify furfuryl alcohol, furfuraldehyde and acetic acid from the volatile oil Prescott ¢¢ al. (1937a) finally preferred to extract a coffee brew with ether. The results were again very disappointing, the identified constituents being ettayl alcohol, acetic acid, furfuraldehyce, furfuryl acet ute, furfuryl alcohol, probably furfuryl formate and a small amount of phenol. In the discussion of thei results the authors agreed with the results of previous investigators: furfuryl alcohol is the chief neutral constituent of aromatic oils of coffee, as ascertained by Erdmann (1902a) and Grafe (1912), However. they disputed the preponderance of valerianic (or methylethy! acetic) acid for which Erdmann argued. since in their view acetic acid was the principal volatile acid present in roasted coffee. Everything suggests that Prescott e/ al. were then unaware of the content of the patent of Reichstein and Staudinger applied for 10 years earlier. They suspected the presence and olfactory contribution of nitrogenous components present in the ‘caffeol’ of Sethness (1924). They also remained intrigued by the hypothetical discovery of ‘methyl derivative of saligenin by Bernheimer (188Ca) nearly 60 years before. In fact, since Prescott eu. really only demonstrated in their analysis the relative abundance of furfuryl alcohol and acetic acid in coffee essence, thev ascribed the rancidity of coffee either to the presence of or the interaction of thesetwo components. Furfuryl alcohol, its formate and its acetate, preserved in sealed test tubes in the presence of damp air, exhibit a marked tendency towards polymerization and resinification in 8 weeks standing. However, the authors did not tule out the contribution of degradation phenomena due to the presence of other unstable products such as certain phenolic components or the decomposition of volatile constituents. They wished to point out the possibility of any group of compounds other than the fat being responsible for the staleness of coffee and concluded that this statement is hased om experimental observations and not on pure speculation. In a second study on the staling of coffee, Prescott et af. (1937b) observed that furfuraldehyde, the oxidation product of furfuryl alcohol, is an odoriferous liquid which promptly undergoes decomposition with the formation of tarry materials. They identiiy diacetyl, diethyl ketone, n-heptacosane, p-vinyl guaiacol and guaiacol (two phenols already identified by Reichstein and Staudinger, 1926b). They suspected, without really being able to confirm it, the presence of p-vinyl catechol. They identified for the first time eugenol and vanilione (4-hydroxy-3-methoxyacetophenone) and isolated the ‘kahweo!’ (the ame was used later for a diterpene, see Section 2.1.3) previously reported by Bengis and Anderson (1932a). Prescott ef al, (1937b) have the merit of being the first to isolate a terpenic fraction and of identifying sylvestrene in it, ie, fimonene, which was only to be confirmed 30 years later by Stoll e¢ ai (1967). They also mentioned that the terpenes of this fraction are volatile substances possessing taste and ‘odor but sensitive in the highest degree to heat, light, air and chemical reagents, They predicted that their identification would be a matter of great importance where the flavor and aroma of coffee are in question. In view of the hydrocarbon nature of the terpenes they decided to substitute benzene for ether as extractive. For Prescott ef al, (1937a,b), unlike their predecessors in the Nineteenth century, the evidence indicated that the fats were not responsible for either flavor or aroma, since fat-free coffee extracts also possess a rich full coffze odor. This prompted them to put the question yet again: ‘What types of chemical reaction must occur on roasting? They suggested three different mechanisms, The oxidation of the fats, however, is not in itself responsible for the staleness as distinguished from the rancidity of coffee. It is worth noting that Prescott et al, (1937b), when they published the second part of their work, knew of the English patent of Reichstein and Staudinger, “two of the best organic chemists of the present time. Staudinger is noted for his work in organic qualitative analysis and highly polymerized substances, while Reichstein first synthesized vitamin C’ A Historical Survey of Coffee Aroma Research 65 The results of an analysis of the volatile constituents was published by Johnson and Frey (1938), the aim of their study being to prevent or retard the staling of roasted coffee. The authors declared that this, effort would have been undertaken with better prospects of success if the identity and nature of the substances responsible for coflee flavor and aroma had been more completely ascertained. Johnson and Frey believed that the treatment used by Reichstein and Staudinger was not ideal because previous work on staling had shown that very small amounts of oxygen and moisture were sufficient to cause and accelerate deterioration. They decided to use a high-vacuum apparatus and to avoid using water during distillation. They were mercly able to isolate, as actual constituents of coffee, several of the substances reported by Reichsiein and Staudinger, and 1o detect acetylmethylearbinol G-hydroxy-2-butanone), @ constituent of roasted coffee previously identified by Schmalfuss and Barthmeyer (1929). From Prescott's results and their own work they concluded that most, if not all, the compounds previously identified by Reichstein and Staudinger (1926), suggested by these authors to be split products formed during the analytical treatment, were in fact actually present in coffee. They are in complete agreement with Prescott etal, (1937a,b) with respect to the probable changes occurring during the stuling process, which probably involves volatilization, onidation, hydrolysis, and polymerization of the various flavor and aroma constituents, and has little if anything to do with the fat of coffee. Johnson and Frey (1938) identified furan. In 1948, Neu identified nonacosane after extraction and column chromatography. A study on the volatile constituents of roasted coffee was published by Hughes and Smith (1949) but without the identification of any original component, The authors considered that the substance, or substances, giving the characteristic aroma to roasted coffee was as yet unidentified and that it was probable that less volatile and readily oxidizable substances, were involved in the flavor of roasted coftee. Their main problem was to investigate the changes that take place during staling. With the relatively pooranalytical tools then available, essentially colorimetric methods, they tried to quantity the main—alrcady identified volatile constituents which they considered to be typical. They particularly observed the evolution of pyridine, furfural, aldehydes, diacetyl, acetylmethylcarbinol, zcetone, volatile phenols and hydrogen sulfide In the first part of his review on the aroma of coffee, Moncrieff (1950a) did not reveal the identification of any new flavor constituent, but summarized 120 years of research. He reminded us obviously that the ily substance, the lighter fraction of the distillate which gives much aroma to roasted coffee, initially named ‘caffeol’ and whose identification had been the dream of generations of chemists, is in fact a complex mixture of substances. Kaufman (1951) summarized some recent developments in fundamental research on coffee flavor, reviewing the evolution of knowledge in the fields of green coffee production, roasted coffee processing, methods of control and packaging. He reviewed the developments of the powdered o soluble coffee extracts technology and made some general comments on the fundamental analytical research into coffee aroma, insisting that the sulfur compounds in coffee were the most important contributors to the flavor. Alter the identification of pyruvic acid by Mabrouk and Deatherage (1956), Clements and Deatherage (1957) proposed a chromatographic study of some compounds of roasted coffee but their results are somewhat ambiguous because the lists of oxo acids characterized as 2.4-dinitrophenylhydrazones are given as reference compounds with the risk of persuading the reader that they are effectively present in coffee. In fact, only trivial compounds were tentatively identified, and propionic and butyric acids mentioned for the first time as being present in roasted coffee flavor. Lockhart (957) identified ethanol and furfuryl formate, Hégl (1958) analyzed by paper chromatography an extract of non-volatile constitu ents of roasted coffee. He identified hydroxymethylfurfurol, pyrogallol, and rediscovered hydroquinone (Bernheimer. 18802.b), as well as maltol and catechol previously identified by Reichstein and Staudinger (1926b). 66 Coffee Flavor Chemisiry 4.2. MODERN TIMES: THE ADVENT OF GAS CHROMATOGRAPHY From 1956, gas chromatography (GC) was introduced as a separation technique, and throughout the sixties, identification systems like mass spectrometry (MS) and nuclear magnetic resonance (NMR) progressively equipped analytical laboratories. The use of these instruments has dramatically accelerated the identification of volatile food constituent Including the previous noteworthy contribution of Reichstein and Staudinger (1926a,b), 10 outstand- ing research groups have actively participated, mainly between 1965 and 1985, in the development of roasted coffee flavor chemistry. by discovering more than 600 original constituents (Pable 44) The identification of original characteristic compounds is becoming more and more difficult, but thanks to very sensitive detectors, typical trace constituents continuc to be discovered, as will be discussed later. Dimick and Corse (1956) were probably among the first authors to analyze food flavor volatiles by GC. They published an article entitled: ‘Gas chromatography, A new method for the separation and identification of volatile materials in foods’, Unfortunately, lacking a satisfactory identification device, the authors did not discover uny new constituents. Using a home-made chromatograph, constructed by his colleague E.Palluy, for the analysis of coffee flavor, Gautschi (1958) separated, within 35 min, and identified seven volatile aliphatic acids: formic, acetic, propionic, isobutyric, butyric, isovaleric and valerie acids. The less polar and more volatile corresponding methy/ esters were prepared by the action of diazomethane and separated within 25 min. ‘Table 4.4 The ten research groups having identified 77% of the roasted collee flavor constiuents (1) Firmenich et Cie, Geneva, Switzerland Stoll M., Winter M., Gautschi F., Willhalm B., Flameat 1. and Goldman 1. M., Scibl J. (Massachusetts Institute of Technology, Cambridge, USA): 149 identifications (1988-67). ® ‘Jacobs-Suchard Corporate R&D, Kraft Jacobs Suehard, Bremen, Germany _Vitzthum, O.G.. Werkholl P., cosa et yar eet g e er cre me, oi eet eee) Wegner H., Gutmann W., Barthels M., Beadig L., Wollman R, and Nitz S., Dochla B. (Technische Universitat Munchen, Freising- Weihenstephan, Germany): 145 idemifications (1974-97), (3) Res. Inst. Chem. Tech. Anal., Technische Universitit, Berlin, Germany Tressl R., Silwar R., Képpler H.. Griinewald K.G.. Bahri D., Jensen A., Kamperschroer and Rewicki D. (Inst for Org. Chem., Free University, Berlin): 99 identifications (1978-87), (4) The Coca-Cola Company, Atlanta, Ga., USA Gianturco M.A., Friedel P., Giammarino A.S., Renner J.A., Shephard E.W., Krampl V., Radford T.. Bondarovich H.A.. Flanagan V., Pitcher R.G.. 61 identifications (1963-71). (5) Institute for Food Chemistry, Technical University, Berlin, Germany Baltes W., Bochmann G., Heinrich L.: 56 identifications (1986-87) (6) US Army Natick Laboratories, Natick, Mass., USA. Merritt C. Jr., Robertson D.H., McAdoo D.J.. Bazinet C., Sullivan J,H.: 38 identifeations (1963-70). (7) International Nahrungs-und Genussmitte! A.G., Schaffausen, Switzerland Reichstein, T., Staudinger H.: 31 identifications (1926), (8) Polak’s frutal Works N. V., Amersfoort, Holland Stoffelsma J. Kettenes D.K. and Pypker J. (Douwe Egberts Koninklijke Tabakstabriek- Koffiebranden jen-Thechandel N, V., Utrecht, Holland), 30 identifications (1968), (9) Nestle Techn. Assist. Co., Nestec Ltd, Lausanne, Switzerland Reymond D.. Egli R.J.. Liardon R.. Viani R., Miggler-Chavan F., Vuataz L. Out U., Aesehbacher H.U. Wolleb U., Laliger J., Spadone J.C., Woodman J'S., Giddey A., Pollien P., Krebs. Y., Chaintreau A.. Takeoka G25 identifications (1963-98) (10) University of Houston, Texas, USA Zlatkis A., Sivetz M., Wang T-H., Shanfield H.: 18 identifications (1960-83) A Historical Survey of Coffee Aroma Research 67 ‘Another innovative GC application is due to Rhoades (1958). He used. in isothermal conditions at 45°C, one of the first commercial chromatographs (Vapor Fractometer Perkin-Elmer. Model 154), equipped with an external gas sampling valve and 2-meters of copper tubing filled with driod fire brick coated with 20% Carbowax 1500. The sampling procedure consisted in sweeping heated ground coffee with a stream of warm and moist helium. The chromatogram of the volatiles exhibited 16 distinct peaks and the author identified acetaldehyde, acetone and methanol, compounds which were unfortunately already known coffee components. Dimethyl sulfide, propionaldehyde, ethy! methy! ketone and diacety were also tentatively identified. Though the results were somewhat meager, they where promising. Two years later, Rhtoades (1960) compared seven varieties of coffee and noted the modifications of volatile concentrations with roasting. He observed, for example, that hydrogen sulfide reaches a maximum concentration at ca 175°C, methyl formate near 200°C and acetyl propionyl (2.3-pentanedione) near 220°C usually the final roast temperature. He identified and quantified 18 volatile compounds and confirmed, for the first time, the presence of isoprene, propanal, butanal, isobutyraldehyde, isovaler- aldehyde, 2-butanone, methyl formate and 2-methylfuran, He observed that the ratio of diacetyl to aectyt propionyl increases with roasting, a value which he proposes be used for measuring the roasting degree of trade coltee varieties. Zlatkis and Sivetz (1960) (University of Houston, Texas, USA, sce Table 4.4) isolated and analyzed two types of extract: the very volatile ‘coffee aroma essence’ (CAE). collected from commercial percola- tor vent gases (200 ppm from roasted coffee) and the less volatile ‘dry vacuum aroma’ (DVA) fraction, prepared by steam distillation and ether extraction. Using packed columns mounted with a catharom- eter, the classical equipment at that time, the authors observed that the CAE, which has an extremely powerful odor and gives the true coffee aroma when sufficiently diluted in the air, is largely composed of aldchydes and sulfides, the main constituents being acetaldchyde, acetone, diacetyl, valeraldehyde and 2- and 3-methylbutyraldchydes. In spite of the low resolution of the GC columns then available, the authors obtained an outstanding chromatogram of coffee aroma essence. MS helped them to characterize 30 volatile components, some of them not yet identified in coffee, such as 1.3-pentadiene, valeraldehyde, ethyl formate, thiophene and carbon disulfide. More than 20 years later, a group from the same laboratory analyzed the trace volatile compounds in coffee by headspace concentration. heart cutting separations, Tenax adsorption and GC-MS identification, Wang ef al. (1983) identified 13 other constitu~ ie and observed, that the characteristic compounds of coffee aroma formed a complex mixtureincluding esters, ketones, alcohols, phenols. sulfur-contaming compounds, pyrazines, furans ane’ pyt= «sthormiller erat (1961) identified and quantified organic weds in « commercial instant cofTee powder whose origin was not revealed. To be noted particularly were lactic, succinic, fumaric and 2-oxoglutarie mn 196, the Institut Frangais du Café, du Cacao et autres Plantes Stimulantes (ICC) organized in Pais the first “International Colloguim on the Chemistry of Green and Roasted Coflee and other Stimulant Plans’ A biennial fequeney was planned. The second meeting was agai held in Paris in 196, Then from 1967, in Trieste, the name became ‘Internationa Callan oe the Chemistry of Coffee ani Association Scientifique Internationale du Café (see Section 1.2) oie tow-boling Nolatiles of cooked foods, Sell e? al. (1963) observed that, after boiling, « numberof foods contained at least six ofthe 13 components of the volatile fraction from boiled potatos. The authors concluded that most cooked foods probably produce 2 similar pattern of low-boling volatiles varying only in the relative quantities present. This suggested the formation of the common volatiles by degradation of metabolites that are normally present in ll biological material. Most of them were indeed produced from amino acids in model systems, The authors concluded that the di cia flavor among certain foods, do not lie in the presence of one or more unique components, but in the 6s Coffee Flavor Chemistry relative quantitative pattern of low-boiling substances. In an instant coilee powder, Self ei al. (1963) identified ethenethiol, In 1963 appeared the publications of ¢ second group actively involved in coffee aroma research (Merritt etal, US Army Natick Laboratories. Natick, Mass.. USA, see Table 4.4). Merritt er af. (1963) applied more or less routinely the ‘modern technique’ of GC/MS coupling as an identification tool. After having discussed the possibility of chemical changes or contamination of the volatile components, by the use of heat, steam, water or solvent extraction, they considered that the most typical compounds may he sampled by xn entirely static transfer procedure, that is to say by distillation under vacuum at room temperature from dry and ground roasted coffee, with condensation into a receiver at liquid nitrogen temperature. Merritt ef af, (1963) give a table summarizing the functions of coffee aroma constituents: aldehydes (50.7%), ketones (21.5%), esters (10.3%), heterocyclic compounds (7.0°%), sulfur compounds (4.3%), alcohols (2.4%), and nitriles (1.7%). According to these authors, the occurrence of 12 compounds: previously reported was confirmed and 18 additional compound were identified, six of these tentatively In fact, eight products were really original: propenal, 2,4-pentanedione, 2,5-dimethylfuran, 2-propeneni- trile and 3-butenenitrile, ethyl methyl sulfide, dimethyl disulfide and ethyl methyl disulfide. The criterion for an identification was the concordance of the mass spectrum of the sample component with that of a pure reference compound. This method of confirmation had, in effect, been too often neglected by authors, complicating the validity of numerous ‘identifications’. Merritt and Robertson (1966) presented the continuation of this study at the Second International Colloguiurn on the Chemi try of Coffee in Paris in 1965. They described in detail a special headspace-sampling device and a combined gas chromato- graph-rapid scanning mass spectrometer analysis system which looked very impressive and was evidently only available to a restricted number of analytical laboratories. The mass spectra of eluted components Were monitored visually with an oscilloscope and recorded. when appropriate, on an oscillograph recorder. OF the 56 compounds identified, 14 were original, particularly the hydrocarbons. The authors had already insisted that the investigation of natural flavors must incorporate organoleptic evaluation with the analytical methods employed, otherwise there is no assurance that what is detectable is significant for the aroma, it is therefore neither presumed that each compound is flavor significant nor that the analytical results allow the establishing of quality control for the Finished market products, the target of these studies being, among other things, to improve the acceptability of soluble coffee products. In 1969, at the 4th ASIC symposium in Amsterdam, Merritt e¢ al. (1970) asserted that the object of research on the composition of the constituents of coffee aroma is not the mere compilation of lists, but the relationship of the compounds to their precursors, in order to establish a mechanism for their formation, and ultimately for controlling the quality of the product. Merritt e7 a/. (1970) tried to correlate the composition of green and roasted coffees and gave a list of some pyrolysis products of various amino chit. plameutnes thay -evenpalites eranentint ear ahes icea wetness Ck ee eaeaving id pel 18. action mevrasti eae ae ee having seated 16 a constituents, the authors hoped that new techniques will lead to more direct elations between the aroma and their precursors. providing @ m fe basi aluate 3 Sorreaony heen thea PI 1g a more sccure basis to evaluate and Gianturco e* al. (Coca-Cola Company Laboratories, Atlanta, Ga, USA, see Table 4.4) elucidated 1963, the siructure of five cyclic diketones isolated from a coffee oil obtained by extrusion-pressing of whole roasted coffee beans. They prepared an ‘aroma complen' with a yield of (025% hased on the weight of roasted coffee, by stripping the volatile fraction, at room temperature and at a pressure of 0.1 0.2 micrometers, from the mixture of glycerides. The separations and identifications were made by fepeated gis chromatography, measurement of the retention times ona polar and non-polr eolann gether with the pertinent UV data. The syntheses of the original eompounds vere published Giantureo and Friedel (1963). In 1964, Gianturco ef al (19642) described the structures of two hetero- eyclic compounds and the synthesis of some tetrakydrofuranones. They pointed out that the instability ‘A Historical Survey of Coffee Aroma Research 69 and complexity of the ‘aroma complex’ mixture and the very low concentration of many of its constitu- ents caused some difficulty in the analysis of the aromatic fraction. They isolated by gas chromatography ‘and identified by IR and MS, 2-methyltetrahydrofuran-3-one and confirmed the structure of V-furfur- yipyrrole, a substance previously found in roasted coffee by Reichstein and Staudinger (19266) and synthesized by Reichstein (1930a), Gianturco ef al. (19646) reported the isolation from roasted coflee of 25 simple furanic and pytrolic compounds, all identified by IR spectroscopy. They considered that the ecent investigations had been focused exclusively on the very volatile constituents of the “aroma complex’, that this choice was somewhat arbitrary, and that this approach had too narrow a base. They intended to go deeper into the identification results by focusing on only the furanic and the pyrrolie compounds. Through this specific approach, the authors obtained efficient results with the new identifi- cations of six furans and cight pyrroles, In a short article, Giantureo er al, (1966) gave a list of 85 compounds including 12 new identifications. Bondarovich ef a. (1967), recalling the results of their previous investigations (Giantureo er al., 1966; Gianturco, 1967), realized that continuation of the analysis of the complex roasted coffee flavor became futile without the aid of mass spectrometry. They estimated the total number of reported volatile components to be 284 but considered rightly that although it could not be claimed that all of the constituents had already been identified, the information collected so far should be sufficient to attempt to determine the relative organoleptic significance of the various compounds. They particularly insisted on. the fact that the pyrazines might contribute in an important way to the favor of a variety of roasted, or otherwise cooked, foods. The article includes the IR spectra, MS and UV data of 22 alkyl- and alkenylpyrazines present in coffee, five of them being new identifications: 2,3-diethyl-S-methylpyrazine, 2,6-diethyl-3-methylpyrazine, 2-(trans-1-propenyl)pyra- Zine, 2-methyl-5-(irans-1-propenyl)pyrazine and 2-methyl-6-(rans-L-propenyl)pyrazine, Twelve other identifications of constituents were mentioned: tetradecane, pentadecane, four cyclopentenones, two eyelohexenones, I-hydroxy-2-butanone acetate, acetol propionate, 2,4,5-trimethyl-2H-Furan-2-one and ‘N-amylpyrrole. Friedel et ul. (1971) reported 12 new identifications: one ketone, two furans, eight pyrazines and butyraldoxime. Reymond ef «i, (1963) (Nestlé Technical Assistance Co, Nestec, Lausanne, Switzerland. See Table 4.4) inaugurated a series of articles by publishing the results obtained by the separation on a steel capillary column (with a sensitive B-ionization Sr detector) of samples obtained by the gas sweeping method. from the gus surrounding roasted coffee, from an aqueous suspension and from an aqueous extract of ground cofice. The identifications were made by measurement of the retention times relative to acetone, chosen as an internal standard, and for 13 of the 23 constituents detected the structures were elucidated by IR speetroscopy. Among the highly volatile constituents were methane, ethylene and acetaldehyde dimethyl acetal, The authors also measured the changes occurring during the staling of coffee and observed, following extraction of roasted coffee by means of hot water, the variations in the equilibrium of the aroma constituents. The relative amounts of methyl acetate, acetaldehyde, propionaldehyde, isobutyraldehyde, butyraldehyde, isovaleraldehyde and methyl cthy! acetaldehyde increase and the methulfugan, methvl formate, methyl ethyl ketone and diacetyl decrease. Viani pee er atet al, (1965) enumerated the previous analytical works on coffee flavor. Using preparative gas chroma- ography on several stationary phases, measurement of Kovats indices and IR. spectrography, they identified 38 compounds and confirmed the structures of four new substances: 2.3-dimethylpyrazine, y-butyrolactone, acetol acetate and 2,3-hexanedione, With high resolution mass spectrometry they explained the fragmentation of 2-methyltctrahydrofuran-3-one, previously isolated by Gianturco ef al (1964a), and confirmed the structure, In order to enrich the samples in higher boiling components not detected hy the classical head-space procedure, Reymond ef ul. (1966b) proceeded to a steam distillation of an aqueous beverage under reduced pressure, Followed by extraction of the condensate with dichlor- methane. Using ethyl butyrate as internal standard, they converted head-space data into Tiquid- 70 Coffee Flavor Chemistry composition data and calculated the approximate composition of an aqueous solution from the compos- ition of the head space. 3,4-Dimethoxystyrene was among the 18 steam-volatile constituents of a Nicaragua coffee. At the 3rd ASIC symposium in Trieste in 1967, Woodman e¢ al. (1968) presented a work on the carboxylic acids of brewed cofiee with the identification of non-volatile acids such as fumaric, citraconic, mesaconic und itaconie acids. An exhaustive review of the role and quantification of carboxylie acids in green and roasted coffee, their origin and their changes on roasting and storage was published by Woodman (1985). Liardon and Ott (1984) analyzed 15 coffees from three different origins and attempted, with the help of multivariate statistics. to classify the headspace profiles into coffee variety and/or roasting categories. A success rate of 98% was obtained using 18 components. Profile ;nment appeared to result from the combined information provided by components belonging to four families: aliphatic aldehydes, furans. pyrroles or pyrazines, and thiophenes. GC/MS analysis led to 92 identifications, of which five were new: 2-hexanone, 3-cthylfuran, 3.4-cimethylfuran, acctonitrile and S-methyl ethanethioate, Spadone and Liardon (1988) identified specific volatile components in ‘Rio’ green cofiee beans. They quantified particularly 2,4,6-trichloranisole which is responsible for the phen lic, medicinal and musty off-flavor of green coffee beans, originating mainly from Brazil and Puerto Rico. Although this compound is destroyed at 50% by roasting. it remains perceptible in the final beverage. Aeschbacher er al. (1989) studied the contribution of 40 coffee aroma constituents to the mutagenicity of coffee. They mention, for the first time, the presence.of glyoxal in coffee and observed that only aliphatic dicarbonyl compounds showed notable direct mutagenic activity. Pollien et al, (1997, 1998) developed a new GC-olfactomeirie method (SNIF-analysis) for tracking the key components of coffee. They compared a brew and an instant coffee and observed that 3-methyl-2-butenal (erroneously described as the 2-methyl isomer), never previously mentioned in coffee, has « high aroma impact in the brew: however it could be an artifact. They also identified |-nonen-3-one, a compound exhibiting a typical mushroom odor and a very potent odorant. ‘Among numerous already known components, Radtke (1964) identified two phenols (2- and 4- methylphenol) and two picolines (Q- and 3-methylpyridine). In 1965, at the Second Internation Colloquium on the Chemistry of Green and Roasted Coffees. Heesterman and Weurman (1966) presented u ‘preliminary report on the tentative organization of an international cooperation in the field of research into the aroma of coffee and especially by gas chromatography’ and ‘on coffee-odor research methods’, I was good advice from a scientific point of view, but utopian on an economical basis. The coffee market presents such an enormous capacity that coffee and flavor companies involved in the field were not enthusiastic about sharing the results of the considerable investments in their own research laboratories in order io improve the quality of their products. The use of the indices proposed by Kovats in 1958 was timidly applied by Chassevent er al, (1966) who recognized that they were not able at that time to analyze ‘very complex mixtures containing 24 to 28 constituents’! At the same meeting, Gianturco and Giammarino (1966) report that the ‘aroma complex’ of coffee may be made up of approximately 250 components and one of the chromatograms that he had shown presented 150 peaks. Joking during his presentation. the first author affirms that he would be only too happy if someone will be able to show that, under certain chromatographic condi- tions, one can indeed achieve complete—or almost complete—resolution of the "aroma complex”, even if this involves the use of up-to-one-mile long capillary columns! The 11 polycyclic aromatic hydrocarbons identified by Grimmer and Hildebrandt (1966), will be discussed in Section 5 4.For the first time, Heins et af, (1966) analyzed the head-space vapors by their direct injection on a GC capillary columa coupled to a mass spectrometer, eliminating by this procedure the danger of artifact formation occasionally due to concentration procedures. Unfortunately & number of technical imper- fections had to be worked on, particularly the blocking of ihe capillary by dry ice during the injection. A Historical Survey of Coffee Aroma Research 7 The results were indeed relatively medivere. Of the 56 peaks present in the chromatogram only 31 were indubitably identified, among them 12 new identifications: benzene, styrene. |-hexanol, 2-butenal. methyl propionate, butyl acetate, 3-methylfuran, 3-propylfuran, tetrahydrofuran, 2-thiophenemethanol. 2-cthylpyrrole and diethyl! disulfide. Walter e7 ul. (1967) recommended the use of derivatization techniques, as a complement to gas chromatography, for the identification of less volatile compounds, in spite of the danger of producing artifacts. They characterized 1,2-propanedione (methylglyoxal) and 1,2-cyclopentanedione, eheady identified by Gianturco et af, (1963), through the 2.4-dinitrophenylhydrazones using IR spectroscopy and thin-layer chromatography. Volatile acids in the beverages of three varieties of coffee (Columbia, Santos and robusta) were determined by Kung ef al, (1967), formic acid by NMR, the remainder by gas chromatography Heptanoic, nonanoie and decanoic acids were identified. Parliment et al. (1973) discovered the presence of rans-2-nonenal, a compound which imparts ‘a fresh-brewed woody character to roasted and ground coffee’. Kung (1974) described the identification of 3-hydroxy-3-penten-2-one with a pleasant, buitery caramel aroma. This ‘compound’ [Section 5, D.45B] is, in fact, the tautomer form of the already found 2,3-pentanedione [Section 5, D.45A] Chronologically, the results for the reseurch eam (1) of the ‘modern period’ appeared only in 1967. As reminded in the publication of Stoll e¢ al. (1967) (Firmenich et Cie, Geneva, Switzerland, see Table 4.4, the analytical project on coffee Mavor started in 1957 but, for commercial reasons, the publication of results already obtained in 1965 was postponed to 1967. In the meantime, first results on the GC separation of acids were published by Gautschi (1958). At the 3rd International Colloquium on Coffee Chemistry (Trieste, 1967), Gautschi et al. (1968) presented a survey of the research of coffee arom discussing the concentration steps, the separation procedures and the identification techniques. An original, simple and efficient analytical method coupling gas chromatography with thin-layer chroma~ tography (Flament et a, 1967b) was also mentioned. Flament ¢f af. (19682) developed some chemical and spectroscopical aspects of the furanic components. The various origins of the furanic cycle were considered, as well as its transformation into other heterocyclic substances already identified or still undefined. Two complementary papers published successively by Gautsehi er af. (1967) and by Stoll e¢ al (1967), mention the identification of 203 cofiee constituents, nearly 150 being identified for the first time. The most represented classes were hydrocarbons (13), ketones (22), phenols (10), furans (37), thiophenes (17) and pyrroles (9). The same research team (Goldman er ul., 1967) found 18 pyridines and pyrazines The publications of Stoll and Goldman, which revealed the results of intensive analytical work carried out over more than 7 years, were immediately followed by the disclosure of numerous other identifica- tions obtained during the same period by other research groups, The syntheses. the physicochemical and spectroscopic characteristics, as well as the organoleptic properties of identified and potentially active constituents of coffee favor have been published in 42 patents between 1968 and 1980 under the names of Winter er al, or Firmenich & Cie (sce References). Biichi et al, (1971) confirmed by synthesis the structure of kahweofuran, a key constituent of coffee flavor discovered by Stoll e7 ai. (1967). For Broderick (1968) a new approach was needed in spite of the very promising results obtained at that time. Considering that attempts to reproduce coffee aroma synthetically had failed, the author rightly affirmed that isolation and separation techniques could alter, modily, or inactivate the key aroma compounds. He assumed that some unknown constituents, particularly among the highly volatiles. are probably extremely unstable and rapidly deteriorate, He refered to other fugacious components probably present in onion and meat and gives, as an example, the identification of propanethial (propy! thioalde- hyde) in fresh onion flavor, This compound, which is particularly unstable, polymerizes rapidly and cannot he isolated or used by flavorists, By analozy he deduces and feels that 2-furancarbothialdehyde(improperly named furfuryl thioaldehyde) has all the theoretical flavor properties of a freshly brewed n Coffee Flavor Chemistry coffee flavor. He does not minimize the difficulty in preparing and stabilizing 2-furancarbothialdehyde and recognizes that it may be impossible. He concludes that the production of an instant coffee with a full fresh roasted flavor is a practical impossibility without added complementary flavors. In two publications, Stoffelsma and Pypker (1968) and Stoffelsma er al, (1968) (Polak’s Frutal Works, Amersfoort & Douwe Egberts, Utrecht, Holland, See Table 4.4) give a list of 158 compounds, 30 being reported for the first time in coffee, Among them figure five esters, five lactones and five furans. The aroma compounds were isolated from a solvent extract of the steam condensate of roasted and ground coffee. They were identified by comparison of their IR spectra, of their retention times on two GC columns and, in a number of cases of their mass specira, with those of reference samples. Other contributions were published by Feldmann es al, (1969) who added two esters, and by Haberer (1970) and Kiécking e7 af, (1971) who added five phenols to the lists, From the seventies, Hag AG, Jacobs-Suchard Corp. R&D, Kraft Jacobs Suchard, Bremen, Germany (Table 4.4) shared in the knowledge of coffee aroma. In their important work, Vitzthum and Werkhof? (1974a,b) enriched. by adsorption on Porapak Q, and analyzed, by combination of glass capillary GC- MS, the basic and neutral components of a steam-volatile extract of roasted coffee. The number of identified components increased by 80, among which were 7 pyrroles, 20 oxazoles, 25 thiazoles. 6 pyridines and 20 pyrazines. These publications point to the role played by oxazoles and thiazoles in food flavors. Nevertheless, the authors realized that, although the number of volatile components of coffee reported until then way relatively high, they only represented the tip of the iceberg. Also in 1975, the same authors identified 17 alkylated cyclopenta- and eyclohexapyrazines (tetrahydroquinoxalines) in the basic fraction of a normal pressure steam-distilled coffee extract. Possible precursors of these heterocycles were discussed and the principal MS fragmentation pathways were demonstrated. Among the 120 components identified by Vitzthum and Werkho'T (1976b) in the neutral fraction, 29 were not yet known components. Methylated vinylfurans, alkenylfurans and V-acylated pyrroles, which had not been previously found in roasted foodstuffs, are specially mentioned. Mass-spectral data for the newly identified substances were tabulated and discussed. An exhaustive table of the constituents or raw and roasted coffee identified 25 ycars ago was published by Vitzthum (1976). Gutmann er ai. (1979) compared the volatile components of roasted arzbusta (hybrid from the Ivory Coast) coffee with arabica and robusta by enrichment of the headspace volatiles on Tenax polymers and GC/MS identification. The aroma profile of arabusta is very similar to that of arabica, confirming the close relationship of their flavors, Moreover, the robusta shows more GC peaks than the other varieties. The authors mentioned LL furanic compounds which are more abundant in arabica and arabusta and four sulfur constituents which predominate in roasted robusta, Becker ef al, (1988) investigated the origin of a ‘peasy’ off-flavor note present in Central African coffees, defective green beans impairing considerably the flavor of the final roasted product. They used a eryogenic headspace enrichment device coupled with a double-oven GC) MS system. They located two peaks, one having a “peasy” note, the other a “bell pepper” character, which they respectively identified as 3-methoxy-2-isopropylpyrazine and 3-methoxy-2-isobutylpyrazine, Both compounds resist roasting conditions and are present, although in lower concentration, in ordinary ‘non- peasy’ Colombian coffee. Holscher ef af, (1990) investigated aroma extracts from roasted Colombian coffee, by means of a HRGC-sniffing technique recently introduced by Grosch ei af. (Ullr Grosch, 1988a; Gasser and Grosch, 1988, Blank ef af., 1989). Of the 61 peaks present in the GC, 49 constituents were characterized by an odor description and a flavor dilution factor. This technique confirmed that only a comparatively small proportion of the volatiles is significant for coffee flavor. In fact, only nine compounds have a very powerful impact. some not yet having been reported in the literature. Among them figure 2-methy!-3-furanthiol and methional which exhibit a strong meat-like flavor, and 3+mercapto-3-methylbutyl formate which js described as having a ‘catty" or black-currant-like odor. 2-Furanmethunethiol, discovered by Reichstein and Staudinger (1926b), was again recognized asA Historical Survey of Coffee Aroma Research 73 one of the most typical constituents of roasted coffee. Furaneol, B-damascenone and, surprisingly. 3-methylbutanoie acids were recognized as contributing remarkably to the flavor. Typical 3 methoxy-2- isopropy!- and -isobutylpyrazines, previously mentioned as being present in green beans by Vitzthum er ul. (1976) and by Becker et al, (1988), were easily located. Holscher er al. (1990) also observed the presence, at the GC-snilfing port, of several other interesting spicy and catty odors whose nature remained unidentified. Vitzthum ¢1 af, (1990) compared the GC profiles and the sensorial differences of arabica versus robusta. They observed that robustas have a typical harsh, rubbery. often earthy and mouldy character, whereas arabicas exhibit a milder. finer and more aromatic flavor. The authors reviewed the analytical studies which had shown that robusta coffees contain higher concentrations of phenols, pyrroles and sulfur compounds. A GC peak corresponding to the very typical earthy musty note of robusta varieties was located and identified as that of 2-methylisoborncol. This key compound has an exceedingly low threshold, corresponds to the descriptor ‘robusta’ in sensorial profile tests and is present at higher concentrations in robusias than in arabices by a factor of at least 3. Bade-Wegner ef al. (1998) tracked the ‘stinker’ constituents of over-fermented green coffee which are associated with an intolerable fruity, silage-like, rotten’ Navor. Their work revealed that three ethyl esters of short chain or alicyclic acid, ethyl 2- and 3-methylbutanoates (2-MBEE, 3-MBEE) and ethyl cyclohexanecarboxylate acid ester (CHEE, improperly named cyclohexanoic acid ethy! ester), are the actual or potential key aroma compounds responsible for the over-fermented flavor defect. Critical factors, such as elevated moi levels and temperatures, might support growth of microbes that generate the off-flavor during shipping or storage. The roasting of the spoiled beans reduces but does not eliminate the light volatile 3 MBEE. and it does not affect the levels of CHEE. The threshold value of CHEE is particularly low (0,005-0.01 pg/L = 510 ppb). Consequently its aroma value in the beverage is relatively high and it may be directly related (o the ‘stinker’ bean phenomenon. Boosfeld and Vitzthum (1995) identified four aklehydes from green coffee, some also being present in roasted coffe Singer and Lijinsky (1976), Neurath et af. (1977) and Amorim er al, (1977) identified some amines and polyamines, Rahn and Kénig (1978, 1980a,b) and Rahn er al. (1979) some original phenols and acid Heyns (1979) reviewed the scientific aspects of coffee research. Tresst ef al. (1978a,b) and Tress! (1979a) (Institute for Biotechnology, Technical University. Berlin, Germany. See Table 4.4) isolated 30 phenols from roasted cotfee by distillation-extraction. Separation from the other constituents was carried out by adsorption chromatography on aluminum oxide and the identification by capillary GC/MS. The authors observed that the robusta contains the largest amount of phenols (mainly phenol, guaiacol and 4-cthyleuaiacol) and maltol, followed by arabusta and arabica varieties. They also note that the quantity and the types of phenols depend on the coffee variety as well as on the roasting conditions. New identified components comprised 15 phenols, three furans and two pyrans. Arabica contained higher amounts of these heterocycles than the other varieties. 4-Hydroay-2.5 dimethyl-3(2)-furanone (Furancol®), the 2- or S-cthylhomologue (ambiguously called ethylfuraneol), jsomaltol (2-acetyl-3-hydroxyfuran), 5-hydroxymaltol (3,5-dihydroxy-2-methyl-4-pyrone) and S-hydroxy +5,6-dihydromaltol (2,3-dihydro-3.5-dihydroxy-6-methyl-4-pyrone) have an agreeable and characteristic caramel note (see Section 5.1). ‘The presence of pyrroles in roasted coffee was studied by Tressl ef af, (1981a) who added a list of 27 coffee components. After a special separation process involving distillation -extraction of freshly roasted coffees. adsorption chromatography on silica gel, and capillary GC/MS, more than 20 N-alkylpyrroles and 15 furfurylpyrroles were identified and (semi)quantified. Robusta coffees contain higher amounts of alkylpyrroles and lower amounts of furfurylpyrroles than arabica varieties. The authors proposed possible formation mechanism by reaction of furfurylamine with sugar fragmentation products Tressl and Silwar (1981) investigated the sulfur-containing components of roasted coffee. Twen| three mercaptans, sulfides, die and trisulfides were characterized, 15 of these products being newcomers.4 Coffee Flavor Chemistry The structures were confirmed by synthesis. Among the most original compounds figure homokahweo- furan (2.4-dimethyl-3-oxa-8-thiabicyclo[3.3.0]-1 d-octadiene), a homologue of the typical kahweofuran identified by Stoll er uf. (1967), 3,3-dimethyl-l,2-dithiolane and 3,3-dimethyl-4-oxo-1,2-dithiolane Silwar (1982) identified numerous components which have been published with Tress in various quoted references. Baltes and Bochmann (1986a, 1987b.¢.d.c) (Institute Food Chemistry, Technical University, Berlin, Germany. Sce Table 4.4) intensively studied model reactions on roast aroma formation. They particularly reported the reaction of serine and threonine with sucrose, which are present in green coffee. under the conditions of coffee roasting. From the reaction mixture they identified about 350 compounds including furans, pyrroles, pyrazines, pyridines and oxazoles. By comparison of the compounds with the compon- ents of an original coffee aroma, thcy identified $3 aroma components. Among them were present, sometimes in very low concentrations, 12 furans, 9 pyridines, 15 pyrazines (alky]- and alkenylpyrazines, pyrrolo[l,2-alpyrazines and furfurylpyrazines) and 2 oxazoles. Heinrich and Baltes ( 1987a,b) identified 14 phenolic constituents in four commercial coffee samples and particularly in robusta varictics. Nishimura and Mihara (1990) investigated 2-hydroxy-2-cyclopenten-|-ones in a steam-distilled coffee extract and identified ten compounds of this elass, seven of them being reported for the first time in a natural flavor and some cven being original chemicals. Procedures used for their syntheses are described. All the 2-hydroxy-2-cyclopenten-I-ones were also identified in a model reaction, when sucrose, the major component of green coffee, was heated with alkali in aqucous solution, Some had been previously found in maple syrup or tobacco. We are reminded that Giantureo e¢ al (1963) mentioned for the first time the presence of these typical, sweet, caramel-like and burnt compounds in a roasted coffee extract Other analytical works including original identifications have been published by Maier (1987, 1988), Guyot e1 al. (1988b). Kallio er al. (1989) and Shimoda and Shibamoto (1990a), Hills er al. (1991), Blank er al, (1992a), and Leino et al, (1992) have also contributed to the identification of roasted caffe constituents Lee et al. (1992) monitored by GC-MS the compositional changes in 44 volatile components of a coffee beverage as @ function of brewing time, concluding that the maximum time should be 10-1 min. Ho e¢ ai, (1993) reviewed the relationship of Maillard reactions with aroma generation in coflec. Ten constitu- ents, mainly furans, were identified for the first time, Wahrmann ef al. (1997a,b). particularly interested in the rancid and sweaty smelling notes. identified five volatile minor acids in raw and roasted coffees Ramos ef ul, (1998) compared supercritical Muid extraction (SFE) of a brewed coffee with other extraction techniques, liquid-liquid extraction and headspace solid-phase microextraction (HS- SPME). There was a high olfactory resembiance of the supercritical-CO: extract to the original brewed coffee. Fifteen components were identilicd. For about 15 ycars, analytical research has turned to more selective approaches, essentially based on sensory methods. The research group of Grosch (Garching, Germany) has been particularly aetive in this field by turning the xttention 10 the qualitative aspects of the already identified and most-contributing constituents, The Nestlé group has also been very much involved in this field. More details are given in Section 3.3. This overview of the most recent studies on coffee arom chemistry is inevitably full of matter and tediously enumerative. The reader will find more specific details and discussions in Chapter 5. 43 RESULTS Figure 4.1 illustrates the progressive identification of the nearly 850 volatile constituents discovered in roasted coffee flavor. It shows that at the advent of chromatography (1956) 60 compounds were already known thanks 10 the remarkable contribution of Reichstein and Staudinger (1926) (see Table 44) A Historical Survey of Coffee Aroma Research 75 a a a el470 sane ota 1987] 5 oo Trea ane Sor (9KI-82) E 500 | é Vitathum and Werkhot¥ (1974-76) = aon 5 Tina eal. 1967) — a Stoll eral. (1967) 2 200 Giantreo etl (1964-6) ‘Merritt eral. (1963-70) Reichstein and Staudinger (1926) 0 se 1960 1962 1964 1966 1968 1970 ora 1976 1938 1972 Fig. 4.1 Identification of roasted collec flavor constituents Nearly 100 scientific papers mentioning original structure discoveries have been published during the last 40 years. Table 4.4 showed the research groups who, during the last half century, found the structures of nearly 80% of the volatile constituents identified up to now. Table 4.5 details the functions of the constituents identified by these groups. Under reserve of possible statistical omissions or crrors, five each discovered more than 50 volatiles. Group | (Stoll ef al., Firmenich) found a wide variety of ketones and of numerous furans, pyrazines and thiophenes. Table 4.5 Functions of the compounds identified by the different research groups Function 1 2 3 4 6 7 8 9 — W Tolal/Function A Hydrocarbons «B07 2 GCG 4s B Alcohols 8 2 tT 0 6 3 0 1 @ 4 19 © Aldehydes 4 3 - 1 0 4 2 0 3 3 21 D Ketones 2 4 5 HB 1 3 3 3 3 2 2 E Acids 1 © 0 2 0 O@ t 3 7 0 4 F sters 5 1 1 3 6 2 2 5 1 1 2 G —_Lactones 6 0 0 F 1 0 O09 5 too 8 H Phenols we | 6 2 Ww © 5 0 2 9 a To Fornsand 37D aD 126 pyrans J Thiophenes 7 5 0 1 0 0 0 2 oO 1 %6 K Pyrroles 9 4 B 8 2 3 2 2 oO 1 9 L_— Oxazoles 0 ~@ 0 0 2 © oo t 0 0 2B M — Thiazoles 2 3 0 © 0 6 6 8 0 Oo Pi N Pyridines 1 8 0 1 9 0 0 0 0 9 » © Pyrazines 17 3% 0 8B B 0 4 2 1 9 a1 Po NCompounts 0 0 0 4 0 2 9 O 1 1 5 Q SCompounds 3 3 8 O GO 4% 4 FT 4 4 23 TotavGrou 149459 LH BB 648 6 Coffee Flavor Chemistry Group 2 (Vitathum et al., Hag-Kraft-Jacobs-Suchard) identified numerous heterocyclic compounds, mainly pyrazines, thiazoles, oxazoles and furans. Group 3 (Tressl et al. Berlin) identified mainly furans, pyrroles, phenols and sulfur-containing compounds, emphasizing particularly their formation pathways. Group 4 (Gianturco et al., Coca-Cola) was the first to identify caramel compounds such as cyclicketones and to reveal the organoleptic interest of pyrazines, furans and pyrroles. Group 5 (Baltes ef ai., Berlin) contributed to the isolation of many pyrazines, phenols, furans and pyridines by the systematic study of mode! reactions. Figure 4.2 shows the distribution according to their functions or structure of the constituents of roasted coffee flavor. Furanic compounds are predominant. The other main constituents are pyrazines, Ketones, pyrroles, phenols and hydrocarbons. Figure 4.3 compares the number of compounds identified in green and in roasted coffee beans Green coffee contains a larger number of identified alcohols (B) and nearly the same number of identified aldehydes (C) and esters (F) than roasted beans. On the contrary, the latter are richer in furans (D), pyrazines (0), ketones (D) and phenols (H). Thiophenes (J), oxazoles (L) and thiazoles (M) have only been identified in roasted coffee. The roasting effect is also revealed by the increased number of pyrroles (K) and sulfur-containing compounds. The distribution of aroma volatiles is shown in Figure 44. gr ee, 2 Enctnedbscs : re 3 meee 2 eo 4 nate 2 Navin Ne. Wenning cours Mise Semuinrg saps ABCDEFGHISKLMNOPQ Functions Fig 4.2 Volatile constituents of roasted coffee flavor 160 — = 140) 120 100 0 Green B Roasted Number of identitieations 4.3 Green and roasted coffee volatiles A Historical Survey of Coffee Aroma Research 7 Green 100 Total in green: 300 Green and roasted Roasted eo Total in roasted: 850Fig. 4.4 Distribution of aroma constituents in green and roasted cofleeA coffec-fountain in the streets of London (in about 1860) The Individual Constituents: Structure, a NN eR eee lll BD eee ee ee,INUTTIGTIGIOLUIG, VIII, VERO aris Organoleptic Properties In the following sections, SA to 5Q, of this chapter, the compounds will be identified by a letter corresponding to the section followed by a digit for the individual structures. The compounds identified in roasted coffee are given upright bold characters and are in bold italies when identified in green coffee. Each identification number will be followed by the name with which it is indexed in Chemical Abstracts, in bold characters, then one or more of the names more generally accepted by IUPAC (International Union of Pure and Applied Chemistry) rules with the 1993 recommendations, in izalies, together with other synonyms and trivial names under which the compounds can be found in literature and whieh may be sometimes confusing, The Register Number in Chemical Abstracts will follow (very few have not been found), in bold characters and, where possible, the FEMA number in bold italies. We remind readers that the IUPAC agrees that indexes, like Chemical Abstracts or Beilstein, need more ‘stringent requirements’ than are usually necessary. Moreover the application of the principles ‘will not necessarily lead to a unique name, but the name obtained should be unambiguous’. For cach compound, we have tried to identify the original discoverer in coffee and we also give, in a non-exhaustive manner, a list of some other references, sometimes with methods of identification, concentration in coffee and thresholds. Some of the model reactions in which the compounds are formed will be quoted, then some syntheses, and eventually the organoleptic properties. We will avoid acronyms as much as possible. Anyway a list is given in Chapter | 5.A HYDROCARBONS Lower alkanes and alkenes with straight or branched chains have been found in coffee volatiles. For example, Zlatkis and Sivetz (1960) drew attention to the presence of low molecular weight hydrocarbons in a coflee aroma essence (Cy-Cy paraffins and olefins), but propane and heptane, which have not yet boon identified in coffee, are probably also present. These small molecules are certainly produced during the roasting of the beans, an operation generally conducted at temperatures near 200°C, relatively higher than those used for the thermal treatment of cocoa beans and tea leaves that are, consequently, poorer in 82 Coffee Flavor Chemistry hydrocarbons. Other mechanisms can be suggested since a few hydrocarbons are actually present in favors as primary metabolites (in green tea, for instance). and others can be formed enzymatically (in black tea and cocoa) or by various pyrolysis transformations. According to the observations of Merritt ef al, (1970), bydroearbons could be formed by oxidation of fais before roasting, the process itself having little effect, although some compounds may disappear and some olefins may be produced. The organoleptic character of hydrocarbons has received little attention in spite of the fact that compounds such us hexane or cyclohexane have a detectable odor. Boelens (1974) reported that the members of a panel could not make any distinction between Cy)- to Cy ~alkanes and the corresponding aliphatic aleohols, On the contrary, polyunsaturated hydrocarbons possess typical odor qualities and may therefore be important contributors to food flavors (Ohloff, 1978a) but their presence in coffee is limited to aliphatic volatile compounds, such as pentadiene (A.41) and isoprene (A.44), and to S-methyl- 3-cyclohexadiene (A.47), not forgetting the lerpenes mentioned later. Nevertheless the flavoring power of these paralfins is certainly negligible as compared with the most characteristic constituents of coffee, cocoa, and tcu. Monoterpenes, on the other hand, possess particularly desirable flavor notes. Three of them, + myrcene (AS), limonene (A.46) and p-cymene (A.59) are common to coffee, cocoa and tea, and to numerous plant families: 6-myreene for example constitutes 63% of hop oils. It is worth remembering that they particularly contribute to the flavor of spices. In a study of the effluvia of fresh red cofiee berries, Mathieu e¢ af. (1996) identified, in two varieties of robusta, a scries of mono- and sesquuiterpenes, a blend also: present in cold pressed oil from pommelos. In fact, the effluvia of red berries attracts thefemales of the coffee berry borer, a very important pest. Without giving a list of the hydrocarbons identified in this study, let us note that limonene is the most abundant of them, caryophyllene (4.11.1 + trimethyl-8-methylenebicyclo[7.2.0]undec-4-ene), followed by humulene (2,6,6,9-ietramethylundeca- triene) and e-pinene (2,6,6-trimethylbicyclo[3.1.I}hept-2-ene), are rather important in two robusta So Co i COO ‘Muoranihene pyrene ehrysene ‘ene{elanthracene otto od a ‘eral -Jacephenanthrylene pervlene benzofelnyrene O ak ov~ CO ibena{a.Aanhacene ‘benzo hilperslene \ é The Individual Constituents 83 varieties. In one arabica variety, only caryophyllene and humulene are identified and no monoterpenes. The amount of terpenic hydrocarbons decreases drastically when the berries are leit on the tree after the red stage (Mathicu et af., 1998), and only limonene is still emitted in appreciable quantity by a robusta, accompanied mainly by p-cymene and pinenes and the two sesquiterpenes quoted ahove. With an arabica (grown in a greenhouse) some limonene was detectable after longer ripening on the tree but no sesquiterpene was identified. Parallel to the decrease in these hydrocarbons is the increase in oxygencited compounds by ripening. Recently, Cantergiani et af. (2001) identified kaur-I6-ene, a tetracyclic diter- pene, in the volatiles of a green Mexican arabica coffee. Being interested here in the volatile components of coffee aroma, we shall arbitrarily limit the list of the aromatic hydrocarbons to tricyclic structures. The higher fused polycyclic hydrocarbons (fluoranthene [206-44-0], pyrene [129-00-0}. chrysene [218-01-9}. benzfuJanthracene (1,2-benzanthracene) [56-55-3), ben7{elacephenanthrylene (3,4-benzofluoranthene) [205-99-2}, benzofajpyrene (3,4-benzopyrene, 3,4-BP) 150-32-8], denzole]pyrene (1,2-benzopyrene) [192-97-2}, perylene [198-55-0], benzolg.h,/perylene (1,12- benzopyrenc) [191-24-2], and dibenz[a,hJanthracene (1,2,5,6-dibenzanthracenc) [53-70-3}) cannot be considered as a part of the aroma. However, as some of these, specially benzofu}pyrene, are known for carcinogenic properties, they have been particularly analyzed in food subject to roasting or smoke-curing, ‘A work of Kuratsune and Hueper (1958) quotes concentrations of benzolalpyrene as high as 0.44 ppm in soots of torrefaction but the same authors (1960) identified (by extraction and chromatog- raphy) nine hydrocarbons including benzofa|pyrene (with a large predominance of pyrene and fluoranthene} only in a very dark roasted coffee that was nearly black at the surface, In other commercia! brands only pyrene and fluoranthene were identified. Chassevent and Héros (1963) found no benzolal- pyrene or concentrations lower than 0.1 ppb in arabica and robusta coffees from New Caledonia, green, or light to dark roasted. After grinding, they extracted with petroleum ether, and used chromatographyEe ER RN, ae a EE Mace SMa See) See, Stee nee ee, Saeeterer ee (0.15 ppm) and in soots of industrial torrefaction (0.26 0.28 ppm), After analyses of brewed coffee. Grimmer and Hildebrandt (1966) found that the main aromatic polycyclic hydrocarbons were phenan- threne (4.77). pyrene and fluoranthene but that the concentration of benzojajpyrene was less than 0.4 ppb, ten times lower than in Indian black tea. These concentrations were similar 10 those found in other food, like bread or smoked products. Calzolari and Pertoldi-Marletta (1968) also analyzed the aromatic hydrocarbons (extraction with petroleum ether and separation of unsaponifiable matter) in roasted coffee at different temperatures, and in the fumes, The concentrations were similar in Columbian, Haiti and Santos samples, At 180 C, they only detected the 3.4-benzofluoranthene (0.4 ppb) but, for the seven of the eight hydracarbons studied, the concentrations increased at 210°C and even more so at the over-reasting lemperature of 240°C, benzofluoranthene being still preponderant (ca 5 ppm at 210°C). Benzofalpyrene (3,4-BP) was only detected at 240 °C (ca 5 ppb) and in the fumes of the corresponding roasting (ca 0.5 ppb). Decafl mn apparently brings a diminution in the concentration in hydrocar- bons (of the order of 50%). Baltes (1979b), in a publication on the effect of roasting on coffee composition, noted an increase of polycyclic hydrocarbons with roasting. At medium roast, phenan- threne (A.77) is by far the most important (28-30 ppb), when benzolalpyrene is only present at 0.3 0.5 ppb; at extremely dark roast, phenanthrene is still the most abundant (ca 70 ppb), followed by fluoranthene (ca 35 ppb) and pyrene (ca 30 ppb), the concentration of benzola}pyrene being about 6 ppb; other hydrocarbons are in the range of 0.6-15 ppb. Strobel (1974) proposed a method for the determination of benzo[a]pyrene in coffee products (brew, oils and commercial extracts) and other foodstuffs, A combination of extraction, paper chromatography and fluorometry enabled the detection of less than 10°! g of benzofalpyrene and measurement of 10-?g. Such a sensitive method demands special attention for possible contaminations. Strobel measured 0.003 ppb in a brewed roasted coffee, 0.02-0.06 ppb in commercial coffee extracts and 0.5ppb in coffee oil. The amount is relatively low 84 Coffee Flavor Chemistry compared to other foodstuiTs. It is suggested that the reason could be that coffee is grown in places where there is not too much pollution, and also that the roasting process does not involve severe pyrolysis. Van der Stegen and van Overbruggen (1983) also studied coffee roasting and benzo[a]pyrene. They concluded. that if burning is avoided, the content is about 0.2 ppb. The 47 commercial samples analyzed were all under | ppb. Husks which are left or added contain more benzopyrene, therefore that should be avoided Eventually, consumption of coffee could contribute to less than 0.4% of total daily consumption, including other food and drinks without even taking pollution into account. Holscher and Steinhart (1995) considered a surprising amount of a great number of hydrocarbons in roasted coffee and attributed this to contamination by the bays used for storing and transportation of the beans, the jute or sisal being treated with batching oils before spinning in order to make the fibers more clastic. The work of Grob ef al. (1992) shows that this contamination is not negligible, coffee beans sometimes containing about 100 mg of oil/kg (using jute bags). Nevertheless it depends on the oil used. Some raw mineral oils contain more aromatics (>10 %). with aliphatic hydrocarbons, nearly exclusively isoalkanes. from C7 to at least Cy; (mainly C5~-Cp). Other oils contain mainly isoalkanes and about 3% of aromatics and others are in fact waxes consisting of alkanes, mainly C2, to Cy. The authors made several recommendations: reduction of the batching oil on the bags, use of oils free of aromatics and use of oils not containing alkanes with fewer than 25 carbon atoms, the transfer to the packed food being more efficient for volatile compounds. In analysis of a brew, Ramos er al. (1998) detected an appreciable amount of hydrocarbons in a liquid-liquid extract with pentane, and in an extract with supercritical CO>. They agrced that this presence was probably due to contact with materials used for transportation or storage. In any case, these compounds do not contribute to the genuine coffee aroma, Remarks concerning the validity of some identifications The following lists are deliberately limited to confirmed identifications, the compounds mentioned in the literature with the mention ‘T (tentatively identified), *” (questionable identification) or *unkn.str.” (unknown structure) being usually commented on but the structures deliberately omitted trom the illustrations. Sometimes, nevertheless, doubts remain concerning the actos! structures of many oe te es