The chemistry of coffee flavor formation during roasting has long been a mystery. This is mainly due to the multitude of possible reactions taking place at high temperatures when coffee biopolymers are degraded to smaller, usually more reactive entities. Today thanks to the development of more sophisticated analytical methods allowing for higher sensitivity, many key aroma and taste components have been identified and their sensory relevance evaluated.
However, still much needs to be further discovered. Although the myriad of aroma compounds in roasted coffee has been elucidated and their impact on the overall aroma quite well described, the taste dimension still requires more molecular understanding. The coffee taste cannot be fully reconstituted by using known taste-active compounds, indicating that there are additional tastants to be identified, in particular those contributing to bitterness, harshness, and acidity. In addition, still, a lot of experimentation is needed to better understand the reactions leading to flavor formation, i.e., from precursors to aroma and taste. With a focus on aroma and taste formation, it seems desirable to determine the amounts of precursor molecules of aroma compounds (such as amino acids, sugars, CGA, etc.) before and after roasting to be able to provide objective statements about the roasting degree, which are transferable from one roasting profile to another and do not depend on raw material fluctuations of the green beans.
The complexity of the green coffee composition as well as the chemical and physical transformations of the bean during roasting are difficult to reproduce in model systems. The approach of in-bean experiments using green beans as mini reactors is a more realistic reaction environment.
The combinations of omission, spiking, and mechanistic experiments under real food matrix conditions are very useful in providing further and more precise insights into Maillard-type reactions and formation mechanisms. The results of different studies clearly indicate that due to the great diversity of precursors and other co-reaction agents present in the green bean, competing and even completely different pathways may take place in the formation of flavor compounds. More research is needed on flavor formation kinetics under controlled conditions, potentially using real-time analytical approaches combined with characterization of the changes in flavor precursors to relate green coffee composition to flavor profile as affected by roasting parameters. Science is helping to understand the mystery around coffee roasting. Description of chemical changes occurring in the coffee bean at a molecular level is of increasing value for adapting the roasting conditions, which today is largely based on empirical knowledge of the barista. We are convinced that a better understanding of science is of equal importance to the barista and relevant to “the craft of coffee”.
The flavor of roasted coffee depends on (1) the green bean constituents and (2) the way the roasting operation is conducted. In general, the precursor composition determines which flavor compounds are formed, whereas the physical parameters mainly influence the formation kinetics.
The quality of the green coffee is the main determinant of the aroma and taste developed during the roasting process, and it is a snapshot at a defined roasting degree. However, there is no concise definition of the roasting degree, although expressions such as “optimum degree of roast” are frequently used in the literature. It seems obvious that the optimum degree of roast is a function of green bean origin, intended brewing method, and personal taste preference. It is widely accepted that characterizing the quality of roasted coffee only by means of weight loss and/or roast color is not sufficient since these attributes do not make a statement about the individually obtained aroma profiles.
The experienced coffee roaster knows that the aroma evolves from sweet, fruity, floral, bread, and nutty character in light roasts, through more complex aroma profiles in a medium roast. Darker roast levels are characterized by cocoa, spicy, phenolic, ashy, pungent, and dark roast flavors. The bitter taste increases during roasting, whereas the acidity decreases during the initial stages of roasting (Fig. 1). This sensory perception is substantiated by sensory and instrumental analysis of aroma and taste components throughout the roasting course. The flavor composition continuously changes all along with the roasting progress. This means, at each time a new aroma/taste profile is delivered. However, the perceived aroma goes through an optimum, i.e., over-roasting will lead to unbalanced burnt, harsh off-notes comprising both aroma and overall flavor (body). In general, the sensory perceivable total acidity is decreasing during the course of roasting, whereas bitterness is steadily increasing. Similarly, color development is increasing with a roast degree from light brown [color test number (CTN) 150 to almost black (CTN 50).
Taking the main constituents of coffee beans into account, Table 1. compares the composition of the green beans to the roasted beans and points out which compounds of the green beans are particularly degraded during roasting. Sucrose and free amino acids are immediately available and highly reactive. Their high reactivity is explained by the presence of a free functional amino group and the rapid thermal hydrolysis of sucrose into reducing sugars. Arabinose branches and CGA are degraded in a later stage. This delayed reactivity can be explained by the additional energy needed for depolymerization or hydrolysis to liberate the reactive functions. Other polymeric carbohydrates (i.e., galactans, mannans, and cellulose) or bound amino acids are less prone to hydrolysis and depolymerization, and thus only contribute at a later stage of roasting to the Maillard reaction.
Excessive roasting (CTN<50) generally leads to decreasing in many volatile substances such as diketones, furfurals, or 4-vinyl guaiacol. These compounds are formed in the early stage of roasting, mainly parallel to free sugar and amino acid consumption, with a maximum at a medium roasting degree. For example, the fruity, blackcurrant-smelling sulfur odorant 3-mercapto-3-methyl butyl formate can completely disappear in a dark roasted coffee. Alkyl pyrazines (e.g., 2-ethyl-3,5-dimethyl pyrazine) responsible for earthy, roasty, nutty character of coffee are as well generated in an early stage, but levels remain almost constant throughout the final stage. In contrast, 2-furfurylthiol, the trigonelline degradation products pyridine and-methyl pyrrole, as well as dimethyl trisulfide were shown to continuously increase with roasting time. All these molecular changes have their impact on the sensorial characteristics. A masking effect by odorants formed in the later stages of roasting, covering the sweet and earthy notes, was described by Gretsch et al. (1999), and fruity, floral notes appearing at beginning of roasting are replaced by roasty and burnt notes.
As discussed before, precise control of roasting time and the temperature is required to reach specific flavor profiles. The generation of flavor compounds depends on the duration and the final bean temperature reached. Both parameters are constituted by the slope of the roasting curve, i.e., the time-temperature (t-T) profile. Based on the same raw material and same roaster, roasting coffee to the same roasting degree leads to different flavor profiles depending on the time-temperature roasting conditions. Short-time roasting at high temperatures has been shown to result in considerable differences in the physical properties and kinetics of aroma formation compared to a long time roasting at lower temperatures. Fast roasting yields more soluble solids while causing less degradation of CGA, and lower loss of volatiles. This is accentuated in higher quantities of roasty, buttery diketones, and furfurals, whereas much lower concentrations of phenols are formed, leading to less burnt, smoky flavor. On the other side, fast roasted coffee is presumed to be more affected by lipid oxidation due to higher oil migration from the inner bean to the surface.
The impact of different roaster technologies on aromatic composition is less well understood. A similar temperature profile applied on two roasters (laboratory scale fluidizing-bed roaster and traditional drum roaster) resulted in similar physical properties and aroma formation in the assessed coffees.
The evaluation of aroma formation in Arabica and Robusta coffees upon roasting is quite similar, but concentration differences between the two coffee varieties are crucial for the final organoleptic characteristic. Green Arabica coffee contains more oligosaccharides, lipids, trigonelline, and organic acids. On the other hand, Robusta is significantly richer in caffeine and CGA and also exhibit a larger amount of free amino acids than Arabica (see also Table 1).
Slightly lower amounts of the precursor isoleucine and leucine in Arabica green beans lead to smaller final amounts of the corresponding Strecker aldehydes (i.e., 2-methylbutanal and 3-methylbutanal) in roasted Arabica beans compared to Robusta coffee. Accordingly, higher amounts of amino acids in Robusta green beans result in higher final amounts of earthy, roasty, nutty smelling pyrazines. Together with the high level of phenols resulting from the CGA degradation, Robusta coffee exhibits their typical smoky, earthy, roasty, and phenolic aroma profile. Tryptophan, an amino acid strongly present in Robusta green beans, resulting in higher final amounts of undesirable, animalic smelling 3-methyl indole (skatole), which is almost not present in Arabica coffee. The higher levels of diketones, furfurals, and cyclic enolones (i.e., furaneol) in roasted Arabica coffees are a consequence of the higher abundance of sucrose in the initial green beans. Interestingly, variability in aroma formation kinetics is not only found between Arabica and Robusta formation, but also within a coffee species as recently reported for different Arabicas from Colombia, Guatemala, and Ethiopia. One way of explaining the varying behavior of coffee of different origins under the same roasting conditions might be the individual changes in physical structures.
The initial moisture content represents another important parameter, which has an influence particularly on light roasted coffee, whereas in dark roasted coffee most differences in the aroma are leveled off. Steam treatment of green coffee beans represents a method to improve the flavor quality of Robusta coffees. They get sweeter, more acidic, and less bitter than untreated coffee, lowering significantly the Robusta character. Water treatment with saturated steam at elevated pressure provokes some changes in the precursor composition, mainly on sugars and amino acids, influencing the color formation and diminishing formation of undesired substances (e.g., catechol, pyrogallol, and hydroquinone, pyrazines, volatile phenols. These changes can be explained by mobilization of precursor resulting in partial extraction and removal of water-soluble such as free amino acids, or CGAs, or partial hydrolysis of the sucrose to fructose and glucose. Decreased amounts of feruloyl quinic acid result in the lower generation of 4-vinyl guaiacol; extraction of free amino acids lowers the number of alkyl pyrazines in roasted coffee. Steam treated coffee reaches target color faster, and hence, roasting time for the same degree of roast is shorter. Actually, monsoon coffee passes through similar mechanisms of precursor changes. The partial hydrolysis of CGAs and loss of low molecular weight components in the humidity result in an increased spicy character of the product. The air-to-bean ratio as a further roasting parameter was discussed as well. It seems that a higher air-to-bean ratio leads to a rather less complex and flat flavor, which was linked to an increased aroma stripping at higher airflows. The major compositional changes and chemical processes are summarized below that affect the development of flavor compounds in coffee upon roasting:
Monitoring the roasting process in view of repeatable and sustainable quality supply and a better understanding of the roasting kinetics has been a target for a long time by scientists as well as coffee roasters. The generation of flavor upon coffee roasting is a highly dynamic process, in particular in the exothermic phase in which many reactions take place with increasing molecular complexity. One way of getting a more precise insight into the formation processes are kinetic studies describing the flavor composition as a function of time.
The evaluation of specific chemical compounds has been proposed to monitor the roasting degree, apart from physical attributes. As examples, the ratios of free amino acids, alkyl pyrazines, or CGA were used to monitor the course of roasting. Research on the formation of volatile organic compounds (VOCs) in coffee roast gases has traditionally relied on chromatographic techniques, most often GC-MS, but also high-performance liquid chromatography-MS (HPLC-MS). Continuous monitoring of the roasting course, however, is difficult with off-line GC-MS or LC-MS techniques, where sample preparation, analyte isolation, and analysis are typically decoupled. As previously mentioned, coffee roasting results in bitterness represented by caffeine and many other compounds belonging to the chemical classes of CQA lactones, phenylindanes, and DKP. Sensory evaluations indicate a distinct change in the bitterness note from mild bitterness at the lower roasting degree to harsh bitterness at a very high roasting degree (Fig. 2). This is due to changes in the composition of bitter tastants during the roasting process. As shown in Fig.2, harsh-bitter tasting compounds (phenylindanes and DKP) are continuously formed whereas CQA lactones having a mildly bitter note go through a maximum being degraded at around CTN 100 (CTN stands for Color Test Neuhaus determined with the help of Color test II based on infrared reflectance expressed relative to a standard sample, Neuhaus Neotec, Reinbek, Germany). Actually, CQA lactones with mild bitter notes are transformed into harsh-bitter phenylindanes. Therefore, it is crucial to control the late stage of the roasting process.
Online mass spectrometric techniques were developed to follow the fast dynamics of flavor development during the roasting process. In particular, online chemical ionization mass spectrometric techniques (e.g., proton-transfer-reaction mass spectrometry, PTReMS) and online photoionization mass spectrometric approaches (i.e., photon ionization, resonance-enhanced multiphoton ionization time-of-flight mass spectrometry; have been applied for real-time monitoring of coffee roasting to predict the roast degree of coffee by online analysis of the roast gas. Even the aroma formation in a single bean was studied, which allowed the assessment of chemical reaction pathways, for instance, the degradation of CGA. The authors used the photon-ionization techniques to study the permeability of the cell walls with regards to different chemical compounds and suggested the methodology for the rapid determination of the relative Arabica content in coffee blends. PTR-MS techniques were also applied to follow the formation of VOCs indifferent origins (Colombia, Guatemala, Ethiopia, Indonesia), and different time-temperature roasting profiles, interestingly observing varied release dynamics of the online monitored VOCs depending on the origin or the different precursor compositions and physical aspects in beans from different origins, meaning a timely shifted start of VOCs formation.
Wieland et al. measured online the concentrations of volatiles released by the beans into the headspace during roasting. The released gas obtained from the roasting chamber was fed into a highly sensitive PTR time-of-flight (ToF) mass spectrometer for real-time measurement. The time-intensity patterns of compound traces confirmed two typical formation behaviors. The first pattern showed strong formation rates in the second half of roasting time, an intensity peak at medium degree of roast, followed by a fast decrease toward the end of the process when roasting to dark degree. The second pattern showed stepwise continuous increase of intensity during the second part of total roasting time. A principle component analysis discriminated the degree of roast along the roasting process and predicted successfully the bean color.
Fischer et al. (2014) used a similar methodology to compare the formation of organic flavor compounds in Arabica and Robusta beans, respectively. Although the basic average single photon ionisation-TOFMS spectra obtained from roast gas of both species looked similar, a more detailed data analysis permitted to discriminate species, degree of roast, and roasting conditions.
Alternatively, the application of chemosensor array monitoring suitable marker substances (2-furfurylalcohol and hydroxy-2-propanone) was proposed, as well as coupling e-nose technique to artificial neural networks (ANNs) to evaluate the roasting degree. The latter may represent an effective possibility to roasting process automation and to set up a more reproducible procedure for final coffee bean quality characterization. A different approach was applied by Wei (Weietal.,2012). Based on this composition-based holistic method they suggested different appropriate chemical markers to control and characterize the coffee-roasting process.