Index Table of contents Recent Developments in Coffee Roasting Technology R. PERREN 2, R. GEIGER 3, S. SCHENKER 4, F. ESCHER 1 1 Institute of Food Science, Swiss Federal Institute of Technology (ETH), CH-8092 Zurich, Switzerland 2 Present address: Keme Food Engineering AG, CH-5001 Aarau, Switzerland 3 Present address: Bühler AG, CH-9240 Uzwil, Switzerland 4 Present address: Nestec Ltd., CH-1350 Orbe, Switzerland SUMMARY Roasting of coffee beans leads to considerable changes in microstructure and structure resulting in a remarkable bean volume expansion. The extent of the volume expansion influences mass transfer during storage and has to be considered as important for bean quality changes during storage. Therefore, a concept for the mechanism of volume expansion was developed. During roasting, structural cell wall polymers such as mannan, arabinogalactan and cellulose are changing to a rubbery state allowing limited changes in volume, but preventing the beans from disintegration. From DMTA data, a state diagram of cell wall polymers and coffee beans was developed and the conditions for an elastic behaviour allowing volume expansion were identified. Besides structure resistance forces, driving forces are required in order to achieve a volume expansion. During roasting, a considerable quantity of moisture and dry mass is evolved. The quantity of carbon dioxide and moisture evolved during High-Temperature-Short-Time (260 C 170 s) and Low-Temperature-Long-Time (228 C 720 s) roasting process was determined online in the exhaust roasting air using Near-Infrared-Absorption technique (NIR). In both processes a maximal moisture evaporation rate caused by the evaporation of initially present moisture was observed before the roasting process was stopped. Under HTST conditions, carbon dioxide evolution rate increased exponentially at temperatures above 180-200 C, whereas under LTLT conditions, carbon dioxide evolution rate was rather constant above 180-200 C until the end of the roasting process. Finally, the weight ratio of carbon dioxide and moisture in the total roast loss were calculated and a mass balance for the roasting process was developed. INTRODUCTION Roasting presents an important process step. The influence of roasting on coffee bean and cup quality was put in research focus only since several years. In our work, we were focussing on the influence of the roasting process on changes of microstructure and on the mechanisms involved in volume expansion. In the work of Geiger (2004), the influence of structure resistance forces and driving forces on volume expansion was studied in detail. STRUCTURE RESISTANCE Dynamic mechanical thermal analysis DMTA was applied for the identification of transition phenomena of polymeric cell wall compounds. With DMTA texture modifications such as 451
softening and hardening of materials can be identified and related to the changes of the physical state of materials. For the analysis coffee beans of different moisture content were cross-sectioned to slices of 3mm thickness and mounted on a DMTA Solids Analyzer RSA II (Rheometrics, Piscataway USA) with plate-plate configuration. Then, the specimen were heated linearly from 30 C to 250 C with a heating rate of 5 C/min. A DMTA thermogramme of a coffee bean slice is presented in Figure 1. Therein, the storage modulus G is shown. The fast drop of storage modulus G between 130 and 170 C represents a transition of polymers from the glassy to the rubbery state, the coffee bean texture is softening. The storage modulus G increase between 200 and 230 C represents the reversion of the transition, interrupted by a melting phenomenon of a compound between 212 and 217 C. The coffee bean texture is hardening again. Figure 1. DMTA-thermogramm of a coffee bean. There is a strong relationship between initial moisture content and the temperature range of the glass transition. Because DMTA analysis was performed under non-moisture-controlled conditions, the moisture content as determined immediately at the beginning of the transition was taken into account. By variation of the initial moisture content a state diagram for coffee beans was developed (Figure 2). Because no sharp and pronounced transition was found in DMTA, the on- and offset of the transition was used to describe the state transition. In order to identify the compounds involved in the state modifications in green coffee beans an adapted DMTA method was applied to analyse the behaviour of pure amorphous or semicrystalline polymers prevailing in coffee bean cell walls, namely cellulose, arabinogalactan and mannan. The observed state diagram of coffee beans correlates to cellulose and mannan, whereas arabinogalactan melted completely above 210 C. Probably, the melting phenomenon observed in DMTA analysis was caused by arabinogalactan. Finally, a mixture of cell wall compounds representing the approximative composition of coffee beans was analysed (Figure 3). In fact, the changes of storage modulus G were comparable to structured coffee beans in quality, which allows to conclude that the overall state changes in coffee beans are composed by the state changes of individual cell wall polymers. 452
Figure 2. State diagram for coffee beans. Figure 3. Thermograms of coffee cell wall polymers, a mixture of cell wall polymers and coffee beans. 453
DRIVING FORCES During roasting coffee beans are exposed to hot temperatures, where flavor compounds are formed during non-enzymatic browning and pyrolytic reactions. Moisture as present in the raw beans is evaporated, and reaction products from non-enzymatic and pyrolytic reactions evolve. Geiger et al. (2004) studied driving forces during coffee roasting in detail, where they applied a high-temperature-short-time (HTST) and a low-temperature-long-time (LTLT) process with air temperatures of 260 C for 170 s (HTST) and 228 C for 720 s (LTLT). Roasting was carried out on a laboratory fluidized bed roaster as described in detail by Schenker (2000) and Schenker et al. (2000) to an equal bean color, which corresponded to a L*-value of 22-23. Sample size was 100 g coffee beans, the initial moisture of the beans was 8.3 g/100g wb. In addition coffee beans were roasted which had been pre-dried in an air-dryer at 85 C for 6 days to a residual moisture content of 1.1 g/100 g wb. In order to identify driving forces the evolution of gases such as carbon dioxide and moisture vapour were analysed during roasting by linking a LI 6400 Portable Photosynthesis System (LI-COR Inc., Lincoln, Nebraska) to the exhaust air tube of laboratory roasting machine. An aliquote of the air (< 0.2%) was lead from the exhaust air tube to the detector. In order to compensate for potential variations of air velocity over the exhaust tube cross section (r = 0.05 m) the sampling tube (r = 0.0025 m) was positioned in the exhaust tube. Figure 4. Bean center temperature, moisture content and roast loss during HTST and LTLT roasting. Figure 4 shows the development of coffee bean core temperature, roast loss and moisture content of beans during the isothermal high-temperature-short time (HTST) and low- 454
temperature-long time (LTLT) roasting process to equal bean color and roast loss. The increase of roast loss and the decrease of moisture content were almost linear at HTST conditions. Only towards the end, the curve for moisture content started to level off. In contrast, exponential curves over the roasting time were observed at LTLT conditions for both roast loss and moisture content. As a consequence, the roasting temperature is the most decisive parameter in controlling overall changes in coffee beans. The evolution of carbon dioxide is presented in Figure 5. The carbon dioxide concentration increased sharply in the end phase of HTST roasting while the concentration in LTLT roasting stayed much lower and levelled off in the end phase. One could imagine that HTST roasting moved towards pyrolytic conditions at the end of the process. The cumulative value show that more carbon dioxide evolves in the LTLT process than in the HTST process due to the much longer roasting time. Figure 5. Evolution rate and cumulative evolution of carbon dioxide during HTST and LTLT roasting. The cumulative values in Figure 5 do not reflect the quantity of carbon dioxide formed completely. A substantial part of this quantity is trapped in the coffee beans and is released only during storage. Therefore, cumulative values of evolved carbon dioxide during roasting and storage were combined in Figure 6. Roast loss and storage time, respectively, were chosen as independent variables for the two steps. The lower gas evolution during HTST roasting is more than compensated during storage by a much higher cumulative gas release. Under the given roasting conditions, HTST and LTLT roasting result in an almost equal total formation and evolution of carbon dioxide. Total carbon dioxide formation and evolution seems to be in particular dependent from roast degree and to a reduced extent from roasting temperature. 455
Figure 6. Cumulative evolution of carbon dioxide during LTLT and HTST roasting and during subsequent storage. Figure 7. Moisture evaporation rate and cumulative moisture evaporation during LTLT and HTST roasting. Figure 7 presents the averaged data for water evaporation during roasting. Both roasting processes lead to a peak evaporation rate and a subsequent decrease. The HTST and LTLT 456
process differ primarily in the extent of the peak rate. Moisture evaporation rate depends on the roasting temperature. The cumulative evaporated quantity of water is composed of water, which evaporates due to dehydration of initial moisture of coffee beans, and of water which is generated by chemical reactions. To evaluate the formation of moisture in chemical reactions, HTST and LTLT roasting trials with pre-dried beans (1.1 g/100 g wb) were carried out. It was assumed that in this case the detected moisture in the exhaust air was exclusively formed by chemical reactions due to a negligible initial water content of the green coffee. In Figure 8 the chemical reaction water and the initial water evaporation rates are shown. The evaporation rate of initial water was calculated from the difference between the evaporation rate of total moisture during roasting of non pre-dried coffee (water content: 8.30 g/100 g wb) and the evaporation rate of total moisture during roasting of pre-dried coffee. Figure 8. Evaporation rate of total moisture, initial moisture and chemical reaction water during HTST and LTLT roasting. Water evaporation is temperature dependent. The peak rate for HTST roasting was higher than the peak rate under LTLT roasting conditions, whereas more moisture evaporates during LTLT roasting due to the much longer roasting time. In an early roasting phase of both roasting processes only initial water has been released from the beans. With increasing roasting time, in particular to see for LTLT roasting, total moisture evaporation mainly consists of evaporating chemical reaction water. Initial water evaporation became almost negligible after approximately 300 s and the formation and evaporation of chemical reaction water were equaled. The peak rate of total moisture evaporation is due to the overlapping effect of initial and chemical reaction water. The decrease of chemical reaction water evaporation could be the consequence of running out of the substrate or formed chemical reaction water will be used in the cells for other chemical reactions and will therefore not 457
occur in the exhaust air. It seems in Figure 8 that chemical reaction water was already produced from the beginning of the process. In fact, most of the residual moisture of 1.1 g/100g wb in the pre-dried green beans had to be removed first before the effective chemical reaction water could be detected. The evaporated moisture from the pre-dried coffee, considered as chemical reaction water, amounted for 41% (HTST) and 36% (LTLT) respectively of total moisture evaporation from the non pre-dried green coffee. In Table 1, a mass balance over roasting and storage is presented on the basis of evolved carbon dioxide, evaporated water and losses of solids in the form of silver chaffs. An estimation of the loss of gases during the cooling step was done by linear extension of the cumulative carbon dioxide and moisture evaporation values for another 20 s after the end of roasting. A standard deviation of the evolved carbon dioxide and moisture was calculated from the deviation of air velocity and gas concentration measurements. Table 1. Mass balance for the roasting and storage of coffee beans and comparison with overall roast loss. Step Weight [g] HTST LTLT Initial beans Total solids 091.70 091.70 Moisture 008.30 ± 0.2 008.30 ± 0.2 Sum 100.00 100.00 Roasting loss (on-line) Carbon dioxide 000.4 ± 0.0 000.50 ± 0.3 Total water 010.20 ± 1.2 011.40 ± 1.4 Silver chaff 001.0 001.0 Sum 011.60 ± 1.2 012.90 ± 1.7 Cooling loss (calculated) Carbon dioxide 0.1 ± 0.0 0.0 ± 0.0 Water 1.6 ± 0.2 0.1 ± 0.0 Sum 1.7 ± 0.2 0.1 ± 0.0 Total weight loss (on-line/calculated) 013.30 ± 1.4 013.00 ± 1.7 Roast loss (gravimetric) 015.38 ± 0.05 015.86 ± 0.02 Storage loss Carbon dioxide 0 0.99 ± 0.02 0 0.83 ± 0.01 Taking the variations of the data in Table 1 into account, approximately 93% (LTLT) and 96% (HTST) of the gravimetrically determined roast loss could be explained by measuring gas evolution and determination of silver chaff. It must be pointed out that the balance values in Table 1 still do not account for all of the roast loss. Material from abrasion, e.g. tippings, and evolving gases other than CO 2 and water vapor also contribute to the total roast loss. The remaining difference for both roasting processes could be explained by inaccuracies of air velocity and gas concentration measurements as well as by inaccuracies of the raw material. In Figure 9 the volume expansion rate is plotted in combination with total gas evolution rate and moisture evaporation rate. It becomes obvious that the evaporation of moisture presents a very strong driving force for volume expansion. It is worth mentioning, that the highest volume expansion rate is reached even before moisture evaporation has peaked. Carbon dioxide evolution is of minor importance only, but in case of HTST roasting contributes to an 458
increase in volume expansion rate at the end of the roasting process because it restrengthens the total gas evolution. Figure 9. Characterisation of the driving forces during HTST and LTLT roasting. PROCESS OPTIMIZATION The observed volume expansion is of major concern for storage stability of roasted coffee beans. The degradation of cellular and subcellular microstructure represents a sink for mass transfer from the environment. Furthermore, the increase in porous volume and porosity accelerates mass transfer. Both are a consequence of volume expansion and allow oxidation to start. Furthermore, carbon dioxide released from the coffee beans after roasting may lead to an aroma stripping. Therefore, it will be crucial to achieve an optimal balance of volume expansion allowing optimal extractability and minimal mass transfer during storage. As a next step in the investigations on coffee roasting processes we will investigate the influence of evolution of carbon dioxide and water vapor during roasting on the development of coffee been structure, the aroma retention and the aroma release during roasting and storage. REFERENCES Schenker S. 2000. Investigations on the hot air roasting of coffee beans. [DPhil thesis]. Zurich, Switzerland: Swiss Federal Institute of Technology (ETH); Number 13620. Schenker S, Handschin S, Frey B, Perren R, Escher F. 2000. Pore structure of coffee beans affected by roasting conditions. J Food Sci 65 (3): 452-57. Geiger R. 2004. Development of coffee bean structure during roasting Investigations on resistance and driving forces. [DPhil thesis]. Zurich, Switzerland: Swiss Federal Institute of Technology (ETH); Number 15430. Geiger R., Perren R., Kuenzli R., Escher F. 2004. Carbon dioxide evolution and moisture evaporation during roasting of coffee beans. Journal of Food Science (in press). 459