Evolution of Coffee Roasting
Evolution of Coffee Roasting
Evolution of Coffee Roasting
In order of presentation, this paper describes: batch coffee roasting in general; batch
roasting systems used in the 1950's; conventional methods for controlling coffee roasting; new
batch roasting systems introduced in the past 35 years; the development of fast-roasted, high-
yield coffee; fluidized-bed roasters; continuous roasters; pollution control for roasting; new
knowledge about roasting chemistry; and new roasting control methods..
To prevent out leakage of roaster gas and minimize in leakage of air during roasts, the
roasting chamber is maintained at a slightly negative gage pressure (e.g. - 6 mm of water) The
furnace is kept at slightly positive gage pressure to prevent air in leakage and minimize out
leakage of furnace gas. Therefore, the gas entry to the discharge stack is located just
downstream from the furnace.
The fuel and burner air control valves are effectively linked so as to provide a constant
fuel:air ratio, roughly 11 to 12 volumes STP of air per volume STP of natural gas. 10% to 20%
excess air is used to minimize formation of CO and NOX. For 10% excess the flame
temperature is roughly 1710oC; for 20% excess air, roughly 1580oC.
Beans char (Backer et al. 1938) or even spall at their tips if roaster gas temperatures
greater than 510oC are used. Inlet gas temperatures, Tgi, in batch roasting chambers vary
during roasting (Schwartzberg, 2003), and at their peak can readily reach 500oC. To provide
Tgi that beans can tolerate, recirculated roaster gas is mixed with the combustion products
generated in the burner in the furnace. For most recycling roasters, the ratio of recycled gas to
products of combustion should be more than 3.3:1 at the start of roasting and more than 6.2:1
Figure 1 Schematic of coffee roasting system in which gas recirculation is used.. a and b
respectively are the burner air and fuel gas control valves, e is the green bean feed damper, f
the quench water valve, t the bean thermocouple and w the quench water volume meter.
at the end of roasting. Recycled gas is mixed with combustion products right after those
products leave the burner so as to cool them rapidly and thereby minimize NOX production. If
recirculating roaster gas contained only products of combustion, its weight % composition
would be: 73.2% N2, 10.2% H2O, 12.8% CO2 and 3.7% O2 for a typical natural gas burned with
20% excess air. However beans emit CO2, H2O, CO and volatile organic compounds (VOC)
during roasting. Therefore the average weight % composition of roaster gas in a recirculating
batch roaster is roughly 50.6% N2, 30.2% H2O, 16.6% CO2 and 2.6% O2. These percentages
vary during roasting, with the percentages of N2 and O2 approaching those in combustion gas
at the start of roasting. The H2O percentage peaks in the middle of roasting and those of CO2,
CO and VOC peak near the end of roasting. At the end of roasting, the VOC content of roaster
stack gas can reach 0.2% by weight. Since VOC partly break down when roaster gas passes
through the furnace, the % VOC for gas in the roasting chamber is probably even higher.
Fig. 2 schematically depicts the roasting chamber for a Jabez Burns roaster, a widely
used batch roaster. It consists of a motor-driven, horizontal, rotating drum with a perforated
Figure 2. Roaster (Jabez Burns roaster) in which a horizontal, rotating drum with a perforated
wall is used as a roasting chamber
wall. The perforations are small enough to retain coffee beans but large enough to pass chaff.
Two oppositely-pitched sets of helical flights attached to the drum wall mix the beans and lift
them. When the beans drop off flights they fall across the axially-flowing hot gas stream. The
gas then turns and passes through the pile of beans at the drum’s bottom and through the
perforations on the bottom 1/3 of the drum wall, the only part of the drum wall not closely
apposed by the insulation lining the roaster shell. While heat transfers to beans as they fall,
much more heat transfers as the gas passes through the bean bed.
In other widely used horizontal, rotating-drum roasters, e.g. Probat type R roasters, the
roasting chamber has solid walls.(see Fig. 3). Hot gas from the furnace first flows around the
exterior of the drum wall and then enters the rear of the drum and flows through it parallel to
the drum’s axis. The drums contain oppositely pitched sets of segmented, helical flights that
mix and lift the beans. The beans drop off the flights and fall across axially-flowing hot gas.
Most heat transfer occurs as the beans fall through hot gas, but significant amounts of heat
transfer to beans from the drum wall by conduction and radiation. To prevent excessive
conduction heating and localized burning of beans, some roasting drums have double walls
with an air gap between them. In the roaster shown in fig. 3, the whole front opens by moving
forward to permit bean discharge. In other solid-walled drum roasters, a smaller, normally-
closed discharge port opens for unloading.
Figure 3 Roaster (Probat Type R roaster) in which horizontal rotating drum with a wall without
perforations is used as a roasting chamber
To prevent centrifugal force from counterbalancing gravity and causing beans to stay at
the drum wall without falling, horizontal-drum roasters have to rotate with an RPM smaller than
(30/π)(g/r)½, where g is the gravitational acceleration and r is the drum radius. The larger the
radius of a horizontal-drum roaster is, the slower the drum must turn. Therefore, larger,
horizontal-drum roasters provide poorer mixing than smaller roasters of the same type. This
probably is a major reason why the maximum green bean batch capacity for commercial,
horizontal, rotating-drum roasters is roughly 500 kg. As RPM increases, beans fall off flights at
higher angles of rotation. The flights impart their velocity to beans falling off them. Thus beans
fall in curved paths that rise slightly before descending. Bean exposure to flowing hot gas is
maximized at an optimal RPM 10% to 25% less than (30/π)(g/r)½.
Till the late 1950's, horizontal, rotating-drum roasters were they only types of industrial-
scale, batch coffee roasters available and roast times ranged from 15 to 18 minutes. Similar
roasters still are widely used and made today and now normally provide in 10 to 12 minute or
10 to 15 minute roasts, with the faster roasts being obtained with smaller bean loads.
Figure 4 Roaster in which single-pass gas glow is used
Single-Pass Roasters
Small roasters usually are small, horizontal, solid-wall, rotating-drum roasters in which
single-pass, hot-gas flow is utilized,. Fig. 4 depicts a schematic of a typical roaster of this type.
The figure shows only two lance-like burners in the roaster shell below the drum, but three
burners are usually used. As roasting progresses, first one, then two of the burners are turned
off. The damper in the gas discharge line from the drum can be used to adjust roaster gas
flow, but is rarely adjusted after installation. Operators of small roasters usually do not use
water quenching and rely solely on air cooling in the cooling tray to stop roasts.
Somewhat lower inlet gas temperatures are used in small, single-pass roasters than in
recirculating roasters. Products of combustion in single-pass roasters are diluted with air
instead of recirculated roaster gas to provide gas inlet temperatures suitable for roasting. On
an average-weight-percent basis, gas passing through the roasting chamber of a single-pass
roaster contains much more O2 (22.8% in and 21.8% out), less H2O (2.1% in and 5.2% out)
and less CO2 (2.7% in and 3.6% out) than gas in roasters employing gas recirculation. Thus, in
single-pass roasters, there is a greater potential for oxidation at bean surfaces than in
recirculating roasters but less exposure to tars that build up in recirculating roaster gas. The
flavor consequences of these differences in roasting environment do not appear to be known.
Small single-pass roasters emit VOC in much more dilute form than recirculating roasters, but
the amount of VOC emitted per unit mass of green beans is as great as and probably greater
than in recirculating roasters
Industrial-scale roasting was and is controlled primarily by stopping a roast when the
measured bean temperature Tbm reaches a set value, (Tbm)f , that experience shows has
provided roasted coffee with the desired flavor and color. Because of thermometric lags, Tbm
differs the true bean temperature Tb. Thus (Tbm)f are 10oC to 20oC lower the true final bean
temperature, Tbf. While thermometric lags differ form roaster to roaster, (Tbm)f for rotating-
drum roasters are usually roughly 190oC for a very light roast and 260oC for a very dark roast..
(Tbm)f depends on bean type, being roughly 4oC higher for Robustas and roughly 8oC higher
for Columbians than for Brazilian Arabicas. During roasting, coffee beans lose weight due to
evaporation of their water content (8% to 12% of the green bean weight) and destruction of dry
matter, 1-5% for a light roast to > 12% for a very dark roast. The total percent weight loss and
percent dry matter loss are used as after-the-fact measures of extents of roasting.
Operators of small custom roasters observe the color of bean samples periodically
withdrawn from the roaster (every 15 seconds near the end of a roast) and stop the roaster
when the beans reach the right color. After roasts, operators of large roasters measure the
percent of light reflected from ground, roasted coffee at a standard wavelength, e.g. 595 mm
for Photovolt roast color meters, (Stefanucci et al. 1982). Lighter roasts (50 to 60 on the
Photovolt) are preferred in Northern Europe and the U.S. and darker roasts (32 to 36 on the
Photovolt) in Southern Europe. The standard deviation in reflectance color of coffee roasted to
the same (Tbm)f in the same type and model of roaster is roughly two Photovolt units. The
public can detect color and taste differences in roasts whose colors differ by four Photovolt
units (Stefanucci et al. 1982). In conventionally operated batch roasters, increasing total
roasting time, tf, causes (Tmb)f to increase, causes larger roasting losses and causes roast
color to progressively decrease. Nevertheless, Mahlman (1986) and Schecter (2000) show
that the same roast color can be obtained at both a lower and a higher (Tbm)f by using roasting
conditions providing a markedly longer tf at the lower (Tbm)f .
Other reflectance colorimeters, Agtron, Neuhaus Neotec and Lange, are used by
roasters today. Some of these meters employ infra-red or near infra-red light sources, e.g. 640
nm light is used in the Agtron meter. Colorimeters employing L (lightness),a (red-green) and b
(yellow-blue) scales are used frequently in academic studies of roasting and have been used
to a limited extent in industry. Though a and b values vary markedly during roasting (Severini
et al 1991, Schecter 2000), L values alone usually are used to characterize degrees of
roasting. Changes in roast color (including L changes) for the colorimeters cited appear to be
linearly proportional to one another, but the zero of each color scale is shifted with respect to
the others. Light roasts are 50 to 60 on the Photovolt, 48 to 68 on the Agtron and 36 to 38 on
the L scale; dark roasts: 32 to 36 on the Photovolt, 25 to 33 on the Agtron and 16 to 18 on the
L scale.
High pressures, developed due to CO2 and water vapor generated inside coffee bean
cells, cause beans to expand during roasting. Bean expansion is accompanied by cracking o
Figure 5 New types of roasting chambers: A) Asymmetric spouted bed; B) Rotating bowl; C)
Chamber with rotating scoops; D) Swirling bed roaster
popping sounds. Some operators of small roasters and many amateur roasters base the
stopping of roasts on these cracking sounds. Stopping a roast shortly after the “first crack”
starts provides a light roast; stopping during the “second crack” a dark roast.
Tgo, the temperature of gas flowing out of the roasting chamber, rises as roasting
proceeds,;and ρg, the density of the gas leaving the chamber and entering the blower,
progressively decreases. Roaster system blowers tend to provide a nearly constant
volumetric-flow rate, V. G, the mass-flow rate of roaster gas, = Vρg, and thus decreases as
roasting proceeds (Schwartzberg, 2003). Unless QF, the heat-delivery rate of the furnace, is
reduced, Tgi will rise excessively. Control systems In older roasters decrease QF in two steps,.
With the first step occurring when Tbm reaches a first set point, e.g. 170oC, and the second at
a higher Tbm set point, e.g.180oC. Reducing QF also permit greater control over the end of
roasting.
New types of batch-roasting chambers came into use after 1971 and are depicted
schematically in Fig. 5. Sivetz (1976) patented a roaster where the hot gas caused beans to
circulate as an asymmetric spouted bed, often called a rotating fluidized bed, (Fig. 5A). He
used gas temperature between 232oC and 277oC and roasted in 10 to 20 minutes. Neuhaus
Neotec makes an industrial-scale version of that roaster (the RFB) in which higher gas
temperatures, e.g. 360oC, are used and faster roasting is obtained, e.g. tf = 4 minutes. High
gas velocities and high gas pressure drops are needed to induce spouting. Therefore, unlike
other roasters, roasting chambers in spouted-bed roasters operate at a positive gage pressure.
In the Probat RZ (Fig. 5B) , centrifugal force causes beans slide rapidly outward across the
upper surface of a rotating bowl. The beans are then redirected inward across the flow of hot
gas by stationary vanes. Arndt (1971) patented a stationary, solid walled roaster with a hemi-
cylindrical bottom and rotating scoops, now the Gothot Rapido-Nova or Probat RT roaster (Fig.
5C).. Wireman and Wireman (1991) patented a roaster, now the Burns System 90, in which
gas injected through louvers on an inverted, vertical, truncated conical wall passes though a
thin bed of coffee beans causing then to swirl around the wall’s inside (Fig. 5D). The beans are
held radially in place by centrifugal force. The systems shown in Figure 5 all utilize gas
recirculation and can provide batch roasting as short as one to three minutes. Three to five
minute roasts are commonly used. When fast roasts are used, most of these systems can
roast up to 2,000 to 4,000 kg of green coffee per hour. If desired, they can provide eight minute
roasts, albeit at reduced rates of throughput.
MacAllister and Clifford (1964), at General Foods, roasted small batches of coffee
(probably Robustas) with an earthy off-taste in a small fluidized bed. Using 316oC “air” the
roast was completed in 2 minutes; and with 399oC “air” in 35 seconds. These roasts produced
coffee that contained more residual moisture than normal, were more acidic, had a less earthy
taste and which yielded significantly more solubles upon brewing. Other inventors also devel-
oped processes that also employed fast roasting and provided lower density roasted coffees
that contained more solubles. Depending on the inventor: faster roasting was obtained by
using much higher gas flow rates either in batch fluidized beds (Mahlman et al.1985),
continuous fluidized beds (Mahlman et al.1985), (Brandtlein et al. 1988) or in continuous
Table 1. Comparison of roasting conditions used in normal roasts and in fast roasts and
resulting roasted coffee bulk densities
BRD = batch rotary drum, BFD = batch fluidized bed, CFB = continuous fluidized bed, CMJ =
continuous roaster where gas from multiplicity of jets impacts on moving bed of beans, *1 = Tgi
in first stage of two-stage roaster, *2 = Tgi in second stage # bulk density data for Columbian
beans, that for Brazilians is roughly 3% lower and that for Robustas roughly 8% higher.
roasters in which hot gas flowing out of an arrays of jets impinged on a vibrating bed of coffee
(Hubbard et al. 1979), (Hubbard 1982), or by employing higher gas temperatures, moderately
higher gas flow rates, lower bean charge weights and a (Tmb)f 1.7oC higher than normal in a
conventional drum roaster (Stefanucci at al. 1982). Price et al. (1991) claim that when
compared to coffee roasted for 840 seconds, coffee roasted in 60 seconds contains more low
molecular weight solubles, more residual sucrose (0.8% versus 0.3%), provides 42% more
head-space aroma and tastes less bitter both when freshly brewed and after the brewed coffee
is held for an hour.
Table 1 lists fast-roasting conditions, conditions for conventional batch roasts yielding
product with a similar roast color, and the resulting bulk densities for the products obtained. As
before, Tgi is the temperature of the gas entering the roasting chamber,
G is the gas mass-flow rate, t the elapsed roasting time, Ro the initial weight of beans and the
integral is evaluated from t = 0 to t = tf the total roasting time. The data in Table 1 show that
SGR decreases as Tgi increases. By modifying an equation developed by Schwartzberg
(1999), one can show that:
where Cgm and Cbm are the respective mean heat capacities of roaster gas and beans, E is the
mean efficiency of heat transfer in the roaster and Tbo and Tbf respectively are the initial and
final values of the true bean temperature. After taking into account cooling due to water
evaporation, exothermic heat generation and effects of bean temperature change, Cbm ≈ 2.41
kJ/(kg⋅oC); and after taking into account roaster gas composition and effects of gas
temperature change, Cgm ≈ 1.31 kJ/(kg⋅oC) for gas in rotating drum roasters.
One can use Eq (2) to roughly assess the heat transfer efficiency of the roasters,
including those listed in Table 1. On this basis, E roughly equals: 0.058 for the fluid bed roaster
used by Mahlman et al., 0.224 for the multi–jet roaster used by Hubbard, 0.61 for asymmetric
spouted bed roasters and 0.72 for conventional horizontal, rotary-drum roasters.
U is the overall gas-bean heat-transfer coefficient and Abg the bean-gas heat transfer contact
area,. As G increases, U increases less rapidly than G does. Therefore E will be lower for
roasters employing very high G. E should not be confused with thermal efficiency, i.e. (heat
received by beans)/(heat input to the roasting system). The thermal efficiency of recirculating
roasters can be high even when E is low.
Less time is available for CO2 and water vapor to diffuse out of coffee beans during fast
roasting. Therefore, during storage, considerably more CO2 diffuses out of fast-roasted coffee
than out of slow-roasted coffee (Schecter 2000). Fast-roasted coffee beans also have a higher
water content than slow-roasted coffee (MacAllister and Clifford 1964). The extra retained CO2
and water vapor cause higher pressures to develop inside coffee cells. Thus. the cells expand
more, thereby providing lower bulk densities. The higher residual water content of fast-roasted
coffee may also provide greater plasticity that facilitates greater expansion..
Green coffee solubles are partly destroyed or made insoluble during normal roasting
(Schechter 2,000). Similar destruction occurs during fast roasting, but that loss is partly or
sometimes slightly more than completely compensated for by the creation of more solubles
(Fig. 6). Fast roasted coffees still contain roughly 4% water when bean temperature reaches
225oC (Geiger at al, 2005). At similarly high temperatures, the water activity of popcorn
containing 4% water is 0.6 (Schwartzberg et al., 1994). If beans containing 4% water behave
similarly at 225oC, the water vapor pressure inside itheir cells would be roughly 1.4 Mpa (14
Figure 6 Extraction yield versus percent roasting loss for high-temperature, short-time roasr
and for low-temperature, long-time roast (Figure by Schecter 2000)
atm.). Thus, the created solubles might have been produced by hydrolysis of polysaccharides
present in green coffee. Similar hydrolysis occurs in steam roasting of coffee and during high
temperature extraction of coffee.
Several types of batch, fluidized-bed coffee roasters have been patented, e.g. Heimbs
and Kangro (1953), Schytil (1954), Rios et al. (1980). Because of the large size of coffee
beans, gas-induced fluidization of coffee beans normally is of the class D type. In class D
fluidization, gas bubbles rise slowly through a bed which is otherwise relatively dense; and
bubble-induced bed mixing is poor. In class B fluidization, which occurs with much smaller
particles, gas bubbles rise rapidly through the bed, more gas bypassing occurs, but bed mixing
is good. Thus non-uniform roasting can occur when coffee is roasted in deep fluidized beds.
Shallow beds, 6 to 50-mm deep in the unexpended state, are used for fast roasting in
continuous multi-jet roasters. The jets induce circulation due to local spouting. The Lurgi
Aerotherm roaster, a roaster using a deep fluidized bed, was introduced in 1957, but is no
longer marketed or used (Clarke and Macrae 1987). Numerous internet items show that
swirling-bed popcorn poppers and other types of small, fluidized-bed roasters are used to roast
coffee at home. Some small fluidized-bed roasters contain baffles that cause bean circulation,
e.g. the small, fluidized-bed roaster described in a patent by Gell (1985).
Continuous Roasters
Continuous roasters were first used commercially in the 1940's. Newer types of
continuous roasters came into use in 1970's (Nutting et al. 1971, Bannar 1978) and 80's
(Clarke and Macrae 1987). Continuous roasters can roast up to 5460 kg green coffee per hour.
Many can provide roasting in as little as 60 seconds, and thus are suitable for producing high-
yield coffee. Continuous roasters can also provide 8 minute roasts if desired. In batch roasters,
bean temperatures change with time; in continuous roasters, bean temperatures change with
axial position; and temperatures are more uniform from bean to bean as roasting progresses.
In continuous roasters, beans are moved either by helical flights attached to a long, rotating,
perforated, cylindrical shell, or by shafts in a rotating, impermeable, cylindrical shell (Probat
Conti-Roaster) or by motion imparted by a vibrating impermeable plate (Wolverine Jet Zone
Roaster). In continuous roasters where a perforated cylinder is used, hot gas passes through
cylinder perforations both in entering and in leaving the cylinder. In some cases, the gas
passes upward, fluidizing the beans (Brandtlein et al, 1988); in others, the gas flows diagonally
downward (Burns continuous roaster). When an impermeable vibrating plate conveyor or an
impermeable cylinder is used, hot gas flowing out of a multiplicity of tubular jets impacts on a
moving bed of beans. Continuous roasters are used only by very large coffee roasting
companies, and then only for producing standardized types of roasted coffee in very long runs.
In the past thirty years, diverse types of roasts employing wide varieties of coffees and coffee
blends and different degrees of roasting have come into use. These roasts are most effectively
produced in batch roasters.
Custom roasters and specialty roasters have proliferated in the past thirty years. Some,
like Starbucks, have become very successful. Thus there are a great many more small and
medium sized roasters today than there were thirty years ago. Some medium-size roasters
cater mainly to the hotel and restaurant trade; others, e.g. Green Mountain Roasters, are
regarded as gourmet roasters. Gourmet roasted coffee produced in small- and medium-size
roasting plants is made from green beans from many different growing areas and is available
in different kinds of roasts. Many of these roasts are markedly darker than used to be the norm
for the U.S. The impact of small- and medium-size gourmet roasters on the roasted-coffee
market has prompted major roasters to offer more varieties of coffees or coffee blends and
different degrees of roasts.
Pollution Control
The VOC and CO contained in stack gas that roasters discharge are environmental
pollutants that nowadays usually have to be nearly completely destroyed. In most cases, VOC
and CO destruction is accomplished in either direct or catalytic afterburners. In direct
afterburners, roaster stack gas is heated to 760oC; and in catalytic afterburners to 482oC. To
provide enough O2 to burn VOC, the air:fuel ratios used for afterburner have to be greater than
those used in roaster furnaces. Though catalytic afterburners use less energy, they are more
costly; and the catalyst has to be replaced from time to time. Afterburners greatly increase
energy costs for roasting and produce NOX. Energy costs for afterburning are far greater for
single-pass roasters than for recirculating roasters. Felip (1999) claims that afterburning
temperatures, energy use and NOX discharge can be reduced by placing the roaster blower
before the cyclone and sending recycle gas taken from the cyclone rather than from the
roaster furnace to the afterburner.
Systems for recovering energy from hot gas discharged from afterburners and roaster
systems have been developed. Rothfos (1986) describes some of these systems.. Most
involve exchange of heat between after-burner-discharge gas and air flowing to the burner for
the roaster. Farina (1994) obtains all the heat required for roasting in a single-pass roaster by
transferring heat from afterburner gas to fresh air entering the roaster. Finken et al. (1998) heat
warm air discharged from the roasted bean cooler with afterburner-discharge gas to supply
burner air and thereby provide additional heat recovery. Exchange of heat between
afterburner-discharge gas and roaster gas is more troublesome than heat exchange with
incoming air because roaster gas contains tars that foul heat-exchangers.
Roasting Chemistry
Green coffee contains roughly 300 volatile compounds that are also present in roasted
coffee. Roughly another 650 volatile compounds form during roasting. The combined 950
compounds include: 97 sulfur compounds derived from cysteine and methionine; 66 pyrroles
and 11 pyridines derived mainly from trigonelline, 126 furans derived mainly from sucrose and
polysaccharides; 70 pyrazines and 35 oxazol(in)es derived from amino acids by Maillard
reactions, 48 phenols derived from chlorogenic acids, linalool and myrcen formed from
terpenoids, 102 ketones and aldehydes largely produced by Strecker degradations, 31 esters,
25 acids and 20 alcohols.
There are also many non-volatile products, e.g.: caramels derived from sucrose and
polysaccharides; brown, condensed, end-stage Maillard reaction products; nicotinic acid and
its methyl ester derived from trigonelline; quinic acid and lactones and esters of quinic, caffeic
and ferulic acid derived from chlorogenic acids; phenols, e.g. catechol, pyrogallol and
hydroquinone derived from quinic acid and caffeic acid; glycolic acid and lactic acid.
Many of these compounds are produced via complex, multi-step reactions involving
competitive parallel reactions as well as series reactions, e.g. Maillard reactions,
caramelizations, and the breakdowns of trigonelline and chlorogenic acid. Thus the reactions
are difficult to wholly characterize from a kinetic point of view. Aside from kinetic data derived
from studies on model systems, the most we can do now is characterize kinetics in terms of
changes in end-product concentration with time at different Tb. Fig. 7 by Gianturco (1967)
shows how the chromatographically measured concentration of some volatile roasting
products change with time during roasting. Concentrations of some products e.g. acetic acid,
furfural, furfuryl alcohol, peak then decline, with the peak coming either before, at, or after the
normal end of a conventional commercial roast (roughly 10.5 minutes). Concentrations that
peak and decline indicate the presence of reactions that cause product breakdown. The
concentrations of other products, e.g. phenol, pyridine, continue to rise well after the normal
end of roasting. These products may also break down, but, if so, the cumulative extent of
product breakdown has not yet exceeded the extent of product formation.
Schecter (2000) carried out fluidized-bed coffee roasts, some of which provides
constant Tb over most of the roast. Product concentration versus time data for these roasts can
be used to characterize the kinetics for some roasting reactions. Thus it may be possible
to obtain effective activation energies and prefactors for Arrhenius equations characterizing
some reactions. Most product concentration versus time curves for Schecter’s isothermal
roasts were initially concave upward. It can be shown that these curves must be due to
product formation through a series of reactions. Schecter’s data (2000, 2002a, 2002b) shows
that, even when roasts provide the same roast color, some products appear in greater
Figure 7 Chromatographically measured concentrations versus roasting time for some volatile
compounds found in roasted coffee (Figure by Gianturco 1967).. CR is the roasting time for a
normal conventional roast.
concentration in fast roasts and others appear in greater concentration in slow roasts. It is
instructive to compare reaction product yields for a fast and a slow roast with similar bean
temperature profiles (Tb versus t records), the temperature profile for the fast roast being a
temporally shrunken version of the profile for the slow roast. It can be shown that if a product’s
concentration still is rising at the end of the slow roast, its concentration will always be lower
for the fast roast. Further, if a product’s concentration peaks and then declines during a slow
roast, that product usually (but not always) will be obtained in greater yield during a fast roast.
Thus consumers with sufficiently sensitive tastes are likely to be able to taste differences
between fast-roasted, high-yield coffee and conventionally roasted coffee.
Schecter et al. (2002a, 2002b) also showed that product compositions differed markedly
when the same type of coffee beans were roasted to same roast color by means of roasts
employing a variety of different bean temperature profiles. This shows that roast color is a far
from adequate indicator of roasted coffee composition. It also again illustrates the principle that
outcome of incomplete reactions depends on their temperature versus time history.
Well before Schecter even began his research, it had become clear to a number of
operators of coffee roasting plants and to suppliers of roasting control equipment that bean
temperature profiles for a given type of green beans should be controlled so as to be the same
roast after roast in order to obtain the same roast quality roast after roast. Further, the bean
temperature profiles used should be tailored for the type of blend of coffee beans involved.
Rd is the dry mass of beans, QR is the rate of exothermic heat generation per unit mass of
beans, X is the dry basis moisture content of beans and λ is the latent heat of evaporation of
water from beans. Schwartzberg (2002) provides equations providing: QR as a function of Tb
and the cumulative previous extent of exothermic heat generation; CB as a function of X and
Tb; and dx/DT as a function of Tb, X and bean size.
In most profile-control systems Tgi is indirectly adjustrf by varying QF, the heat output
rate of the roaster furnace or burners. In Praxis International’s Logofile system, G is adjusted .
Changes in G tend to change Tgi. Therefore, the Praxis system usually automatically maintains
Tgi constant by Tgi - based feedback control of QF. The Logofile system can also provide
programmed adjustment of the Tgi set point if Tb approaches Tgi closely enough to cause dTb /dt
to become relatively insensitive to changes in G. Regardless of whether QF or G adjustment is
used, the object of adjustment is to make Tbm conform to or to home-in on a Tbm set point or a
series of Tbm set points. Most profile control systems rely on feedback control employing a
limited number of Tbm set points, e.g. 3, and adjust the gain, differential and integral response
of the controller so that homing in on the successive Tbm set points provides a Tbm versus t
profile that conforms to the desired profile as well as possible. In some cases, such adjustment
causes a marked shift in the control profile itself, e.g. as in graphs 1, 2 and 3 in an article by
Davis and Ribich (2005). Praxis International’s Logofile controller relies on the Tbm versus t
records of control roasts that have produced desired products. A programmable logic controller
(PLC) then changes the Tbm set point every second so as to provide a Tbm versus t profile that
closely conforms to the profile for the selected control roast.
Roasters are preheated before a load of beans is added to start a roast. Thus, at the
start of a roast, the thermocouple probe that measures Tbm contacts hot roaster gas rather
than beans; Tbm is much higher than the true, average bean temperature, Tb. As t increases,
Tbm first decreases, then passes though a minimum, where Tbm = Tb, and then increases.
Because of preheating and thermometric lag Tbm may be more than 200oC higher than Tb at
the start of a roast; and because of thermometric lag, Tbm is usually 10oC to 20oC lower than Tb
near the end of roasts. Tbm versus t profiles must be converted into Tb versus t profiles in
order to specify needed G versus t and Tgi versus t behavior and to provide bases for
predicting the kinetic consequences of choices of Tbm - based profiles. If the bean temperature
probe is immersed in a dense stream of beans, Tb versus t behavior can be determined from
Tbm versus t profiles with aid of Eq (5)
K, the thermometric lag coefficient, = UtAt/(MtCt) where Ut, At, Mt and Ct respectively are the
effective bean to probe heat-transfer coefficient and the thermal contact area, the mass and
heat capacity of the thermometric probe. Eq (5) works well except during and immediately after
beans loading.
Conclusions
In the past fifty years, advances in roasting equipment and control methods have
permitted roaster operators to roast faster and more effectively. Fast roasting decreases
roasted bean densities, increases extraction yields from roasted beans and changes the
results of roasting chemistry. Improved knowledge about heat-transfer during roasting and
roasting kinetics and improved methods for monitoring and controlling Tb versus t behavior
soon should permit roaster operators to provide some measure of predictive control of roasting
conditions and some degree of control over roasting reactions.
References
Allington, R.A., Torbet, P.A. and Granger, E.M. (2000) Method of terminating roasting of bulk
materials in real timen the basis of objectively determined color of materials, U.S. Patent
6,106,877
Arndt. H.U. (1971), Device for batch roasting and cooling particulate material, U.S. Patent
3,608,202
Backer, L.H.. Keenan, J.H. and Kopf, J.I. (1935), Roasting method, U.S. Patent 1,991,190
Bannar, R. (1978). New roasting process for coffee increases yield per bean by 20%, Food
Engineering
Brandtlein, L.S. , Schecter, S.M. and Mahlman, J.P. (1988) Coffee roasting method, U.S.
Patent 4,737,736
Clarke, R.J. and Macrae, R. (1985), Coffee, Vol 1, Chemistry, Elsevier, Applied Science,
London
Clarke, R.J. and Macrae, R. (1987), Coffee, Vol 2, Technology, Elsevier, Applied Science,
London
Davis, T. and Ribich, P. (2005) Taking control, Roast. (March/April), pp. 53-61
Farina, S. (1994) Roasting sysren and method, U.S. Patent 5,372,833
Finken, H., Jansen, G and Naves, E. (1998) Apparatus for roasting vegetable bulk material,
such as coffee beans, U.S. Patent 5,718,164
Geiger, R., Perren, R. Kuenzli, R. and Escher, F., (2005), Carbon dioxide evolution and
moisture evaporation during roasting of coffee beans, J. Food Science, 70, No. 9. p. E124-
E130
Gianturco, M. (1967) in The Chemistry and Physiology of Flavor, eds. H.W. Schulz, E.A. Day
and L.M. Libbey, AVI, Westport Conn., pp. 431-449
Heimbs, E. and Kangro, W. (1953) Treatment of coffee and other materials, U.S. Patent
2,859,116
Hubbard, M.H. (1982) High speed process for roasting coffee U.S. Patent 4,322,447
Hubbard, M.H. Phipps, R.M. and Thompson, R.L. (1979) Low density coffee roasting process,
U.S. Patent 4,169,164
Illy, A. and Viani, R. (eds) (1995), Espresso Coffee ., Academic Press, London, p. 96
Lee, C.M. (1999) On the cutting edge, a rundown of today’s process control technology, Tea
and Coffee Trade Journal, 171, No. 11, pp. 32,34,36-42
Mahlman, J.P., Schecter, S.M. and Scher, L. (1985) Controlled coffee roasting, U.S. Patent
4,501,761
Nutting, L., Chong, G.S., Miller, J.T. and McCheney, J.G. (1971) Continuous coffee roasting
apparatus, U.S. Patent 3,572,235
Price, S.E., Kusain, R.F., Fruhling, R.J. and Harris, M.B. (1991) Ultrafast roasted coffee, U.S.
Patent 4,988,590
Rios G.M., Manguio, G.H., Gibert, H, Crouzet, J. and Vincent, J.C., (1980), Process for
roasting an agro-food product in a fluidized bed of inert particles, U.S. Patent 4,215,151
Rothfos, B. (1986) Coffee Consumption, Gordian Max Rieck GmbH, Hamburg, Germany, pp.
118-173
Schecter, S. (2000), Investigations on the Hot Air Roasting of Coffee Beans, Doct. Tech. Sci.
Thesis, ETH, Zurich
Schecter, S., Heineman, C., Huber, M., Pompozzi, R., Perren, R. and Escher, F. (2002a),
Impact of roasting conditions on the formation of aroma compounds in coffee beans, J. Food.
Sci., 57, No. 1, p. 60-66
Schecter, S., Heineman, C., Huber, M., Pompozzi, R., Perren, R. and Escher, F. (2002b),
Impact of roasting temperature profiles on chemical reaction conditions in coffee beans,
Institute of Food Science, ETH, Zurich, 9 pages
Schwartzberg, H.G. (2002) Modelling bean heating during batch roasting of coffee beans,
Chap. 52 in Engineering and Food for the 21st Century, J. Welti-Chanes, G.V. Barbosa-
Canovas. J.M. Aguilera (eds), CRC Press, Boca Raton, p. 871-890
Severini, C., Nicoli, M.C., Mastrocola, D. and Lerici, C.R. (1991) Influence of heating rate on
some physical and physico-chemical properties of coffee beans during the roasting process,
14th ASIC Colloquium, San Francisco, p. 641-647, Association Scientifique, International de
Café, Paris
Stefanucci, A., and Protomastro, M.G. (1982) Low density coffee process, U.S. Patent
4,349,575
Wireman, J. and Wireman, D.R. (1991) Apparatus for conditioning particulate materials, U.S.
Patent 5,068,979