$5::j class and home problems )

The object of this column is to enhance our readers' collections of interesting and novel prob-
lems in chemical engineering. Problems of the type that can be used to motivate the student by
presenting a particular principle in class, or in a new light, or that can be assigned as a novel home
problem, are requested, as well as those that are more traditional in nature and that elucidate dif-
ficult concepts. Manuscripts should not exceed 14 double-spaced pages and should be accompanied
by the originals of any figures or photographs. Please submit them to Professor James 0. Wilkes
(e-mail: [email protected]), Chemical Engineering Department, University of Michigan, Ann
Arbor, Ml 48109-2136.

TEACHING TRANSPORT PHENOMENA

AROUND A CUP OF COFFEE

JEAN STEPHANE CONDORET

Ecole Nationale Superieure d'lngenieurs en Arts Chimiques et Technologiques • Toulouse 31078
tank, after a batch transformation, is let to cool freely before

W
e are all aware that teaching scientific matter is
much more accepted by students when it can be discharge. Another very important characteristic of the study
related to situations they can experience in their is that experiments to assess the modeling are easy to perform
everyday life. A good example is the cooling of a cup of cof- with very simple tools, such as a thermometer, a stopwatch,
fee, whose scientific analysis is much more instructive than and a balance (to estimate the loss by evaporation). Such ex-
we could have thought at first sight. Indeed, we will see that periments could even be done in a kitchen, in full accordance
all heat transfer mechanisms (conduction, convection, and with the "everyday life" aspect of the situation. The method
radiation), as well as those of mass transfer, (because of the to approach the problem, and the reflection about transport
evaporation of the coffee) are involved. This problem was phenomena that it induces, make it a good basis for discus-
often addressed as "leisure in science" or "first approach sion between students and teachers. To avoid a lengthy paper,
of science," and a quick search on the Web shows that this all equations given here are not discussed deeply and, for a
problem has been proposed at all levels of education, from student, may deserve an additional look into textbooks or,
beginning to university. The approach presented here is aimed better, a discussion with teachers.
at being rigorous, but because we do not intend to use very
powerful numerical modeling, simplifications will be made.
Jean Stephane Condoret is a professor
An important quality for an engineer is to make the "right" of chemical engineering at the lnstitut Na-
simplification, i.e., which results only in slight inaccuracies, tional Polytechnique of Toulouse (France).
He graduated in 1977 from the lnstitut de
while respecting the correct hierarchy for the parameters. In Genie Chimique in Toulouse. His Ph.D. thesis
the case chosen here no chemical reaction is present, but the dealt with heat and mass transfer in packed
beds. Since 1987, he has been involved in
coupling of heat and mass transfer in a nonstationary process supercritical technology for chemistry and
is a common situation in chemical engineering. It can be biochemistry.
encountered, for instance, in small industrial units when a

© Copyright ChE Division of ASEE 2007

Vol. 41, No. 2, Spring 2007 137

PRESENTATION OF THE PROBLEM AND radiation. Moreover, at the liquid surface, evaporation of the
HYPOTHESES liquid simultaneously takes place. This evaporation induces
an extra heat loss, corresponding to the heat needed for
The problem: we put a cup of hot coffee on a table. Its
vaporization of water, that is provided from a decrease in the
initial temperature, 8 0 , is around 80 °C, and ambient air is internal energy of the liquid and the cup. Forced convection
at temperature 8 with, for instance, a relative humidity of
3
,
by blowing air is not considered here, although it could be
50% (that means half-saturated). What is the temperature of very easily implemented through adapted computation of
the coffee after 10 min. for instance? Or, more widely, when the coefficients of convection. It is important to mention here
will I be able to drink it safely and what parameters influence that, at temperatures below 200 °C, chemical engineering
this duration? A scheme of the situation is given in Figure calculations usually neglect radiation fluxes because they
1 and the different fluxes will be discussed in paragraph 2. are competing with forced convection fluxes, which are much
To solve the problem we have to make a list of simplifying larger. When dealing only with free convection, this omission
would lead to significant errors, even at low temperature.
hypotheses:
1. Temperature 0 is homogeneous through the liquid in 5. The coffee cup is simulated by a cylinder, external
the cup. There is no temperature gradient in the liquid, and height H,, internal diameter D,, with constant wall thick-
the inner wall temperature of the cup is equal to that of the ness ew' and with a thermal conductivity Aw. The area of the
liquid, because internal free convection is sufficiently high. external vertical wall surface is Aw, and that of the horizantal
These are very important hypotheses and we will devote a liquid surface is A,. Also, this hypothesis upon the geometry
specific paragraph to assess them. of the cup is not very restrictive and can be adapted for other
cases. Liquid is supposed to fill the cup almost entirely.
2. Even if our system is time dependent, we will use
steady-state equations to model the heat and mass transfer 6. Coffee is similar to water, and properties are evalu-
fluxes. This "pseudo steady-state approximation" is very ated at 60 °C.
often proposed and is fully justified here, because establish- Description of the Equations for Modeling
ment of transfers is more rapid than evolution of tempera-
ture of the liquid. It is always difficult to demonstrate this The heat loss through the wall and at the liquid surface
statement, and intuition is often the only indicator. Such results in a temperature decrease that may be described by
ambiguity is rarely addressed, but it has been discussed by the instantaneous heat balance equation, where accumulation
Cussler in his book about mass transfer.m of internal energy in the water and the cup (considering ho-
3. There is no heat loss through the bottom of the cup, mogeneous temperature) equals the sum of all instantaneous
because the table blocks the heat flux. Nevertheless, we may heat losses. Because water evaporates, we also need an in-
foresee that putting the cup on a massive metallic surface stantaneous mass balance:
will speed up the cooling. In this case, the bottom heat flux d0
(mCp +MCp ,)-=-~heatlosses (1)
would not obey the steady-state law, ( see, in textbooks, c wa dt ~
the chapter devoted to conductive transfer in semi-infinite
medium). This will not be considered here. dM
- - = -evaporative. fl ux
(2)
dt
4. At the vertical cylindrical wall of the cup and at the
surface of the liquid, heat loss occurs by free convection and where m and Cpc refer, respectively, to the mass and specific
heat of the cup, and Mand Cpwat' to mass and specific heat of
the water. We can now describe the different heat losses and
express them using steady-state equations of heat and mass
transfer, as stated in hypothesis 2.
Heat Loss at the Vertical Wall of the Cup, Qw
This is a transfer, in series, by conduction through the wall,
then, in parallel, free convection and radiation to ambient
air. As stated in hypothesis 1, internal convection at the in-
ner wall is not considered. This global transfer is accounted
for by a global coefficient Uwe' referred to the external area,
given by:
-1

.
area,g1ven b y: U we = 1 + Dew D (3)
hnv +hR A e+ i
w 2D e
No heat loss at the bottom and
(4)
Figure 1. Schematics of the fluxes.
138 Chemical Engineering Education
 We need to evaluate the coefficient h nv for free convection surface. From Reference 2, for air, it is given by:
at the vertical wall. An equation for such a heat transfer coef-
ficient can be found in:[ 2J
hns
( r
= 1.31l 0 ~eea) (9)

hnv = 1.35l ew; ear (5)

Heat loss resulting from the evaporation, Qevap
We must first estimate the evaporative molar density of
where Eq. (5) is adapted to be used directly for free convec-
flux, Nwat" At the interface, air is saturated at the surface
tion in air. SI units are used throughout.
liquid temperature 8, and water partial pressure is equal to
Radiation transfer is accounted for by a radiation coefficient its vapor pressure Pv(8) at this temperature. Far from the
hR. To estimate hR' we can approximate our case by a situation surface, for half-saturated ambient air, the water partial pres-
in which a small gray surface at 8 radiates toward a large gray sure is 0.5P (8 ). In the case of a single component we can
enclosure, the room at 8 a. In this case, and if 8 and 8 a are not find explan;tio~s in mass transfer textbooks (see for instance
very different, it can be shown (see any heat transfer textbook, Reference 5):
for instance Reference 3), that hR is proportional to the third
1
power of the mean absolute temperature: Nwa, = kcCT-(Pv ( 0)-0.5Pv (ea)) (10)
F
3
~ ((0+273)+(ea +273)1
(6)
k is the mass transfer coefficient referred to a molar concen-
hR = 4Gsl
2
)
tration difference at low or equimolar transfer fluxes.Fis the
logarithmic mean of the partial pressure of air, Pair= PT - P wat
where £ represents the emissivity of the surface and a is the
at the surface and far from the surface, and it accounts for the
Stefan-Bolzman constant. This linearization of radiation
influence of the bulk flow of air. So:
fluxes is very convenient and is a great help to account for
the radiation without adding complex equations.
(11)
An important quality for an engineer is to make
the "right" simpl-ification, i.e., which results
CT is the total molar concentration. Pv(8) can be computed
only in slight inaccuracies ....
from a vapor pressure law for water, such as Clapeyron's or
Antoine's law. Here we have used, from Reference 4:
Note that the convection coefficient, as well as the radiation
105 7 9668_1668 21
coefficient, depends on the outer wall temperature 8w. Indeed, P (0) = --10 228 0
+ whereP isinPa,0in °C (12)
it is not convenient in the computation to evaluate the outer V 760 V

wall temperature, so, for estimation of these coefficients, we An important feature is now to estimate the mass transfer co-
will equate the outer wall temperature to that of the liquid. efficient k . This can be done using the analogy between heat
It results in some inaccuracy for hnv and hR. Eventually, this and mass transfer, as first proposed by Chilton and Colburn. [6l
inaccuracy is likely to be weak because, for usual materials For the air-water system, because the Lewis number, Le, is
and thicknesses, the thermal resistance of the wall is low in close to 1, it gives[7l:
respect to the outer thermal resistance, and the outer wall
temperature is actually not very different from the inner wall (13)
temperature. It does not mean that the thermal resistance
of the wall is neglected here, because it does appear in the The molar density of the flux is then
equation of U [Eq. (3)]. The extreme case of an insulating
wall (as for an expanded polystyrene cup, see paragraph 4) Nwat = h~ ;(Pv(0)-0.5Pv(0a)) (14)
9)1:air Pair
where the inaccuracy is maximum is well described because
the "inaccurate" term has a weak numerical influence in the After some rearrangements, using the perfect gas law, mass
computation of U [Eq. (3)]. flux is:
Heat loss, by Heat Transfer Only, at the Surface of W = hns9J1wat l A. (P (0)-0.5P (0 )) (15)
the Liquid, o. wat 9)1:airCPair F s1 v v a

It also occurs by free convection and radiation, in parallel, Now, knowing the evaporative mass flux, the heat loss by
and is accounted for by a global coefficient hs with: evaporation, Q evap , is given by:
hs = hns +hRs (7) (16)
Qs = hsAsJ0 - 0a) (8) where Af-lv is the heat of vaporization of water per kg of
hns represents the coefficient for free convection at a horizontal water.

Vol. 41, No. 2, Spring 2007 139

Finally the system of differential equations to solve is:
(mCpc +MCpwa,) de= -hs (0)Ase (8-8 3
)- Uw(0)Aw(0-0a)-6Hv hns9J1wat l Asi (Pv (0)-O.SPv (0a)) (17)
ili 9J1~~~F
dM hns9J1wa, 1 A (P (0)-O.SP (0 )) (18)
dt 9J1airCPair F s1 v v a
85
with initial conditions at t = 0, 8 = 8 0 , and M = M 0 • This system can Gt)
be solved numerically by the variable step Runge-Kutta method, for
instance. For all our computations, we have used a very convenient
80
"~
75
commercial software, Mathcad 13, where automatic resolution of • "-,

such system of equations is implemented. Listing of the program -10 ~ ---

e •"-._

r
can be found at <http://lgc.inp-toulouse.fr/intemet/pers/condoret.
·~
6

htm>.
60
RESULTS OF THE MODELING AND
.. 55
~ >,-...
COMPARISON WITH EXPERIMENTS •
The experimental apparatus, including a numerical thermometer, 50
a balance, and a stopwatch, is seen in Figure 2. Three different
45
porcelain cups (No.'s 1, 2, and 3) were used filled with water. The
cups were put on the balance plate, hot water from an electric kettle 40
was poured in, and the temperature and mass variation of the liquid 0 200 400 600 800 1000
were recorded. A piece of insulating material was set under the cups tlme(s)
to prevent direct contact with the balance plate. It proved to be ~)85
useful with respect to hypothesis 3. Physical and geometrical data
80
are given in Table 1. Figures 3 a, b, and c presents the comparison r,.~........ _

between experimental temperature and the modeling as described 75

"' •-i. - ......
- £.
- --- -
-·. ·.....,.--- --
.
above. The modeling appears very good, although it slightly un- "' -... _
-10
e --- --- --
r
derestimates the cooling rates in all cases. A simple explanation
could be that area of the handle was not taken into account in the <.,

computations (indeed, cup No. 2, which gave the best results, had ;; ,.·.,..,__
60
a small handle). Figure 4 also presents good agreement between ........!(....... __

experimental and modeled mass variation. As an example, Table 2 • 55 ....... _

gives computed values of different terms of the equation for experi-
50
"
ments of cup No. 2, and relative contribution of each flux can be
appreciated. It can be seen, for instance, that the evaporative flux, 45
except at the end, is quite significant (see below, paragraph 4). Also
40
note that radiation and free convection coefficients are in the same 0 200 400 600 800 1000
range (around 7Wm 2 °C 1) time(sl
85

Figures 3. Variation of th
© e 80
temperature of the liquid
75 ~

-~
for the 3 different cups :
a) cup No. 1, M = 78.6 g -10
"'.
- ~---
ea= 22.3 °C, ea= 82.5 °C'
0

e; 65
(• : experiment;--: model) -
b) cup No. 2, M = 102.9 g
ea= 21.8 °C, ea= 79 °C'
0 ! 60 -- --- ~
{J..: experiment;
--:model;
i 55 -- --- ---I-~
-• ....____
-------: simplified model; 50 -- ---
-----: model withou t
45
evaporation )
c) cup No. 3, M = 87.2 g 40
ea= 21.1 °C, ea= 80 °C' 0 200 400 600 800 1000
Figure 2. Experimental apparatus. {• :experiment;--: model) tlme(s)

140 Chemical Engineering Education

USE OF A SIMPLIFIED ANALYTICAL SOLUTION
One may feel frustrated to need a numerical solution for the system of differential equations. First note that the mass of water
varies only slightly (less than 3%), so we can suppress the mass balance equation and consider the mass of water as a constant,
equal to M 0 • Furthermore, Table 2 shows that variation, in respect to the temperature, of heat and mass transfer coefficients
is not very large. We have found that use of parameter values computed at the mean temperature leads to very similar results.
Eventually, we can only consider one simplified differential equation, using averaged values:

(19)

110 Nevertheless, even with the proposed averaging, Eq. (19) has still no
obvious analytical solution, due to the exponential term in the expression
105 of Pv(8). But, if the function Pv(8) is approximated by a parabolic equa-
100
..... -
-
·--- ~ tion, Pv(e) = be 2 + ce + d, we can propose an analytical solution. We found
by numerical fitting, that
95 b =18.367 c =-1237.2 d = 27753 in the range 40 °C to 80 °C
In this case, Eq. (19) is a differential equation with separated variables,

80 .
- - • -
-·--- ---·- - •-
whose solution is:

t ( 8) (
= M Cp
Owat
+ mCp )( 2 (2bB8+A+cB) 2
- arctan - - - - - - - arctan - - - - -
cw w w w
(2bB8+A+cB)) (20)
r---
75 - - with
2 2 2 2
w=.J-4bBA0" +4bB dm-A +2AcB-c B (21)
70
0 200 400 600 800 (22)
lime(s)

Figure 4. Loss of mass (g) for the 3 TABLE 1

different cups: Numerical Values of the Parameters of the Three Different Cups
a) cup No. 1, M = 78.6 g, 0 Values are in SI units as given in the nomenclature.
ea= 22.3 °C, ea= 82.5 °C
H ew m Cp
{• : experiment;--: model)
b) cup No. 2, M = 102.9 g, cup
D,
0.0520
'
0.0495 0.0040 0.1092
"1
w

970
E

0.924
0

ea= 21.8, °C, ea= 79 °C N°l

(..__.experiment;--: model) cup 0.0512 0.0610 0.0020 0.0642 1 970 0.924
c) cup No. 3, M = 87.2 g, 0
N°2
ea= 21.1 °C, ea= 80 °C cup 0.0520 0.0635 0.0040 0.1278 1 970 0.924
(• :experiment;--: model) N°3

TABLE2
Numerical Values Given by the Model for Experiment of Cup No. 2 (M 0 = 0.1029 kg, 0 0 = 21.8 °C)
Values are in SI units as given in the nomenclature.
time temp. mass(g) Qevap Qw Qs hnw hns hRw hRs Uwe
0 79.0 102.9 12.0 8.6 2.1 7.5 7.4 7.1 7.6 14.1
90 75.2 102.5 9.5 7.8 1.9 7.3 7.3 7.0 7.5 13.9
180 71.9 102.1 7.8 7.2 1.7 7.2 7.2 6.9 7.3 13.7
270 69.0 101.8 6.6 6.7 1.6 7.1 7.1 6.8 7.2 13.5
360 66.3 101.5 5.6 6.3 1.5 7.0 7.0 6.7 7.2 13.3
450 64.0 101.3 4.9 5.9 1.4 6.9 6.9 6.6 7.1 13.2
540 61.8 101.1 4.3 5.5 1.3 6.8 6.8 6.5 7.0 13.0
630 59.8 101.0 3.8 5.2 1.2 6.7 6.7 6.5 6.9 12.9
720 58.0 100.8 3.4 4.9 1.2 6.7 6.6 6.4 6.9 12.7
810 56.3 100.7 3.1 4.6 1.1 6.6 6.6 6.4 6.8 12.6
900 54.7 100.6 2.8 4.4 1.0 6.5 6.5 6.3 6.8 12.5

Vol. 41, No. 2, Spring 2007 141

 convection in a horizontal cell. Equations for such situation
B=- h ns,v 9J1 wat 1 A L:.H (23) can be found in Reference 9:
81
9)1:airCPair Fav v

(24) Nu= hL = 0.069Ra 033 Pr 0074

\
Figure 3b compares the numerical and the analytical so- H3 PL,0
lution of the equation, and shows that the simplification is with 3xl0 5 <Ra= gf-1 < 7xl0 9
av
quite valid.
and Pr= Cpµ (29)
The case may be further simplified if we consider that \
evaporation does not occur (insulating cover on the cup). In
In our case, if we want to accept a temperature difference
this case Eq. (18) becomes:
Li8 = 1 °C, between bottom and surface of the liquid, Eq.
d0 (29) predicts a conductivity enhancement of around seven-
(M 0 Cpwa, +mCpc) dt =-UwavAwe(0-0a) (25)
fold that now allows a better fulfillment of the Biot criterion.
which is very easily integrated to Remember that this very simplified approach aims only at es-
0 - 0 u.,yA., t
timating if we are in the acceptable range. If we now consider
_ i_ _ =e (M 0 Cp.,,+mCp,)
(26) the case of the industrial tank with a characteristic length of
ei -ea 1 m, Eq. (29)-which gives a conductivity enhancement of
Results of this analytical solution are presented in Figure 120-fold-allows maintaining the Biot number at a low value,
3b, showing that the final temperature is significantly higher in and the hypothesis of homogeneity is still valid.
this case. This situation exists in real life. It corresponds to the
"fast food coffee," which is served in expanded polystyrene CONCLUSION
cups with a cover that insulates and blocks evaporation. This The agreement between modeling and experiments (Figures
absence of evaporation combined with an increased thermal 3 and 4) was surprisingly good. Indeed, every experienced
resistance of the wall (expanded polystyrene has a very low researcher knows that a totally predictive model is often
conductivity) results in very slow cooling. This explains why disappointing and parameter adjustment is common practice
we often bum our lips at the end of a fast food meal when we (conversely, students are very confident in these predictive
drink our coffee without precaution, as we cannot imagine it models!). Nevertheless, be aware of the numerous simplifica-
is still so hot after the duration of the meal! tions we used that here proved to be reasonable. As a practical
conclusion, note that when preparing a cup of coffee another
COMMENTS ON THE HYPOTHESIS OF scenario is possible: hot coffee from the pot is poured into the
HOMOGENEOUS LIQUID TEMPERATURE cup. In this case there is first cooling of the coffee by exchange
of enthalpy with the cup. The cup and the liquid quickly reach
We can use the analogy with the well known case of heating
or cooling of a solid. The homogeneity of the solid tempera- an equilibrium temperature, 8eq' given by the equation:
ture is usually assessed by considering the Biot number, e = MoCPwa,0i + mCpc0a (30)
eq MoCPwa, + mCpc
Bi= hL (27)
\ Indeed, the temperature decrease is significant and this
where L is a characteristic length of the system. The Biot speeds up considerably the desired cooling. Evaluation of the
number evaluates the ratio between inner conductive trans- kinetics of this process is not easy, but is useless because its
fer and outer convective transfer. When the Biot number is rapidity (a few tens of seconds) can be easily demonstrated.
much smaller than 1, homogeneity of the solid temperature is So, an even more efficient cooling process would be to pour
insured. In our case, the Biot number can be written as: the coffee again into anew cup (as massive as possible), and
repeat if necessary. Because everyday life situations are an
(28) unlimited source of scientific questions, what will happen if
we add sugar to the liquid? Will this influence the cooling
So with Uw"" 14 Wm 2 °C 1,D =4x 10 2 m, and Awa,= 0.67W rate? This is another story, worth being discussed-around
m 1 °C 1, we obtain Bi= 0.8. This value is not "much" smaller a cup of coffee!
than 1, but we have considered here that only thermal con- NOMENCLATURE
duction occurs in the liquid, while free convection is actually A area (m 2)
present, and greatly increases the inner transfer. For instance, A term defined by Eq. (22)
we can estimate the enhancement of the "apparent" conductiv- B term defined by Eq. (23)
ity by the value of the Nusselt number, Nu. To evaluate this Bi Biot number, Eq. (27)
value, we can use a simplified sketch and consider inner free Cp specific heat (J kg 1 °C 1)

142 Chemical Engineering Education

 CT total concentration (Mo! m 3) 8 temperature (C)
D diameter (m) subscripts
e thickness (m) 0 initial
logarithmic mean of partial pressures of air (Pa) a ambient
height (m) a!f a!f
0
free convection heat transfer coefficient (Wm- 2 C- 1)
0
av average
radiation heat transfer coefficient (Wm 2 C- 1) C cup
low or equimolar flux mass transfer coeff. (kg s- 1m- 2)
e external
characteristic length (m) internal
Lewis number= ratio of thermal and massic diffusivities s surface
mass of water (kg) V vertical
m mass of the cup (kg) w wall
9J1 molecule weight (kg MoJl) wat water
N molar density of flux (Mo! s 1m 2)
Nu Nusselt number, Eq. (29) REFERENCES
Q heat flux (W m 2)
Pr Prandt number, Eq. (29) 1. Cussler, E.L., Diffusion, Mass Transfer in Fluid Systems, p. 24, Cam-
pv vapor pressure (Pa) bridge Univ. Press
2. Coulson, J.M., andJ.F. Richardson, Chemical Engineering, Vol 1,205,
Qevap
evaporative heat flux (W m 2)
Pergamon Press
Ra Rayleigh number, Eq. (29)
3. McCabe, L., and J.C. Smith, UnitOperationsofChemicalEngineering,
t time (s) 422, McGraw Hill
u global heat exchange coefficient (Wm 2 C- 1)0

4. Hirata, H., S. Ohe, and K. Nagahama, Vapor Liquid Equilibria, Else-

w mass flux (kg s 1) vier
b.Hv massic latent heat of water (J kg 1) 5. Treyball, R., Mass Transfer Operations, p. 49, McGraw Hill
a thermal diffusivity (m 2 s 1) 6. Chilton, TH., andA.P Colburn, Industrial and Engineering Chemistry,
26(1183) (1934)
~ thermal expansion coefficient (K 1)
7. Coulson, J.M, and J.F. Richardson, Chemical Engineering, Vol. 1, p.
E emissivity of the surface
303, Pergamon Press
>, thermal conductivity
8. Parulekar, S.J., "Numerical Problem Solving Using Mathcad," Chem.
V kinematic diffusivity (m 2 s 1) Eng. Ed., 40(1) 2006
Q density (kg m 3) 9. Incropera, F.P, and D.P DeWitt, Fundamentals of Heat and Mass
o Stefan Boltzman constant= 5.67 x 10 8 (W m 2 K 4 ) Transfer, Wiley, New York (1985) 0

Vol. 41, No. 2, Spring 2007 143