Pergamon

PII: SOO38-092X(96)00133-8

SIMULATED

Printed in Great Britain. All rights reserved

0038-092X/97 $17.oO+O.C@

PERFORMANCE
OF THERMAL
SOLAR COOKER

P. KARIUKI NYAHORO,* RICHARD R. JOHNSON**

*Appropriate

Solar Energy Vol. 59, Nos. 1-3, pp. 11-17, 1997

0 1997 Elsevier Science Ltd

Technology

STORAGE IN A

and JOHN EDWARDS**

Centre, Kenyatta University, Box 43844, Kenya and **North

University, Raleigh, NC 27695-7910, U.S.A.
(Communicated

by Joachim

Carolina

State

Luther)

Abstract-An
explicit finite-difference method is used to simulate the thermal performance
of short-term
thermal storage for a focusing, indoor, institutional,
solar cooker. The cooker storage unit consists of a
cylindrical solid block. The block is enclosed in a uniform layer of insulation except where there are
cavities on the top and bottom surfaces to allow heating of a pot from storage and heating of the storage
by solar radiation. A paraboloidal
concentrator
focuses solar radiation through a secondary reflector onto
a central circular zone of the storage block through the cavity in the insulation. The storage is charged
for a set period of time and heat is subsequently
discharged to a pot of water. In these simulations a pot
of cold water is placed on the hot storage block and the time then estimated until the water either boils
or the temperature
of the water reaches a maximum value. Simulations are made for a given pot capacity
with the storage block made from either cast iron or granite (rock). The effects on cooker performance
are compared for a variety of height to diameter ratios of the storage block and size of the area of solar
input zone. 0 1997 Elsevier Science Ltd. All rights reserved.

1. INTRODUCTION

installed (Fig. 1). The concentrator

in this
cooker is parabolic-shaped, mild steel, grid supporting panels of aluminized plastic sheet. Both
pot and secondary reflector are mounted inside
a kitchen. If needed during non-sunny periods
the pot is moved laterally over an adjacent
combustion chamber where burning wood or
charcoal may be used to sustain the cooking
process. In the study reported here, a numerical
model is used to predict the behavior of this
type of cooker when it is operated with shortterm sensible heat storage. Since the addition of
a heat storage component makes the cooker
more expensive and complicated, it is necessary
to study the effects of thermal and geometric
parameters on the cooker performance in order
to provide guidance for the effective design of
such cookers.

Cooking with fuel wood consumes the bulk of

total national energy demand in many developing countries (Nandwandi, 1988). Due to the
increasing population
in these countries,
demand for both wood and agricultural land
has risen to such an extent that there is now a
net depletion of wood resources with some
serious present and potential consequences such
as soil erosion, food shortage, and fuel wood
shortage. In addition to improving efficiencies
of wood-based cooking devices and introducing
other biomass-based fuels, the use of solar cookers is one way of reducing the demand for
firewood. Advantages of using solar cookers
include free availability of solar radiation, accessibility to this radiation even in remote places
with little access to conventional energy supplies, and its higher abundance in hot, dry or
arid areas with the greatest need for fuel wood
substitutes.
Current designs of solar cookers normally
used are box cookers, concentrators, and flat
plate collector cookers. Common problems
associated with cooker designs (Kuhnke, 1987;
Grupp, 1990) include limitation to daytime and
outdoor use during sunny weather, small capacity, and slow cooking speed. In an attempt to
overcome some of the problems, indoor institutional hybrid cookers of the type designed by
Scheffler (Oehler and Scheffler, 1994; Otieno
and Scheffler, 1989) have been constructed and

pot
storage

tmct

mi

kitchen
WOII

-primary reflector
enlcentrot00

Fig. 1. Layout of components

of an indoor focusing hybrid
cooker similar to the design by Scheffler (1994) but including
a thermal storage block.
11

12

P. K. Nyahoso

A layout of components
for the solar cooker
is shown in Fig. 1. The layout of all but the
storage block is consistent with the basic Scheffler
talled at the Appropriate Technology
nyatta University,
Nairobi (Otieno
and Scheffler, 1989). The parabolic concentrator
is located outside the kitchen building. It is a
truncated
section of a parabola
of revolution
such that the focal point is well away from the
concentrator
itself. The rays from the sun are
directed through a hole in the kitchen wall on
to an angled reflector, finally focusing on an area
on the bottom of the storage block.
The concentrator
is on a two axis mounting
such that the dish can be manually
reset to
follow the path of the sun.
ecause the system
is not precisely focused, a slight adjustment
every half hour works quite adequately. In fact,
in cooking experience without storage, when the
pot has already been
t-ought to the boil, the
concentrator
is often left untracked
so as to
limit heating of the pot. The concentrator
is
constructed
from many flat, reflective panels
mounted in a shaped frame. The rays from the
concentrator
strike a secondary reflector in the
kitchen which redirects them on to the bottom
of the storage block. Simulation
studies are
carried out for only one concentrator
size.
The storage block is a cylindrical shaped solid
that is enclosed in insulation
At the bottom
center of the block the insulation
has been
removed so that the solar rays can impinge on
the storage surface. The solar input zone is set
back in a cavity because of the thickness of the
insulation.
The surface at the solar input zone
may be coated so as to be a selective absorber.
The solar input zone cavity is g!azed, just as in
solar collectors, to reduce the heat lost from the
Rot surface. At the top surface of the storage
block there is another centrally located cavity
in the insulation
to allow for placement of the
pot on the storage
block.
For simulation
purposes this area was assumed insulated during
solar heating of storage, and in intimate contact
ot bottom when cooking. The layout
of pot, storage block, and insulation
are shown
in Fig. 2. The 0.05 m3 (50 1); steel pot is completely fihed with water.

3.1. Finite

&fSeerence model

storage
cal coordinate
The

block is discretized as a cylindrisystem with axial symmetry. The

et d

Fig. 2. tieat storage elements of a solar cooker showing

storage block, insulation, and glazing.

PO;.

insulation is a!so discretized in the same coor&nate system, but with special nodal matchmg
near the interface between storage and insula-.
tion. The pot wall and bottom have a specia!
set of one-layer nodes. The water in the pot is
represented
by one node at the bulk water
temperature.
The interfacial
surfaces between
pot, storage and insulation
were assumed to
have contact resistance. The temperature
T(r,z,t)
is solved using an explicit finite difference
scheme based on a heat balance at each node
and the heat conduction
equation. The boundary conditions
on the outside surface of the
insulation
are specified as free convection
and
radiation
to the surroundings
at 20C. At the
radiation input zone there Is a specified solar
radiation
input uniformly distributed. over the
heat loss by convection
and
radiation through the glazed cavity to the surroundings
at 20C. Heat transfer between the
interior pot wall and the water in the pot was
either described by equations for convectlon or
equations
for boiling depending
on the pot
surface and water temperatures.

The
heat transfer mechanism
between the
interior pot surfaces and the water in the pot
may either be natural convection
or boihng.
The descriptions
of these heat transfer types,
and the rest&ant heat transfer rates, are quite
diEerenf, necessitating
a criterion to select the
right mode, and protection
against numerical
instability that may resuit from the step change
in values.
The coefficient for natural convection
heat
transfer inside the pot, h,, was evaluated from
the correlation
suggested by Evans and Stefany

Simulated

performance

of thermal

Ra

in a solar cooker

13

bottom of the pot is assigned a boiling heat

removal rate of:

(1966) and given as:

hi&
Nui = k,

storage

= 0.55Ra:4

gH;C,,d, CL(t) - T,l

hkl

(1)

(2)

Here Nu, and Rai are the instantaneous Nusselt

and Rayleigh numbers, T,(t) is the timedependent, area-averaged, instantaneous temperature on the surface in contact with water,
and T, is the initial temperature of water in the
pot. The value of C =0.55 in eqn (1) is the value
suggested by Evans and Stefany (1966). The
temperature-dependent
properties of specific
heat, conductivity, and viscosity for water were
taken from expressions given by Yaws (1977)
and the density from an expression given by
Perry (Nyahoro, 1992) and evaluated at temperature T,.
When the temperature of a hot surface
immersed in a liquid exceeds the saturation
temperature of the liquid by a few degrees, the
mechanism of heat transfer changes to one of
boiling. For water at atmospheric pressure this
occurs when the surface reaches a threshold
temperature of about 105C (Incropera and
Dewitt, 1990). The transition occurs even if the
liquid is sub-cooled relative to the saturation
temperature. For surface temperatures between
105C and 130C there is likely to be pool
nucleate boiling. During the boiling phase the
vertical wall of the pot and the water are lumped
into a compound node which communicates by
conduction with the lid and side insulation, and
by the boiling mechanism with the bottom
surface of the pot. The boiling flux q,,, used for
sub-cooled nucleate boiling in water, from the
heating at the bottom of the pot, is that by
Rohsenhow and given by Incropera and Dewitt
(1990) as:

The excess temperature is calculated as AT,,=

(K - 100) where T, is the inside surface temperature at the bottom of the pot. The empirical
constant CSf is assigned a value of 0.132 for a
combination of water and polished stainless
steel (Vahon et al., 1968). The exponent n on
the Prandtl number has a value of 1.0 for water
and 1.7 for other liquids. Property terms with
subscripts 1 and v are, respectively, evaluated
for liquid and vapor phases at saturation temperature. During boiling, each node at the

where T,, is the temperature of the compound

node (water temperature) and A,(m,n) is the area
of node (m,n) in contact with the water. This
equation ensures that no node at the bottom of
the pot is assigned a heat removal rate that is
so high as to cause its temperature to drop
below T,, at the end of a simulation time step.
The criterion for transition to boiling, and the
expression for heat transfer during boiling, contain the area-weighted surface temperature T,,
that covers the heated zone of the pot bottom.
The criterion for transition from free convection
to boiling was taken as T,= 105.5C. When
boiling was determined to occur at the heated
zone, it was assumed to occur throughout the
water in the pot.
3.3. Modelfor heat transfer at the input zone
The zone where the input solar radiation is
incident on the storage block is a geometrically
complicated region in which both radiation and
convection are important. There is a cavity
formed in the insulation to allow access to the
storage block. The cavity may be left open or it
may be closed with a transparent glazing which
allows radiation to pass easily into the cavity,
but restricts radiation and convection of heat
out of the cavity. The open cavity is treated as
a heat transfer zone with gray surfaces, view
factors, and free convection coefficients. The ray
tracing technique is used to determine the effective fraction of input radiation that is absorbed
at the elemental surfaces in the input cavity.
The glazed cavity has the same features, but
includes the multiple reflections, refractions and
absorptions at the glazing and within the cover
that are analyzed in a way that is similar to the
techniques used in flat-plate solar collectors as
described by Duffie and Beckman (1991).
3.4. Convection and radiation from the exposed
outer surfaces

The dimensions and temperature range of the

cooker system satisfy the condition proposed
by Churchill and Chu to calculate natural convection heat transfer from exposed surfaces
(Incropera and Dewitt, 1990) as given in the
expression for Nusselt number based on effective

P. K. Nyahoro

14

height H of the vertical

Nu,=

=O.6&+
a

surface:
0.67Ra~4
419

(5)

Property terms for air in this correlation

are
evaluated at the arithmetic mean of the surface
and
room
temperatures
using
expressions
applied by Nyahoro (1992).
Radiation heat loss from the outside surface
is dealt with by considering the surroundings
to
be at 20C. At the beginning of a new time step,
the emissive power is calculate
temperature
of the previous
time step. The
matrix method (Incropera and Dewitt, 1990) is
used to analyze
ra iation exchange
at the
surface.

The time step was initially set according to

ility condition
based on the grid size
and boundary
condition.
The time step was
then dynamically
reduced at the time of the
transition to boiling as described below. At the
time of transition from convection in the pot to
a mechanism of boiling, there is quite a change
in the heat balance on the elements that make
up the pot wall The thermal resistance associated with boiling is only a fraction of the thermal
resistance associated with natural convection.
ecause eqns (I) and (3) are different in structure, there is not a logical mechanism for developing a smooth transition from one to the other.
If the time step is too large it can lead to a
situation which is unstable. The description
of
the procedure employed will explain the problem and the solution. After each time step in
the convection mode, the area-weighted
average
temperature
of the pot bottom in the heating
zone is calculated and compared to 1055C. If
the temperature
has risen above 105.5C, indicating the initiation
of boiling, then the next
time step is calculated according to the boiling
equation, It is quite possible that the temperature at the pot bottom may decrease because
of the increased heat loss to the water. There is
also a time delay before there is a change in
heat conducted to the pot from storage because
of the explicit formulation
of the numerical
equations. If the temperature falls below 105.5C

et ui

then it either means that there is a reversion to

convection or that the time step was too large.
In the case that the tern~~rat~~~e goes from being
above lO5.5C to being
elow 1055C the
calculation
is redone, but with a shorter time
step. Should the time step be less than a preset
rn~~~rn~rn limit and the ~~rn~erat~re is still less
than IO.YC then it is assumed that t~a~sit~~~
did not take place and the calculation is replaced
c;alculation.
by a shorl
Bime
see ) cenvection
CXherwise
hbe calculation
will progress with the
assumption
of boiling heat transfer. The preset
minimum time step for these ~irn~lat~~~s is 6 s.
The procedure for trans
convection
is similar but

than 1055C for the present ~~~~~~~

time
step, then it is assumed Ihat the -mode will
change to convection
on the subsequent
tim~e
step.

Shown

Insulation

radiation

in Fig. 2 is t

surrounds

input

the solid to form an in

f below the solid an

area OF zone.

4.9. Storage charging phase

Starting with the pot cavity filled with a
of insulation (no pot), concentrated
solar
tion at a constant
rate of ar strikes the so
ough the open focal cavity for
time of charge. ~adiatio~
is
eat by a coating of solar abso
tance z!f and thermal
e~missiv~t~
Ed at the
heating zone. At the end of the charging phase,
the input cavity is sealed with a block of insulation. For simulations
reported here the storage
charge time is fixed at 140 min.
4.2. Pot

heating

phase

After the storage

(storage

char

discharge)

base is compl
ed from the
replaced with a metallic pot of co
water of diameter D, and height HP. An insur the pot. The pot is
the water either boils

Simulated
Table 1. Nominal

dimensions

Specific
heat
(J,$k)
Cast iron
Granite
Pot
Insulation
Input zone
Input cavity
Pot cavity
External surface

420
820
480
loo0

performance

and constant

Density

of thermal

properties

assumed

in a solar cooker

for the storage

Conductivity

(kg?m)

(W&Q

1212
2640
8055
160

52
3
15.2
0.1

Absorptance
lx

Emissivity
E

0.10
0.05
0.95
0.57

0.95
0.10

or reaches a maximum temperature below that

of boiling.
5. SIMULATED

storage

CASES AND RESULTS

5.1. Scope of simulations

Simulations are carried out for granite and

cast iron storage blocks of equal size. For one
set of three simulations, the shape of the storage
block was varied by changing the height to
diameter ratio while maintaining the same mass
of storage. For another set of four simulations,
the focal area on a nominally sized storage
block was varied while the total rate of heat
input remained the same. The properties and
sizes of the components are presented in Table 1.
The following parameters were fixed for all
simulations: 50 1 pot size, input radiation rate
of 7 kW (estimated from two solar concentrators
around midday in Nairobi), initial start-up temperature of 20C storage charge time of 140 min,
mesh size of 0.02 m, and tlrermal contact conductance of 3.600 W/(m.C) (Incropera and
Dewitt, 1990).
5.2. Typical time history and energy inventory
Typical time histories for the temperature
within the nominally sized storage block, at a
node located on the solid side of the solid-pot
interface, are shown in Fig. 3. The diameter of
the focal zone is 0.46 m. At the 140 min mark,

Fig. 3. Time history of the pot-storage

interface temperature for a cast iron (CI) storage unit and for a granite (Gr)
storage unit. Charging occurs for the first 140 min, followed
by discharging
for the remaining time.

15

materials

and for the 50 1 pot

Nominal
radius

Nominal
height

Thickness

(Z

(i)

(5

0.346
0.346
0.22

0.1
0.1
0.32

0.2225

0.3225

0.0025
0.06

the solar heating is cut off and the pot is placed

on the storage block. The maximum temperatures are 570C for granite and 430C for
cast iron.
The temperatures from 140 min to the end of
the simulation represent the time to bring the
water to boil. From Fig. 3 it can be seen that it
takes 56.0 min for granite and 5.5 min for cast
iron. The temperature values do not represent
the average storage temperature. They represent
the local temperature on the storage side of the
storage-pot interface. As can be seen from
the times, the heating rate of the pot in the cast
iron case is an order of magnitude faster than
that of the granite (heat added to the pot is the
same in each case). The higher conductivity of
the cast iron allows free flow of heat from
throughout the storage to the storage-pot interface. The difference of temperature values
between the cast iron and granite at the storage-pot interface is enough to produce the
difference between natural convection (low rate
of heat transfer for granite) and boiling heat
transfer (high rate of heat transfer for cast iron)
over a significant part of the water heating
phase.
Shown in Fig. 4 is a bar chart which gives
some perspective on what happens to the radiation incident on the storage block. For the
140 min charge at 7.128 kW, the total incident

Cast IrOn

GK3Mtf?

Fig. 4. Fractions of the 60 MJ energy input that are lost,

left in storage, or used to heat water for two cases using
heat storage in either cast iron or granite.

16

P. K. Nyahoro

energy is 60 MJ.
f that, 16.8 NlJ goes into
heating the 50 1 of water in the pot from 20C
to lWC. Part of the balance of 43.2 MJ is lost
to the surroundings, and part is left as stored
energy in the storage block. The figure shows
that for granite the proportion that is lost is
much higher than that for cast iron because of
the longer time to heat the water an
temperature in the focal zone. The 29.4 MJ of
energy remaining in the cast iron storage at the
end of the simulation suggests that it is at an
and that there
average temperature of 300
ation of storage
should be an alternative corn
size, pot size, and heating schedule to better
utilize the energy.
5.3. Eflect of storage block shape on performance
The values in Table 2 reflect the results of
simulations for three height to diameter ratios
of the storage block while maintaining the same
block volume. For these simulations there was
no selective coating in the focal zone and CQ=
er= 0.95. The focal diameter is 0.46 m. The time
to heat the water from 20C to a final temperature and the heat fraction lost both decrease
with mcreasing height to diameter ratio for the
cast iron, suggesting that H/D =0.2 is better
than a ratio less than 0.2. The granite data
suggests that at values ofH/D ~0.2, it is doubtful
that the water ever reaches the target 108C for
the nonselective coating in the focal area.

5.4. E#ects offocal zone area on performance

ratio of the solar collector

Table 2. Simulation

results for cooker

and: therefore, the cost and complexity of the

concentrator. The values presented in Table 5

these simulations

ere is a selective coating in

mum tern~e~at~~~ seen in the focal zone for

granite. With a small foca? area j&=%15 m),
the temperature is as high as l35OC. This value
nreasonably high for any selective coat@,
produces high thermal stresses jn the granite. At a focal radius of 0.25 rn~ the temperature
is much reduced, but still very high at 78EC.
The cast iron shows much less variation in
gemperature with focal radius because the heat
can be electively ~onducte away from the focal
zone. The cast iron shows little variation in
absorption efhciency with changes in focal size,
suggesting that there is an even tradeoff
between the additional area fm beat loss and
the reduced focal temperature. The focal area is
important because it irectly relates to the
sophistication (and therefore cost) of the solar
concentrator needed to deliver the solar radiation to the cooker. A larger focal area means a
less sophisticated concentration system.

mulation of a focusing type solar

thermal storage remonstrates
d
characteristics for granite and cast iron as the
storage material. Cast iron is shown to have
Wi

The size of the input zone area will determine

the concentration

et al

with granite

or cast iron for three storage

Cast

Granite

Granite

Height, H, (m)
Diameter, D, (m)
Ratio, NJD,
Time to heat water, t (min)
Final temperature
(C)
Neat fraction lost
Heat fraction left
Heat fraction in water

Table 3. Overall

Focal radius, Rr (m)

Efficiency gab-s
Time to heat water, t (min)
Focal temperature.
Z fC)

0.06
0.892
0.07
? 70
77
0.70
0.10
0.20

simulation

Granite

0.1
0.692
0.15
295
93
0.62
0.12
0.26

results for cooker

Granite

Granite

0.15
0.55
66.3
1350

0.19
0.63
56.0
1053

with granite

(thin)

Cast
iron
(medium)

Cast
iron
(thick)

0.06
0.892
0.07
15.0
.#
0.30
0.42
0.28

0.1
0.692
0.15
7.2
IO@
0.24
0.48
0.28

0.126
0.616
0.20
6.8
100
0.23
0.49
0.28

II-on

0.126
ii.616
0.20
253
100
0.58
0.14
0.28

shapes

or cast iron For four focal radii

Cast

Cast

Granite

Cast
Iron

Cast

Granite

II-OX2

IiQn

h-on

0.23
0.68
56.2
890

0.25
0.70
59. !
781

O.i5
0.87
5.0
555

0.19
0.86
5.2
538

0.23
0.86
5.5
525

0.25
0.86
5.6
522

Simulated

performance

of thermal

shorter cooking times and less heat loss than

granite. With cast iron much of the heat remains
unused in storage by the end of the simulation,
but is still available to sustain boiling thereafter.
The results indicate that the height of the storage
block should be at least one-fifth of the diameter
of the block.
For the same total heat input, the temperature
at the heating zone will decrease as the area of
the focal zone increases. This result is important
because an increase in the focal zone area
implies a reduction of the cost and complexity
of the concentrator that results from a less
critical concentration ratio. It is also an important design parameter that may be selected in
order to limit temperature, limit thermal stress,
or allow for the use of an absorber coating.
Overall changes in performance, because of
different focal radii and shape, are all relatively
small. Because the maximum temperatures for
the granite are beyond the use of normal
absorber coatings, and the time taken to heat
storage and cook the meal are lengthy, the
cooker with storage has limited appeal in the
mode simulated here, but short-term storage
could serve in practice to store incoming energy
in moments when the cooker is not in use, or
to dampen temperature fluctuations due to
changing weather conditions. Whereas the simulations done here reflect the ability to boil water,
a cooker with solid storage may be better suited
to a baking situation, where the intention is to
extend the baking period and not necessarily to
time shift the cooking operation.
NOMENCLATURE

CP specific

heat (J/kg.K)
diameter (m)
height (m)
heat transfer coefficient ( W/m.K)
latent heat of vaporization
(J/kg)
thermal conductivity
( W/m.K)
Nusselt number
Prandtl number
rate of heat transfer (W)
rate of heat transfer per unit area (W/m)
R4a Rayleigh number

D
H
h
hfp
k
Nu
Pr
4
I,

storage

in a solar cooker

17

temperature
(C)
time (min)
thermal absorptance
thermal coefficient of expansion (l/K)
thermal emissivity
storage efficiency: fraction of heat that approaches
the absorber that ends up in storage
dynamic viscosity (N.s/m2)
kinematic viscosity (m/s)
density ( kg/m3)
surface tension at the liquid-vapor
interface

Subscripts
a
b
f
i

1
s

air
boiling
focal area (or input zone area)
insulation
liquid (water)
surface

REFERENCES
Duffie J. A. and Beckman W. A. (1991) Solar Engineering
of Thermal Processes, 2nd Edn. Wiley Interscience,
New York.
Evans L. B. and Stefany N. E. (1966) An experimental study
of transient heat transfer to liquids in cylindrical enclosures. Chemical Engineering Progress, Symp. Series, 64,
209-215.
Grupp M. (1990) Solar cooking: lessons from the past, hope
for the future. Proc. First World Renewable Energy ConPress, Oxford, Vol. 2,
gress, Sayigh (Ed.). Pergamon
pp. 1325-1327.
Incropera
F. P. and Dewitt D. P. (1990) Introduction to
Heat Transfer, 2nd Edn. John Wiley, New York.
Kuhnke K. (1987) Solar cooking: obstacles and opportunities. Gate: Questions, Answers, Information, 1, i&14.
Nandwani S. S. (1988) Experimental and theoretical analysis
of a simple solar oven in the climate of Costa Rica-I.
Solar and Wind Technology, 5(2), 159-170.
Nvahoro P. K. (1992) The effects of thermal, ootical and
* geometric paiametkrs on the performance
bf &I institutional focusing solar cooker. M.Sc. Thesis, Kenyatta
University.
Otieno H. and Scheffler W. (1989). Drawings for a Solar
Technology
Center,
Kenyatta
Cooker. Appropriate
University.
Oehler U. and Scheffler W. (1994) The use of indigenous
materials for solar conversion.
Solar Energy Materials
and Solar Cells, 33, 379.
Vahon R. I., Nix G. H. and Tanger G. E. (1968) Evaluation
of constants for the Rohsenhow pool boiling correlation.
J. Heat Transfer, 9llC, 239-247.
Yaws C. L. (1977) Physical Properties: A Guide to the Physical, Thermodynamic and Transport Property Data of Industrially Important Chemical Compounds. McGraw-Hill,
New York.