Chapt 17
",
Crystallization
f1
lUrystallization is a solid-fluid separationoperation in
which crystallineparticlesare formed from a homogeneous
fluid phase.Ideally,the crystalsale a pure chemical,obtained
in a high yield with a desirableshapeand a reasonablyuni-
form and desirable size. Crystallization is one of the oldest
known separationoperations,with the recovery of sodium
chloride as salt crystals from water by evaporation dating
back to antiquity. Even today, the most common applications
arethe crystallizationfrom aqueoussolutionof variousinor-
ganic salts,a short list of which is given in Table 17.1.All
thesecasesare referred to as solution crystallization because
the inorganic salt is clearly the solute,which is crystallized,
and water is the solvent,which remainsin the liquid phase.
The phasediagram for systemssuitablefor solution crystal-
lization is a solubility curve, such as shown in Figure 17.1a
and describedearlier in Chapter 4.
For the formation of organic crystals, organic solvents
such as acetic acid, ethyl acetate,methanol, ethanol, ace-
tone, ethyl ether, chlorinated hydrocarbons, benzene, and
petroleum fractions may be preferred choices, but they must
be used with great care when they are toxic or flammable
with a low flash point and a wide range of explosive limits.
For either aqueousor organic solutions,crystallizationis
effectedby cooling the solution, evaporatingthe solvent,or
a combinationof the two. In somecases,a mixture of two or
more solventsmay be best,examplesof which include water
with the lower alcohols, and normal paraffins with chlori-
nated solvents.Also, the addition of a second solvent is
sometimesused to reducethe solubility of the solute.When
water is the additionalsolvent,the processis calledwatering,-
out; when an organic solvent is added to an aqueoussalt
solution, the processis called salting-out.For both of these
casesof solvent addition, fast crystallizationcalledprecipi-
tation can occur, resulting in large numbers of very small
crystals.Precipitationalso occurs when one product of two
I7.O INSTRUCTIONALOBJECTIVES
After completing this chapter,you should be able to:
. Describedifferent types of crystallization.
. Explain how crystalsgrow.
. Explain how crystal-sizedistribution can be measured,tabulated,and plotted.
644
reactingsolutionsis a solid with low solubility.For example,
when aqueoussolutions of silver nitrate and sodiumchloride
are mixed together, insoluble silver chloride is precipitated
leaving an aqueous solution of mainly soluble sodium
nitrate.
When both componentsof a homogeneous,binarysolu-
tion have melting (freezing) points not far removedfiom
eachother, the solution is referred to as a melt. If, asin Figure
17.1b, the phase diagram for the melt exhibits a eutectic
point, it is possibleto obtain, in one step calledmelt crystal-
lization, pure crystals of one component or the other,de-
pending on whether the composition of the melt is to theleft
or right of the eutecticcomposition.If, however,solidsolu-
tions form, as shown in Figure ll .lc, a processof repeated
melting and freezingsteps,called/ractional melt crystalliza'
tion, is requiredto obtain nearly pure crystallineproducts. A
higher degree of purity can be achieved by a technique
called zone melting or refining. Examples of binary organic
systems that form eutectics include metaxylene-paraxylene
and benzene-naphthalene.Binary systems of naphthalene-
beta naphthol and naphthalene-B naphthylamine, which
form solid solutions,are not as common.
Crystallizationcan also occur from a vapor mixture by a
processmore properly called desublimation.A numberof
pure compounds,including phthalic anhydride and benzoic
acid, are produced in this manner. When two or more com-
pounds tend to desublime, a fractional desublimation
processcan be employedto obtain near-pureproducts.
Crystallization of a compound from a dilute, aqueous
solution is often precededby evaporation in one or more
vessels,called effects.to concentratethe solution, and fol-
Iowed by partial separationand washing of the crystalsfrom
the resulting slurry, called the magma, by centrifugation or
filtration. The processis completedby drying the crystalsto
a specifiedmoisturecontent.
 17.0 InstructionalObiectives 645

Explain the importance of supersaturationin crystallization.

Differentiate between primary and secondarynucleation of crystals.
Use mass-transfertheory to determine rate of crystal growth.
a Describe major types of batch and continuous solution-crystallization equipment.
Apply the MSMPR model to design of a continuous, vacuum, evaporating crystallizer of the draft-tube baffled
(DTB) type.
a Understand precipitation.
Describe equipment for melt crystallization.
a Apply mass-transfertheory to a falling-film melt crystallizer.
Apply the ideal zone-melting model.
a Differentiate between crystallization and desublimation.
Describe evaporation equipment.
Derive and apply the ideal evaporator model.
Design multiple-effect evaporation systems.

Industrial Example A representativecommercial processfor producing 4,205

lbftr (dry basis)'of MgSOo .'lHzO crystalsfrom a l0 wtvo
Considerthe crystallizationof MgSOo .7HzO (Epsom salt)
aqueoussolution at 1 atm and 70'F is shown in Figure 17.3.
from an aqueoussolution. The solid-liquid phasediagram for
This solution is first concentratedin a double-effect evapo-
the MgSOo .H2O systemat 1 atm is shown in Figure 17.2.
ration system with forward feed and then mixed with recy-
Depending on the temperature,four different hydrated forms
cled mother liquors from the hydroclone and centrifuge. The
of MgSO+ are possible: MgSOo .H2O, MgSOa .6H2O,
combined feed of 14,326lb/h containing 31.0 wtTo MgSOa
MgSOo .7HzO, and MgSOa ' l2HzO. Furthermore,a eutec-
at 120"F and I atm enters an evaporative, vacuum crystal-
tic of the latter hydrate with ice is possible. To obtain the
lizer constructedof 316 stainlesssteel and shown in more
usually desired heptahydrate, crystallization must occur in
detail in Figure 17.4.
the temperaturerangefrom 36"F to 118"F(Pointb to Point c).
The crystallizer utilizes internal circulation of 6,000 gpm
Within this range, the solubility of MgSOa (anhydrous or
of magma up through a draft tube equipped with a 3-Hp
hydrate-freebasis) increasesalmost linearly from about 21 to
marine-propeller agitator to obtain near-perfect mixing of
33 wl%o.

Table 17.1 Some Inorsanic Salts Recovered from Aqueous Solutions

Chemical Name Formula Common Name Crystal System

Ammonium chloride NH+Cl sal-ammoniac cubic

Ammonium sulfate (NH+)zSO+ mascagnlte orthorhombic
Barium chloride BaClz.2HzO monoclinic
Calcium carbonate CaCO: calcite rhombohedral
Copper sulfate CuSOa.5H2O blue vitriol triclinic
Magnesium sulfate M g S O a. 7 H 2 O Epsom salt orthorhombic
Magnesium chloride MgCl2 '6H2O bischofite monoclinic
Nickel sulfate NiSO+ '6HzO single nickel salt tetragonal
Potassiumchloride KCl muriate of potash cubic
Potassiumnitrate KNOg nitre hexagonal
Potassiumsulfate KzSOa arcanite orthorhombic
Silver nitrate AgNO3 lunar caustic orthorhombic
Sodium chlorate NaCIO: cubic
Sodium chloride NaCl salt, halite cubic
Sodium nitrate NaNO: chile salt petre rhombohedral
Sodium sulfate NazSO+.10HzO glauber'ssalt monoclinic
Sodium thiosulfate NazSzO: . 5HzO hypo monoclinic
Zinc sulfate ZnSOa'7H2O white vitriol orthorhombic
646 Chapter17 Crystallization

qt
o
l
950
o
6
E
o
o
JU

o
o
I 15"
o Eu t e c t i c
o 10'
100% 100%
o ortho Composition para
E ( b ) E u t e c t i c - f o r m i n gs y s t e m o f o r t h o - a n d p a r a c h l o r o -
0)
E n i t r o b e n z e n es y s t e m s u i t a b l e f o r m e l t c r y s t a l l i z a t i o n

= 218
3
o
o
7
-4,: uid-
= O
o
; = 160
/s
a tid
o

o
F 120

97.5'
z
0
10 20 30 40 50 60 70 80 90 100 80
Temperature,
"C 100mol% 50 100mol%
p h e n a n t hr e n e anthracene
(a)Aqueoussystemssuitablefor solution
crystallization Composition
(c)Solid-solution system suitable for fractional melt crystallization

Figure L7.1 Different types of solubility curves.

lFromHandbookof SeparationTechniques for ChemicalEngineers,2nd ed,.,PA. Schweitzer,
Editor-in-chief,
McGraw-Hill,NewYork (1988)with permission.l

the magma. Mother liquor, which is separatedfrom crystals
during upward flow outside of the skirt baffle, is circulated
externally at the rate of 625 gpm, by a 10-Hp stainless-steel
pump, through a 300-ft2 stainless-steel,plate-and-frameheat
exchanger, where 2,052,000 Btu/hr of heat is transferred to
the solution from 2,185 lb/h ofcondensing 20 psig steamto
provide supersaturationand energy to evaporate 2,31I lblh
of water in the crystallizer.
The vapor leaving the top ofthe crystallizer is condensed
by direct contact with cooling water in a barometric
condenser,attachedto which are ejectors to pull a vacuum of
0.867 psia in the vapor spaceofthe crystallizer.The product
magma, at a temperatureof 105'fl consistsof 7,810 lb/h of
mother liquor saturated with 30.6 wt%oMgSOa and 4,205
lb/h of crystals. This corresponds to a magma containing
357o crystalsby weight or 3O.2Vocrystals by volume, based
on a crystal density of 1.68 g/cm3and a mother liquor density
of 1.35 g/cm3. The boiling-point elevation of the saturated
mother liquor at 105'F is SoF.Thus, the vapor leaving the
crystallizer is superheatedby the same 8oF.The magmares-
idence time in the crystallizer is 4 hours, which is sufficient
to produce the following crystal-size distribution:
35 wtVoon 20 mesh U.S. screen
80 wtVoon 40 mesh U.S. screen
99 wtVoon 100 mesh U.S. screen
The crystallizer for the representativeprocessis 30 ft high
with a vapor-spacediameter of 5-l 12 ft and a magma-space
diameter of I 0 ft. The magma is thickened to 5Owt%ocrystals
in a hydroclone, from which the mother-liquor overflow is
recycled to the crystallizer and the underflow slurry is sentto
a continuous centrifuge, where the slurry is further thickened
to 65 wtTo crystals and washed. Filtrete mother liquor from
the centrifuge is also recycled to the crystallizer. The cen-
trifuge cake is fed to a continuous direct-heat rotary dryer to
reduce the moisture content of the crvstals to 1.5 wt%.
 17.0 Instructional Obiectives 647

b' t+o

E
o
tzo
E
F9 1oo + MgSO
Eut + MgSOa.12H2O 20
tl
tt Nol l - N

| | | t\ tTl -r"
S o l n + M g S O a . 1 2 H 2 O\ \ lNl
I
. -----.]-
+
t{l
t v l
o
t@l
t c e* ' s o t n | / t o ---J-
I -h
t> Figure 17.2 Solid-liquid phase diagram for the MgSOa-
f M g S O o .1 2 H zH2O system at I atm.
g l+ MgSOa I
201 [FromW.L. McCabe,J.C.Smith,andP.Harriott,Unit Operations
0 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50 of ChemicalEngineering,5thed.,McGraw-Hill,New York
Concentration,mass fraction, MgSOa ( 1993)with permission.l

Double-effectevaporator system Crystallizer

Vapor

2,311lblh H2OVapor
Combinedfeed
14,326lblh
31 wt% MgSOa
Condensate Condensate 120F

=
LO

Feed E3
10 wt% MgSOa
1 atm -o_O
70'F F+E.
o)x6N
JYFFS

Centrifugalfilter

Hydroclone

a
Underflow
Recyclefiltrate 50 wt% crystals
s
;
|l)
(o

Rotarydryer Figure 17.3 Processfor

productionof MgSOa. 7HzO.
648 Chapter17 Crystallization

crystal. This led to the concept of a spacelattice as a regular

3-Hp Cooling arrangementof points (molecules, atoms, or ions) such that
marine water * if a line is drawn between any two points and then extended
propeller
2,311-tblh
vapor t
A i r ejector
in both directions, the line will pass through other lattice
points with an identical spacing.In 1848, Bravais showed
that only the 14 spacelattices shown in Figure 17.5are pos-
Barometric
conoenser sible. Based on the symmetry of the three mutually perpen-
dicular axes with respectto their relative lengths (a, b, c) and
the angles (a, 9, ^y) between the axes, the 14 lattices can be
Draft classified into the sevencrystal systemslisted in Table l1 .2.
tu be For example, the cubic (regular) system includes the simple
cubic lattice, the body-centered cubic lattice, and the face-
centered lattice. Examples of the seven crystal systemsare
included in Table 17.1. The five sodium saltsincludedin that
20 psig steam table form three of the sevencrystal systems.
2Jes tblh Actual crystals of a given substanceand a given crystai
Heat exchanger
2,052,000Btu/h system can exhibit markedly different appearanceswhen the
3ooft2 faces grow at different rates, particularly when these rates
vary markedly from stunted growth in one direction, so asto
Magma
product give plates, or by exaggeratedgrowth in another direction,to
12,015tblh give needles.For example,potassiumsulfate,which belongs
Condensate 35 wt% crystals
105"F to the orthorhombic-crystal system, can take on any of the
Magma
circulating shapes (crystal habits) shown in Figure 17.6, including
pump Combined feed plates, needles,and prisms. When product crystals of a par-
625 gpm 14,326tblh
1 0H p 31 wtY" MgSOa ticular crystal habit are desired, experimental researchmay
120"F be required to find the necessary processing conditions.
Figure L7.4 Crystallizer for production of MgSOa ' 7H2O crystals. Modifications of crystal habit are most often accomplished
by deliberate addition of impurities to the solution.

I7.1, CRYSTAL GEOMETRY Crystal-SizeDistributions

In a solid, the motion of molecules,atoms,or ions is restricted
largely to oscillations about fixed positions. If the solid is
amorphous, these positions are not arranged in a regular or
lattice pattern; if the solid is crystalline, they are.Amorphous
solids are isotropic, suchthat physical properties are indepen-
dent of the direction of measurement;crystalline solids are
anisotropic, unless the crystals are cubic in structure.
When crystals grow, unhindered by other surfacessuch as
container walls and other crystals, they form polyhedrons
with flat sides and sharp corners. Crystals are never spheri-
cal in shape.Although two crystalsof a given chemicalmay
appear quite different in size and shape, they always have
something in common, known as the Law of Constant Inter-
facial Angles, proposed by Hauy in 1784. This law states
that the anglesbetween correspondingfaces of all crystals of
a given substanceare constant,even though the crystals vary
in size and in the development of the various faces (called
the crystal habit). The interfacial angles and lattice dimen-
sions can be measured accurately by x-ray crystallography.
As discussedby Mullin [1], early investigatorsfound that
crystals consist of many units, each shaped like the larger
Typical magmas from a crystallizer contain a distribution of
crystal sizes and shapes.It is highly desirable to characterize
a batch of crystals (or particles in general) by an average
crystal size and a crystal-size distribution. This is often ac-
complished by defining a characteristic crystal dimension.
However, as shown in Figure 17.6, some crystal shapes
might require two characteristic dimensions, while one
might suffice for others. One solution to this problem, which
is particularly applicable to the correlation of transportrates
involving particles, is to relate the irregular-shapedpafiicie
to a sphereby the sphericity, r]l, defined as
surface area of a spherewith the same volume as the particle
ilt:
surface area of the oarticle
(17-1\
For a sphere,Q : l, while for all otherparticles,rf < 1.For
a spherical particle of diameter, Do, the surface area,sp,t0
volume, uo,ratio is
rc) : a1o o
: (T D3)I 6 o3o
(solup)spnere
Therefore,(17-1) becomes
,v: +(f),*,",. (r1-z)
 17.l CrystalGeometry 649

Table 17.2 The Seven Crvstal Svstems

CrystalSystem SpaceLattices Length of Axes Angles between Axes

Cubic (regular) Simple cubic a:b:c o:F:^y:90o

Body-centered cubic
Face-centeredcubic
Tetragonal Squareprism a:b<c gt:9:"y:90o
Body-centered square prism
Orthorhombic Simple orthorhombic alblc a:9:"y:90'
Body-centered orthorhombic
Base-centeredorthorhombic
Face-centeredorthorhombic
Monoclinic Simple monoclinic alblc o: I :90"
B ase-centeredmonoclinic 't +90
Rhombohedral (trigonal) Rhombohedral a:b:c o:9: "l+90
Hexagonal Hexagonal a:b*c cr:9:90"
1 I120
Triclinic Triclinic alblc e+9+'y +90"

MM
tt'ttlKj ffi
Itr Simple
D)/
Body-centered
)tN
Face-centered
cubic cubic cubic

M
I l-Kl I
Simple
tetragonal
Ire
Body-centered
tetragonal
Simple
orthorhombic
Body-centered
orthorhombic

Base-centered Face-centered Rhombohedral Hexagonal

orthorhombic orthorhombic

Simple Base-centered Triclinic Figure 17.5 The 14 spacelattices of

monoclinic monoclinic Bravais.
650 Chaoter17 Crystallization

Table 17.4 U.S. Standard ScreensASTM EII

Q Mesh Number

3-r/2
in.

0.220
Opening of SquareAperture

5.60 5600

A
4 0 . 18 7 4.75 4750
5 0.157 4.00 4000
6 0.132 J.J.) 3350
7 0.110 2.80 2800

V
Figure 17.6 Somecrystalhabitsof orthorhombic,
potassium-sulfate
crystals.
8
10
l2
l4
t6
0.0929
0.0787
0.0669
0.0551
0.0465
2.36
2.00
1.70
1.40
1.18
2360
2000
1700
1400
1180
:'i
j

i
18 0.0394 1.000 1000
20 0.0335 0.850 850
EXAMPLil 1?.1 25 0.0280 0.710 710
30 0.0236 0.600 600
Estimate the sphericity of a cube of dimension a on each side.
35 0.0197 0.500 500
40 0.0167 0.425 425
SOLUTION
45 0.0140 0.355 355
Ucube: a3 50 0.0118 0.300 300
scube:6a2 60 0.00984 0.250 250
0.00835 0.212 2t2
From(17-2), 80 0.00709 0.180 180

\':+(#):fr 100
r20
140
0.00591
0.00492
0.00417
0.150
0.125
0.106
150
Ili

106
Becausethe volumes of the sphere and the cube must be eoual. r70 0.00354 0.090 90
200 0.00295 0.075 /)
no]/a: a3
230 0.00248 0.063 63
Solving, 270 0.00209 0.053 53
DP : l'241a 325 0.oor77 0.045 45
400 0.00150 0.038 38
Then, 450 0.00126 0.032 JL

*:a/(1.241a):0.806

The most common methods for measuring particle size standard wire-mesh screens IASTM Ell (1989)] derived
are listed in Table 17.3 together with their useful particle- from the earlier Tyler standardscreens.The U.S. standardis
size ranges. Because of the irregular shapesof crystals, it based on a l-mm (1000-p,m)-squarcaperture-opening
should not be surprising that the different methods can give screencalled Mesh No. 18 becausethere are 18 aperturesper
results that may differ by as much as 50Vo.Crystal-size dis- inch. The standard Mesh numbers are listed in Table 17.4,
tributions are most often determined with U.S. (or British) where each successively smaller aperture differs from the
preceding aperture by a factor of approximately (2)t/4.Me-
Table 17.3 Methodsof MeasurinsParticleSize chanical shaking is applied to conduct the sieving operation,
using a stack of ordered screens.
Method Size Range, Microns When wire-mesh screensare used to determine crystal-
Woven-wire screen 32-5600 size distribution, the crystal size is taken to correspondto the
Coulter electrical sensor 1-200 screen aperture through which the crystal just passes.
Gravity sedimentation 1-50 However, becauseof the irregularity of particle shape,this
Optical microscopy should be consideredas a nominal value only. This is particu-
0.5-150
Laser-light scattering 0.04-2000 larly true for plates and needles,as illustrated in Figure 17.7.
Centrifugal sedimentation 0.01-5 Particle-size-distribution data, called a screen analysis,
Electron microscopy 0.001-5 are presentedin the form of a table, from which differential
and cumulativeplots can be made,usually on a mass-fraction
 l'l .l CrystalGeometry 651

Screen
apertu re qp aa
-i F
il tl o o o
tl tl
ll tl

W W
M @
W W
Plate

Sphere Cube Rough

particle

Oblong
Parallel- rougn
epiped particle

Figure 17.7 Different particle shapesthat just pass through the same screen

basis. Consider the following laboratory screen-analysis D,Average

Mesh Range Particle Size,mm Mass Fraction, r;
data presented by Graber and Taboada I2l for crystals of
Na2SO4. 10HzO (Glauber's salt) grown at about 18'C dur- -50 + 70 0.256 0.0448
ing an averageresidence time of 3'7.2 min in a well-mixed' -70 + 100 0.181 0.0147
laboratory,cooling crystallizer. The smallest screenused was
-100+ 140 0.128 0.0025
-140 + (r70) 0.098 0.0011
140 mesh, with particles passing through that screen being
retained on a pan. r.0000

Mesh Aperture, Mass Retained 7o Mass A plot of the differential screen analysis is shown in Fig-
Number Dormm on Screen. Grams Retained ure 17.8 both as (a) an x-y plot and as (b) a histogram' If a
l4 1.400 0.00 0.00 wide range of screen aperture is covered, it is best to use a
16 1.180 9.12 1.86 log scale for that variable.
18 1.000 32.t2 6.54
20 0.850 39.82 8.11
0.600 235.42 47.95 Cumulative ScreenAnalysis
30
40 0.425 89.14 18 . 1 5 Screen analysis data can also be plotted as cumulative-
50 0.300 54.42 11.08 weight-percent oversize or (which is more common) under-
0.212 22.02 4.48
'1.22 size as a function of screen aperture. For the above data of
100 0.150 1.47
Graber and Taboada [2], the two types of cumulative screen
r40 0.106 1.22 0.25
0.11 analysis are as follows:
Pan 0.50
491.00 100.00 Aperture, Cumulative wtTo Cumulative wtTo
Dormm Undersize Oversize

Differential ScreenAnalysis 1.400 100.00 0.00

1.180 98.14 1.86
A dffirential screen analysis is made by determining the 1.000 91.60 8.40
arithmetic-average aperture for each mass fraction that 0.850 83.49 16.51
passesthrough one screenbut not the next screen.Thus, from 0.600 35.54 64.46
the above table, a mass fraction of 0.0186 passesthrough a 0.425 17.39 82.61
screen of 1.400-mm aperture, but does not pass through a 0.300 6.31 93.69
0.2t2 1.83 98.17
screen of 1.180-mm aperture. The average of these two
0.1s0 0.36 99.64
aperturesis 1.290 mm, which is taken to be the nominal par-
0.106 0.11 99.89
ticle size for that mass fraction. The following differential
analysis is computed in this manner, where the designation Because0.11 wtVopassedthrough a 0.106-mm aperturebut
-I4 + 16 refers to those particles passing through a was retained on a pan with no indication of just how small
l4-mesh screenand retained on a 16-meshscreen. these retained particles were, the cumulative wtTo undersize
D'Average and oversize cannot be taken to 0 and 1007o,respectively.
Mesh Range Particle Size,mm MassFraction, r; The above cumulative screen analyses are plotted in
Figure 17.8c. The two curves, which are mirror images of
-14 + 16 1.290 0.0186
- 1 6+ 1 8 each other, cross at a median size where 5O wt%ois larger in
r.090 0.0654
-r8 + 20 0.925 0.0811 size and 50 wtTo is smaller. As with differential plots, a 1og
-20 + 30 0.725 0.4795 scaleis preferredif a large rangeof screenapertureis covered.
-30 + 40 0.513 0 . 181 5 Alog scaleforthe cumulativewt7o may also be preferredif an
-40 + 50 0.363 0 .l 1 0 8 appreciablefraction ofthe datapointslie below 107o.
 652 Chapter17 Crystallization

0.5 A number of different mean particle sizesthat are derived

from screen analysis are used in practice, dependingupon
the application. Of these, the most useful are: 0) surface-
mean diameter, (2) mass-mean diameter, (3) arithmetic-
mean diameter, and (4) volume-mean diameter.

Surface-Mean Diameter
The specific surface area(arealmass)of a particle of spheri-
cal or other shapeis defined by

A.: s p fl t r p : s o fu p p p (17-3)
Combining (I7 -2) and (I7 -3)

A, :6/,ltpoDo (17-4)
0.4 0.6 0.8 1.0
Averageparticlesize,mm For n mass fractions, .r;, each of average aperture Do,, from
(a)Differential
analysis a screenanalysis,the overall specific surfaceareais givenby

4
.\\ 6x;' = 6
: :-
D
?i *ppDp, Vpp
The surface-mean
diameteris definedby

,/l r6
, : - (17-6)
UpoDs
Combining(17-5)and(I7-6),

D,: (r7-1)
$"t
7:, bo,
which can be used to determine D5 from a screen analysis.
This mean diameter is sometimes referred to as the Sauter
mean diameter and as the volume-surface-mean diameter.
0.4 0.6 0.8 1.0 1.2 1.6 It is often used for skin friction, heat-transfer, and mass-
Aperturesize,mm transfer calculations involving particles.
(b)Histogramof differentialanalysis

Mass-MeanDiameter
The mass-meandiameter is defined by

80 D w: l * , 0 0 , (17-8)

s
st Arithmetic-Mean Diameter
360
E The arithmetic-mean diameter is defined in terms of the
o

G
number of particles, N;, in each size range:
f,+u n
E
(J
LN'Do,
r_zN : --\_-ii- (11-e)
20 / rrt

The number of particles is related to the massfraction of par-

ticles by
0.0 0.2 0.4 0.6 0.8 1.0 1.2
- mass of particles of average sizeDo,
Aperturesize,mm X "; :
(c)Cumulativeanalysis total mass
.1
Figure 17.8 Screen analysesfor data of Graber and Taboada [2]. Nif,(oo)'co (17-10)
_
_M
 17.2 ThermodynamicConsiderations
653

where From(17-8),
Dw :0.666mm
,f, : volume shapefactor defined by uo : f,D3o,
Mr : total mass (17-11) From(17-12),
-. 4.0032
For sphericalparticles,fr: r/6. l)x :
"
-: U.Jl6mm
12.5909
If (17-10) is solved for N;, substitutedinto (17-9), and
simplified, we obtain From(17-15),

ir+\ ), (17-r2)
-
D "' : l
/ r rl/l
| :0.430mm
\ 12.s909l
or :E'\'3
i {+)
7t \ o3r,/
Thus,the meandiametersvary significantly.

Volume-MeanDiameter I7.2 THERMODYNAMIC CONSIDERATIONS

The volume-meandiameter,Dr, is definedby
Solubility and Material Balances
; * : f, U,u',,)
(r,b',) ;-t
*, ;-l
(11-r3) Important thermodynamic properties for crystallization
operations include melting point, heat of fusion, solubility,

'":(#)
Solving (I7-13) for Dy for a constantvalue of /, gives heat of crystallization, heat of solution, heat of transition,
and supersaturation.For binary systemsof water and soluble
inorganic and organic chemicals, Mullin [1] presentsexten-
(r1-r4) sive tables of solubility, as a function of temperature, and
heat of solution at infinite dilution and room temperature
(approximately 18-25'C). Data in water are listed in Table
The corresponding relation in terms of x; rather than N, is 17.5 for the inorganic salts of Table 17.1, where solubility

'":(+)"'
obtainedby combining (I7-14) with (17-10), giving data are given on a hydrate-free basis.
Solubilities are seen to vary widely from as low as
4.8 9/100 g of water for Na2SO4(as the decahydrate)at 0'C
to 952 glTOOg of water for AgNO3 at 100'C. For KNO3, the
(17-1s)
solubility increasesby a factor of 18.6 for the same temper-
ature increase.
The solubility of an inorganic compound can be even
EXAMPLE I7.2 much lower than that shown for Na2SO4.Such compounds
are generallyconsideredto bejust slightly or sparinglysolu-
Using the screenanalysis data of Graber and Taboadagiven above,
ble or almost insoluble.The solubility of suchcompoundsis
compute all four mean diameters.
usually expressedas a solubility product, K., in terms of ion
concentrations. Data for several compounds are given in
SOLUTION
Table 17.6. For example,consider.AI(OH):, which is spar-
Since the data are given in weight (mass) fractions, use (17-7), ingly soluble with a solubility product of K, : I.1 x l0-r5
(l'7-8), (l'7-r2), and (17-15). at 18"C and dissolvesaccordingto the equation
Do,mm x x/Dp xDp ,/4 ,14 <+Ari;, + 3OHaq)
Ar(oH)r",
1.290 0.0186 0.0144 0.0240 0.0112 0.0087 By the law of mass action, the equilibrium constant,called
1.090 0.0654 0.0600 0.0713 0.0550 0.0505
the solubility product for dissolution, is given by
0.925 0.0811 0.0877 0.0750 0.0948 0.r025
0.725 0.4795 0.66t4 03476 0.9122 r.2583 - : -( c o , , - ; i c o s -) 3 : (cop*)(cag-)3: 1.1 x 10-15
0.513 0.1815 0.3538 0.0931 0.6897 r.3444 r\c
aA(oH)3
0.363 0.1108 0.3052 0.0402 0.8409 2.3164
0.256 0.0448 0.1750 0.0115 0.6836 2.6703 where the activity of AI(OH) I solid is taken as 1.0. Since,by
0.181 0.0147 0.0812 0.0027 0.4487 2.4790 stoichiometry,
0.128 0.0025 0.0195 0.0003 0.1526 1.t921
0.098 0.0011 0.otl2 0.0001 0.1145 1.1687 (con-) : 3(cer,*) and K,: ( c ^ r * ) 4 ( 3 ) 3: 1 . 1 x 1 0 - 1 5
1.0000 r.7695 0.6658 4.0032 t2.5909 then,
From(17-7),
1
A(oH)3: 8 x 10-5 gmoles/L
(cett*) : cdissolved
D "s - _ : 0.565mm
1.7695 which is a very small concentration.
654 Chapter 17 Crystallization

Table 17.5 Solubility and Heat of Solution at Infinite Dilution of Some Inorganic Compounds in Water
(A Positive Heat of Solution Is Endothermic)

Heat of
Solution of
Stable
Hydrate
(at Room Stable
Temperature) Hydrate
Solubility (Hydrate-freeBasis) 9/100 gH2O atT,"C
kcaVmole at Room
Compound Compound 10 20 30 40 60 80 100 Temperature
NH4CI +3.8 29.7 33.4 5 t.z 41.4 45.8 55.2 65.6 77.3 0
(NH4)2S04 +t 1 7t.o 73.0 75.4 78.0 81.0 88.0 95.3 103.3 0
BaCl2 -r4 5 3r.6 35.7 38.2 40;7 46.4 52.4
3.J.2 58.3 2
CuSO+ +2.86 14.3 l7.4 20.7 25.0 28.5 40.0 55.0 75.4 5
MgSOa +3.18 22.3 27.8 33.5 39.6 44.8 55.3 56.0 50.0 7
MgC12 -3.1 52.8 53.5 54.5 56.0 57.5 61.0 66.0 73.0 6
Nis04 +4.2 26 32 JI 43 47 55 63 7
KCI +4.4 27.6 31.0 34.0 J /.U 40.0 45.5 51.1 56.7 0
KNO3 +8.6 13.3 20.9 3r.6 45.8 63.9 110 169 247 0
K2SOa +6.3 7.4 9.3 11.1 1 3 l. 14.9 18.3 21.4 )4) 0
AgNO3 122 t70 222 300 376 525 669 952 0
NaClO3 +5.4 80 89 101 113 t26 r55 189 233 0
NaCl +0.93 35.6 35.7 35.8 36.1 36.4 3'7.1 38.1 39.8 0
NaNO3 +5.0 72 78 85 92 98 I JJ 163 0
Na2SOa +18.7 4.8 9.0 19.4 40.8 48.8 45.3 43.7 42.5 10
NazSzO: +11.4 52 6l 70 84 103 207 250 266 5
Na3POa +15.0 1.5 4 1l 20 JI 55 81 108 12

Table 17.6 Concentration Solubility Products Although the solubility of most inorganic compounds
of Some Sparingly Soluble Inorganic Compounds increaseswith increasing temperature, a few common com-
pounds exhibit a so-called negative or inverted solubility, in
Compound T,"C Kc
certain ranges of temperature, where solubility decreases
Ag2CO3 25 6 . 1 5x l0-r2 with increasing temperature. These compounds are the
AgCl 25 1 . 5 6x 10-ro so-called hard salts, which include anhydrous Na2SOaand
A(OH)3 l5 4x 10-13 CaSO+.
A(OH)3 18 1 . 1x 10-15 A considerable change in the solubility curve can occur
BaSOa 18 0.87x 10-10 when a phase transition from one stable hydrate to another
CaCO3 l5 0.99x 10-8 takesplace.For example,in Table 17.5,NazSO+. 10H2Ois
CaSOa 10 1.95x lO-a the stable form from 0'C to about32.4C. In that tempera-
CuSOa r6-18 2x lo-aj ture range, the solubility increasesrapidly from 4.8 to 49.5 g
Fe(oH): t8 1 . 1x 10-36 (hydrate-freebasis)/100g HzO. From32.4C to 100"C,the
MgCO3 t2 2.6 x l}-s stable form is Na2SOa, whose solubility decreasesslowly
ZnS 18 1.2x 10-23 from 49.5 to 42.5 9/100 g H2O. In the phase diagram of
Figure 17.2 for the MgSOa-water system, the solubility-
For less sparingly soluble compounds, the equilibrium temperaturecurves of each of the four hydrated forms hasa
constant, called Ko, is more rigorously expressedin terms of distinctive slope.
ionic activities or activity coefficients: The solubility characteristicof a solute in a particular sol-
(aap*)(ass-' )3 vent is, by far, the most important property for determining:
Ko: : (^yer,*
)(cerr*)(^ysH-)3(csH-)3 (1) the best method for causing crysrallization, and (2) the
4Al(oH)3
ease or difficulty in growing crystals. Crystallization by
In general,^y ! 1.0 for c < 1x 10-3gmoles/L.As c rises cooling is only attractive for compounds having a solubility
above 1 x 10-3 gmoles/L, 1 decreases,but may pass that decreasesrapidly with decreasing temperature above
through a minimum and then increase. Mullin [1] presents ambient temperature. Such is not the case for most of the
activity-coefficient data at25"C for soluble inorganic com- compounds in Table 17.5. For NaCl, crystallization by cool-
pounds over a wide range of concentration. ing would be undesirable because the solubility decreases
 I7.2 ThermodynamicConsiderations 655

only by aboil 107o when the temperature decreasesfrom balance,

100 to 0'C. For most soluble inorganic compounds, cooling 4,446:0.28L+0.4886S (1)
by evaporation is the preferred technique. where
Solid compounds with a very low solubility can be pro-
L : lb/h of liquid
duced by reacting two soluble compounds. For example, in
Table 17.6, solid AI(OH)3 can be formed by the reaction S: lb/h ofcrystals
Also,
AlCl3luq;* 3NaOHlnq;() Al(OH)3(oo9* 3NaC\aq;
s : 0.208(s
+ r) Q)
However, the reaction is so fast that only very fine crystals,
Solvingt I I and(2) simultaneously,
called a precipitate, are produced, with no simple method to
causethem to grow to large crystals. S:2,856lb/h
L : r0,876lblh
EXAMPLE 17.3 By a total materialbalancearoundthe crystallizer,
F:V-TL-IS
The concentrate from an evaporationsystemis 4,466lblh of 37.75
wtToMgSOaat 170"Fand20 psia.It is mixedwith 9,860lb/h of a whereF : total feedrateand V: evaporationrate.Therefore,
saturated aqueous recyclefiltrateof MgSO+at 85"F and20 psiaand
14,326:V +10,876+2,856 (3)
sentto a vacuumcrystallizer,operatingat 85"F and0.58psiain the
vapor space,to producewater vapor and a magmaof 20.8 wt%o Solving,
crystalsafi79.2 wtTosaturatedsolution.The magmais sentto a V :5941b/h
filter, from which filtrate is recycledas mentionedabove.Deter-
The resultsin tabularform are:
mine the lb/h of water evaporatedand the maximumproduction
rateof crystalsin tons/day(dry basisfor 2000lb/ton). lb/h for crystallizer

Component Feed Vapor Liquid Crystals

SOLUTION
MgSO+,,, 4,446 0 3,045 0
For the saturatedfiltrate at 85'F, the weight fraction of MgSOa,
MgSOa.7H2O1e 0 0 0 2,856
from Figure 17.2,is 28 wt%. Therefore,MgSOain the recyclefil-
HzO 9,880 594 7,831 0
trateis 9,860(0.28): 2,760lblh. By materialbalancearoundthe
mixingstep, 14,326 594 10,876 2,856

lb/h The maximumproductionrateof crystalsis

? R56
-'"""
Component Feed RecycleFiltrate Crystallizer Feed (24\: 34.3tons/dav
2,000
MgSOa 1,686 2,760 4,446
HzO 2,780 7,100 9,880 A large number of organic compounds, particularly or-
4,466 9,860 14,326 ganic acids with relatively moderate melting points
(125-225"C), are also soluble in water. Some data are given
The material balance around the crystallizer is conveniently made
in Table 11.7, where it is seen that the solubility often in-
by a balance on MgSOa. At 85'R from Figure 17.2, the magma is
20.8 w t%oMgS Oa . 7H2O crystals and 79.2 wtvo of 28 w tVoaqueous creasessignificantly with increasing temperature.For exam-
MgSOa liquid. Becausethe MW of MgSOa and MgSOa.TH2O ple, the solubility of o-phthalic acid increasesfrom a very
arc 120.4 and 246.4, respectively, the crystals are 120.4/ low value of 0.56 to 18.09/100 g H2O when the temperature
246.4:0.4886 mass fraction MgSOa. Therefore, by a MgSOa increasesfrom 20 to 100'C.

Table 17.7 Solubility and Melting Point of Some Organic Compounds in Water

Solubility,9/100g HzO atT,"C

Melting
Compound Point,'C 10 20 30 40 60 80 100

Adipic acid 153 0.8 1.0 1.9 3.0 5.0 18 70 160

Benzoic acid 122 0.17 0.20 0.29 0.40 0.56 1.16 2.72 5.88
Fumaric acid (trans) 287 0.23 0.35 0.50 0.72 1.1 2.5 5.2 9.8
Maleic acid 130 39.3 50 70 90 115 r78 283
Oxalic acid 189 3.5 6.0 9.5 14.5 2r.6 44.3 84.4
o-phthalic acid 208 0.23 0.36 0.56 0.8 t.2 2.8 6.3 18.0
Succinic acid 183 2.8 4.4 6.9 10.5 16.2 35.8 70.8 r27
Sucrose d t79 r90 204 219 238 287 362 487
Urea I JJ 6'7 85 105 135 165 250 400 730
Uric acid d 0.002 0.004 0.006 0.009 0.012 0.023 0.039 0.062
656 Chapter 17 Crystallization

EXAMPLE 1?.4 I
MgSOa
-20,000
Oxalic acid is to be crystallized from a saturated aqueous solution E
initially at 100'C. To what temperaturedoes the solution have to be o Z n ii o a
-1 5.000 l,,so
cooled to crystallize 95Voof the acid as the dihydrate?
MnSOoi
-10,000
;
SOLUTION a

-5,000
Assume a basis of 100 g of water. From Table 77.7, the amount of NarSOo
o
c
dissolvedoxalic acid at 100'C is 84.4 g. 0
Amount to be crystallized: 0.95(84.4) :80.2 g. LiruHo)rsoo
o
Amount of oxalic acid left in solution :84.4 - 80.2 - 4.2 s.. c +5,000
KrS6o-i
MW of oxalic acid : 90.0.
MW of water: 18.0. 020406080
^ 2(18.0) g H:O Vlolsof water/mol of salt
Water of hydration for2H2O : : O.On"** .
,* -- r-* Figure 17.9 lntegral heats of solution for sulfates in water
Theretbre water of crystallization : 0.4(80.2) 32.I gH2O. at25"C.

Liquid water remaining : 100 - 32.1 : 67.9 e.

4.2 g
. mustbe -6 7 . 9x 1 0 0: 6 . 1 91 0 0 g H 2 O
Finalsolubilitv
From Table 17.7, by linear interpolation, temperature : 10.6'C.
Enthalpy Balances
When an anhydroussolid compound, whose solubility in-
creaseswith increasingtemperature,dissolvesisothermally
in water or some other solvent, heat must be absorbedby the
solution. This amount of heat per mole of compound in an
infinite amount of solvent varies with temperature and is
referredto as the heat of solution at infinite dilution (AHffi).
For example,in Table 17.5,the solubility of anhydrousNaCl
is seento increase slowly with increasing temperaturefrom
10 to 100'C. Correspondingly,the heatof solution at infinite
dilution in Table 17.5 is modestly endothermic (*) at room
temperature. In contrast, the solubility of anhydrous KNO3
increasesmore rapidly with increasing temperature, result-
ing in a higher endothermic heat of solution. For compounds
that form hydrates, the heat of solution at infinite dilution
may be negative (exothermic) for the anhydrous form, but
becomesless negative and often positive as higher hydrates
are formed by the reaction:
A.nH2O19 -+ A(aq)I nH2O
For example, the following heats of solution at infinite dilu-
tion in kJ/mol of compound at 18"C for four hydrates of
MgSOa clearly show this effect:
MgSOa -88.3
-58.6
MgSOo.H2O
M g S O o. 6 H z O -2.3
M g S O o. l H z O +13.3
Heats of a solution for a number of hydrated and anhydrous
compoundsare listed in Table 17.5.
As a solid compound continues to dissolve in a solvent,
the heat of solution, which is now referred to as the integral
heat of solution, varies somewhat,as shown in Figure 17.9
for several compounds as a function of concentration.The
integral heat of solution at saturation is numerically equal,
but opposite in sign, to the heat of crystallization. The
difference between the integral heat of solution at saturation
and the heat of solution at infinite dilution is the heatof
dilution:
AH,Ti-411ff:AF1oil-
with
aHffl : -AH.,y,
As indicated in Figure l1 .9, the heat of dilution is relatively
small: therefore. it is common to use:
A11"rv.r -AH'T
An energy-balance calculation around a crystallizer is
complex becauseit can involve not only the integral heatof
solution and/or heat of crystallization, but also the specific
heats of the solute and solvent and the heat of vaporization
of the solvent. The calculation is readily made if an
enthalpy-mass fraction diagram is available for the system,
including solubility and phase-equilibria data. Mullin []
lists 11 aqueous binary systems for which such a diagram
has been constructed.A diagram for the MgSO*-H2O system
is shown in Figure 17.10.The enthalpydatum is pure liquid
water at 32'F (consistent with steam tables in American
Engineering Units) at Pointp and solid MgSOa at 32"F (not
s h o w ni n F i g u r e1 7 . 1 0 ) .
Points a to l, n, p, and q in the enthalpy-mass fraction
diagram of Figure 17.10 correspondto the samepoints in the
phasediagramof Figure 17.2.In Figure 17.10,the isotherms
in the region above Curve pabcdq pertain to enthalpies of
unsaturated solutions of MgSOa. The straightnessof these
isotherms indicates that the heat of dilution is almost negli-
gible. In this solid-free region, a 30 wtVo aqueous solution
of MgSOa has a specific enthalpy at 110"F of -31 Btu/lb
solution.
 17.2 ThermodynamicConsiderations657

140 from solubility curve bc to MgSOa .7HzO solid line ih.The

120 relative amounts of the two equilibrium phasescan be com-
puted from a MgSOa balance.For a basis of 100 lb of mix-
100 ture, let S be the pounds of crystals and A be the pounds of
saturatedaqueoussolution. Thus, the MgSOa balance is
0 . 3 0 ( 1 0 0 ): 0 . 4 9 s + 0 . 2 6A
where
100:S*A
Solving, S : I7.4lb and A :82.6Ib. The enthalpyof the
OU
{= mixture at 70'F is -65 Btu/lb, which is equivalent to
f

E -2o enthalpiesof 46 and -155 Btu/lb, respectively,for the solu-

-40
tion and crystals.
;

> -ou
o
-80
EXAMPTE17.5
E
U For the conditions of Example 17.3, calculate the Btu/h of heat
-1 00
addition for the crystallizer.
-120

-1 40 SOLUTION
-1 60 An overall energy balance around the crystallizer gives
-180 IrL feedH f eed I Q in : m, uporHvapor + mtiquia F/tiquio * fl?crystatsllcrystals
(1)
-200
where liquid refers to the saturated-liquid portion of the magma.
-zzu
From the solution to Examole 17.3. the feed consistsof two
0 0.050.100.150.200.250.300.350.400.450.50 sffeams:
Concentration-massf raction,M gSO,
mreedt: 4,466lblhr of 37.'75wt%o MgSOo at 170'F
Figure 17.10 Enthalpy-concentration diagram for the
: 9,860 lb/hr of 28.0 wtVo MgSOa at 85"F
mreed2
MgSOa-H2O systemat I atm.
From Figure 17.10,
l4eedt : -20Btu/lb
In the region below the solubility curve pabcdq, in both
Hfeed2: -43 Btu/lb
Figures 17.2 and 17.10, the following phases exist at equi-
librium: Therefore,
trlfeedHfeed: 4,466(-20) + 9,860(-43) : -513,000 Btulh
Temperature
Region Range, oF Phases : 594 lblh at 85'F and 0.58 psia
ffiyapor
pae 25-32 ice and aqueoussolution of MgSOa The vapor enthalpy doesnot appearon Figure 17.10,but enthalpy
ea 25 ice and eutectic mixture tables for steamcan be used since they are basedon the samedatum
ag 25 eutectic and MgSOo . l2HzO (i.e., liquid water at 32'F).
abfs 25-37.5 saturatedsolution and Therefore, Huupor: 1099 Btu/lb from steam tables and
MgSOa .12H2O
muupo,H,apo,:594(1099): 653'000 Btu/h
bcih 37.5-118.8 saturatedsolution and
MgSOa .7H2O The liquid plus crystals can be treated together as the magma. From
cdlj 118.8-154.4 saturatedsolution and the solution to Example 17.3,
MgSOa'6H2O : ffimagma:10,876 +2,856
mtrqtn]- ncrystals
dqrk 154.4- saturatedsolution and :13,732l'b|hof 32.4 wtToMgSOa ar 85"F
MgSOa.H2O
From Figure 17.10,
Pure ice exists at Point e, where in Figure 17.10 the spe- /I-un.u : -67 Btu4b
cific enthalpy is -I4l Btu/lb, which is the heat of crystal-
and
lization of water at 32"F. If a 30 wtEo aqueons solution of
MgSOa is cooled from 110"F to 70'F, the equilibrium ffi^up uH^us u: 13'732(-6'7) : -920,000 Btu/lb
magma will consist of a saturated solution of 26 wt%o From (1),
MgSO+ and crystals of MgSO. .lHzO (49 wtVo MgSOa) as
Oin :653,000 - 920,000- (-513,000) :246,000Btu/h
determined from the ends of the 70"F tie line that extends
658 Chapter17 Crystallization

In the absence of an enthalpy-mass fraction diagram, a lattice structure. Collectively, these phenomenaare referred
reasonably accurate energy balance can be made if data for to as crystallization kinetics. Experimental data show that
heat of crystallization and specific heats of the solutions are the driving force for all three stepsis supersaturation.
available or can be estimated and the heat of dilution is
neglected,as shown in the next example.
Supersaturation
The solubility property discussed in the previous section
EXAMPLE 17:6 refers to relatively large crystals of the size that can be seen
by the naked eye, i.e., larger than 20 pm in diameter.As
The feedto a coolingcrystallizeris 1,000lb/h of 32.5wtTaMgSOa
in waterat 120'F.This solutionis cooledto 70'F to form crystals crystal size decreasesbelow this diameter, solubility notice-
of the heptahydrate.
Estimatethe heatremovalratein Btu/h. ably increases,making it possible to supersaturatea solution
if it is cooled slowly without agitation. This phenomenon,
based on the work of Miers and Isaac l4l in 1907, is repre-
SOLATION sentedin Figure 17.11,where the normal solubility curve,cr
Material balance is representedas a solid line. The solubility of very small
crystals can fall in the metastableregion which is shownto
From Figure 17.2,the feed at 120'F containsno crystals,but the
have a metastable limiting solubility, c., given by the
magmaat 70F consistsof crystalsof the heptahydrateand a
motherliquor of 26 wtVaMgSOa.By materialbalancein the man- dashedline.
ner of Examplel7 .3, thefollowing resultsareobtained: Consider a solution at a temperature,21, given by the ver-
tical line in Figure l7.ll.If the concentrationis given by
Point a, the solution is undersaturated and crystals of all
lb/hr
sizesdissolve.At Point b, equilibrium existsbetweena satu-
Feed MotherLiquor Crystals rated solution and crystals that can be seenby the nakedeye.
In the metastableregion at Point c, crystals can grow but can-
HzO 675 530 0
not nucleate. If no crystals are present, none can form. For
MgSOa 325 186 0
Mg2SOa.7H2O that concentration,the difference between the temperatureat
0 0 284
'716 Point e on the solubility curve and Point c in the metastable
1,000 284 region is the supersaturationtemperature difference, which
may be about 2'F. The supersaturation,A,c : c - c", is the
Take a thermodynamic path consisting of cooling the feed from
difference in concentration between Points c and b.
120'F to 70'F followed by crystallization at 70"F. From Hougen,
Watson, and Ragatz [3], the specific heat of the feed is approxi At Point d, spontaneousnucleation of very small crystals,
mately constant over the temperature range at 0.72 Btunb-.F. invisible to the naked eye, occurs. The difference in temper-
Therefore, the heat that must be removed to cool the feed to 70.F is ature between Points/and d is the limit of the supersatura-
tion temperature difference. The limiting supersaturationis
r,000(0.72)(120 - 70) : 36,000 Btuftr A c 1 1 6 1: C m - c r .

For data presented earlier in this section, the heat of crystalliza- The relationship between solubility and crystal size is
tion can be taken as the negative of the heat of solution at infinite given quantitatively by the Kelvin equation (also known as
dilution:
- 13.3 kJ/mole of heptahydrare

or -23.28tu/lb of heptahydrate
Metastable limit, c.
Therefore, the heat that must be removed during crystallization of
the heptahydrateis
o
284(23.2): 6,600Btu/h ; S o l u b i l i t y c L l r v ec, "
The total heat removal is 36,000 + 6,600 : 42,600 Btu/h.
If this example is solved with Figure 17.10, in the manner of ."1-------r f
Example 17.5,the result is 44,90OBtu/h,which is 5.4VohisheL o I
cl---
o

Unsaturated
region
O
I7.3 KINETIC AND TRANSPORT
CONSIDERATIONS
Crystallization is a complex phenomenonthat involves three T1
steps'.nucleation, mass transfer of the solute to the crystal Temperature
surface, and incorporation of the solute into the crystal Figure 17.11 Representativesolubility-supersolubility diagram.
 17.3 Kinetic andTransportConsiderations 659

the Gibbs*Thompson and Ostwald equations): Dp, Fm c/c, c, g KCI/100 g H2O

-(;):(ffi) (r7-16)
0.01
0.10
1.00
1.0887
1.0085
1.00085
38.65
35.80
3s.s3
10.00 1.000085 35.50
where 100.00 r.0000085 35.50
us : rflolar volume of the crystals
trs,t : interfacial tension Nucleation
u : number of ions/molecule of solute
To determine the volume or residencetime of the magma in
c/c, - supersaturationratio : ,S a crystallizer, the rate of nucleation (birth) of crystals and
their rate of growth must be known or estimated. The rela-
Measured values of interfacial tension (also called surface
tive rates of nucleation and growth are very important
energy) range from as low as 0.001 J/m2 for very soluble
because they determine crystal size and size distribution.
compounds to 0.170 for compounds of low solubility.
Nucleation may be primary or secondary depending on
As might be expected, (l'7-16), in a more general form,
applies to the effect of droplet diameter on vapor pressure whether the supersaturatedsolution is free of crystalline sur-
faces or contains crystals, respectively. Primary nucleation
and solubility in another liquid phase.
requires a high degreeof supersaturationand is the principal
It is common to define a relative supersaturation,s, by
mechanism occurring in precipitation. The theory of primary
nucleation is well developed and applies as well to the con-
s-c-c":'-1 :s-r (n-ri)
cs cs densation of liquid droplets from a supersaturatedvapor and
the formation of droplets of a second liquid phase from an
In practice,s is usually less than 0.02 or 2Vo.For such small initial liquid phase. However, secondary nucleation is the
values, ln(c lc,) can be approximated by s with no more than principal mechanism in commercial crystallizers, where
a IEo error. crystalline surfaces are present and large crystals are
desirable.

EXAMPLE 17.7

Determinetheeffectof crystaldiameteron the solubilityof KCI in Primary Nucleation

waterat 250C. Primary nucleation can be homogeneous or heterogeneous.
The former occurs within the supersaturatedsolution in the
SOLUTION absenceof any foreign matter, such as dust. Molecules in the
solution first associateto form a cluster, which may disso-
FromTable17.5,by interpolation, cs : 35.5g/100g H2O.
ciate or grow. If a cluster gets large enough to take on the
BecauseKCI dissociates into K+ autdCl- , u :2.
appeamnceof a lattice structure, it becomesan embryo. Flur-
MW of KCI :74.6.
ther growth can result in a stable crystalline nucleus whose
Densityof KCI crystals: 1980kgim3. size exceedsthat given by (17-16) for the prevailing degree
of supersaturation.
74.6
u, : : 0.0376m'/kmol The rate of homogeneousnucleation is given by classical
I 9g0-
chemical kinetics in conjunction with (17-16), as discussed
T :298K
by Nielsen [5]. The resulting expressionis
R :83l4Jlkmol-K

FromMullin [], page200, -16ru!o!,rN"

Bo:Aexp (17-18)
o''t :0'O28Jlm2 3u2(Rr)3[',(;)]'
F r o m( 1 7 - 1 6 ) ,
where
/ 4u"o"
" 'r- \ I Bo : rate of homogeneousprimary nucleation, number
c : c'. e x D|
\uRTDr/ of nuclei/cm3-s
- r).) expI 4(0.0376X0.028 for
L2(831s)(298)Dp il Do in m, or (1) A : frequency factor
No : Avogadro'snumber:6.022 x 1026molecules/
: 35.5exp(0.00085/Do,
pm) kmol
660 Chapter17 Crystallization
Theoretically, A :1030 nuclei./cm3-s;however, observed
values are generally different due to the unavoidable
presenceof foreign matter. Thus, ( 17-18) can also be applied
to heterogeneousprimary nucleation, where A is determined
experimentally and may be many orders of magnitude
different from the theoretical value. A value of 1025is often
quoted.
The rate of primary nucleation is extremely sensitive to
the supersaturationratio, S, defined by (11-I7), as illustrated
in the following example.
EXAMPLN 17.8
Using the data in Example 17.7, estimatethe effectsof relative
supersaturationon the primary homogeneous nucleationof KCI
from an aqueoussolutionat25"C. Usevaluesof s corresponding to
valuesof c/c" of 2.0,1.5,and1.1.
SOLUTION
For cf c,:2.0, ln(c/c"):0.693. From (17-18), using data in
Exarryle 17.7,
- r6(3.14X0.0zt af g.ozty3(6.022x :o2\ 1
," : tO'o
"*nI
I uration. Nernst [7] proposedthe existenceof a thin stagnant
:2.23 x l02s nuclei/cm3-s
Calculations for the other values of cf c, ue obtained in the same
manner with the following results:
c /ct Bo, nuclei/cm3-s
'rn
2.23 x 1025
1.5 2.60 x 1016
1.1 0
Since large values of the supersaturation ratio (clc, > 1.02) are
essentially impossible for crystallization of solutes of moderate to
high solubility (e.g., solutes listed in Tables 17.5 and 17.7), primary
nucleation for these solutes never occurs. Howeveq for relatively
insoluble solutes(e.g.,soluteslisted in Table 17.6),largevaluesof
cfc, can be generatedrapidly from ionic reactions causing rapid
precipitation of very fine particles. If A: 1025 is used, Bo is
, ^s
o1v10eo Dv IU".
Secondary Nucleation
Nucleation in industrial crystallizers occurs mainly by sec-
ondary nucleation caused by the presenceof existing crys-
tals in the supersaturatedsolution. Secondarynucleation can
occur by: (1) fluid shear past crystal surfacesthat sweeps
away nuclei, (2) collisions of crystalswith each otheq and
(3) collisions of crystals with metal surfaces such as the
crystallizer vessel wall or agitator blades. The latter two
mechanisms,which are referred to as contact nucleation are
the most common types since they can occur at the low val-
ues of relative supersaturationthat are typically encountered
in industrial applications.
In the absenceof a theory for the complex phenomenaof
secondary nucleation, the following empirical power-law
function, which correlatesmuch of the experimental data.is
widely used:
Bo : kusb utrW' (r1-1e)
whete M7: mass of crystals per volume of magma and
N : agitation rate (e.g., rpm of an impeller). The constants,
ku, b, j, and r, are determined from experimental data, on the
system of interest, as discussedbelow in the section on a
crystallizermodel.
Crystal Growth
In 1897, Noyes and Whitney [6] presenteda mass-transfer
theory of crystal growth based on equilibrium at the crystal-
solution interface. Thus, they wrote
dmldt:k,A(c-c,) (r7-20)
where dmf dt : rate of mass depositedon the crystal surface,
A : surfaceareaof the crystal, ft, : psss-transfer coefficient,
c : masssolute concentrationin the bulk supersaturatedsolu-
tion, and c": masssoluteconcentrationin the solutionat sa1-
film of solution adjacent to the crystal face through which
molecular diffusion of the solute took place. Thus, k, : 276,
where 2 : diffusivity and 6 : film thickness, where the lat-
ter was assumedto dependon velocity of the solution pastthe
crystal as determinedby the degreeof agitation.
The theory of Noyes and Whitney was challenged by
Miers [8], who showed experimentally that an aqueoussolu-
tion in contact with crystals of sodium chlorate was not sat-
urated at the crystal-solution interface, but was supersatu-
rated. This finding led to a two-step theory of crystal growth,
referred to as the diffusion-reaction theory, as describedby
Valeton [9]. Mass transfer of solute from the bulk of the so-
lution to the crystal-solution interface occurs in the first
step,as given by a modification of (17-20):
dm/dt:k,A(c-ci) (11-2r)
where c; is the supersaturatedconcentration at the interface.
In the second step, a first-order reaction is assumedto occur
at the crystal-solution interface, in which solute molecules
are integrated into the crystal-lattice structure.Thus, for this
kinetic step,
dmldt:kiA(ci-c,) (r7-22)
lf (11-21) and (17-22) are combined to eliminare c,, we
obtain
dm/dt: .l!,' . 9.',', (r7-23a)
llk,+t/tci
Typically, ft. will depend on the velocity of the solution as
shown in Figure I7.12. At low velocities, the growth rate
 17.3 Kinetic andTransportConsiderations 661

spherical crystal. Rewriting (11-23a) in terms of an overall

coefficient,

Controlled by kinetics dmldt:K,A(c-c,) (r7-23b)

c of integration into
3 crysta l-lattice
structu re Since 4: nD| and * :*o
=
o Equation (17 -23b) becomes
@

- c')
dDp
:2K,(c _ 2K,(A'c) 0j_24)
o
dtpp

If the rate of growth is controlled by ft;, which is assumedto

be independentof Do, then

o D o - 2k;Lc
Solutionvelocityrelativeto crystal
(r'7-2s)
Lt
Figure 17.12 Effectof solutionvelocitypastcrystalon the rateof
crystalgrowth. and the rate of increase of crystal size is linear in time for a
constant supersaturation.If the rate of growth is controlled
by k, at a low velocity, then, from (15-60),
may be controlled by the first step. The second step can be
important, especially when the solution velocity past the K, : k, :2D/DP (r7-26)
crystal surfaceis high, such that k; is large compared to ft;. In
where D is solute diffusivity.
adsorption, the kinetic step is rarely important. It is also
unimportant in dissolution, the reverse of crystallization. Substitutionof (17-26) into (17-24) gives
The mass-transfer coefficient, k., for the first step, is
independent of the crystallization process and can be dDp
:4D(Lc) (r'7-27)
estimated from general fluid-solid particle mass-transfer- dt Dpp
coefficient correlations described in Chapters 3 and 15. The
Integration from Do" to Do gives
kinetic coefficient, k;, is peculiar to the crystallization
process.A number of theories have been advanced for the Di - Dl 4D(L,c\
kinetic step, as discussedin Myerson [10]. One prominent (r7-28)
2:p
theory is that of Burton, Cabrera, and Frank [11], which is
basedon a growth spiral starting from a screw dislocation, as If Dp. <<Dp, (l'7-28)reduces
to
shown in Figure 17.13 and verified in some experimental
studiesusing scanning-electronmicroscopy. A dislocation is Do: (8D(e^dt)t/' (r1-29)
an imperfection in the crystal structure. The screw-disloca-
tion theory predicts a growth rate proportional to: 1c; - cr)2
In this case,the increasein crystal diameter slows with time.
at low supersaturationand to (c; - c") at high supersatura-
At higher solution velocities where k still controls, the
tion. Unfortunately, the theory does not provide a means to
use of (15-62) resultsin
predict fr;.Accordingly, (17 -23a) is generally applied with k
estimated from available correlations and ft; back-calculated Kr:kr:CtlD)/2 (17-30)
from experimentaldata.
Although crystals do not grow as spheres,let us develop For this case,the increasein crystal diameter also slows with
an equation for the rate of increase of the diameter of a time, but not as rapidly as predicted by (17-29). It is

Screw Growth Resulting

dislocation spiral crystal Figure 17.13 Screw-dislocation mechanism of
growth crystal growth.
662 Chapter 17 Crystallization

common to assume that the rate of crystal growth is con- Initial charge:
trolled by ft; and, thus, is not dependent on crystal size and is, 900 lb aq MgSOa
therefore, invariant with time. This assumption will be uti- 2,100lb H2o
lized in a latter section of this chapter to develop a useful 3,000Ib
crystallizer model. *2 lb MgSOo. 7H2Ocrystalseeds
(a) Crystalsgrow from 50 pm to 400 pm in diameter:
EXAMPLE 17.9 /400 \ 3
y' i e l d : 2 l = | : r , O 2 4 l b M g S O o . T H 2 O
The heptahydrateof MgSOa is to be crystallized batchwise from a \50/
seeded aqueous solution. A low supersaturation is to be used to (b) Numberof crystals,basedon the crystalseeds,
avoid primary nucleation and a mild agitation is to be applied to
mass of seeds
minimize secondarynucleation. The temperature in the crystallizer
mass/seed
will be maintained at 35'C, at which the solubility of MgSOa in
water is 30 wt%o.The crystallizer will be charged with 3,000 lb of a
2(4s4)
saturatedsolution at that temperature.To this solution will be added
/ ?- ' -r 4' \
1 . 6 8 1 l ( s 0x t 0 - + ) :
2 lb of heptahydrate seed crystals of 50 pm in diameter. A super- \ol
saturation of 0.01 gm of heptahydrate per gm of solution at 35'C
will be maintained during crystallization by operating the crystal- :8.26 x l0e, where : ppVp: ,+
mass/seed
lizer at vacuum and using heat exchange and the heat of crystal-
lization to gradually evaporate water from the solution. Based on (c.) Pounds of heptahydratecrystallized : 1,024 - 2 : l,022lb
the assumptions and data listed below, determine the following if
the final crystal size is to be 400 pm. H2o or hydration: r,022 ('!#) : szl u
(a) Yield of crystals of heptahydratein pounds. MgSOain crystals: 1,022- 523: 499\b
(b) Number of crystals. MgSOaleft in solution: 900 - 499: 401lb
(c) Amount of water in pounds that will have to be evaporated.
H z o l e f ti n s o l u t i o n :* t (rA) : e36lb
(d) Product magma density in pounds of crystals per pound of
solution. : 2,100 - 523 - 936 : 641lb
H2O evaporated
(e) Crystallizer volume in gallons if the volume occupied by the
(d) Finalmotherliquor: 936+ 401: 1,337lb
magma during operation is at the most 50Voof the crystallizer
volume.
to:,o':1'
(f) Crystallizer pressurein psia and the boiling-point elevation in "F.
: f :-)
Magmadensiry :0.766(,, )
\ 1.337
/ \ lb mother liquor/
(g) Time in minutes to grow the crystals to the final size.
(e) Initially, usingthe factorof 8.33lb/gal for 1.0g/cm3,
(h) Amount of heat transfer in Btu.
3'ooo
(i) Also, determine whether crystal growth is controlled by mass Solutionvolume- = g33( :269 Eal
134)
transfer, by surface reaction (incorporation into the lattice), or Finally,
by both.
1'337
Solutionvolume- ' : l2o gal
Assumptions and Data: 83H134)
1. No primary or secondarynucleation.
crystalvotume: :73 sal
2. Properties of aqueous30 wt%oMgSOa at 35'C: *ffi.,
Total volume : t93 gal
densitY: 1.34 glcm3
viscosity : 8 cP Therefore, initial conditions before crystallization will control
the volume.
diffusivityof MgSOo: l.l0 x 10-s cm2ls
Make crystallizervolume :'# :538 gal
3. . Density of the crystals: 1.68 g/cm3.
4. Crystal shapecan be approximated as a sphere. (f) Calculate H2O partial pressure by Raoult's law applied to
mother liquor:
5. Average crystal-face growth rate, including effects of both mass
transfer and surface reaction : 0.005 mm/min. l b m o l M g S O a: L : 3 . 3 3
6. Solution velocity past crystal face : 5 cm,/s.
H2o:
lbmol
#h:51.e5
SOLATION Total - 55.28 lbmol
Molecular weights: MgSOa 120.5
xH2o: : o.no
MgSOa .7H2O 246.6 ffi
Hzo 18.02 At 35'C :95"F, Piro : 0.8153psia.
 17.4 Equipment for Solution Crystallization 663

400
Therefore,Pszo: rHzoPizo : 0.94(0.8153): 0.766psia' Forr: :200 p.m,
2
This correspondsto a saturation temperature of 93'F.
,oP) : r.ro
Therefore, boiling-point elevation : 2oF. Nn.:0.42[ .
(g) Growth rate : 0.005 mm/min. "i0'l
Nsn: 2 + 0.6(33qt t26,430)t /3
Therefore, diameter grows at 0.01 mm/min. : 2 - t 1 9 . 3: 2 1 . 3
Must grow from 50 pm : 0.05-mm diameter to 0.40-mm ( 2 1 ' 3 )l(' I x l 0 - 5 )
diameter. t- - : 0 . 0 0 6c m / s
2(200x l0 a)
0.40 - 0.05
Time: : J5 mtn Thus,ft" changesby a factorof 3 as crystalsizechangesfrom
001
25 to2OOp.m.
(h) Enthalpy balance:
Assumea Ac basedon the given supersaturation of 0.01g
FHr-lQ:VHv+MHM (1) g
crystalper solution.

Assume charge is at 35'C :95'F; then, from Figure 17.10' o'ollt'34)

:54.3
- Jr'J x 10-6lq
Lc : '"
Hp : -40Btlllb. 24e6
"tr,
From the steam tables, Hv :1,102.2 * 0.9 (for the boiling- : x lo-'"*
pMofcrystals
point elevation): 1,103.1Btu/lb. ffi:6.81
( 4 9 9 + 4 0 1+ l ) 901
WtToMgSO4in magma: From (4).
( 3 , 0 0+
02-64r) 2,361
dr - x lo-6-:
: 0.38. ' 6 . 8 1x l 0 - r o.oogk"
n^54'3
dr
From Figure 17.10,Hu : -90 Btuilb. for the lowest value of k : 0'006 cm/s'
From (1),
- 3'002(-40)
Q : 64r(r,103.r) + 2,36r(-90) t : O.OOS(0.006)
:48 x 10-6 cm/s
: 615,000 Btu : heat transferred to crystallizer dt
(i) Assume mass-transfercontrolled, using molar amount of crys- or
tals, n, and molar concentrations,c. 480 x 10_6(60): 0.029 mm./min.
The molar form of ( l7-2 I ) is
But this is six times faster than the growth rate. Therefore, crystal
dn
: :4rrr2kr(L,c), (2) growth is largely controlled by surface reaction.
dt
where r is the crystal radius
A
n : -rrr Pu,using amolar densitY.
J
17.4 EQUTPMENTFOR SOLUTTON
Therefore, CRYSTALLIZATION
dn"dr
--- :4nr'pu-- (3)
dt at Before developinga crystallizermodel in Section 17.5' it is
Equating (2) and (3), useful to describe the most widely used equipment for solu-
tion crystallization. Such equipment may be classified
: +nr'prfi
4trr2k,(Lc) according to the three schemes shown in Table 17.8.
dr
d,: fr
k,L'c
(4)
Table 17.8 Classificationof Equipmentfor Solution
From(15-62), Crystallization
k"D" 2k.r
N* : ry : : 2+ 0.6(Nn";r/21ry..;1/3 Methods for Crystallizer Features
+
Operation Achieving for Achieving Desired
-5l0
Forr: :25 p'm, Modes Supersaturation Crystal Growth
z
D ^ u o ( 5 0 x l 0 - 6") (1 0 0' 'x 5 x 1 . 3 4:)0 . 4 2 Batch Cooling Agitated or nonagitated
-v*.:z:'-""-- Evaporation Baffled or unbaffled
p 0.08 Continuous
o ' 0 8 Circulating liquor or
M"^: P - - :5.430
circulating magma
pD 1 . 3 4 ( 1 . x1 0l 0 - ) )
Classifying or
/26,43cil1/3
Nsn: 2 + 0.6(0.42)r nonclassifying
Controlled or
2k.r
:2+6.8: *.8: uricontrolled
;
Cooling jacket or
( 8 . 8 )lt. l x l 0 - s ) cooling coils
o-_ : 0 . 0 1 9c m / s
2Q5 x l0-4)
 664 Chapter17 Crystallization
Although industrial equipment is available for batch or
continuous operation, the latter is generally preferred. The
choice of method for achieving supersaturation depends
strongly on the effect of temperatureon solubility. From the
data in Table 17.5, it is seen that for many inorganic com-
pounds in the near-ambient temperature range, e.g.,
10-40"C, the changein solubility is small and may be insuf-
ficient to utilize the cooling method. This is certainly the
case for MgCl2 and NaCl. For KNO3 and Na2SOa,crystal-
lization by cooling may be feasible. The majority of indus-
trial crystallizers use the evaporation method or a combina-
tion of cooling and evaporation. Direct-contact cooling with
agitation and evaporation can be achieved by bubbling air
through the magma.
To produce crystals of a desired size distribution, consid-
erable attention must be paid to the selection of features of
the design of the crystallizer. The use of mechanical agita-
tion can result in smaller and more uniformly sized crystals
of a higher purity that are produced in less time. The use
of vertical baffles can promote more uniform mixing. Super-
saturation and uniformity can be controlled by circulation
between a crystallizingzone and a supersaturationzone. In
a circulating-liquor design, only the mother liquor is cir-
culated, while in the circulating-magma design the mother
liquor and crystals are circulated together. The circulation
may be limited to the crystallizer vessel or may include
pumping through an external heat exchanger.In a classify-
ing crystallizeq the smaller crystals are separatedfrom the
larger and retained in the crystallizing zone for further
growth or removed from the zone and redissolved. In a con-
trolled design, one or more techniquesare used to control the
degree or supersaturation to avoid undesirable nucleation.
Coolant
( a )C i r c u l a t i o n o f m a g m a t h r o u g h
a n e x t e r n a l ,c o o l i n g h e a t e x c h a n g e r
Cooling crystallizers may use a vessel jacket or internal
coils, with the latter preferred becauseof the easeof wiping
the crystals off the cooling surface.
A large number of patented commercial crystallizer
designshave been developed.Many of them are describedin
Myerson [10] and by Mullin []. Only four of rhe more
common solution crystallizers are described and illustrated
here. The designs suffice to illustrate most of the features
found in the many other designs.
Circulating, Batch Crystallizers
Although batch crystallizers can be operated without agita-
tion or circulation by simply charging a hot solution to an
open vesseland allowing the solution to standas it coolsby
natural convection, the resulting crystals may be undesirably
large, interlocked, impure becauseof entrapment of mother
liquor, and difficult to remove from the vessel. Therefore,if
the desired product of the crystallization operation is the
crystals, it is preferable to use a more elaborate crystallizer
configuration similar to either of the two batch crystallizers
shown in Figure 17.14.In the design with externalcircula-
tion, a high magma velocity is used for the flow of magma
through the tubes of the heat exchanger to obtain a reason-
able heat-transfer rate with a small temperature-driving
force and minimization of crystal formation on the heat
exchanger tube surfaces. This design can also be used for
continuous crystallization, when the solubility-temperature
curve dictatescooling crystallization.
In Figure l7.I4b, crystallization is accomplishedby
evaporation under a vacuum pulled by a two-stage, steam-
jet-ejector system though a water-cooled condenser. The
Ejectors
Thermocomoressor
Barometric
conoenser
Magma
discharge
Agitator
drive
{b} Internalcirculationwith a draft tube Figure 17.14 Circulating, batch cooling
crvstallizers.

Joint between sections
Long-pitch
Cooling-water inlet
Figure 17.15 Swenson-Walkercontinuous,cooling crystallizer.
magma is circulated internally through a draft tube by a pro-
peller. The energy for evaporation is supplied from the hot
feed. A typical cycle, which includes charging the feed to the
vessel, crystallization, and removal of the magma, may
range from 2 to 8 h.
Continuous,CoolingCrystallizers
Figure 17.15 is a schematicdiagram of a typical scraped-
surfacecrystallizer known as the Swenson-Walker continu-
ous, cooling crystallizer,describedin detail by Seavoy and
Caldwell [12]. The feed flows through a semicylindrical
trough, typically I m wide x 3-12 m long. The trough has
a water-cooledjacket and is provided with a low-speed
(3-10 rpm) helical agitator-conveyorthat scrapesthe wall
and prevents growth of crystals on the trough wall and pro-
motescrystalgrowth by gentleagitation.Standard-sizeunits
can be linked together. The crystallization process may be
controlled by the rather slow rate of heat transfeq with the
major resistance due to the magma on the inside. Overall
heat-transfer coefficients of only 10-25 Btu/h-ft2-'F
(51_142 Wim2-K) are typically observed,based on a log-
mean temperature difference between the magma and the
coolant.Productionratesofup to 20 tons per day ofcrystals
of Na:PO+ . 12 HzO and Na2SOa. 10 H2O of moderatesize
and uniformity have been achieved. Both salts show a sig-
nificant decreasein solubility with decreasing temperature
making the cooling crystallizera viable choice.
EXAMPLE 17.10
Thecoolingcrystallizer ofExample17.6is to be a scraped-surface near the bottom of the elutriation leg where the largest crys-
unit with 3 ft2 of coolingsurfaceper foot of runninglengthof crys- tals are present.Otherwise, the product magma may be with-
Iallizer. Cooling will be provided by a countercurentflow of
chilledwaterenteringthe coolingjacketat 60'F and leavingat
17.4 Equipment for Solution Crystallization 665
85'F. The overall heat-transfer coefficient, U, is expected to be
20 Btu/hr-ft2-"F. What length of crystallizer is needed?
SOLUTION
From Example 17.6, using Figure 17.10,the required rate of heat
transfer is 44,900 Btu/h. The log mean temperature driving force
for heat transfer is
(r20-85)-(70-60)
AII-1a : :20.F
. /120-85\
'n\70-60/
The area for heat transfer is
o 44,900
: 112 ft2
U ATLM 20(20)
The crystallizerlength: I 12l3 : 37 ft.
Continuous,Vacuum,Evaporating Crystallizers
A large number of designs for continuous, vacuum, evapo-
rating crystallizers have been developed.A particularly suc-
cessful and widely used design of this type is the Swenson
draft-tube, baffled (DTB) crystallizer, describedby Newman
and Bennett [13] and shown in one of severalvariationsin
Figure I1 .16.ln the main body of the crystallizer,evapora-
tion occurs, under vacuum, at the boiling surface, which is
located several inches above a draft tube that extends down
to within several inches of the bottom of the main body of
the crystallizer vessel. Near the bottom and inside of the
draft tube is a low-rpm propeller that directs the magma
upward through the draft tube toward the boiling surface
under conditions of a small degreeof supercooling and in the
absenceof any violent flashingaction.Thus, nucleationand
buildup of crystals on the walls are minimized. Surrounding
the draft tube is an annular space where the magma flows
back downward for re-entry into the draft tube. The outer
wall of the annular spaceis a skirt baffle, surroundedby an
annular settling zone, whose outer wall is the wall of the
crystallizer. A portion (perhaps lOTo)of the magma flowing
downward through the first annular space turns around and
flows outward and upward through the settling zone where
larger crystals can settle, leaving a mother liquor containing
fine crystals, which flows to a circulating pipe where it is
joined by the feed and flows upward through a pump and
then a heat exchanger. The circulating solution is heated
several degrees to provide the energy for feed preheat and
subsequentevaporationand to dissolve the finer crystals.
The circulating magma re-entersthe main body of the crys-
tallizerjust below the bottom of the draft tube. Further clas-
sificationofcrystals by size can be accomplishedby provid-
ing an elutriation leg, as shown at the bottom of the main
body of the crystallizerin Figure I7 .16. In that case,product
magma is withdrawn through a pipe from a nozzle located
drawn from the lower part of the annular region surrounding
the draft tube.
666 Chapter17 Crystallization

Barometric condenser

Cool
i nl e t

Propeller
drive

B o i l i n gs u r f a c e

Magma

Drafttube
Skirt baffle
Clarified M.L.
Settl er -------------__,
Settling zone Circulating
pipe

Propeller

Feed inlet

Elutration leg
Productcrystals
Productslurry
d i s c h ar g e

Heating element

Figure 17.16 Swensondraft-tube,baffled

C o n d e n s a t eo u t l e t (DTB) crystallizer.

I7.5 THE MSMPR CRYSTALLIZATION 7. No crystal breakage.

MODEL
Becauseof the popularity of the DTB crystallizer, a mathe-
matical model, due to Randolph [14], for its design and
analysis is quite useful and is now found in the process
equipment model libraries of a few continuous, steady-state,
computer-aidedsimulation programs. This model is referred
to as the Mixed-Suspension, Mixed-Product-Removal
(MSMPR) model and is basedon the following assumptions:
1. Continuous,steady-flow,steady-stateoperation.
2. Perfect mixing of the magma.
3. No classificationof crystals.
4. Uniform degree of supersaturation throughout the
magma.
5. Crystal growth rate independentofcrystal size.
6. No crystals in the feed, but seedsare added initially.
8. Uniform temperature.
9. Mother liquor in product magma in equilibrium with
the crystals.
10. Nucleation rate constant and uniform and due to sec-
ondary nucleationby crystal contact.
11. Crystal-sizedistribution (CSD) uniform in rhe crys-
tallizer and equal to that in the magma.
12. All crystalshave the sameshape.
Modifications to the model to accountfor: (1) classification
of crystalsdue to settling, elutriation, and fines dissolving,
and(2) variable growth rate are discussedby Randolph and
Larson [5]. The central part of the MSMPR model is the
estimation, by a crystal-population balance, of the crystal-
size distribution (CSD), which is determinedin practiceby
the rpm of the propeller in the draft tube and the external
circulation rate. It is relatively easy to conduct experiments
 17.5 The MSMPRCrvstallizationModel 667

in a small laboratory crystallizer that approximates the
MSMPR model and can provide some of the necessarycrys-
tal nucleation rate and growth rate data to design a large-
scale,industrialcrystallizer.
Crystal-PopulationBalance
The crystal-population balance accounts for all crystals in
the magma and, together with the mass balance, makes pos-
sible the determination of the CSD. Let L : a chatacteristic
crystal size (e.g., from a screenanalysis),N: cumulative
number of crystals of size L and smaller in the magma in the
crystallizer, and V7a7: volume of mother liquor in the
magma in the crystallizer. A typical cumulative-numbers
undersize plot based on these variables is shown in Fig-
ure 17.17, where the slope of the cvrve, n, at a given value of
t is the number of crystals per unit size per unit volume,
given by
where t1: residence time in the magma in the crystallizer
for crystals of size ,L. Equation (17-32) is referred to as the
L,L law of McCabe [16], who found that the law correlated
experimental data on the growth of crystals of KCI and
CuSOa' 5H2O surprisinglywell. Although McCabe'sexper-
iments were conducted in a small, laboratory crystallizer
under ideal conditions, his resulting AL law is applied to
commercial crystallization even though conditions may be
far from ideal.
From (17-31),dN : nVyaldL: numberof crystalsin
the size range dL. Now withdraw (An dZ) crystals per unit
volume in time increment A/. Becauseof the perfect mixing
assumption for the magma,
number of crystals withdrawn
*"th*-ttq*. *t"tdt"*"
""1"..
number of crvstals
mother-liquor volume in the crystallizer

tdN Therefore,
d(N /vML)
(17-31) number of crystals withdrawn
dL VTay dL
number of crystals in crystallizer
The limits of n, as shown in Figure 17.I7, vary from no at mother-liquor volume withdrawn
I : 0 to Oat L : L7,the largestcrystal size.In the MSMPR
mother-liquor volume in the crystallizer
model, the cumulative plot of Figure 11.I'7 is independent
of time and location in the magma. The plot is, in fact, the LndL A,n Qvr Lt (r'7-34)
numbers-cumulative CSD for the crystals in the product
ndL n Vur
magma.
For a constant, crystal-size growth rate, independent of where Qlap: volumetric flow rate of mother liquor in the
withdrawn product magma and Vur: volume of mother
crystal size, G : dLldt, or
liquor in the crystallizer. Combinine 0'7-32) and (17-34)
LL: GA,t (r1-32) and taking the limit,
and dn Qurn (r1-3s)
L: Gtr (r7-33) dL GVur

Slope =

=
o
=
E
o
E
f

o
q

\
o
E
:c
0)

=
E
f

LT Figure 17.17 Typical cumulative-

Crystalsize.L ( l a r g e s tc r y s t a l ) numbers-undersizedistribution.
668 ChaPter17 Crystallization

given by
which is a simplified version of the following more-general'
for crys-
transient-populationbalance equation that allows ( l7-45)
but does as-
tals in the feed and a nonuniform growth rate'
sumea constant volume of mother liquor in the crystallizer:
(17-43) is
where k is the order of the moment' Thus'
moment of
0n
*- E ( n G ) *'( Q u r ) o t n - ( Q n t - ! ) i " n i "- o (rj-36) obtained by setting /c : 0 and thus, is the zeroth
Vur and the cone-
at aL vttr the distribution. Results for this moment
(area)' and third
is sponding first (length or size), second
The retention time of mother liquor in the crystallizer
(17-35) can be rewritten as (volume or mass) moments are summarized in Table 17'9'
r : VurlQur. Therefore, plots are
Corresponding cumulative and differential
dL
_ -dn
:_ (r1-3',7) includedin Figure 17.18.
of a
nGr Of particular interest in the design and operation
Lpa' in terms of
and resi- crystaliizer is the predominant crystal size,
If (17-3'7)is integrated for a constant growth rate thl volume or mass distribution. This size corresponds
to the
dencetime, we obtain peak of the differential-mass distribution and is derived
as
n : no exp(-LIGT ) (17-38) follows: From Table 17.9,
distribu- dx* ldz : 173l6)e-' (t't -46)
This equation is the starting point for determining
size or length'
tion currvesfor crystal population, crystal At the peak,
crystal surface area, and crystal volume or mas-s'
of
For example, to obtain crystal population' the number '(*)
crystalsper unit volume of mother liquor below
size ,L is z"e -
(r't-41)
rL dz 6
tt lVur : ndL ( l7-39)
Jo Solving(I1-4'7)for z,
L
mother (17-48)
The total number of crystals per unit volume of c- --7-, -- -
G.r
liquor is
Therefore,
Nr I Vvr = ,ar (17-40) Lp7:3Gr (r7-4e)
lo* in
number of Similar developments using the differential expressions
Combining (17-38) to (17-40), the cumulative Table 17.9 show that the most populous sizes in terms of the
of the total is
crystals oisire smaller thanL, as a fraction number of crystals, the size of crystals, and the
surfacearea
2Gr respectively' If Lpa ts
[ o t r o r - t / G ' d L : 1 _ e x p( _ L / G r ) of crystals are O, Gr, and ,
^- ' - _
(r1-4r) -41) -43) be used to estimate cumu-
selected,(17 and (I1 can
dL
f f n oe - L l G a
lative and differentialscreenanalyses'
Or if we definea dimensionless crystalsizeas To utilize the distributions of Table 17'9, values of G
and
(r1-42) r are needed.The growth rate depends on the supersatura-
7:LlGr depends
tion and the degreeof agitation' The residencetime
crystallizer'It is also use-
then, on the design and operation ofthe
(r7-43) ful to know Bo and ro, which are related as follows:
xn:l-e-z
to I dN I dN (dL\
A plot of (l'7-43), shown in Figure 17'18a' is referred : (17-s0)
V"t d, vrrdL\dt )
asthecumulativedistributionorcumulativecrystalpopula-
hav-
tion. For a given value of z, xn is the fraction of crystals 4{:go
of Also of interest is the corresponding 1_,1
ing a smalllr value e. L-o V741dt
where
diiferential plot of dx,ld'2, shown in Figure 17'18b' dL:G
from (17-43), dt
us ^- n 1 d!- :r'
(r1-44) lim
), L-o V747dL
in a Therefore,
The differential plot gives the fraction of the crystals ( 1 7 - l5)
the fraction is seen Bo : Gno
given interval of z. .q't small values of z,
is small'
io be large, while at large values of z, the fraction Combining(17-51)with (17-38)'
distribution plots
In Figures 17.18a and b, four different
one of a number of -s2)
(1,'7
are sho'in. From statistics,(17-41) is n:Lexp(-LlGr)
G
n : ate
moment equations, which for a relation f(z)'
 17.5 The MSMPRCrystallizationModel 669

0.9

0.8

c o.7
F

0.6
6

X o.t
' .2
0.+
E
<i o.s
0.2

0.1

(a)Cumulativedistributions

0.9

0.8

0.7

0.6
q

0.5
o

o 0.4

o 0.3

0.2

0.1

0
o2 468
Dimensionless
length,z Figure 17.18 Crystal-sizedistributions
(b)Differential
distributions predictedby theMSMPR model.

Table 17.9 CumulativeandDifferentialPlotsfor Momentsof CrvstalDistributionfor Constant

Growth Rate

Moment DistributionBasis Cumulative Differential

Zercth Number xn: | - e-z dx,fdz - e-z
First Size or length xL:l-(l-tz)e-z dxL/dz: ze-z
/ .2\ ,
7-
Second Area xa:t-lr+z+ilr-, dxo/dz: \e-'
\ z/ 2
/ -2 -3\ 1

Third Volume or mass xn:t-('*z+i+Z)u, d x ^ /' d

oz :
l-e-'

7: L/Gt
670 Chapter17 Crystallization

This equation can be used with experimental data, as shown in Estimate with the MSMPR model:
the following example, to obtain values of nucleation and (a) Required crystal growth rate in ft/h
growth rates for a given set of operating conditions when the (b) Necessary nucleation rate in nuclei/h-ft3 of mother liquor in
assumptions of the MSMPR model hold. The effect of operat- the crystallizer
ing conditions on Bo can be expressedas an empirical, power-
(c) Number of crystals produced per hour
law function of the form given by (17-19). However, since the
growth rate can be proportional to the relative supersaturation, (d) Tablesand plots of estimatedcumulative and differential screen
analysesof the product crystals on a mass or volume basis
s, raisedto an exponent,(17-19) can be rewritten as
Also explain how the required growth and nucleation rates for the
Bo : k'uci uJ, N' (17-s3) operating crystallizer might be achieved.
Unfortunately,/cl,canbe sensitiveto thesizeof theequipment
andis, therefore,bestdeterminedfrom datafor a commercial SOLATION
crystallizerasdiscussed by ZumsteinandRousseau [17].The (a) From(17-48),
j,
exponentsi, and r can be determined from small-scale
experiments. - " Lod :- - Q.417
-- : 3 /304'8):2.28 x t0-4ft/h
3(2)
The necessary nucleationratefor a crystallizationopera-
(b) Need volume of mother liquor in the crystallizer.
tion is relatedto thepredominantcrystalsizeby theMSMPR
model.From(17-40)and(17-38), : 1990 : 19.1ft3h.
produced
volumeof crystals
lu5
f@roo
Volumeof crystalsin crystallizer: l9.l(2) : 38.2ft3.
n,:NrlVut: I ndL:norG I t-'dz:norG
Jo Jo
Volumeof magmain crystallizer
- : = : 255ft3.
07-54) u.l5
The massof crystalsper unit volumeof motherliquor is Volumeof motherliquor in crystallizer: Vut
: 2 5 5 - 3 8 . 2: 2 1 7 f t r .
f@
mr: I mondL (17-ss) From(17-60),assuming/,- 0.5,
JO
9(2,000)
where mr: rnssSof particle, given by Bo: :3.1 x 108try
2(0.5X105)(2 / 3O4.8)3
r7)(0.4r7 h-ft'
mo: frL3po (r7-s6) (c) Numberof crystalsproduced: 3.1 x lO8Ql7)
where/, is definedin (17-11). : 6.7 x 10ro/h.
Combining (17-55), (17-56), and (17-38), followed by L
(d) -7 : -
integration gives Gr (2.28x 10-4)(2)
mc :6frPrno(Gr)a (r7-s7) : 2 . 2 x 1 0 3Z , f t o r 7 . 1 9L , m m

Combining(17-54) and(17-57),the numberof crystalsper FromTable17.9,

unit massof crystalsis / ,3\
x m : r - ( t + z + + +"z2i l r -o' / (1)
nr- | \
(17-s8) zt ^-'
m, 6fu po (Gr)3 4*^ \z)
dr:6t'
Combining (17-48) with (17-58),
Using (1) and (2) with a spreadsheetfor the older Tyler mesh sizes,
n"- 9 the following results are obtained for the cumulative and differen-
(r7-s9) tial screenanalyses:
m, 2f, pp L3pa
Cumulative Differential
or the corresponding required nucleation rate is
T$er Opening, Dimensionless Screen Screen
- n.C gC Mesh mm Length, z Analysis, 7o Analysis7o
4u - _____:_
'?
(17-60)
m"Vu t 2frp oVu rL"rl 8 2.357 16.96 100.00 0.00
10 1.667 11.99 99.77 0.18
where C : mass rate of production of crystals. 1.t79 8.48 96.95 2.ll
l4
20 0.833 5.99 84.82 8.95
28 0.589 4.24 61.16 18.31
:,$iil 35 0.417 3.00 35.20 22.40
A continuous,vacuum-evaporating crystallizerof the DTB type 48 0.295 2.12 16.50 19.05
is to be used to produce2,000 lb/hr of Alz(SO+)g'l8HzO 65 0.208 1.50 6.54 12.53
(pp : 105lb/ft'). Th" magmacontains0.15ft3 crystals/fCmagma 100 o.r47 1.06 2.29 6.87
andthe residencetime of the magmain the crystallizerin2h.The 150 0.104 0.75 0.73 3.31
desiredpredominantcrystal size on a massbasis is 0.417 mm. 200 0.074 0.53 0.22 1.46

80.00
70.00
o 60.00
o 50.00
o
40.00
30.00
20.00
10.00
0.00 L
0 2.5
Aplot ofthe cumulative and differential screen analyses is shown
in Figure 17.19.
Both growth and nucleation ratesdepend on the supersaturation.
The growth may also depend on the relative velocity between the
17.6 PRECIPITATION
Solution crystallization,as discussedabove, occurs when a
solution containing a moderately to strongly soluble solute is
cooled or partially evaporated,causing the solute to exceed
its solubility sufficiently to partially crystallize.The degreeof
supersaturationis small, primary nucleation is avoided, and
crystal growth is slow. Ifthe processis carried out under con-
trolled conditions, reasonablylarge and desirablecrystals can
be grown. In many respects, the process of precipitation is
just the oppositeof solution crystallization.precipitation,as
discussedby Nielsen [ 18], involves solutesthat areonly spar-
ingly soluble. The precipitate is formed by changing solution
pH, solvent concentration, solution temperature, or most
commonly by the addition of a reagentto a solution, resulting
in a reaction to produce anotherchemical that is almost insol-
uble in the resulting solution. The latter is commonly referred
to as reactive crystallization. The degree of supersaturation
produced by the reaction is very large, causing a high degree
of primary nucleation. Although some growth occurs as the
supersaturationis depleted, precipitates generally consist of
very small particles that may be crystalline in nature and are
frequently Ltggregates and agglomerates. Aggregates are
masses of crystallites that are weakly bonded together.
Agglomeration can follow, cementing aggregatestogether.
Typical chemicals that are produced by precipitation are the
sparinglysoluble compoundslisted in Table 17.6.precipita-
tion may be thought of asfast crystallization.
17.6 Precipitation 671
Figure 17.19 predicted cumulative and
differential screenanalysesfor Example 17. 11.
crystals and the mother liquor. The nucleation rate dependsupon the
degreeof agitation. With a DTB crystallizer, the agitator rpm and the
magma circulation rate through the external heat exchanger can be
adjustedto achieve the required growth and nucleation rates.
In precipitation, particle size is related to the solubility by
(17-16), the Kelvin equation. However, for precipitates
formed from ionic reactions in solution, the supersaturation
ratio, S : c/cs, is replaced by (T/K)1/,, where tr : the
ionic concentration product for the reaction, 6, : (equilib-
rium) solubility product, and u : sum of the number of
cations and anions that form the precipitated compound.
Thus, for aluminum hydroxide at 15.C,
n: (cff)(co")3
with K. given in Table 17.6and u : I * 3 : 4.The Kelvin
relation is of great importance in precipitation because
nucleation can be due to homogeneousprimary nucleation.
The extent of supersaturationis a very important factor in
precipitation. When a reagent is added to a solution to form
a sparingly soluble compound, a very high supersaturation
can be developed, depending upon the ionic concentrations
in the reagentand solution before mixing. For example, con-
sider the formation of a precipitate of BaSOa from an aque-
ous reagentsolution containing Ba++ ions and an aqueous
solution of, e.g., sulfuric acid.
From Table l'7.6,the solubility product of BaSOaat 1g.C
is 0.87 x 10-10.Figure 17.2}a,rakenfrom Nielsen [19], is a
plot of sulfate ion concentration versus barium ion concen-
tration, both in the solution just after mixing and before pre-
cipitation, with contours ofconstant supersaturationratio, S.
The dashed lines refer to an ideal situation where activity
672 ChaoterlT Crvstallization

o precipitation generally follow one of the following laws,

)- which exhibit different dependencieson the relative super-
2 " saturation,s : S - 1:

$
cr trr
j

"la
*? Linear
a
6.
tn
6z
G:kts (r7-61)
o
T ft
Parabolic
0
012345 60123456 G : kzsz (r7-62)
- l o g l B a 2 * 1m, o l / L - l o g l B a 2 * 1m, o l / L
(a)Extentof supersaturation ( b )I n d u c t i o np e r i o d Exponential
6 G : kzf{sl (r7-63)
The linear rate law, which often applies for G > 10 nm/s,

@
l J

trr may be due to mass-transfer control or surface-adsorption

*c control, where the latter dependsmore strongly on tempera-
a
ture. The parabolic rate law, for which G may be < 10 nm/s,
-a2
applies to screw-dislocation-controlled growth. The expo-
ft
nential law, where /{s} can involve a complex exponential,
0! log, and/or power-law dependency corresponds to growth
01 2345
- l o g l B a 2 * 1m, o l / L control by surface nucleation. The latter two mechanisms
( c )N u m b e ro f p a r t i c l e sf o r m e d can occur in parallel. Rate constants for many electrolytes
Figure 17.20 Precipitation diagrams for BaSOa. for the three different rate laws are given by Nielsen [20].
When growth is rapid, coprecipitation of soluble electrolytes
may occur by entrapment,thus making it difficult to obtaina
pure precipitate product.
Becausea precipitate is formed at considerable supersat-
coefficients are 1.0 (concentrations : activities) and no
uration, the resulting particle shapes may be far from the
complexesof Ba++ and SO;, are formed. The solid, curved
shape corresponding to a minimum Gibbs energy, which
contours take these deviations from the ideal situation into
depends on the particle surface area and interfacial tension.
account. Note the very large supersaturationratios that can
However, if the precipitate and mother liquor are allowed to
occur.For example, if the SOt concentrationin the solution is
age,then, as discussedby Nielsen, [20], the precipitatepar-
0.01 mol./Land the Ba++ concentrationis 4.7 x 10-a mol/L,
ticle sizesand shapescan tend toward equilibrium valuesby:
the supersaturation ratio: 100.
(1) flocculation and sintering of fine particles (2) transportof
Typically, precipitation does not take place immediately
ions over the surface, and (3) ripening by dissolution and
upon development of a large supersaturationbecauseof the
redeposition. Ripening can result in the releaseofcoprecipi-
slow growth of very small particles. However, after a certain
tates, thus increasing precipitate purity.
period of time, called the induction period, Z1 visible precip-
The small particles produced in abundanceduring precip-
itation begins. As shown in Figure 17.20b, this period is
itation have a tendency to cluster together by interparticle-
highly dependent on the initial-ion concentrations and, by
collision phenomena,variously referred to as agglomeration,
superposition of Figure l7.2Oa, on the supersaturationratio.
aggregation, and flocculation. Such clusters, often called
For example, at a supersaturationratio of about 300, the in-
agglomerates,are common when the number of particles/cm3
duction period is only one second,while at a supersaturation
of solution exceeds107.For BaSO+,Figures l'7.20a andc
ratio of about 10, the induction period is more than one
show that agglomeration may require a very large initial
minute.
supersaturation ratio. Agglomeration is important mainly
As shown in Figure 11.20c, the number of particles
when particles sizesare between 1 pm and 50 pm.
formed per unit volume of solution, NlV, also depends
strongly on the initial concentrationsofthe anion and cation
and, therefore, on the supersaturationat high ion concentra-
tions. A minimum number of particles per unit volume or a EXAMPLE 17.12
corresponding maximum particle size is frequently
FitchettandTarbell[21] studiedthe effect of impellerrpm on the
observed.For BaSOa,this maximum size occursfor a super-
crystal-sizedistributionobtainedfor the continuousprecipitation
saturation ratio of about 300. The number of particles of barium sulfatewhen mixing solutionsof sodium sulfateand
formed also depends on particle growth rate, which may barium chloride.The contentsof the 1.8-L crystallizerwere as-
be controlled by mass transfer of the ions to the particle sur- sumedto be perfectlymixed.In their Run 15,theyusedfeedswith
face and/or integration of ions into the particle crystalline dissolvedsalts in stoichiometricratio to give a sodium-chloride
structure (surface reaction). Particle growth rates during concentration of 0.15mol/L andan averaseresidencetime of 38 s.
 l7.7 Melt Crystallization 673

Results for two different impeller speedswere as follows: (f) In a similar manner,the following resultsare obtainedfor
400rpm:
ln n, Number Density of Crystals,ln (pm-1 t-l)
Intercept: 25.53,slope: -0.313
Size, 950rpm, 400 rpm, n" :1.22 x l0rrcrystals/pm.L
p"m 0.361J/s.kg Feeds 0.028Jls'kgFeeds
G : 0.0841 p.m/s
7 22.07 23.45 Bo : 1.03 x lOlonuclei/L. s
9 21.66 22.75
Lmean: 3'20 Pm
1l 21.35 22.08
l3 20.97 21.36 n, :3.91 x 108crystals/m3
l5 20.77 20.75 Comparing the two sets of results, we see that a higher agitator
17 20.41 20.14 speed gives a larger mean crystal size, a larger growth. rate, and a
t9 20.04 19.57 lower nucleation rate.
21 19.77 18.94
z3 19.48 18.43
25 19.09 t7;15 I7.7 MELTCRYSTALLIZATION
2'7 18.85
29 t8.49 Solution crystallization refers to crystallization of a solute
3l 18.11 from a solvent. It is most commonly conductedwith aqueous
33 17.87 solutions ofdissolved inorganic salts.The phaseequilibrium
diagram for a water-salt system, e.g., Figure 17.2, typically
Using the MSMPR model, determine from the two sets of data: includes temperaturesranging from the eutectic temperature
(below 0"C) to a temperatureexceeding the melting point of
(a) no, number density of nuclei, nuclei/pm-L
ice, but not greater than 200'C. Since the melting points of
(b) G, linear growth rate, prn/s
inorganic salts greatly exceed 200'C, the pure-salt melting
(c) Bo, nucleation rate, nuclei/L-s point is, accordingly, not included on the diagram.
(d) mean crystal length, pm For mixtures of organic compounds, the situation is quite
(e) n", number of crystals/volume of mother liquor, crystals/m3
different. An analysis by Matsuoka et al. l22l found that
more than 70% of the common organic compounds had
melting points between 0 and 200"C. For binary mixtures of
SOLATION such compounds, the phase-equilibrium diagrams will typi-
(a) From (17-38), if each set of data is fitted to cally include the melting points of both compounds.Typical
diagramsare shown in Figure 17.1.In (b), crystalsof ortho-
ln n: ln no - Lf Gr (1)
chloronitrobenzenecan be formed if the feed composition is
the best straight line yields an intercept of ln no and a slope of less than the eutectic composition based on para-chloroni-
-l/Gr, from which no and G can be determinedfor r : 38 s. trobenzene;otherwise, pure para-chloronitrobenzenecan be
Using a spreadsheet,the results for 950 rpm are formed. The exception is the eutectic composition. Eutectic-
intercept: 23.13,slope: -0.1601 forming systemsconsist of compoundsthat can not substi-
tute for each other in the crystal lattice. Thus, the eutectic
Therefore,no : exp(23.13): 1.1I x 1010crystals/pm . L
mixture consists'of two different solid phases.The two solu-
(b) G: 1/t(38X0.1601)l: 0.164 u.m/s bility curves separatingthe liquid-phase region from the two
(c) The nucleationrate is given by (17-51): solid-liquid regions are sometimes referred to as freezing-
B o : G n o : 0 . 1 6 4 ( 1 . 1 1x 1 0 1 0 :) 7 . 8 2x l O e n u c l e i / L .s point curves,and mixtures at conditions in the liquid-phase
region are referred to as melts.
(d) The mean particle length is determined from the value of z
Much less common are solid-solution-forming systems
for the maximum value of dxp/dz, which, from Table 17.9
tsze '. of the type shown in Figure 17.1c for the phenanthrene-
anthracenesystem. These systemsconsist of compounds so
'(ze-'1 :U nearly alike in structure that they can substitute for each
d::e'-Z.e
dz
other in the lattice structureto form a single solid phaseof the
Therefore.z: | : L^runf Gr trom(17-42\ two compounds over a wide range of composition. With this
L^"un: Gr :0.164(38) : 6.23 pm type of system, a mixture is crystallized. The liquid-solid
phase diagram resemblesthat for vapor-liquid equilibrium,
(e) The number of crystals per unit volume of mother liquor is
where the freezing-point and melting-point curves replace
given by (17-54).
the dew-point and bubble-point curves, respectively. Mix-
f r c : n ot G : B o r : 1 . 8 2 x 1 0 9 ( 3 8 ) tures in the liquid-phase region above the freezing-point
: 6.92 x 1010crystals/L :6.92 x 107crystals/m3 curve are also referred to as melts. A mixture in the resion
674 Chapter17 Crystallization
between the two curves separatesinto a liquid phase and a
solid phase,neither of which is pure.
Crystallization of melts from either eutectic-forming or
solid-solution-formingmixtures is called melt crystalliza-
tion. Although theoretically, melt crystallization of eutectic-
forming systems, like solution crystallization of such sys-
tems, can produce pure crystals, the crystalline product from
commercial single-stage crystallizers may not meet purity
specificationsfor reasonsdiscussedin detail by Wilcox [23].
In that case, repeated stages of melting and crystallization
may be necessary.With eutectic-forming systems,multiple
stagesare mandatory to produce crystals of high purity.
The separation of organic mixtures is most commonly
achievedby distillation. However, if distillation: ( 1) requires
more than I 00 theoretical stages,(2) cannotproduce products
that meet specifications (e.g., purity and color), (3) causes
decomposition of feed components, or (4) requires extreme
conditions of temperature or pressure (e.g., vacuum), other
separation methods should be considered. According to
Wynn 1241,if the compounds to be separatedare: (1) disub-
stituted benzenes,diphenyl alkyls, phenones,secondaryor
tertiary aromatic or aliphatic amines, isocyanates,fused-ring
compounds,heterocycliccompounds,and carboxylic acids
of MW < 150; (2) have a melting-temperature range from 0
to 160'C; (3) required to be very high-purity products;or if
(4) a laboratory test produces a clearly defined solid phase
from which the liquid phase drains freely, then melt crystal-
lization should be consideredas an alternative or supplement,
as in a hybrid process,to distillation.
Equipment for Melt Crystallization
As with solution crystallization, a large number of crystal-
lizer designs have been proposed for melt crystallization.
Crystalslurry
Only some of the more widely used commercial units are
discussedhere.Myerson [10] gives a detaileddiscussion.In
all cases, crystallization is caused by cooling the mixture.
The two major methods used in melt crystallizers are.rur-
pension crystallization and layer crystallization by progres-
sive freezing. In the former, crystals of a desired size distri-
bution are grown slowly in a suspension by subcooling a
seeded-feedmelt. In the latter, crystalsof uncontrolled sizeare
grown rapidly on a cooled surface,wherein subcooling is sup-
plied through the crystallized layer. In suspensioncrystalliza-
tion, the remaining melt must be separatedfrom the crystalsby
centrifugation, filtration, and/or settling. In layer crystalliza-
tion, the remaining melt or residual liquid is drained from the
solid layer, followed by melting of that layer.
Figure I 7.2 I shows a two-stage, scraped-wall-crystallizer
systemusedfor suspensioncrystallization. A cooling medium
is usedto control the surfacetemperaturesof the two scraped-
wall units, causing crystals to grow, which are subsequently
scrapedoff by screws.The melt-crystal mixture is circulated
through a ripening vessel.The two scraped-wallunits of a typ-
ical system are3.6 m long with 3.85 m2 of total heat-transfer
surfacearea.The screwsare driven by a l0-kW motor.
Of more importance in commercial applications of melt
crystallization is the falling-film crystallizer developed by
Sulzer Brothers Limited and shown in Figure 17.22. This
equipment is particularly useful for producing high-purity
crystals(>99 .9Vq)at high capacity(> 10,000tons/y).A large
pair of units,each4 m in diameterand containing1,10012-m-
high tubes can produce 100,000tons/yr of very pure crystals,
with typical layer growth rates of 1 in./tr. The feed melt flows
as a film down the inside of the tubes,over a crystal layer that
forms and grows by progressive freezing because the wall
of the tube is cooled from the outside by a coolant. When a
predetermined crystal layer thickness, typically 5-20 mm, is
Circulationpump
Figure 17.21 Two-stage,
scraped-wall, melt crystalhzer.

F a l l i n gf i l m
heat carrier
F a l l i n gf i l m
melt
Collector
Meltoutlet
Figu.re17.22 Sulzerfalling-film melt crystallizer.
reached,the feed melt flow rate is stopped.The tubes are now
slightly warmed to causepartial melting, called sweating, to
remove impurities that may be stuck or bonded to the crystal
layer. This is followed by complete melting of the remaining
layer, which is of high purity. During the initial crystallization
phase,melt is circulated at a high rate compared to the crys-
tallization rate so that a uniform temperatureand melt com-
position is approached down the length of the tube. The
coolant also flows as a film down along the outside surfaceof
the tubes.
Consider the freezing step in a falling-film crystallizer. A
typical temperatureprofile is shown in Figure 17.23.The
melt entersat the top of the tube and flows as a film down the
inside wall. A coolant, at a temperature below the freezing
point of the melt, also enters at the top of the tube and flows
as a film down the outside wall. Heat is transferred from the
melt to the coolant, causing the melt to form a crystal layer
at the inside wall of the tube. As the melt and coolant flow
down the tube, their temperaturesdecreaseand increase,re-
spectively. The thickness of the crystal layer increaseswith
time, with the latent heat of fusion transferred from the
crystal-melt interface to the coolant.
17.7 Melt Crystallization 675
Figure 17.23 Temperatureprofile for melt crystallization from a
fallins film at a vertical location along the tube.
For a eutectic-forming, binary melt, only one compo-
nent will crystallize, although small amounts of the other
component may be trapped in the crystal layer, particularly
if the rate of crystal formation is too fast. The temperature
at the interface of the crystal layer and the melt can be
assumed to be the equilibrium temperature corresponding
to the saturated melt composition on the solubility curve.
If mass transfer of the crystallizing component from the
melt film to the phase interface is rapid, the interface tem-
perature will correspond to the melt-film composition at
that vertical location. Furthermore, if the thermal resis-
tances of the coolant film. tube wall. and melt film are
negligible compared to that of the crystal layer, and if the
heat capacity of the crystal layer and metal wall are neg-
ligible, then the following simple model for the rate of
increase of thickness of the crystal layer with time can be
constructed.
At a particular vertical location, T, N T*, the melting
point, and Ti- N Tr, the coolant bulk temperature.The rate
of heat releasedby freezing, A,Hy, is equal to the rate of heat
conduction through the crystal layer. Thus, for a planar wall,
referring to Figure 17.23, the heat evolved from fusion is
equated to the rate of heat conduction through the crystal
layer to give
dm dr. k, A(T* - T,)
L H 'r - : - ! p , ( A , H 1 ) i (r1-64)
dt " ' d t fi-fs
with an initial condition rs - r; &t t :0. Integration of (17 -64)
gives
(ri-rr)z
_kr(T*-7,)t (17-65)
2 p, LHf
 676 Chapter77 Crysrallizarion

or crystal layer thicknessis 75.2Voof the crystallayer. The time required for this thickness
of growth for the cylindrical tube is only g\Vo of the time for
2k,(T* - T,)t the planar wall. Thus, a conservative result is obtained by
lri - rt) : (r7-66)
p' LHf using the much simplerplanarwall caseof (17-66).
During operation of the falling-film crystallizer, the tem_
A similar derivation for a cylindrical tube wall, where r; : perature of the melt film decreasesas it flows down the tube
inside radius of the tube, gives becauseof heat transfer from the melt to the colder crystal
layer. Based on the earlier assumptions,the melt temperature
7,, 1\ r!. / r i '\l k.(T--T"\t at any elevation will be the temperaturecorresponding to the
n"l r i - r : ) - + l n2 l
4 p"LHr e7_67) solubility curve for the melt composition at that elevation. If
\r,/
we: (1) assumethat the sensibleheat from the cooling of the
As would be expected, the value of the left-hand side of melt layer is negligible compared to the latent heat of fusion,
(l'7 -67) approachesrhe value of the left-handside of ( I 7_65) (2) neglect any sensible-heatstoragein the crystal layer and
as the value of r" approachesthe value of 4. For the planar tube wall, (3) neglect vertical conduction in the tube wall,
wall, (17-66) showsthat the crystallayer grows as the square crystal layer, and melt layer, and (4) assume a constant
root of time. Thus, for a given period of time, the growth dur_ coolant temperature, we can couple (lj-66) with a material
ingthefirsthalf of thetimeperiod is{o/J2): j0.7Voof the balance for the depletion of the crystallizing component as it
total growth. For the cylindrical-tube wall, if the growth is flows down the inside tube wall. The result, which is left as
from the wall to half of the radius, 67.9Vo of that growth an exercise,gives the crystal-layer thickness as a function of
occurs during the flrst half of the time period to produce time and vertical location.

EXAMPLil 1?.13
A melt of 80 wt%onaphthalene and 20 wt%obenzene at the satura_ The estimatecan be madewith (17-67),basedon the following
tlon temperature is fed to a falling-film crystallizer, where coolant propertiesfor naphthalene:
enters at 15'C. Estimate the time required for the crystal_layer
thickness near the top of the 8-cm i.d. tubes to reach 2 cm. P" : 71.41b/ft3
k" :0.17Btu/h-ft-'F
SOLATION LHf :63'9Btunb
From Figure 17.24, by extrapolation the saturation temperature of
From(17-67),solvingfor time,r,
the melt is 62'C. Therefore, naphthalenewill crystallize, with
, _ (7r.4)(63.e)
T- - T" : (62 - 15)1.8: 84.6"F (0.l7)(84.6)
ri :8/2:4cm: 0.131ft
rs:4-2:2cm:0.0655ft
x I l ^ - - - , 0.06552)
2 . 0 . 0 6 5 5' "2/,o . l 3 l
( ffi) l
\l
:0.54eh
lo( 0.13r ' -

40
Homogeneous solution

e
f
so
o
o

Fro
0)

10

o f s o l i d C 5 H ^a n d s o l i d

40 50 60 70
Figure 17.24 Solid-liquid phase diagram for
Weight, YoCrcHBin solution
the naphthalene-benzenesystem.

I7.8 ZONE MELTING
When the melt consists of two components that form a solid
solution,such as shown in Figure 11.lc,the liquid and solid
phasesat equilibrium contain amounts of both components,
similar to the behavior previously discussedfor vapor-liquid
equilibrium. Accordingly, multiple stagesof crystallization
are required to obtain products ofhigh purity. Aparticularly
useful technique for doing this, especially when impurities
are to be removed to achieve a very high-purity product, is
zone melting, as developedby Pfann I25,261 and discussed
further by Zief and Wilcox [27].
The zone-melting technique, when carried out batchwise,
is illustrated in Figure 17.25. Starting with an impure crystal
slab, a melt zone, which is thin compared to the length of the
slab, is passedslowly (typically at 1 cmlh for organic mix-
tures) through the slab from one end to the other by fixing
the slab and using a moving heat source or, less commonly,
by moving the slab through a fixed heat source. Radiofre-
quency (RF) induction heating is particularly convenient
and causesmixing in the melt. The slab can be arrangedhor-
izontally or vertically, with the latter orientation preferred
becauseit can take advantageof any density difference be-
Refined
- - - ' C o n t a i n e r- - = - - - - - - - . r . - - - - -
)
tube zone movestnts way
Figure 17.25 Zone melting.
[FrcmPerry'sChemicalEngineers'Handbook,6thed.,R.H. perry,D.W
Green,andJ.O.Maloney,eds.,McGraw-Hill,New York (1984)with
permission.l
Refined portion
of crystal layer
ws
l-,----*l
Direction of movement
of melt zone
I7.8 ZoneMeltins 677
tween the crystals and the melt.Zone melting can be applied
to the composition region near either end of the phase-equi-
librium diagram. For example, for the binary mixture of
phenanthrene(P) and anthracene(A) in Figure lJ.1c, zone
melting could be used to remove small amounts of A from p
or small amounts of P from A.
A solid-liquid equilibrium distribution coefficient, K, can
be defined in a manner analogous to that for liquid-liquid
equilibrium. Thus,
z _ concentration of impurity, i, in the solid phase
" concentration of impurity, i, in the melt phase
(17-68)
For example, if the concentration is expressed in mole
fractions and the impurity is anthracenein the system of Fig-
ure 17.lc, thenat 120'C,Ki :0.30/0.12:2.50.Ar 200'C,
with phenanthreneas the impurity Kr : 0.1 | /0.28: 0.393.
Thus, K; can be greater than or less than one. When greater
than one, the impurity raisesthe melting point and the impu-
rity concentratesin the solid phase; when less than one, the
impurity lowers the melting point and concentrates in the
melt phase. However, when K; approaches a value of l,
purification by zorrcmelting becomesvery difficult.
Near either end of the composition range, the equilibrium
curves for the solid and liquid phasesapproachstraight lines,
and therefore, the values of K; become constants. In these
Unrefined
composition ranges,an equation is readily developed to pre-
dict concentrationof the impurity in the solid phaseupstream
of the moving melt zone as a function of distance down the
crystal layer in the direction of movement of the melt zone.
Figure 17.26 shows the position of the melt zone during zone
melting. The melt zone is shown to move a distance dz.
Assume that: (1) the melt zone of width, l, is perfectly mixed
with impurity concentratiotr, w1, in weight fraction; (2) no
diffusion of impurity in the solid phasesoccurs; (3) the initial
concentration of the impurity is uniform atwo; and (4) the
concentration of the impurity in the melt zone is in equilib-
rium with that in the solid phaseupstreamof the melt zone.A
mass balance on the impurity for a dz movement of the melt
Melt Portion of initial
zone crystal layer
wL
Figure L7.26 Ideal, zone-melting
model.
678 Chapter17 Crystallization

zone is given by zll ws

(Mass of impurity addedto melt zone) 0.0085

- (massof impurity removed from melt zone) 5 0.0089
: (increasein the mass of impurity in the melt zone) 6 0.0093
7 0.0095
Thus, 8 0.0096
9 0.0097
wo pc Ac dz - w, p, A, dz - p1 Ar.(.dw y Q7-69)
From (17-68), let In the melt zone,w L : w, / K : 0.0097/0.36 : 0.0269. A plot of
the predicted profile is shown in Figure 17.27. To further refine the
K : ws/wr o1-10) anthracene,additional passesof zone melting can be made to move
more of the impurity into the melt zone. However, for each pass
Therefore,
after the first, (17-73) is not valid becauseat the beginning ofeach
dwr - dw,/K (r7-71) of the additional passes, w, is not uniform. For example, at the
beginning of the second pass, )?obecomes w" in Figure 77.27. For
Combining (17-69) to (17-71) to eliminate wTand assuming the additional passes,it is necessaryto numerically solve the fol-
p r : p c ,w e o b t a i n lowing differential-equation form of (17-72).

XI"'n,:I;:" dw,
wo-ws
(r7-72) 0.03

Integration gives
0.025

\:t-il- K\exp(-Kz/t) (r1-73) c

0.02
Refinedportion Melt
for of crystallayer layer
!
L
z/t :0 to z/t - - - 1
(. 0.01

EXAMPLE 17;14 0.005 Directionof movementof melt laver

A crystal layer of I wtVo phenanthreneand 99 wt%oanthracene is

subjectedto zone melting with a melt-zone width equal to 0. I of the
3
012345678910
length of the crystal layer (l / L : 0. 1). The distribution coefficienr Ratioof positionalong layerto length of meltzone,zl(
for phenanthrene in the dilute composition region is 0.36. Deter- Figure 17.27 Predicted impurity profile for Example 17.14.
mine the phenanthreneconcentration profile in weight fractions at
the conclusion of zone meltins when the melt reachesthe last 107o
of the crystal layer length. 102

101
SOLATION
From Figure 17.1c and the value of K, the phenanthrenefavors dis- 1
tribution to the melt phase. Therefore, as the melt zone moves
--
down the crystal layer, the phenanthrenewill migrate from the up- 10-'
stream portion of the crystal layer into the melt zone. When the F
leading edge of the melt zone reaches the end of the crystal layer, S ro-'
all of the phenanthrenethat has migrated will be in the melt layer at o

a uniform concentration equal to wr{z/l:0.9}/K. Assume 3 to-'

a
instant freezing of this melt zone so as to maintain the uniform 0)
'!
composition.From (17-73),with K: 0.36 and wo:0.01, i n-4
G '-
6
w " : 0 . 0 1 [ 1 - 0 . 6 4e x p ( - 0 . 3 6 2 / l ) l (t) 10-5

Solving (1) for valuesof z / I : 0 to 9 gives

10-6
Ultimatedistribution
zlt ws
10_7
0 0.0036 0 0.2 0.4 0.6 0.8 1.0
I 0.0055 Relativeposition,/t
z 0.0069 Figure 17.28 Concentrationprofiles for multiple zone-melting
3 0.0078 passes.
 17.9 Desublimation 679

dw" K (17-74) If naphthalene is present in an ideal gas mixture with a

__;-:;\wo-ws)
az 1 noncondensibleinert gas, sometimescalled an entrainer' at a
where luo : wo{z: z} from the resultsfor w, for the resultsfrom
the previouspass,and w s{z : 0l : Kw o{z Oi ' : of the naphthaleneto a crystalline solid will occur if the par-
iypical impurity-concentration profiles for multiple melt-zone
pur."t ." shown in Figure 17.28 from calculations by Burris'
Stockman, and Dillon [28] for the caseof K : 0' 1 and a melt-zone
width of 2OVoof the crystal-layer length (l I L :0'2) ' Howeveq
their calculations, the melt zone was allowed to diminish to zero as
crystal layer, resulting in a steep
the heater moved away from the
gradient from z I L : 0'8 to 1'0. It is seenthat an ultimate distribu-
iion of the impurity is approached after 9 passes'The number of
passes required to approach the ultimate distribution is given
approximatelyby 2(Ll0 * 1, as observedby Herington I29l'
thi, e*urnpt", a highly pure crystalline product can be achieved
portion of the more pure end of the final crystal layer is taken
I7.9 DESUBLIMATION
When crystallization occurs from a vapor phase,rather than
from a liquid solution or melt, the operation is referred to
desublimation.The reverseoperation,i'e', the vaporization
of a solid directly to a vapor, is sublimation' To understand
how such a phase change can occur without going through
the liquid phase,consider the phasediagram for naphthalene
shown in Figure 11.29. As with most chemicalsthat have
vapor pressureof much less than 1 atm at the melting tem-
p"iutoi", that temperature coincides, within a fraction
d"g."" Celsius, with the triple point. Below the triple-point
teriperature of 80.2'C, corresponding to a vapor pressureof
7.8 ton, naphthalenecannot exist as a liquid, regardlessof
the pressuri. Below 80.2'C, naphthaleneexists as a solid
provided that the pressureis greater than the vapor pressure
of naphthaleneat the prevailing temperature'However, if the
pr"r*." falls below the vapor pressure, solid naphthalene
will sublime directlY to a vapor.
temperaturebelow the triple point of 80.2'C, desublimation
tial pressureof naphthalenein the gas is increasedto a value
that exceeds its vapor pressure. For example, consider a
vapor mixture of 5 mol%onaphthalene in nitrogen at 70'C
in
and a total pressureof 40 torr. The partial pressureof naph-
thalene is 0.05(40) : 2 ton. From Figure 17'29' this partial
pressureis less than the vapor pressureof 3'8 torr' Ifthe total
p."rru." is increased, desublimation will begin when the
iotal pressureis increasedto 3'8/0'05 :'76 torr' Alterna-
In tively, if the pressureis maintained at 40 torr' but the tempera-
if a ture is reduced,desublimationwill begin to take place at 61'C,
where the vapor pressure of naphthalene is 2 torr' Unless
nitrogen is trapped,the naphthalenecrystals will be pure'
The most common occurrence of desublimation is the
formation of snow crystals by condensationof water vapor
from air at temperaturesbelow 0'C. At temperaturesabove
-40'C, the snow crystalsform by heterogeneous nucleation
mm in
as on fine mineral particles of perhaps 10-5-10-2
-40oC,
diameter. At temperatures below homogeneous
nucleation can take place. As the snow crystals fall through
The
the atmosphere, they cluster together into snowflakes'
difference between the melting point of ice and the triple
a point of water is less than 0.01'C, with a corresponding
vapor pressureof onlY 4'6 torr.
of a As discussedby Nord [30] and Kudela and Sampson[31]'
a number of chemicals are amenable to purification by
desublimation, precededperhapsby sublimation' In general'
they are chemicals that are solids at ambient temperatureand
pressure and have solid vapor pressuresgreater than 5 pm
Hg at moderateoperating temperatures.These chemicals are
listed in Table 17.10.Major applicationsof desublimation'to
obtain near-pure solid chemicals or pure-chemical solid
films on substrates,include crystallization of water-insoluble

1000.000

100.000

10.000
o-
f
q
o 1.000
o

G
0 . 10 0

0.010

0.001 1s(
100
Temperature,'C Figure 17.29 Vapor pressureof naphthalene.
680 Chapter 17 Crystallization

Table 17.1.0 Chemicals Amenable to Purification by The number, size, and shape of the crystals producedin
Desublimation the snow of the third technique depend on the relative rates
Aluminum chloride Molybdenum Trioxide of nucleation, crystal growth, and crystal agglomeration.
Anthracene Naphthalene Frequently, as commonly observed with snow crystals,den-
Anthranilic acid dritic growth occurs, producing undesirable crystals of the
B-Naphthol
Anthraquinone Phthalic anhydride shapeshown in Figure 17.30.The main stem grows rapidly,
Benzanthrone o-Phthalimide followed by slower rates of growth of primary branches,and
Benzoic acid Pyrogallol much slower growth of secondarybranches.If the dendritic
Calcium Salicylic acid crystals can be suspended in the vapor long enough, the
Camphor Sulfur spacesbetween the branches can fill in to produce a more-
Chromium chloride Terephthalic acid desirable dense shape.
Cyanuric chloride Titanium tetrachloride
Ferric chloride Thymol Desublimationin a Heat Exchanger
Hafnium tetrachloride Uranium hexafluoride
Iodine Zirconium tetrachloride This technique is used in industry to recover, from a gasmix-
Magnesium ture, a number of organic chemicals, including anthracene,
maleic anhydride, naphthalene,phthalic anhydride, and sali-
cylic acid. The crystals are deposited on the outside of the
tubes while a coolant flows through the inside of the tubes.If
organic chemicals from mixtures with inert, nonvolatile the rate of desublimation is controlled by the rate of conduc-
gasesand vapor-deposition of metals.
tion of heat through the deposited crystal layer, then a
Desublimation is almost always effected by cooling the relationship for the time to deposit a crystal layer of given
gas mixture at nearly constant pressure.As discussedby
thickness is derived in a manner similar to that for the
Holden and Bryant [32], cooling may be achieved by any of falling-film melt crystallizer in Section 17.8. The result,
the following techniques: which is similar to (17-67), except that the crystal layer
1. Heat transfer from the gas through a solid surface, on grows outward from the outside radius of the cylindrical
which the crystals form. tube, ro, insteadofinward from radius, riin(17-61),is
2. Quenching by addition of a vaporizable liquid.
3. Quenching by addition of a cold, noncondensablegas. ,:ffayl|"(:) - 7\ri
l, " - ".1
r;) | (r7-15)
4. Expansion of the gas mixture through a\ozzle.
where /,sis the radius to the crystal layer-gasinterface. Equa-
The first technique, which is widely used, has the disad-
tion (17-75) ignoresany sensibleheat associatedwith cool-
vantage that the crystals must be removed by scraping of or
ing of the gas. If the sensibleheat is not negligible, then it
by melting from the heat-transfer surface. The second tech-
must be addedin (17-15) to the heat of sublimation,AI1,. If
nique, which is less common, must use a liquid in which the
the heat-transferresistancein the gas phaseis negligible, the
sublimateis not soluble. If excesswater is used, this tech-
temperature, Tr, in (17-75) corresponds to the temperature
nique will produce a slurry of the crystals and nonvaporized
where the partial pressureof the desubliming solute equals
liquid. The third technique, which is also common, produces
its vapor pressure.The derivation of (17-75) including the
dry crystals,sometimescalled a snow.The fourth techniqueis
sensible-heat correction is given in detail by Singh and
much less cornmon, becausethe degreeof cooling necessary
Tawney [33].
may require either a high pressureupstreamofthenozzle or a
vacuum downstream.

Figure 17.30 Dendritic growth of a crystal.
EXAMPLE 17.15
A desublimationunit of the heat-exchanger type is to be sizedfor
the recoveryof 100 kg/h of naphthalene(N) from a gas stream,
wherethe othercomponentsarenoncondensable underthe condi-
tions ofoperation.The heat-exchanger tubesare I m long with an
outsidediameterof 2.5 cm. Tubespacingis suchthat N canbuild
up to a maximumthicknessof 1.25cm. The gasentersthe unit at
800ton and80'C with a molefractionfor N of 0.0095.Thecoolant
is coolingwater,which flowsthroughthe insideof thetubes,coun-
tercurrentlyto the flow of gas.The inlet andoutlettemperatures of
the cooling waterare 25"C and 45'C, respectively.Determinethe
numberof tubesneededandthetimerequiredto reachthemaximum
thickness,if thegasleavesat 60oC.
 17.10 Evaporation 681

SOLATION Evaporatorsdiffer in configuration and the degreeof agita-

tion of the liquid phase.The five most widely usedcontinuous-
Assume the properties
flow evaporators are shown schematically in Figure 17.31.
Cp of gas :0.26 cal/g:C Their main characteristicsare as follows:
MW of naphthalene: 128.2
(a) Horizontal-tube evaporutor. This unit consists of a
k. of solid naphthalene: 1.5 caUh-cm-'C horizontal cylindrical vessel, equipped in the lower
A.F1"of naphthalene: ll5 callg section with a horizontal bundle of tubes, inside of
: 1.025g/cm3
p. of solid naphthalene which steam condensesand outside of which the so-
lution to be concentratedboils. The vapor produced
When the maximum thickness of N is achieved, the amount of sub-
leaves the surface of the liquid solution. Agitation is
limate per tube, if uniform. is
provided only by the movement of the bubbles
m:rr(r?-t)rp" formed. Therefore, this type of unit is only suitable
.25+ | .2r2 - 1.252]
: 3.14[(r 100(
1.025) for low-viscosity solutions that do not deposit scale
: 1 , 5 1 0g : 1 . 5 1k g on the heat-transfersurfaces.
The entering gas has a partial pressurefor N of 0.0095(800) : 7.6 (b) Short-vertical-tube evaporator. This unit differs sig-
ton. From Figwe 17.29, the saturation temperature for this partial nificantly from the horizontal-tube evaporator. The
pressureis79.7"C. This is just slightly below the entering-gastem-
tube bundle is arranged vertically, with the solution
perature of 80'C, which is less than the triple-point temperature.
Therefore, N will condenseas a solid. At the exit temperatureof the
gas, 60'C, the vapor pressure of N is 1.8 torr. Assuming saturation
at the exit and no pressure drop, the mole fraction ofN in the exit
gas will be 1.8/800 :0.00225.Thus, per mole of entering gas,
0.0073 moles of N will be condensed.If the gas is assumedto be
nitrogen and N, then M,, the ratio of the mass of gas mixture to the
mass of N condensed,will be
28(0.990s)
+ 128.2(0.0095)
M,: :30.94
(128.2X0.0073) ( a ) H o r i z o n t a lt u b e
The sensibleheat plus the heat of fusion is
M,Cp"(Tia - %u,)e+ AH" :30.94(0-26)(80 - 60) + l15
:276 caUg

Thus, for this example, with a very low mole fraction of the chem-
ical undergoing desublimation, the sensible heat effect is large.
From (17-75), the time to reach the maximum thickness of N, using
the temperature driving force acrossthe solid N of 35"C, is
(r.02s)(276)
lQ.!h(- \ t o2) -_ lrz.s'_ r.zs2tl
': /
(li)(3t | 2 4\z'r I
P
F
The numberof tubesrequiredis
(c) Long verticaltube
(100)(5.37) (d)Forced circulation
_ ?{A
l.-51

I7.IO EVAPORATION LEGEND

C = condensate
Before crystallization of a solute from an aqueous solution F = feed
takes place, it is customary to bring the concentration of the G = vent
solute close to the solubility curve. This is accomplished by P = product
evaporation ofthe solvent in an evaporator.Such a device is S = steam
also usedto concentratesolutions even when the solute is not V = vaPor
subsequentlycrystallized,e.g., solutionsof sodium hydrox-
ide. When the vapor formed is essentially pure, there is
no mass-transfer resistance in the vapor phase. When the ( e )F a l l i n gf i l m
liquid phase is agitated, mass-transferin the liquid phase is Figure 17.31. Common types of evaporators.
sufficiently rapid that the rate of evaporation of solvent can
lFrom Perry's Chemical Engineers'Handbook, 6th edition, R.H. Perry,
be determined by the rate of heat transfer from the heating D.W. Green,and J.O. Maloney, eds.,McGraw-Hill, New York (1984) with
medium, usually condensingsteam,to the solution. permission.l
682 Chapterl7 Crystallization

inside the tubes and steam condensing outside. Boil- 450

ing inside the tube causesthe solution to circulate, thus
providing additional agitation and, therefore, higher ;4 7
w
heat-transfer coefficients. Nevertheless, this type of
evaporator is not suitable for very viscous solutions. F 350
i
(c) Long-verticql-tube eveporator. By lengthening the
vertical tubes and providing a separatevapor-liquid E soo \
disengagementchamber, as shown in Figure I7.3Ic,
o //
a higher tube-enteringliquid velocity can be achieved f; zso
and, thus, an even higher heat-transfercoefficient.
b 200
(d) Forced-circulati.on evaporator. To handle very vis-
E
cous solutions, a pump is used to force the solution o
; 150
upward through relatively short tubes. =
(e) Failing-film evaporator. This unit is widely used ro o 100
concentrate heat-sensitive solutions such as fruit
juices. The solution entersat the top and flows as a film 50
down the inside walls of the tubes. The concentrate
and the vapor produced are separatedat the bottom.
0 50 100 150 200 250 300 350
For a given pressurein the vapor spaceof an evaporator, Boilingtemperatureof water,oF
the boiling temperatureof an aqueoussolution will be equal Figure 17.32 Diihring chart for aqueoussolutions of sodium
to that of pure water if the solute is not dissolved in the hydroxide.
water, but rather consists of small. insoluble. colloidal mate- [FromW.L. McCabe,J.C.Smirh,andP.Harriott,Unit Operationsof
rial. If the solute is soluble, the boiling temperature will be ChemicalEngineering,5thed.,McGraw-Hill,New York (1993)with
permission.l
greater than that of pure water by an amount known as the

Sucrose
o
oCi
acid
o Kraft rtubot u
o Glycerol
o,-
a
6 L 100
o NaNO. c
o(NH+)zSO+
o K,CO.
Ca(N03)2o =
o 66i '6
dl
o HNOg
o L iN O s 260 r
r
0)
o
qan
' - , v 240 6
o
c

o
o)60 220z
E F
3
(Il

50 200 a
NaOHo 6669

Figure 17.33 Nomograph

At 270"F a 22To CaCl, solution for boiling-point elevation
has a boiling-point rise of 9.7'F, of aqueoussolutions.
Note: Points shown o based mainly lFromPerry'sChemical
on atmospheric boiling point Engineers' Handbook,6th ed.,
R.H. Perry,D.W Green,and
J.O.Maloney,eds.,McGraw-
Hill, NewYork(1984)with
permission.l
 17.10 Evaporation 683

boiling-point elevation of the solution. If, as is usually the Evaporator Model

case,the solute has little or no vapor pressure,the evapora-
tor pressureis equal to the partial pressureofthe water in the
solution. Then, by a modifled Raoult's law,
P : P:aro: JH2otHroPIl,o (r7-76)
Thus, for a given P and xsr6, a temperature can be deter-
mined to give the required Pfr,o provided that lsrs can be
estimated.For solutions dilute in the solute, TH2oapproaches
a value of 1.0. For concentratedsolutions,the value ofTHro
can be estimated from correlations for electrolyte solutions
as discussedby Reid, Prausnitz,and Poling [34].
Alternatively, the boiling temperatureof the solution caa
be estimated by using a Diihring-line chart if it is available
for the particular solute. Such a chart is shown in Fi gwe 17.32
for the sodium hydroxide-water solution. The straight lines
on this chart for different mass fractions of NaOH obey
Dilhring's rule, which statesthat as the pressureis increased,
the boiling temperature of the solution increases linearly
with the boiling temperature of the pure solvent. A nomo-
graphfor other solutesin water is given in Figure 17.33.The
use of Figures 17.32 and 17.33, and (11-76) is illustratedin
the following example.
EXAMPLE 17.16
An aqueous,NaOH solutionis being evaporatedat 6 psia. If the
solutionis 35-wt%aNaOH,determine:
(a) The boiling temperature
of the solution.
(b) The boiling-pointelevation.
(c) The activity coefficientfor waterfrom (77-76\.
The following mathematical model is widely used to make
material balance, energy balance, and heat-transferrate cal-
culations to size evaporators operating under continuous-
flow, steady-state conditions. The model is based on the
schematicdiagram in Figure 17.34.A so-called thin liquor at
temperature, 7y, with weight-fraction solute, w7, is fed to
the evaporator at mass flow rate, my. Aheatingmedium, e.g.,
saturated steam, is fed to the heat-exchangertubes at tem-
perature and pressure, Z" and P", and mass flow rate, ms.
Saturatedcondensateleaves the heat exchanger at the same
temperatureand pressure.Heat-transferrate,e,tothe solution
in the evaporator causes the solution in the evaporator at
temperature, 7", to partially evaporateto produce a vapor at
temperature, Zr, with flow rate, mu. Tlte so-called thick-
liquor concentrate leaves at temperature, Tp, with weight-
fraction solute, l?p,at massflow rate, mo.Theheat exchanger
has a heat-transfer area, A, and overall heat-transfer coeffi-
cient, U, basedon that area.
Key assumptionsin formulating the mathematical model
are:
1. The thinJiquor feed has only one volatile componenr,
e.9., water.
2. Only the latent heat of the heating steam at Z" is avail-
able for heating and vaporizing the solution in the
evaporator.

SOLUTION
Fromsteamtables,purewaterhasa vaporpressure of6 psiaat 170.F.
(a) From the DiihringJinechartof Figure 17.32,theboiling tem-
peratureof the solutionis 207.F.
(b) The boiling-pointelevationis: Vaporspace
pressure,
P
2 0 7- 1 7 0 : 3 7 ' F Thin-liquor feed
Alternatively,the nomographof Figure 17.33maybe usedby
drawing a straightline through the point for NaOH and a
solution (boiling-point)temperatureof 207"F. That line is
extendedto theleft to theintersectionwith theleftmostvertical
Iine.A straightline is thendrawnfrom that intersectionpoint Heatingsteam
throughthe lower,inclinedline labeledweightpercentsolids ated vapor)
at 35 (wtvo NaOH). This secondline is extendedso as to
intersectthe right, verticalline. The valueofthe boiling-point
elevationat this point of interactionis readas 36"F,which is
closeto the valueof 37'F from the Dtihringchart.
(c) The mole fraction of water in the solutionis, for complete
ionizationof NaOH:
0.65l18
xH2o: :0.674
0 . 6 5 1 1+82 ( 0 . 3 s ) 1 4 0
Vaporpressureof purewaterat 207"F:13.3 psia.From (lj -76),
To''o' Ho
6
TH.^: :n^, Figure 17.34 Continuous-flow, steady-statemodel for an
0 . 6 7 4 (1 3 . - 3 )
evaporator.
684 Chapter 17 Crystallization

%
zz
'ry z
-/r

!
f
fi-.
350

300
\
-l-.
.-1. i:i -
400'F
:==i
,*
4
-zz 'z
'/
% 7
E ::: ^----
o
a ::: :::
55U -

tt=t ,ry,
='-'-
d zso - * :
::=
:ii
:== 300,:
==7=
'2 ,ry 7
=2=z' z v

=q 2oo
=
c
-
1bo
254- -

zAO
=
7
ryv z Figure 17.35 Enthalpy-

ffi V"i t'd sol concentration

diagramfor
100 - 50 sodiumhydroxide-water
system.
50
: = [From W.L. McCabe, J.C.

0.1 0.2
v 0.4 c 0.5 0.6
Smith, and P. HarrioIt, Unit
Operations of Chemical
Engine ering, 5th ed., McGrar,"-
Hill, New York (1993) with
Mass fraction of NaOH permission.l

3. The boiling action on the heat-exchangersurfacesagi- Heat-transfer rate:

tates the solution, in the evaporator, sufficiently to
achieve perfect mixing. Therefore, the temperatureof
the solution in the evaporator equals the exiting tem-
peratureof the thick-liquor concentrate.Thus, T" : fo.
Also,7, : 4.
4. Becauseof Assumptions 2 and 3, the overall tempera-
ture driving force for heat transfer : AZ : T, - Tp.
5. The AZ is high enough to achieve nucleate boiling
and not so high as to causefilm boiling.
6. The exiting vapor temperature.T,: Tp - Tr, corre-
sponds to evaporator vapor-spacepressure,p, taking
into account the boiling-point elevation of the solution
unlessthe solute is small, insoluble particles,such as
colloidal matter.
7. No heat loss from the evaporator.
Q:UA(7,-Tp) (17-81)
The procedure used to solve these five equations depends
upon the specificationsof the problem. The following exam-
ple is just one illustration. The solution of the energy-
balance equations is greatly facilitated if an enthalpy-
concentration diagram is available for the solute-solvent
system.A diagram of this type for the NaOH-water system
is given in Figure 17.35.Inthis diagram,the enthalpydatum
for water is the pure liquid at 32'fl so that the diagram has
the same datum as the steam tables. For NaOH, the datum
is NaOH at infinite dilution in warer at 2C C (68.F). Fig-
we I7.35, togetherwith the Diihring chart of Figure 17.32,
were first preparedby McCabe [35].

Based on the above assumptions,the mathematical model is EXAMPI,E 1?.17

as follows: An existingforced-circulation evaporatoris to be usedto concen-
trate44 wtToNaOH to 65 wt%oNaOH using steamat 3 atm gage
Total massbalance:
(barometerreadsI atm). The feed temperaturewill be 40.C and
mf:mp+mu (r7-tt) the pressurein the vaporspaceof the evaporatorwill be 2.0 psia.
The evaporatorhas a heat-transferareaof 232 m2.The overall
Mass balanceon the solute:
heat-transfercoefficientfor the givenconditionsis estimatedto be
wfmf :Wpmp (n-78) 2,000Wm2-'C. The densiryof the feedsolutionis 1,450kg/m3.
Neglectheatlossesfrom the evaporator. Determine:
Energy (enthalpy) balance on the solution:
(a) The temperature of the solutionin the evaporatorin oC.
Q : mrH, * moHo - mf Hf (r'7-'/9)
(b) The heatingsteamrequirementin kg/h.
where H : enthalpyper unit mass
(c) The m3/hof feed solutionthat can be sentto the evaporator.
Energy (enthalpy) balance on the heating steam: (d) Thekg/hof concentrated
NaOHsolutionleavingtheevaporator.
Q = mrLHvaP (17-80) (e) The rateofevaporationin kg/h.
 l7.lO Evaporation 685

SOLUTION From(17-78),

(a) At 2.0 psia, the boiling temperature of pure water is 126'F. 0A4m7:O'65mP (2)
From Figure 17.32, for 65 wt%oNaOH, the solution boiling Substituting(2) into (1) to eliminatemp,we obtal.n
point is 240"F or 116'C. This is a considerableboiling-point
44 qOOOOO
elevationof 114'F.
"'r- 1 0.44\ f0.44 I
(b) In American engineering units, I r - i- - ltr ,r oa) + li- :t:+o.l-lr sI
\ 0.65/ L0.65 )
A :232/0.0929 :2.49'7 ft2
: 91,170lb/h
U : 2,000/ 5.674 : 352.5 Btu/h-ft2-'F
T, :291"F for 4 atm saturatedsteam or
The driving force for heat transfer is : 41,350kg/h
my : (91,110X0;4536)
LT :7, -7,:291 -240:51"F (d) From(2),
From (17-81), 0.44
mo : : 27,990keth
*r@1,350)
Q : U A LT :352.5(2,497)(5r): 44,900,000 Bttt/h (e) rnr : mf - ffip : 41,350- 27,990: 13,360kg/h
The heat of vaporization of steam at 291"F is 917 Btti/lb.
From (17-80),
Multiple-Effect Evaporator Systems
m, : - 44,900,0001917: 48,950 lblh
Q/Ll/vap
When condensing steam is used to evaporate water from
or
an aqueoussolution, the heat of condensationof the higher-
: 22,200ks/h
0 : 48,9s0(0.4536) temperaturecondensing steamis less than the heat of vapor-
It should be noted that the heat flux is 44,900,000/2,49'l : ization of the lower-temperature boiling water. Conse-
18,000 Btu/h-ft2. This heat flux is safely in the nucleate-boiling quently, less than 1 kilogram of vapor is produced per
region. kilogram condensationof heating steam.This ratio is called
(c) From (17-79), using Figure 17.35 for NaOH solutions and the the economy.In Example 17.17, the economy is 13,360/
steam tables for water vapor, the energy balance on the 22,200 :0.602 or 6O.2Vo. To reduce the amount of steam
solution is as follows, where the evaporated water is required and, thereby,increasethe economy, a seriesofevap-
superheatedsteam at 240oF and 2.0 psia., andmu - mf - mp: orators, called effects,can be used as shown in Figure 17.36.
44,900,000 : (mt - m)(1168)-l mo(340) - m7(ll5) (1) The increasedeconomy is achieved by operating the effects

Condensate

Concentrate Concentrate Concentrated

from first from second product
effect effect
(a)Forward-feed,triple-effect

Figure 17.36 Multiple-effect

(b) Backward-feed, triple-effect evaporator systems.
686 Chapter17 Crystallization

at different pressures,and thus at different boiling tempera- If the heat-transferareasof each of the three effects are to be equal,
tures, so that vapor produced in one effect can be condensed determine:
to supply the heat in another effect. Evaporation temperatures,Ty andT2, in the first two effects.
In Figure l1 .36a, referred to as a forward-feed, triple-
Heating steam flow rate, ms.
effect-evaporalor system, approximately one-third of the
total evaporation occurs in each effect. The fresh feed solu- Solution flow rates, mr, m2, and,m3leaving the three effects.
tion and steam both enter the first effect, which operatesat
pressure P1. The concentrate from the first effect is sent to SOLUTION
the second effect. The vapor produced in the first effect is The unknowns,which number seven,are A(: At : Az : A),Tr
also sent to the second effect where it condenses,giving up 72, tr1,,m; m2, and n3. Therefore, seven independentequationsare
its heat of condensation to cause additional evaporation of needed. Because the solute is colloids (insolubles), boiling-point
the solution. To achieve a temperature-driving force for heat elevations do not occur.
transfer in the second effect, the pressure of the second It is convenient to add three additional unknowns and, there-
effect, P2, is lower than that of the first effect. This procedure fore, three additional equations,making a total of l0 equations.The
is repeatedin the third effect. For three effects, the flow rate added unknowns are the heat-transfer rates, Q1, Qz, and Q3 in the
three effects. The l0 equations, which are similar to (17-78) to
of steam entering the first is only about one-third of the
(17-81),are
amount of steam that would be required if only one effect
were used. However, the temperature-driving force in each Overall colloid mass balance
of the three effects is only about one-third of that in a single (1)
effect. Therefore, the heat transfer area of each of the three Energy
balances "*t
evaporators in a triple-effect system is approximately the "'.n",Jiff:
sameas for the one evaporator in a single-effect unit. There- et : (mf _ mr)Hr, * mlHy _ mf HJ (2)
fore, the savings in annual heating-steamcost must offset the e z : ( m r _ m z ) H r , * m 2 H 2_ m r H r (3)
additional capital investment for equipment. Qt : (mz - mz)Hr, I m3H3 _ mzHz (4)
When the temperature of the fresh feed is significantly Energy balanceson steam and water vapors
below its saturation temperature corresponding to the pres-
sure in the first efTect,backward-feed operation is desirable, Qt : m,LHI^P (5)
as shown in Figure 1"1 .36b.The cold fresh feed is sentto the ez: (mf - m)A,HluP (6)
third effect, which operatesat the lowest pressureand, there- ez:(mr-m)A,HluP (1)
fore, the lowest temperature. Unlike the forward-feed sys- Heat-transfer rates
tem, pumps are required to move the concentrate from one
Q r : U r A r ( T '- T r ) : U t A r L ' T r (8)
effect to the next because Pr > Pz > P3. However, unlike
gas compressors,liquid pumps are not high-cost items. Qz: UzAz(Tr- T) : UzAzA'Tz (e)
Calculations for multieffect evaporator systems involve Qz : UzAz(Tz- Tt) : U3A3 A'73 (10)
the same types of mass-balance,energy-balance,and heat- From (1),
transfer equations as for a single-effect system. These equa- m3 : (w y f w 3)mf : (0.08/ 0.45)(44,090) : 7, 838 lb/tr
tions are usually solved by an iterative method, especially
Also, the flow rate of colloids in the feed is (0.08X44,090) :
when boiling-point elevations occur. The particular iterative 3,527lbth.
procedure depends on the problem specifications. The fol-
lowing example illustrates a typical procedure. Initial estimatesof solution temperature in each effect:
With no boiling-point elevation, the temperature of the solution in
the third effect is the saturation temperature of water at the speci-
EXAMPLE 17.18 fied pressureof 1.94 psia or 125'F. The temperature of the heating
steam entering the first effect is the saturation temperatureof 249.F
A feedof 44,090lb/h of an aqueoussolutioncontaining8 wt%ocol- at 29.3 psia. Thus, if only one effect were used, the temperaturedri-
loids is to be concentrated to 45 wtTocolloids in a triple-effect- ving force for heat transfer, AT, would be 249 - 125 : 124"F.
evaporatorsystemusing forward feed. The feed entersthe first With no boiling-point elevations, the AZs for the three effects must
effect at 125'F, and the third effect operatesat 1.94psia in the sum to the value for one effect. Thus,
vaporspace.Freshsaturatedsteamat29.3 psiais usedfor heating
in the first effect.The specificheatof the colloidscanbe assumed LTr* LTzl A'73:124"P (11)
constantat 0.48 Btu/lb-'F. Overall heat-trahsfercoefficientsare As a first approximation, assumethat the AZs for the three effects
estimatedto be: are, using (8)-(10), inversely proportional to rhe given values of
Uy U2, and U:. Thus,
Effect 4 Btu/h-ft2-'F
LT,:!1LT^
"Ut (r2)
I 350
2 420
3 490
tr -":91tn
Uz (1 3 )
 17.10 Evaporation 687

Solving(11),(12),and(13),we obtain Corresponding

valuesof colloid massfractionsare
AIr :48.6'F
wr :0.108
LT2: !Q.$oP
wz:0.172
A7: :34.8'F
It may be notedthatthesevaluesof m1,m2,tu1,and1r2arecloseto
Tr : T, - LTt :249 - 48.6: 200.4.F
the first approximations.
This is oftenthe case.
Tz : Tt - LT2 : 200.4- 40.6: 159.8.F
Step2
Tt : Tz - LT3 : 159.8- 34.8: 125.F
Using the valuescomputedin Step l, valuesof Q aredetermined
Initial Estimatesof m1and m2: from (5), (6), and(7); andvaluesofA aredeterminedfrom (8), (9),
The total evaporationrate for the three effects is mf - m3 : and(10).
44,090- 7,838:36,252 lb/h.Assume,as a first approximarion,
equalamountsof vaporproducedin eacheffect.Then, h : 15,070(946.2)
: 14,260,000Btu/h
: -
Q2 @4,090 32,770)(977.6) : 11,070,000Btu/h
mf - mt :36,252/3 : l2,O84lblh
Q3 : Q2,770- 20,500X1,002.6)
: 12,400,000
Btu/h
mt : 44,090- 12,084:32,0061b1h
mz : 32,006- 12,084: 19,922lblh 14.260.000
------:- -R?Rffz
Al:,
mz : 19,922- 12,084: 7,838lb/h (350)(48.6)

Corresponding
estimatesof the massfractionsof colloidsare I 1.070.000- 6 4 g f t '
A2:,
(420X40.6)
w t : 3 , 5 2 7/ 3 2 , 0 0 6 : 0.110
I 2.400.000- 1 1 ' 7 r + z
w z : 3 , 5 2 7 / 1 9 , 9 2 2: 0 . 1 7 7 h:,
r490x34.8)
w3 :3,527 /7,838: 0.450(given)
The remainingcalculationsare iterativein natureto obtain cor- Step3
rectedvaluesof Ty 72,m1,and,m2,as well as valuesof A, m,, Q1, Becausethe three areasare not equal,calculatethe arithmetic-
Qz, wtd Q3.Thesecalculationsare best carriedout on a spread- average,heat-transfer areaanda new setof Af driving forcesfrom
sheet.Eachiterationconsistsof the following steps: (8), (9), and (10).Normalizethesevaluesso rheysumro the over-
all LT (124"Fin this example).From the AZ values,computeZt
StepI
andT2.
Combine(2) through(7) to eliminateQr Qz, andQ3.Usingthe ap-
proximationsfor 71,72,73,wy andw2,the speciflcenthalpiesfor 838+ 649+727 -1^5 6^I-t -)
,A. a- v g :
the resultingequationsarecalculatedandthe equationsaresolved
for new approximations of ms,m1,dndm2.Corresponding
approxi-
mationsfor wl andw2arecomputed. 67,, : !'ho,loo : --'-
-s-s.2.F
-
For the first iteration,the enthalpyvaluesare (3s0x738)
ll'o7o'ooo-3s.7'F
Al?"uuP:946.2Btunb ar' - --' -
(420)(738)
LH;^P : 977.6Btunb
LH;"P:1,002.6 Btu/lb l2'4oo'ooo
an. - : --r4.3'F
'- -
Hq: l,746Btu/Ib (490X738)
H,, - 1,130Btunb Thesevaluessum to 125.2'F.Therefore.they are normalizedto
F1,.: 1,116Btu4b
Hr :0.92(92.9) + 0.08(0.48X125 - 32) : 39.s3,r7tt A,T1:55.21124/125.2): 54.7"F
flr : 0.89(168.4)+ 0.110(0.48)(200.4- 32) : 158.8Btu/lb A'72: 35.71124
/ 125.2): 35.3"F
Hz : 0.823(127 .7)+ 0.177(0.48)(159.8- 32) : 116.0Btu/lb A ' \ : 3 4 . 3 1 1 2 4 / r z s . 2: ) 3 4 . 0 ' F
Hz :0.55(92.9)+ 0.45(0.48X125 - 32) :71.291u11O
Tr :249 - 54.7: 194.3'F
When these enthalpy values are substitutedinto the combined Tz: 194.3- 35.3: 159.0.F
energybalances,the following equationsareobtained:
Steps1 through3 arenow repeatedusingthe new valuesof T1and
ms :49,250 - 1.043mt (14)
?2 from Step3 andthe new valuesof w1 dndw2from Stepl. The
4 4 , 0 9 0 : 1 . 9 9 4m 1- 1 . 0 3 7
m2 (1s) iterationsarecontinueduntil the valuesof the unknownsno longer
mz - mr
8,168: 1.99'7 (16) changesignificantlyandAl : Az : Az.The subsequent iterations
for this exampleareleft asan exercise.Basedon the resultsof the
Solving(14),(15),and(16),
first iteration,theeconomyof thethree-effectsystemis
m" : l5,07Olblt'
mr :32,77OIblh 44,090- 7,838
:2.41 or 24l%o
mz:20,500lb/h 15,070
688 Chapter17 Crystallization
Overall Heat-Tiansfer Coefficients in Evaporators
In an evaporator,the overall heat-transfercoefficient, U, de-
pendsmainly on the steamside,condenseningcoefficient, the
solution-sidecoefficient,and a scaleor fouling resistanceon
the solution side. The conduction resistanceof the metal wall
ofthe heat-exchanger tubesis usually negligible.Steamcon-
densationis generally of the film, rather than dropwise, type.
When boiling occurs on the surfaces of the heat-exchanger
tubes, it is of the nucleate-boiling, rather than film-boiling,
regime. In the absenceof boiling on the tube surfaces,heat
transfer is by forced convection to the solution. Local coeffi-
cients for film coridensation, nucleate boiling, and forced
convection ofaqueous solutions are all relatively large, ofthe
orderof 1,000Btu/h-ft2-'F (5,700Wm2-K). Thus,rheoverall
coefficient would be about one-half of this. However, when
fouling occurs, the overall coefficient can be substantially
SUMMARY
1. Crystallization involves the formation of solid crystalline par-
ticles from a homogeneous fluid phase. However, if the fluid is a
gas, the process is usually referred to as desublimation.
2. In crystalline solids, as opposed to amorphous solids, the
molecules, atoms, and./orions are arranged in a regular lattice pat-
tern. When crystals grow unhindered, they form polyhedrons with
flat sides and sharp corners. Although the faces of a crystal may
grow at different rates, referred to as crystal habit, the Law of Con-
stant Interfacial Angles restricts the angles between corresponding
faces to be constant. Crystals can form only seven different crystal
systems, which include 14 different space lattices. Because of
crystal habit, a given crystal system can take on different shapes,
e.g.,plates,needles,and prisms,but not spheres.
3. Crystal-size distributions can be determined or formulated in
terms of differential or cumulative analyses,which are convertible,
one from the other. A number of different, mean-particle sizes can
be derived from crystal-size distribution data.
4. The most important thermodynamic properties for crystalliza-
tion calculations are melting point, solubility, heat of fusion, heat of
crystallization, heat of solution, heat of transition, and supersatura-
tion. Solubilities of inorganic salts in water can vary widely from a
negligible value to a concentration of greater than 50 wtVa. Many
salts crystallize in hydrated forms, with the number of waters of
crystallization of the stable hydrate depending upon temperature.
The solubility of sparingly soluble compounds is usually expressed
in terms of a solubility product. When available, phase diagrams
and enthalpy-concentrationdiagrams are extremely useful for mak-
ing material- and energy-balancecalculations.
5. Crystals smaller in size than can be seen by the naked eye
(<20mm) are more soluble than the normally listed solubility.
Supersaturation ratio for a given crystal size is the ratio of the
actual solubility of a small-size crystal to the solubility of larger
crystals that can be seen by the naked eye. The driving force for
nucleation and growth of crystals is supersaturation.
6. Primary nucleation, which requires a high degree of supersatu-
ration, occurs in systems free of crystalline surfaces, and can be
homogeneous or heterogeneous. Secondary nucleation occurs
Table 17.1I Typical Heat-Transfer Coefficients in Evaporators
II
Type Evaporator Btu/h-ft2-'F Wmz-K
Horizontal-tube 200-500 1,100-2,800
Short-tube vertical 200-500 I,100-2,800
Long-tube vertical 200-700 1,100-3,900
Forced circulation 400-2,000 2,300-l1,300
less.Table 17.11,takenfrom Geankoplist361,lists rangesof
overall heat-transfercoefficients for different types ofevapo-
rators. The higher coefficients in forced-circulation evapora-
tors are mainly a consequenceof the greatly reduced fouling
due to the high liquid velociry in rhe tubes.
when crystalline surfaces are present. Crystal growth involves the
mass transfer ofthe solute up to the crystal surface followed by in-
corporation of the solute into the crystal-lattice structure.
7. Equipment for solution crystallization can be classified accord-
ing to operation mode (batch or continuous), method for achieving
supersaturation(booling or evaporation), and featuresfor achieving
desired crystal growth (e.g., agitation, baffles, circulation, and clas-
sification). Of primary importance is the effect of temperature on
solubility. Three of the most widely used types of equipment for so-
lution crystallization are: (1) batch crystallizer with external or in-
ternal circulation, (2) continuous, cooling crystallizer, and (3) con-
tinuous, vacuum, evaporating crystallizer.
8. The MSMPR crystallization model is widely used to simulare
the often used continuous, vacuum, evaporating draft-tube, baffled
crystallizer. Some of the assumptionsof this model are perfect mix-
ing of the magma, no classification of crystals, uniform degree of
supersaturation throughout the magma, crystal growth rate inde-
pendent of crystal size, no crystals in the feed, no crystal breakage,
uniform temperature, equilibrium in product magma between
mother liquor and crystals, constant and uniform nucleation rate
due to secondarynucleation by crystal contact; uniform crystal-size
distribution, and all crystals with the same shape.
9, For a specified crystallizer feed, magma density, magma resi-
dence time, and predominant crystal size, the MSMPR model can
predict the required nucleation rate and crystal-growth rate, num-
ber of crystals produced per unit time, and crystal-size distribution.
10. Precipitation, leading to very small crystals, occurs with
solutes that are only sparingly soluble. The precipitate is often pro-
duced by reactive crystallization from the addition of two soluble
salt solutions, producing one soluble and one insoluble salt. Unlike
solution crystallization, which takes place at a low degree of super-
saturation, precipitation occurs at a very high supersaturationthat
results in very small crystals.
ll. When both components of a mixture can be melted at reason-
able temperatures (e.g., certain mixtures of organic compounds),
melt crystallization can be used to separatethe components. If the

components form a eutectic mixture, pure crystals of one of the
componentscan be formed. If the componentsform a solid solution,
repeatedstagesof melting and crystallization are required to achieve
high purity.
12. A large number of crystallizer designs have been proposed for
melt crystallization. The two major methods are suspensioncrys-
tallization and layer crystallization. Of particular importance is the
falling-film crystallizer, which can be designed and operated for
high production rates when the components form eutectic mixtures.
For components that form solid solutions, the zone-melting tech-
nique developed by Pfann can be employed to produce nearly pure
compounds.
13. A number of chemicals are amenable to purification by
desublimation, preceded perhaps by sublimation. Desublimation is
almost always achieved by cooling the gas mixture at constant
pressure. The cooling can be accomplished by heat transfer,
quenching with a vaporizable liquid, or quenching with a cold, non-
condensiblegas.
14. Evaporation can be used to concentratea solute prior to solu-
tion crystallization. Common evaporators include the horizontal-
REFERENCES
1. MULLIN, 1.W' Crystallization, 3rd ed., Butterworth-Heinemann,
B o s t o n( 1 9 9 3 ) .
2. GRABER,S., T.A., and M.E., TABoADA,M' Chem. Eng. Ed., 25,
1 0 2 - 1 0 5( 1 9 9 1 ) .
3. HoucpN, O.A., K.M. WlrsoN, and R.H., Recerz, Chemical Process
Principles, Part I, Material and Energy Balances, 2nd ed., John Wiley &
S o n s( 1 9 5 4 ) .
4. MIEns, H.A., and F. Iseec, Proc. Roy. Soc., 479,322-351 (1907).
5. NIELsEN,A.E., Kinetics of Precipitation, Pergamon Press, New York
( I 964).
6. Noves, A.A., and W.R. WHrrNEy, J. Am. Chem. Soc., 19,930-934
( 18 9 7 ) .
7. NERNST,W' Zeit. fiir Physik. Chem., 47,52-55 (1904).
8. MIERS,H.4., Phil. Trans., 4202,492-515 (1904).
9. VereroN, I J.P., Zeit. fiir Kristallographie, 59, 483 (1924).
10. Mvensott, A.S., Ed., Handbook of Industrial Crystallization,
Butterworth-Heinemann, Boston ( 1993).
11. BunroN, W.K., N. Cnennnl., and F.C. FnnNr, Plrjl. Trans., 4243,
2 9 9 - 3 5 8 ( 1 9 51 ) .
12. SEAvoy, G.E., and H.B. CALDwELL,Ind. Eng. Chem., 32,627-636
(1940).
13. NowrvtnN,H.H., and R.C. BsNNerr, Chem.Eng. Prog., 55 (3), 65-70
(19s9).
14. RANDoLIH,A.D., AIChE Journal, 1I,424430 (1965).
15. RANDoLIH,A.D., and M.A. LnnsoN, Theory of Particulate Processes,
2nd ed.,Academic Press,New York (1988).
16. McCABE,W.L., Ind. Eng. Chem.2l,30-33 and ll2-19 (1929).
1'1. Zuusrrru, R.C., and R..W.RoussEAU,AIChE Symp. Ser,83 (253),
1 3 0( 1 9 8 1 .
18. NTELSEN, A.8., Kinetics of Precipitation, Pergamon Press, Oxford,
England (1964).
References 689
tube unit, short-vertical-tube unit, long-vertical-tube unit, forced-
circulation unit, and the falling-film unit. For a given evaporation
pressure, the presence of the solution can cause a boiling-point
elevation.
15. The most widely used evaporator model assumes that the
liquor being evaporated is well-mixed such that the temperature
and solute concentration are uniform and at exiting conditions.
16. The economy ofan evaporator is defined as the mass ratio of
water evaporated to heating steam required. The economy can be
increased by using multiple evaporator effects that operate at dif-
ferent pressuressuch that vapor produced in one effect can be used
as heating steam in a subsequenteffect. The solution being evapo-
rated can progress through the effects in forward, backward, or
mixed directions.
17. Evaporators typically operate so that solutions are in the nu-
cleate-boiling regime. Overall, heat-transfer coefficients are gener-
ally high becauseboiling occurs on one side and condensation on
the other side.
19. NmrsnN, A.E., Chapter 27 in Treatise on Analytical Chemistry, Part 1,
Volume 3, 2nd ed., Editors I.M. Kolthoff and P.J. Elving, John Wiley &
Sons,New York (1983).
20. NELSEN,A.E., J. Crys. Gr.,67,289-310 (1984).
21. FrrcHETr, D.E., and J.M. TARBELL,AICLE J., 36, 511-522 (I99O).
22. Mrrsuorn, M., M. OnsHr, A. SurraruNr, and K. Osonr, World
Congress III of Chemical Engineers, Tokyo, Sept. 21-325, 1986,
pp. 980-983.
23. WLcox, W.R., /nd Eng. Chem.,60 (3), 13-23 (1968).
24. WYNN, N., Chemical Engineering,9S (7) 149-154 (1991).
25. Pru.NN,W.G., Trans. AIME., 194,'747 (1952).
26. PFANN,WG., Zone Melting,2nd ed., John Wiley and Sons, New York
(1966).
27. ZrEF,M., and W.R. Wtcox, Fractional Solidification, Marcel Dekker,
New York (1967).
28. BuRRrs,JR.,L., C.H. SrocrvrN, and I.G. DrLLoN,Trans.AIME,203,
1 0 1 7( 1 9 5 5 ) .
29. HERINGToN,E.F.G., Zone Melting of Organic Compounds,John Wiley
& Sons,New York (1963).
3 0 . N o n o , M . , C h e m .E n g . , 5 8 ( 9 ) , 1 5 7 - 1 6 6( 1 9 5 1 ) .
31. Chem. Eng.,93 (12),93-98 (1986).
Kunslr, L., and M.J. SerrapsoN,
32. HoLDEN,C.A., and H.S. BRyANr,Sep. Sci.,4 (1), 1-13 (1969).
33. SNGH,N.M., and R.K. TAwNEy,Ind. J. Tech.,9, M5447 (1971).
34. PoLNG, B.E., J.M. PRAUSNITz, and J.P.O'CoNNELL,The Properties of
Gases and Liquids, 5th ed., McGraw-Hill Book Co., New York (2001),
p.8.191.
35. McCess, W.L., Trans. AIC4E, 3-1,,129-164 (1935).
36. GnlNropr-rs, C.J., Transport Processesand Unit Operations, 3rd ed.,
Prentice Hall, Englewood Cliffs, NJ (1993).