3-246 PREOICTlON ANO CORRELATION OF PHYSICAL PROPERTIES

hígh-pressure ~2 data, and average errors of 1.0 lo 3.2 per cent where, in turn, Ze is obtained from Eq. (3-19) or (3-20) or from
for the other díatomics. For the 11 gases tabulated below the chart, Reid' s suggestion: 167
k may be calculated directly al any T and P within its range. For
other gases, ke may be obtained if one experimental value of k is Zc = 0.291 - O.08Ow (3-116)
Imown. In the absence of any data. kc may be approximated with Alternatively (and preferably), a e may be predicted with Miller's
Eq. (3-22). eguation: 168

VAPOR PRESSURE ac = 0.9076 (1 + To• In Pe) (3-117)

1 ­ T),r
If only !he normal boiling point and the critical temperature and
in whieh Pe is in atm. abs. Other prediction methods for vapor
pressure are Imown, a linear two-point plot of In Pe vs. liT, is often
pressure are discussed in detail by Reid and Sherwood. 122
sufficiently accurate. Alternatively, one may plot P" of the com­
pound vs. Pe as a reference substance (for which vapor-pressure data
are known) at the same T, as a straight line on log-Iog paper with VISCOSITY
good accuracy.85
The best of the simpler (three constant) vapor-pressure correla­ 1. Liquids. Except for the first members of homologolls series
tion equations is probably Antoine's86 (and quantum liquids), the viscosity of most pure liquids, organic
and inorganie, at their normal boiling points is -0.30 centipoise,
B within average and maximum deviations of ::t:25 and 70 per cent.
10gP. = A (3-105)
Arrhenius's relation,91 as modified below (Gambill, unpublished),
permit5 more aceurate predietion:
where A, B, and e are constanls for a particular 5ubstance. lf only
two boiling points are known, let e 230 and evaluate A and B 1'. = 0.324pl{2 (3-118)
from
where the units of l' and pare centipoise and g./ee. Equation
B (3-llB) gave average and maximum deviations of 17 and 39 per cent
(3-106) for 34 liquids (organic and inorganic), and represents a signifil'ant
improvement over the original proposal,91 in which the eoefficient
and B=' ~.
(3-107) was 0.275. As noted by Bretsznajder,133 the Arrhenius equa­
tiOlí is almast as accurate as other more complicated methods. lf
wherein uníts of mm. Hg and K. are most often u5ed. One
Q
the thermal eonductivity i5 known, estímates of l' may be made
Antoine equation will accurately fit vapor-pressure data from TI to for simple non-polar Iiquids within -::t:23 per cent with McLaugh­
Tr = 0.85 to 0.95. Accuracy is best when all three constants are Jin's dimensionless ratio,156 given by Eq. (3-84).
At temperatures other than tb' Iiqnid viscosity is probably best
evaluated from data; when Eqs. (3-106) and (3-107) are used with
two boiling points, the method loses aCCllracy for T. 2500K. or < estimated with Thomas's equation92
for highly polar compounds at T, 0.65. 33 l' 0.1I67PV21OY (3-119)
Within a maximum error of per eent for most eompounds
where
(but not for NHa or H 20, for example), Pv may be predicted with
Riedel's analytical eorrelation 87
y (3-120)
log ( Pe)
Pv
= 0.118B - 7 log Tr + (a e - 7)(O.0364B log Tr )
and where l' is in eentípoise, PI i5 in g,jce., and B is a' viseosity
(3-108) constant to be ealeulated by summation of the atomic and group
eontributions of Table 3~'321. Equation (3-119) i5 most valid when
36
where B = - - 35 (T,)6 + 42 In T (3-109) t :s:; lo and the ealculated vi8Cosity i5 less than about 15 cp. Average
T, ' and maximum errors of -20 and 90 per cent may be expected,
The parameter a e (the Riedel factor) is deflrd by though for many Iiquids predktions may be made within 5 per
cent. 92 ,169
d(ln Pe,) The effect of ten¡peralure on liquid visL'Osity, which is largest
ae =' d(lri T ) (at the eritical'point) (3-110)
r
for Ihe more complexly struetured Iiquids, may be correlated within
the accuracy of most experimental data (1 to 2 per eent) with the
and may in turn be calculated fr(lm 88,122 de Guzman-Andrade equation 93
a c = 5.808 + 4.93w (3-111)
l' == Ae B1T (3-121)
The acentric factor w, which usually varies from O to 0.3, is defined which requires knowledge of two or more values of l' fOf evaluation
by of the constants A and B. Equivalent to the use of Eq. (3-121) i5
w == -logPer(at T, == 0.7) 1.000 (3-112) a linear plot of log l' vS. liT. With the Andrade equation, excellent
correlation ha~ been oblained for many pure inorganic and organic
and may be estimated within -5 per cent with the relation 17 Iiquids, fus'ed sahs, liquid metals, glasses, and salt solutions. Though
3[ log Pe ] (3-113)
w == '1 (Te/T,) 1 Table 3-321. Strudllral Contribulions for Calculating B in Eq.
13-1201
where Pe is in atm. abs. Mixture w vallles are calculahle from the
Kay-type relation: Carbon. -0.462
Hydrogen. +0.249
Oxygen +0.054
"'m = ¿ J.li"'i (3-114) Chlorine +0.340
Bromine +0.326
ladine +0.3.15
The parameter a e may be estimated from Ze with Riedel's egua­ Sulfur +0.043
tion 89 C.H, . +0.385
Douhle bond +0.478
ca (ketones esters) +0.105
= 3.846 7.308 (3-115) j

(le Ze eN (cyanides) +0.381

 VISCOSITY 3-247
100,000 cent. 154 lf at least one liquid-mixture viscosity is known, the Ta­
mura-Kurata relation 98
¡Lm X¡¡Ll<1>1 + Xzl'z<1>z + 2¡L12(X,X:t!>,<1>2)1/2 (3-123)
\ relates mix.ture viscosity with composition within -7 per cent
10,000 (maximum) even when a polar component is present. The interac­
tion vÍScosity ¡L12 of Eq. (3-123) may be regarded as au adjustable
constant to be determined from known I'm data. ¡L12 is constant
only at a giveu temperature level and may be consídered to vary
1,000 \ -~
with temperature according to Eq. (3-121). lf several mixture data
points are availahle, the semi-thcoretical equatious of McAllis­
f---. .~

r---' r--- ter I22 . 172 may be applied with good results to binary and ternary
+-- systems.
I The viscositíes of ímmisci'ble liquid mixtures may be calculated
100 \ with Taylor's semitheoretical equation99

¡L.. ="1 + 2.5<j>d [~:t:_Oo4¡L.".] (3-124)

-- ~
¡Le ¡Ld + ¡Le
I I where subscripts e and d denote the continnous and dispersed
10

- _.
" .~ ~.
L phases, resp~ívely, Equation (3-124) is based on the existence of
small spherícil droplets and probably does no! apply for <1>. greater
than about 0,03. For larger values of <1>., Olney and Carlson 1OO
have suggested use of Arrhenius's equation101 for certain appli­
I cations:
1 "- ¡Lm (¡Ll)'l(¡L2)'z (3-125)
,---f-- The general subjce! of the viscosity of particulate systems is
"- reviewed in sorne detail by Happel and Brenner. '73 The ¡t of
'--' ' - - f--.
"'- 1":-:-,. Iiquid-solid suspensions may be estimated for <1>, (volume fraction
.1 sol ids) < 004, in the absence of mixture data, with the rclation of
o foc 100oC,+ 100°C-+ 100°C.-+ 100°C. + 100.C.-1 Kunitz;tOz.I03
fto. 3-61. Approximate temperature variation uf Iiquid viscosity,
¡Lm 1 + 0.5<j>,
(3-126)
-;; = (1 - <1>,)4
the error is relatively large for some highly polar Iiquids, especially
for T, > 0.8, the equation is usually very accurate between 'f¡ and or that of Mori and Ototake: 114
T~. 1.5&/>,
\Vhen only one value of ¡L is known, the temperature dependen~-e 1+--..:c.... (3-127)
0.52 - <1>.
may be obtained within -20 per cent with the generalized chart 94
showll as Fig. 3-61, which is based mainly OR data for organic Iíquids Equatíolls (3-126) and (3-127) are most accurate for <1>, ::; 0.1. Estí­
and does not apply to mercury, suspensions, or emulsions, To use mates are reasonable only when the solíds are free-Howing when
the chart, merely locate the known value of ¡t on the viscosíty scale wet (e.g., metal powders and g1ass beads). For solíds such as clay,
and then follow the mrve the necessary amount-as indicated on chalk, starch, and graphite, which are usually not free-Howing, the
the relative abseissa scale-to reach the temperature at which ¡L viscosities are uearly always larger than those calculated. If mixture
is desired. 'data are available, Moouey's correlation 175
Estimation methods for the ¡t of liqnid metals have been proposed
by Grosse1 70 and by Chapman,171 who developed a corre­ In ¡Lm = (3-128)
sponding-stat~s correlation based on atomic parameters; these tech­ ¡tI - C<l>,
niques are based on data for -20 molten metals. ÍS 50metimes applicable to <1>, as large as 0.50. The coeflicient e,
Though liquid viscosity exhibits a greater pressure variation than evaluated from data, usually varíes from 1 to 1.5 for monodisperse
almost any pure-liquid property, the eHect is rarely significant for systems, Other similar correlations which are useful have been
pressures less than about 40 atm. As with temperature sensitivity, proposed by Ting aud Luebbers lo4 (al5O requires liquid and solid
the pressure dependenee is largest for those liquid. of complex deusíty data), Kynch l76 (best fOf concentrated suspensions of uni­
molecular .tructure-for Hg, for example, the ratio of ¡L at pressure form spheres), and Ford177 (applíed to mílk fat, latex, and asphalt
to ¡L at 1 atm., ¡tpl¡LI' is ouly 1.32 at P:= 11,700 atm., whereas for emulsions, for example). ­
CHpH it is 10, and for eugenol (4-allyl-2-methoxy phenol), 107. 2. Gases. The Bromley and Wilke modification lO5 of the theo­
Tnough Audrade' s pressure-effect correlation95 allows approximate retical Hirschfelder method l06 for computiug the viscosíties of
estimates to be made to a P of -2000 atm., considerable auxiliary low-pressure pUTe gases and vapors gíves good results (-3 and 13
data are required, iucluding one value of ¡L at high pressnre. Bondi per cent average and maximum errors) if the crítical volume is
has notedOO !hat, on the average, a pressure íncrease of 480 to 500 known or can be accurately calculated. In the equation
Ib./sq. in. will increase the ¡L of liquid lubricating oils to approxi­
mately the same extent as a temperature decrease of l"C. 33.3(MTy/2
¡Lo -----vv3~ [f(1.33T,)] (3-129)
lt is not possible, in the general case, to relate, the viscosities e
of miscible Iiquid mixtures to !he pure-component viscosities alone;
but, on an empírical basis, the Kendall-Monroe equation91 appears ¡LO is expressed ín micropoise, Te in 'K., and Ve in cc./g.-mole. The
, to be the most successful of this type; variatíon of the function [f(I.33T,.)] is shown iu Table 3-322 and
has been represented analytically with accuracy by Scheibepo1 with
¡L!{3 = :t¡¡L¡;" + :t2¡L~/3 (3-122) the equatiou
Equation (3-122) applies to non-electrolytic, non-associatcd similar
liquid pairs whose component molecular weight and viscosíty diHer­
f(1.33T)
,
= L058To,645 _ _ _0.~1__ _
'(1.9T,)0,910g(L9T~
(3-130}
cnces are smal! [say (¡Lz - ¡t¡) <
15 centipoisesJ. Equation (3-122)
has been showll to be accurate to within -3 per cent for oil blends whích applies for T, > 0,3. 'Ibe group (I.33T,) in Table 3-322 is
aud to givc an average deviation fer binary mixtures of -9 per the same as kT/f 12 in Table 3-309 (used fOf Do estimatíon) if .Ik