Professional Documents
Culture Documents
Search
Collections
Journals
About
Contact us
My IOPscience
This article has been downloaded from IOPscience. Please scroll down to see the full text article.
1967 Br. J. Appl. Phys. 18 965
(http://iopscience.iop.org/0508-3443/18/7/312)
View the table of contents for this issue, or go to the journal homepage for more
Download details:
IP Address: 152.23.116.246
The article was downloaded on 14/10/2010 at 21:43
GREAT BRITAIN
Abstract. Various empirical equations used for expressing the compressive properties
of liquids have been studied. Despite its popularity, the well-known equation
attributed to Tait has several undesirable features; moreover, it is in fact not Taits
original equation, but is the result of an accidental misquotation by a later writer.
Taits original equation, when rearranged, leads to an equation of the form
_ _ _ _ - K~ + mP
vo - v
where KOis the bulk modulus at zero pressure and m the slope of the bulk-moduluspressure curve.
The equations of Tumlirz and of Tammann are merely rearrangements of the above
equation. The spurious version of the Tait equation, Hudlestons equation,
MacDonalds equation and the Van der Waal equation of state are all asymptotic to
it at zero pressure, and are practically equivalent to it over the normal range of
application.
The above equation is both the most accurate and the most convenient two-constant
compressibility equation available. By the addition of a term in P2 the equation can
be adapted to fit data for water up to 12 kb; a further term in P3 is needed to accommodate a wide range of organic liquids up to 12 kb.
1. Introduction
The subject of compressibility equations for liquids is really a very simple one, Unfortunately, it has been so badly treated in the literature that it has been made to appear
unnecessarily complex, and the resulting confusion has had at least one serious consequence.
Two generations of workers have been misled into using what they have come to call
Taits equation, without realizing that this well-known equation is not Taits original
equation, and that another equation closer to Taits original is available which is simpler,
more convenient to use and fits experimental data at least as well as (or, in some cases,
much better than) the so-called Tait equation.
The purpose of this paper is therefore threefold: to clarify an unnecessarily obscure
subject, to show the advantages of using the modified form of the true Tait equation instead
of the popular equation wrongly attributed to Tait and to discuss the use of compressibility
equations at high pressures, where all the existing two-constant equations fail to express
experimental data adequately.
2. Basic principles
Many attempts have been made to derive a compressibility equation from molecular
theory, but none of them has resulted in a convenient equation expressing the results of
experiments with adequate accuracy. To meet this need it is necessary to employ some
empirical equation, the sole justification for which is that it works.
The construction of such a n equation is a relatively simple matter, on account of the
shape of the compression curve of a typical organic liquid, which is illustrated in figure 1.
When the applied pressure is plotted against the consequent volume change, for either an
isothermal or an isentropic pressure change, the result is a smooth curve of relatively small
965
966
A . T. J. Hayward
curvature. Because of this small curvature the curve over the first few tens of bars may,
for all practical purposes, be regarded as a straight line. Within this range a liquid may
be regarded as obeying Hookes law, and only one constant is needed to define its compressive properties, namely the bulk modulus of elasticity, which is equal to the (constant)
slope of the compression curve.
Within the range of hundreds of bars the departure from linearity of the compression
curve can no longer be neglected, but it is still so small that the curve can be represented
adequately by a quadratic passing through the origin. It is not until pressures of the
order of kilobars are encountered that the departure from linearity becomes sufficiently
great to necessitate the use of a cubic equation.
Figure 1. Compression curve for a liquid. Ranges and equations applicable: A, tens of bars,
linear; B, hundreds of bars, quadratic; C, kilobar range, cubic.
This is the only true definition of bulk modulus, but to distinguish the quantity so defined
from certain broadly similar quantities it is often referred to as the tangent bulk modulus.
To derive the value of K at P = PIfrom a series of measured values of P and Vit is necessary
first to plot P against VO- V, and then to derive the slope of the curve at the point
(PI, V I )and finally to multiply this slope by V I . If the measurement of the slope is done
graphically the result is likely to be rather inaccurate, and it is usually necessary to use a
numerical process, which involves either a considerable expenditure of effort or the use
of a computer.
967
It is very much easier to derive another quantity, which may be regarded as an average
value of bulk modulus over the range from 0 to P. It is defined by
K- E - VOP
vo - Y
Because it corresponds to the slope of the secant (or chord) cutting the compression curve
at the origin and at the point (P,V), this quantity is usually referred to as the secant bulk
modulus.
A third quantity has also been introduced to which the term bulk modulus is sometimes
applied. This quantity is a hybrid between the two quantities defined above as K and R,
and is defined by
dP
dV
K/e-vo-
This quantity has no commonly accepted distinctive name, but it can fittingly be termed
the mixed bulk modulus. It serves no useful purpose, since it is neither applicable to
thermodynamic analysis nor convenient for engineering calculations, and it seems a pity
that it was ever invented.f
All three forms, K,R and K, coincide when P is zero, but diverge at an increasing rate
as P increases.
At pressures up to several hundred bars with any liquid, K, R and K are all practically
linear functions of P. If the results of a compressibility experiment within this range are
expressed as a plot of either K, R or K against P,the deviation of the experimental points
from a straight line is likely to be random, and ascribable to experimental error in the
apparatus used. This means that any one of the following three equations provides an
adequate way of expressing experimental results within this pressure range, so it is not
possible to distinguish between them on grounds of accuracy.
(a) The linear tangent-modulus equation (used by Moelwyn-Hughes (1957) and Anderson
(1966)) :
dP
K = - V--Ko-mlP.
(4)
dV
(b) The linear secant-modulus equation (used by Klaus and OBrien (1964), Holland
(1966) and Hayward (1964)):
(e) The so-called Tait equation (used by a very large number of authors), which is
identical with the linear mixed-modulus equation :
dP
- YOdV
- = KO+ nzsP.
(6)
It will be observed that whereas KO,the value of bulk modulus at zero pressure, is the
same in all three equations, the slope m of the bulk-modulus-pressure line is different in
each equation.
Since all three equations are equally accurate within the pressure range that concerns
the great majority of workers, the choice between them can be governed by convenience.
On this ground the linear secant-modulus equation (5) is unquestionably superior because
R can very easily be evaluated from the experimental data. The use of this equation is
also in harmony with the philosophical principle of Occams razo?, which insists that the
simplest possible hypothesis should always be adopted first, and abandoned for a more
VP/(VO- V )is, of course, also possible. It is equivalent to - P.
Fortunately no writer so far appears to have confused the issue still further by using this form Qf
bulk modulus, and so its hypothetical existence can be ignored.
f The other hybrid, K
968
A . T. J. Hayward
201
0.2
0.4
0.6
0.8
P (kbl
There is, however, a third reason for preferring this equation, in that it is the only compressibility equation which can be used to express the result of tests on water at pressures
up to several kilobars. Figure 2 shows the linear relationship between R and P given by
the results of some recent measurements up to 1 kb by Kell and Whalley (1965), who took
elaborate precautions to obtain the highest possible accuracy. The most accurate values
available for the range 0-3 kb are generally conceded to be those of Amagat (1893), which
also give a linear relationship between R and P, as is shown in figure 3. Although the
results at one temperature only are given in each case, similar graphs have been drawn
for tests at other temperatures and similar linear relationships were always obtained.
In both figures 2 and 3 there is a small departure from linearity at very low pressures,
but that this is without physical significance may be deduced from the fact that at the
lowest pressures the Kell and Whalley points deviate upwards, while those of Amagat
deviate downwards, In both cases this deviation at low pressures is almost certainly due
to the increased proportional error which is inevitable when very small values of volume
change have to be measured.
Only when a compression test on water is extended to about 10 ltb does the curvature
of the ($P) graph become clearly apparent. Bridgman (1958) worked over this range, and
although his results when expressed in this form show a departure from linearity, this is
still less than 2 % of the mean value. This is not very much greater than the probable
error in the data.
The superiority of the linear secant-modulus equation (5) can be seen by comparing
figure 3 with figure 4, in which comparable values of K and K derived from Amagats
results have been plotted against P. The departure from linearity amounts to about 1* 5 %
201
0
969
P (kbl
P (kbl
of the mean value in the K graph and about 2 % in the K' graph, thus showing that neither
(4) nor (6) is valid for water over the range 0-3 kb.
Yet another important advantage of the linear secant-modulus equation is that, because
of its simple form, it can be used to derive several important parameters, such as the specific
volume and density at any pressure P and the true or 'tangent' bulk modulus K.
Whatever the relationship between K and P, it follows from (2) that
v=
x-P
Vo-
(7)
poR
p=-- K -
K=
R(R - P )
R - P dR]dP
(9)
and
For the special case of the linear secant-modulus equation, which may be expressed in
the form
= KO - J??P
(1 1)
it is additionally true that
K = -(I?KO
and
R2
K' = -,
KO
P)
A. T. J. HayM.ard
970
5. The original Tait equation
This equation is of identical form with (ll), but with II/A = KOand A-1 = m. Thus
the equation propounded by Tait was actually the linear secant-modulus equation, expressed in
reciprocal form.
Why Tait chose to express it in this form, and then laboriously to fit his experimental
points to a hyperbola, when he could have inverted the equation and fitted his points to a
straight line, will always remain a mystery. All that can now be said is that if only he
had done so, he might have saved himself from being misquoted and two generations of
physicists from being led astray in consequence.
6. The spurious Tait equation
It will never be known for certain who it was that first misquoted Tait. The earliest
occurrence of the spurious Tait equation discovered by the author, however, is in a book
published by Tammann (2907). But Dr. G. S. Kell (who independently reached the same
conclusion as the present author) has said in a private communication that Tammann
(1895) actually attributed the wrong equation to Tait as early as that year, in a paper
published during Tait's lifetime.
Whether Tammann himself misquoted Tait, or whether he was merely following some
unknown earlier writer who had misquoted Tait, is of little consequence. The fact is
that hundreds of authorities during the past sixty years have blindly followed Tammann
and used this inferior equation instead of the true Tait equation, so chat Tammann probably
deserves to be regarded as the perpetrator of one of the most far-reaching misquotations
in the history of physics.
The equation which he attributed to Tait was, in the notation of the present paper.
dV
dP
A
B - P'
Comparison with (14) reveals that Tammann has replaced Tait's 'average compressibility'
'O (which could be written as - PVO
vo AP
'
971
by the corresponding differential coefficient, dV/dP. The resulting equation is the linear
mixed-modulus equation (6), as may be seen by inverting both sides of (16) and multiplying
throughout by the constant VO.
Tammann then proceeded to integrate this equation, obtaining
It was left to later writers to convert this integrated equation into the form in which it is
generally used today:
v o - v - Clog B + P
--
(B).
VO
It cannot be emphasized too strongly that this equation has no advantages over the linear
secant-modulus equation, and has a number of serious disadvantages. It is much less
convenient to use, its constants have no obvious physical meaning, it does not give rise
to simple expressions from which other important constants can be obtained and it does
not fit experimental data for water nearly as well as the other equation, as may be seen
from a comparison of figures 3 and 4.
This last statement is, of course, in conflict with the views of various modern writers
who have extolled the virtues of the spurious Tait equation. Some of them (see for example
Rowlinson 1959, p. 32) have quoted the following table from Tait's papers as an alleged
proof of the accuracy of the spurious Tait equation.
501
1001
Pressure (atm)
1
1501
2001
2501
3001
Relative volume of water 1 .OOOOO 0,97668 0.95645 0.93924 0,92393 0.91065 0.89869
at 0"c observed by
Amagat
Relative volume of water 1-00000 0.97657 0.95652 0.93916 0.92399 0.91062 0.89875
at 0"c calculated by
Tait
Reference to Tait's original paper (Tait 1900, pp. 334-8) reveals, of course, that he based
his calculations upon his own equation (equation (14)), so that t h s table is really a confirmation of the linear secant-modulus equation, and not of the spurious Tait equation.
7. Equivalence of the constants in the spurious Tait equation
Although there is no reason why the spurious Tait equation should continue in use in
the future, a great deal of work in the past has been published in terms of the two constants
in this equation. It is therefore fortunate that these two constants can easily be converted
to the two constants in the linear secant-modulus equation.
When the spurious Tait equation is expressed in the usual form (equation (18)) the corresponding differential form is
dP
B 2.3026p.
K' E - Vo- = 2.3026 - $- (19)
dV
C
C
At zero pressure K' = X = KO,and consequently it follows from (19) that
Expressing m in terms of B and C is not quite so simple, since (11) and (19) are incompatible. That is to say, if K' is a linear function of P, then Rcannot also be linear (although
it is very nearly linear when P K). Equation (13), however, although strictly true only
when R is linear, is a very close approximation in the case of (19), provided that P <I?.
From this it follows that for a (K,
P ) curve corresponding to a set of points satisfying
equation (17)
dl? _
--1 E 1 3
B jli2
(21)
dp
B-P
'
<
(
~
9 72
A . T. J. Hapsaid
This set of points will form a very flat curve. and if the best straight line is drawn through
them its slope will be practically the same as the value of dK/dP at the mean pressure.
Hence the required relationship is
where Pmas is the upper extreme of the pressure range over which the experiments were
carried out. (It is assumed that the lower extreme is at or near atmospheric pressure.)
Over fairly small pressure ranges the factor in parentheses can be neglected, with a
consequent simplification of (22). Even over the greatest pressure ranges for which the
spurious Tait equation can be expected to hold, the factor in parentheses will have only a
small effect on the value of 112, which remains nearly proportional to the reciprocal of C.
Consequently, the observation which has frequently been made about C in the spurious
Tait equation-that it is approximately the same for nearly all organic liquids at any
temperature-applies also to the constant in in the linear secant-modulus equation.
where p and t' denote pressure and volume, P and a are arbitrary constants and the righthand side of the equation is (for isothermal compression) another constant. This is
expressed in the present notation as
(P - A)( V - B ) = C.
(23)
Tammann
Equations (25) and (26) are both clearly of the linear secant-modulus form, thus showing
that both Tumlirz's and Tammann's equations are nothing more than rearrangements of
the linear secant-modulus equation.
973
where L = L o ( V / V O )and
~ / ~Lo, C1 and C2 are arbitrary constants.
Substituting for L and rearranging gives
-+
A . T. J. Hayward
974
Such a curve, however, has not been drawn through the points because, for several reasons,
it seems unlikely to be justified.
In the first place, the graphs for every other liquid that has been examined are curved
in the opposite direction, and there is no reason to expect that mercury would prove
exceptional in this respect. Secondly, it is physically impossible for the upward curvature
to continue very far, since it is bound to lead to an infinite value of K (though not of K)
at a finite pressure; a second-degree polynomial drawn though the points for mercury in
figure 5 would lead to a n infinite value of K at about 40 kb. Thirdly, Bridgman criticized
his own experiments with mercury, on the grounds that his piezometers appeared to be
deforming anisotropically under pressure. This caused his results to be widely scattered,
and he employed an elaborate smoothing technique to deal with the scatter. The upward
curvature in figure 5 is probably a direct consequence of Bridgmans smoothing technique.
Figure 5. Secant bulk modulus of four liquids at pressures up to 12 kb. Curve A, water (50c,
quadratic); 3, di(2-ethylhexyljsebacate (25 c, cubic) ; C, iso-amyl alcohol (20c,cubic); D, mercury
(20c,linear).
Consequently a straight line has been drawn through the points. The line is within
1 % of every point, and this is considerably less than the probable error of the data. In
the present rather inadequate state of knowledge, therefore, it is evident that the linear
secant-modulus equation (equation (1 1)) is capable of expressing the compressive properties
of mercury up to at least 12 kb.
Although with water, as has already been mentioned, the value of dzK/dP* is practically
constant, with all organic liquids it appears to decrease with increasing pressure. Two
typical sets of data for organic liquids, one from each source, are given in figure 5, in
which the increased curvature at the lower pressures can be clearly seen. Because the
rate of decrease of d2K/dP2 may, within the limits of accuracy of the available data, be
regarded as constant, a third-degree polynomial can be used to express the compressive
properties of any organic liquid up to 12 kb. This will take the form
K=
KO+ w P - I z P ~ + ~ P ~
(31)
975
K = R(R - P)
KO nP2'
Similarly, with the cubic secant-modulus equation, for use with organic liquids up to
very high pressures, it follows from (9) that when
then
R(R - P)
K = KO-+
nP2- 2qP3'
(33)
Both the expressions within parentheses in (35) are constants, and since V changes only
very slowly with P the factor Vo2/V2can be ignored at low pressures. Consequently this
equation also coincides with the linear secant-modulus equation at zero pressure. At
higher pressures, however, the factor Vo2/V 2 becomes increasingly significant, and the
equation gives much higher values of i? at high pressures than are given by the linear
secant-modulus equation-while real liquids at high pressures deviate from the linear secantmodulus equation in the opposite direction.
12. Conclusions
(a) Several two-constant empirical equations may be used to express the compressive
properties of liquids over the most commonly used pressure range, from zero to several
hundred bars. Of these, the linear secant-modulus equation is for severzl reasons the
best. By the addition of one or two extra terms it can be extended to cover all liquids at
pressures up to 12 kb over a wide range of temperature (see $13).
(b) The well-known equation to which Tait's name has been attached is not a desirable
equation, and was certainly not propounded by Tait. It appears to have originated
through an unfortunate misquotation by Tammann of Tait's original equation (which was
actually the linear secant-modulus equation expressed in reciprocal form). Relationships
have been derived between the so-called Tait constants and the constants in the linear
secant-modulus equation, thus enabling results expressed in one form to be re-expressed in
the other (see $13).
(e) The equations proposed by Tumlirz and by Tammann both fit experimental data
well, but only because they are both identical with the linear secant-modulus equation,
which these writers have expressed in less convenient forms. The so-called Tait equation,
and the equations proposed by Hudleston and by MacDonald, are all asymptotic to the
linear secant-modulus equation at zero pressure and, within the limits of experimental
error, may be regarded as coinciding with it over the normal range of application (0-1 kb);
these three equations, however, are much less convenient to use than the linear secantmodulus equation. Van der Wad's equation of state is also asymptotic to the linear
9 76
A . T. J. Hayward
secant-modulus equation at zero pressure, but does not fit experimental data at all well
at moderately high pressures. No two-constant equation is satisfactory at very high
pressures.
13. Summary of useful equations
13 . 1. DeJinitiom and relatioriships folloir.ingfr.oiii them
K- r - YOP
K = - - VdP
dV'
vo - V '
Y = Yo-K - P
K
K=
K(K - P)
K - P dRldP'
K = KO- WIP
from which it follows that
K=
X(K - P )
KO
'
X
from which it follows that
- nP2
= KO- UTP
K(K - P)
K = KO nP2 *
= KO- nzP
- 11P2 - qP3
(where KO,m,72 and q are all positive constants), from which it follows that
K(K - P )
K = KO- nP2 - 2qP3'
13.5. The so-called Tait equation
Because this equation is in several ways inferior to the linear secant-modulus equation,
there is no reason why it should continue to be used. Where the results of previous work
have been expressed in terms of the constants B and C in the equation
they can be re-expressed in terms of the constants KOand in in the linear secant-modulus
977
KO= 2.3026
and
where Pm,, is the upper extreme of the pressure range over which the experiments were
carried out, the lower extreme being assumed to be at or near zero.
Acknowledgments
The painstaking work of Mrs. A. C. Findlay in fitting a very large volume of experimental
data t o various compressibility equations is gratefully acknowledged.
This paper is published by permission of the Director of the National Engineering
Laboratory of the Ministry of Technology and the Controller of Her Majestys Stationery
Office.
References
AMAGAT,
E.-H., 1893, Ann. Chim. Phys., 29, 68-136, 505-74.
ANDERSON,
0. L., 1966, J. Phys. Chem. Solids, 27, 547-65.
A.S.M.E., 1953, Viscosity and density of over forty Iubricatingfluids of known composition atpresslire
to 150 000 p.s.i. and temperatures to 425% (New York: Amer. Soc. Mech. Engrs).
BET, K. E., 1953, J. Imp. Coll. Chem. Engng Soc., 7 , 35-51.
BRIDGMAN,
P. W., 1958, The Physics of High Pressure (London: Bell and Sons).
- 1964, Collected Experimental Papers (Cambridge, Mass. : Harvard Univ. Press).
HAYWARD,
A. T. J., 1964, Acta Imeko 1964, Part 2 (Budapest: Hungarian Scient. Soc. for
Measurement and Automation), pp. 249-72.
HOLLAKD,
C. D., 1966, Engng Mater. Des., 9, 1646-9.
HUDLESTON,
L. J., 1937, Trans. Faraday Soc., 33, 97-103.
KELL,G. S., and WHALLEY,
E., 1965, Phil. Trans. Roy. Soc. A, 258, 565-617.
KLAUS,E. E., and OBRIEN,J. A., 1964, J. Bas. Engng, 86, 469-74.
MACDONALD,
J. R., 1966, Rev. Mod. Phys., 38, 669-79.
MOELWYN-HUGHES,
E. A., 1957, Physical Chemistry (Oxford: Pergamon), pp. 323-4.
PARTIKGTOK,
J. R., 1951, An Advanced Treatise on Physical Chemistry, Vol. 2: The properties of
liquids (London: Longmans, Green), pp. 66-8.
ROWLINSON,
J. S., 1959, Liquids and Liquid Mixtures (London: Butterworths), pp. 31-2.
TAIT,P. G., 1888, The Voyage of H.M.S. Challenger, Vol. 2, Part 4(London: H.M.S.O.),pp. 1-73.
-1900, Collected ScientiJic Papers, Vol. 2 (Cambridge: Cambridge University Press), pp. 334-8,
339-42, 343-8, 354-5.
TAMMAAN,
G., 1895,Z. Phys. Chem., 17, 620.
- 1907, Uber die Beziehungen zwischen den inneren Kriifferi und Eigenschaften der Losungeiz
(Hamburg: Leopold Voss), p. 32.
- 1911, Ivachr. Akad. Wiss. Gottingen, Mathematisch-physikalischer Klasse, pp. 527-62.
TUMLIRZ,
O., 1909, Akad. Wiss. W e n , Mathematisch-natumissenschaftlicheKlasse, IIa, 1318,
203-41.