Professional Documents
Culture Documents
An Intensified Analytic Solution For Finding The Roots of A Cubic Equation of State in Low Temperature Region
An Intensified Analytic Solution For Finding The Roots of A Cubic Equation of State in Low Temperature Region
a r t i c l e i n f o a b s t r a c t
Article history: Because of the effect of the enlarging process of errors in low temperature region, the analytical solutions of a
Received 19 December 2014 cubic equation of state (CES) lead to illogical results. To overcome this problem, we recommend an enhanced an-
Received in revised form 3 February 2015 alytic method, in this paper. Up to now, there is not any analytical method for this problem. In the present study,
Accepted 16 February 2015
the general cubic equation x3 + Bx2 + Cx + D = 0 is transformed into simple equation γ3 − βγ + β = 0. For
Available online 19 February 2015
β ≥ 27/4, then the three roots are real. To compare, as an example, for 1-butene at T = 112.3 (K) and P =
Keywords:
3.79 × 10−9(Pa), the empirical value of the molar volume is 72.00 (m3mol−1). By using Patel–Teja (PT) CES,
Analytic the molar volume value is obtained as −342.3 from traditional analytic method and 73.88 from both the new in-
Cubic tensified analytic method and the iterative Newton–Raphson method.
Equation of state © 2015 Elsevier B.V. All rights reserved.
Low temperature
Organic
1. Introduction CES cannot provide simultaneous accurate predictions for all the proper-
ties of pure fluids and their mixtures. But, the most CES models (e.g.,
The CES models are useful formulas for analyzing and correlating the Peng-Robison CES in 1976 as P(T, v) = RT/(v − b) − α(T)/(v2 + 2bv −
thermodynamic equilibrium state of simple fluids and their mixtures, b2) being used widely today for practical design purposes have been de-
with agreeable approach [1–5]. Generally, for a given substance, the rived from van der Waals CES [7].
CES is an empirical relation of the pressure P, the temperature T and For finding the roots of Eq. (2), the analytical methods can solve their
the molar volume v consist of a repulsive term and an attractive term three roots simultaneously. One of the relevant math formulas is the
as follows [6]: well-known formula due to Ferro and Tartaglia, communicated by
Gerolamo Cardano in 1545 [8,9].
RT α ðT Þ Cardano's method involves division. Thus it can encounter near “0/0”,
P ðT; vÞ ¼ − ð1Þ
v−b ðv−cÞðv−dÞ resulting insignificant numerical errors. Thus it may encounter a numeri-
cally unstable case. Lagrange in 1770 gave the new method for solving the
where R is the gas constant. b, c and d are constant parameters which are three roots of a cubic equation, with 18 possible interpretations [10]. The
obtained based on the kind of equation. In Eq. (1), α(T) is a function and Lagrange formula, does not require division and thus avoids the “0/0”
it has many empirical definitions in the literature. By expressing Eq. (1) case. In the Lagrange formula some interpretations are correct (yielding
in terms of molar volume, the following equation is obtained: the solutions), But the others are not. Zhao et al. [11] proposed a convec-
tion which provided correct interpretations of the Lagrange formula for all
3 RT 2 RT RT α ðT Þ cubic equations with real coefficients.
v − bþcþdþ v þ bc þ bd þ cd þ cþ dþ v
P P P P Zhi and Lee [12] revealed by several examples that, in low tempera-
ð2Þ
RT α ðt Þ ture region, the analytical solutions of CES lead to irrational results,
− bcd þ cd þ b ¼ 0:
P P while the iterative solutions of the CES, using Newton–Raphson method
produced valid results. They appeared that errors caused by the limita-
Eq. (2) is referred to as cubic in molar volume. Additionally we can tion of significant figures of the computer languages are revealed, and a
substitute the definition of compressibility factor (i.e., Z = Pv/RT) into magnification of errors is defined which a main factor is bringing out the
Eq. (2) and obtain different cubic equations in Z. irrational results of the analytical solution of CES. Salim [13] showed
The first CES was recommended by van der Waals in 1873 as P(T, v) = that, in the case of low temperature region, the coefficient D in cubic
RT/(v − b) − α(T)/v2. It has been proven that the original van der Waals equation x3 + Bx2 + Cx + D = 0 is very small. Therefore the main
root is also small and becomes so close to zero. Thus, by eliminating
E-mail address: behrooz_m_ziapour@yahoo.com. the term of x3 in x3 + Bx2 + Cx + D = 0, then the main root can be
http://dx.doi.org/10.1016/j.molliq.2015.02.026
0167-7322/© 2015 Elsevier B.V. All rights reserved.
166 B.M. Ziapour / Journal of Molecular Liquids 206 (2015) 165–169
found via the approximated equation of the form Bx2 + Cx + D = 0. formula as:
Loperena [14] proposed an iterative refinement of the solution obtained 8 0 sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi11 0 sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi11
by the analytical method. This method allowed one to take advantage of >
>
3 3
>
> @ β β 2
β 3
A @ β β2 β3 A 27
the calculations fulfilled in the application of the analytical solution. His >
< γ ¼ − þ − þ − − − ; for β ≤
2 4 27 2 4 27 4
numerical results show that the proposed iterative refinement proce- rffiffiffiffi !
>
>
dure is an attractive and numerical inexpensive option and can be easily >
> β 1 −1 9 27
>
: γ ¼ −2 cosθ; θ ¼ cos pffiffiffiffiffiffiffiffiffi ; for β ≥ :
incorporated into any process simulation program, that use Cardano's 3 3 12β 4
method, without significant modifications. ð5Þ
Up to now, as mentioned above discussions, there is not any
completely analytical method for this problem. To notice to complexity It is clear that (see Fig. 1), when β → − ∞, then γ → 1. Also, when
and overhead of analytical methods, it has been recommended to use β → + ∞, then γ → − ∞.
the numeric iterative methods. In this paper, we describe an enhanced To obtain the two formulas in Eq. (5), an analytic method was
completely analytic method for solving the general cubic equation completely used. This method has been described in Appendix A. There-
x3 + Bx2 + Cx + D = 0 consists of real coefficients. We verify this fore, through an intensified analytic solution, the real root of the cubic
new method with the miscellaneous examples. Eq. (3) is captured.
The other roots of Eq. (3) i.e., x2, x3 can be calculated analytically
2. New analytic solution and the miscellaneous examples from the following quadratic equation as:
3 2 Note that, in order to reach the high precision of the solutions, espe-
x þ Bx þ Cx þ D ¼ 0: ð3Þ
cially in low temperature region, the two other roots may be obtained
from Eq. (6). In this paper, all solutions were conducted by Microsoft
C3 Excel, 2003.
In Eq. (3), one of the three roots is real. To find it, let β ¼ − D12 ,
1
where C1 = 9C − 3B2 and D1 = 2B3 − 9BC + 27D. Make the substi- Example 1. In equation: x3 − 3x2 + 4x − 2 = 0, the coefficients are:
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi B = − 3, C = 4 and D = − 2. Then, C1 = 9 and D1 = 0. Therefore,
tution x ¼ − 13 3 ð1−γÞD1 þ B or x ¼ − 13 DC 11 γ þ B . By this substitu-
β = − ∞. Thus, from Fig. 1: γ = 1. Therefore, x1 = − B/3 = 1. Finally
tion, Eq. (3) is transformed as the following simple equation for γ: from Eq. (6): x2 − 2x + 2 = 0. Then, x2 = 1 + i and x3 = 1 − i.
3
Example 2. In equation: x3 + 6x2 + 3x − 10 = 0, the coefficients are:
γ −βγ þ β ¼ 0: ð4Þ B = 6, C = 3, D = −10. Then C1 = −81 and D1 = 0. Therefore, β = + ∞.
Thus, from Fig. 1: γ = − ∞. Therefore, x1 = − B/3 = −2. Finally from
γ 3 Eq. (6): x2 + 4x − 5 = 0. Then, x2 = 1 and x3 = −5.
To find γ, we arrange Eq. (4) into the functional form as βðγ Þ ¼ γ−1.
Fig. 1 shows the sketch of β(γ). It is seen that for β ≥ 27 Example 3. In equation: x3 − 2x2 − 5x + 6 = 0, the coefficients are:
4 , then all three
roots are real. Therefore, Eq. (3) has three real roots (i.e., x1, x2, x3). B = − 2, C = − 5, D = 6. Then C1 = − 57 and D1 = 56. Then, β =
59.0539. Thus, from Eq. (5):γ = −8.1429. Therefore, x1 = −2.0000. Fi-
Also, for β b 27
4 , Eq. (3) has one real root (i.e., x1). In this case, the others
nally from Eq. (6):x2 − 4x + 3 = 0. Then, x2 = 1.0000 and x3 = 3.0000.
are conjugate imaginary roots (i.e., x2, x3).
Therefore, like a new analytic method, for finding the one real root of Example 4. The cubic equation: (x − 1)2x = x3 − 2x2 + x = 0. The co-
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
a cubic equation, let x1 ¼ − 13 3 ð1−γ ÞD1 þ B or x1 ¼ − 13 DC 11 γ þ B be efficients are: B = −2, C = 1, D = 0. Then C1 = −3 and D1 = 2. Then,
β = 6.75. Thus, from Eq. (5):γ = −3. Therefore, x1 = 0. Finally from so-
one real root of Eq. (3), where γ is related with the left branch of the
lution of the equation:x2 − 2x + 1 = 0, we have:x2 = x3 = 1.0000.
curves β in Fig. 1 (i.e., γ ≤ 1); and it is obtained from the following
Example 5. The molar volume of carbon dioxide (CO 2 ) at P =
10 atm and T = 300 K, by using the van der Waals CES (P(T, v) =
RT/(v − b) − α(T)/v 2 ), is the cubic equation as: v 3 −
2.5046v 2 + 0.3598v − 0.0155 = 0. This equation and another
equation in the form of v 3 − 7.8693v 2 + 13.3771v − 6.5354 = 0
Table 1
Results for example 5.
were solved by the some numerical iteration methods [15]. The ini- Table 3
tial guess for starting the iteration was selected from the ideal gas Comparison among experimental, old and new analytical methods calculated liquid vol-
ume for 1-butene (TC = 419.5 K, PC = 4020 kPa, ω = 0.194).
law. We collected both these numerical solutions and our new an-
alytical calculations, in Table 1. As shown from the new solutions, P (KPa) P (MPa) VExp PR PT
the vales of β are less than 6.75. Thus, the selected cubic equations VNew VOld VNewton VNew VOld VNewton
have only one real root (i.e., v1 ≡ x1). The other imaginary roots are
87.8 3.56E−13 70.09 70.32 1685.9 70.32 72.68 – 72.68
unusable. 91.3 1.87E−12 70.35 70.47 1603.6 70.47 72.84 1621.3 72.84
94.8 8.58E−12 70.62 70.62 – 70.62 73.01 1544.6 73.01
98.3 3.51E−11 70.89 70.78 −92,975 70.78 73.17 – 73.17
3. Results and discussions 101.8 1.29E−10 71.16 70.94 −25,216 70.94 73.34 1408.3 73.34
105.3 4.32E−10 71.44 71.11 1332.8 71.11 73.52 1347.3 73.52
For substances, it is clear that, in the low temperature and pressure 108.8 1.33E−09 71.71 71.28 – 71.28 73.69 1290.5 73.69
112.3 3.79E−09 72.00 71.45 0.460 71.45 73.88 −342.3 73.88
ranges, their CES will have three real roots (i.e., β ≥ 6.75). In order to
115.8 1.01E−08 72.28 71.62 116.28 71.62 74.06 129.1 74.06
compare our new method with the traditional methods, taken from 119.3 2.51E−08 72.57 71.80 78.07 71.80 74.25 72.17 74.25
Zhi and Lee [12], we assume that CES be formed versus the compress- 122.8 5.92E−08 72.86 71.99 72.18 71.99 74.45 73.20 74.45
ibility factor z as: z3 + c1z2 + c2z + c3 = 0. The smallest root (i.e., the 126.3 1.32E−07 73.16 72.17 72.91 72.17 74.64 74.45 74.64
compressibility factor of liquid phase:z2) is obtained from the tradition- 161.3 4.98E−05 76.34 74.30 74.30 74.30 76.89 76.89 76.89
196.3 1.80E−03 80.00 76.98 76.98 76.98 79.73 79.73 79.73
al method as follows:
231.3 1.90E−02 84.31 80.45 80.45 80.45 83.39 83.39 83.39
0 pffiffiffiffiffiffiffiffiffiffi 1 266.3 9.86E−02 89.52 85.10 85.10 85.10 88.27 88.27 88.27
pffiffiffiffiffiffiffiffi arccos r= −q3 ∘ c
z2 ¼ 2 −q cos@ þ 120 A− 1 ð7Þ
3 3
where, q ¼ 19 3c2 −c21 and r ¼ 16 c1 c2 − 12 c3 − 27
1 3
c1 . Then the molar vol-
3 −1 infinity. The magnification of error analysis for present intensified ana-
ume of the liquid phase is given as: V Old ¼ z2 RTP (m mol ). Suppose
that, VExp, VNew, VOld and VNewton be denoted to the molar volume of the lytic method was done as the following process:
pffiffiffiffiffiffiffiffiffi
liquid phase obtained from the experimental, the new method, the old Let s ¼ 9= 12β, so the second formula of Eq. (5) can be arranged as:
method and the iterative Newton–Rephson method, respectively.
These values for three pure substances (propylene, 1-butene and 1- −3 1 −1
γ¼ cos cos ðsÞ : ð8Þ
pentene), have been listed in Tables 2–4, by using both the PR and the s 3
PT CES models. In these tables, to find the VNewton, we consider Newton
3 2 2 When Δs tend to zero, the magnification of error (M) can expressed
scheme as: zkþ1 ¼ zk − zk þ c1 zk þ c2 zk þ c3 = 3zk þ 2c1 zk þ c2 .
as [12]:
The recorded results for VNewton in Tables 2–4, were obtained due to
the three iterations as k = 3. In this iterative process, the initial guess dγ s 3 1 −1 1 1 −1 s
M¼ ¼ cos cos ðsÞ þ pffiffiffiffiffiffiffiffiffiffiffi
ffi sin cos ðsÞ : ð9Þ
for zk was as z1 = 0. As shown in Tables 2–4, the results indicate that ds γ s2 3 s 1−s 2 3 γ
both the present intensified analytic method and the iterative New-
ton–Rephson method have identical accuracies. But, because of the ef- In low temperature region, the values of β, γ and s do tend to 27/4, 3
fect of the enlarging process of errors, especially in lower temperature and 1 respectively. When s = 1, then the second term in Eq. (9) gets
region, the traditional analytic solutions of CES lead to illogical results. p1ffiffiffiffiffiffiffi sin 1 cos−1 ð1Þ ¼ 0. Let
1 1−1 3 0
The magnification of error was analyzed for traditional method (i.e.,
Eq. (7)) by Zhi and Lee [12]. Through this analysis, it was found that the dγ s
pffiffiffiffiffiffiffiffiffiffiffiffi M¼
critical term in the magnification of error is 1= 1−s2 , where, the value ds γ
pffiffiffiffiffiffiffiffiffiffi rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi !
of s is as: s ¼ r= −q3. They showed that in low temperature region, the 3 1 −1 1 1 −1 s
¼ 2 cos cos ðsÞ þ pffiffiffiffiffiffiffiffiffiffiffiffi 1− cos2 ð cos ðsÞÞ : ð10Þ
values of s do tend to 1, so the error magnification (M) will become s 3 s 1−s2 3 γ
87.9 9.18E−10 55.07 53.91 1002.3 53.91 56.62 – 56.62 As mentioned above, in low temperature region, we have as:
89.4 1.60E−09 55.18 53.97 980.71 53.97 56.69 995.94 56.69 γ → − 3. Finally, by inserting this value into Eq. (11), then the value
90.9 2.74E−09 55.29 54.04 −158.8 54.04 56.76 −395.28 56.76 of M is obtained as follows:
92.4 4.59E−09 55.39 54.10 108.41 54.10 56.83 275.55 56.83
93.9 7.56E−09 55.50 54.17 82.69 54.17 56.91 96.98 56.91
dγ s 3 1 −1 1 s
95.4 1.22E−08 55.61 54.24 67.72 54.24 56.98 81.71 56.98 M¼ ¼ cos cos ðsÞ þ : ð12Þ
96.9 1.95E−08 55.72 54.30 53.97 54.30 57.05 42.68 57.05 ds γ s2 3 s γ
98.4 3.06E−08 55.84 54.37 61.62 54.37 57.13 56.99 57.13
99.9 4.73E−08 55.95 54.44 55.69 54.44 57.20 60.93 57.20 Therefore, as seen from value of M in Eq. (12), in low temperature re-
101.4 7.21E−08 56.06 54.51 54.45 54.51 57.28 58.77 57.28
gion, there is not any infinity term for value of M, which is the reason
102.9 1.08E−07 56.17 54.58 54.92 54.58 57.35 58.24 57.35
127.9 2.08E−05 58.17 55.87 55.87 55.87 58.76 58.76 58.76 why this present method can give corrective calculation results.
152.9 6.03E−04 60.39 57.42 57.42 57.42 60.44 60.44 60.44 For a better presentation, we have listed the new solution procedure
177.9 6.09E−03 62.91 59.31 59.31 59.31 62.48 62.48 62.48 in Tables 5 and 6. Tables 5 and 6 have been listed based on the PR model
202.9 3.24E−02 65.79 61.64 61.64 61.64 65.00 65.00 65.00 for propylene and 1-butane, respectively. In the following example,
227.9 1.14E−01 69.16 64.58 64.58 64.58 68.18 68.18 68.18
more details of the new method calculations have been introduced.
168 B.M. Ziapour / Journal of Molecular Liquids 206 (2015) 165–169
Table 4
Comparison among experimental, old and new analytical methods calculated liquid volume for 1-pentene (TC = 464.4 K, PC = 3555 kPa, ω = 0.233).
Example 6. From Table 5, for propylene using PR model, at T = from the following equation as:
227.9(K) and P = 114(kPa), then the following CES is obtained:
3
2c1 −9c1 c2 þ 27c3 c1
z1 ¼ − : ð13Þ
9c2 −3c1 2 3
3 2
z −0:9969224047840840z þ 0:0303480538062392z−0:0001029289169739 ¼ 0:
Please note that, Eq. (13) has been adjusted based on the cubic equa-
D1
tion as: x3 + Bx2 + Cx + D = 0 and its one real root (x1 ¼ − 13C 1 γ þ B ),
Therefore, C1 = − 2.7084303592252900 and where C1 = 9C − 3B2,D1 = 2B3 − 9BC + 27D and γ = −3.
D1 = − 1.7120783876879200. Thus:
β = 6.7780650055531100. Then, θ = 0.0214638985371484. Hence 4. Conclusions
from Eq. (5):
γ = − 3.0055377417229900. Then, x1 = z1 = For substances, it is clear that, in the low temperature and pressure
0.9656036987339030. From Eq. (6): ranges, their CES will have three real roots. It is shown that, in the low
x2 = z2 = 0.0038856563864182 and x3 = z3 = temperature region, the traditional analytical solutions of the CES lead
0.0274330496637624. Finally we have: to irrational results, while the iterative solutions of the CES, produced
−1 valid results. In this work, as a new analytic method, by making the sub-
V 2 ¼ V New ¼ z2 RT
P ¼ 64:5849974569685000 m3 mol .
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
According to the recorded results in Tables 5 and 6, one can see that, stitution x ¼ − 13 3 ð1−γÞD1 þ B or x ¼ − 13 DC 11 γ þ B , the cubic equa-
in lower temperature region, the value of γ is exactly closed to −3. A tion of x3 + Bx2 + Cx + D = 0 was converted to equation γ3 −
cheap analytic method may be introduced. Therefore, for finding the βγ + β = 0, where γ was related with the left branch of the curves β
real root of the CES in lower temperature region, one can get as in Fig. 1 (i.e., γ ≤ 1); and it was obtained from Eq. (5). The magnification
γ = −3; then the one root of z3 + c1z2 + c2z + c3 = 0 is easily obtained of error analysis for present intensified analytic method was done. Also,
Table 5
The details of the new method solutions, for propylene using PR model.
Table 6
The details of the new method solutions, for 1-butene using PR model.
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
a new analytic method was tested for three pure substances as follows:
If β ≤ 27/4, then the sign of the square root β2 =4−β3 =27 is posi-
propylene, 1-butene and 1-propylene. Then the results were listed in
tive; and the formula in Eq. (A-5) can be directly solved. If β N 27/4,
Tables 2–6. As shown in these tables, the appreciable solutions have qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
been obtained for the abovementioned new analytic method. Also, it then the sign of the square root β2 =4−β3 =27 is negative. In this
was found that, in lower temperature region, the value of γ is exactly case, we split Eq. (A-5) into the two complex terms as follows:
closed to −3 so that the one root of z3 + c1z2 + c2z + c3 = 0 was easily
obtained from Eq. (13). 0 sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi11
3
γ @ β β2 β3 A iθ
¼ − þ − þ ib ¼ re
Appendix A. Analytic extraction of the two formulas in Eq. (5) 2 2 4 27
0 sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi11 ðA 6Þ
3
We start with the following cubic equation as: γ @ β β2 β3 A −iθ
¼ − − − −ib ¼ re :
2 2 4 27
3
y þ Py þ Q ¼ 0: ðA 1Þ