Professional Documents
Culture Documents
Carrier W.D. (2003) - Goodbye Hazen Hello Kozeny-Carman
Carrier W.D. (2003) - Goodbye Hazen Hello Kozeny-Carman
Abstract: The century-old Hazen formula for predicting the permeability of sand is based only on the D 10 particle size. Whereas, the
half-century-old Kozeny-Carman formula is based on the entire particle size distribution, the particle shape, and the void ratio. As a
consequence, the Hazen formula is less accurate than the Kozeny-Carman formula. It is recommended that the former be retired and the
latter be adopted.
DOI: 10.1061/共ASCE兲1090-0241共2003兲129:11共1054兲
CE Database subject headings: Permeability; Sand; Particle size; Particle size distribution; Silts.
Hazen Formula Kaufman 共1962兲 and Mitchell 共1976, 1993兲. Mansur and Kauf-
man do not actually report the values of C H shown above but
About a century ago, Hazen 共1892, 1911兲 developed the follow- instead plot D 10 versus permeability 共based on field tests in sands
ing empirical formula for predicting the permeability 共or hydrau- along the middle and lower Mississippi River兲. Their log-log plot
lic conductivity兲 of saturated sands: 共Fig. 314兲 shows a rough correlation between D 10 and k; for
k⫽C H D 210 (1) 0.01 cm⬍D 10⬍⬃0.07 cm, the implied value of C H varies from
100 to 1,000. Furthermore, for D 10⬎⬃0.04 cm, k is not propor-
where k⫽permeability 共cm/s兲; C H ⫽Hazen empirical coefficient; tional to D 210: at D 10⫽0.1 cm, k⬀D 0.710 . Mitchell does not refer-
and D 10⫽particle size for which 10% of the soil is finer 共cm兲. ence Hazen, and only discusses Kozeny-Carman.
Although Hazen developed his formula for the design of sand None of the geotechnical textbooks referenced herein mention
filters for water purification 关i.e., loose, clean sands with a coef- that the Hazen empirical coefficient corresponds to a water tem-
ficient of uniformity D 60 /D 10 , less than about 2 共Terzaghi and perature of 10°C. Perhaps all of the writers have tacitly concluded
Peck 1964兲兴, it is frequently used to estimate the permeability of that the Hazen formula is so inaccurate that to correct for tem-
in situ soil. perature would be superfluous. Hazen himself used a compound
The value of C H is usually assumed to be equal to 100, but the coefficient equal to C H (0.70⫹0.03 T), where T is the tempera-
following ranges have been reported in geotechnical textbooks: ture in degrees celsius. In his own words, ‘‘I have found that the
• 41 to 146: Taylor 共1948, p. 112兲, friction 关i.e., resistance to flow兴 also varies with the temperature,
• 100 to 150: Leonards 共1962, p. 119兲, being twice as great at the freezing point as at summer heat ...’’
• 100 to 1,000: Mansur and Kaufman 共1962, p. 260–261兲, 共Hazen 1892, p. 553; 1911, p. 200兲. Thus, at 20°, the compound
• 100 to 150: Terzaghi and Peck 共1964, p. 44兲, coefficient is equal to 1.3 C H , and the permeability would be
• 90 to 120: Cedergren 共1967, p. 42兲, 30% greater than at 10°.
• 1 to 42: Lambe and Whitman 共1969, p. 290兲,
• 40 to 120: Holtz and Kovacs 共1981, pp. 209–212兲,
• 50 to 200: Terzaghi et al. 共1996, pp. 73–74兲, Kozeny-Carman Formula
• 100 to 150: Das 共1997, p. 153兲, and
• 80 to 120: Coduto 共1999, pp. 226 –227兲. About a half-century ago, Kozeny 共1927兲 and Carman 共1938,
Thus, the published value of C H ranges from 1 to 1,000. The 1956兲 developed the following semiempirical, semitheoretical
formula’s applicability is generally limited to 0.01 cm⬍D 10 formula for predicting the permeability of porous media:
⬍0.3 cm 共Hazen 1892, 1911; Holtz and Kovacs 1981; Coduto k⫽ 共 ␥/ 兲共 1/C K-C兲共 1/S 20 兲关 e 3 / 共 1⫹e 兲兴 (2)
1999兲.
Nearly all of the geotechnical textbooks cited herein reference where ␥⫽unit weight of permeant; ⫽viscosity of permeant;
the Hazen formula, with two notable exceptions: Mansur and C K-C⫽Kozeny-Carman empirical coefficient; S 0 ⫽specific surface
area per unit volume of particles 共1/cm兲; and e⫽void ratio.
1
When the permeant is water at 20°, ␥/⫽9.93⫻104 1/cm s.
President, Argila Enterprises, Inc., 76 Woodside Dr., Lakeland, FL
关At 10°, ␥/⫽7.64⫻104 1/cm s. Thus, the ratio of 20° to 10° is
33813. E-mail: dcarrier@tampabay.rr.com
Note. Discussion open until April 1, 2004. Separate discussions must
1.3, the same as Hazen used in his compound coefficient. For the
be submitted for individual papers. To extend the closing date by one effect of temperature on the viscosity of water, see Lambe 共1965,
month, a written request must be filed with the ASCE Managing Editor. p. 148兲 or Terzaghi et al. 共1996 p. 73兲.兴 Carman 共1956, p. 14兲
The manuscript for this technical note was submitted for review and reported the value of C K-C as being equal to 4.8⫾0.3 for uniform
possible publication on May 3, 2002; approved on November 26, 2002. spheres; C K-C is usually taken to be equal to 5. Thus, Eq. 共2兲
This technical note is part of the Journal of Geotechnical and Geoenvi- becomes
ronmental Engineering, Vol. 129, No. 11, November 1, 2003. ©ASCE,
ISSN 1090-0241/2003/11-1054 –1056/$18.00. k⫽1.99⫻104 共 1/S 20 兲关 e 3 / 共 1⫹e 兲兴 关 for 20°C兴 (3)
and Kovacs, Terzaghi et al., and Coduto. And yet, the Kozeny- ⫻关 e 3 / 共 1⫹e 兲兴 (10)
Carman formula is known to be more accurate than the Hazen
formula 共e.g., Loudon 1952兲. The explanation for this paradox is
at least twofold: First, most geotechnical engineers do not regu- Hazen versus Kozeny-Carman
larly measure specific surface area 共SSA兲, and, thus, they are not
used to it. 关SSA is usually expressed in m2/g; by also measuring As noted above, the published value of the Hazen coefficient C H
the specific gravity of the soil particles, S 0 in Eqs. 共2兲 and 共3兲 can varies from 1 to 1,000; that is, three orders of magnitude. This
be simply calculated from the SSA.兴 Although soil scientists rou- alone should have disqualified the Hazen formula. Nonetheless, it
tinely measure SSA using nitrogen gas adsorption 共e.g., Hillel has continued to be used because of its simplicity and ease of
1980兲, there is presently no ASTM standard for soil 共there are memorization.
several standards for measuring the SSA of other materials兲. In the past, calculating the effective particle size diameter D eff
Second, none of the geotechnical textbooks referenced herein required for the Kozeny-Carman formula was considered by some
explain that S 0 can be simply estimated from the particle size geotechnical engineers to be time consuming, and this led to at-
distribution. For example, if a soil consists of uniform spheres of tempts to improve the Hazen formula. For example, Amer and
diameter D (cm), S 0 ⫽area/volume⫽(D 2 )/ 关 (D 3 /6) 兴 ⫽6/D. Awad 共1974兲 developed a permeability formula in which k
Thus, Eq. 共3兲 becomes ⬀(D 60 /D 10) 0.6D 2.32
10 关 e /(1⫹e) 兴 ⫽D 60 D 10 关 e /(1⫹e) 兴 . But the
3 0.6 1.72 3
冒 冋兺 册
incorporated into software that calculates and plots sieve analy-
D eff⫽100% 共 f i /D ave i兲 (5) ses, such that the estimated permeability of a sandy layer could be
automatically output. There is no longer any reason not to use the
where f i⫽fraction of particles between two sieve sizes; larger 关l兴 Kozeny-Carman formula.
and smaller 关s兴 共%兲; and
D ave i⫽average particle size between two sieve sizes 共cm兲 Limitations of Kozeny-Carman Formula
冒 再兺
creases, turbulent flow and the inertia term must be taken
k⫽1.99⫻104 100% 冉 关 f i / 共 D li0.5⫻D si0.5兲兴 冎冊 2
共 1/SF2 兲
into account.
Hazen 共1892, p. 554兲 also recognized this: ‘‘For gravels
with effective sizes above 3 mm the friction varies in such a
⫻ 关 e 3 / 共 1⫹e 兲兴 (9) way as to make the application of a general formula very
difficult. As the size increases beyond this point, the 关quan-
Eq. 共9兲 is similar to versions of the Kozeny-Carman formula tity of flow兴 with a given head does not increase as rapidly as
presented in Fair and Hatch 共1933—note the date兲, Loudon, Car- the square of the effective size; and with coarse gravels the
man 共1956兲, and Todd 共1959兲 关interestingly, Eq. 共9兲 does not ap- 关quantity of flow兴 varies as the square root of the head in-
pear in Todd 共1980兲兴. Fair and Hatch suggested the following stead of directly with the head as in sands. The influence of
values for the shape factor, SF: spherical—6.0; rounded—6.1; temperature also becomes less marked with the coarse
worn—6.4; sharp—7.4; and angular—7.7. Loudon suggested the gravels.’’
following values: rounded—6.6; medium angularity—7.5; and Thus, in gravels and coarser soils, a more advanced for-
angular—8.4. mula must be used. As an example, Scheidegger 共1974, pp.
Eq. 共9兲 can be improved slightly by noting that although 165–166兲 and Åberg 共1992兲 use an expression in which the
D ave i⫽D 0.5
li ⫻D si 关Eq. 共6兲兴, S 0 i /SF⫽(1/D i) ave⫽(1/D ave i). As-
0.5
apparent permeability is a function of the hydraulic gradient:
suming the particle size distribution is log-linear between each The maximum value of k apparent is equal to k Kozeny-Carman ,
pair of sieve sizes, it can be shown that (1/D i) ave⫽1/(D 0.404 li which occurs when the hydraulic gradient is zero; and as the
⫻D si0.595). Thus, Eq. 共9兲 becomes hydraulic gradient increases, k apparent decreases.