Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
BNFL, Research and Technology, Building 709, Springfields, Preston PR4 0XJ, United Kingdom, and
Department of Engineering, University of Exeter, North Park Road, Exeter EX4 4QF, United Kingdom
theory confirms that the dominant deformation mechanism is flexure of the honeycomb ribs.
The auxetic membrane has been challenged with single-sized glass chromatography beads such
Downloaded via BOCHUM LIBRARIES on November 8, 2019 at 03:46:23 (UTC).
that the beads were initially resting on the re-entrant cells. Subsequent tensile loading of the
membrane showed the auxetic cells opening during deformation, enabling the beads to pass
through the membrane. We have modeled the pore-opening properties of both types of
membranes, and the observed behavior for the auxetic membrane is consistent with the model.
This is a clear proof-of-concept demonstration of the potential of auxetic materials and structures
in filter defouling or cleaning operations. This paper, therefore, demonstrates the successful
design and fabrication of a micromachined auxetic structure having specifically tailored
mechanical properties that show enhanced functional performance over the conventional filter
structure.
2. Membrane Fabrication
(tl)
3
derived elsewhere12,13,24 and are simply quoted here: Kf ) Esw (4)
cos2 R
νxy ) νyx-1 ) (1) where Es is the intrinsic Young’s modulus of the rib
(hl + sin R) sin R material and t is the rib thickness defined in Figure 1.
Similarly, if hinging is considered to be due to shearing
K cos R of the material at the cell wall junction, Kh has been
Ex ) (2) shown to take the form24
h
( )
w + sin R sin2 R
l
h
Kh ) Gsw (tl) (5)
Ey )
l (
K + sin R ) (3) where Gs is the shear modulus of the cell wall material.
w cos3 R The shear mechanism for hinging is expected to be
important for cells where the rib aspect ratio l/t is small.
where Ex and Ey are Young’s moduli due to loading in Note that Poisson’s ratios for honeycomb networks are
the x and y directions, respectively, h, l, and R are the identical when deformation is due to hinging or flexure,
geometrical parameters defined in Figure 1, w is the depending only on the geometrical parameters of the
depth of the honeycomb ribs, and K is the force con- network (eq 1). The expressions for Young’s moduli
stant for the hinging (Kh) or flexure (Kf) deformation associated with the hinging and flexure mechanisms
mechanism. From standard beam theory it has been differ only by the form of the respective force constants.
658 Ind. Eng. Chem. Res., Vol. 39, No. 3, 2000
Table 1. Values for ν and E for Conventional and Auxetic HoneycombssTheory and Experiment
νxy νyx Ex, GPa Ey, GPa
Conventional Membrane
experimental 0.86 ( 0.06 0.6 ( 0.1 0.23 ( 0.04 0.130 ( 0.001
theory (flexure)12 1.4 ( 0.3 0.7 ( 0.1 0.2 ( 0.2 0.12 ( 0.08
theory (concurrent)13,24 1.0 ( 0.2 0.6 ( 0.1 0.15 ( 0.08 0.09 ( 0.05
Re-entrant Membrane
experimental -1.82 ( 0.05 -0.51 ( 0.01 0.127 ( 0.006 0.032 ( 0.006
theory (flexure)12 -2.0 ( 0.2 -0.49 ( 0.05 0.10 ( 0.03 0.024 ( 0.006
theory (concurrent)13,24 -1.8 ( 0.1 -0.44 ( 0.04 0.08 ( 0.02 0.021 ( 0.005
Figure 7. Video stills of re-entrant membrane defouling when challenged by glass chromatography beads: (a) ∼60% bead coverage on
an undeformed membrane; (b) ∼50% bead coverage at a strain of 0.5% in the x direction; (c) ∼40% coverage at a strain of 0.75% in the
x direction; (d) ∼30% coverage at a strain of 1% in the x direction.
()
tion mechanisms act concurrently.13,24 An example of
Ex 1/2
the concurrent model expression is given here for νxy:
|νxy| e (7)
Ey
ablation with specifically tailored mechanical properties. To check the validity of the force constant expressions
Auxetic and nonauxetic structures can be made from (eqs 4, 5, and 9), the concurrent model calculations for
the same material by modifying the geometry of the the membranes considered here are also presented in
structure, for example, by changing R or h/l. Table 1. Again, good agreement is achieved with experi-
5.1. Deformation Mechanisms and Force Con- ment, as expected because flexure dominates and we
stants. In a comparison of the experimental mechanical have already seen that the experimental data are in
properties with theory, it was assumed that deformation good agreement with the flexure model. The concurrent
of the micromachined structures was via flexure of the model predictions are slightly different from those of the
honeycomb ribs. This is vindicated by comparing the flexure model (because of the influence of the hinging
force constants for rib flexure and hinging given in eqs and stretching mechanisms), but the experimental
4 and 5, which leads to errors are too large to enable a definitive statement as
to which model best represents the experimental data.
Kf Es t 2 t2
)
K h Gs l ()
) 2(1 + νs)
l () (8) Note, though, that the experimental νxy value for the
conventional membrane agrees within error with the
concurrent model predictions. This was found not to be
where νs is Poisson’s ratio of the rib material which we true using the flexure model alone.
660 Ind. Eng. Chem. Res., Vol. 39, No. 3, 2000
x ) ln ( )
X
X0
) ln(1 + ex) (11)
δ
ex ) tan R (12)
l
where δ is the deflection of the rib undergoing flexure.
In the flexure model, therefore, strains are calculated
by varying δ, while maintaining l and R constant. This
is applicable to low strains where beam theory26 is valid,
as is the case for the experimental data presented in
Figures 5 and 6.
In the case of the hinging model x is given by
( )
Figure 9. (a) Theoretical (hinging and flexure models) and
cos R experimental x vs y data for the re-entrant honeycomb mem-
x ) ln (13)
cos R0 brane; (b) theoretical (hinging and flexure models) and experi-
mental x vs y data for the conventional honeycomb membrane.
Hinging model ) solid line; flexure model ) dashed line.
where R and R0 are the deformed and undeformed
honeycomb angle, respectively. In the hinging model than for the re-entrant membrane (Figure 9a). However,
strains are therefore calculated by varying R while we note that the calculated uncertainties in the slope
maintaining all other geometrical parameters constant. of the model curves are of the order of 15-20% (see
In Figure 9a we show the comparison of the experi- flexure model νxy and νyx data in Table 1).
mental x versus y data with the flexure and hinging Hence, the experimental strain data agree within the
models for the re-entrant membrane. The two models calculated uncertainties with the flexure model data.
are almost coincidental throughout the entire strain It also appears reasonable to use the rib hinging model
range (0 e y e 0.010). Furthermore, the agreement as a first approximation to model other strain-depend-
with the experimental data is also good, with the models ent properties (e.g., particle transmission) for these
having only a slightly lower gradient than that shown membranes.
by the experimental data. This is of course reflected in 5.3. Strain-Dependent Particle Transmission. In
the excellent agreement between the theoretical and the simple particle transmission tests performed here
experimental values of νyx for the re-entrant honeycomb we have demonstrated the potential of auxetic materials
membrane (see Table 1). or structures in filtration applications, specifically in the
The experimental and theoretical (flexure and hinging cleaning of fouled membranes. This has been achieved
models) x versus y data for the conventional membrane by the application of a uniaxial load or displacement to
are shown in Figure 9b. Again, the hinging and flexure open up the pores in a manner unique to auxetic
models give almost identical results throughout the materials or structures, thereby allowing the passage
strain range considered in Figure 9b. The experimental of glass beads from one side of the re-entrant membrane
trends are also reasonably well reproduced, although to the other. This was not observed in the comparative
there is slightly more discrepancy in the model and tests on the conventional membrane. This may be due
experimental slopes for the conventional membrane to the fact that the conventional honeycomb cells simply
Ind. Eng. Chem. Res., Vol. 39, No. 3, 2000 661
dx
νyx ) - (18)
dy
(d(n/n0)
dx ) (
σx
)
d(n/n0)
dx )
σy
(19)
then, from eqs 17 and 19, the ratio of the slopes of the Figure 11. Averaged normalized fraction of blocked pores vs
strain in the x direction for the re-entrant membrane challenged
two curves in Figure 8 is given by by glass beads. Data for loading in the x (crosses) and y (circles)
( )
directions are shown. Also shown is the calculated curve assuming
d(n/n0) a normal cumulative distribution of glass bead sizes having a mean
value of R/l ) 0.415 and a standard deviation of 0.019. All
dx σx 1 membrane geometrical parameters were as observed experimen-
( )
)- (20) tally.
d(n/n0) νyx
dy σy where [d(n/n0)/dx)]σx is the gradient of the straight line
()-30.00 in Figure 8). Similarly, for loading in the y
Similarly, it can also be shown that the ratio of the direction
( )
slopes is related to νxy by
d(n/n0)
( )
n
d(n/n0) ) y + 1 (23)
n0 dy σy
dx σx
( )
) -νxy (21) Equations 20 and 21 provide conversion between the
d(n/n0)
gradients of the straight lines in Figure 8, defined by
dy σy eqs 22 and 23. Taking the average of the rate of
defouling with respect to strain for loading in the x
In the case of the flexure or hinging models, eq 1 is valid direction given by eqs 20 and 21 (due to the assumption
and eqs 20 and 21 are equivalent. However, for models that the rate of defouling with strain in a given direction
or structures where νxy * νyx-1, the assumption that the is independent of the direction of loading being invalid
rate of defouling with strain in a given direction is in this case; see above), we have
independent of the direction of loading may not be valid.
1
( )
〈( ) 〉 ( )
This is the case for the concurrent model and mem- + νxy
branes discussed in this paper. Furthermore, the mem- d(n/n0) νyx d(n/n0)
brane used in the particle-transmission experiment )- (24)
dx σx 2 dy
contained a small number of damaged ribs, further σy
invalidating the assumption in eq 19. In such cases, eqs
20 and 21 give different values, which we take to be Rearranging eq 22 and substituting eqs 23 and 24 yields
the limiting cases.
Taking the average of eqs 20 and 21 for Poisson’s 2y
x ) -
( )
ratios associated with the membrane during the particle (25)
1
transmission experiment (νxy ) -1.4 and νyx ) -0.18) + νxy
gives a predicted ratio of the slopes for Figure 8 of 3.5, νyx
which is in good agreement with the observed value of
3.2. Hence, the magnitude of Poisson’s ratio is a Equation 25 provides the necessary conversion factor
measure of the sensitivity of the membrane to changes to enable the particle transmission data due to loading
in the pore size. in the y direction to be plotted versus the same strain
To check for consistency in the particle transmission as the data due to loading in the x direction. (Note that
data for loading in the x and y directions, plotted versus if the assumption of the rate of defouling being inde-
the loading strain (i.e., x and y, respectively) in Figure pendent of the loading direction was valid, then νxy )
8, it is instructive to plot the data versus the strain in νyx-1 and so eq 25 would simplify to x ) -νyxy.)
one direction only. To a first approximation each set of This is illustrated in Figure 11 where the experimen-
data in Figure 8 can be fitted by a straight line which tal normalized fraction of blockages (n/n0) are plotted
in the case of the x-loaded data is given by versus x for uniaxial loading along both the x and y
directions. In Figure 11, the data due to loading along
n
n0
) (
d(n/n0)
dx )σx
x + 1 (22)
the x axis (crosses) are as given in Figure 8. The data
due to loading along the y axis (circles) have also been
included by using eq 25 to convert the strain data. Using
Ind. Eng. Chem. Res., Vol. 39, No. 3, 2000 663
this method, it is readily seen that the data for loading bending moments leading to curvature (“bowing”) of the
along the x and y directions are consistent, because they membrane in the direction of flow. It is known that
now overlap once the effects of Poisson’s ratios and materials and structures possessing a negative in-plane
loading direction have been taken into account. Poisson’s ratio undergo synclastic curvature when sub-
For a bead/membrane system consisting of beads of ject to a bending moment, whereas those possessing a
exactly identical shape and size, and a membrane positive in-plane Poisson’s ratio undergo anticlastic
having exactly identical pore structure and rib geometry curvature.4,10 In other words, auxetic materials and
throughout the membrane, it might be expected that structures can be deformed into a doubly curved “dome”
complete defouling would occur at the instant the pore shape more easily than nonauxetic materials and struc-
size exceeds the bead diameter. In reality, of course, the tures. Such deformation will lead to an increase in pore
beads will have a distribution of sizes, and there will size which, as we have demonstrated in this paper, may
be some variation in the pore geometry across the be used for membrane defouling or for reducing a
membrane, leading to particle transmission occurring pressure increase during fouling.
across a range of strains, as observed in Figure 8. However, it should be noted that the bowing action
Therefore, we have modeled the particle transmission due to pressure build-up could also introduce detrimen-
through the re-entrant membrane assuming a normal tal effects such as the release of initially trapped
distribution of particle sizes going through pores of a particulates into the downstream. The transmission of
particular size governed by eq 16 for the membrane particles through the filter due to bowing may also
geometry considered in this paper. In the calculations, hinder the establishment of the more effective filtration
therefore, those particles in the distribution having a phase where the initial deposit acts as part of the
radius less than that determined from eq 16 are filtering media to trap finer particles.
considered to be transmitted through the membrane. In the case of thick filtering media then the bowing
Particle transmission was calculated for varying R (i.e., mechanism may not be applicable. In this case the
assuming membrane deformation is due to rib hinging). pressure build-up across the filter would tend to close
The normalized fraction of blockages calculated in this the pores which would oppose the effect of tensile lateral
way are also plotted in Figure 11 (solid curve), using forces that may be applied to open up the pores for
eq 13 to convert the variations in R to x. defouling purposes. Clearly, the anticipated forces due
The model curve is normalized to n/n0 ) 1 at the to pressure build-up need to be considered in the design
undeformed geometry to ensure consistency with the and selection of the appropriate filter material and
experimental data. A normal cumulative distribution of control system to ensure satisfactory defouling occurs
particles having a mean value of R/l ) 0.415 and a in reality.
standard deviation of 0.019, with all other membrane The pore-opening properties characteristic of auxetic
geometrical parameters as observed experimentally, materials and structures may also be of benefit in
gives the best fit to the experimental data in Figure 11. applications where, for example, tuneable pore size
The mean value of R/l employed in the calculations is, variations or oscillations are required, or where dynamic
therefore, in excellent agreement with the value of R/l alterations of pore size are needed.
) 0.40 ( 0.04 calculated using eq 16 from the experi- Auxetic materials and structures have potential in
mentally determined geometrical parameters for the smart filtration applications requiring both active and
undeformed membrane, which we have already seen passive control of the pore size. Future work will involve
possesses a pore size which corresponds well with the the use of alternative fabrication techniques to enable
size of beads used in the test. auxetic structures having smaller pore sizes to be
In this paper the potential of employing auxetic fabricated and tested. For example, the LIGA (Lithog-
materials in filter defouling operations has been con- raphie, Galvanoformung, Abformung) micromachining
sidered under uniaxial loading conditions. Of course, technique, which employs lithography, electroplating,
biaxial loading operations may be possible in certain and moulding processes to manufacture finely defined
filtration applications. Tensile loading in either the x microstructures for use in microelectronic and micro-
or y direction leads to an increase in pore dimensions filtration components, for example, has been used to
of an auxetic membrane, and so the effects are additive fabricate conventional honeycomb structures having
in the biaxial loading case, leading to further enhanced pores up to 2 orders of magnitude smaller than the
pore-opening possibilities. However, for a conventional membranes fabricated and tested here. Hence, this
membrane the effects due to tensile loading in each technique enables the possibility of auxetic membranes
direction oppose each other from a pore-opening point for use in practical air filtration systems to be consid-
of view (i.e., one tends to open the pore in the y direction ered. Scaled-down auxetic honeycombs may also find
while closing the pore in the x direction and vice versa). use in biomedical applications such as ultrafiltration
This is demonstrated by considering eq 20 or 21 where and drug-release membranes where the filtering or
positive Poisson’s ratio values lead to a negative ratio release of a clearly defined dose of similarly sized cells/
of the slopes of the defouling with strain curves for molecules is required.
loading in the x and y directions, indicating that The fabrication, characterization, and testing of aux-
defouling may be possible by loading in one direction etic forms of more realistic filter materials than the
but not the other. Hence, once again we expect auxetic membrane studied here for the proof-of-concept dem-
materials or structures to offer improvements in filter onstration are also now required. Auxetic forms of
defouling operations. polymeric foams used in air filtration are known,4 which
An alternative mechanism for effecting pore variation are obvious choices for study in this respect. Auxetic
may be as a result of the passive response of a foams could also be used to support auxetic membranes,
membrane subject to a pressure-drop build-up across similar to the one studied here, for large-scale filtration
the membrane as a result of membrane fouling. This operations. In this case the support would be required
may be considered equivalent to the application of to be highly deformable and have a larger effective pore
664 Ind. Eng. Chem. Res., Vol. 39, No. 3, 2000
size and matching in-plane Poisson’s ratios with respect νxy ) Poisson’s ratio for uniaxial loading along the x
to the membrane for the auxetic membrane to exhibit direction
the full benefits in filter defouling and size selectivity νyx ) Poisson’s ratio for uniaxial loading along the y
demonstrated here. Auxetic metallic foams4 should also direction
be useful in this respect, particularly in applications σx ) applied stress along the x direction
where increased support stiffness is required. Alterna- σy ) applied stress along the y direction
tively, deeper honeycomb monoliths having larger ef- τp ) laser pulse length, fs
fective pore sizes could be used to support honeycomb
membranes. Because of the scale-independent nature
of the auxetic effect, the analytical models described in Literature Cited
this paper would enable a close match of the in-plane (1) Evans, K. E. Tailoring the Negative Poisson’s Ratio. Chem.
mechanical properties of the membrane and the mono- Ind. 1990, 20, 654-657.
lith to be achieved in this scenario. (2) Simmons, G.; Wang, H. Single-Crystal Elastic Constants
and Calculated Aggregate Properties: A Handbook, 2nd ed.; MIT
Press: Cambridge, MA, 1971.
6. Summary (3) Love, A. E. H. A Treatise on the Mathematical Theory of
Elasticity, 4th ed.; Dover: New York, 1944.
In summary, we have used femtosecond laser ablation (4) Lakes, R. S. Foam Structures with a Negative Poisson’s
to fabricate micromachined auxetic and nonauxetic Ratio. Science 1987, 235, 1038-1040.
polymeric membrane structures. The structures have (5) Caddock, B. D.; Evans, K. E. Microporous Materials with
mechanical properties in good agreement with the Negative Poisson’s Ratios: I. Microstructure and Mechanical
flexure model for the deformation of honeycombs, and Properties. J. Phys. D: Appl. Phys. 1989, 22, 1877-1882.
also in good agreement with a more detailed model in (6) Alderson, K. L.; Evans, K. E. The Fabrication of Microporous
which deformation due to rib hinging and stretching are Polyethylene Having a Negative Poisson’s Ratio. Polymer 1992,
33, 4435-4438.
also incorporated. The experimental data are consistent
(7) Yeganeh-Haeri, Y.; Weidner, D. J.; Parise, J. B. Elasticity
with both models because of flexure being the dominant of R-Cristobalite: A Silicon Dioxide with a Negative Poisson’s
mechanism in the real structures. Simple tests have Ratio. Science 1992, 257, 650-652.
been performed, demonstrating that auxetic membranes (8) Alderson, K. L.; Pickles, A. P.; Neale, P. J.; Evans, K. E.
have beneficial properties over conventional membranes Auxetic Polyethylene: The Effect of a Negative Poisson’s Ratio
in filter defouling and controlled pore size applications. on Hardness. Acta Metall. Mater. 1994, 42, 2261-2266.
The strain-dependent defouling is seen to be dependent (9) Lipsett, A. W.; Beltzer, A. I. Reexamination of Dynamic
on the value of Poisson’s ratios of the membrane. Problems of Elasticity for Negative Poisson’s Ratio. J. Acoust. Soc.
Am. 1988, 84, 2179-2186.
(10) Evans, K. E. The Design of Doubly Curved Sandwich
Acknowledgment Panels with Honeycomb Cores. Compos. Struct. 1991, 17, 95-111.
(11) Alderson, A.; Evans, K. E.; Rasburn, J. International
This work has been funded by BNFL and the Engi- Patent No. WO 99/22838, May 1999.
neering and Physical Sciences Research Council of the (12) Gibson, L. J.; Ashby, M. F. Cellular Solids: Structure and
United Kingdom. Properties; Pergamon Press: Oxford, 1988.
(13) Evans, K. E.; Alderson, A.; Christian, F. R. Auxetic Two-
Dimensional Polymer Networks: An Example of Tailoring Geom-
Nomenclature etry for Specific Mechanical Properties. J. Chem Soc. Faraday
Trans. 1995, 91, 2671-2680.
ex ) engineering strain along the x direction (14) Jacobs, P. F. Rapid Prototyping and Manufacturing:
Es ) intrinsic Young’s modulus of honeycomb rib material, Fundamentals of Stereolithography; Society of Manufacturing
Engineers: Dearborn, MI, 1992.
GPa
(15) Aksay, I. A.; Groves, J. T.; Gruner, S. M.; Lee, P. C. Y.;
Ex ) Young’s modulus in the x direction, GPa Prud’homme, R. K.; Shih, W. H.; Torquato, S.; Whitesides, G. M.
Ey ) Young’s modulus in the y direction, GPa Smart Materials Systems Through Mesoscale Patterning. SPIE
f ) flat field lens focal length, mm 1996, 2716, 280-291.
Gs ) intrinsic shear modulus of honeycomb rib material, (16) Larsen, U. D.; Sigmund, O.; Bouwstra, S. Design and
GPa Fabrication of Compliant Micromechanisms and Structures with
h ) vertical honeycomb rib length, mm Negative Poisson’s Ratio. Proceedings, IEEE Micro-Electro-
Kf ) flexure force constant, N/m Mechanical Systems; IEEE: New York, 1996; pp 365-371.
Kh ) hinging force constant, N/m (17) Gower, M. C.; Rumsby, P. T.; Thomas, D. T. Novel
Applications of Excimer Lasers for Fabricating Biomedical and
Ks ) stretching force constant, N/m
Sensor Products. SPIE Excimer Lasers 1993, 1835, 133-142.
l ) diagonal honeycomb rib length, mm
(18) Dyer, P. E.; Farley, R. J.; Giedl, R.; Karnakis, D. M.
n ) number of blocked pores in the field of view Excimer Laser Ablation of Polymers and Glasses for Grating
n0 ) initial number of blocked pores in the field of view Fabrication. Appl. Surf. Sci. 1996, 96-98, 537-549.
R ) radius of inscribed sphere, mm (19) Kuper, S.; Stuke, M. Ablation of UV-Transparent Materials
t ) honeycomb rib thickness, mm with Femtosecond UV Excimer Laser Pulses. Mater. Res. Soc.
w ) honeycomb rib depth, mm Symp. Proc. 1989, 129, 375-383.
X ) deformed unit-cell length along the x direction, mm (20) Kruger, J.; Kautek, W. Femtosecond-Pulse Visible Laser
X0 ) undeformed unit-cell length along the x direction, mm Processing of Transparent Materials. Appl. Surf. Sci. 1996, 96-
98, 430-438.
R ) honeycomb angle, deg
(21) Kumagai, H.; Midorikawa, K.; Toyoda, K.; Nakamura, S.;
R0 ) undeformed honeycomb angle, deg
Okamoto, T.; Obara, M. Ablation of Polymer Films by a Femto-
δ ) deflection of honeycomb rib due to flexure, mm second High-Peak-Power Ti Sapphire Laser at 798 nm. Appl. Phys.
x ) true strain along the x direction Lett. 1994, 65, 1850-1852.
y ) true strain along the y direction (22) Strickland, D.; Mourou, G. Compression of Amplified
νs ) intrinsic Poisson’s ratio of honeycomb rib material Chirped Optical Pulses. Opt. Commun. 1985, 56, 219-221.
Ind. Eng. Chem. Res., Vol. 39, No. 3, 2000 665
(23) Alderson, K. L.; Alderson, A.; Evans, K. E. The Interpreta- (27) Beatty, M. F.; Stalnaker, D. O. Poisson Function of Finite
tion of the Strain-Dependent Poisson’s Ratio in Auxetic Polyeth- Elasticity. J. Appl. Mech. 1986, 53, 807-813.
ylene. J. Strain Anal. 1997, 32, 201-212.
(24) Masters, I. G.; Evans, K. E. Models for the Elastic Received for review August 2, 1999
Deformation of Honeycombs. Compos. Struct. 1997, 35, 403-422.
Revised manuscript received November 23, 1999
(25) Lempriere, B. M. Poisson’s Ratio in Orthotropic Materials.
Am. Inst. Aeronaut. Astronaut. J. 1968, 6, 2226-2227. Accepted November 24, 1999
(26) Roark, R. J.; Young, W. C. Formulas for Stress and Strain,
5th ed.; McGraw-Hill: New York, 1976. IE990572W