You are on page 1of 11

A numerical study of the 2 and 2 structural parameters for a class of ber-reinforced composites

Theoretical Physics, Royal Institute of Technology, S-100 44 Stockholm, Sweden NADA, Royal Institute of Technology, S-100 44 Stockholm, Sweden

Henrik Christiansson

Johan Helsing

Abstract
Linear elasticity for a ber reinforced two-component periodic composite is treated numerically. Given volume fractions and elastic moduli of the two components, which are the realizable combinations of the e ective moduli? We solve a reduced problem for a system of concentric cylindrical shells arranged in a hexagonal unit cell. The e ective moduli are estimated via thirdorder bounds, which in turn are determined by the structural parameters 2 and 2, trivially bounded by the inequality 0 2; 2 1. Using an interface integral technique, we compute these parameters for hundreds of millions of di erent con gurations. Realizable regions in the 2 - 2 plane, are presented together with cross-sections of the composites that corresponds to points on the boundary of the realizable region. This will highlight the geometric signi cance of 2 and 2 . For composites with roughly the same amount of the two components we realize over 60% of the area 0 2 ; 2 1.

I. INTRODUCTION
The determination of e ective properties of macroscopically isotropic composites is of both practical and theoretical importance. A calculation of the e ective elastic properties of a composite means solving the governing elastic equations and performing certain volume averages of the resulting elds. In general this is di cult. So far, for elasticity, this has been done for simple structures with standard numerical methods such as the FDM (Finite Di erence Method) 1{3] and the BEM (Boundary Element Method) 4]. These methods are in present implementations time consuming which may lead to low precision of the result. Rayleigh multipole coe cients has recently been calculated for a square array of disks which could be used to calculate its elastic properties 5]. A reduced problem, which in recent years has attracted a lot of interest, is the construction of upper and lower bounds between which the e ective properties must lie 6,7]. The 1

bounds depend on the component properties and on structural parameters that contain information about the composite geometry. As the amount of structural information and the order of the bounds is increased, the bounds become tighter. The best third-order bounds currently available for the three independent elastic moduli of ber-reinforced composites, are due to Silnutzer 8] and were simpli ed by Milton 9]. These bounds involve the structural parameters and . These parameters have been calculated for certain structures, using analytical methods 10,11] or exact expressions of the correlation functions 12,13]. For general geometries, Monte Carlo computer simulations 14,15] are used. Monte Carlo calculations are accurate but slow and they are often performed for a few di erent con gurations, which are parameterized by volume fractions. Recently, one of us 16] presented an interface integral formulation for the and parameters of two-component composites. This formulation is well suited for rapid calculations. In this paper we will use this formulation to calculate and for hundreds of millions of di erent ber-reinforced composites. As we shall see, one can reach a large set of - points even for a xed volume fraction. The purpose of this paper is to illustrate the relation between and for various geometries. This should highlight the geometric and physical signi cance of and 9]. In section II we review third-order bounds and relations between the independent elastic parameters of a ber-reinforced composite. In section III expressions for the and parameters are stated as sums over certain con guration parameters, for a hexagonal concentric arrangement of cylindrical shells. Finally, in section IV we present results for the parameters and gures of the cell structures that corresponds to points on the boundary of the realizable region.
2 2 2 2 2 2 2 2 2 2 2 2 2 2

II. BOUNDS AND RELATIONS FOR ELASTIC PARAMETERS


The elastic properties of transversely isotropic ber-reinforced composites are characterized by ve e ective elastic moduli. For a two-component ber material, with isotropic components, Hill 17] showed that there are only three independent moduli. Choosing Te, the e ective transverse bulk modulus, Te, the e ective transverse shear modulus, Ae , the e ective axial shear modulus, as the three independent moduli, one can write the ve elastic moduli in contracted notation 18] C = EAe + 4 Ae Te C = 2 Te Ae C = Te + Te (1) C = ? Te + Te C = Ae : The uniaxial Young's modulus EAe and Poisson's ratio Ae are related to Te through 17] ! 1 ? 1 ; ? (2) E =3 +
11 2 12 22 23 66

Ae

T f
2

T1 ? T2
2 1

T f

Te
2

Ae

= ( ? 2) Te(+ (? )? Te
1 1 2

)? ( ? ) : 3( T ? T )
1 1 2

(3)

Here fi is the volume fraction of component i and

haif = f a + f a
1 1

2 2

(4) (5)

is the volume average for the quantity within angular brackets. Further,
Ti

= i + i=3;

i = 1; 2:

Only two independent elastic moduli are needed for the components since these are isotropic. Milton 7] simpli ed Silnutzer's 8] third-order bounds on Te and Te and fourth-order bounds on Ae for ( ) materials that are statistically isotropic in the transverse plane. These bounds are given by
1 2

AL AL

Ae

AU ;

(6)
2 1 2

+
2

? +2 ? 2
1 1

f1

f2

f1

? ?
+

2 1

!;

(7)

AU

+
1

? +2 ? 2
2 2

f2

f1

f2

1 1

2 2

!;

(8) (9) (10) (11) (12) (13) (14) (15)

TL

Te

TU ;

# f1 f2 (1= T 2 ? 1= T 1 )2 ?1 ; TL = h1= T if ? h1=~T if + h1= i


TU

"

= h

"

# f1 f2 ( T 2 ? T 1)2 ; T if ?

h~T if + h i
TU ;

TL

Te

# f1 f2 (1= 2 ? 1= 1 )2 ?1 ; TL = h1= if ? h1= ~if + "


TU

= h if ?

"

f1 f2(

h ~ if +

? )
1

where = 2h1= T i + h1= i ; 3

2h i h i + h i h i = T h f+ 2 iT f ; T f
2 2 2

(16) (17) (18) (19)

and

hai = hai =
where = 1 ? and
2

1 1

a + 2 a2 ; a + 2 a2;
2 1

1 1

haif = f a + f a ~
1 2 1 1 2

= 1 ? . Crude bounds on 0
2

and

are (20)

1:

The requirement of transverse isotropy of the composite excludes, for instance, the simple square structure to be used in our numerical investigation. A simple structure that is transversely isotropic is the hexagonal array of ber which will be considered in the next section.

III. EXPRESSIONS FOR THE AND PARAMETERS


2 2

The interface integral method 16] is a fast way to compute structural parameters for composites. The implementation is particularly easy when the interfaces are circular surfaces arranged in an array. Then it is possible to write the structural parameters as rapidly converging sums, involving volume fractions and lattice sums. This was done in 16] and applied to a hexagonal suspension of circular ber. The only free parameter for this model is the volume fraction of bers. Using circular shells instead of circular ber makes it possible to increase the number of free parameters and still have the advantage of circular interfaces. Consider N concentric shells, centered in a hexagonal array. Shell k is de ned by the inner and outer radii, Rki and Rko , where k is between 1 and N (N corresponds to the inner most shell). De ne three parameters km, km and m as
km

= (Rko)m ? (Rki)m;
N X l=k+1 km :
2

(21) (22) (23) and


2

km

= (Rki)m ?
m= N X k=1

lm ;

Following the method outlined in 16], the structural parameters expressed as 4

can then be

N X = p2 3f f k
1 2

=1

k2

k(?2) +

1 3f X (6n ? 1)S ; n 2 f
2 1

n=1

2 6

(24)

4 k k ? +3 k k ? ?6 k k k ? = p2 3f f k p X( ) 3f 1 n? n? n? + CD Cn 2 f n 6n ? 3 + n n 3n ? 1 + Dn 6n ? 1 ;
2 1 2 2 ( 2) 2 4 ( 6) 2 4 ( 4) =1 2 1 =1 2 4 2 2 2 12 6 4 12 4 2 12 2 2

N Xn

(25)

where

6n ? 1 S ? 2 Cn = 3 n 3
6

2 4

6n ? 2 T ; n 2
6 6

(26)

6n ? 1 S : (27) Dn = 2 n 2 S n and T n are hexagonal lattice sums that are tabulated with double precision up to n = 9 in 16]. Fast algorithms for lattice sum evaluation are discussed by Berman and Greengard 19].
6 6

IV. RESULTS AND DISCUSSION


We have computed and for three types of shell structures; inclusions consisting of one, two or three shells. By varying the volume fractions and the radii of the shells, we obtained realizable regions in the - plane. The result is presented in Fig. 1. The dash-dotted lines bound the region we realized using one shell. The dashed and solid lines bound the regions we realized using two and three shells, respectively. The lower boundaries of these regions coincide except for a small interval around 0:6. We have found a simple analytic relation between and , valid for the con guration with one shell when f 0,
2 2 2 2 2 2 2 2 2

( )= ?3 +3 :
2 2 2 2 3 2

(28)

Naturally, three shells give the largest set of parameters. Then our realizable region ll 64% of the area 0 ; 1. The achieved improvement going from two to three shells is relatively small. Therefore no shell structures with more than three shells are considered. Torquato 20] observed that and for di erent distributions of d-dimensional spheres obeyed the inequality . Figure 1 shows that this inequality does not hold for the shell structure we have investigated. The shell model presented here makes it possible to x the component volume fraction and still vary the radii of the shells. Figures 2-4 show realizable regions in the ? plane for three di erent volume fractions using the three-shell model. The regions are obtained by calculating the parameters for 13 000 000 di erent shell con gurations, using Eq. (24) and Eq. (25). Nine points are marked with letters on the graphs. The shell structures that correspond to these points are given to the right in the gures. Figure 2 shows the
2 2 2 2 2 2 2 2

realizable region and shell structures for f = 0:004. It is noteworthy, that it is possible to ll 41% of the - plane even with a small volume fraction. The lower boundary is approximately described by Eq. (28) since the cross-sections F-I are one-shell con gurations. Figures 3-4 show regions and shell structures for f = 0:20 and f = 0:46, respectively. For the curves connecting A and E, it seems favorable to have a solid ber surrounded by two almost equally thick shells. The radii of the shells increase, moving from A to E. No general features can be seen in the shell structures F-I. Figure 4 illustrates a relation between structural parameters and inversion of components. Some pairs of points, P and Q, in the graph are almost inverses of each other in the sense that they have (f )P 1 ? (f )Q , ( )P 1 ? ( )Q and ( )P 1 ? ( )Q. The pairs of points are: B and F, C and G, D and H. The physical interpretation of the transformation is an exchange of the components. For example, F is approximately obtained from B by switching black and white in B. Interchanging components will improve on the realizable region. Since the interchange means an inversion in ( ; ) = (0:5; 0:5), an inversion of the lower boundary, which is closer to = 0 than the upper boundary is to = 1, will improve the upper boundary. The realizable region in Fig. 1 will be 72% of the area 0 ; 1. This paper demonstrates the possibility of obtaining di erent combinations of the e ective elastic moduli for a ber reinforced composite. The interface integral technique prove to be well suited for rapid and accurate calculations of structural parameters that enters into bounds on the elastic moduli.
2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2

ACKNOWLEDGMENTS
This work was supported in part by TFR grant 92-691 and 93-859 to Jesper Oppelstrup and by the Center for Computational Mathematics and Mechanics C M at the Royal Institute of Technology. HC gratefully acknowledges support from the Swedish research councils NFR and NUTEK and the Goran Gustafsson foundation.
2 2

REFERENCES
1] D. F. Adams and D. R. Doner, J. Compos. Mater. 1, 152 (1967). 2] C. H. Chen and S. Cheng, J. Compos. Mater. 1, 30 (1967). 3] J. E. Flaherty and J. B. Keller, Comm. Pure and Appl. Math. 26, 565 (1973). 4] J. W. Eischen and S. Torquato, J. Appl. Phys. 74, 159 (1993). 5] R. C. McPhedran and A.B. Movchan, J. Mech. Phys. Solids 42, 711 (1994). 6] Z. Hashin, J. Mech. Phys. Solids 13, 119 (1965). 7] G. W. Milton and N. Phan-Thien, Proc. R. Soc. London Ser. A 380 305 (1982) 8] N. Silnutzer, Ph.D. thesis, University of Pennsylvania, PA, 1972. 9] G. W. Milton, J. Mech. Phys. Solids 30, 177 (1982). 10] R. C. McPhedran and G. W. Milton, Appl. Phys. A 26, 207 (1981). 11] M. Beran and N. Silnutzer, J. Comp. Mater. 5, 246 (1971). 12] C. G. Joslin and G. Stell, J. Appl. Phys. 60, 1607 (1986). 13] S. Torquato and J. D. Beasley, Int. J. Engng. Sci. 24, 435 (1986). 14] S. Torquato, G. Stell and J. D. Beasley, Lett. Appl. Engng. Sci. 23, 385 (1985). 15] J. G. Berryman, J. Comput. Phys. 75, 86 (1988). 16] J. Helsing, J. Mech. Phys. Solids 42, 1123 (1994). 17] R. Hill, J. Mech. Phys. Solids 12, 199 (1964). 18] R. Christensen, Mechanics of Composite Materials (Krieger, Florida, 1979). 19] L. Berman and L. Greengard, Comm. Math. Phys. (submitted 1994). 20] S. Torquato, Appl. Mech. Rev. 44, 37 (1991).

1.0

2 0.5

0.0 0.0

0.5

1.0

FIG. 1 Region boundary for the structural parameter vs the structural parameter for a hexagonal array of concentric shells. Solid lines bound the region using three shells. Dashed lines show the boundary using two shells and nally dash-dotted lines for one shell.
2 2

1.0

D C

2 0.5

G A 0.0 0.0 I 0.5 1.0 H

FIG. 2 Upper and lower boundaries for vs using the three-shell model for f = 0:004. Figures A-I show the cell structures that correspond to the points given in the graph.
2 2 2

1.0

E D C

2 0.5

A 0.0 0.0 I H 0.5

1.0

FIG. 3 Upper and lower boundaries for vs using the three-shell model for f = 0:20. Figures A-I show the cell structures that correspond to the points given in the graph.
2 2 2

10

1.0

E C D

2 0.5
A I

G H

0.0 0.0

0.5

1.0

FIG. 4 Upper and lower boundaries for vs using the three-shell model for f = 0:46. Figures A-I show the cell structures that correspond to the points given in the graph.
2 2 2

11

You might also like