hr.

J. Solids

Strucrures,

1968, Vol.

4. pp.

109 to 124.

Pergamon Press.

Printedin Great

Britain

ON FIRST

STRAIN-GRADIENT THEORIES LINEAR ELASTICITY

R. D. MINDLIN

IN

Department of Civil Engineering, Columbia University and

N. N. ESHEL
Department of Civil Engineering, The City College, University of the City of New York

Abstract-This paper contains a study of the linear theory of elasticity in which the potential energy-density depends on the gradient of the strain in addition to the strain. In the first part of the paper, three forms of the theory are compared and the relations connecting the stresses in the three forms and the boundary conditions in the three forms are derived. In the second part of the paper, ambiguities in the form of the moment-equation of equilibrium and the definition of couple-stress are resolved by a derivation based on conservation principles rather than the variational principles employed previously.

INTRODUCTION
THE purpose of this paper is to clarify certain aspects of the linear theory of elasticity in which the potential energy of deformation is a function of the six components of the strain and the eighteen components of the gradient of the strain. Stress-equations of equilibrium, constitutive equations and boundary conditions of the strain-gradient theory were first given in a general, non-linear form by Toupin [l, $71. Subsequently [2], linear versions of the theory were given in three forms-distinguished by different groupings of the eighteen additional variables in the potential energy-density: I, the eighteen components of the second gradient of the displacement; II, the eighteen components of the first gradient of the strain; III, the eight components of the gradient of the rotation and the ten components of the fully symmetric part of the second gradient of the displacement (or of the gradient of the strain). The components in the second and third sets are simply linear combinations of those in the first. The third form of the theory is the most convenient one for reduction to the theory in which the potential energy-density is a function of the strain and the gradient of the rotation [l, 3-61. In [2], the three forms of the strain-gradient theory were shown to lead to the same displacement-equations of motion for isotropic materials. However, the general identity of the stress-equations of motion and the general relations among the stresses in the three forms and among the traction boundary conditions for the three forms were not exhibited. These results are derived in the first part of the present paper. The second part of the paper is concerned with the moment equation and the couplestress. In the derivation of the equations of the strain-gradient theory by the variational 109

110

R. D. MINDLIN and N. N. ESHEL

methods employed previously, the moment equation does not appear explicitly. Although a moment equation can be deduced subsequently from the condition of invariance of the potential energy-density in a rigid rotation of the deformed body [7], the equation can be produced in a variety of forms. As a result, the identification of the couple-stress is uncertain to a constant factor. To clarify the situation, the complete equations of the linear strain-gradient theory are rederived, here, starting from principles of conservation of linear momentum, angular momentum and energy. The moment equation and couplestress thereby are displayed without ambiguity. A theorem of uniqueness of solutions leaves the spherical part of the couple-stress undetermined just as in the theory in which the potential energy-density depends on the strain and the gradient of the rotation.

1. KINEMATIC

VARIABLES

The kinematic variables to be employed components of displacement as follows :

Eii =

are defined in terms of derivatives

of

$(uj.i + ui.j) = uoiI = aji = strain,

= rotation,

oij

= $(u~,~- u~,~)= uIj,il = -wji

wi =

+eijkulr,j= vector rotation,

lZijk = u,,~~= Ejik = second gradient of displacement, Rijk = tij = Fijrc =

$(Uk,ji+Uj,ki) YlejlkUk,h = = (k,j)i = Qikj =

gradient of strain,

(1.1)

gradient of rotation (Kii = 0),

= = u(k,ij)

!dUk.ij+ Ejki =

Ui,jk+Uj,ki) Tkij = Pkji

symmetric part of Cijkor Rijk,

where eijk is the alternator. The following relations among the variables are found by eliminating the displacement from the definitions (1.1) :
wij = eijkwk,
1 wi = Teijkcojk,

(1.2)
-

ic"ijk

fiijk

gjki

i?kij = ?ijk+
-

$iileljk

@jlelik,

f?ijk

Tijk -$ijlekil
Wj,i =

$kklejil

$(iiYijk Rikj), +
A

Kij

$f?jk&Okl,iycinejrk = = l-

(1.3)

K,ikejaT

2. EULER

EQUATIONS

AND NATURAL

BOUNDARY

CONDITIONS

The derivations of the three forms of the strain-gradient summarized, here, with some minor alterations.

theory, given in [2], are

On first strain-gradient theories in linear elasticity

111

Hamiltons principle is written for independent ui at times t, and t, :

variations 6ui between fixed limits of

(2.1) where F and w are the total kinetic and potential energies in a volume V :
F =

TdV, sV

YY=
sV

WdV

and SW1 is the variation of work done by external forces. The kinetic energy-density is taken to be T = $&. (2.2)

In [2], velocity-gradient terms were included in the kinetic energy-density. Such terms are appropriate if the strain-gradient equations are r,egarded as a low frequency approximation to the equations of a certain elastic material with a deformable micro-structure [2]. However, if the strain-gradient equations are viewed as a moderately long wavelength limit of the finite difference equations of a simple, crystal lattice, the velocity-gradient terms must be omitted. The latter view is adopted, here, in order to avoid carrying along complicated terms which are not germane to the present study. Three forms are considered for the potential energy-density : W = -all
jir(&ij, Fiji) = ~(Eij, a,) = W(&ij, ~ij, ~ijk)

(2.3)

for the same displacement field. In the case of isotropic materials,

w = +A&ii&jj+
~ =
~~ZEiiEjj +~Ei,~ij+

pEij

+a,Kii$~jj + a&j$ikk (2.4)

+ d3~ii~jjk + a^,Kij~ijk+ d,RijLKkji,

B*RiiklZkjj+CiZlZijjlZikk

(2.5)
+ ci,rZi,Rjj~ + B,rZij,Izij~ +Ci5iZijkRlrji,

where B, = 2a,-4a,, cij = 4ii3, and (2.7) where 18& = -2Et,+4&+&+6B,-3ri,, 3& = 2(ri,+ci,+(i3), 5, = ci,+(i,, 18& = 2ri,-4&-ci3, 3f= ci,+4&--2ci3, (2.8) ci, = -d,+&+d,, ci, = -2ii,+2ii,,

(id = 3a,-ii,,

(2.6)

112

R. D. MINDLIN and N. N. ESHEI.

The variation of the work done by external forces is given a separate form for each form of W :
cJ*; = &g; = si; = &jq (2.9)

where, for W, SK= for W, &; and, for W, S%= sV F,Gu,dV + ss [&Yu, + Qk(Bkj n,nj)bwj + R&,,]dS + PC E,Gu,ds; (2.12) = sV F,Su,dV + sS (&BuL + &DGu,)dS + Y! C &bu,ds ; (2.11) sV F,du,dV +
ss

(&6u, + i?,DGu,)dS +

@u,ds
c

(2.10)

where S is the bounding surface, ni are the components of the outward unit normal to S, C is an edge of S, s is the coordinate along C, D is the normal derivative :
DV = ni(D,i;

(2.13) (2.14) was employed inadvertently in degrees about the normal to S. principle lead to the following and natural (traction) boundary (2.15)

and Enn =
ninj&ij,

In (2.12), Qk(Bkj- n,nj)6wj replaces Qkn,Jwieijk which [2, (12.12)]. The change constitutes a rotation of ninety As shown in [2], the three forms of the variational three sets of Euler equations (stress-equations of motion) conditions : I: ?jk.j- ~ijk,ij+ F, = Pii,,

Pk = nj(zjk fiijk,i)- Dj(ni &j/J + (Dln&injbijk 9 -

d, =

nin+Tiijk,

(2.16)

where the components

of stress, Ti> and double stress, iiijk, are defined by i7W

ij = G =

_
jil

(2.17) (2.18)

the Di are the components of the surface gradient :

DiV = V,i-niDp,

the sP are the components of the unit vector tangent to C and the bold face brackets in the third of (2.16) indicate that the enclosed quantity is the difference of its values, at C,

On first strain-gradient on

theories in linear elasticity

113

the two portions of S that intersect at C. Equations (2.15H2.18), without the acceleration term, are linear forms of Toupins results [l, 571. II :
Pk = It,= 8, =

Qjk,j-fiijk,ij+ Fk = pii,,

(2.19)
3

nj(?jk -Pijk,J- Dj(ni@ijk)(Dln,)ninjfiijk +

WjPijk,

(2.20)

where (2.21)

&k =
III :
q = nj[fjk +$fili,l Qk
ijk,j-~~i,,ije,jk-~ijk,ij+

a%
__ aeijk =
fiikjr

(2.22) Fk = pQ& (2.23)

+ npnqnk&jh

-&&jik =

-&jk,il

-(Dj

njDFJ(Gjk

ni jiij(6,, - njnk) + 2n,?l$tj&jpe,,k,

(2.24)

R = ninjnk&jk,

where
f.. = 11
-

?E i.. = 11'
c7eij

(2.25) (2.26)

pijk = -

aw a;z,,

=
pjki =

=
pkij =

pkji

(2.27)

and p,, = ninj& In [2], ,Gijwas designated the deviator of the couple-stress. If u~k,ij) set equal to is zero in w, fiij does, in fact, reduce to the deviator of the couple-stress tensor ~ij which was defined in [6] by means of the angular momentum principle and appears in the angular momentum equation [6, (1.9)]. That kij remains the deviator of the couple-stress when w depends on utk,ij, is verified in Section 5.

3. RELATIONS

AMONG

THE THREE

FORMS

From (2.3) and the definitions (2.17), (2.21) and (2.2% it follows that the stresses, in the three forms of the theory, are the same :
fij = Qij = fij

(3.1)

114

R. D. MINDLIN and N. N. ESHEL

To find the relations among the double stresses, we first form the relations

ati ac,,, _ __~ a&. aw aW aK,, kik aR,,, - aEp,,, aiqjk +G = G aqjk

(3.2)

Then we replace the derivatives of the energy, in (3.2), with the definitions (2.18), (2.22), (2.26) and (2.27) and we use the relations (1.3) in evaluating the derivatives of the kinematic quantities. The results are
hjk bijk = = #ijk +$jik) = &jk = = Sip* ++&$,jk+$lijleCk, fiijkf+fljle)ik++Pklelijr +PpiJejpp = !dflijk+bjki+@kij).

(3.3) (3.4) (3.5) (3.6)

fiijk+fikij-fijki

@ii iijk

= =

%ipqejp~

!dhjk+?jki+bkij)

A useful, alternative form of the second of (3.3) is

~ijk = $ijk ++i_i$,jk +-&&,ij

(3.7)

From (3.7) and the first of (3.3) we find

L?ijk,ij = fiijk,ij = $&ijeljk +Lijk,ij

(3.8)

Accordingly, the three stress-equations of motion, (2.15), (2.19) and (2.23), are the same and they lead to the same displacement-equations of motion-as shown in [2] for isotropic materials. The relations among the boundary conditions may be obtained by equating coefficients of like variations in (2.10) (2.11) and (2.12). From (2.10) and (2.1 l), it follows immediately that Pk = Pk, R, = R,, To find the corresponding by noting that, in (2.10), Ek = I?,. (3.9)

relations between the first and third forms, we can proceed

&D&k = 2&nJdw,e,jk i- W,njD&j -I-iikn&&,,,,,

2R,nJ6w,e,j, = 2njRk(dli nl?li)dWielj,,

and sS
=

ii,njD,GuJdS = 5

[D,(&nJSuj)[(D,n,)n,&nj-

D&nj)Guj]dS D,(i?,nj)]GuJdS + $ s,[niWjn,]e,,,6u,ds

C

On first strain-gradient

theories

in linear elasticity

115

by the surface divergence theorem. Thus, (2.10) takes the form SR= sV
+

F,Gu,dV +
ss

[& - (Dj - njD,n,)(~jn,)]Gu,dS &&,,dS ss (3.10)

2nj&e,j,(6,i - n,n,)6w,dS + ss (E, + sI[niBjn,]e,ij)Gu,ds.

L! C

Equating coefficients of like variations in (3.10) and (2.12), we have s = p,-(Dj_njDini)(8jnk), QI, = 2niR,eijk, R = niR, E, = E,+ or, in vector notation, P = B-n.Vx(nxftn), 0 = 2nxR, iTi= R-n, E = E+s.[nxfin]. Conversely, P = P++nqVx(Qn) B(= nxWxn+n.Rn) E=E-+s.[Qn]=& The formulas (3.12) and (3.13) show how the same traction boundary conditions can be set in each of the three forms of the theory. The same results may also be found by substituting (3.3H3.8) in (2.16), (2.20) and (2.24). = B, = +Qxn+Rn = 8, (3.13) (3.12)
S,[?liRj?l&,ij

(3.11)

4. MOMENT EQUATIONS The assumption that the first gradient of the displacement enters the potential energydensity only in the symmetric form .sij makes W ab initio invariant in a rigid rotation of the deformed body and, in the variational derivation, precludes the display of a differential equation expressing the equilibrium of moments. Toupin [7] has shown how a moment equation can be found by assuming an energy function

w = W(Uj$

lzijk)

(4.1)

and applying the requirement of rotational invariance subsequently. It is shown, here, how such a procedure can lead to a variety of equivalent forms of the moment equation. To exhibit this result in a simple context, we shall omit time dependence and edges.

116

R. D. MINDLIN and N. N. ESHH

Hamiltons principle then reduces to

s
V

GWdV = sV

F,&

dV +
s s

(&u,

+ ii;DGu,) dS.

(4.2)

Starting with

we find

GWdV s = -~v[(g),i-(g),ijl~ukdv
V

after application of the chain rule and the divergence theorem. Define
@ijk

-,

SW
= G =

_,
pjik?

(4.4) (4.5)

Sjk =

----&jk,i.
k,j

iiW

Then

I
Now,

GWdV = sV

~)k,j&k dV +

njzJksUk

dS +
s S

nifi:j&k,j

dS.

nififj&k,j

dS =

ni &jkDJ&dk

dS +

ss

ninj/IijkD>sUk dS

and nijI;jkDj&k dS = Also, Is Dj(ni~:jkSUk)dS s S

Dj(ni~:i,)6Uk dS.

s
s sV

Dj(niPIjkduk) dS =

(Dlnl)ninjb:jkbuk dS

by the surface divergence theorem for a smooth surface. Assembling these results, we have 8WdV = sV
+ s s [nj+ Dj(ni$ijJ

7;ksJ&kdV+

ninj~lj,DGUk

sS
+ (D,n&injfi~jk]hUk

dS
dS .

(4.6)

On first strain-gradient theories in linear elasticity

117

Upon equating coefficients of like variations in (4.6) and (4.2), we find

Tjk,j+

Fk = 0,

(4.7) (4.8)

PA = nj+ - Dj(nipijJ + (D&rinj&, R; = ninjjiijk,

which are equivalent to Toupins results (10.13) and (10.14) in [7]. The conditions of invariance of the potential energy-density in a rigid rotation of the deformed body are, in the present linear case, ZW ___ = 0. %j,il With (4.9) ,Eijkbecomes iiijk and the definition (4.5) yields
;jk] +Pi[jk],i = 0,

(4.9)

(4.10)

which, allowing for differences in notation, is Toupins moment equation (10.20) in [7]. Now, return to the definition (4.5) and replace it with (4.11) where A is a constant, and note that jIktijl,ij = 0. Then, instead of (4.7) and (4.8) we have
T;k.j+

F, = 0,

(4.12) (4.13)

P; = nj(+ - 2A~;,ij,,i) - Dj(ni&jk) + (D$l&linj&jk,

I?; = ninjjijjk.

In terms of W, (4.12) and (4.13) are the same as (4.7) and (4.8). The invariance conditions (4.9), applied to the definition (4.1 l), produce
T;;kl+

(1 + A)~i2iIjkl,i= 0.

(4.14)

Again, (4.14)and (4.10) are equivalent equations; but derivatives of the potential energydensity are distributed differently between the two terms in each equation. By the first of (3.5)
pi[jk] = 3#ifejkl.

(4.15)

Hence (4.14) can be written as

T;Jk] + ~

1 + A)~i~,iejk~ = 0.

(4.16)

This is to be compared with the moment equation derived from the principle of angular momentum [6, (1.9)] :
T[jk]+hil,iejkr
= 0.

(4.17)

Noting that A is arbitrary and that bij is a deviator whereas pij is not, it is apparent that the invariance conditions (4.9) do not contribute to the identification of the couple-stress.

118

R. D. MINDLIN and N. N. ESHEL

5. DERIVATION

BASED ON CONSERVATION

PRINCIPLES

In this section, principles of conservation of linear and angular momentum and of energy are employed in the derivation of the equations of strain-gradient theory. For completeness, body double forces, with and without moment, are included along with body forces, edges and time dependence. Let ti and rni be the components of force and couple, per unit area, acting on the surface S of a body occupying a volume V ; and let Fi and Ci be the components of force and couple, per unit volume. in V. Then principles of linear and angular momentum are expressed by IsridS +S,&dV = j-v&dV, (5.1)

(xjt,eijk + ~1;)dS + (xjFkeijk + Ci) dV = pXjiike,j~dV. (5.2) sV sV s Application of (5.1) to an elementary tetrahedron leads, in the limit as the tetrahedron shrinks to zero, to the definition of stress, r, such that tj = nirij Substitution of (5.3) into (5.1) and application stress-equations of motion : (5.3) of the divergence theorem lead to the (5.4) leads to the definition of (5.5)

sjk,j+ F, = pii,.

Similarly, application of (5.2) to an elementary couple-stress, p, such that

mj = nipij

tetrahedron

which, with (5.2). yields the moment-equation

pij.i + rkieijk+ Cj = 0

(5.6)

or (5.7) Now, in (5.4), write ~~~= rcjk,+ttjkl Then, with (5.7), (5.4) becomes

In (5.8), separate pij into its deviatoric and spherical parts

~ij = ~~+S~ij~kk.

(5.9)

But the curl of the divergence of a spherical tensor vanishes. Hence (5.8) is (5.10) As for energy densities, we suppose, again, that the kinetic energy-density is
T =

&xiilii.

(5.11) (5.12)

Then i- = piiiici.

On first strain-gradient theories in linear elasticity

119

For the potential energy-density, we assume, again, w = W(Eij, Rij, rijk); so that

VL= ?i,iij + jlijkij + Fijkftijk,

where ii, bij and Pijk are defined, as before, by (2.25), (2.26) and (2.27). We now adopt the following principle of conservation of energy :

(5.13)

s
V

(~+~)dV

=
s V

(Fjlij+Cj~j++~ij,~ij)dV+
s s

(tjlij+mj~tj+ni~ijk~jk)dS,

(5.14)

where the symmetric components @cij, are the densities of body double forces without moment. The antisymmetric part @tiJ1 the body couple per unit volume: ieijkC,. is With (5.3), (5.5), the divergence theorem and the chain rule, the surface integral in (5.14) becomes

But

v (Tij,i~j+~ij,i\ifj+Fijn,i~j~ + 7ijtij.i+pijGj.i+fiij#jk,i) dV.

(5.15)

7ij~j,i =

7cij,fi,j,i,+ 7~ijli41j,il7cij+ij+ zkieijktij =

and

so that (5.15) becomes

s
V

[rij,#j + bij,i + r,ieij&tj + @ok,+ Fij~,i)~j~ ~~kij + Fijti?ijk] dV. +

Hence, with (5.4), (5.6) and (5.12), the principle of conservation of energy (5.14) is converted to (5.16) Finally, inserting (5.13) in the left hand side of (5.16) and equating coefficients of like kinematic variables on both sides of the equation, we find
7(jk) = fjk -Fijk.i @(jk)

(5.17) (5.18)

,u; = /iij

Thus, bij is indeed the deviator of the couple-stress. Upon substituting (5.17) and (5.18) into (5.10), we recover (with the addition of body double forces) the stress-equations of motion (2.23) which were obtained from Hamiltons principle : (5.19)

120

R. D. MINDLIN and N. N. ESHEL

In the case of an isotropic material, F;i is given by (2.7). Then, from (2.25), (2.26), (2.27) and (5.18) + %r = E.S,,E,, Zeus,,, .7.P Fq = 4&?,,+4d,ri,,+~e Pql IJl? =
fiLpqr =
Si*(ziirhpq + F;iipaqr + iziiq6,p) +

(5.20) (5.21)

2Zi*Zpqr (5.22)

Sf~ij(Spqeij,

Sqr@ijp

b,peijq).

contributes to the couple-stress-contrary to the statement in [Note that, in (5.21) Zijijk [2] following (12. l)]. When (5.20), (5.21) and (5.22) are inserted in (5.19) and sij, kij and Fijk are replaced by their expressions in terms of ui, we find the displacement-equation of motion @+2~)(1-I:V)VV. where 1: = (3ti, +2ti2)/(E,+2p), 1: = (32, +a, +2ti2 -f)/3~. (5.24) u-~(l-I;V2)VxVxu+F+)VxC-V.@ = pii, (5.23)

Necessary and sufficient conditions for positive definiteness of W are jl > 0, 6, > 0, 31.+2~ > 0, -2, < d2 < d,, (5.25)

5U, +2ti, > 0,

5f2 < 6(d, -d2)(5ti, +2&,),

which replace (12.18) of [2]. From (5.25) it follows immediately that I: > 0. To show that 1: > 0, note first that [2(c(, +2ti2)-3& -;f]+&J6(d, -d,)(5& +2ti2)-5f2]+yd1ti2 (5.26)

++&f,+d2)(5ti,+2L12)

= (3d,+ti,+2ti2-f)(3d,+4ti,+8ti2+2f).

Since, by (5.25), the left hand side of (5.26) is the sum of positive terms, the right hand side is positive. The two factors on the right must both be positive because, if they were negative, the sum of the second factor and twice the first would be negative-in violation of (5.25). Hence 3d,+ti,+2G2-f and, consequently, 1: > 0. 6. UNIQUENESS OF SOLUTIONS > 0

As deduced from conservation principles in the preceding section, the equations of the strain-gradient theory for isotropic materials are sixty-three in number: (3) (31
(6;

tij.i + Fj = /liij, jjij.i +

T&?ijk + cj = O,

(6.1)
(6.2)

= (jk) =

rjk

-pijk,i

-@(jk)v

(6.3)
(6.4)

(6: (31

&ij

u(j,i),

wj

iUf,,cjk[?

(6.5)

On first strain-gradient theories in linear elasticity Kij =

Kijk = Wj$

121

(6.6)
(6.7) (6.8) feijBTpg4,
+ zppp,dki)+2z2k?ijijk

(k,ij),

sij = Ui,ckk+2/fsij, & = 4d,Kij+4d,Kji+

jijk = 7i,(Eppk6ij+
~~,,,~jk

(6.9)

+3fKJBijep,k

djke,,i

akie,,j).

(6.10)

Whereas there are sixty-three equations, there are sixty-four dependent variables : 3 of Ui, 3 of wi, 6 of Eij, 8 of Rij, 10 of rijk, 9 of rij, 9 of pij, 6 of fij and 10 of The additional variable is the spherical part of the couple-stress which, because it does not contribute to the change of potential energy-density (5.13), is indeterminate within the framework of the theory as represented by the sixty-three equations. With this understanding, we proceed to the proof of a theorem of uniqueness of solutions in the usual manner. Consider two sets of the sixty-four dependent variables ui . . . and uy. . . (with ,L&and ,ui\ arbitrary) and their differences ui = u; - u: . . . Similarly, define body force and body double force differences : Fi = F; - 8; . . . If each set of variables and body forces is a solution of (6.1H6.10), so is the difference set and, from (6.1), we may form the equation
Fijk.

fs s
dt
10 V

(Ti,i+Fj-ptij)tijdV

= 0,

(6.11)

where t, is an initial time and V is a volume bounded by a surface S with an edge C. Now

Hence, with (6.2) and (6.3),

ij,ij = (Tijtij),i++eijk@[k,l+ ck)ib,i] (zij-Fkij,k @~ij&j,iJ ;

or, with (6.4) and (6.Q

Further,

or, with (6.6) and (6.7),

~(lk,l~k+~kij,k~ij =
(Clij~j+Fijkgjk),i-~~kij-Pijk~ijk,

where we have used cii = 0 and the symmetry of Fijk. Also, by (6.8), (6.9) and (6.10), Qij + ,L@, + FijkZijk = W,

122

R. D. MINDLIN and N. N. ESHEL

where W is defined by the same form as (2.7) but in terms of the difference variables. Thus, the integrand of (6.11) becomes

where T is given by (2.2). Accordingly, application integration with respect to time convert (6.11) to

of the divergence theorem and an

(Fjtij+ CjWj+ @~ij,&j) dV

t + s dt ni(Tijtij s + pij@j + /Gijkijk)

10

dS.

(6.12)

Thus, starting with the system of sixty-three equations (6.1)-(6.10), we have recovered the principle of conservation of energy for the difference system. We have now to reduce the twelve variables icj, tij, ijk, in the surface integral of (6.12), to six as only six are independent of each other on S. For example, if the uj are known on S, so is the normal component, njwj, of wj; and, if the uj and the tangential component, (aij-ninj)wj, of wj are known on S, so are all the components of &ijexcept the normal component ninjcij It is also convenient to express the coefficients of ti, (dij--ninj)tij and ninjtij in forms that are independent of tfij, and pii so that the coefficients can be computed directly from the constitutive equations (6.8W6.10). From (6.2) and (6.3),
zjk = = 1 tjk-~~li,leijk-~ijk,i

-$Cieijk - OcjkJ

(6.13)

Also,

and

s
S

nkeki,(pnntij),i dS =
P

c by,]sjtij ds.

Hence (njtjktik + nipi@j) dS =

s
+

nj[fjk +%&

-,&,i)eja

-Pijk,J~k dS

nj+ciejik

@(jk))hk

dS dS + +jj~;~-Js~ti~ ds, (6.14)

ni,$(djk -

njnk)tik

On first strain-gradient theories in linear elasticity

123

in which we have used

PijCBjkwnjnk) = P~lbjk-njnk)3 Pli,l-P(nn,i = Pk-P!h

As for the last term in (6.12), we write ni&Ejk = ni,Gijktik,j= nibijkDjzik+ ninj&+Dtik. But Dir, = 2ti,n,qjk + njD,lij + nkknn Hence

ni&ijk

= (niF+ + n,n,fi,jn,)Djtik + 2n,ninj&jpe,,kG~ + ninjnkFijk~nn

= Dj[(ni& + npn4Fpqjn&] - Dj(ni Eijk+ nf14 &,,&k + 2n,n$Ij~ijpe,,~kk + ninjnkFij&

Then, with the surface divergence theorem, ni&j&jk dS = [(njDlnr - Dj)(nicijk + n,n,~,,jnk)]tik dS dS (6.15)

n,nj(

2n,Fijpeqp~+t,

,&j~n&,,)

P
C

s,[n,ej,(ni,iiijk + nlnink&ij)Jrik ds.

Upon inserting (6.14) and (6.15) in (6.12), we obtain, finally, [T+W]:OdV= sV

dt * (FjUj + CjWj + ~cij,~ij) dV s f0 sV

[Is,@,+ Qk(Bkj n,nj)ij+ I&,,,] dS -+ dt s 10 s S

+

E,ir, ds,

(6.16)

where

pk=
Qk =

nj[?jk

+it/&

-&,i

kja

Fij/c,J

Pj

njDlnJ(niPijk

np,pk

!pqj)

n&$iejik

@tj/c))r

?li&{6j, - nj&) + 2n$Ijn,&jpeqpk,

(6.17)

R = ninjn,jiijk,

It will be observed that (6.17) have the same form as (2.24) with the addition of the body double forces. Note that the body couple C has been placed in the expression for P, following Koiter [S].

124

R. D. MINDLIN and N. N. ESH~I

Conditions sufficient for a unique solution of (6.1H6.10) are now obtained from (6.16) in the usual manner-based on the assumption of positive definiteness of T and W : 1. At each point in V: F, C, Q, and initial values of u and ti. 2. At each point on S : (a) a component of u (or p), in any direction, and the resultant of P (or u) in the plane at right angles; (b) in the tangent plane, a component of w (or Q) and the component of 0 (or w) at right angles; (c) E,, or R. 3. At each point on C: a component of u (or E), in any direction, and the resultant of i? (or u) in the plane at right angles. Note that, according to the linear momentum principle, nj~jk and not P, are the components of force per unit area. Also, by the angular momentum principle, nipij6jk - njn,) and not Q, are the components of tangential couple per unit area. As in classical elasticity, the uniqueness theorem is subject to the regularity restrictions implied by the forms of the divergence theorems employed in the proof. In particular, if a singularity is present, an additional condition, generally obtained through a limit process, is required. Sternberg and Muki [9] have shown how failure to observe such a requirement can lead to physically irrelevant pseudo-solutions.
Acknowledgemenrs-During this investigation, one of the authors (N.N.E.) was a Visiting Research Scientist at the Courant Institute of Mathematical Sciences, New York University. The work was supported by the Office of Naval Research, under a contract with Columbia University, and by the National Science Foundation, under a contract with New York University.

REFERENCES
[l] R. A. TOUPIN, Elastic materials with couple-stresses. Archs ration. Mech. Analysis 11, 385-414 (1962). [2] R. D. MINDLIN, Micro-structure in linear elasticity. Archs ration. Mech. Analysis 16, 51-78 (1964). [3] G. GRIOLI, Elasticiti asimmetrica. Anna/i Mat. puru appl. Ser. IV, 50, 389417 (1960). [4] E. L. AERO and E. V. KUVSHINSKII, Fundamental equations of the theory of elastic materials with rotationally interacting particles. Fizika tverd. Tela 2, 1399-1409 (1960); Translation: Soviet Phys. solid St. 2, 1272-1281 (1961). [5] E. S. RAJAGOPAL, The existence of interfacial couples in infinitesimal elasticity. Annln Phys. 6, 192-201 (1960). [6j R. D. MINDLIN and H. F. TIERSTEN, Effects of couple-stresses in linear elasticity. Archs ration. Mech. Analysis 11,415-448 (1962). [7] R. A. TOUPIN, Theories of elasticity with couple-stress. Archs ration. Mech. Analysis 17, 85-l 12 (1964). [8] W. T. KOITER, Couple-stresses in the theory of elasticity. Proc. K. ned. Akad. Wet. B67, 17-44 (1964). [9] E. STERNBERGand R. MUKI, The effect of couple-stresses on the stress concentration around a crack. 1~. J.
Solids Struct. 3, 69-96 (1967).

(Received 25 July 1967)

A~cTIBBHT-B
3HCpTAK

pa6ore
YilcTH

UCClleJljeTCB J%iHe%HaR TeOpUB yflpyrOCTU,

.I&OSlM~Wf CpaBlViBFiWTCSl H OT ,l&OXMaUHH. TpH BHAtl BHABX. TeOPHH @OTbI

B KOTOpOti rIJlOTHOCTb nOTeH4UaSlbHOk

SSLBHCHMOCTH,

38BHCHT

OT rpaJJHeHTa

B
Bo

IlCpBOti

&i BbIBOAIlTCX

KacaroujWecR

HNIPXXKeHH#

M rPaHHqHblX

YCJIOBHti B 3TWX TWX

xemx

Moraemme

BTOpOfi WCTH pemaIOTCB HeOAH03HaSHOCTH BbdpBXeHHI )WB MOhIeHTHOrO ypaBHeHHn H OnpeAenHanpmrceHne, Hcnonb3yr BALBOAOCHOB~~B~WU&~C~I B 6onbmeft creneBH Ha npwHqHnax
IeM Ha ISipHaUEiOHHbIX llpl?HIUKIaX, KOTOPble HCIlOJIb30BaHbI palI,%.

COXpaHeHWI,