Beruflich Dokumente
Kultur Dokumente
hit. J, SolkA StructuresVol. 32. No. 22, pp. 3327 3340, 1995
Copyright I 1995ElsevierScienceLtd
Printed in Great Britain. All rights reserved
0020 7683/95$9.50 + .00
Pergamon
H. X I A O
D e p a r t m e n t o f M a t h e m a t i c s & I n s t i t u t e o f M a t h e m a t i c s , P e k i n g University, Beijing 100871,
China
(Received 17 M a y 1994; in revised f or m 20 November 1994)
A b s t r a c t - - T h i s p a p e r p r o v i d e s simple a n d unified explicit basis-free expressions for time rate and
c o n j u g a t e stress o f a n a r b i t r a r y Hill's strain for the first time, which are valid for all cases o f the
eigenvalues o f the right stretch tensors.
1. I N T R O D U C T I O N
Let U be the right stretch tensor and {2i} and {Ni} be the eigenvalues of U and the
subordinate orthonormal eigenvectors, respectively. The following class of strain measures
is known as Hill's strains [cf. Hill (1968, 1978)] "
d
d = 3,
(la,b)
i=1
where f ( - ) is a strictly-increasing scalar function satisfying f ( l ) = 0 and f ' ( 1 ) = 1 [henceforth we shall assume that f ( . ) is at least of C~]. The following subclass of Hill's strains,
indexed by the parameter m, is known as Seth's strains [cf. Seth (1964)]"
1 3
l(um
E(m) = m ~ (2,.'- 1)N, N, = ---I).
a
FF/
(2)
Hill's strains, even Seth's strains, are broad enough to include almost all the commonly
used Lagrangian-type strains such as the nominal strain E ~1), Green's strain E ~2), Almansi's
strain E ~-2) and the logarithmic strain E ~) = In U, etc.
On the other hand, by means of the notion of work conjugacy, introduced by Hill
(1968) and Macvean (1968), a class of stress measures may be derived in a natural way.
Let E be a Lagrangian-type strain. A symmetric second order tensor T is the conjugate
stress of the strain measure E if T : 1~ offers the stress power ~i, per unit reference state
volume, i.e.
~=IIIa'D=T:I~,
(3)
where a, D and III are the Cauchy stress, the stretching tensor and the third principal
invariant of U, respectively.
The conjugate stresses o f E (j), E ~2) and E (-2) a r e well known [cf. Hill (1978) and Guo
(1984)] : they are the second Piola-Kirchhoff stress tensor,
T (2) = I I I F - l a F -v,
the weighted convected stress tensor,
3327
(4)
3328
H. Xiao
T (2) = IIIFT~rF,
(5)
(6)
In a deforming body, the Lagrangian triad {Ni) rotates with respect to a fixed orthogonal triad. The angular velocity of {Ni} can be described by a skew-symmetric second order
tensor ~ , i.e.
l~lg = ~Ni.
(7)
~ , called the twirl tensor by Hoger (1986) and Guo et al. (1992), will be used in the next
section. If we can justifiably differentiate the spectral representation
3329
(8)
U = 2 iNi@Si
i=l
N,+2~Ni (g~N~))
[~J = Z (,(~Ni @ N i + 2 , ( n N , )
i=l
3
i.e.
3
nU-Un
= U-
~ 2,N~ N,.
(9)
i=1
U= ~. 2~P(7; E= ~ f(2.)e~,
(7=1
(lOa,b)
a=l
with
P~P~ = 6(7~P~(3~ is the Kronecker delta)
Po = I,
(11)
(12)
tr=I
where I is the second order identity tensor. The eigenprojections are expressible in terms of
U and its eigenvalues. In reality [cf. Luehr and Rubin (1990)],
P~= fi U-2,I
hence,
m > 1;
P~=I,
m = 1;
(13)
H. Xiao
3330
1
m-
P~ = -
I "m
(14)
I r,
r ~J
Par=0
IP*=*l-II(2a-;~O~O;F=l;
m - l ' z ~andfrs=l
=
(I,
(-I)'
~.
1 <~1
.....
(15)
2~,... 2~,(I - - ~ o , ) . . . (i - - 5 , , ) .
<'"<e.~<~m
(16)
a=l
(17)
P,(U-,=~,
)o,P,)P~ = P ~ ( f l U - U f ~ ) P ,
= O,
r=l
. . . . . m,
i.e.
J,~P~ = P~UP~,
o = 1. . . . . m.
(18)
By m e a n s of eqn (1) it can be readily proved that the above m conditions are equivalent to
the following"
o-=
cy= 1
The latter is the necessary and sufficient condition for the existence of the solution of the
tensor equation (16). We have the following result.
Theorem 1. The tensor equation (16) has a solution iffeqn (19) or (18) holds and under
this condition the general solution ~ e skw of the tensor equation (16) is of the following
form :
P~WP~ + ~ ~
a=
P~UP.
V W G skw.
(20)
~r~:r
VX6Lin.
3331
(21)
= ~
or,r=
(22)
cr,~
{ P o I ~ P ~otherwise
,a=a',r=r'O,
P,~ B] P~ = 1 [] I,
ry.r=
we conclude that the right-hand side of eqn (22) offers the spectral representation of the
second order tensor over Lin, I [] U - U [] I, and hence this tensor is a symmetric secondorder tensor. The tensor equation (16) can be written into the following vector equation
over Lin :
(I[]U-U~I)'~=
(23)
7=I
Hence, the tensor equation (16) has a solution iff C is contained in the image space of
= I [ ] U - U [ E ] I, i.e. CeIm(fJ). Since the image space Im(l~) and the kernel space
Ker(U) of the second-order symmetric tensor fJ is orthogonal and Lin = Ira(U) @ Ker(U),
as well as the orthogonal projection of Ker(U), is Y~' ;P~ [] P~, we infer that C e Im(fJ)
iff C is perpendicular to the kernel space Ker(U), i.e.
(P~ ~ P~)-C = O,
ff=l
i.e. eqn (19) holds. Moreover, since the restriction Ul~m(0~ : Im(l~) ~ Ira(U) is a nonsingular
linear transformation over Im(U) and by means of the spectral representation (22) its
inverse is readily available, hence eqn (16), i.e. eqn (23), has a unique solution contained
in the image space Im(U), given by
~'~Im ( E
Thus, the general solution ofeqn (16), i.e. eqn (23), is as follows"
(24)
3332
H. Xiao
Remark. Equation (20) offers the general solution of the tensor equation (16) in explicit
form, valid for all cases of the eigenvalues of U. This result, as it will be seen, plays a key
role in the succeeding discussion. As far as we know, it is given for the first time.
Remark. If U has three distinct eigenvalues, then for each W ~ skw,
3
PoWP. = ~ (N~WN,)N~ N, = 0.
~r~l
(25)
i=1
Thus, the solution of eqn (16) is unique and hence eqn (20) offers the required twirl tensor.
However, if U has repeated eigenvalues, eqn (25) no longer holds and hence eqn (16) has
an infinite number of solutions. Guo et al. (1992) have pointed out that the skew-symmetric
tensor f~lm may be justifiably taken as the twirl tensor for the case when U has repeated
eigenvalues. In our succeeding analysis, the multivaluedness of the solution of eqn (16) is
of no consequence and hence such further considerations are not essential, so we do not
bother to list further related results. Moreover, we obtain eqn (9) by differentiating eqn (8)
with respect to time and by using eqn (7), the defining equation for ~ . There remains the
question of whether and under what conditions we may justifiably proceed as indicated.
This problem has been solved by Guo et al. (1992).
3. TIME RATE OF AN ARBITRARY HILL'S STRAIN
Carlson and Hoger (1986) and Man and Guo (1993) provided distinct basis-free
expressions for the time rate of an arbitrary Hill's strain for distinct cases of the eigenvalues
of U [cf. Wang and Duan (1991) and Scheidler (1991b, 1992)]. In this section, we provide
new and unified basis-free expressions, which are valid for all cases of the eigenvalues of U
and from which the conjugate stress of an arbitrary Hill's strain can be derived immediately.
By differentiating eqn (1) and by using eqns (7) and (10a), we obtain
=
J (2.)2.P. + f ~ E - EFL
(26)
o'=1
(27)
P, W P , E - E
=
\a=
P,WP~
a
= 0,
V W ~ skw.
(28)
Hence, by substituting eqn (20) into (27) and then using eqns (28) and (11), we get
!~ =
f (2,)PoUP, +
a=l
= ~
Denoting
f'(2,~)P,,ldP~,+ ~. f(2~2*)P,,IJP~.
f(2o)Po
0
(29)
for -
3333
(30)
f o r a , z = 1. . . . . m,
where for a = r = 1. . . . . m, the limiting process 2~ ~ 2~ is meant, we m a y recast the preceding expression into
1~ =
~ f,~Pfl]P~ = L(U):IU
(31a,b)
O',Z = 1
L(U) =
~ f , , P , [] P,.
(32)
o,T~ 1
Finally, substituting eqn (13) into (31a), we obtain the following result.
Theorem 2. T h e time rate o f an arbitrary Hill's strain possesses the following basisfree expression"
m-I
~"
g~,=
E '~rsUr~Jus
r,s= 0
(33)
/" _ - I _P~
- 1 i om _ , _ ~ I*, . _ l _ s = e s . r , s = O , . .
J,~P.
,m--1.
(34)
Ir,~ = I
23 ~ '~1"
~22 U 2 1 ~ J U 2 -{-/~01 ( U O - - { - ~ J U )
+eo2(U2(d+f3U2)+g,2(U(dU2 +U2(3U)
(35)
2III 2
+-X-- 12.23,31
E
glt
=--
1
A2
(2~2,)-'(2~--2.)-2(f(k~)-f(2~))
(36)
(37)
12,23,3 I
2F
I 2,23,31
e22=~
~
12.23.31
(2o--2,) ( f ( 2 , ) f ( 2 J )
-1
2 E
~ + X12.23,3, (20 - 2,)- 2(f(2,,) - f ( 2 , ) )
(38)
~01
A2 I 2,23,3
E I 2~2, (2o + 2,) (2~ -- 2~) 2 ( f ' ( 2 ~ ) f ' ( 2 , ) ) - '74
- - - -
III
E
12,23,31
(2+III2922g2(2o+2,))(2,--2~)-2(f(2,,)--f(2,))
(39)
3334
H. Xiao
1
802
--
A2 12,23,31
2j'2~'(2~+2~)(2~-2~)2(f(2.)-f(2~))
(40)
12,23.31
1
gl2 ~
----
A2 12,23,31
--~
(2.+2,+2III2g'2;')(Z.-2~)-2(f(2.)-f(2,)),
(41)
12,23,31
where
~"
J.~ = J,2
+J23 +J3l
and
12.23.31
~--- ( A 1 - - 2 2 ) ( 2 2 - - 2 3 ) ( 2 3 - - 2 1 )
{i1
.=f'(2,)f'(22)f'(23)
(42)
(2,+22)(22+23)(23+2,) = I I I - I I I
(43)
too
1
2~/'(22)+2~e'(21)
(2,_22)2
22,22
2~Z~ 2 (/(Z,)-f(Z2))
1
g,l -(2, -22)2(f'(2,)+f'(22)-2f(~!-~2(222)
)
eo, - (2,
--
22)
21
2,f'(i2)+2zf'(2,)
Z1+22
)
2,--22(f(2')-f(22)) .
(44)
(45)
(46)
(47)
(48)
Let T be the conjugate stress of the Hill's strain E. According to the notion of work
conjugacy [cf. eqn (3)], we have
T:I~ = T ~1) :10.
(49)
Here, the Jaumann stress T m, given by eqns (4) and (6), is conjugate to the nominal strain
E ) = U - - I .
Substituting eqn (31) into (49), we get
3335
(50)
(51)
a,z=l
(52)
. . . . . m.
L(U) =
~' f . , P , [] P,
0",1:=
is a nonsingular symmetric second-order tensor over Lin. The crucial point is that the righthand side of the above formula is just the spectral representation of L(U) (cf. the p r o o f for
Theorem 1). From this fact and eqn (51), we derive the conjugate stress of E immediately:
T = L ( U ) - I .T(l) =
f 1 p , [] p,
a,
:T(l)
~.~ f~,lp,,T<l)P,.
(53)
a,~ = 1
Substituting eqn (13) into the above expression we get the following result.
Theorem 3. The conjugate stress of an arbitrary Hill's strain possesses the following
basis-free expression :
m-I
T = ~" P r s U r T ( l ) U s
(54)
r,s = 0
Prs =
~, f gtp~lp~-l I~m-1. . P. .
1--s
(55)
-~- P s r "
#,,r = 1
It can be seen that the coefficients Prs can be obtained merely by replacing the coefficients
f ~ on the right-hand side of eqn (34) w i t h f g I . Specifically, we have the following.
(i) U has three distinct eigenvalues, i.e. m = 3 ; ;-t 4:22 4:~'3
# '~1"
21112
A 2 12,23,31
2~2~(2*--2")7'(2~)f'(202i'-~ - - - A
(56)
1
Pll
A2 12,23,31
Y~
2F
Z
1
l + ~ - - ' 2,23,31 ('~* + '~*) -- OC(J"r) - - f ( 2 , ) )
--1
(58)
H. Xiao
3336
----
A2 1 2 , 2 3 . 3 1
(59)
A 1 2 ,22 3 , 3 1
(60)
P02
--
(61)
I 2.23,31
P12
- - -
- ~
(20+2~+2III2,-'2U')(f(2n)--f(2,))-',
(62)
12.23.31
where Xt, A and F, I, II and III are given by eqns (42) and (43).
(ii) U has only two distinct eigenvalues, i.e. m = 2 ; 2~ 4: 22.
T = PooT(t) + p t iUT(J)U+pol (UT (l) +T(I)U)
(63)
(65)
f(2,)-f(x2))
')
-'
Po~ - ( 2 , ~ 2 ) 2 \
(64)
- :(--(TS-,)-f~ )
/(2,)'f(22)
(66)
(iii) U = I, i.e. m = 1.
1
--T
T = PooT (') = f'(2)
(I) "
(67)
5. E X A M P L E S
As an example, we derive the basis-free expressions for the time rate and the conjugate
stress of the logarithmic strain E () = In U by using the results given in the previous sections.
We have
f(2)=ln2;
f'(2)=2-'.
(68)
eO0
3337
1
~ 2`'2,(2,;--2~)
3 3
2 4_ _21112
_
~ (2,'2,)-1(2`--2,)-21n_~
IIIA 2 12,23,3 I
A 12,23,31
(69)
_ 1
Z 2 ` ' 2 , ( 2 ~ - - 22 , 2) + ~2- F
el l -- IIIA 2 12,23,31
/~22
~01
_
_
IIIA 2 12,23,31
(2-+2,)-'(2~--2,)-21n)~
(70)
12,23,31
2`'2~(2o-2,)2+S2 ~
12.23,31
2 2`"
(2`'-2,)-ln~
(70
-1
y, 2`'2,(2,,+2,)(2,,_2,)222
IlIA 2 ,2,23,31
III
A
(72)
I 2,23,3 I
1 2 12,23,31
Z
eo2 -- IIIA
~12
2 + XIII
2222(26 - - 2 , )
12,23.31
2g127,(2`'+2j(2,_2j_Zln ~
(73)
~ 2`'2,(2,, + L ) ( 2 o - L) 2
IlIA 212,23,31
--~
(2`'+2,+2III2g'2;-')(2`'-2,)-21n
(74)
12,23,31
Equations (69)-(74) and (35) offer the time rate of In U when U has three eigenvalues
21 ~ 22 ~ 23 ~ 21"
(ii) U has only two distinct eigenvalues.
1
Eli
--
1
(21 - - 2 2 )
2~221 +2~271
eO0 - - (21 - - 2 2 ) 2
22.22 ln~_2
-1
(75)
21--22
(2,2' +22271
~01 - - (21 - - 2 2 ) 2 k
21 +22
- - 21 - 2 2
21\
ln~).
(76)
(77)
Equations (75)-(77) and (44) offer the time rate of In U when U has only two distinct
eigenvalues.
(iii) U = 21.
1~ = 2 - ' 13.
(78)
Substituting eqn (68) into eqns (57)-(62) and (64)-(67), we obtain the following.
(i) U has three distinct eigenvalues.
Ill
POO--A2
i,i
12,23,31
21112
2`'2,(2`'--2,)2+ _
A
I 2,23,31
2F
2,,2,
in 2`"] -'
2~-/
(2`'+2,)1n
(79)
(80)
3338
H. Xiao
P ~ ' : ) x -~ 1_~233,
Y <'x"(<-<)2+X,:
III
pOl
III
a2
(2o+}~)()o_},)2 -
12,21,~1
p02-A2
jO12 =
(81)
.... 3,
I 2,23,3 I
(2+III/..
. ~- 22 ~ - (2~+}~))_\(In2~</
(82)
A 12.23.31
(,<-5.02+- A
III
A 2 12.23.3,2 2 f l ~ :
I 2.23.3 I
0.,+2,),~,)~ln
(83)
(,;.~+)o.+2III2g'2U l) In
(84)
Equations (79) (84) and (56) offer the conjugate stress of In U when U has three distinct
eigenvalues.
(ii) U has only two distinct eigenvalues.
(85)
Do
( t - , - 2 2 ) 2 5~, +22-2(5.1
-I
(21 _22) 2
,~.) In
2)~22--(}~--/'~)l n 2 ,
(86)
i)
(87)
Equations (85) (87) and (63) offer the conjugate stress of In U when U has only two
eigenvalues )q 4- 2~.
(iii) U = 21.
T = 2T ll}.
(88)
With considerable labor, one can obtain Hoger's (1986) and Man and Guo's (1993)
results for the time rate of In U and Hoger's (1987) result for the conjugate stress of In U
from the above results, and vice versa.
6. CONCLUDING REMARKS
In the previous sections, basis-free expressions for time rate and conjugate stress of an
arbitrary Hill's strain are derived by a simple method. The expressions (33), (34) and (54),
(55) are valid for all cases of the eigenvalues of U. By comparing our results for the time
rate of Hill's strains with the corresponding results presented by Carlson and Hoger (1986)
and Man and Guo (1993), one can see that the former is simpler than the latter. On the
other hand, our result for the conjugate stress of Hill's strains is the first one.
To ensure the validity of our results, some condition should be imposed on the function
J(.). This problem was investigated by Carlson and Hoger (1986) and Man and Guo (1993).
Let U be a symmetric second-order tensor over an n-dimensional Euclidean space.
Then for d = n, eqns (1 a, b) offer a general form of isotropic tensor-valued function of a
symmetric tensor U. Carlson and Hoger (1986) were the first to obtain the derivative of
such a function. They presented explicit results for n distinct eigenvalues and remarked that
for any choice of the dimension n and any particular type of coalescence, the corresponding
result could be derived by means of the continuity. Moreover, in a remark they conjectured
that, "in general, the formula for DF(X)[T] depends only on the number of distinct
eigenvalues of X and is independent of the dimension of the underlying space" [see Carlson
3339
and Hoger (1986, p. 421)]. It can be easily understood that eqns (33) and (34) hold for the
n-dimensional case and hence that they provide the derivative of the tensor-valued function
E(U) that is valid for all cases of the eigenvalues of U. From eqns (33) and (34) it can be
seen that the aforementioned Carlson and Hoger conjecture is true.
The basis-free expressions (33) and (54) for the time rate and the conjugate stress of
the Hill's strain E(U) contain only the integral powers of the right stretch tensor U and
depend linearly on the rate U and the Jaumann stress T ('~, respectively; moreover, their
coefficients e,~ and Prs [cf. eqns (15), (29), (34) and (55)] depend on the distinct eigenvalues
of U only. To apply these expressions, determination of U and its distinct eigenvalues is
required. As we know, U is related to the right Cauchy-Green tensor C by
U 2 = C = FTF.
(89)
Presuming that the deformation gradient F is known, from the above we see that C is easy
to calculate, but the calculation of U is considerably more difficult, since it is the square
root of C. This difficulty can be circumvented by means of the works of Hoger and Carlson
(1984a) and Sawyers (1986) [cf. Marsden and Hughes (1983, p. 55) and Ting (1985)]. These
works enable us to calculate U, U - l and the rotation R etc. in terms of the integral powers
of C and the principal invariants of C. In the following we indicate that the coefficients er.~
and P,s in eqns (33) and (54) may be calculated directly in terms of the principal invariants
of C.
In reality, from eqn (89) we know that the eigenvalues of U are the positive roots of
the sextic
26 - Ic24 + IIc)~2 -- IIIc = 0,
(90)
IIc = ((trC) 2 - t r C 2 ) / 2 ,
IIIc = detC.
(91)
i = 1,2,3,
(92)
where
(2I 3 - 9Iclic + 27IIIc)
qb = c o s - ' \
2(I2-311c) 3/2 j.
(93)
From the above all the distinct eigenvalues of U, 2~. . . . ,Zm, are known and hence the
coefficients er, and Pr~ are determined. For the latter, we mention that as the functions of
the distinct eigenvalues 2, . . . . . J-m, the coefficients ers and p,, are independent of the order
of 2, . . . . . Am, since er.~and Pr, are symmetric functions of 21. . . . . 2m.
Acknowledgements--The author dedicates this paper to the memory of his mentor, Professor G u o Zhong-Heng,
who offered him much advice, help and encouragement, who passed away suddenly on 22 September 1993. The
author is sincerely obliged to Professor C. S. M a n for sending him his papers. Moreover, the author greatly
appreciates the referee's valuable comments, which drew his attention to determination of the coefficients er~ and
p~ in terms o f the principal invariants of C. This research was supported by National Natural Science Foundation
of China and the National Postdoctoral Science Foundation of China.
REFERENCES
Carlson, D. E. and Hoger, A. (1986). The derivative of a tensor-valued function of a tensor. Q. Appl. Math. 44,
409-423.
3340
H. Xiao