ON SAINT VENANT'S PRINCIPLE

R. v . MISES

The so-called principle of statically equivalent loads, due to Saint

Venant, has been referred to for the last fifty years in almost all texts
on elasticity. The statements in different books vary only slightly.
Let us quote A. E. H. Love's Treatise on the mathematical theory of
elasticity (4th éd., p. 132) : "According to this principle the strains that
are produced in a body by the application, to a small part of its sur-
face, of a system of forces statically equivalent to zero force and zero
couple, are of negligible magnitude at distances which are large com-
pared with the linear dimensions of the part." In this form the state-
ment is not very clear. Forces applied to a body at rest must be in
equilibrium in any case. It would not make sense to speak of adding
or subtracting a system of forces that is not an equilibrium system.
What is meant may be correctly expressed in this way : If the forces
acting upon a body are restricted to several small parts of the sur-
face, each included in a sphere of radius e, then the strains and stresses
produced in the interior of the body at a finite distance from all those
parts are smaller in order of magnitude when the forces for each
single part are in equilibrium than when they are not. If this state-
ment is true, it must be capable of a mathematical proof, that is, it
must be a consequence of the fundamental differential equations of
elasticity theory. But no attempt is made in the usual textbooks to
supply a demonstration. Most texts give Boussinesq as a reference
for the proof. What Boussinesq really dealt with was the infinite body
filling the half space z>0 and subjected to normal forces at its bound-
ary 0 = 0. If the forces are applied to points £, 77, 0 where £2+rç22§€2,
Boussinesq proved that the stress at a point x, y, z is of order c when
the sum of forces is zero and of order e2 when their moments also
vanish. It will be shown in the following that this is not the case, in
general, if tangential components of the forces at 2 = 0 are admitted.
Moreover we shall consider a body of finite dimensions and see that
there too Saint Venant's principle in its traditional form does not hold
true. The main result, from a practical point of view, is that Saint
Venant's principle can be applied if all forces involved are parallel
and not tangential to the surface of the body, but not under more
general conditions. No objection is raised in the present paper against
using the principle in the case of bodies with one or two infinitesimal
dimensions, like thin plates, shells or beams, although a proof of its
Received by the editors January 18, 1945.

555
556 R. v. MISES [August

validity in these cases or even a precise statement has not yet been
given.
1. The infinite half space. If the force X, F, Z is applied in the
origin to the body z > 0 , the displacements «, v, w at a point x> y, z
can be expressed, as it follows from Boussin esq's results, 1 in the fol-
lowing way. We denote by U and V the harmonic functions
xX + yY
(1) U —, F=Zlog(r + s)
r + z
where r2~x2+y2+z2, by K the value 1 —2a- where a is Poisson's ratio,
and by JJ, the shear modulus. Then the displacements are given by
2 / d\dU ( d\dV
Airpu s* — X + [K — 1 + z — ) 1 K+ z— ) ;
r \ dz/ dx \ dz) dx
2 / d\dU ( d\dV
(2) 4TM*« — F + I J K - l + « —) [K + Z — ) ,
r \ dz) dy \ dz) dy
2 ( d\dU ( d\dV
4xju w = — Z + l — K + Z — J ( 1—K + z — J •
r \ dz) dz \ dz) dz
I t can easily be shown t h a t these values satisfy the differential equa-
tions of elasticity theory :
l dd l do i do
AM H = 0, Av -\ = 0, Aw -\ == 0;
K dx K dy K dz
(3)
du dv dw
e*n—-\ 1
dx dy dz
From (2) we derive by differentiation
K xX + yY + zZ
(4) e ' •
2*n fs
On the other hand, it follows from (2) if the strain-stress relations
are introduced t h a t the stress vector for an element parallel to the
x-y-plane is directed toward the origin and has the magnitude
1
Boussinesq, J., Applications des potentiels à Vétude de l'équilibre et du mouvement
des solides élastiques, Paris, 1885. A. E. H. Love, Treatise on the mathematical theory of
elasticity, 4th éd., Cambridge, 1934, p. 191, gives only the results for a force acting
normal to the boundary. The complete expressions can be derived from the formulae
developed by E. Trefftz in: Frank-Mises, Differential- und Integralgleichungen der
Mechanik undPhysik, 2nd éd., vol. 2, Braunschweig, 1935 (Reprint, New York, 1944),
p. 303.
'9451 SAINT VENANTS PRINCIPLE 557

3z xX + yY + zZ

Thus the stress is zero in all points 2 = 0, x2+y2^0 and across each
plane 2 = c o n s t . > 0 the stresses are equivalent to a single force —X,
— F, —Z passing through the origin.
The mean normal stress is given by
1 3-/c K- 3 xX + yY + zZ
(6) a = - (crx + <ry + <TZ) = ——fxd = —
3 3K 6W r8
If the force attacks at £, 77, 0 instead of at the origin, the same
formulas can be used, where only x, y have to be replaced by x — £,
y-~V, and r2 by (x — £)2-\-(y — rj)2-\-z2. Developing with respect to £,
rj we obtain for the mean normal stress due to a system of forces
Xvy YV} Zv attacking at points £,, rjvi 0 (*> = 1, 2, 3, • • • ) :

/c — 3

(7) + — [(3s 2 - r2) 2 £,X, + 3xyT, &F,

r2
+ 3*2:22 ^ + 3 x ^ X ^ >
+ (3y' - ' 2 ) E * 1 % + 3 y « E ^ > ] + • • •
and similar expressions for other stress values.
If all £„ and T]V are of the order of magnitude €, we can conclude :
The stresses (and strains) in a point x, y, z are of the order c, if the
sums of force components 22^"> 22 ^V» 2 2 ^ " a r e zero; they are of the
order e2, if and only if the six linear moments 22£"^> 22&^> 2 2 ^ ^ "
an
d 22 1 /^"» 2277"F„, 2277"-2*" a l s o vanish. The case of a system in equi-
librium, that is, YL^v—^^vZv^2,{^vYv'-7]vZv) = 0, is, in general, in
no way distinguished. Only if all forces are parallel to each other,
either normal to the boundary surface or inclined under an angle
different from zero, the three equilibrium conditions entail the other
three conditions. In general, the order of magnitude of the inner
strains is reduced to e2 if and only if the external forces acting upon a
small part of the surface are such as to remain in equilibrium when
turned through an arbitrary angle (astatic equilibrium).

-0

(a) (b) « (c) (d)

finite stress stress of order of magnitude e stress of order ê
FIG. 1
558 R. v. MISES [August

T h e results are illustrated in the four simple examples of Fig. 1.

All forces are here parallel to the ^-direction ( F = Z = 0). In the case
(a) we have a single force which provides a finite stress value accord-
ing to the first term of (7). In both cases (b) and (c) the sum of forces
is zero, but in (b) the sum of T)VXV and in (c) the sum of £„X„ is differ-
ent from zero. It follows from (7) that in either case the stress has the
order of magnitude e. If the Saint Venant principle were correct, the
stress should be small of higher order in the case (c) where all equi-
librium conditions are fulfilled. In fact, in the case (d) only where the
three forces form a system in astatic equilibrium with all linear mo-
ments zero the stress has the order of magnitude e2.
2. Circular disk. It may be doubted whether the infinite half space
gives an adequate picture of what happens in the case of a finite body.
Here one cannot speak of applying a single force to the body since the
forces as a whole have to form an equilibrium system. I t will be suffi-
cient, however, for the present purpose, to analyze one particular
shape of finite boundary.
Let us consider the two-dimensional stress distribution in a circular
disk of radius R subjected to external forces Fi, F 2 , • • • , F n which

FIG. 2

attack at points P i , P 2 , • • •, P n of the circumference. The stresses in a

point P are determined by Airy's stress function 5 ( P ) . If (Fig. 2) r is
the distance of P from the center C, rv the distance between P and P„,
<t>P the angle of Fv with CPvy and 8P the angle of Fv with PPvy the stress
function is given by 2
1 n / r2 \
(8) S(P) » — ]C M r,09 sin 09 cos 4>, ).
T „»i \ 4R /
8
Cf. Love, loc. cit. p. 217. H. Hertz, Gesammelte Werke, vol. 1, p. 283.
19451 SAINT VENANTS PRINCIPLE 559

The stress components with respect to an x-^-coordinate system

can be expressed as follows, if ft designates the angle of F , with the
#-axis :
1 _ f 2 cos ó , l
<r, = — 23 F* — cos ft cos2 (ft - j8,) - — — ,
* Lr, 2R J
1 x_^ f 2 cos ó , l
(9) <rtf » — 23 *V — cos ft sin2 (ft - ft) —- ,
7T Lr, 2JR J
1 _ 2
r — 2 ^ F, — cos ft cos (ft — ft) sin (ft — ft).
v r,
It can easily be shown that S(P) fulfills the differential equation
AAS = 0 and t h a t the stresses on the boundary r = R are zero, except
in the points P,(r, = 0). In these points the stresses become infinite
and combine so as to balance F,.
In the center of the circle we have r, = i? and ft = # , and introducing
a , = f t —<£„ the polar angle of P „ we find for the stress components
in C
<rx = — 23 F» cos (ft - «0(1/2 + cos 2a,),
TR

(10) <rv =* X) J?, cos (ft - a,) ( 1 / 2 - cos 2a,),

7ri£

r = 23 F , cos (ft — a,) sin 2a,.

Let us now assume t h a t there are two groups of forces F , and F , ' ,
forming angles ft and ft' with the #-axis. The points of attack P , of
the first group may lie close to a point P 0 with polar angle a0 and the
points Pi of the second group near to PÓ with polar angle ai so as to
have av = ao+t;v1 ai = a o + £,' (£, £' small). If we develop the expres-
sion for <xz with respect to £ and £', we find
TTR(TX = (1/2 + cos 2a0) 23 ^ cos (ft - a0)
+ (1/2 + cos 2a 0 ') ] £ F,' cos (ft - a 0 ')
— 2 sin 2a 0 23 &»%» cos (ft — ao)
(ID
- 2 sin 2a„' 23 W cos (ft - a„')
+ (1/2 + cos 2a0) 23 F& sin (ft - a 0 )
+ (1/2 + cos 2a 0 ') 23 WE/ sin (ft - a 0 ') + • • • .
T h e first two terms vanish if in each group the sum of forces is zero,
since ft—«o and ft' — ai are the angles the forces form with CPo and
560 R. v. MISES [August

CPi respectively. In this case, therefore, the highest terms in the ex-
pression for <TX are of the order £, £'.
If the sums of moments in each group are also zero, we have
Z PpR sin (ft - av) « RJ2 F, sin (ft - « 0 - fc) - 0,
E F / J R sin 08/ - al) - *y; F ; sin (ft' - ai - tf) - 0.

If these expressions are developed with respect to £ and £', it is seen

t h a t with the resultant forces and moments vanishing the sums
E F& cos (ft - av) and £ F,' tf cos (ft' - a! )
are zero except for quantities of higher order. In this case the third
and fourth terms in (11) vanish, but the fifth and sixth still supply a
quantity of the order of magnitude £, £'. Only if the two additional
conditions are fulfilled t h a t
£ FA sin (ft - «o) and £ FI Hi sin (ft' - ai )
vanish, except for terms of higher order, the development of ax will
start with terms of the order £2, ££', £' 2 .
Again the results may be illustrated on four simple examples shown
in Fig. 3. In the case (a) the two single forces of magnitude F produce

(a) finite stress (b) and (c) stress of order of magnitude «

(d) stress of order ê

FIG. 3
1945] SAINT VENANTS PRINCIPLE 561

in the center a finite normal stress <rx~3F/TrR according to (11). In

(b) we have two groups of parallel forces with a 0 = 45°, ai =135°.
If the angular distance of the two neighboring points of application
is called e, one has £i = £i = 0 , & = - € , £2' = e and with j8i=j32' =180°,
j32=j8/ = 0 , equation (11) supplies <rx~5Fe/21,2TrR. In the case (c) the
two forces attacking in the points ai = 90° —€, a 2 = 90° + € form an
equilibrium system. Equation (11) gives here the value Vx — Fe/irR
which is of the same order as in (b). Were Saint Venant's principle
correct, the stresses in (c) should be small of a higher order. In fact,
this is the case with the three forces in example (d) only which form
an astatic system (up to terms of higher order). All terms written
down on the right-hand side of (11) vanish in the case (d).
3. Conclusions. In order to obtain a precise and sufficiently general
statement let us consider a finite simply connected body, supported at
an adequate number of distinct surface points Si, 5 2 , 53, • • • . Let
P i be a point of the surface where the load Fi is applied and P
an inner point of the body at finite distances from P i and from
•Si, S2, 5s, • • • . Let, finally, a be some well defined strain or stress
quantity in P , for instance, the normal stress in ^-direction, or any
component of the distortion. Then, with constant Fi, this a will be a
function of the coordinates of P i . If P i is a regular surface point
(tangential plane, finite curvature) the function will have finite de-
rivatives. T h a t means, if P i moves through a small distance e the
change in a will be of the order of magnitude €. Consequently, two
equal and opposite forces attacking at points of distance e will pro-
duce a (7-value of the order €. On the other hand, the load Fi can be
replaced by several loads that have the vector sum Fi, all attacking
a t Pi. Each of them can be shifted to the neighborhood and then re-
versed. The system of these reversed forces combined with the origi-
nal Fi will still produce a cr-value of order e. Thus our first statement
reads :
(a) If a system of loads on an adequately supported body, all applied
at surface points within a sphere of diameter e, have the vector sum zero,
they produce in an inner point P of the body a strain or stress value <r of
the order of magnitude €.
To this statement we add the results reached in the preceding sec-
tions by way of direct computation for two particular cases, the infi-
nite half space and the circular disk. The general proof following the
same lines can be given without difficulty.
(b) If the loads, in addition to having the vector sum zero, fulfill three
562 R. y. MISES

further conditions so as to form an equilibrium system within the sphere

of diameter e, the a-value produced in P will, in general, still be of the
order of magnitude €.
(c) If the loads, in addition to being an equilibrium system, satisfy
three more conditions so as to form a system in astatic equilibrium, then
the cr-value produced in P will be of the order of magnitude e2 or smaller.
In particular, if loads applied to a small area are parallel to each other
and not tangential to the surface and if they form an equilibrium system,
they are also in astatic equilibrium and thus lead to a a of the order e2.
In this whole argument the loads as well as the supporting reactions
were supposed to be concentrated, finite forces acting at distinct
points of the surface. No difficulty arises if, instead, continually dis-
tributed surface stresses are assumed with the provision that all in-
tegrals of such stresses over finite regions (and the regions that tend to
zero) remain finite.
A final remark is in order about the legitimate application of St.
Venant's principle (or some equivalent statement) in cases of thin
rods, shells, and so on. The only precise and consistent way to deal
with thin elastic rods is the theory of the so-called one-dimensional
elastica. In this theory the forces acting on the ends of the rod enter
the computation only with their resultant vector and resultant mo-
ment. This implies, evidently, a principle of "statically equivalent
loads." What Saint Venant originally had in mind was doubtlessly
the case of a long cylinder with infinite ratio of length to diameter.
The purpose of the present paper was to show that an extension of the
principle to bodies of finite dimensions is not legitimate.
HARVARD UNIVERSITY