The torsional rigidity of a homogeneous, isotropic, linearly elastic prismatic shaft under uniform torsion is the ratio of applied torque to twist per unit length
(Baksa 2020, p. 85). It equals , where
is the modulus of rigidity and
(sometimes also denoted
) is a geometric multiplier determined by the shape
and size of the shaft's cross section. The angular
twist
of the shaft is given by
|
(1)
|
(Roark 1954, p. 174), where is the twisting moment (commonly measured in units of inch-pounds-force),
is the length (inches),
is measured in pounds-force per square inch, and
in inches to the fourth power. Note that the quantity
is sometimes denoted
(e.g., Timoshenko and Goodier 1951, p. 264).
To characterize
by a boundary-value problem, represent this
cross section by a bounded,
connected plane domain
with a piecewise smooth boundary.
Let
be the outward unit normal vector. The unknown
is a warping function
describing displacement along the shaft's axis. If the
twist per unit length is
, then this axial displacement is
(Lipton 1998, pp. 86-87). Mechanical equilibrium
requires
to be a harmonic function, satisfying
in
, where
is the two-dimensional Laplacian.
The condition that the sides of the shaft are free of applied traction gives the
Neumann boundary conditions
|
(2)
|
on , where
is the gradient of
. These conditions determine
up to an additive constant. The
two engineering shear strains are
and
. The elastic energy per unit
length is therefore
, where
is the double integral
|
(3)
|
(Lipton 1998, pp. 87-89; Freddi et al. 2010). Differentiating this energy with respect to
gives the torque
, identifying
with the torsional rigidity in the opening definition. This
formulation also applies to cross sections with
holes, such as an annulus. For a slit
disk, the Neumann boundary conditions
apply on each bank with its own outward unit normal
vector.
For a simply connected cross section, the same shear deformation can be represented by a normalized Prandtl
stress function
defined by
|
(4)
| |||
|
(5)
|
Since
is a harmonic function, these definitions are
compatible. Differentiating them and canceling the mixed partial
derivatives gives
|
(6)
| |||
|
(7)
|
Thus the right-hand side follows from the factor 2 chosen in the definition of
. The traction-free condition makes
constant on the boundary.
Since the boundary is connected,
this constant can be set to zero, giving zero Dirichlet
boundary conditions. Substitution into the energy integral and integration
by parts then give
|
(8)
| |||
|
(9)
| |||
|
(10)
|
With holes, the boundary values of need not all be zero (Freddi et al. 2010).
Values of
are known exactly only for a small number of cross sections,
and as closed-form solutions for even fewer.
The following table lists exact expressions and approximate values for some common
shapes (Timoshenko and Goodier 1951, pp. 258-280; Roark 1954, pp. 174-179).
Here,
is the radius for circular cross
sections and circular sectors, the side length for the equilateral
triangle and square, and the equal leg length for the isosceles
right triangle. For an annulus,
and
are the outer and inner radii with
;
for an ellipse, they are positive semiaxes; and for a
rectangle, they are positive full side lengths. The exact expressions in the table use the convergent
series
|
(11)
| |||
|
(12)
|
Here . The table gives the full geometric multiplier
, so the physical torsional rigidity is
. Timoshenko and Goodier (1951, p. 279) tabulate
approximate coefficients calculated by Saint-Venant
(1878ab) from the classical circular sector solution.
The closed-form expressions in the table
for the half-disk and slit
disk were obtained by E. Weisstein (Aug. 27, 2010).
| cross section | OEIS | references | ||
| annulus | A019669 | Roark (1954, p. 175) | ||
| circle | A019669 | Roark (1954, p. 174) | ||
| ellipse | A000796 | Timoshenko and Goodier (1951, pp. 263-265); Roark (1954, p. 174) | ||
| equilateral triangle | A180317 | Timoshenko and Goodier (1951, pp. 265-267); Roark (1954, p. 175) | ||
| half-disk | A180310 | E. Weisstein (2010) | ||
| isosceles right triangle | A180314 | Galerkin (1919) | ||
| quarter-disk | A401616 | Timoshenko and Goodier (1951, pp. 278-280) | ||
| rectangle | aspect-ratio dependent | Timoshenko and Goodier (1951, pp. 275-277) | ||
| slit disk | A180311 | E. Weisstein (2010) | ||
| square | A180309 | Timoshenko and Goodier (1951, pp. 275-277) |
Exact series representations are known for the rectangle and square
|
(13)
| |||
|
(14)
|
(Timoshenko and Goodier 1951, pp. 275-277), isosceles right triangle
|
(15)
|
(Galerkin 1919, p. 116, correcting to
in equation (18)).
Applying Ramanujan's transformation for hyperbolic cotangent sums (Berndt and Straub 2017, p. 6, equation (3.1)), with analytic continuation, gives
|
(16)
|
and hence the exact relation
|
(17)
|
where the square has side length and the isosceles right
triangle has equal legs of length
.
A faster convergent series is obtained by putting
|
(18)
|
(OEIS A255702) and writing
|
(19)
| |||
|
(20)
|
where
is the Riemann zeta function. These are series representations, with exponentially decaying summands
in
.
For a circular sector, the solution is
|
(21)
|
where
are polar coordinates,
,
, and
|
(22)
|
(Dinnik; Saint-Venant 1878ab, Greenhill 1879, Föppl and Föppl 1928; Timoshenko and Goodier 1951, pp. 278-280). Here is the radial coordinate,
and
is the fixed outer radius. The first term has Laplacian
, and the remaining series
is a harmonic function. The separated expression
for
is interpreted by continuous limits at
and
. At
, the cross section
is a slit disk formed by cutting a disk
along one radius from its center to its circumference.
The two sides of the cut remain separate free edges. The ideal slit has zero area,
but changes the boundary conditions; its multiplier
is therefore smaller than the uncut disk multiplier
.
Integrating the circular sector solution gives the finite expression
|
(23)
|
where
is the digamma function and
is the trigamma function.
This formula is valid for
, including the
resonant values above. At
,
, and
, it reduces to the elementary
quarter-disk, half-disk,
and slit disk expressions.
For a quarter-disk, integrating the circular sector solution at opening angle gives
|
(24)
| |||
|
(25)
|
as given in the table above.
The circular sector series for can be written using a generalized
hypergeometric function (Cetinkaya et al. 1996, p. 7). Put
and
|
(26)
|
and define
|
(27)
|
Then
|
(28)
|
Here
and
. The complete expression for
is extended to the exceptional angles
specified above and to
by continuous limits.
The domain-integral formulation also gives a sharp geometric lower bound using distance to the boundary.
Let
solve
in a bounded open convex set
,
, with zero Dirichlet
boundary conditions, and put
. For
, the convex set
is simply
connected, and
is precisely the normalized Prandtl stress
function defined above. Consequently,
, connecting
this notation to the same geometric multiplier used throughout the entry. Perelmuter
(2026) proved
|
(29)
|
Here
denotes the boundary of
, and
is the
distance from
to that boundary, measured with
the Euclidean norm
. Balls attain equality.
For a shaft with this plane cross section, multiplying
the bound by
gives a lower bound for its torsional rigidity
.
Bogosel et al. (2026, Theorem 1) proved the polygonal Saint-Venant inequality. Among simple polygons of prescribed area
with at most sides,
, the regular
polygon with
sides uniquely maximizes the geometric torsional multiplier
. Equality is up to congruence,
and the competing polygons need not be convex.