Tensor eld

Not to be confused with the Tensor product of elds. ters being in a manifold M. For example, a vector space
of one dimension depending on an angle could look like
a Mbius strip as well as a cylinder. Given a vector bun-
In mathematics, physics, and engineering, a tensor eld
assigns a tensor to each point of a mathematical space dle V over M, the corresponding eld concept is called a
section of the bundle: for m varying over M, a choice of
(typically a Euclidean space or manifold). Tensor elds
are used in dierential geometry, algebraic geometry, vector
general relativity, in the analysis of stress and strain in
materials, and in numerous applications in the physical vm in Vm,
sciences and engineering. As a tensor is a generaliza-
tion of a scalar (a pure number representing a value, like the vector space 'at' m.
length) and a vector (a geometrical arrow in space), a ten- Since the tensor product concept is independent of any
sor eld is a generalization of a scalar eld or vector eldchoice of basis, taking the tensor product of two vector
that assigns, respectively, a scalar or vector to each pointbundles on M is routine. Starting with the tangent bun-
of space. dle (the bundle of tangent spaces) the whole apparatus
Many mathematical structures called tensors are actu- explained at component-free treatment of tensors carries
ally tensor elds. For example, the Riemann curvature over in a routine way again independently of coordi-
tensor is actually not a tensor, as the name implies, but a nates, as mentioned in the introduction.
tensor eld: It is named after Bernhard Riemann, and as- We therefore can give a denition of tensor eld, namely
sociates a tensor to each point of a Riemannian manifold, as a section of some tensor bundle. (There are vector
which is a topological space. bundles which are not tensor bundles: the Mbius band
for instance.) This is then guaranteed geometric content,
since everything has been done in an intrinsic way. More
1 Geometric introduction precisely, a tensor eld assigns to any given point of the
manifold a tensor in the space
Intuitively, a vector eld is best visualized as an 'arrow'
attached to each point of a region, with variable length
V V V V
and direction. One example of a vector eld on a curved
space is a weather map showing horizontal wind velocity where V is the tangent space at that point and V* is the
at each point of the Earths surface. cotangent space. See also tangent bundle and cotangent
The general idea of tensor eld combines the requirement bundle.
of richer geometry for example, an ellipsoid varying Given two tensor bundles E M and F M, a map A:
from point to point, in the case of a metric tensor with (E) (F) from the space of sections of E to sections
the idea that we don't want our notion to depend on the of F can be considered itself as a tensor section of E F
particular method of mapping the surface. It should exist if and only if it satises A(fs,...) = fA(s,...) in each argu-
independently of latitude and longitude, or whatever par- ment, where f is a smooth function on M. Thus a tensor
ticular 'cartographic projection' we are using to introduce is not only a linear map on the vector space of sections,
numerical coordinates. but a C (M)-linear map on the module of sections. This
property is used to check, for example, that even though
the Lie derivative and covariant derivative are not tensors,
2 The vector bundle explanation the torsion and curvature tensors built from them are.

Main article: Tensor bundle

3 Notation
The contemporary mathematical expression of the idea The notation for tensor elds can sometimes be confus-
of tensor eld breaks it down into a two-step concept. ingly similar to the notation for tensor spaces. Thus, the
There is the idea of vector bundle, which is a natural idea tangent bundle TM = T(M) might sometimes be written
of 'vector space depending on parameters the parame- as

1
2 6 TENSOR CALCULUS

similarly be expressed as mappings of covector elds into

functions (namely, we could start natively with covector
T01 (M ) = T (M ) = T M elds and work up from there).
to emphasize that the tangent bundle is the range space In a complete parallel to the construction of ordinary sin-
of the (1,0) tensor elds (i.e., vector elds) on the mani- gle tensors (not elds!) on M as multilinear maps on
fold M. This should not be confused with the very similar vectors and covectors, we can regard general (k,l) ten-
looking notation sor elds on M as C (M)-multilinear maps dened on l
copies of T (M ) and k copies of T (M ) into C (M).
Now, given any arbitrary mapping T from a product of
T01 (V ) k copies of T (M ) and l copies of T (M ) into C (M),
it turns out that it arises from a tensor eld on M if and
in the latter case, we just have one tensor space, whereas only if it is a multilinear over C (M). Thus this kind of
in the former, we have a tensor space dened for each multilinearity implicitly expresses the fact that we're re-
point in the manifold M. ally dealing with a pointwise-dened object, i.e. a tensor
Curly (script) letters are sometimes used to denote the set eld, as opposed to a function which, even when evalu-
of innitely-dierentiable tensor elds on M. Thus, ated at a single point, depends on all the values of vector
elds and 1-forms simultaneously.
A frequent example application of this general rule is
Tnm (M ) showing that the Levi-Civita connection, which is a map-
ping of smooth vector elds (X, Y ) 7 X Y taking a
are the sections of the (m,n) tensor bundle on M which pair of vector elds to a vector eld, does not dene a
are innitely-dierentiable. A tensor eld is an element tensor eld on M. This is because it is only R-linear in Y
of this set. (in place of full C (M)-linearity, it satises the Leibniz
rule, X (f Y ) = (Xf )Y + f X Y )). Nevertheless, it
must be stressed that even though it is not a tensor eld, it
4 The C (M) module explanation still qualies as a geometric object with a component-free
interpretation.
There is another more abstract (but often useful) way of
characterizing tensor elds on a manifold M which turns
out to actually make tensor elds into honest tensors (i.e.
single multilinear mappings), though of a dierent type
(although this is not usually why one often says tensor 5 Applications
when one really means tensor eld). First, we may con-
sider the set of all smooth (C ) vector elds on M, T (M ) The curvature tensor is discussed in dierential geometry
(see the section on notation above) as a single space &3; and the stressenergy tensor is important in physics and
a module over the ring of smooth functions, C (M), by engineering. Both of these are related by Einsteins the-
pointwise scalar multiplication. The notions of multilin- ory of general relativity. In engineering, the underlying
earity and tensor products extend easily to the case of manifold will often be Euclidean 3-space.
modules over any commutative ring.
It is worth noting that dierential forms, used in dening
As a motivating example, consider the space T (M ) of integration on manifolds, are a type of tensor eld.
smooth covector elds (1-forms), also a module over the
smooth functions. These act on smooth vector elds to
yield smooth functions by pointwise evaluation, namely,
given a covector eld and a vector eld X, we dene
6 Tensor calculus
((X))(p) = (p)(X(p)).
In theoretical physics and other elds, dierential equa-
Because of the pointwise nature of everything involved, tions posed in terms of tensor elds provide a very general
the action of on X is a C (M)-linear map, that is, way to express relationships that are both geometric in
nature (guaranteed by the tensor nature) and convention-
((fX))(p) = f(p) (p)(X(p)) = (f)(p)(X(p)) ally linked to dierential calculus. Even to formulate such
equations requires a fresh notion, the covariant derivative.
for any p in M and smooth function f. Thus we can regard This handles the formulation of variation of a tensor eld
covector elds not just as sections of the cotangent bun- along a vector eld. The original absolute dierential cal-
dle, but also linear mappings of vector elds into func- culus notion, which was later called tensor calculus, led to
tions. By the double-dual construction, vector elds can the isolation of the geometric concept of connection.
 3

7 Twisting by a line bundle What is usually spoken of as the 'classical' approach to

tensors tries to read this backwards and is therefore
An extension of the tensor eld idea incorporates an extra a heuristic, post hoc approach rather than truly a foun-
line bundle L on M. If W is the tensor product bundle of dational one. Implicit in dening tensors by how they
V with L, then W is a bundle of vector spaces of just the transform under a coordinate change is the kind of self-
same dimension as V. This allows one to dene the con- consistency the cocycle expresses. The construction of
cept of tensor density, a 'twisted' type of tensor eld. A tensor densities is a 'twisting' at the cocycle level. Ge-
tensor density is the special case where L is the bundle of ometers have not been in any doubt about the geometric
densities on a manifold, namely the determinant bundle of nature of tensor quantities; this kind of descent argument
the cotangent bundle. (To be strictly accurate, one should justies abstractly the whole theory.
also apply the absolute value to the transition functions
this makes little dierence for an orientable manifold.)
For a more traditional explanation see the tensor density 10 See also
article.
One feature of the bundle of densities (again assuming Ricci calculus
orientability) L is that Ls is well-dened for real number Jet bundle
values of s; this can be read from the transition functions,
which take strictly positive real values. This means for Spinor eld
example that we can take a half-density, the case where
s = . In general we can take sections of W, the tensor
product of V with Ls , and consider tensor density elds 11 References
with weight s.
Half-densities are applied in areas such as dening The Geometry of Physics (3rd edition), T. Frankel,
integral operators on manifolds, and geometric quantiza- Cambridge University Press, 2012, ISBN 978-1-
tion. 107-60260-1

McGraw Hill Encyclopaedia of Physics (2nd Edi-

tion), C.B. Parker, 1994, ISBN 0-07-051400-3
8 The at case
Encyclopaedia of Physics (2nd Edition), R.G.
Lerner, G.L. Trigg, VHC publishers, 1991, ISBN
When M is a Euclidean space and all the elds are taken (Verlagsgesellschaft) 3-527-26954-1, ISBN (VHC
to be invariant by translations by the vectors of M, we Inc.) 0-89573-752-3
get back to a situation where a tensor eld is synonymous
with a tensor 'sitting at the origin'. This does no great Gravitation, J.A. Wheeler, C. Misner, K.S. Thorne,
harm, and is often used in applications. As applied to W.H. Freeman & Co, 1973, ISBN 0-7167-0344-0
tensor densities, it does make a dierence. The bundle
of densities cannot seriously be dened 'at a point'; and Relativity DeMystied, D. McMahon, Mc Graw
therefore a limitation of the contemporary mathematical Hill (USA), 2006, ISBN 0-07-145545-0
treatment of tensors is that tensor densities are dened in Relativity, Gravitation, and Cosmology, R.J.A.
a roundabout fashion. Lambourne, Open University, Cambridge Univer-
sity Press, 2010, ISBN 978-0-521-13138-4

9 Cocycles and chain rules

As an advanced explanation of the tensor concept, one
can interpret the chain rule in the multivariable case, as
applied to coordinate changes, also as the requirement
for self-consistent concepts of tensor giving rise to ten-
sor elds.
Abstractly, we can identify the chain rule as a 1-cocycle.
It gives the consistency required to dene the tangent bun-
dle in an intrinsic way. The other vector bundles of ten-
sors have comparable cocycles, which come from apply-
ing functorial properties of tensor constructions to the
chain rule itself; this is why they also are intrinsic (read,
'natural') concepts.
4 12 TEXT AND IMAGE SOURCES, CONTRIBUTORS, AND LICENSES

12 Text and image sources, contributors, and licenses

12.1 Text
Tensor eld Source: https://en.wikipedia.org/wiki/Tensor_field?oldid=743255168 Contributors: Loren Rosen, Charles Matthews, Dys-
prosia, Jitse Niesen, Phys, Choni, Tea2min, Tosha, Giftlite, Lethe, Mboverload, DemonThing, LiDaobing, Fg2, Paul August, Splat, Oleg
Alexandrov, Postrach, BD2412, Juan Marquez, Mathbot, YurikBot, Wigie, Segv11, SmackBot, Incnis Mrsi, Commander Keane bot, Silly
rabbit, Mr Stephen, Woodshed, Marek69, RobHar, Jondaman21, Lantonov, LokiClock, Caidh, BOTarate, DumZiBoT, Addbot, Latitude-
Bot, Luckas-bot, AnomieBOT, MattTait, J04n, Point-set topologist, Sawomir Biay, Steve Quinn, Tkuvho, Rausch, Slawekb, Hhhippo,
Quondum, Jadzia2341, Maschen, Maxdlink, F=q(E+v^B), Terrell8763, Enyokoyama, Volker Siegel and Anonymous: 20

12.2 Images

12.3 Content license

Creative Commons Attribution-Share Alike 3.0