“Variations, Geometry and Physics”

in honour of Demeter Krupka’s sixty-fifth birthday

O. Krupková and D. J. Saunders (Editors)
Nova Science Publishers (2008)

Remarks on the history of the

notion of Lie differentiation1

Andrzej Trautman2

1. The derivative X(f ) of a function f , defined on a smooth manifold, in the

direction of the vector field X and the bracket of two vector fields, introduced by
Sophus Lie himself, are the first examples of what is now called the Lie derivative.
Another early example comes from the Killing equation. David Hilbert [1], in his
derivation of the Einstein equations, used the expression
X ρ ∂ρ g µν − g µρ ∂ρ X ν − g ρν ∂ρ X µ
and stated that it is a tensor field for every tensor field g and vector field X. Around
1920, Élie Cartan defined a natural differential operator L(X) acting on fields of
exterior forms. He noted that it commutes with the exterior derivative d and gave,
in equation (5) on p. 84 in [2], the formula3
L(X) = d ◦ i(X) + i(X) ◦ d, (1)
where i(X) is the contraction with X.

2. Wladyslaw Ślebodziński, in his article of 1931 [5], wrote an explicit formula for
the Lie derivative (without using that name) in the direction of X of a tensor field
1
Dedicated to Demeter Krupka on the ocasion of his 65th birthday
2
Instytut Fizyki Teoretycznej, Uniwersytet Warszawski, Hoża 69, Warszawa, Poland
email: andrzej.trautman@fuw.edu.pl
3
In this note, I transcribe all equations from the form given by their authors to the notation in
current usage. All manifolds and maps among them are assumed to be smooth. Good references
for my notation and terminology are [3] and [4].

297
298 Remarks on the history of the notion of Lie differentiation

A of arbitrary valence. He gave also an equation equivalent to

L(X)(A ⊗ B) = (L(X)A) ⊗ B + A ⊗ L(X)B
and noted that L(X) commutes with contractions over pairs of tensorial indices. He
then applied his results to Hamilton’s canonical equations of motion. For a function
H(p, q), p = (pµ ), q = (q µ ), µ = 1, . . . , n, Ślebodziński defined the vector field
∂H ∂ ∂H ∂
XH = µ
− µ ,
∂pµ ∂q ∂q ∂pµ
introduced the symplectic form A = dq µ ∧ dpµ , the Poisson bivector B = ∂/∂q µ ∧
∂/∂pµ and showed that L(XH )A = 0 and L(XH )B = 0. This allowed him to
generalize some results of Théophile de Donder in the theory of invariants [6].

The priority of Ślebodziński in defining the Lie derivative in the general case was
recognized by David van Dantzig who wrote, in footnote on p. 536 of [7], Der
Operator [the Lie derivative] wurde zum ersten Mal von W. Ślebodziński eingeführt.
It was van Dantzig who introduced, in the same paper, the name Liesche Ableitung.
Also Jan Arnoldus Schouten, in footnote 1 on p. 102 of [8], lists the 1931 paper by
Ślebodziński as the first reference for the notion of Lie differentiation. Van Dantzig
complemented the approach of Ślebodziński by pointing out that the Lie derivative
can be defined as the difference between the value of a geometric object A at a point
and the value of that object at the same point obtained by an infinitesimal ‘dragging
along’ a vector field. In contemporary notation this is expressed by the formula
d ∗
L(X)A = ϕ A , (2)
dt t |t=0
where ϕ∗t A is the pull-back of A by the flow (ϕt , t ∈ R) generated by X. In view of
the equation
d ∗
ϕ A = ϕ∗t L(X)A,
dt t
the vanishing of L(X)A is equivalent to the invariance of A with respect to the flow
generated by X; see, e.g., §24 in [9].

3. For quite some time, physicists had been using Lie derivatives, without reference
to the work of mathematicians. Léon Rosenfeld [10] introduced what he called a
‘local variation’ δ ∗ A of a geometric object A induced by an infinitesimal transforma-
tion of coordinates generated by X. He noted that δ ∗ commutes with differentiation.
It is easily seen that his δ ∗ A is −L(X)A; see, e.g., [11]. Assuming that A is a tensor
of type determined by a representation ρ of GL(4, R) in the vector space RN and
denoting by ρνµ ∈ End RN the matrices of the corresponding representation of the
Lie algebra of GL(4, R), one can deduce from Rosenfeld’s equations the following
formula for the Lie derivative
L(X)A = X µ ∂µ A − ∂ν X µ ρνµ A.
 Andrzej Trautman 299

In particular, assuming that L is a Lagrange function

R
depending on the components
of A and on their first derivatives and such that Ld4 x is an invariant, Rosenfeld
showed that
L(X)L = ∂µ (LX µ )
and used the formula
∂L ∂L
L(X)L = L(X)A + ∂µ (L(X)A)
∂A ∂(∂µ A)
to derive a set of identities of the Noether type, and the conservation laws of energy-
momentum and of angular momentum. One of the main results of that paper was
the symmetrization of the canonical energy-momentum tensor t achieved by adding
to it an expression linear in the derivatives of the spin tensor s.

Incidentally, it is remarkable that this symmetrization, derived independently also by

F. J. Belinfante, is a natural consequence of the Einstein–Cartan theory of gravita-
tion. In that theory, based on a metric tensor g and a linear connection ω µ ν = Γµνρ dxρ
which is metric, but may have torsion, there are field equations relating curvature
and torsion to t and s, respectively; see [12] and the references given there. If
these Sciama–Kibble field equations are satisfied and X is a vector field generating
a symmetry of space-time so that

L(X)g = 0 and L(X)ω = 0

then, denoting by tµ and sµν the 3-forms (densities) of energy-momentum and spin,
and the covariant derivative with respect to the transposed connection ω̃νµ = Γµρν dxρ
e one has the conservation law dj = 0, where
by ∇,
e ν X µ sµν .
j = X µ tµ + 12 ∇

In the limit of special relativity, if X generates a translation, then j reduces to the

corresponding component of the density of energy-momentum; for X generating a
Lorentz transformation, one obtains a component of the density of total angular
momentum.

4. The Lie derivative defines a homomorphism of the Lie algebra V(M ) of all vector
fields on an n-dimensional manifold M into the Lie algebra of derivations of the
algebra of all tensor fields on M ,

L([X, Y ]) = [L(X), L(Y )].

L
The Cartan algebra C(M ) = np=0 C p (M ) of all exterior forms on M is Z-graded by
the degree p of the forms. A derivation D of degree q ∈ Z maps linearly C p (M ) to
C p+q (M ) and satisfies the graded Leibniz rule,

D(α ∧ β) = (Dα) ∧ β + (−1)pq α ∧ Dβ for every α ∈ C p (M ).

300 Remarks on the history of the notion of Lie differentiation

Derivations of odd degree are often called antiderivations. The vector space
Der C(M ) of all derivations of C(M ) is a super Lie algebra with respect to the
bracket
′
[D, D′ ] = D ◦ D′ − (−1)deg D deg D D′ ◦ D. (3)
The degree of [D, D′ ] is the sum of the degrees of D and D′ and there holds a
super Jacobi identity; see [13] for an early review of super Lie algebras, written for
physicists. In particular, d is a derivation of degree +1 and, if X ∈ V(M ), then L(X)
and i(X) are derivations of degrees 0 and −1, respectively. The Cartan formula (1)
represents L(X) as a bracket, as defined in (3), of d and i(X).

The contraction i(X) generalizes to fields of vector-valued exterior forms. Let X ∈

V(M ), ξ ∈ C p (M ), p = 0, . . . , n, and Y = X ⊗ ξ, then Y is a vector-valued p-form
and i(Y ) is a derivation of the Cartan algebra, of degree p − 1, defined by

i(Y )α = ξ ∧ i(X)α, α ∈ C(M ).

By linearity one extends i(Y ) to arbitrary vector-valued p-forms. The bracket

[d, i(Y )] is now a derivation of degree p; by the super Jacobi identity its bracket
with d is zero and every derivation (super) commuting with d is of this form. If
Y and Z are vector-valued forms of degrees p and q, respectively, then the bracket
[[d, i(Y )], [d, i(Z)]] super commutes with d and, therefore, there exists a vector-
valued (p + q)-form [Y, Z] such that

[d, i([Y, Z])] = [[d, i(Y )], [d, i(Z)]]. (4)

The Fröhlicher–Nijenhuis [14] bracket [Y, Z], defined by (4), generalizes the Lie
bracket of vector fields; it is super anticommutative,

[Z, Y ] = −(−1)pq [Y, Z],

and makes the vector space of all vector-valued forms into a super Lie algebra.
For example, an almost complex structure J on an even-dimensional manifold is a
vector-valued 1-form and [J, J] is its Nijenhuis torsion.

5. A convenient framework to generalize the definition (2) of Lie derivatives is

provided by natural bundles. A natural bundle is a functor F from the category
of manifolds to that of bundles such that πM : F (M ) → M is a bundle and if
ϕ : M → N is a diffeomorphism, then F (ϕ) : F (M ) → F (N ) is an isomorphism
of bundles covering ϕ. If A is a section of πN : F (N ) → N , i.e. a field on N of
geometric objects of type F , then ϕ∗ A = F (ϕ−1 ) ◦ A ◦ ϕ is its pull-back by ϕ to
M . All tensor bundles are natural, but spinor bundles are not. The vertical bundle
V F (M ) is the subbundle of the tangent bundle T F (M ) consisting of all vertical
vectors, i.e. vectors that are annihilated by T πM . Let (ϕt , t ∈ R) be the flow
generated by X ∈ V(M ) and let A be a section of πM . The curve t 7→ (ϕ∗t A)(x) is
 Andrzej Trautman 301

vertical for every x ∈ M and the Lie derivative L(X)A is now defined as the section
of the vector bundle V F (M ) → M such that (L(X)A)(x) is the vector tangent to
t 7→ (ϕ∗t A)(x) at t = 0. The monograph by Kolář, Michor and Slovák [15] contains
a full account of this approach and, in Ch. XI, an even more general definition of
Lie differentiation.

Bibliography
[1] D. Hilbert, Die Grundlagen der Physik (Erste Mitteilung) Nachr. Göttingen
(1915) 395–407.

[2] É. Cartan Leçons sur les invariants intégraux based on lectures given in 1920-21
in Paris (Hermann, Paris 1922; reprinted in 1958).

[3] Y. Choquet-Bruhat, C. DeWitt-Morette and M. Dillard-Bleick Analysis, Mani-

folds and Physics 2nd ed. (North-Holland, Amsterdam 1982).

[4] I. Agricola and Th. Friedrich, Global analysis: Differential forms in analysis,
geometry and physics transl. from the 2001 German edition, Graduate Studies in
Mathematics, vol. 52 (American Mathematical Society, Providence, RI, 2002).

[5] W. Ślebodziński Sur les équations de Hamilton Bull. Acad. Roy. de Belg. 17
(1931) 864-870.

[6] Th. de Donder Théorie des invariants intégraux (Gauthier–Villars, Paris 1927).

[7] D. van Dantzig Zur allgemeinen projektiven Differentialgeometrie Proc. Roy.

Acad. Amsterdam 35 (1932) Part I: pp. 524–534; Part II: pp. 535–542.

[8] J. A. Schouten, Ricci-Calculus 2nd ed. (Springer-Verlag, Berlin 1954).

[9] A. Lichnerowicz Géométrie des groupes de transformations (Dunod, Paris 1958).

[10] L. Rosenfeld Sur le tenseur d’impulsion-énergie Mémoires Acad. Roy. Belg.,

Classe des Sciences 18 Fasc. 6 (1940) 1–30.

[11] A. Trautman Sur les lois de conservation dans les espaces de Riemann In: Les
Théories Relativistes de la Gravitation: Royaumont 1959 (Éd. du CNRS, Paris
1962) pages 113–116.

[12] A. Trautman, Einstein–Cartan theory, In: Encycl Math. Phys., edited by J.-
P. Françoise, G.L. Naber and Tsou S.T. (Elsevier, Oxford 2006) vol. 2, pages
189–195.

[13] L. Corwin, S. Sternberg and Y. Neeman Graded Lie algebras in mathematics

and physics Rev. Mod. Phys. 47 (1975) 573–603.
302 Remarks on the history of the notion of Lie differentiation

[14] A. Fröhlicher and A. Nijenhuis Theory of vector-valued differential forms Part

I Indag. Math. 18 (1956) 338–359.

[15] I. Kolář, P. W. Michor and J. Slovák Natural Operations in Differential Geom-

etry (Springer-Verlag, Berlin 1993).