cphysics.

org article 03519 2023-12-11

Covariant Derivatives and Curvature

Randy S

Abstract Partial derivatives of tensor fields are generally

not tensor fields. The concept of a covariant derivative is a
modification of the concept of a partial derivative, defined
so that covariant derivatives of tensor fields are still tensor
fields. Curvature tensors are defined in terms of covariant
derivatives. General relativity is formulated with the help of
a special covariant derivative that is metric-compatible and
torsion-free.

Contents

1 Introduction 3

2 Notation and prerequisites 4

3 The commutator of two vector fields 5

4 Covariant derivatives 6

5 Torsion and curvature 8

6 Metric-compatible covariant derivatives 9

7 Uniqueness of the Levi-Civita connection 10

1
© 2018-2024 Randy S
For the latest version and the revision history, visit [Link]/article/03519
[Link] article 03519 2023-12-11

8 Coordinate representations 11

9 The product rule 14

10 Example: covariant derivative of a metric field 15

11 The Levi-Civita connection coefficients 16

12 Definition of the curvature tensor 17

13 Components of the curvature tensor 18

14 The Ricci tensor and curvature scalar 19

15 Sign conventions 20

16 Geodesics 21

17 A covariant version of integration-by-parts 22

18 References 24

19 References in this series 24

2
[Link] article 03519 2023-12-11

1 Introduction
Article 33547 previewed the concept of a geodesic, which generalizes the idea of
a “straight” path to curved spaces (or curved spacetimes). This article introduces
the concept of a covariant derivative, which is the foundation for the concept of
a geodesic. The concept of a covariant derivative is a modification of the concept of
a partial derivative, and it allows writing the equations of motion in classical field
theories (like general relativity) exclusively in terms of tensor fields, which avoids
showing favoritism for any specific coordinate system (article 09894).
The first several sections introduce the concepts using a coordinate-free ap-
proach, and the remaining sections show how to express the same things in a
coordinate representation.

3
[Link] article 03519 2023-12-11

2 Notation and prerequisites

This article assumes familiarity with the coordinate-free definitions and coordinate
representations of tensor fields that were reviewed in article 09894. As in that
k


article, a tensor field is said to be of type m if its inputs are k one-form fields and
m vector fields and its output is a scalar field.1
For clarity, this article uses different brackets (round and square) for different
purposes. Square brackets are used to indicate grouping. Example: x[y + z] means
“first add y and z and then multiply the result by x.” Round parentheses enclose
the input to a function or map. Example: ω(V + W ) is the result of using V + W
as the input to the map ω. Juxtaposition, writing symbols next to each other with
no brackets or parentheses, denotes ordinary multiplication. One exception will be
made: the result of applying a covariant derivative ∇V to an object X will often
be written ∇V X instead of ∇V (X).

1
Lee (1997) uses the opposite notation, with meanings of the upper and lower integers exchanged.

4
[Link] article 03519 2023-12-11

3 The commutator of two vector fields

Recall that a vector field V is a linear derivation from the set of scalar fields to
itself, so it satisfies the same product rule that derivatives satisfy:

V (f g) = V (f )g + f V (g) (1)

for all scalar fields f, g. Consider two vector fields, V and W , applied in succession
to a scalar field f :

V W (f ) .


The composition V W (· · · ) is a map from scalar fields to scalar fields, but it
doesn’t qualify as a vector field because it’s not a derivation (it doesn’t satisfy the
product rule):



V W (f g) 6= V W (f ) g + f V W (g) .
On the other hand, the commutator [V, W ] defined by

[V, W ](f ) ≡ V W (f ) − W V (f )

does qualify as a vector field, because it is a derivation:

[V, W ](f g) = [V, W ](f )g + f [V, W ](g)

for all scalar fields f, g. The commutator will become useful in section 13.

5
[Link] article 03519 2023-12-11

4 Covariant derivatives
A vector field acts like a derivative on scalar fields (equation (1)), but the action
of a vector field on another vector field is undefined.2 This section introduces the
concept of a covariant derivative ∇V , through which a vector field V can act on
other vector fields (and, more generally, on other tensor fields) in a derivative-like
way.3
For any given vector field V , the effect of ∇V on another tensor field is required
to satisfy the following conditions. When applied to a scalar field f , ∇V is required
to act just like the vector field V itself:

∇V f ≡ V (f ). (2)

The effect of ∇V on another vector field is required to have this derivative-like

property when applied to a product f W , where f is a scalar field and W is a
vector field:
∇V (f W ) = [∇V f ]W + f ∇V W. (3)
The action of a covariant derivative on a 1-form field ω is determined by its action
on scalar fields and vector fields through this condition:4



[∇V ω](W ) ≡ ∇V ω(W ) − ω ∇V W . (4)

The action of the covariant derivative on an arbitrary tensor field K is then given
2
The naı̈ve definition V (W ) = V m (∂m W k )∂k is not what we want, because it doesn’t behave like the coordinate
representation of a tensor field.
3
The symbol ∇ here should not be confused with the ordinary gradient with respect to the “space” coordinates,
for which some other articles in this series use the same symbol ∇.
4
The notation [∇V ω](W ) means “apply ∇V to ω to get the new one-form field ∇V ω and then use W as the input
to this new one-form field.”

6
[Link] article 03519 2023-12-11

by the similar condition

[∇V K](X1 , X2 , ...) ≡ ∇V K(X1 , X2 , ...)


− K ∇V X1 , X2 , ...

(5)
− K X1 , ∇V X2 , ...
− ···

where each argument Xn is either a vector field or a 1-form field, whichever is

implied by the tensor field K. In words: the effect of ∇ on a tensor field K is given
by its effect on the scalar field K(X1 , X2 , ...) with its effects on the arguments Xn
subtracted. Finally, when acting on any tensor field X, the covariant derivative is
required to satisfy
∇f V +gW X = f ∇V X + g∇W X (6)
for all scalar fields f, g and all vector fields V, W .
These conditions don’t completely determine ∇, but after we specify its effect
on vector fields, its effect on all other tensor fields is uniquely determined by these
conditions.

7
[Link] article 03519 2023-12-11

5 Torsion and curvature

If the quantity
∇V W − ∇W V − [V, W ] (7)
is not zero for some pair of vector fields V, W , then the covariant derivative is said
to have non-zero torsion. Beware that


[∇V W ](f ) 6= ∇V W (f ) .

In particular, the quantity

∇V W (f ) − ∇W V (f ) − [V, W ](f )

is always identically zero even though the quantity

 
∇V W − ∇W V − [V, W ] (f )

is not necessarily zero (the torsion may be non-zero).

If the quantity  
∇V ∇W − ∇W ∇V − ∇[V,W ] (U ) (8)
is non-zero for some triple of vector fields V, W, U , then the covariant derivative is
said to have non-zero curvature.5 This quantity may be non-zero even though the
quantity  
∇V ∇W − ∇W ∇V − ∇[V,W ] (f )
is identically zero for every scalar field f .
General relativity uses a covariant derivative whose torsion is zero, so the com-
bination (7) is zero in general relativity. The curvature may still be non-zero.

5
According to the proof of theorem 7.3 in Lee (1997), this is consistent with the way the word curved is used in
articles 21808 and 48968.

8
[Link] article 03519 2023-12-11

6 Metric-compatible covariant derivatives

Given a metric field g, a covariant derivative is called metric-compatible if

∇V g = 0

for all vector fields V . According to the definition (5), the preceding condition may
also be written




∇V g(U, W ) = g ∇V U, W + g U, ∇V W . (9)

A metric-compatible derivative that also has zero torsion is called a Levi-Civita

connection.6 This is the covariant derivative used in general relativity. The next
section shows that, given a metric field, the corresponding Levi-Civita connection
is unique.

6
The concept of a covariant derivative specializes the more general concept of a connection, which is a (path-
dependent) way of comparing objects at different points of a manifold. The Levi-Civita connection is an example of
a covariant derivative, so it is also an example of a connection (hence the name).

9
[Link] article 03519 2023-12-11

7 Uniqueness of the Levi-Civita connection

Given a metric field, the corresponding Levi-Civita connection satisfies this identity:





2g ∇V U, W = V g(U, W ) +
U g(V, W ) −
W g(V, U )

+g [V, U ], W − g [V, W ], U − g [U, W ], V . (10)

To derive this, start by re-writing the first term on the right-hand side like this:





V g(U, W ) = ∇V g(U, W ) = g ∇V U, W ) + g U, ∇V W .

The first step holds because g(U, W ) is a scalar field (equation 2), and the second
step holds because the covariant derivative is assumed to be metric-compatible
(equation (9)). The first three terms on the RHS7 of (10) may all be re-written this
way, and the last three terms may be re-written using the zero-torsion condition
like this (using one term as an example):



g [V, U ], W = g ∇V U − ∇U V, W .

After re-writing things in this way, most terms on the RHS of (10) cancel, leaving
only the remainder indicated on the LHS. This completes the derivation.
The right-hand side of (10) does not involve any covariant derivative, so it can
be used to prove that the Levi-Civita connection is unique: use the ∇-independence
of the right-hand side of (10) to see that (10) implies

g ∇V U − ∇0V U, W = 0

for any two Levi-Civita connections ∇ and ∇0 , and all vector fields W . Since g is
a metric field, this implies ∇V U − ∇0V U = 0, so ∇ = ∇0 . This uniqueness result is
called the Fundamental Theorem of Riemannian geometry.8

7
RHS = right-hand side, LHS = left-hand side.
8
Lee (1997), theorem 5.4

10
[Link] article 03519 2023-12-11

8 Coordinate representations
Recall (article 09894) that the coordinate representation of a vector field V is
a combination of partial derivatives, V = V a ∂a , with smooth functions V a as
coefficients. According to the condition (2), the coordinate representation of a
covariant derivative acting on a scalar field is
∇V f = V a ∂a f. (11)
In words: at any given point, the covariant derivative of f along V is just the
partial derivative of f in the direction of the vector field V at that point.
The effect of covariant derivative on other tensor fields is uniquely determined
after we specify its effect on vector fields. One way to do this is to choose a
particular coordinate system and a three-index collection of functions Γbac , and
then impose
∇V W ≡ V a ∂a W b + W c Γbac ∂b .
 
(12)
This clearly satisfies the constraints (3) and (6), so it qualifies as a covariant deriva-
tive. The connection coefficients Γbac are also called Christoffel symbols.9
Every covariant derivative on a vector field can be written this way.10
The connection coefficients are different in different coordinate systems. The
effect of a coordinate transformation on these coefficients may be deduced from the
preceding equations. The result (not shown here) demonstrates that the connection
coefficients Γbac are not the components of a tensor field.  
Equation (12) says that the components of ∇V W are V a ∂a W b + W c Γbac . We
can also write this relationship as
∇a W b = ∂a W b + W c Γbac . (13)
This abbreviation can be misleading, because the left-hand side suggests that ∇a is
something that acts on individual compenents W b , but the right-hand side demon-
strates that it really involves all of W ’s components (a sum over the index c is
9
Lee (1997), chapter 4, page 51, and Wald (1984), section 3.1, page 34. The name Christoffel symbols is often
reserved for a special case, namely for the connection coefficients of the Levi-Civita connection (section 11).
10
Lemma 4.4 in Lee (1997), or equations (3.1.7)-(3.1.15) in Wald (1984)

11
[Link] article 03519 2023-12-11

implied). The abbreviation is very convenient, though: it is standard in texts

about general relativity, and I’ll use it, too. Just remember that ∇ acts on the
whole vector field, not on individual components.11
The coordinate representation of a 1-form is

ω = ωa dxa .

The effect of ∇V on a 1-form can be derived using equations (2), (4), and (12).
According to (2), the first term on the right-hand side of (4) is

∇V ω(W ) = V ω(W ) = V b ∂b ωa W a .


 

According to (12), the second term on the right-hand side of (4) is

   
b a a
ω ∇V W = ω V ∂b W + Γbc W ∂a = V b ∂b W a + Γabc W c ωa .
c

 

When these are combined in (4), the terms involving ∂b W a cancel each other,
leaving
∇V ω(W ) = V b W c ∂b ωc − Γabc ωa .

 

Therefore,
∇V ω = V b ∂b ωc − Γabc ωa dxc .
 
(14)
This is the coordinate representation of the effect
b
 of ∇V ona a 1-form
 field. Equation
(14) says that the components of ∇V ω are V ∂b ωc − Γbc ωa . We can also write
this relationship as
∇a ωb = ∂a ωb − Γcab ωc . (15)
Notice that the Γ term has the opposite sign compared to equation (13).
The coordinate representation of the effect of ∇V on other tensor fields may
now be deduced from the condition (5). The result, expressed in words, is that the
11
In the abstract index notation, the notation W a represents the whole vector field, using the superscript as
a “connector” to indicate how the vector field is used to construct other tensor fields. I’m not using that notation
here.

12
[Link] article 03519 2023-12-11

effect of ∇V on any tensor field consists of one partial-derivative term, plus one
Γ-term for each of the tensor’s superscripts, and minus one Γ-term for each of the
tensor’s subscripts.12 The next section shows an example.
For a given vector field
V , the covariant derivative ∇V is a map from the set
of tensor fields of type mk to itself. If the vector field V is left unspecified, as in
equations (13) and (15), then ∇ converts a tensor field of type mk to a tensor field


k


of type m+1 , with an extra slot to accept the vector field V as an input specifying
the direction of the derivative.

12
Lee (1997), lemma 4.8

13
[Link] article 03519 2023-12-11

9 The product rule

Let A···
··· be the components of a tensor field A, writing “· · · ” in place of a list of
indices. Condition (5) implies the product rule13

∇a [A··· ··· ··· ··· ··· ···

··· B··· ] = [∇a A··· ]B··· + A··· ∇a B··· . (16)

This is an easy consequence of the result that was described in words at the end
of section 8, which says that the left- and right-hand sides of (16) both have the
same Γ-terms. After canceling the Γ-terms, the remainder is the familiar product
rule for the partial derivatives ∂a . The next section shows an example.

13
Wald (1984) uses this as an axiom. This article didn’t do it that way because I didn’t want the definition to
refer to components and I didn’t want to use the abstract index notation.

14
[Link] article 03519 2023-12-11

10 Example: covariant derivative of a metric field

To illustrate the result that was described in words at the end of section 8, consider
a metric field g. For any three vector fields V, U, W , the definition (5) gives




[∇V g](U, W ) = ∇V g(U, W ) − g ∇V U, W − g U, ∇V W .

Use equations (11), (13), and (15) to get the coordinate representation

(∇a gbc )U b W c = ∂a (gbc U b W c ) − gbc (∂a U b + Γba• U • )W c − gbc U b (∂a W c + Γca• W • ).

(To make the pattern more evident, I’m using the symbol • for one of the indices.)
This holds for all U, W , and ∂a satisfies the familiar product rule, so it implies

∇a gbc = ∂a gbc − Γ•ab g•c − Γ•ac gb• . (17)

In words: the effect of ∇V on gab consists of one partial-derivative term minus one
Γ-term for each of gab ’s subscripts. This illustrates the general pattern described
at the end of the previous section.
This result, together with equations (11), (13), and (15), immediately shows
that ∇ satisfies the product rule when acting on gbc U b W c .

15
[Link] article 03519 2023-12-11

11 The Levi-Civita connection coefficients

Equation (12) implies
h i
∇V W − ∇W V − [V, W ] = Γbac − Γbca V a W c ∂b .

This shows that a connection whose coefficients are symmetric in the two subscripts
has zero torsion (and conversely). For the Levi-Civita connection, equation (17)
implies
∂a gbc = Γ•ab g•c + Γ•ac gb• . (18)
By cyclically permuting the index-variables a → b → c → a, we get these equivalent
ways of writing (18):

∂b gca = Γ•bc g•a + Γ•ba gc• . (19)

∂c gab = Γ•ca g•b + Γ•cb ga• . (20)

Subtract equation (18) from the sum of equations (19) and (20) and use the sym-
metry of g and Γ in their subscripts to get

∂b gca + ∂c gba − ∂a gbc = 2Γ•bc g•a ,

and use the fact that the metric is invertible to get

1
Γabc = g a• (∂b gc• + ∂c gb• − ∂• gbc ). (21)
2
These are the conection coefficients for the Levi-Civita connection.

16
[Link] article 03519 2023-12-11

12 Definition of the curvature tensor

Given a covariant derivative ∇, define the curvature endomorphism RX,Y by

RX,Y Z ≡ ∇X ∇Y Z − ∇Y ∇X Z − ∇[X,Y ] Z (22)

for all vector fields X, Y, Z. This satisfies14

RX,Y (f Z) = f RX,Y Z

for any scalar field f . Section 13 mentioned that the quantity (22) is related to the
curvature of the covariant derivative ∇. More precisely, the curvature tensor is
1


the tensor field R of type 3 defined by the condition15

R(X, Y, Z, ω) ≡ ω(RX,Y Z) (23)

for all one-form fields ω.

14
Lee (1997), exercise 7.1
15
Proposition 7.1 in Lee (1997) shows that this does indeed give a tensor field as defined in article 09894.

17
[Link] article 03519 2023-12-11

13 Components of the curvature tensor

In a coordinate representation,

R(X, Y, Z, ω) = Rabc d X a Y b Z c ωd .

The definition (22) clearly implies that the components Rabc d are antisymmetric in
the first two subscripts: Rabc d = −Rbac d . To derive an explicit expression for the
components, we can use a matrix notation in which Z c are the components of a
matrix Z of size N × 1, and for each b, we can define a matrix Γb with components
[Γb ]ac ≡ Γabc where the superscript a is the row index and the subscript c is the
column index.16 In particular, the components of the matrix product Γb Z are
Γabc Z c . With this notation, the coordinate representation of equation (22) may be
written
RX,Y Z = X a (∂a + Γa ), Y b (∂b + Γb ) Z = X a Y b Rab Z
 

with17
Rab ≡ [∂a + Γa , ∂b + Γb ] (24)
with [A, B] ≡ AB − BA, so the components of the curvature tensor are the com-
ponents of the matrix Rab for each a, b:18

Rabc d ≡ [Rab ]dc . (25)

More explicitly,
Rabc d = ∂a Γdbc − ∂b Γdac + Γda• Γ•bc − Γdb• Γ•ac (26)
with an implied sum over the index •.

16
The indices for the components of a matrix are all written as subscripts, using the standard convention in which
the first index labels rows and the second index labels columns.
17
Martin (1988), equation (6.14)
18
Conventions for the overall sign and the order of the indices vary throughout the physics literature.

18
[Link] article 03519 2023-12-11

14 The Ricci tensor and curvature scalar

The Ricci tensor is19
Rab ≡ R•ab • (27)
with an implied sum over the index •. Use equation (26) on the right-hand side of
(27) to get20
Rab = ∂• Γ•ab − ∂a Γ••b + Γ••× Γ× • ×
ab − Γa× Γ•b (28)
with implied sums over • and ×. The Ricci tensor is symmetric:

Rab = Rba .

The first, third, and fourth terms on the right-hand side of (28) are manifestly
individually symmetric (given the symmetry Γcab = Γcba that is evident in equation
(21)), and the symmetry of the second term may be deduced from equation (21)
with a little more effort.
The curvature scalar (or scalar curvature or Ricci scalar) is21

R ≡ g ab Rab . (29)

If we take the covariant derivative ∇ to be the Levi-Civita connection, then the

curvature scalar is a scalar field constructed entirely from the metric field. In gen-
eral relativity, the equationsR governing the behavior the metric field come from the
Einstein-Hilbert action dN x | det g|1/2 R, plus the action for the other fields.
The curvature scalar R is the simplest nontrivial scalar field that can be constructed
entirely from the metric field, so general relativity is the simplest generally covari-
ant model22 in which the behavior of the metric field is governed by a nontrivial
condition.
19
Don’t confuse this with the matrix Rab defined in equation (24).
20
The overall sign of the relationship (28) appears to be standard the physics literature, despite variations in the
conventions used for the curvature tensor (footnote 18).
21
The overall sign of the relationship (29) appears to be standard the physics literature, despite variations in the
overall sign used for the metric tensor itself. Section 15 will explain why this is useful. p
22
General covariance is defined in article 37501, which also mentions the reason for the factor | det g|.

19
[Link] article 03519 2023-12-11

15 Sign conventions
Conventions for curvature tensor vary,23 both in the overall sign and the order of
the indices, but the sign conventions used in equations (21), (28), and (29) appear
to be standard in the literature about general relativity.24 Those equations imply
that the overall sign of the curvature scalar R depends on the sign convention used
for the metric tensor gab (mostly-plus or mostly-minus), but the more important
quantity in general relativity is the Einstein tensor
1
Rab − gab R,
2
and equations (28) and (29) ensure that the sign of this quantity doesn’t depend
the sign convention used for the metric tensor gab . Article 80838 explains why this
is beneficial in the context of the gravitational field equation.

23
The sign convention used in equations (24)-(25) avoids a minus sign in equation (23), and it agrees with the
convention used in Lee (1997).
24
Article 80838 quantifies this statement by citing several sources that all use equations (21), (28), and (29) even
though they differ in their conventions for the curvature tensor.

20
[Link] article 03519 2023-12-11

16 Geodesics
Given a covariant derivative ∇, a vector field U is said to be parallel transported
if
∇U U ∝ U. (30)
Covariant derivatives are defined on vector (and tensor) fields, not on individual
worldlines, but given a worldline xa (λ), we can think of its tangent vectors ẋa ≡
dxa /dλ as part of a spacetime-filling vector field U a that is equal to ẋa along the
worldline.25 In this case, the identity

U a ∂a U b = ẍa

holds everywhere along the worldline, where each overhead dot represents a deriva-
tive with respect to the parameter λ along the worldline. Using this in (30) gives

ẍa + Γabc ẋb ẋc ∝ ẋa . (31)

Most worldlines don’t satisfy this condition, because most worldlines are not parallel-
transported along themselves with respect to the given covariant derivative ∇. A
worldline that is parallel-transported along itself is called a geodesic. Given a
geodesic, we can always parameterize it in such a way that the proportionality
factor in equation (30) is zero. This is called an affine parameterization. When
an affine parameterization is used, the geodesic equation (31) reduces to

ẍa + Γabc ẋb ẋc = 0. (32)

Article 33547 highlights the geometric significance of this equation when the co-
variant derivative ∇ is taken to be the Levi-Civita connection corresponding to a
given metric field.

25
Wald (1984), page 34

21
[Link] article 03519 2023-12-11

17 A covariant version of integration-by-parts

In this section, ∇ denotes the Levi-Civita connection. Let | det g| denote the mag-
nitude of the determinant of the matrix with components gab (x), the components
of the metric tensor. This section derives the identity26

p  p
a
∂a | det g| V = | det g| ∇a V a , (33)

which holds for the components V a of any vector field V . Taken together with
the product rule (equation (16)), this identity justifies an analog of integration-
by-parts, but with covariant derivatives instead of ordinary derivatives. As an
example, suppose V a = Aab Bb . Then equations (16) and (33) imply
Z p  Z p Z p
dN x ∂a | det g| V a = dN x | det g| (∇a Aab )Bb + dN x | det g| Aab (∇a Bb ),

so if the left-hand side is zero, then the two terms on the right hand side are each
other’s negatives.
To derive the identity (33), start with this identity from article 18505:
δ
det M = (det M )(M −1 )ab ,
δMab
which implies X
∂a det M = (det M ) (M −1 )bc (∂a Mbc ).
b,c

This holds for any invertible matrix M whose components are functions of the
coordinates. Specializing this to the matrix whose components are the components
gab of the metric tensor gives
∂a | det g| = | det g| g bc ∂a gbc , (34)
26
This is equation (3.4.10) in Wald (1984).

22
[Link] article 03519 2023-12-11

which implies
p 1p
∂a | det g| = | det g| g bc ∂a gbc .
2
This implies
p  p 1p
∂a | det g| V = | det g| ∂a V a +
a
| det g| (g bc ∂a gbc )V a .
2
The explicit expression for Γcab in terms of the metric (section 21) implies
1
Γbba = g bc ∂a gbc ,
2
and using this in the preceding equation gives
p  p
| det g| V = | det g| ∂a V a + Γbba V a .
a


∂a

On the right-hand side, the quantity in parentheses is the coordinate representation

of ∇a V a , so this completes the derivation of the identity (33).

23
[Link] article 03519 2023-12-11

18 References
Lee, 1997. Riemannian Manifolds: An Introduction to Curvature. Springer

Martin, 1988. General Relativity: A guide to its Consequences for Gravity and
Cosmology. John Wiley & Sons

Wald, 1984. General Relativity. University of Chicago Press

19 References in this series

Article 09894 ([Link]
“Tensor Fields on Smooth Manifolds” (version 2024-04-16)

Article 18505 ([Link]

“Matrix Math” (version 2023-02-12)

Article 21808 ([Link]

“Flat Space and Curved Space” (version 2023-11-12)

Article 33547 ([Link]

“Free-Fall, Weightlessness, and Geodesics” (version 2024-02-25)

Article 37501 ([Link]

“Why the (Hilbert) Stress-Energy Tensor is Covariantly Conserved” (version 2022-02-18)

Article 48968 ([Link]

“The Geometry of Spacetime” (version 2024-02-25)

Article 80838 ([Link]

“Sign Conventions in General Relativity” (version 2024-03-04)

24