Index juggling is the final remaining element of the tensor notation. While subtle, it is powerful
and leads to an even greater economy of notation and, therefore, an even higher level of algebraic
organization.
As we have established in the preceding chapters, the tensor framework relies on two flavors
of indices: superscripts, also known as contravariant indices, and subscripts, also known as
covariant indices. We will explain the substantive difference between the two types
of indices in Chapter 14. However, we have already
had occasion to observe that in the course of a disciplined analysis, the placement of indices is
naturally dictated by the rules of the tensor notation. Furthermore, once the natural placement of
indices is determined, it, in turn, reveals to us the proper combinations of objects we ought to
consider in order to arrive at geometrically meaningful results. This ability of the notation to
lead, rather than follow, the algebraic intuition is a unique and powerful feature of Tensor
Calculus. Therefore, in some ways, the tensor notation and its implications are actually more
important than the tensor concept itself.
11.1The definition of index juggling for first-order systems
Index juggling refers to the practice of denoting a contraction with the metric tensor by changing
the placement of the participating index. In fact, we have already encountered this practice on two
occasions. Indeed, in Chapter 9, we established
that the covariant and contravariant bases and
are
related by a contraction with the metric tensor, i.e.
(Notice that the first equation was, in fact, the definition of the contravariant basis .)
Similarly, the contravariant and covariant components and
of a
vector are similarly related by contraction with the metric tensor,
i.e.
In general, denoting the result of contracting a system
with the contravariant matric tensor ,
i.e.
is referred to as raising the
index. Meanwhile, denoting the result of contracting a system with
the covariant metric tensor ,
i.e.
is referred to as lowering the
index.
Note that the last two equations are equivalent by the inversion of the metric tensor
described in Section 9.5.2. This, if is
obtained from by
lowering the index, then is
obtained from by
raising the index. In other words, the operations of lowering and raising indices are the inverses
of each other. Had this not been the case, using the same letter to denote objects related by index
juggling would lead to ambiguities as, for instance, the symbol would
denote both an original system as well as the result of lowering and subsequent raising of
the index.
11.2Index juggling and higher-order systems
For a higher-order system, any of the indices can be individually raised or lowered. For
example, consider a third-order system
with three subscripts. To raise the first index, contract
with the contravariant metric tensor to
produce the system
T_{jk}^{r}=Z^{ri}T_{ijk}.tag{} end{equation} Note that in this situation, the use of
the dot placeholder described in Section 7.2 is essential.
Otherwise, it is not possible to say whether the symbol, say,
represents the raising of the first index on the system , the
second index on , or
the third index on . The
use of a placeholder removes this ambiguity by making the order of the indices clear. With the help
of the placeholder, the above identity reads
and it is clear simply by looking at
the symbol
that it represents the result of raising the first index of a system .
Contracting
with
-- notice the superscript in
-- raises the second index, i.e.
The indicial signature of the symbol
clearly indicates that the superscript is the second index. Similarly, the equation
represents the raising of the third index of .
Once one index of a multi-dimensional system has been juggled, another can be juggled, as well. For
example, raising the second index of
yields
When two indices are juggled one
after another, the order in which the two steps take place is immaterial. This is a direct
consequence of the commutative property of elementary multiplication since
is also given by a double contraction, i.e.
Thus, two juggling operations may be
viewed as simultaneous.
Finally, any number of indices can be juggled at once. For example, raising all three indices of
yields a system with three superscripts, i.e.
11.3Juggling dummy indices
Recall that, according to the Einstein convention, dummy indices represent a contraction.Thus,
dummy indices, being already involved in a summation, are not available for contraction with the
metric tensor. However, index juggling allows such indices to exchange flavors. For a most basic
example, let us show that
The idea is to "extract" the metric
tensor from by
noting that :
and subsequently let it get
"reabsorbed" into :
Finally, renaming into , we arrive at the desired result
In summary, the flavors of the dummy
indices in any contraction can be switched at will.
A word of caution is in order for situations where one of the systems appears under a differential
operator of some kind. For example, suppose that systems and
are
defined along a trajectory , and thus these systems may be
considered functions and of . Then it would be wrong to let the dummy indices exchange
flavors in the contraction
To see why this is so, let us
attempt to repeat the logic used in the previous example. The product in
occurs under the derivative
operator. Therefore, the next step is an application of the product rule:
Since the metric tensor varies from
one point to another, the derivative is not zero and we thus have a new term compared to our
previous analysis. Multiplying out the right side, we find that
and, therefore,
or, having renamed into ,
Thus, we find that
due to the variability of the
covariant basis. We can describe the effect we just observed by saying that dummy indices cannot
be juggled across derivatives. This is something that we should be cognizant of when we begin
to study the covariant derivative in Chapter 15.
11.4Juggling free indices in tensor identities
In a tensor identity, a free index can be consistently raised or lowered on both sides of the
equation. For example, the identity
implies a similar one in which the
free index appears as a superscript:
To see why this is true, contract both sides of the original identity with :
which yields
Now, rename into to obtain the final form
We will refer to the net effect of
this manipulation as raising or lowering an index on both sides of the equation.
The same word of caution regarding index juggling in the presence of differential operators must be
repeated here: free indices cannot be juggled across derivatives. For example, the
superscript cannot be lowered in the identity
Indeed, contracting both sides with
yields
On the left,
combines to
produce .
On the right, however, is
separated from by
the derivative. It can be brought inside the derivative by an application of the product rule
Therefore, we find that
and, once again, there is an extra
term due to the variability of the metric tensor, leading to the conclusion that
does not imply
11.5The equivalence of the metric tensor and the Kronecker delta
Index juggling can be applied to any index in any system. Let us now apply it to the Kronecker
delta .
As we discussed in Section 7.2, the order of the indices of
the Kronecker delta does not matter. Nevertheless, in order to avoid any potential ambiguities, we
will count the superscript as the first index and the subscript as the second, and therefore write
as
.
Let us now lower the superscript by contracting
with .
Consistent with our convention of using the same letter for objects related by index juggling,
denote the result by :
We must emphasize the fact that the
elements of no
longer equal when and otherwise. Those are the values of the proper Kronecker
delta .
The values of ,
on the other hand, depend on the choice of coordinates and vary from one point to another.
As a matter of fact, we can determine the values of the elements by
looking at the combination
from a different perspective. Instead of thinking of as
lowering the index on ,
let us think of it as
renaming the index into on and
thus yielding :
Matching up the equivalent outcomes
of the two interpretations, we discover that the elements of
are precisely the same as those of :
Thus, the Kronecker delta
and the metric tensor
are related by index juggling and are therefore thought of as two manifestations of the exact same
object.
This conclusion can be reinforced by raising one of the indices of .
Let us raise the first index of by
contracting it with and
denote the result by
By the very definition of the
contravariant metric tensor as the matrix inverse of the covariant metric tensor, we know that
Therefore
and we once again observe the
equivalence of the Kronecker delta and the metric tensor.
Summing up, the systems
and
are related by index juggling and are therefore different manifestations of the same object. Thus,
the letters and in the symbols
and
can be used interchangeably. For example, the symbols
and
are legitimate representations of the covariant and the contravariant metric tensors. Meanwhile,
is
a legitimate representation of the Kronecker delta. Nevertheless, in practice, we will always use
the symbol
for the Kronecker delta and
and
for the metric tensors. Additionally, we will go back to using the symbol
without a placeholder since, as we discussed in Section 7.2, this cannot lead to ambiguity.
11.6The effect of index juggling on the tensor notation
Index juggling is a powerful feature of the tensor notation, as it greatly increases the
succinctness of tensor expressions. Just compare the expressions and
that represent the dot product of the vectors and . It is hardly debatable that the
former expression is more appealing than the latter: it has fewer terms, fewer indices, fewer
contractions, a sense of algebraic transparency, and a general sense of lightness and dynamism.
Thanks to index juggling, these attractive features are found in many tensor expressions. For
example, it is rare for any formula to contain an explicit reference to the metric tensor since any
instance of the metric tensor engaged in a contraction with another system gets absorbed into that
system. For a similar reason, the identity matrix rarely appears in Linear Algebra
expressions. After all, if it is featured in a matrix product, such as , it simply gets absorbed into the matrix that it
multiplies, i.e. .
The only way a metric tensor can explicitly appear in a tensor identity is when both of its indices
are free. Two notable examples of such expressions appear in this book. The first is the statement
that two systems
and
are matrix inverses of each other, i.e.
The second is the formula
found in the next volume where we
will give a tensor description of embedded surfaces. This formula states that the sum of the two
projection operators, onto the surface and away from the surface, is the identity
operator. In other words, a vector is a sum of its orthogonal projections onto and
away from the surface.
11.7Exercises
Exercise 11.1Show that if we lower the index on to produce and then raise the index on to produce , then
This exercise confirms that the operations of lowering and raising indices are the inverses of each other. Furthermore, it shows that using the same letter for systems related by index juggling cannot lead to ambiguities.
Exercise 11.2Show that the result of raising both indices on the covariant metric tensor is the contravariant metric tensor
Exercise 11.3Consider a coordinate system in which the metric tensor at a given point corresponds to the matrix
Suppose that is a fourth-order system whose sole nonzero element at the same point is
Find the values of the elements and at that point.