Contraction
Relevant parts to questions...
- Set two indices of a tensor equal and invoke the summation convention. Each contraction drops the rank by 2.
- In generalised coordinates, contract one upper with one lower index only.
- Outer product followed by contraction = inner product.
- Orthogonality is what makes the result a tensor.
Contraction takes a rank- tensor (with ) and reduces it to a rank- tensor by summing over a pair of its indices. Setting two indices equal and invoking the summation convention is the whole operation.
For a rank-3 covariant tensor , there are three distinct ways to contract:
- (contract first two),
- (contract last two),
- (contract first with third).
Each is a rank-1 tensor (a vector).
Why Contraction Preserves Tensor Character
In Cartesian coordinates, . Contracting the first two indices by setting :
which is exactly the transformation law of a rank-1 tensor. The orthogonality relation collapses the two -factors into a Kronecker Delta, dropping the rank by 2.
Generalised Coordinates: Contract Upper with Lower
Only valid with mismatched index positions
In curvilinear coordinates, contraction only produces a tensor when one index is covariant and the other contravariant. Contracting two upper (or two lower) indices does not give a tensor, because in general.
So is fine (one up, one down), but contracted over and fails.
Repeated Contraction
Since each contraction drops the rank by 2:
- A rank- tensor can be contracted times.
- If is even, repeated contraction eventually gives a scalar.
- If is odd, it eventually gives a vector.
The Inner Product
The inner product of two tensors is an outer product followed by a contraction:
All are inner products. They combine two tensors and immediately reduce the rank by contracting matching (upper-lower) index pairs.
Properties
- Rank drop::one contraction removes 2 from the rank; two contractions remove 4; etc.
- Preserves tensor character when indices are at opposite levels (upper ↔ lower).
- Familiar identities from the Kronecker Delta and Alternating Tensor::
- (contract alternating against symmetric)
- .
Applications
- Reducing rank to produce a simpler tensor - e.g. the trace of a matrix is the full contraction of a rank-2 tensor.
- Forming scalars and invariants from higher-rank objects - e.g. is a coordinate-invariant trace.
- Computing inner products cleanly in any coordinate system by pairing upper with lower indices.
Trace as a contraction
Contract the rank-2 tensor : . This is a scalar - the trace.
Set two indices equal; rank drops by 2; upper-with-lower in curvilinear coords.