1.11 Tensor Algebra Convention Foundations

Tensor Algebra Convention Foundations is the set of arbitrary but fixed choices, agreed upon within a given context, that make tensor expressions unambiguous even though the underlying mathematics does not force any one particular choice. Where notation supplies the symbols and formatting rules for writing tensors, convention supplies the specific decisions, which index range to use, how to order indices, which sign to attach to an antisymmetric quantity, that must be settled before a notation can be applied consistently, since more than one equally valid choice is always available in principle.

These conventions are foundational precisely because they are invisible when followed correctly and a persistent source of error when mismatched: two documents using the Einstein summation convention with index notation can still disagree in their results if one assumes indices range from 0 while the other assumes they range from 1, or if one places the row index of a matrix as upper while the other places it as lower. Tensor algebra convention foundations catalog the specific points at which such choices arise and states the most common resolutions adopted across the field.

Index Range Conventions

Starting at One Versus Starting at Zero

Indices in tensor expressions are conventionally taken to range over 1 through n, matching classical linear algebra and differential geometry usage, though computer science and some physics contexts instead range indices from 0 through n − 1, matching array-indexing conventions used in most programming languages. Both choices are internally consistent, but mixing them within a single derivation, using 1-based ranges for one tensor and 0-based ranges for another, produces off-by-one errors that are easy to introduce and difficult to trace.

Greek Versus Latin Index Letters

In many physics contexts, a further convention distinguishes Greek indices, such as μ, ν, which are taken to range over all coordinates including a distinguished time-like coordinate, from Latin indices, such as i, j, which are restricted to range only over the remaining spatial coordinates. This letter-based convention communicates, at a glance, whether a distinguished coordinate is included in a given sum, without requiring an explicit restatement of the range each time.

Ordering Conventions

Row Index Before Column Index

For a matrix regarded as a (1, 1) tensor, the standard convention places the upper index first, corresponding to the row, and the lower index second, corresponding to the column, so that M^i_j names the entry in row i, column j. This ordering is what makes the standard formula for matrix-vector multiplication and matrix-matrix multiplication produce results consistent with ordinary matrix arithmetic; reversing the convention without adjusting the corresponding formulas would produce the transpose of the intended result.

Left-to-Right Reading of Multi-Index Tensors

For a tensor with several indices of the same type, such as T_ijk, the convention is that the order in which the indices are written corresponds to the order of the arguments in the associated multilinear map: T_ijk denotes the value of a trilinear form evaluated with its first argument matched to i, second to j, third to k. Because many multilinear operations are not symmetric in their arguments, this ordering convention is essential to interpreting the tensor correctly, and swapping the written order of the indices, without an explicit symmetrization, generally changes the value being described.

Sign Conventions

The Antisymmetric Sign Convention

When defining the antisymmetric part of a tensor's indices using bracket notation, the standard convention assigns a plus sign to permutations of even parity and a minus sign to permutations of odd parity, normalized so that the operation is idempotent, applying it twice reproduces the same antisymmetric part rather than doubling it.

Sign Convention for the Levi-Civita Symbol

The Levi-Civita symbol, used to express determinants and cross-product-like operations in index notation, is conventionally defined to equal +1 for an even permutation of its indices, −1 for an odd permutation, and 0 whenever any two indices repeat. Some sources additionally distinguish the Levi-Civita symbol, a fixed array of numbers independent of any metric, from the Levi-Civita tensor, which incorporates a factor of the square root of the metric determinant; conflating the two, or forgetting which convention a given source uses, is a common source of sign errors in computations involving orientation or volume.

Basis and Frame Conventions

Orthonormal Versus General Bases

A significant convention decision is whether components are expressed relative to an orthonormal basis, in which the distinction between upper and lower indices collapses because the metric components equal the Kronecker delta, or relative to a general, possibly non-orthonormal, basis, in which upper and lower indices remain genuinely distinct and must be tracked carefully. Many introductory treatments default to orthonormal bases specifically to avoid this distinction, while more advanced treatments require the general case, and confusion often arises when a formula learned under one assumption is applied under the other.

Active Versus Passive Transformation Convention

When describing how a tensor changes under a transformation, a passive convention interprets the change as relabeling the same fixed object using a new coordinate system, while an active convention interprets the change as physically moving or altering the object itself while the coordinate system stays fixed. The transformation law for tensor components is usually stated under the passive convention, and care must be taken when switching to an active interpretation, since the direction of certain Jacobian factors reverses between the two conventions.

Type-Ordering Convention for Mixed Tensors

Interleaved Versus Grouped Index Placement

For a tensor with multiple upper and multiple lower indices, one convention groups all upper indices together followed by all lower indices, as in T^{ij}_{kl}, while another convention interleaves them in an order reflecting a specific sequence of arguments, as in T^i_k{}^j_l. Interleaved notation is used when the relative position of upper and lower indices carries additional meaning, such as tracking which argument slot corresponds to which index, while grouped notation is simpler to read when no such distinction is needed. Consistency in choosing one style throughout a derivation is essential, since the two are not always trivially interchangeable once symmetrization or contraction is involved.

Why Explicit Convention-Setting Matters

Conventions Are Arbitrary but Not Optional

None of the conventions described here are forced by the underlying mathematics; each represents one workable choice among several equally valid alternatives. What is not optional is picking one and adhering to it consistently throughout a given piece of work, since tensor algebra's compact notation offers few internal safeguards against a silently inconsistent convention, unlike, for example, dimensional analysis, which can catch some errors automatically.

Convention Statements as a Prerequisite for Communication

Because different sources, textbooks, and fields adopt different conventions for index ranges, sign, and ordering, careful treatments of tensor algebra typically state their conventions explicitly at the outset, and reading unfamiliar tensor material productively requires first identifying which conventions are in force before attempting to reconcile its formulas with those learned elsewhere.