📐 Math · Differential Geometry · Foundations Foundations 2 October 24, 2026

Superscripts and Subscripts: Coordinates $x^i$ vs $x_i$

Is $x^2$ the square of $x$, or the second coordinate? In this series, it will mean the second coordinate. That tiny raised number is a label, and whether we write it up or down will turn out to carry real information.

Last time, we quietly wrote some indices upstairs and others downstairs. Today we explain the rule.

1. An Index Is a Label, Not a Power

In a space with $n$ coordinates, we name them $x^1, x^2, \dots, x^n$, or all at once, $x^i$. The $i$ simply says which coordinate we mean. When we really want a power, we will say so or use parentheses, as in $(x^i)^2$.

2. Where Do the Indices Go?

From last time, we already know the pattern. The basis vectors $\partial_i$ carry a lower index, while the components $v^i$ of a vector carry an upper one. For covectors, it is reversed: the components $\alpha_i$ are lower, and the basis $dx^i$ is upper.

ObjectSymbolIndexTransforms by
Basis vectors$\partial_i$down$A$
Vector components$v^i$up$A^{-1}$
Dual basis (differentials)$dx^i$up$A^{-1}$
Covector components$\alpha_i$down$A$
Gradient components$\dfrac{\partial f}{\partial x^i}$down$A$

The last row shows a useful trick: an index in the denominator counts as a lower index. That is why $\dfrac{\partial}{\partial x^i}$ is written $\partial_i$.

So the position of an index is a code. Upstairs means “transforms like the components of a vector.” Downstairs means “transforms like the basis.”

Why do coordinates themselves get upper indices? Because their small changes, $dx^i$, transform like vector components. In general, the coordinates $x^i$ are not components of a vector. Only their differentials are.

3. The Summation Convention

Writing out sums gets tedious, so we drop the sum sign:

$$v = \sum_{i=1}^{n} v^i\,\partial_i \quad\longrightarrow\quad v = v^i\,\partial_i$$

The rule: when an index appears once upstairs and once downstairs, we sum over it. A covector acting on a vector, for example, is:

$$\alpha(v) = \alpha_i\, v^i$$

Three rules keep the bookkeeping honest:

  • An index appears at most twice in a term.
  • A repeated index must be once up and once down. It is then a dummy index, and its name does not matter: $\alpha_i v^i = \alpha_k v^k$.
  • An index that appears once is free, and it must sit in the same position (up or down) in every term of an equation.

Let’s test the rules on a formula from last time:

$$\tilde{v}^i = \frac{\partial \tilde{x}^i}{\partial x^j}\, v^j$$

The index $i$ is free. It is upstairs on the left, and upstairs (in the numerator) on the right. The index $j$ appears once downstairs (in the denominator) and once upstairs (in $v^j$), so it is summed. Everything balances, which is a good sign the formula is correct.

4. The Kronecker Delta

One small symbol shows up constantly: $\delta^i_j$, which is $1$ when $i=j$ and $0$ otherwise. It tells us that coordinates do not depend on each other, and that the dual basis really is dual to the basis:

$$\frac{\partial x^i}{\partial x^j} = \delta^i_j, \qquad dx^i(\partial_j) = \delta^i_j$$

The second equation says that $dx^i$ measures the $i$-th component of a vector. It returns $1$ on its own basis vector and $0$ on the others.

5. Why $x_i$ Is Not the Same as $x^i$

In school, we never wrote both $x^i$ and $x_i$, because with perpendicular axes they are equal. To see them differ, we need axes that are not perpendicular.

Take two unit basis vectors $e_1, e_2$ at sixty degrees, so that $e_1\cdot e_1 = e_2\cdot e_2 = 1$ and $e_1\cdot e_2 = \tfrac12$. There are two natural ways to describe a vector $v$:

  • Upper components $v^i$: how much of each basis vector we add up, so $v = v^1 e_1 + v^2 e_2$.
  • Lower components $v_i$: how much $v$ lines up with each basis vector, so $v_i = v\cdot e_i$.

Take $v = e_1 + e_2$, so its upper components are $v^i = (1, 1)$. Its lower components are:

$$v_1 = v\cdot e_1 = e_1\cdot e_1 + e_2\cdot e_1 = 1 + \tfrac12 = \tfrac32, \qquad v_2 = \tfrac32$$

The two lists differ: $(1,1)$ versus $(\tfrac32, \tfrac32)$. With perpendicular axes, the cross terms vanish and they would agree.

Now pair them up. One up, one down, so we sum:

$$v^i v_i = 1\cdot\tfrac32 + 1\cdot\tfrac32 = 3 \;=\; |v|^2$$

Check directly: $|v|^2 = (e_1+e_2)\cdot(e_1+e_2) = 1 + 2\cdot\tfrac12 + 1 = 3$. It matches.

Notice that going from $v^i$ to $v_i$ required a dot product, which means a notion of length and angle. That extra structure, which we will formalize later, is not part of the manifold itself. Without it, vectors have upper components, covectors have lower ones, and there is no canonical way to turn one into the other.

6. Up Meets Down: Invariants

The pairing of one upper and one lower index is not just notation. Last time, we saw that components and basis change in opposite ways, so their combination is unchanged. The same happens here:

$$\alpha_i v^i \quad\text{is the same number in every coordinate system.}$$

A sum over two upper indices, or two lower ones, has no such guarantee. If you ever see it, treat it as an alarm.

7. One-Line Summary

Upper and lower indices are not decoration. Upstairs means “transforms like vector components,” downstairs means “transforms like the basis,” and a sum over one up and one down index always gives a number that does not depend on coordinates.
Next episode. We have used $\partial_i$ as a basis vector without asking what it really is. In Foundations 3, we look at the tangent space it lives in. The question from Foundations 1, how to compare vectors at different points, returns in Foundations 5.
index notation summation convention Kronecker delta covectors