Covariant and contravariant components

WHY do we even need two kinds of components?

WHAT is the problem? A vector $\mathbf{v}$ is a real geometric arrow — it doesn't care about coordinates. But numbers describing it only exist relative to a chosen basis $\{\mathbf e_1, \mathbf e_2\}$ . If the basis vectors are not orthonormal (different lengths, not at $90°$ ), there is genuine ambiguity in "how much of $\mathbf e_1$ is in $\mathbf v$ ".

WHY ambiguity? Two reasonable rules:

Parallelogram rule — slide along the axes to reach the tip. These coefficients $v^i$ are the contravariant components, written with an upper index: $\mathbf v = v^1 \mathbf e_1 + v^2 \mathbf e_2 = v^i \mathbf e_i$
Orthogonal-projection rule — drop a perpendicular onto each axis and take the dot product. These give the covariant components $v_i$ (lower index): $v_i = \mathbf v \cdot \mathbf e_i$

Start from the definition of covariant components and plug in the contravariant expansion. Why this step? Because covariant components are defined by dotting with $\mathbf e_i$ , and we already know $\mathbf v$ as a sum of $\mathbf e_j$ 's — so substitute.

$v_i = \mathbf v \cdot \mathbf e_i = (v^j \mathbf e_j)\cdot \mathbf e_i = v^j\,(\mathbf e_j \cdot \mathbf e_i)$

The quantity $\mathbf e_i\cdot\mathbf e_j$ appears naturally. Name it:

WHY is $g_{ij}$ the identity when orthonormal? Then $\mathbf e_i\cdot\mathbf e_j = \delta_{ij}$ , so $v_i = v^i$ . The whole distinction collapses. That's your sanity check.

The dual (reciprocal) basis — the deep reason for "covariant"

WHAT object do covariant components multiply? There is a second basis $\{\mathbf e^i\}$ , the dual basis, defined by: $\mathbf e^i \cdot \mathbf e_j = \delta^i_j$ i.e. $\mathbf e^1$ is perpendicular to $\mathbf e_2$ and scaled so $\mathbf e^1\cdot\mathbf e_1=1$ . Then the same vector has two clean expansions: $\mathbf v = v^i\mathbf e_i = v_i \mathbf e^i$ So covariant components are just contravariant components with respect to the dual basis. Beautiful symmetry.

Figure — Covariant and contravariant components

WHY the names "co-" and "contra-"? (transformation law)

Change basis $\mathbf e_i \to \tilde{\mathbf e}_i = A^j{}_i\,\mathbf e_j$ .

Basis vectors transform with $A$ . For $\mathbf v$ to stay the same arrow, its contravariant components must transform with the inverse $A^{-1}$ — they go against (contra) the basis change. Meanwhile covariant components transform with (co) the basis, like the basis vectors do. That naming is the entire point of tensor calculus.

Worked examples

Worked example Example 1 — a concrete 2D skew basis

Let $\mathbf e_1 = (1,0)$ , $\mathbf e_2 = (1,1)$ (Cartesian background). Take vector $\mathbf v = (2,1)$ .

Step 1 — metric. $g_{11}=\mathbf e_1\!\cdot\!\mathbf e_1=1$ , $g_{12}=\mathbf e_1\!\cdot\!\mathbf e_2=1$ , $g_{22}=\mathbf e_2\!\cdot\!\mathbf e_2=2$ . Why? Geometry first; everything follows from $g$ . $g=\begin{pmatrix}1&1\\1&2\end{pmatrix},\quad g^{-1}=\begin{pmatrix}2&-1\\-1&1\end{pmatrix}$

Step 2 — contravariant (solve $v^1\mathbf e_1+v^2\mathbf e_2=\mathbf v$ ): $v^1+v^2=2,\ v^2=1\Rightarrow v^1=1,\ v^2=1$ . Why? These are the parallelogram coefficients.

Step 3 — covariant by dotting: $v_1=\mathbf v\!\cdot\!\mathbf e_1=2$ , $v_2=\mathbf v\!\cdot\!\mathbf e_2=2+1=3$ . Why? Perpendicular projections, the definition.

Step 4 — check the metric link. $v_i=g_{ij}v^j$ : $v_1=1(1)+1(1)=2$ ✓, $v_2=1(1)+2(1)=3$ ✓.

Common mistakes (steel-manned)

Recall Feynman: explain to a 12-year-old

Imagine drawing a vector on stretchy, slanted graph paper instead of normal square paper. To describe your arrow you can either say "walk this far along the slanted lines" (contravariant) or "how much shadow does the arrow cast straight down onto each line" (covariant). On normal square paper both answers are the same. On slanted paper they're different — so we keep two sets of numbers, and a little table called the metric translates between them.

Flashcards

Contravariant components

v^i

are defined as

the coefficients in the parallel (parallelogram) expansion

\mathbf v = v^i\mathbf e_i

; index up.

Covariant components

v_i

are defined as

the orthogonal projections

v_i=\mathbf v\cdot\mathbf e_i

; index down.

The metric tensor

g_{ij}

is defined as

\mathbf e_i\cdot\mathbf e_j

— it encodes all lengths and angles of the basis.

Formula to lower an index

v_i = g_{ij}v^j

Formula to raise an index

v^i = g^{ij}v_j

, where

g^{ij}

is the inverse metric.

When do covariant and contravariant components coincide?

When the basis is orthonormal, since then

g_{ij}=\delta_{ij}

The dual basis

\mathbf e^i

satisfies

\mathbf e^i\cdot\mathbf e_j=\delta^i_j

; covariant components are coefficients in this basis.

Invariant squared length of

\mathbf v

|\mathbf v|^2 = v^iv_i = g_{ij}v^iv^j

(must pair up with down).

Under basis change with matrix

A

, contravariant components transform with

A^{-1}

(against the basis — "contra").

Why is the gradient naturally covariant?

\partial_i f=\partial f/\partial x^i

carries a lower index; raising needs

g^{ij}

Connections

Metric tensor
Dual (reciprocal) basis
Tensors and index notation
Change of basis and transformation laws
Inner product spaces
Curvilinear coordinates (polar, spherical)
General relativity — raising and lowering indices

Concept Map

Hinglish (regional understanding)

Intuition Hinglish mein samjho

Dekho, normal square graph paper pe to koi tension nahi — ek vector ke components nikalo, sab clear. Lekin jab basis skewed ho (axes 90° pe nahi, ya alag-alag length ke), tab ek hi vector ko padhne ke do tareeke ban jaate hain. Pehla: axes ke parallel chal ke tip tak pahuncho — yeh hain contravariant components $v^i$ (upar index). Doosra: vector ka perpendicular shadow har axis pe daalo, yaani dot product $\mathbf v\cdot\mathbf e_i$ — yeh hain covariant components $v_i$ (neeche index). Orthonormal basis me dono same nikalte hain, isiliye school me kisi ne bataya hi nahi.

Inke beech translator hota hai metric tensor $g_{ij}=\mathbf e_i\cdot\mathbf e_j$ . Yeh matrix basis ki saari geometry (lengths, angles) store karta hai. Index ko neeche laana ho to $v_i=g_{ij}v^j$ , upar le jaana ho to inverse metric se $v^i=g^{ij}v_j$ . Bas yaad rakho: lower karne ke liye $g_{ij}$ , raise karne ke liye $g^{ij}$ .

Sabse important practical baat: jab bhi koi invariant (jo coordinates pe depend na kare) banana ho, jaise length, to ek upar wala index ek neeche wale se contract karo — $v^iv_i$ . Agar tum dono upar wale jod doge $v^iv^i$ , to galat answer aayega kyunki tum chupke se metric ko identity maan rahe ho. Yahi gravity (General Relativity) aur curved coordinates me sab kuch chalata hai.

"Contra" ka matlab basis ke against transform hona ( $A^{-1}$ ), aur "co" ka matlab basis ke saath ( $A$ ). Naam se hi unka behaviour pata chal jaata hai — yeh tensor calculus ka asli dil hai.

WHY do we even need two kinds of components?

The dual (reciprocal) basis — the deep reason for "covariant"

WHY the names "co-" and "contra-"? (transformation law)

Worked examples

Common mistakes (steel-manned)

Flashcards

Connections

Concept Map

Hinglish (regional understanding)

Go deeper — visual, from zero

Test yourself — Advanced Topics (Elite Level)

Connections