Visual walkthrough — Vector fields — definition, visualization

Before any symbol appears, two words we will lean on:

---

Step 1 — Start with what you already know: a number at each point

WHAT. A familiar function $f(x,y)$ takes a place and hands back a single number. Think of a temperature map: stand anywhere, read one number (how hot).

$$\underset{\text{the rule}}{\underbrace{f}}\left(\underset{\text{the place}}{\underbrace{x,y}}\right)=\underset{\text{one number out}}{\underbrace{z}}$$

• $f$ — the rule, the machine.
• $(x,y)$ — the place you feed in.
• $z$ — the single number that comes out (a height, a temperature, a price).

WHY start here. A vector field is the next idea up. If you can see clearly what "a number at each point" looks like, the jump to "an arrow at each point" is one honest step, not a mystery.

PICTURE. Below, each grid point is colored by its number — bright = big, dark = small. No directions anywhere. Just a value sitting on each dot.

---

Step 2 — Replace the number with a two-number output

WHAT. Instead of returning one number, our rule now returns two numbers at each point. Call them $P$ and $Q$:

$$\mathbf{F}(x,y)=(\underset{\text{across-amount}}{\underbrace{P(x,y)}},\underset{\text{up-amount}}{\underbrace{Q(x,y)}})$$

• $\mathbf{F}$ — the new rule (bold letter = it outputs an arrow, not a number).
• $P(x,y)$ — the first output number: how far the arrow reaches across (right if positive, left if negative).
• $Q(x,y)$ — the second output number: how far the arrow reaches up (up if positive, down if negative).

WHY two numbers? Because an arrow needs exactly two numbers to pin it down in the plane: one for across, one for up. That is the entire reason the output dimension matches the input dimension — the arrow has to live in the same flat plane as the point it is glued to.

PICTURE. One point $(2,1)$. Its two output numbers $P=1,Q=2$ become the sides of a little right triangle, and the arrow is the slanted side (the hypotenuse). Watch how $P$ lays out the horizontal leg and $Q$ stacks the vertical leg.

---

Step 3 — Glue the arrow onto its point (the key move)

WHAT. We take the arrow $(P,Q)$ and root its tail at the point $(x,y)$. The tail sits on the point; the tip lands at $(x+P,y+Q)$.

$$\text{tail at }(x,y)⟶\text{tip at }(\underset{\text{go right by }P}{\underbrace{x+P}},\underset{\text{go up by }Q}{\underbrace{y+Q}})$$

WHY this exact placement. This is what makes a field a field and not a loose bag of arrows. Every arrow is anchored — it means "if you stand here, you get pushed that way." Move the tail somewhere random and the meaning is lost.

PICTURE. The same arrow drawn twice: once "floating" at the origin (grey, meaningless), once rooted at its home point $(2,1)$ (amber, meaningful). Only the rooted one is part of the field.

---

Step 4 — Measure each arrow: length and direction

WHAT. Every rooted arrow carries two readable facts.

Its length (how hard the push) comes from the right triangle of Step 2 via Pythagoras:

$$|\mathbf{F}|=\sqrt{\underset{\text{across leg squared}}{\underbrace{P^2}}+\underset{\text{up leg squared}}{\underbrace{Q^2}}}$$

• $|\mathbf{F}|$ — the length of the arrow (the "magnitude"), always $\ge 0$.
• Squaring $P$ and $Q$, adding, then square-rooting is exactly Pythagoras on the Step-2 triangle: the two legs give the slanted side.

Its direction (which way) is the arrow shrunk to length $1$:

$$\hat{\mathbf{F}}=\frac{\mathbf{F}}{|\mathbf{F}|}=\left(\frac{P}{|\mathbf{F}|},\frac{Q}{|\mathbf{F}|}\right)$$

• Dividing by the length scales the arrow down to length $1$ while keeping its heading — a pure "which way" with the "how hard" stripped out.

WHY these two tools. Pythagoras is the only honest way to turn two perpendicular legs into a straight-line length — that is the exact question "how long is the slanted side?" And dividing by length answers a different question, "just the heading please," which we need when we want to compare directions without long arrows cluttering the drawing.

PICTURE. One arrow with its horizontal leg $P$, vertical leg $Q$, and the hypotenuse labelled $\sqrt{P^2+Q^2}$; alongside, the same heading shrunk to a unit arrow of length $1$.

---

Step 5 — Sweep a whole grid: the field appears

WHAT. Now do Steps 2–4 at many points on a grid. At each grid point $(x,y)$: compute $(P,Q)$, root the arrow there, note its length. The forest of arrows is the vector field.

WHY a grid. We can't draw infinitely many arrows (there's one at every point), so we sample a tidy grid — enough arrows to see the pattern, few enough to read.

PICTURE. We use the simplest interesting rule, the radial field $\mathbf{F}(x,y)=(x,y)$ — the output equals the position. Watch: near the origin arrows are tiny (small $P,Q$), far out they grow (Pythagoras: bigger legs, longer hypotenuse), and every arrow points straight away from the centre.

---

Step 6 — A different rule, a whole new personality: rotation

WHAT. Keep the machinery identical; change only the rule to $\mathbf{F}(x,y)=(-y,x)$. So now $P=-y$ (across-amount) and $Q=x$ (up-amount).

WHY show a second rule. To prove the field comes from the rule, not from the drawing procedure. Same grid, same gluing, same Pythagoras — but a swirl appears instead of an explosion.

Term-by-term at a test point $(1,0)$:

$$\mathbf{F}(1,0)=(\underset{-0=0}{\underbrace{-y}},\underset{1}{\underbrace{x}})=(0,1)\Rightarrow \text{points straight up.}$$

Why is every arrow tangent to the circle through its point? Check with the dot product — a number that is $0$ exactly when two arrows are at right angles:

$$\mathbf{F}\cdot (x,y)=\underset{\text{across}\times \text{across}}{\underbrace{(-y)(x)}}+\underset{\text{up}\times \text{up}}{\underbrace{(x)(y)}}=-xy+xy=0.$$

• The dot product multiplies matching parts and adds them.
• Getting $0$ means the arrow $\mathbf{F}$ is perpendicular to the pointer-from-origin $(x,y)$.
• Perpendicular to the radius $\Rightarrow$ pointing along the circle $\Rightarrow$ pure spin.

WHY the dot product and not something else? We asked a yes/no question — "is the arrow at right angles to the radius?" The dot product is precisely the tool built to answer "are these two arrows perpendicular?" ($=0$ yes). No other single number does that so cleanly.

PICTURE. The rotation field on the same grid: arrows circle counter-clockwise, each sitting tangent to its dotted circle, with the right-angle to the radius marked at one sample point.

---

Step 7 — Where fields come from: the gradient (uphill arrows)

WHAT. Take an ordinary number-function $f(x,y)=x^2+y^2$ (a bowl, lowest at the origin). Its gradient $\nabla f$ collects the two slopes and packages them as an arrow:

$$\nabla f=\left(\underset{\text{slope going right}}{\underbrace{\frac{\partial f}{\partial x}}},\underset{\text{slope going up}}{\underbrace{\frac{\partial f}{\partial y}}}\right)=(2x,2y)$$

• $\frac{\partial f}{\partial x}$ (read "partial $f$ by $x$") — how fast $f$ climbs if you step in the $x$-direction only. For $x^2+y^2$ that slope is $2x$.
• $\frac{\partial f}{\partial y}$ — how fast $f$ climbs stepping in $y$ only; here $2y$.
• Bundling the two slopes into $(2x,2y)$ gives an arrow at every point.

WHY the gradient turns a number-map into an arrow-map. Each partial derivative asks "which way is uphill along one axis, and how steep?" Put the two together and you get a single arrow that points in the direction of steepest climb. That arrow-per-point is a vector field — born from a plain number-function.

Compare with Step 5: $(2x,2y)$ is just the radial field $(x,y)$ scaled by $2$ — so gradient arrows here also point outward, away from the bottom of the bowl, uphill. A field that equals $\nabla f$ for some $f$ is called a conservative (gradient) field; see Gradient and Directional Derivatives and Conservative Fields and Potential Functions.

PICTURE. The bowl's height shown as faint rings (contours); on top, the gradient arrows $(2x,2y)$ all pointing outward across the rings, uphill, longer where the bowl is steeper.

---

The one-picture summary

One frame, three columns — the same drawing procedure feeding three different rules: radial (explode), rotation (swirl), gradient (climb). Read it left to right: point in → two numbers out → arrow glued on → sweep the grid → a personality emerges.

Recall Feynman retelling — the whole walk in plain words

Picture a giant sheet of graph paper. First I show you a color map: every dot has one number (like temperature) — that's an ordinary function. Now I upgrade the machine so at every dot it spits out two numbers instead of one. Those two numbers are the sides of a little right triangle — go across a bit, go up a bit — and the slanted side is an arrow. I glue that arrow's tail right onto the dot it came from: "stand here, get pushed this way." Pythagoras tells me how long each arrow is (the length of that slanted side), and if I only care which way it points I shrink it down to length one. If both numbers are zero, there's no push at all, so I draw a plain dot — no arrow. Then I repeat this at a whole grid of dots and the pattern pops out. Feed the machine the rule "arrow = your position" and everything explodes outward. Feed it "arrow = position spun a quarter-turn" and everything swirls in circles (I can prove it's a perfect swirl because the arrow is at a right angle to the line back to the centre — the dot-product test gives exactly zero). Feed it "arrow = the uphill direction of a bowl" and every arrow climbs away from the bottom. Same gluing recipe every time — only the rule changes, and the rule is the whole personality of the field.

Connections

• Gradient and Directional Derivatives — Step 7's arrows are gradients.
• Conservative Fields and Potential Functions — when a field is some $\nabla f$.
• Divergence and Curl — measure the "explode" vs "swirl" of Steps 5–6 numerically.
• Line Integrals — add up a field's push along a path.
• Flux and the Divergence Theorem — flow of a field through a boundary.
• Parametric Curves and Velocity — a moving point's velocity is arrows along its curve.