Visual walkthrough — Dot product — formula, geometric meaning, work calculation

Before we begin, three plain words we will keep reusing:

---

Step 1 — The honest question: "how much of $\vec{A}$ goes along $\vec{B}$?"

WHAT. We draw two arrows, $\vec{A}$ (blue) and $\vec{B}$ (orange), tail-to-tail. We are not asking their lengths and we are not adding them. We ask one thing: if I only care about the direction $\vec{B}$ points, how much of $\vec{A}$ is helping in that direction?

WHY. This is the physical question hiding inside work, power, and projection. When you push a box, only the part of your push that goes along the floor moves it. So "how much of one arrow lies along another" is the quantity worth naming. Everything below is just answering this cleanly.

PICTURE. Look at the figure. The blue arrow $\vec{A}$ leans away from the orange arrow $\vec{B}$ by the angle $\theta$. Imagine the sun shining straight down onto the line of $\vec{B}$: the blue arrow casts a shadow (green) that lies flat along $\vec{B}$'s direction. That shadow's length is our answer — "how much of $\vec{A}$ goes along $\vec{B}$."

---

Step 2 — The shadow has length $|\vec{A}|\cos \theta$ (a right triangle appears)

WHAT. We find the exact length of that green shadow. Drop a straight line from the tip of $\vec{A}$ perpendicular (at a clean $90^{\circ }$) down onto the line of $\vec{B}$. This makes a right triangle.

WHY. A right triangle is the one shape where the sides and the angle are locked together by fixed ratios. Dropping a perpendicular is the standard move to turn "a lean of $\theta$" into a computable length — it is why trigonometry exists (see Trigonometry — cosine and components).

PICTURE. In the figure:

• The hypotenuse (longest side, blue) is $\vec{A}$ itself — its length is $|\vec{A}|$.
• The adjacent side (green, lying along $\vec{B}$) is the shadow we want.
• The opposite side (gray dashed, the perpendicular drop) closes the triangle.
• The angle at the tail, between hypotenuse and adjacent side, is exactly $\theta$.

Now the definition of cosine — chosen on purpose because it is the ratio linking the angle to the adjacent side:

Rearranging for the side we want:

$$\underset{\text{adjacent, green}}{\underbrace{\text{shadow}}}=\underset{\text{hypotenuse (length of }\vec{A})}{\underbrace{|\vec{A}|}}\underset{\text{shrink factor from the lean}}{\underbrace{\cos \theta }}$$

• $|\vec{A}|$ is the full length of the arrow.
• $\cos \theta$ is a dial between $1$ and $-1$ that shrinks that length down to the shadow.

WHY cosine and not sine or tangent? Sine would give the perpendicular (gray) side — the part wasted, not the part aligned. Tangent is a ratio of the two legs and ignores the hypotenuse length entirely. We want aligned length, and that is adjacent-over-hypotenuse — cosine.

---

Step 3 — Multiply by $|\vec{B}|$: the dot product is born

WHAT. We now combine the shadow of $\vec{A}$ with the length of $\vec{B}$ by multiplying them. Call this product the dot product, written $\vec{A}\cdot \vec{B}$ (a raised dot between the two arrows).

WHY. In physics the aligned amount rarely matters alone — it matters scaled by how much of $\vec{B}$ there is. Work = (aligned force) × (distance travelled). So the natural quantity is shadow-of-A times length-of-B. Multiplying the two lengths gives one plain number — a scalar (no direction).

PICTURE. The green shadow ($|\vec{A}|\cos \theta$) laid flat along $\vec{B}$, then stretched by the full length of $\vec{B}$ (orange). The result is an area-like number, drawn as the shaded rectangle.

$$\vec{A}\cdot \vec{B}=\underset{\text{shadow of }\vec{A}\text{ (Step 2)}}{\underbrace{(|\vec{A}|\cos \theta )}}\times \underset{\text{length of }\vec{B}}{\underbrace{|\vec{B}|}}=\boxed{|\vec{A}||\vec{B}|\cos \theta }$$

• $|\vec{A}|\cos \theta$ — the part of $\vec{A}$ that lies along $\vec{B}$.
• $|\vec{B}|$ — how far that alignment stretches.
• Together: one number, the geometric form of the dot product.

---

Step 4 — The sign of the answer: walk $\theta$ from $0^{\circ }$ to $180^{\circ }$

WHAT. The one moving part in $|\vec{A}||\vec{B}|\cos \theta$ is $\cos \theta$. The lengths are always positive, so the sign of the whole dot product is the sign of $\cos \theta$. We watch it across every case.

WHY. A reader must never meet a scenario we skipped. There are exactly three regions of the lean plus two knife-edges — five cases — and each one has a physical meaning we must show.

PICTURE. The figure sweeps $\vec{A}$ around $\vec{B}$ and plots how the shadow (and its sign) changes.

Case	$\theta$	shadow direction	$\cos \theta$	$\vec{A}\cdot \vec{B}$	Meaning
same way	$0^{\circ }$	full, along $\vec{B}$	$+1$	$+	\vec A
leaning with	$0^{\circ }<\theta <90^{\circ }$	forward	positive	positive	partly helping
dead across	$90^{\circ }$	shrinks to a dot	$0$	$0$	perpendicular
leaning against	$90^{\circ }<\theta <180^{\circ }$	backward	negative	negative	partly opposing
opposite	$180^{\circ }$	full, backward	$-1$	$-	\vec A

Recall The perpendicular knife-edge

At exactly $90^{\circ }$ the shadow is a single point — zero length. So $\vec{A}\cdot \vec{B}=0$ (for non-zero arrows) means, always and only, perpendicular. This is the test used everywhere: two vectors are at right angles $⟺$ their dot product is zero.

---

Step 5 — Degenerate case: what if an arrow has zero length?

WHAT. Suppose $\vec{A}=\vec{0}$ — an arrow of length $0$, just a dot with no direction. What is $\vec{0}\cdot \vec{B}$?

WHY. A formula you can trust must survive the empty case. And $\theta$ isn't even defined when one arrow has no direction, so we must decide the answer from the shadow idea, not from $\cos \theta$.

PICTURE. A dot casts no shadow. There is nothing of $\vec{A}$ to lie along $\vec{B}$, so the aligned amount is $0$, and $0\times |\vec{B}|=0$.

$$\vec{0}\cdot \vec{B}=0\text{for every }\vec{B}.$$

The component form agrees instantly: every component of $\vec{0}$ is $0$, so $0\cdot B_x+0\cdot B_y+0\cdot B_z=0$. Both definitions handle the empty case cleanly.

---

Step 6 — The special arrows $\hat{i},\hat{j}$: dotting the "rulers" together

WHAT. We introduce two special vectors: $\hat{i}$ points one unit to the right, $\hat{j}$ points one unit up. Each has length $1$ (a "unit vector") and they sit at $90^{\circ }$ to each other — the two rulers we measure everything against (see Vectors — components and unit vectors).

WHY. If we can dot these rulers with each other, we can dot any two vectors, because every vector is just so-many-rights plus so-many-ups. This is the bridge from the picture-formula to the component-formula.

PICTURE. The two perpendicular unit arrows, and their four possible dot products.

Feed each pair through the geometric form from Step 3:

$$\hat{i}\cdot \hat{i}=(1)(1)\cos 0^{\circ }=1,\hat{j}\cdot \hat{j}=(1)(1)\cos 0^{\circ }=1,$$ $$\hat{i}\cdot \hat{j}=(1)(1)\cos 90^{\circ }=0,\hat{j}\cdot \hat{i}=0.$$

• Matched ruler with itself → angle $0^{\circ }$ → $\cos 0^{\circ }=1$.
• Mismatched rulers → angle $90^{\circ }$ → $\cos 90^{\circ }=0$.

---

Step 7 — Expand in components: the two formulas become one

WHAT. Write each vector in the rulers: $\vec{A}=A_x\hat{i}+A_y\hat{j}$ and $\vec{B}=B_x\hat{i}+B_y\hat{j}$. Here $A_x$ is "how many rights," $A_y$ is "how many ups." Now dot them by multiplying every term of the first by every term of the second (just like expanding two brackets — the dot product distributes over addition because shadows add).

WHY. We want a formula that needs no protractor — only the numbers $A_x,A_y,B_x,B_y$ you can read straight off a grid. Step 6 gives us exactly the tools: each little product carries a ruler-dot-ruler that is either $1$ or $0$.

PICTURE. The four cross-multiplied terms, colour-coded: two survive (green, the $1$s), two vanish (gray, the $0$s).

$$\vec{A}\cdot \vec{B}=A_x{B}_x\underset{=1}{\underbrace{(\hat{i}\cdot \hat{i})}}+A_x{B}_y\underset{=0}{\underbrace{(\hat{i}\cdot \hat{j})}}+A_y{B}_x\underset{=0}{\underbrace{(\hat{j}\cdot \hat{i})}}+A_y{B}_y\underset{=1}{\underbrace{(\hat{j}\cdot \hat{j})}}$$

The two $\times 0$ terms disappear; the two $\times 1$ terms stay:

$$\boxed{\vec{A}\cdot \vec{B}=A_x{B}_x+A_y{B}_y}(\text{add }{A}_z{B}_z\text{ in 3D})$$

• $A_x{B}_x$ — how the two "rightward parts" line up.
• $A_y{B}_y$ — how the two "upward parts" line up.
• Their sum is the same number Step 3 got geometrically — no angle in sight.

---

The one-picture summary

Everything above compressed into a single frame: the arrows, the shadow, the right triangle giving $|\vec{A}|\cos \theta$, the stretch by $|\vec{B}|$, and the same answer read off the grid as $A_x{B}_x+A_y{B}_y$.

Recall Feynman retelling — the whole walkthrough in plain words

I drew two arrows from the same dot. I asked only one thing: how much of the first arrow goes the same way as the second? To answer, I shone a light straight down onto the second arrow's line and measured the shadow of the first — a right triangle popped up, and its base came out to (length of first) times $\cos$ of the lean. Then I stretched that shadow by the full length of the second arrow, and that product I named the dot product — one plain number. I watched the number as I swung the first arrow around: full and positive when they point the same way, shrinking to zero exactly when they're at right angles, then going negative when the first leans backward. A zero-length arrow casts no shadow, so its dot product is zero — the formula never breaks. Finally I noticed every arrow is just so-many-rights plus so-many-ups; dotting the rights-and-ups rulers gives $1$ for like-with-like and $0$ for cross-terms, so all the mess collapses to $A_x{B}_x+A_y{B}_y$ — the same number, now from the grid alone. Two roads, one destination: how much one arrow goes along another.

---

Connections

• Dot product — formula, geometric meaning, work calculation (parent)
• Trigonometry — cosine and components (the right-triangle shadow)
• Vectors — components and unit vectors (the rulers $\hat{i},\hat{j}$)
• Projection of a vector (the shadow is the projection)
• Work-Energy Theorem (aligned force × distance)
• Circular motion (velocity ⟂ centripetal force ⇒ zero dot)
• Cross product — area and torque (the perpendicular partner operation)