Vector Analysis: the Toolkit

1. The Vector Toolkit

Title

Griffiths, Chapter 1

Gradient, divergence, curl, the three big theorems — everything Chapters 2-4 will lean on

2. What this deck buys you

Objectives

Chapter 1 is not really a chapter about maths. It is Griffiths handing you the five tools he is about to use on every page of electrostatics.

• Read the separation vector notation Griffiths uses in every field formula
• Compute a gradient, a divergence, and a curl — and say in words what each one measures
• Choose the right integral theorem (gradient / divergence / Stokes) for a given problem
• Set up an integral in spherical or cylindrical coordinates from the symmetry of the object
• Use the delta function to write a point charge as a density

Everything here comes back in Chapter 2 wearing a physics costume: divergence becomes Gauss's law, curl becomes the electric field is conservative, gradient becomes E is minus the slope of V.

You almost certainly opened this book for electricity and magnetism, and the first fifty pages are gradients, surface integrals, and theorems with people's names attached. It is a fair moment to wonder whether this part really counts.

It does, and here is why. Griffiths is not teaching vector calculus for its own sake — he is handing you the five instruments he is about to use on every page of electrostatics, and he would rather do it now than stop halfway through a derivation to explain what a curl is.

So here is the ending, in advance. Divergence is about to become Gauss's law. Curl becomes the reason a potential exists at all. Gradient is how you get the field back out of that potential. The three big theorems let you switch between the local and the global version of every law. And the delta function is what lets a point charge sit inside an integral without breaking it.

If you are impatient to reach the physics, hold on to this: every tool in this deck reappears within about two weeks, and the one you skip is the one that will stop you.

3. Vectors and the Two Products

Section

Section 1

4. A vector is a magnitude plus a direction

Concept

A vector has a magnitude (how long) and a direction (which way). Where you draw it on the page is not part of it.

A scalar — temperature, charge, mass — has only a magnitude. Getting this distinction wrong is the single most common source of sign errors in Chapter 2.

The definition is the easy half. What deserves your attention is what it leaves out — a vector is a length and a direction, and nothing else at all. In particular it carries no position. The two arrows above are drawn in different places on the page, and they are the same vector.

That sounds like pedantry right up until the separation vector turns up a few slides from now, which is exactly where position starts getting smuggled back in without anyone noticing.

Now set it against a scalar — temperature, charge, mass, energy — which has a size and nothing more. The distinction feels too obvious to be worth dwelling on, and it is behind most of the sign errors in Chapter 2: a direction gets attached to something that has not got one, or dropped from something that has.

Quick test of whether it has landed. Does it mean anything to say the temperature in this room points north? If that sounds absurd to you, you have the distinction.

5. Why electrostatics cannot be done with numbers alone

Intuition

At every point in space around a charge, something is pushing — and the push has a direction that changes from point to point.

That object — a vector attached to every point of space — is a vector field. Chapters 2 through 4 are the study of two of them.

This is the reason all the apparatus that follows exists, so it is worth a moment.

Around a charge, something is pushing at every point in space — and the push has both a strength and a direction, and both change from point to point. A single number at each point could record the strength, but it could never say which way.

That object — a whole vector attached to every point of space — is a vector field, and it is the central character of the next three chapters. Both the electric and the magnetic field are of this kind.

In the animation, the arrows are longer near the charge and shorter far away, which carries the strength, and they point radially outward, which carries the direction. Two pieces of information at every one of infinitely many points.

6. Decoding Griffiths' separation vector

Notation

This one symbol appears in nearly every formula in Chapters 2-4. If you never decode it, none of them make sense.

\[ \mathfrak{r} = |\mathbf{r} - \mathbf{r}'|, \qquad \hat{\boldsymbol{\mathfrak{r}}} = \frac{\boldsymbol{\mathfrak{r}}}{\mathfrak{r}} \]

Read it out loud every time: from the charge, to the point I care about. The direction of that arrow is the direction a positive test charge would be pushed.

This is the most important slide in the deck for surviving Chapter 2, so it is worth more time than its size suggests.

Here is the problem it solves. In any field calculation there are two different positions in play: where the charge is, and where you are asking about the field. Confusing the two is the single most common setup error in the whole course, and this notation exists to keep them apart.

Griffiths' convention is that primed coordinates are source points — where the charge sits — and unprimed ones are field points, where you want the answer. The separation vector is the difference between them, and it points from the source to the field point.

Say the direction out loud every time you write it: from the charge, to the place I care about. That is the direction a positive test charge would be pushed, which is why this unit vector sits on the end of every field formula in the book.

One presentational note. Griffiths draws this as a script r; these slides render it as a bold fraktur letter, because that is what the typesetting supports. Same object. Worth knowing now, so it does not throw you when you open the book.

7. Adding vectors: head to tail

Concept

Multiplying by a scalar stretches the arrow (a positive) or flips and stretches it (a negative). This is all of vector algebra you need for Chapter 2 superposition.

\[ a(\mathbf{A} + \mathbf{B}) = a\mathbf{A} + a\mathbf{B} \]

A quick slide, with two things worth saying.

Head-to-tail addition is the geometric statement and adding components is the algebraic one. They are the same operation, and it pays to be fluent in both directions: the picture tells you whether an answer is plausible, and the algebra is what actually gets it.

Scalar multiplication stretches a vector, and multiplying by a negative number flips it as well as stretching it. That is the entire vector algebra that superposition needs in Chapter 2 — which is worth saying, because the machinery really is that small.

8. The dot product: how much of A lies along B

Concept

\[ \mathbf{A} \cdot \mathbf{B} = AB\cos\theta \]

It returns a scalar — a plain number, no direction. Two consequences you will use constantly:

• Perpendicular vectors have zero dot product
• A · A is the length squared, so the length of a vector is the square root of its dot product with itself

\[ \mathbf{A} \cdot \mathbf{B} = A_xB_x + A_yB_y + A_zB_z \]

Start with the geometry rather than the formula. The formula is easy to remember and not much use without the picture.

The dot product answers one question: how much of this vector lies along that one? The dashed line drops A onto B, and the length of that shadow, times the length of B, is the dot product.

Two consequences worth drawing out. Perpendicular vectors have zero dot product, because the shadow has no length — that is what makes flux calculations collapse whenever the field skims along a surface. And the dot product of a vector with itself is its length squared, which is how you get magnitudes out of components.

The result is a scalar. No direction at all. If you ever catch yourself asking which way a dot product points, that is the signal something has slipped.

9. Worked example: the angle between two face diagonals of a cube

Worked example

Griffiths' Example 1.2. Watch how a geometry question becomes arithmetic the moment you write coordinates.

\[ \mathbf{A} = 1\,\hat{\mathbf{x}} + 0\,\hat{\mathbf{y}} + 1\,\hat{\mathbf{z}}, \qquad \mathbf{B} = 0\,\hat{\mathbf{x}} + 1\,\hat{\mathbf{y}} + 1\,\hat{\mathbf{z}} \]

\[ \mathbf{A}\cdot\mathbf{B} = (1)(0) + (0)(1) + (1)(1) = 1 \]

\[ A = B = \sqrt{1^2 + 1^2} = \sqrt{2} \quad \Rightarrow \quad \mathbf{A}\cdot\mathbf{B} = 2\cos\theta \]

\[ \cos\theta = \tfrac{1}{2} \quad \Rightarrow \quad \theta = 60^\circ \]

The value of this example is the strategy, not the answer.

The key move happens at the very start: put the cube somewhere convenient and give it side one. The angle between two diagonals cannot depend on where the cube sits or how big it is, so you are free to pick whatever makes the arithmetic trivial. Carrying an unknown side length through the entire calculation and watching it cancel at the end is wasted effort.

After that, the geometry has been converted into arithmetic. Write both vectors in components, compute the dot product two ways — once by components, once by the cosine formula — and set the two equal.

The verification is worth doing because it is completely independent of the algebra. The two diagonals, plus the cube edge joining their far ends, form a triangle whose three sides all have length root two. An equilateral triangle has sixty-degree angles, which confirms the answer geometrically.

10. The cross product: the piece of A that is perpendicular to B

Concept

\[ \mathbf{A} \times \mathbf{B} = AB\sin\theta\, \hat{\mathbf{n}} \]

n-hat is the unit vector perpendicular to the plane of A and B, in the direction your right thumb points when your fingers sweep from A to B.

\[ \mathbf{A} \times \mathbf{B} = \begin{vmatrix} \hat{\mathbf{x}} & \hat{\mathbf{y}} & \hat{\mathbf{z}} \\ A_x & A_y & A_z \\ B_x & B_y & B_z \end{vmatrix} \]

Its magnitude is the area of the parallelogram A and B span. Parallel vectors span no area, so their cross product is zero.

The cross product is the one that feels genuinely alien at first, so anchor it in the picture before the formula.

It produces a vector perpendicular to both inputs — out of the plane they span. That is a strong statement: whatever the answer is, it cannot lie in the plane of the two vectors you started with, which on its own rules out most wrong answers.

Its magnitude is the area of the parallelogram the two vectors span, which explains why parallel vectors give zero: they span no area at all.

The direction comes from the right-hand rule, and it has to be the right hand every time. Do it physically — fingers along A, curl them toward B, thumb gives the answer — rather than trying to picture it in your head. The determinant is the computational route once the geometry is clear.

11. Commit first: which way does it point?

Prediction

Any answer that lies in the page had to be wrong: the cross product is always perpendicular to both inputs, and both inputs live in the page.

Commit before revealing, because the elimination reasoning is worth more here than the answer.

Both input vectors lie in the plane of the page, and the cross product is perpendicular to both. So the answer cannot lie in the page at all, which eliminates two of the four options before you do any thinking about handedness.

Then the right-hand rule settles the rest: fingers along A pointing right, curling up toward B, and your thumb comes out of the page.

Worth noting the reverse case in the same breath. B cross A points into the page — same magnitude, opposite direction — and in Chapter 5 that sign is the difference between two current-carrying wires attracting and repelling.

12. Trap: treating the cross product like ordinary multiplication

Trap

13. Dot versus cross, side by side

Comparison

Nearly every 'I do not know how to start' moment in this course is really 'I do not know which product this is'.

In Chapter 2 you will use the dot product for flux (how much field crosses a surface) and, in Chapter 5, the cross product for magnetic force.

Use this grid as a diagnostic. Nearly every 'I do not know how to start' moment in this course is really 'I do not know which product this is'.

The two rows worth making automatic are the first and the third. The dot product returns a scalar and the cross returns a vector. And the dot is largest when the vectors are parallel, while the cross is largest when they are perpendicular — which is exactly backwards from what most people assume at first.

Look ahead to where each one gets used. Dot products turn up in flux, work and potential differences, all of which are numbers. Cross products turn up in torque, magnetic force and area vectors, all of which have directions.

14. Pattern: picking the right product

Pattern

Read the physical question, not the symbols.

1. Am I asking how much of one thing lies along another? Work, flux, projection — use the dot product
2. Am I asking for something perpendicular to both? Torque, magnetic force, area vector — use the cross product
3. Does the answer need a direction? If yes, it cannot be a dot product
4. Is my answer a single number? If yes, it cannot be a cross product

Then check the trivial cases before you trust the algebra: parallel inputs kill a cross product, perpendicular inputs kill a dot product.

This is the decision procedure, and it works by reading the physical question rather than the symbols.

If the question is 'how much of one thing lies along another' — work, flux, a projection — it is a dot product. If it is 'what is perpendicular to both of these' — torque, magnetic force, an area vector — it is a cross product.

The type check is the fastest filter and worth drilling until it is automatic. Does the answer need a direction? Then it cannot be a dot product. Is the answer a single number? Then it cannot be a cross product.

Then check the trivial cases before trusting the algebra. Parallel inputs kill a cross product; perpendicular inputs kill a dot product. Testing a setup against those two catches a surprising share of errors.

15. Check yourself: products

Check

Try it on paper before you click.

Work is the standard first application of the dot product, and this question is really testing whether the type check has landed.

The arithmetic is straightforward. The reasoning worth reinforcing is that work is a scalar — energy has no direction — which eliminates option D immediately, whatever number comes in front of it.

Option C swaps sine for cosine, which is the cross-product factor. Force times distance times sine is a torque rather than work, and naming that is what stops the two formulas blurring together.

16. The Three Derivatives

Section

Section 2

17. One operator, three ways to use it

Intuition

Ordinary calculus has one derivative. Vector calculus has three — but they are all the same operator applied differently.

\[ \nabla = \hat{\mathbf{x}}\frac{\partial}{\partial x} + \hat{\mathbf{y}}\frac{\partial}{\partial y} + \hat{\mathbf{z}}\frac{\partial}{\partial z} \]

Del is not a vector — it is an instruction waiting for something to act on. But it obeys enough vector rules that treating it like one, carefully, works.

This is what stops the three derivatives feeling like three unrelated things to memorise.

Del is a single object, and it is not really a vector — it is an instruction waiting for something to act on. But it obeys enough of the algebra of vectors that treating it carefully as one works out.

The three branches are just the three things you can do with a vector-like object. Apply it to a scalar and you get a vector: the gradient. Dot it into a vector and you get a scalar: the divergence. Cross it into a vector and you get a vector: the curl.

The type table at the bottom is worth memorising on its own, because it catches errors instantly. Producing a scalar from a curl, or taking the gradient of a vector, gets flagged by the type check before any physics is involved.

18. The gradient: which way is uphill, and how steep

Concept

\[ \nabla T = \frac{\partial T}{\partial x}\hat{\mathbf{x}} + \frac{\partial T}{\partial y}\hat{\mathbf{y}} + \frac{\partial T}{\partial z}\hat{\mathbf{z}} \]

The gradient of a scalar function is a vector: it points in the direction of steepest increase, and its length is the rate of increase in that direction.

\[ dT = \nabla T \cdot d\mathbf{l} \]

That one line is the whole meaning: the change you experience depends on the direction you step. Step along a contour and the dot product vanishes — no change.

The contour-map picture is the one to make permanent, because you almost certainly have the intuition already from hiking maps and it just needs connecting.

The gradient of a scalar function is a vector. It points in the direction of steepest increase, and its length is the rate of increase in that direction. On the map, the arrows point toward the summit and are longest where the contours bunch up.

The most useful line here is the one relating the change in T to a dot product with your step. In words: how much the function changes depends on which direction you step. Step along a contour, perpendicular to the gradient, and the dot product vanishes — no change at all.

That geometric fact, that a gradient is perpendicular to the level surfaces of its function, is what will make electric field lines perpendicular to equipotentials in two decks' time.

19. Where the gradient is zero

Intuition

On a hill, the ground is level at a summit, at the bottom of a bowl, and at a saddle. In all three, every direction you step is flat to first order.

This matters in Chapter 3: Laplace's equation forbids maxima and minima inside a charge-free region, so every flat point of the potential there is a saddle.

Three different shapes give a vanishing gradient, and telling them apart matters later.

At a maximum, a minimum, or a saddle, every direction you can step is flat to first order. The gradient tells you nothing about which of the three you are standing on — that takes second derivatives.

The reason it comes up now is Chapter 3. Laplace's equation forbids maxima and minima of the potential inside a charge-free region, which means every flat point of V there has to be a saddle. That has the striking consequence that a charge cannot be trapped in a stable electrostatic equilibrium.

It is a memorable result, and knowing it is coming gives this discussion a purpose beyond the definition.

20. Worked example: the gradient of the distance from the origin

Worked example

Griffiths' Example 1.3 — and the single most reused gradient in the whole book.

\[ r = \sqrt{x^2 + y^2 + z^2} \]

\[ \frac{\partial r}{\partial x} = \frac{1}{2}\frac{2x}{\sqrt{x^2+y^2+z^2}} = \frac{x}{r} \]

\[ \nabla r = \frac{x\hat{\mathbf{x}} + y\hat{\mathbf{y}} + z\hat{\mathbf{z}}}{r} = \frac{\mathbf{r}}{r} = \hat{\mathbf{r}} \]

Keep this result. In Chapter 2, the gradient of one over the separation distance is what turns the potential back into the field.

This is the most reused gradient in the whole book, so it is worth deriving rather than quoting.

The calculation is a chain rule on a square root, done once and then repeated by symmetry for the other two components. The result assembles into the position vector divided by its own length, which is the radial unit vector.

The verification against the picture is where the understanding lives. If you walk one metre directly away from the origin, how much does your distance from the origin increase? By one metre. So the magnitude of this gradient has to be exactly one — and it is, since the answer is a unit vector. And the direction of fastest increase is obviously straight away from the origin.

Keep this result. In Chapter 2, taking the gradient of one over the separation distance is the step that turns a potential back into a field.

21. The divergence: how much is spreading out

Concept

\[ \nabla \cdot \mathbf{v} = \frac{\partial v_x}{\partial x} + \frac{\partial v_y}{\partial y} + \frac{\partial v_z}{\partial z} \]

divergence — The outflow per unit volume from an infinitesimal box around a point. Positive means the point is a source; negative means it is a sink; zero means whatever flows in flows straight back out.

Vector in, scalar out. Divergence is a number at each point, not a direction.

The animation is doing the teaching here, because the three cases side by side pre-empt the standard misconception.

Divergence measures spreading, not motion. The first panel spreads and has positive divergence. The second is a steady flow and the third is a rotation — both are obviously moving, and both have zero divergence.

It is very easy to equate 'divergence' with 'the field is doing something'. Those two counterexamples are the cure: two of the three panels are visibly in motion, and both have divergence zero.

The formula is the local version — differentiate each component along its own axis and add. Note the type: vector in, scalar out. Divergence is a number at each point, and asking which way it points is a category error.

22. The tiny-box picture

Intuition

Shrink a box around a point. Count what leaves through the six faces, subtract what enters, and divide by the volume. That limit is the divergence.

Hold that image. In Chapter 2 the field is E, the box is a Gaussian surface, and the source inside is electric charge. Gauss's law is this picture with units attached.

This is the definition worth actually carrying, because it is the one that makes Gauss's law obvious later.

Shrink a box around a point. Count what flows out through the six faces, subtract what flows in, divide by the volume. That limit is the divergence.

Positive divergence means more leaves than enters, so something inside is creating the stuff — a source. Negative means a sink. Zero means whatever goes in comes back out.

Hold on to this image, because in two decks' time the field is E, the box is a Gaussian surface, and the source inside is electric charge. Gauss's law is this picture with units attached, and having the image already in place is what makes that law obvious rather than mysterious.

23. Worked example: divergences of three sample fields

Worked example

Griffiths' Example 1.4. Three fields chosen so that the answers land on the three cases you must be able to tell apart.

\[ \mathbf{v}_a = x\hat{\mathbf{x}} + y\hat{\mathbf{y}} + z\hat{\mathbf{z}} \quad\Rightarrow\quad \nabla\cdot\mathbf{v}_a = 1+1+1 = 3 \]

\[ \mathbf{v}_b = \hat{\mathbf{z}} \quad\Rightarrow\quad \nabla\cdot\mathbf{v}_b = 0 \]

\[ \mathbf{v}_c = z\hat{\mathbf{z}} \quad\Rightarrow\quad \nabla\cdot\mathbf{v}_c = \frac{\partial z}{\partial z} = 1 \]

Three fields, chosen so the answers land on the three cases that have to be told apart.

The radial field spreads visibly and gives three. The constant field gives zero, because nothing changes from point to point and no box anywhere gains or loses anything.

The third is the one worth slowing down for. Every arrow points the same direction, so it looks like the uniform case — but the arrows get longer as you move along that direction. More leaves the top of a box than enters the bottom, so the divergence is positive even though the field never turns.

That case is the standard exam trap: a field can have divergence without visibly fanning out. What matters is whether each component changes along its own axis.

24. The curl: how much is circulating

Concept

\[ \nabla \times \mathbf{v} = \begin{vmatrix} \hat{\mathbf{x}} & \hat{\mathbf{y}} & \hat{\mathbf{z}} \\ \frac{\partial}{\partial x} & \frac{\partial}{\partial y} & \frac{\partial}{\partial z} \\ v_x & v_y & v_z \end{vmatrix} \]

Vector in, vector out. The curl points along the axis the paddle wheel spins about, by the right-hand rule, and its length is how fast.

The headline of Chapter 2: the electrostatic field has zero curl everywhere. No paddle wheel anywhere in electrostatics ever turns.

The paddle-wheel test is the definition to carry, and the animation makes it literal — the wheel actually turns.

Drop an imaginary paddle wheel into the field at a point. If it spins, the curl is non-zero there. The curl vector points along the axis it spins about, by the right-hand rule, and its magnitude says how fast.

Type check: vector in, vector out. The determinant is the computational device; the paddle wheel is the meaning.

And here is the headline that makes this whole section matter. The electrostatic field has zero curl everywhere. No paddle wheel anywhere in electrostatics ever turns, and that single fact is what makes the potential exist.

25. Worked example: curls of two sample fields

Worked example

Griffiths' Example 1.5, with the determinant written out once so you can see where each term comes from.

\[ \mathbf{v}_a = -y\,\hat{\mathbf{x}} + x\,\hat{\mathbf{y}} \]

\[ \nabla\times\mathbf{v}_a = \left(\frac{\partial v_y}{\partial x} - \frac{\partial v_x}{\partial y}\right)\hat{\mathbf{z}} = (1 - (-1))\hat{\mathbf{z}} = 2\hat{\mathbf{z}} \]

\[ \mathbf{v}_b = x\,\hat{\mathbf{y}} \quad\Rightarrow\quad \nabla\times\mathbf{v}_b = \frac{\partial x}{\partial x}\hat{\mathbf{z}} = \hat{\mathbf{z}} \]

Two fields, and the second is the important one.

The first circulates around the origin, so the paddle wheel is obviously turned and the algebra simply confirms the picture.

The second does not circulate at all. Every arrow points the same way and they just get longer as you move to the right. Yet the curl is non-zero, because a wheel placed there has faster flow on one side than the other and gets turned.

So shear produces curl just as surely as rotation does. Looking only for visible swirling will miss half the cases, which is why the verification step walks both pictures explicitly.

26. Sort the fields by what the operators say

Sorting

Before the algebra, decide from the picture. Divergence asks does it spread?; curl asks does it swirl or shear?

Item 4 is the one that catches people. The arrows never turn — but the field still has curl, because it shears.

Judge from the picture before computing anything — that is the skill that transfers.

The two questions are simple and independent. Does it spread? That is divergence. Would a paddle wheel turn? That is curl.

Item four is the one that catches people, and it is the shear case from the previous slide. The arrows never change direction, so it feels curl-free, but one side of the wheel sits in faster flow than the other.

If one of these comes out wrong, do not just take the correction. Put an imaginary wheel on it and describe what each paddle feels — the reasoning is what generalises.

27. Trap: the two second derivatives that are always zero

Trap

28. The Laplacian: the second derivative that survives

Concept

Of the five possible second derivatives, two are identically zero and one is the one you will actually use.

\[ \nabla^2 T = \nabla \cdot (\nabla T) = \frac{\partial^2 T}{\partial x^2} + \frac{\partial^2 T}{\partial y^2} + \frac{\partial^2 T}{\partial z^2} \]

Laplacian — The divergence of the gradient. It compares the value of a function at a point with its average on a tiny sphere around that point: negative where the point is a local peak, positive where it sits in a dip, zero when it exactly equals the surrounding average.

Chapter 2 ends with Poisson's equation, which is nothing but the Laplacian of the potential set equal to minus the charge density over epsilon-zero.

Of the possible second derivatives, two are identically zero and this is the one that does real work.

Definition first: the divergence of the gradient. But the meaning is what matters, and it is a comparison. The Laplacian at a point compares the value there with the average over a tiny sphere around it.

Negative where the point is a local peak, positive where it sits in a dip, zero when it matches the surrounding average exactly. The curve on the slide shows all three at once.

Where this is going: Chapter 2 ends with Poisson's equation, which is the Laplacian of the potential set equal to minus the charge density over epsilon-zero. And where there is no charge it becomes Laplace's equation, which is the single most important equation in the book.

29. Pattern: computing any of the three

Pattern

1. Identify what you are handed. A scalar can only take a gradient. A vector can take a divergence or a curl — never a gradient
2. Check what the answer must be. Gradient gives a vector, divergence a scalar, curl a vector. If your answer has the wrong type, you have used the wrong operator
3. Differentiate each component with respect to its own variable for divergence; use the determinant for curl
4. Sanity-check against the picture: does the field spread? does a paddle wheel spin? Your algebra must agree with what you can see

This is the procedural checklist, and the type checks in it are what catch errors fastest.

Identify what you have been handed. A scalar can only take a gradient. A vector can take a divergence or a curl, but never a gradient.

Then check the type of your answer against the table. Wrong type means wrong operator, and that gets caught in seconds rather than after a page of algebra.

Finally, sanity-check against the picture: does the field spread, would a paddle wheel turn? The algebra and the picture have to agree, and when they do not, it is almost always the algebra that is wrong.

30. Check yourself: derivatives

Check

Work it out before clicking.

This is the shear case again, now without a picture to lean on.

The field points along x and varies along y. Divergence differentiates each component along its own axis, and the x-component does not depend on x — so the divergence is zero.

The curl picks up the y-derivative of the x-component, which is not zero. Physically, a paddle wheel here has faster flow on its upper side than its lower one, so it turns.

Option B is the interesting wrong answer, because it reverses exactly the thing this deck has been drilling. Worth a look even if you did not pick it.

31. Integrals and the Three Theorems

Section

Section 3

32. The line integral: adding up along a path

Concept

\[ \int_a^b \mathbf{v} \cdot d\mathbf{l} \]

At each step, the dot product keeps only the component of the field along the path. Field perpendicular to your step contributes nothing.

If the path closes on itself, the integral gets a circle on it and a name: circulation.

\[ \oint \mathbf{v} \cdot d\mathbf{l} \]

The picture carries the idea: chop the path into short steps, and at each step keep only the part of the field that lies along your step.

The dot product is what performs that 'keep only the part along' operation. Field perpendicular to your step contributes nothing at all.

The physical anchor is work, which you already know: force dotted into displacement, summed along a path. Every line integral in this course has the same structure.

When the path closes on itself the integral gets a circle on it and a name — circulation — and the whole zero-curl story in Chapter 2 is stated in those terms.

33. Worked example: a line integral along two different routes

Worked example

Griffiths' Example 1.6. Same start, same end, two paths — and the point is what happens when you compare them.

\[ \mathbf{v} = y^2\,\hat{\mathbf{x}} + 2x(y+1)\,\hat{\mathbf{y}}, \qquad a = (1,1,0) \;\to\; b = (2,2,0) \]

\[ \int_1^2 y^2\,dx \Big|_{y=1} = \int_1^2 1\,dx = 1 \]

\[ \int_1^2 2x(y+1)\,dy\Big|_{x=2} = 4\int_1^2 (y+1)\,dy = 4\left[\tfrac{y^2}{2}+y\right]_1^2 = 10 \]

\[ \int_{\text{path 1}} \mathbf{v}\cdot d\mathbf{l} = 1 + 10 = 11 \]

\[ \mathbf{v}\cdot d\mathbf{l} = y^2\,dx + 2x(y+1)\,dy = \big(x^2 + 2x(x+1)\big)dx = (3x^2 + 2x)\,dx \]

\[ \int_{\text{path 2}} \mathbf{v}\cdot d\mathbf{l} = \int_1^2 (3x^2+2x)\,dx = \big[x^3 + x^2\big]_1^2 = 10 \]

\[ \oint \mathbf{v}\cdot d\mathbf{l} = 11 - 10 = 1 \neq 0 \]

The point of this example is the comparison at the end rather than the arithmetic, so it is worth knowing that in advance.

Both routes start and end at the same two points. The corner route splits into two legs, each with one variable frozen, which collapses the vector integral into two ordinary ones. The diagonal route locks the variables together with y equal to x, reducing everything to a single integral in x.

The results differ: eleven and ten. So for this field, the line integral depends on the route you take. Equivalently, going out one way and back the other gives a non-zero circulation.

Which has a consequence worth stating plainly: a field like this cannot be written as the gradient of anything, so it has no potential. The whole of Chapter 2 turns on the electrostatic field not behaving this way, which is exactly why this example is here.

34. The surface integral: flux through a surface

Concept

\[ \Phi = \int_S \mathbf{v} \cdot d\mathbf{a} \]

da is a vector: its magnitude is the patch's area and its direction is perpendicular to the patch. Field that skims along the surface crosses nothing and contributes nothing.

For a closed surface the convention is fixed for you: da points outward. That convention is what makes Gauss's law come out with a plus sign.

Flux is what the divergence theorem and Gauss's law are both built on, so the geometry needs to be solid.

The area element is a vector: its magnitude is the patch's area and its direction is perpendicular to the patch. Dotting the field into it keeps only the component actually crossing the surface.

The tilted-surface picture is worth looking at directly. Field skimming along a surface crosses nothing and contributes nothing, which is precisely what the dot product gives you.

For closed surfaces there is a convention: the area vector points outward. That is not arbitrary bookkeeping — it is what makes Gauss's law come out with a plus sign rather than a minus.

35. Worked example: flux through five faces of a cube

Worked example

Griffiths' Example 1.7 — the exercise that teaches you to handle a surface integral one face at a time.

\[ \mathbf{v} = 2xz\,\hat{\mathbf{x}} + (x+2)\,\hat{\mathbf{y}} + y(z^2-3)\,\hat{\mathbf{z}} \]

\[ \int \mathbf{v}\cdot d\mathbf{a} = 16 + 0 + 12 - 12 + 4 = 20 \]

This teaches a method rather than a result: break a complicated surface integral into flat pieces.

On each flat face one coordinate is constant and the area vector points along a single axis, so only one component of the field survives. Five easy integrals instead of one impossible one.

The table is the working, and the column to watch is the area vector. On the x equals zero face the outward normal points along minus x, and on the y equals zero face along minus y. Those minus signs are why two of the contributions come out negative.

Getting one of those signs wrong is the classic way this problem fails, so write the outward normal explicitly on every face rather than assuming it.

36. The volume integral: adding up over a region

Concept

\[ \int_V T\, d\tau, \qquad d\tau = dx\,dy\,dz \]

The least glamorous of the three, and the one you will use most: total charge is the charge density integrated over a volume.

\[ Q = \int_V \rho\, d\tau \]

The least glamorous of the three, and the one you will use most often.

The formula that matters is the second one: total charge is the charge density integrated over a volume. That single relationship turns up in nearly every Gauss's law problem in Chapter 2.

The picture — a blob chopped into small boxes with one highlighted — is the mental model. Sum the quantity over every infinitesimal box in the region.

One detail worth noting: when the integrand is a vector, the Cartesian unit vectors are constants and can come outside the integral. The next section's trap depends on exactly that fact.

37. All three theorems are the same idea

Intuition

You already know the pattern from first-year calculus:

\[ \int_a^b \frac{df}{dx}\,dx = f(b) - f(a) \]

Read that in words: the integral of a derivative over a region equals the function evaluated on the boundary of that region. The boundary of the interval from a to b is its two endpoints.

The three vector theorems are that sentence, once for each kind of derivative.

This is the organising insight of the section, and it is worth more time than its length suggests.

Start from something already familiar: the fundamental theorem of calculus. Read it in words rather than symbols — the integral of a derivative over a region equals the function evaluated on the boundary of that region. For an interval, the boundary is its two endpoints.

The three panels show that pattern repeating in higher dimensions. The boundary of a line is two points. The boundary of a surface is a loop. The boundary of a volume is a closed surface. In every case the theorem trades an integral over a region for one over its boundary.

Once the three theorems are one sentence applied to three kinds of derivative, they stop being three separate things to memorise.

38. Theorem 1: the fundamental theorem for gradients

Concept

\[ \int_a^b (\nabla T) \cdot d\mathbf{l} = T(b) - T(a) \]

The integral of a gradient along a path depends only on the endpoints. Every route from a to b gives the same answer.

Two corollaries, both used constantly in Chapter 2:

• The integral of a gradient around a closed loop is zero — you end where you started, so the height change is zero
• A field that is a gradient is called conservative, and 'conservative' and 'zero curl' turn out to mean the same thing

The statement is that the integral of a gradient along a path depends only on the endpoints. Every route from a to b gives the same answer.

The picture of three wildly different routes up the same hill, all with identical net elevation gain, is the intuition. Nobody is surprised that different paths to the summit gain the same height.

Two corollaries get used constantly in Chapter 2. The integral of a gradient around a closed loop is zero, because you end where you started and the height change is zero. And a field that is a gradient is called conservative.

One vocabulary warning: 'conservative', 'path-independent', 'zero circulation' and 'zero curl' all turn out to be the same statement, and expect to see all four used interchangeably.

39. Worked example: checking the gradient theorem

Worked example

Griffiths' Example 1.9. Take a function, take its gradient, and confirm that the path integral really does forget the path.

\[ T = xy^2, \qquad a = (0,0,0) \;\to\; b = (2,1,0) \]

\[ \nabla T = y^2\,\hat{\mathbf{x}} + 2xy\,\hat{\mathbf{y}} \]

\[ \underbrace{\int_0^2 y^2\,dx\Big|_{y=0}}_{0} + \underbrace{\int_0^1 2(2)y\,dy}_{2} = 2 \]

\[ \underbrace{\int_0^1 2(0)y\,dy}_{0} + \underbrace{\int_0^2 (1)^2\,dx}_{2} = 2 \]

This is the counterpart to the two-route example earlier, and running them side by side is what makes the point.

Here the field is a gradient by construction, and both routes give two. The theorem's endpoint formula also gives two. Three independent computations agreeing.

Contrast that with the earlier example, where two routes gave eleven and ten. That field was not a gradient; this one is. The difference is exactly whether the curl vanishes.

Notice the labour-saving structure in each route: on each leg one variable is frozen, which makes one of the two integrals vanish outright. Choosing routes along the axes is a deliberate tactic rather than luck.

40. Theorem 2: the divergence theorem

Concept

\[ \int_V (\nabla \cdot \mathbf{v})\, d\tau = \oint_S \mathbf{v} \cdot d\mathbf{a} \]

Sum the outflow of every tiny box inside the region. Each internal face is shared by two boxes, exits one and enters the other, and cancels. What is left is the flux through the outer skin.

Griffiths calls it Gauss's theorem, and in Chapter 2 it is what turns Gauss's law from a statement about surfaces into a statement about points.

The statement converts a volume integral of the divergence into a flux through the enclosing surface, and the proof idea is visible in the picture.

Chop the region into tiny boxes and sum the outflow of each. Every internal face is shared by two neighbouring boxes: what leaves one enters the other, so the pair cancels. Only the faces on the outer skin have no partner, and those are exactly the enclosing surface.

That cancellation argument is the whole proof in cartoon form, and it makes the theorem feel inevitable rather than magical.

Griffiths calls it Gauss's theorem, and in Chapter 2 it is the tool that converts Gauss's law from a statement about surfaces into a statement about points.

41. Worked example: checking the divergence theorem

Worked example

Griffiths' Example 1.10, on the unit cube. Both sides computed independently; they must agree.

\[ \mathbf{v} = y^2\,\hat{\mathbf{x}} + (2xy + z^2)\,\hat{\mathbf{y}} + 2yz\,\hat{\mathbf{z}} \]

\[ \nabla\cdot\mathbf{v} = 0 + 2x + 2y = 2(x+y) \]

\[ \int_0^1\!\!\int_0^1\!\!\int_0^1 2(x+y)\,dx\,dy\,dz = 2\left(\tfrac{1}{2} + \tfrac{1}{2}\right) = 2 \]

The value here is that both sides get computed completely independently and have to agree.

The left side is quick: take the divergence, integrate over the unit cube. The right side is six face integrals, using the same method as the earlier flux example.

Watch the signs on the faces at zero, which contribute negatively because their outward normals point along the negative axes. Two of the six contributions are negative for exactly that reason.

Both sides give two. A volume integral of a derivative and a sum of six surface integrals, landing on the same number — that is the theorem doing its job, and it is worth a pause rather than rushing on.

42. Theorem 3: Stokes' theorem

Concept

\[ \int_S (\nabla \times \mathbf{v}) \cdot d\mathbf{a} = \oint_P \mathbf{v} \cdot d\mathbf{l} \]

Same cancellation trick, one dimension down: tile the surface with tiny loops and every internal edge is traversed twice, in opposite directions.

Two consequences worth memorising:

• The answer depends only on the boundary line, so any surface with the same rim gives the same result — a soap film and a balloon stretched on the same wire agree
• If the loop is closed up into a surface with no boundary, the integral is zero

Same cancellation trick as the divergence theorem, one dimension down.

Tile the surface with tiny loops. Every internal edge gets traversed twice, in opposite directions, and cancels. Only the outer rim survives, and that is the boundary curve.

Two consequences worth memorising. The answer depends only on the boundary, so any surface sharing that rim gives the same result — a flat disc and a balloon stretched on the same wire agree exactly. And a closed surface, having no boundary at all, gives zero.

That second one is why the flux of any curl through a closed surface vanishes, a fact that turns up repeatedly in Chapters 5 and 7.

43. Worked example: checking Stokes' theorem

Worked example

Griffiths' Example 1.11, on the unit square in the yz plane with corners at the origin, (0,1,0), (0,1,1) and (0,0,1).

\[ \mathbf{v} = (2xz + 3y^2)\,\hat{\mathbf{y}} + 4yz^2\,\hat{\mathbf{z}} \]

\[ (\nabla\times\mathbf{v})_x = \frac{\partial}{\partial y}(4yz^2) - \frac{\partial}{\partial z}(2xz+3y^2) = 4z^2 - 2x \]

\[ \int_0^1\!\!\int_0^1 4z^2\,dy\,dz = 4\left[\tfrac{z^3}{3}\right]_0^1 = \frac{4}{3} \]

\[ \oint \mathbf{v}\cdot d\mathbf{l} = 1 + \tfrac{4}{3} - 1 + 0 = \tfrac{4}{3} = \int_S (\nabla\times\mathbf{v})\cdot d\mathbf{a} \]

Both sides again, and this example carries a specific procedural lesson.

Take the curl, keep only the component along the surface normal, and integrate over the unit square. Then walk the boundary counter-clockwise and add the four legs.

The lesson is Griffiths' own advice at the end of this example. Always write the line element with plus signs and let the limits carry the direction. Leg three comes out negative purely because its limits run from one down to zero, not because anyone inserted a minus sign by hand.

That discipline removes almost all orientation errors from line integrals, and adopting it is the difference between losing marks to sign conventions and not.

44. Match each theorem to the job it does

Matching

This is the whole decision procedure. Look at what kind of derivative is sitting inside the integral, and the theorem is chosen for you.

This is the decision procedure, and it is simpler than it looks.

Look at what kind of derivative sits inside the integral. A gradient means theorem one. A divergence means theorem two. A curl means theorem three. The theorem picks itself.

Each one drops the dimension by one: a line becomes two points, a volume becomes a surface, a surface becomes a loop.

If you are ever unsure which theorem to use, the answer is almost never to think harder about the geometry — it is to look at the integrand.

45. Trap: assuming every line integral is path-independent

Trap

46. Pattern: choosing your theorem

Pattern

1. Look at the integrand. A gradient, a divergence, or a curl? That alone picks the theorem
2. Ask which side is easier. Theorems are used to escape a hard integral — if the volume integral is nasty, do the surface one instead, and vice versa
3. Check the geometry matches. Divergence theorem needs a CLOSED surface. Stokes needs a surface WITH a rim
4. Fix the orientation before integrating. Outward normals for closed surfaces; right-hand rule linking loop direction to da for Stokes

In Chapter 2 you will use the divergence theorem to derive Gauss's law in differential form, and Stokes' theorem to prove that the electrostatic potential is well defined. Both are one-line applications of this table.

Four steps, and the second is the one that saves time in practice.

The integrand picks the theorem. Then ask which side is easier — the entire purpose of a theorem here is to escape a hard integral, so if the volume integral is unpleasant, do the surface one instead, and the other way round.

Check the geometry actually matches. The divergence theorem needs a closed surface; Stokes needs a surface with a rim. Applying one to the wrong kind of region is a common error.

And fix the orientation before integrating rather than afterwards: outward normals for closed surfaces, right-hand rule linking loop direction to the area vector for Stokes.

47. Check yourself: the theorems

Check

Decide before you click.

The reasoning is to notice which derivative you have been handed and let that pick the theorem.

The divergence is known and constant, and the surface is closed — that is the divergence theorem exactly. The flux becomes the constant times the enclosed volume, with no surface work at all.

Option B mixes up the theorems: Stokes relates a surface integral of a curl to a loop, and a closed sphere has no boundary loop to relate to.

Option C reflects a common belief that these theorems only work on simple shapes. The divergence theorem applies to any closed surface whatsoever.

48. Coordinates That Fit the Symmetry

Section

Section 4

49. Spherical coordinates

Concept

Use these whenever the physics only cares about distance from a point — which, given that Coulomb's law is a one-over-distance-squared rule, is most of Chapter 2.

These get used more than any other system in this book, because Coulomb's law depends on distance from a point.

Three coordinates, read against the picture. r is the distance from the origin. Theta is the angle down from the z axis, running from zero to pi. Phi is the angle around from the x axis, running from zero to two pi.

The last column of the table is the useful one: holding each coordinate fixed generates a surface. Constant r is a sphere, constant theta is a cone, constant phi is a half-plane. Recognising which of those matches a problem's symmetry is how you choose coordinates.

One warning. Physicists and mathematicians swap theta and phi. Griffiths uses the physics convention, and mixing sources is a reliable route to confusion.

50. Decoding dl, da and d-tau in spherical coordinates

Notation

This is the single most-consulted equation set in the book. Every term is an arc length, and each one earns its factors geometrically.

\[ d\mathbf{l} = dr\,\hat{\mathbf{r}} + r\,d\theta\,\hat{\boldsymbol{\theta}} + r\sin\theta\,d\phi\,\hat{\boldsymbol{\phi}} \]

\[ d\mathbf{a} = r^2\sin\theta\,d\theta\,d\phi\;\hat{\mathbf{r}} \quad \text{(on a sphere of radius } r\text{)} \]

Learn to reconstruct these from the picture rather than memorising them: an angle is never a length until you multiply it by the radius it turns about.

This is the most consulted equation set in the book, and memorising it without understanding it is exactly how it goes wrong under pressure.

The principle underneath is simple: an angle is never a length until you multiply it by the radius it turns about. Every factor in these formulas is an application of that.

Take the volume element apart. The r-squared is there because a shell at radius r has area growing as r squared. The sine of theta is there because circles of constant latitude shrink toward the poles — at the equator sine is one, at the poles it is zero and the circle has collapsed to a point.

Reconstruct rather than recall. Try deriving the area element on a sphere straight from the picture, and check you can say why there is no dr in it.

51. Worked example: the volume of a sphere from the integral

Worked example

Griffiths' Example 1.13. A result you already know, used to prove the volume element is right.

\[ V = \int_0^{2\pi}\!\!\int_0^{\pi}\!\!\int_0^{R} r^2\sin\theta\,dr\,d\theta\,d\phi \]

\[ V = \left(\int_0^R r^2\,dr\right)\left(\int_0^{\pi}\sin\theta\,d\theta\right)\left(\int_0^{2\pi} d\phi\right) \]

\[ V = \left(\frac{R^3}{3}\right)(2)(2\pi) = \frac{4}{3}\pi R^3 \]

This computes something you already know, which is exactly the point: it validates the volume element rather than the answer.

The integral factorises because the integrand is a product of a function of r, a function of theta, and nothing in phi. Three independent one-dimensional integrals.

The middle factor is the one to watch. The integral of sine from zero to pi is two, not zero — being used to sine integrating to zero over a full period is exactly how this one goes wrong.

And the verification is the whole purpose: the answer is four-thirds pi R cubed, which is the sphere volume everybody knows. Had the sine been forgotten, the result would have been a sphere half its true size, which is the kind of error this check exists to catch.

52. Trap: treating r-hat, theta-hat and phi-hat as constants

Trap

53. Cylindrical coordinates

Concept

\[ d\tau = s\,ds\,d\phi\,dz \]

Use these when the physics cares about distance from a line: a charged wire, a long cylinder, a solenoid. Chapter 2's Example 2.3 is exactly this case.

Note the single factor of s, not s squared. A circle of radius s has circumference proportional to s — one factor, not two.

The second most-used system, and the one that fits wires, cables and solenoids.

The critical detail is in the picture: s is the distance from the axis, not from the origin. It gets conflated with the spherical r constantly, and every step after that goes wrong.

The volume element carries a single factor of s, not s squared, and the reason is geometric: a circle of radius s has circumference proportional to s — one factor, not two. Holding this against the spherical case is a good way to make both stick.

Reach for these whenever the physics depends on distance from a line and is unchanged along it. Chapter 2's charged-cylinder example is exactly that case.

54. Which coordinate system does this object want?

Discrimination

Choosing coordinates badly turns a two-line problem into a two-page one. Match the coordinates to the symmetry, never to habit.

This is the same judgement Gauss's law will demand in Chapter 2, where the choice of Gaussian surface is exactly this choice of symmetry.

A judgement drill, and the judgement matters more than the answers.

Match the coordinates to the symmetry of the source, never to habit. If the physics depends only on distance from a point, use spherical. Distance from a line, cylindrical. Distance from a plane, Cartesian.

Choosing badly does not make a problem impossible, but it can turn a two-line calculation into two pages of unnecessary algebra.

This is the same judgement Gauss's law will demand in the next deck, where picking a Gaussian surface is precisely this choice of symmetry. Getting it right here pays off immediately.

55. The Delta Function and the Theory of Fields

Section

Section 5

56. The Dirac delta function

Concept

\[ \delta(x) = \begin{cases} 0, & x \neq 0 \\ \infty, & x = 0 \end{cases} \qquad \int_{-\infty}^{\infty}\delta(x)\,dx = 1 \]

It is not a function in the ordinary sense — it is infinitely tall and infinitely thin. It is legitimate only inside an integral, and that is the only place you will ever use it.

This is not a function in the ordinary sense, and its strangeness is the point rather than a problem to be smoothed over.

It is zero everywhere except at a single point, infinite there, and yet has total area one. No ordinary function does that — and the animation shows why it is nonetheless well defined: it is the limit of a family of perfectly ordinary bumps, each with area one, getting taller and narrower.

The crucial usage rule is that it is only legitimate inside an integral. Asking for its value at a point is meaningless; asking what it does to an integral is the only well-posed question.

If that feels unsatisfying, the instinct is right — making it rigorous takes distribution theory — but the operational rules are simple and reliable.

57. The delta function is a sampling device

Intuition

Stop thinking of it as a spike and start thinking of it as an instruction: evaluate the other factor at this point.

\[ \int_{-\infty}^{\infty} f(x)\,\delta(x-a)\,dx = f(a) \]

Everywhere except the spike, the delta is zero and kills the integrand. At the spike, the other factor is effectively constant and comes out of the integral, which then contributes its unit area.

That is why a point charge can be written as a charge density: a delta function of the right strength.

\[ \rho(\mathbf{r}) = q\,\delta^3(\mathbf{r} - \mathbf{r}') \]

This reframing is what makes the delta function usable, so it is worth taking seriously.

Stop thinking of it as a spike and start thinking of it as an instruction: evaluate the other factor at this point. That is all it ever does.

The mechanism is visible in the picture. Everywhere except the spike, the delta is zero and kills the integrand. At the spike, the other factor is effectively constant over an infinitesimal width and comes out of the integral, leaving the delta's unit area behind.

And here is the payoff for Chapter 2: a point charge can be written as a charge density. That single move lets discrete and continuous distributions be handled by the same equations, which is why the general formulas in Chapter 2 never need a separate case for point charges.

58. Worked example: delta functions inside integrals

Worked example

Griffiths' Examples 1.15 and 1.16 in miniature. The whole method is: locate the spike, evaluate everything else there.

\[ \int_0^5 x^3\,\delta(x-2)\,dx = 2^3 = 8 \]

\[ \int_3^5 x^3\,\delta(x-2)\,dx = 0 \]

\[ \delta(kx) = \frac{1}{|k|}\delta(x) \]

Three short cases, each drilling one rule.

First, the spike is inside the range: the answer is the other factor evaluated at the spike. Second, the spike is outside the range: the answer is zero, because the delta vanishes everywhere in the region being integrated over.

Third, a scaled argument. A delta of kx is narrower by a factor of k, so its area shrinks by the same factor and a compensating one over k appears. The absolute value is essential — a negative k does not produce a negative area.

The verification frames all three in terms of area captured, which is the unifying idea and much easier to reconstruct under exam conditions than three separately memorised rules.

59. The identity that fixes Coulomb's law

Concept

Here is a genuine paradox, and its resolution is the reason the delta function is in this chapter at all.

\[ \mathbf{v} = \frac{1}{r^2}\hat{\mathbf{r}} \]

Compute its divergence with the spherical formula and you get zero everywhere. But the flux of this field out of any sphere centred on the origin is 4-pi, no matter how small the sphere.

A divergence of zero everywhere would force zero flux. The escape: the divergence is not zero at the origin, where the formula breaks down. All of the source is packed into that single point.

\[ \nabla \cdot \left(\frac{\hat{\mathbf{r}}}{r^2}\right) = 4\pi\,\delta^3(\mathbf{r}) \]

Chapter 2 uses this in one line to get from Coulomb's law to Gauss's law in differential form. Without it, the derivation has a hole at the location of every charge.

There is a genuine paradox here, and it is worth sitting with it before the resolution.

Compute the divergence of the inverse-square radial field using the spherical formula and you get zero everywhere. But the flux of that field out of any sphere centred on the origin is four-pi, no matter how small you make the sphere.

Those two statements are incompatible, because the divergence theorem says the flux equals the volume integral of the divergence — and a divergence of zero everywhere would force zero flux.

The resolution is that the divergence is not zero at the origin. The spherical formula breaks down there because the field is singular, and all of the source is packed into that single point — which is exactly what a delta function describes.

And the payoff: Chapter 2 uses this identity in one line to get from Coulomb's law to the differential form of Gauss's law. Without it, that derivation has a hole at the location of every charge.

60. The theory of vector fields: why potentials exist

Concept

Helmholtz's theorem says a vector field is completely determined by its divergence and its curl (given sensible behaviour at infinity). That is why Maxwell's equations come in exactly four parts: a divergence and a curl for each field.

\[ \nabla\times\mathbf{F} = 0 \;\Rightarrow\; \mathbf{F} = -\nabla V \qquad \nabla\cdot\mathbf{F} = 0 \;\Rightarrow\; \mathbf{F} = \nabla\times\mathbf{A} \]

The minus sign in the first is pure convention, chosen so that positive charges roll downhill in potential. Chapter 2 defines exactly this V for the electric field; Chapter 5 defines exactly that A for the magnetic field.

This is the structural result that explains why Maxwell's equations look the way they do.

Helmholtz's theorem says a vector field is completely determined by its divergence and its curl, given sensible behaviour at infinity. That is why the equations come in exactly four parts: a divergence and a curl for each field. Nothing more is needed, and nothing less would do.

The table gives the two consequences that matter. A field with zero curl everywhere can be written as the gradient of a scalar potential. A field with zero divergence everywhere can be written as the curl of a vector potential.

The minus sign in the first is pure convention, chosen so that positive charges roll downhill in potential. Chapter 2 defines exactly this V for the electric field, and Chapter 5 defines exactly that A for the magnetic field — worth naming both now so the parallel is visible when it arrives.

61. Where each tool shows up in the next three chapters

Real world

Divergence → Gauss's law

The divergence of E equals the charge density over epsilon-zero. Charge is where the field diverges.

Curl → potential exists

The curl of E is zero, so E is a gradient, so a scalar potential V is well defined everywhere.

Gradient → E from V

Solve for one scalar function V, then differentiate to get all three components of E.

Delta function → point charges

Point charges become densities, so discrete and continuous distributions obey the same equations.

These four cards close the loop on the promise at the start of the deck.

Divergence becomes Gauss's law: charge is where the field diverges. Curl becomes the guarantee that a potential exists. Gradient is the rule that recovers the field from that potential. And the delta function makes point charges legal inside integrals.

The written prompt is worth actually doing rather than skipping. If you cannot say in plain words what an operator measures, the formula will not rescue you in Chapter 2 — a formula is only useful once you know which question it answers.

Pick the one you feel least sure about and say it out loud. That is usually the one that will block you a week from now.

62. Pattern: the whole toolkit on one page

Pattern

If a Chapter 2 derivation ever looks like magic, it is almost always one row of this table applied without comment.

This table is the reference to come back to whenever a Chapter 2 derivation looks like magic, because it almost always is one of these rows applied without comment.

Read down the middle column, which is the meaning rather than the formula. Those phrases are the ones worth being able to produce from memory.

The right-hand column is the promise being kept: every tool has a named destination in the physics. Nothing here was included for completeness.

This is the single most useful page to have alongside the next three decks, so it is worth a photograph.

63. Check yourself: the toolkit

Check

Last one — take your time.

This tests whether the divergence-versus-curl distinction is secure, which is the thing most likely to cause trouble later.

Zero curl is exactly the condition for a field to be conservative, so it licenses the potential. That is the whole answer.

Option B is the one worth thinking about, because it confuses the two operators. Charge is controlled by the divergence of E, not its curl — and a point charge has a field with zero curl everywhere and plenty of charge.

Option C is worth a sentence too: a field can have no circulation at all and still be very strong. The radial field of a point charge is the standard example.

64. What you can do now

Recap

• Read Griffiths' separation vector and say which point is the source and which is the field point
• Compute a gradient, divergence and curl, and state in plain words what each one measured
• Pick the right integral theorem by looking at the derivative inside the integral
• Set up an integral in spherical or cylindrical coordinates, with the correct volume element
• Use the delta function to write a point charge as a density, and know why the divergence of r-hat over r-squared is not zero

Next deck: all of this becomes physics. Coulomb's law, the electric field, and Gauss's law.

Worth reframing the deck now that it is done. It looked like mathematics; it was actually a list of physical questions with notation attached.

Go down the one-line summaries and check each is attached to a picture — the gradient to the contour map, the divergence to the tiny box, the curl to the paddle wheel, the theorems to the region-and-boundary idea.

Then look ahead. All of this becomes physics in the next deck. The tiny box becomes a Gaussian surface, the paddle wheel becomes the statement that no electrostatic field circulates, and the delta function becomes the reason a point charge can appear in an integral at all.