Chapter 31: Tensors & the Geometry of Physics

Welcome to Last Minute Lecture.

This free chapter overview is designed to help students review and understand key concepts.

These summaries supplement not replaced the original textbook and may not be redistributed or resold.

For complete coverage, always consult the official text.

Okay, let's unpack this.

Welcome to a deep dive.

Today we're looking at why some pretty fundamental ideas in physics, you know, things involving crystals or even spinning objects, just don't quite work if we stick only to simple vectors.

Right.

So our goal today is to really get into the mathematical language that fixes this problem, tensors.

We're digging into the core ideas, trying to figure out why physics even needed this, well, more complicated way of describing things.

Yeah, because physics often starts out simple, almost deceptively so.

You push something, it accelerates in that direction,

cause and effect seem perfectly aligned.

Exactly.

We often begin with vector fields, right?

Where one quantity is just proportional to another, maybe multiplied by a single number.

Easy.

Uh -huh.

But the moment you look at real materials, materials with internal structure like crystals, that beautiful simplicity kind of breaks down.

The effect just isn't in the same direction as the cause anymore.

And that breakdown is precisely where tensors come in.

We needed a tool, a mathematical description that could handle these direction dependent properties, no matter how you look at it, how you set up your coordinates.

And the main idea we'll explore is how to relate one vector to another in these cases.

You often need this whole grid of numbers, not just one, but nine.

So the classic example to start with is dielectrics, materials that polarize.

You apply an electric field, let's call it E, that's the cause.

And that field causes the charges inside the material to separate a bit, creating polarization.

P, that's the effect.

Right.

Now, in simple stuff, isotropic materials like, say, air or water, P lines up perfectly with E, apply E pointing north, P points north.

Simple proportionality.

Okay, straightforward.

But now,

the crystals.

Yes, this is where it gets interesting.

Crystalline substances, what we call anisotropic materials, their internal structure, the lattice arrangement of atoms, dictates how charge can actually shift around.

So if I apply my electric field, E straight up.

I'd normally expect the polarization P to also point straight up.

You'd think so.

But the crystal lattice can act almost like a filter or maybe a prism for the field,

apply the field straight up, and the structure might make it easier for charges to separate slightly off to the side.

So the resulting polarization vector P might point mostly up, but maybe also a bit sideways, or even forward and back.

Exactly.

The polarization gets physically twisted relative to the field that created it.

The cause is up, the effect is tilted.

Okay, I can picture that physical twist.

But mathematically, why isn't it enough to just have three different constants?

One for how E affects P in the X direction, one for Y, one for Z, Y9.

That's the crucial bit.

If you only use three constants, you'd be assuming that the X part of the field, X, only affects the X part of the polarization, PX.

Same for Y and Z.

Right, that they're independent.

But in these anisotropic crystals, they aren't independent.

The field you apply in the X direction, X can actually contribute to the polarization in the X direction and the Y direction and the Z direction.

Hold on.

So putting a field on pointing, say, east,

could make the material polarize a little bit east, but also a little bit north.

Yes.

Because the crystal's internal structure couples those directions.

To fully figure out the resulting polarization vector P, each of its components,

PX, PY, PS, depends linearly on all three components of the electric field, EX, ES, and ES.

Okay, let me track that.

So P depends on east, E north, and YUP.

That's three numbers, three constants, just for the east component of P.

And then I need another three constants for the P north component, because it also depends on all three Es and another three for PUP.

Exactly.

Three inputs affecting each of the three outputs.

Three times three.

Yeah.

That gives us And that grid, that array of nine numbers, usually written as a three -by -three matrix, that's what we call the tensor of polarizability.

It's known as a second -rank tensor because it connects two vectors, E and P.

And this matrix gives the complete description, regardless of which way I point the E field.

It gives the complete linear relationship, yes.

It maps any input E vector to the resulting P vector for that specific material.

All right.

So we have this three -by -three matrix of nine numbers, but just being a matrix isn't enough to be called a tensor, right?

It has to behave in a special way if we change our viewpoint.

That's absolutely key.

Yeah.

If you measure these nine numbers using one set of coordinate axes, say XYZ, and then your colleague comes along and uses a rotated set, XYZ, the numbers they measure for the tensor components will be different.

Okay.

But they change according to very specific mathematical rules.

It's like describing a sculpture.

From different angles, the coordinates you use to describe points on its surface change.

But the sculpture itself doesn't change.

Exactly.

The underlying physical reality, the sculpture, or in our case, the physical relationship between the electric field E and the polarization P stays the same.

A tensor is the mathematical object whose components transform in just the right way to guarantee this physical invariance, regardless of the coordinate system you choose.

So the tensor is the physical reality in a sense, not just the list of nine numbers in one particular coordinate system.

Okay.

That makes sense.

Now, what about the energy involved here?

Polarizing the crystal takes work.

Yes.

The work done per unit volume, let's call it up a stare, you can derive an expression for this energy density, and it involves the components of the electric field and the polarization.

And this energy calculation gives us something important.

It gives us a really critical physical constraint.

It turns out based on thermodynamic arguments that the polarization tensor must be symmetric.

Symmetric meaning the number that connects x to pi has to be the same as the number that connects a to pi.

Like alpha ti must equal alpha.

Precisely.

Same for the xe and yz pairs.

Why?

What does thermodynamics have to say about this?

Well, think about the energy stored.

If the tensor wasn't symmetric, the amount of energy you'd get back when you removed the field could depend on the order in which you turned the field components off.

Ah, so it would be path -dependent.

You could potentially extract energy just by cycling the fields, which isn't allowed for polarizing a dielectric crystal.

It has to be reversible thermodynamically.

Exactly.

So thermodynamics forces the symmetry onto the tensor.

And that's huge.

If alpha, alpha, and similarly for the other pairs, we don't have nine independent numbers anymore.

Nope.

The nine components are reduced to only six independent values.

The three on the diagonal, alpha, and three off -diagonal ones like alpha, that's a significant simplification purely from physics principles.

Wow.

Okay.

So six numbers capture the full anisotropic behavior and this leads to a nice visualization.

It does.

This symmetric tensor allows us to visualize the energy density, a p.

If you plot all the points in e field space that correspond to the same energy density, you get a shape.

And that shape is?

An ellipsoid.

A three -dimensional ellipse like a squashed or stretched sphere is called the energy ellipsoid.

I like that.

So the shape of this ellipsoid visually tells you about the crystal's polarization properties.

Yes.

If the ellipsoid is, say, really long and thin in one direction, it means it's much easier to store polarization energy if the e field points that way compared to the short directions.

It visually represents the anisotropy.

Like a cigar shape means it prefers to polarize along the cigar's axis.

You got it.

And the beauty of this ellipsoid is that mathematically,

any ellipsoid has three special axes perpendicular to each other, its principal axes.

Think of the longest direction, the shortest, and the one in between.

Right.

And if we align our coordinate system, our x, y, z axes, with these principal axes of the energy ellipsoid.

Then the mathematics simplifies beautifully.

In this special coordinate system, all the off -diagonal terms of the polarizability tensor become zero.

The ones that represented that twisting effect, gone.

So the tensor becomes diagonal.

Just three numbers left on the main diagonal.

Alpha.

Exactly.

When the e field is applied along one of these principal axes,

the resulting polarization p is perfectly parallel to e.

No more tilting.

The description is much cleaner along these natural axes of the crystal.

And what if, for some material, those three principal coefficients, alpha, alpha, alpha, all happen to be the same number?

Ah, well then the ellipsoid wouldn't be stretched or squashed at all.

It would be a perfect sphere.

Which means the material behaves the same way in all directions.

Precisely.

It would be completely isotropic.

And in that case, p is always parallel to e, no matter the direction, and the whole relationship boils down to just that one single number, that scalar polarizability we started with.

The tensor machinery handles the simple case too.

Okay, so this tensor idea, this three by three matrix structure, it's not just for polarization, right?

This seems like a general tool.

Oh, absolutely.

It pops up all over physics.

Anytime you have a situation where a vector input, like e, produces a vector output, like p, but the direction gets modified by the properties of the medium or the object, you'll likely need a second rank tensor.

So give us another example.

What about conductivity?

Current flow.

Good one.

We usually think current flows in the same direction as the electric field that drives it, like in a simple wire.

Ohm's law.

J equals sigma e, where sigma is just a number.

Right.

But in some materials, especially certain crystals or metals in magnetic fields, that's not true.

The current density J might flow at an angle to the electric field e that's pushing the electrons.

Because the crystal structure, again, channels the electrons in preferred directions.

Exactly.

So the relationship between J and E is defined by a conductivity tensor, another three by three matrix, usually denoted sigma error.

Push electrons one way, they drift off somewhat sideways.

Okay.

Conductivity tensor.

Makes sense.

What about mechanics?

Anything outside E and M?

Definitely.

Think about rotation.

If you take an irregularly shaped object, like maybe a potato, and spin it around some axis.

Okay.

Spinning potato.

Got it.

It's angular velocity vector.

Omega points along the axis of rotation.

But it's angular momentum vector, L, which represents sort of the amount of rotational motion.

Well, it generally does not point along the same axis as omega.

It doesn't.

Why not?

Because the mass isn't distributed symmetrically around that arbitrary rotation axis.

More mass might be swinging around further out on one side.

The relationship between the angular velocity omega and the angular momentum L is given by the moment of inertia tensor, often written as I.

Another three by three matrix.

Yep.

Symmetric.

Second rank.

It describes how the mass is distributed relative to the rotation point.

And just like with polarization, you can find principal axes of inertia for the axis,

where if you spin it around one of them, L is parallel to omega.

Rotation is simpler along those axes.

Fascinating.

Okay.

Polarization, conductivity, inertia.

What else?

You mentioned internal forces.

Stress.

This is fundamental in material science and continuum mechanics.

Stress describes the internal forces that parts of a continuous material exert on each other.

Like pressure inside a fluid.

Pressure is the simplest case.

Imagine you have a solid block.

Now, mentally slice it with a plane at some angle.

The material on one side of the cut exerts a force on the material on the other side.

That force vector doesn't necessarily have to be perpendicular to the surface of the cut.

It can have components both perpendicular, normal stress, and parallel, shear stress to the surface.

Ah, shear stress.

Like rubbing.

Kind of.

So, to describe the state of internal forces at a point, you need to know what the force vector would be for any orientation of the cut you might make through that point.

That sounds complicated.

It is.

And it's captured by the stress tensor.

It's another symmetric second -rank tensor.

Its components tell you things like, what is the x component of the force acting on a small area whose normal vector points in the a direction?

That would be sick.

Nine components, again, reduced to six by symmetry, describing all the normal and shear forces internally.

Exactly.

Simple hydrostatic pressure is just the special case where the stress tensor is diagonal and the three diagonal terms are all equal to the negative of the pressure such as size of the juries.

For anything more complex, like a stretched beam or a torqued shaft, you need the full stress tensor.

So far, we've mostly talked about rank zero scalars, just numbers, rank one vectors, and these rank two tensors, the three by three matrices.

Does it go higher?

Oh, yes.

You can have tensors of rank three, rank four, and so on.

For example, in elasticity, the relationship between the stress tensor, rank two, and the strain tensor, also rank two, describing deformation, is given by a fourth -rank tensor, the elasticity tensor, that has 34 Borel -Opper's 81 one -amarty equal components, though symmetries reduce that number significantly.

Wow, okay.

But the material we're drawing from focuses on one particularly important higher -rank example, right?

Something involving relativity.

Yes.

The capstone example is the connection to relativity and electromagnetism itself through the concept of the four tensor.

We move from 3D space to 4D space time.

So now it's not a three by three matrix.

It's a four by four matrix operating in four dimensions, three space, one time.

The specific one discussed is the electromagnetic energy momentum tensor, sometimes called the stress energy tensor, often denoted Mu, where Mu and Nu run from one to four or zero to three, depending on convention.

It has 16 components.

16 components.

What does this mega tensor tell us?

It's really quite profound.

It essentially bundles together all the information about distribution and flow of energy and momentum within the electromagnetic field into a single mathematical object that behaves correctly under Lorentz transformations in special relativity.

Okay.

So how do the pieces we know fit into this four by four structure?

Where's energy density?

Where's momentum?

Let's think about the indices.

If one index is the time index, say index four, and the other is a space index, say x index one, then seven one or some 14 dollars, represents the density of the x component of momentum in the field.

Similarly for y and z momentum.

So the time space components give momentum density.

What about time time?

The second component represents the energy density of the electromagnetic field.

Okay.

Energy density and momentum density are in there.

What about the purely spatial parts?

The three by three block C geller, where i and j are both space indices.

One, two, three.

That three by three block is essentially the Maxwell stress tensor.

It describes the flow momentum through space.

Its components relate to the forces the electromagnetic field exerts, the stresses, the pressures, and shears within the field itself.

So the stress tensor concept we just discussed for materials reappears here, but for the field.

Exactly.

And things like the pointing vector, which describes the flow of energy or related to it, are also neatly packaged within this four tensor structure.

It unifies energy density, momentum density, energy flow, pointing vector, and momentum flow, Maxwell stresses, into one relativistic object.

Hashtag outro.

So let's try to wrap this up.

What does this all mean for you as you're trying to understand physics?

Tensors fundamentally seem to be the language physics absolutely needs when things get directionally complicated.

That's the core idea.

They pull us away from that simplified picture where cause and effect always line up perfectly, and they allow us to describe the real world anisotropic crystals, complex rotations, internal stresses, and materials, even the fabric of space time and fields.

The key takeaway seems to be that a tensor defines a physical relationship or quantity in a way that remains true no matter how you orient your measuring sticks, your coordinate system.

It guarantees the physics is consistent.

Exactly.

We saw the polarizability tensor needed those nine components for crystals, but then symmetry rooted in thermodynamics cut it down to six independent ones.

And that symmetry gave us the energy ellipsoid visualization, which then showed us how choosing special principal axes simplifies the math again, making the tensor diagonal in that preferred frame.

Right.

And we saw this tensor pattern repeat for conductivity, for inertia, for stress, wherever directionality matters in relating to vector -like quantities.

It really shows that sometimes the deepest insights aren't in the simple linear relationships,

but in understanding these more complex interconnected matrix structures that govern how things behave.

Absolutely.

The three -by -three tensor handles directional properties in 3D space, and the four -by -four tensor steps it up to handle energy and momentum flow in spacetime, all while respecting relativity.

So maybe a final thought for you to consider.

We've seen how vectors rank one handle simple motion, and how these second -rank tensors handle things like polarization and rotation.

What other, perhaps even more intricate phenomena in nature might demand the even more complex language of higher -ranked tensors, maybe rank four or beyond, for us to truly grasp them?

What else out there is hiding in that complexity?