Chapter 4: Crystallographic Computations

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.

Welcome back to the Deep Dive.

Our mission here is to cut through complexity and give you

you need, fast.

Today, we're taking on a problem that has really been at the core of material science for centuries.

And that's how do you measure things, distances, angles when the space you're working in is in a neat little box with perfect right angle.

Exactly.

I want to open this deep dive with a quote that I think just perfectly sets the stage.

It's from Erwin Schrödinger.

Oh, this is a good one.

He said, we are told such number as the square root of two worries, few worried Pythagoreans and his school almost a distraction.

Being used to such queer numbers from early childhood, we must be careful not to form a low idea of the mathematical intuition of those ancient sages.

Their worry was highly justified.

And that is so foundational to what we're talking about because it captures that exact moment when your mathematical comfort zone just hits a wall.

The wall of uncomfortable reality.

Precisely.

Our comfort zone is that simple 90 degree world, the Cartesian frame where Pythagoras is king.

But in, you know, the vast majority of real crystal structures, the axes are not at 90 degrees.

And all that high school geometry just spectacularly fails.

It falls apart completely.

So that's the tension we're diving into.

We are really getting into the geometry of materials specifically focusing on crystal graphic computations.

And our mission today is what's ambitious, but it's clear.

We're going to introduce the metric tensor, the metric tensor.

This isn't just some mathematical curiosity.

It is the universal computational tool.

It radically simplifies how you calculate distances, directions, angles across every across all crystal systems from the simplest cubic lattice all the way to the most asymmetric triclinic system.

So the knowledge we're unpacking today is genuinely a shortcut.

It absolutely is.

If you've ever looked at a crystallography textbook and seen, you know, seven different messy equations for the seven different crystal systems.

Oh, they're horrible.

You're about to see how all of that just collapses into one single elegant mathematical framework.

We'll start with the basics, defining direction, and then we'll hit that fundamental problem of distance before we really build up the metric tensor in its full 3D glory.

Okay, let's do it.

Let's start with directions and what we're calling the Cartesian conundrum.

Sounds good.

So where does the geometry begin?

It has to start with directions in the lattice.

Right.

And a crystallographic direction is at its heart, just a vector.

Let's call it T.

And like any vector, it's defined by two things, its length and its orientation.

Its orientation relative to the crystal axis, which we call A, B, and C.

To describe that vector, we use something called crystallographic indices.

The ones you always see in the square brackets.

Exactly.

Written as U, V, W.

These indices, U, V, and W, are just the components of that translation vector T you need to get from the origin to a specific point in the lattice.

Okay, but what's the absolute most critical rule about those three numbers?

This is the non -negotiable part of crystallography.

U, V, and W must be the smallest integers that are proportional to the components of the vector.

Smallest integers.

So for example, if your point is at,

say, one -third of the way along the A axis, two -thirds along B, and one full unit along C.

So your coordinates are 13, 23, 1.

Right.

You can't use those fractions.

You have to multiply everything by three to clear them out, which gives you the direction one, two, three.

It's all about finding that simplest integer ratio.

And that's to avoid ambiguity.

Completely.

And there's also a specific notation for negative directions, right, if you're going backward along an axis.

Yeah, you use the bar over the number.

So a vector with components 1, negative 1, 2 would be written 1 bar 2.

It's just a very clear visual cue.

The chapter also makes a practical point about what to do if the numbers get really big.

Yeah, that's a good point.

If you have something like 12, 3, 4, you might write it with commas like 12, 3, 4, just to make sure no one reads it as 1, 2, 3, 4.

It's a clarity thing.

I think we need to pause here on a really important idea about how we visualize these directions.

Okay.

The book shows a diagram, figure 4 .1, where the same direction, 100, is plotted in two very different lattices.

One is orthogonal and the other is really skewed.

Right, it's non -orthogonal.

What's the core lesson from that visualization?

The lesson is that you have to separate the index definition from the space.

The direction 100 is, by definition, parallel to the A axis, full stop.

It doesn't matter what that A axis is doing.

It doesn't matter if that axis is leaning over at some crazy angle to the B axis.

The 100 vector still just points straight along.

So the indices define directions within the lattice's own coordinate system, not in some absolute external Cartesian space.

Exactly.

And getting that distinction is really the first big step toward understanding all of this non - geometry.

All right.

Directions handle.

Let's get to the real problem.

Distance.

This is where the Pythagorean paradox comes in.

This is where it gets interesting.

Okay, let's take that simple problem.

Find the distance D from the origin, 00, 0 to the point, 111.

If we're in the easiest possible situation, a simple cubic crystal with a lattice parameter A, we just use the 3D version of Pythagoras' theorem.

Right.

D squared is just a squared times one squared plus one squared plus one squared.

So D equals a times the square root of three.

Simple.

And it works perfectly because in a cubic system, your movements along the A, B, and C axes are all mutually perpendicular.

They're orthogonal.

So you can just add the squares of the length.

You can.

But the second you step out of that cubic comfort zone and into, say, a monoclinic or a triclinic system, the whole thing just collapses.

Why?

What's the fundamental reason it fails?

Because that core assumption of orthogonality is gone.

Think about finding the diagonal of a parallelogram instead of a rectangle.

You can't just use a squared plus B squared.

No, you have to use the law of cosines.

You need a correction factor based on the angle between the sides.

In 3D, it's the same idea, but, you know, much more complex.

You have to account for the angles between all three axes.

So that simple formula is just invalid.

We need a universal method, one that only relies on the basic parameters of the lattice.

The three lengths A, B, C, and the three angles, alpha, beta, gamma.

And it has to work for any values of those parameters?

Correct.

And to build that method, we have to go back to the bedrock of vector math,

the dot product.

Okay, let's define it conceptually like the book does.

What is the dot product of A and B?

Geometrically, the dot product is the length of vector A multiplied by the length of the projection of vector B onto A.

It's a measure of how much they point in the same direction.

Exactly.

Mathematically, it's A, B, cos theta.

And that's a key distinction.

In a high school algebra class, you just multiply the components by 1 by 2 plus y1, y2.

But its real power, its geometric meaning comes from that cosine term.

Right.

And that immediately gives us a universal way to define length.

It does.

If you dot a vector A with itself, the angle theta is zero and the cosine of zero is one.

So the length of vector A, A is just the square root of its dot product with itself.

A equals the square root of a dot A.

Now we have our foundation.

So if we have two points, P and Q, we can define the vector between them as the difference, QP.

The distance D is just the length of that vector.

Which gives us our universal formula for distance.

D equals the square root of QP dotted with QP.

This formula is always true.

The challenge is just how do we actually calculate that dot product in a skewed, non -Cartesian system?

That's the million dollar question.

If we can solve the dot product problem, we solve the distance problem.

Universally.

Which brings us to the hero of this story, the metric tensor.

A universal key.

Okay.

So if the lattice is non -orthogonal, how do we calculate that dot product?

The simple component multiplication doesn't work.

We have to build the geometry directly into the calculation itself.

What does that mean?

It means the calculation has to depend on the dot products between the basis vectors themselves.

A, B, and C.

So we're creating a sort of geometric fingerprint of the unit cell.

How many of these relationships do we actually need to define the whole thing?

We need six.

Six unique values.

Okay, what are they?

You need the three self dot products.

A dot A, B dot B, and C dot C.

Which are just the squared lengths.

A squared, B squared, C squared.

Easy enough.

Then you need the three unique cross dot products.

A dot B, B dot C, and a dot C.

And those six values contain all the geometric information.

All three lengths and all three angles are baked into them.

Everything you need to know is right there.

And this is exactly what the metric tensor G is for.

It's a way to store those six relationships in one convenient package.

It's the ultimate shorthand.

It's a three by three matrix where each component, G, is just the dot product of basis vector I with basis vector J.

So G equals

That's the definition.

Let's break down one of the off diagonal components, say G12.

What is that?

G12 is a dot B.

And we know from the geometric definition that this is equal to the length of times the length of B times the cosine of gamma, the angle between them.

And why is that term so critical?

Because if gamma is not 90 degrees, then cos gamma is not zero.

That term, G12, becomes the correction factor you have to apply when calculating distances in the AB plane.

So if we left it out?

If you left it out, your calculation would be assuming A and B are perpendicular and you get the wrong answer.

The tensor literally stores how skewed the axes are.

Okay, let's describe the full three by three matrix verbally.

This is the centerpiece of the whole deep dive.

Absolutely.

So the main diagonal, that's the easy part.

It's just the squared lengths.

G11 is a squared, G22 is B squared, and the D33 is C squared.

And the off diagonal terms are where the angles live.

And since a dot B is the same as B dot A, the matrix is symmetric.

It is.

So we have our three unique cross dot products.

Willis -Lis.

Okay, G12 and G21 are both AB cos gamma.

They correct for the angle between A and B.

And G13 and G31.

Those are ACOS beta, correcting for the angle between A and C.

And finally.

G23 and G32 are BC cos alpha for the angle between B and C.

So when you look at that three by three matrix, you're looking at a complete self -contained geometric description of the unit cell.

Everything is in there.

Now let's revisit the distance equation using this tensor.

The source gives it in matrix notation.

D square equals QPG QP transpose.

Right.

And this is a point that can trip people up if they're new to matrix algebra.

Let's say our difference vector is R.

That's QPP.

So R is a row vector, a one by three matrix.

It is.

And G is a three by three matrix.

The rules of matrix multiplication say the inner dimensions have to match.

So one by three times three by three works.

And that gives you another one by three vector.

It does.

But we need a single number, a scalar, for the distance squared.

So to finish it, we have to multiply that result with a second R vector, which has to be a column vector, a three by one.

And that's what the transpose symbol, the T, does.

It flicks the row vector into a column vector.

Precisely.

It's purely a requirement of the math to make sure you end up with a single one by one scalar value.

So that structure row vector tensor column vector is what makes the equation work for any crystal system.

The complexity is all bundled up inside G.

And just for context, the book mentions this comes from differential geometry.

It's the same concept Einstein used for measuring distances in curved spacetime in general relativity.

Here we're just using it for, you know, a slightly simpler problem.

Flat but skewed space.

It's a mathematical giant repurposed for material science.

So let's really solidify this with the universal definition of the dot product itself.

P dot Q equals pi G Q J.

Right.

Instead of just multiplying the components together, you use this matrix multiplication.

The metric tensor G acts like a geometric filter.

A filter.

I like that.

Yeah.

When you do the multiplication, G makes sure that every component pair multiplication by a Q J gets weighted by the right correction factor G based on the geometry.

This all sounds pretty complicated if you're just dealing with a simple cubic system.

What happens when you apply this universal tool to the easiest case?

It simplifies beautifully and it shows how consistent the whole system is.

In a standard cubic system, A, B, C are all equal.

Let's just say they're one for now.

And all the angles alpha, beta, gamma are 90 degrees.

And the cosine of 90 degrees is zero.

So all of those off diagonal terms and the metric tensor, the Avakos, gamma and so on, they all become zero.

And the diagonal terms, A squared, B squared, C squared, they all just become one.

Exactly.

The metric tensor G collapses into a matrix with ones on the diagonal and zeros everywhere else.

That's the identity matrix.

It's the identity matrix.

And when you run the dot product calculation using the identity matrix, you just get the standard Cartesian dot product back.

The complexity disappears when it's not needed.

That is the ultimate proof of its universality.

It works in the simplest case and the most complex case.

Yes.

Let's spend a minute synthesizing this by walking through the seven crystal systems and seeing what their specific symmetries do to the metric tensor.

This is a great way to see it.

The symmetry dictates the structure of G.

We can group them by how empty or full the matrix is.

Okay, let's start with the most symmetrical, the orthogonal systems.

Cubic, tetragonal and orthohombic.

Right.

In all three of those, alpha, beta, gamma equals 90 degrees.

So all the cosine terms are zero.

Which means all the off -diagonal elements of the tensor are zero.

They're all diagonal matrices.

They are.

In cubic, A, B, C.

So the diagonal is A2, A2, A2.

For tetragonal.

A, B, but not equal to C.

So the diagonal is A2, A2, C2.

And orthohombic is the least symmetric of the orthogonal group.

A, B, and C are all different.

So the diagonal is A2, B2, C2.

The key thing here is that the orthogonality keeps all the complexity on that main diagonal.

Now we start to introduce the angle problems.

Which systems start to fill in those off -diagonal spots?

The next step down in symmetry is hexagonal, and also trigonal.

Here, alpha and beta are 90 degrees, but gamma is 120 degrees.

So cos 120 is not zero, it's negative one half.

Right.

So the terms that involve gamma, that's G12 and G21, they're suddenly not zero anymore.

They become abcos gamma.

The matrix now has non -zero terms off the diagonal, reflecting that skew in the AB plane.

Exactly.

Now for monoclinic, it handles its non -orthogonality differently.

How so?

In monoclinic, alpha and gamma are 90, but beta is the non -90 degree angle.

So now it's the terms related to beta that become non -zero.

Right.

That's G13 and G31.

They get populated with acos beta.

The geometry is skewed, but only in one specific plane.

Which leaves us with the final most complex system, triclinic.

The triclinic system is a wild west.

No restraints.

A, B, C are all different.

Alpha, beta, gamma are all different and not 90.

So since no angles are necessarily 90 degrees.

All six of the unique dot products are non -zero.

The metric tensor is fully populated.

It's a dense matrix, which perfectly reflects the total lack of symmetry in the triclinic unit cell.

I think seeing the matrix fill up like that as the symmetry breaks down is maybe the most illuminating way to understand this.

It really is.

It shows you mathematically how the Okay, so this is where the abstract algebra becomes concrete.

Let's walk through the logic of the book's worked examples.

This is the aha moment section.

You see the same formula work over and over again.

The only thing that changes is the G matrix you feed it.

Let's start with a tough distance calculation.

Finding the length of the main body diagonal.

The vector T equals 111 in a monoclinic crystal.

And the parameters are specific.

A3, B4, C6, and the angle beta is 120 degrees.

Okay, step one.

Step one is to build a specific metric tensor for this system.

It's monoclinic, so we know alpha and gamma are 90.

We only care about beta.

And cos 120 is negative one half.

Right, so the diagonal terms are easy.

A29, B216, C236.

And the off -diagonal terms related to beta?

That's G13 and G31.

They are a cos beta, which is 36, negative 0 .5, or negative 9.

So the tensor is built.

Now, if we were naive and just used the Pythagorean method for 111, we'd get D squared equals 9 plus 16 plus 36, which is 61.

So D would be the square root of 61, about 7 .81.

But that's wrong.

That's wrong.

Because when we use the proper formula, D2 equals TGT transpose,

the calculation automatically includes those negative nines as correction factors.

And the true length squared comes out to be 43.

Giving a final length of the square root of 43, which is about 6 .557.

So the obtuse angle actually made the diagonal shorter than it would be in an orthogonal box?

Exactly.

The 120 -degree angle pushes the A and C axes apart, shortening the diagonal.

The tensor correctly applied a negative correction, that minus 9 term, to shrink the distance.

It does all the heavy lifting.

Okay.

Next problem.

Calculating the distance between two atoms at some tricky fractional coordinates.

Yeah.

P is at 12, 13, 14, and Q is at 13, 12, 34.

And the crystal is to trigonal.

A2, B3, C5.

So the axes are orthogonal, but the lengths are different.

We still need G to handle the scaling.

First step is the difference vector, R equal QP.

You subtract the fractions, you get R, and it gets 16, 16, 12.

Step two, build the metric tensor.

It's to trigonal, so it's a diagonal matrix.

The diagonal terms are just A24, B29, and C225.

All the off -diagonals are zero.

And step three is to compute the length of R using the tensor formula.

And because G is diagonal, the math simplifies.

You just multiply the square of each component of R by the corresponding lattice parameter squared.

The source calculation gives a result for d squared of 89 over 36.

Which is a distance d of about 1 .572 length units.

This example is great because it shows that even with 90 -degree angles, the tensor is still essential for handling the different scaling along the axes.

Okay, we've done distance.

Now let's move to angles.

The problem is finding the angle between two bonds in a cubic crystal.

This is the simplest possible geometry.

We identify the two vectors, R1 and R2.

And because it's cubic, the metric tensor G is just the identity matrix?

Maybe scaled by squared?

It is.

So why do we even bother with the tensor here?

Because we want to show that the universal method works.

We use the general formula for the angle.

Cosine of theta equals the dot product of the vectors divided by the product of their lengths.

Okay.

If we plug our metric tensor definition into the dot product on top and into the length calculations on the bottom, all of the A squared scaling factors just cancel out.

And you're left with the standard Cartesian calculation.

Exactly.

The result is cos theta equals 12, which means the angle theta is 60 degrees.

It confirms the standard math, but it also shows the standard math as just a special case of the more general tensor method.

Right.

Now for the real asset test.

Find the angle between two directions, 101 and 1 bar, 0, 1 bar, and a monoclinic crystal.

This is a calculation that is just a nightmare to do without the metric tensor.

You have skewed axes and you're finding the angle between two vectors in that skewed space.

The parameters are A4, B6, C5, and beta is 120 degrees.

First, we build the monoclinic G.

Diagonal terms are A216, B236, C225.

And the key is that beta angle.

Cos 120 is negative one half.

So G13 and G31 are a cos beta, which is 45 negative 0 .5 or negative 10.

Then we use the universal angle formula.

We need three things, the dot product and the two vector lengths.

And we calculate all three using our new metric tensor.

The dot product 101G1 bar, 0, 1 bar transpose comes out to be three.

And the length?

The length squared of 101 is 21.

The length squared of 1 bar, 0, 1 bar is the 129.

And those numbers are completely dependent on those minus 10 terms in the G matrix.

So the final calculation is theta equals the arc sine of three divided by the product of the square root of 21 and the square root of 129.

Which gives you theta 88 .69 degrees.

So it's close to 90, but it's not 90.

And that's the crucial point.

A naive calculation would have gotten it wrong.

The metric tensor gives you that necessary sub -degree precision.

The chapter then offers a little refinement.

An alternative method that's more about programming efficiency than geometry.

This is the computational optimization layer.

Yeah.

If you're writing code, you want to minimize operations.

The way we just did it required three separate matrix multiplications.

One for the dot product, two for the lengths.

Right.

The alternative method streamlines it.

You take your two vectors, R1 and R2, and you stack them into a single 2 by 3 matrix.

By performing a single, slightly larger matrix multiplication,

you get a 2 by 2 matrix as a result.

And that little matrix contains all three numbers you need.

That's clever.

The diagonal of that result gives you the two lengths squared.

And the off -diagonal terms.

They give you the dot product.

Yeah.

So you do one calculation and get everything you need in one go.

It's just a much more efficient way to program it.

Okay.

We've established this is the universal key.

But to really appreciate it, you have to look at the alternative.

The alternative is, and I don't use this word lightly, horrifying.

The book includes tables showing the explicit formulas for length and angle for each of the seven crystal systems without using the metric tensor.

And if you look at the formula for the length of a vector in the triclinic system,

it's just insane.

It's a massive equation.

15 separate terms with all the squares of the components, all the lattice parameters, all the cosines of the angles.

It's a mess.

And if you use that method, you have to write code.

There's just a giant if -then statement.

If the crystal is monoclinic, then use this horrible formula.

If it's triclinic, then use this even worse formula.

Seven separate complex equations to write and debug.

Right.

The metric tensor approach throws all of that away.

And replaces it with?

One single matrix operation.

You define the geometry once in g, and then the calculation d2 equals our gr transpose does the rest.

That generality is why it's the preferred method for any modern computer program.

It's computationally superior, less prone to error, and just.

Fundamentally, more elegant.

It's a difference between building seven custom clunky machines and one universal machine that takes seven different inputs.

So all this modern computational power rests on centuries of foundational work.

Before we could do the math, people had to figure out that the geometry was even predictable.

That's right.

The concept of the lattice itself had to be discovered first.

Let's go back to the 17th century.

To Nicola Steno.

Steno's law.

The law of constancy of interfacial angles.

This was a huge deal.

Steno noticed that no matter how a quartz crystal grew or where it was from, the angles between the same corresponding faces were always identical.

That suggested some kind of immutable internal geometric design.

The first real hint that geometry dictates structure.

And that physical observation needed a mathematical model.

So we move into the 18th century with people like Torber and Olaf Bergman.

He started deducing crystal forms based on simple rational intercepts.

He was one of the first to say that crystal faces must cut the axes at ratios of small whole numbers.

Which sounds a lot like the precursor to Miller indices.

It's the direct intellectual foundation.

And building on Bergman's idea was René Gistoy.

Often called the father of modern crystallography.

He proposed that all crystals were built by stacking identical tiny little blocks, which he called Molecules Entegrantes.

That's our modern concept of the unit cell and the lattice.

The idea that a huge structure is just the endless repetition of one fundamental block.

Exactly.

Hoy formalized the idea that the regularity Steno saw had to come from a periodic internal structure.

The metric tensor, centuries later, is just the tool to measure the geometry of that single block that Hoy imagined.

While they were focused on geometry, others were classifying based on observable properties.

Yeah, people like Johann Karl Gayler who looked at color, cleavage, hardness.

And his work influenced Abraham Gottlob Werner, who helped move mineralogy into geology.

Though Werner's system was a bit too simple, wasn't it?

Just four primary forms.

It was, but the emphasis on systematic classification was the key.

And then figures like Weiss and Mohs refined those systems, moving away from simple shapes toward classification based on a system of several axes.

The same reference axes that the metric tensor uses today.

Exactly.

Without them developing the idea of ABC and alpha, beta, gamma,

the metric tensor would have nothing to describe.

It's the computational payoff for centuries of geometric insight.

So let's wrap this up.

Okay.

If you take one thing away from this deep dive, it should be that crystallographic calculations don't have to be a nightmare of seven different equations.

We moved beyond Pythagoras.

Discovered the six critical geometric relationships, stored them in the three by three metric tensor,

and learned how that one single matrix is the universal key for distance and angles in any of the seven crystal systems.

The complexity of the specific crystal just gets absorbed into the tensor's definition.

That's the intellectual elegance of it.

It turns a daunting task into a unified matrix operation.

And that leaves us with a final provocative thought for you to explore.

If you were building a material simulation program today in Python or MATLAB, which would you rather do?

Program one single universal function that just takes the metric tensor as its input?

Or?

Or try to program and debug seven individual monstrously complicated formulas, one for each crystal system.

The efficiency gap there isn't just mathematical.

It fundamentally dictates how computational crystallography is done today.

That difference is profound.

Thank you for joining us for this deep dive into crystallographic computations.

We hope this knowledge proves immediately useful.