Boo-AI — Master Artificial Intelligence by Building from Scratch

Learning Objectives

By the end of this section you should be able to:

Explain, in plain words, why a periodic crystal naturally lives in two spaces — the familiar real space of atom positions, and a second, inseparable "dual" space called reciprocal space.
See that a plane wave $e^{i\mathbf{k}\cdot\mathbf{r}}$ is the only kind of function whose "shape" survives the lattice's translation symmetry, and recognise $\mathbf{k}$ as a wavevector with units of inverse length.
Identify special wavevectors — the reciprocal lattice vectors $\mathbf{G}$ — for which the wave is invisible to the lattice (returns the same value at every atom).
Connect three apparently different things — the Fourier decomposition of a periodic function, the Bragg spots in an X-ray diffraction pattern, and the k-points demanded by every VASP input — as facets of the same reciprocal-space picture.
State Bloch's theorem in one sentence and feel why it makes electronic structure of crystals computable in the first place.

One-line preview: a crystal's real-space picture tells you where the atoms are; its reciprocal-space picture tells you which waves the lattice allows. Every electronic, vibrational, and diffraction property of the crystal lives in that second picture.

The Story So Far — and Where It Hits a Wall

In Chapters 1 and 2 we built the real-space toolkit: lattice vectors $\mathbf{a}_1, \mathbf{a}_2, \mathbf{a}_3$ , fractional coordinates, the 14 Bravais lattices, and the symmetry groups that organise them. That picture is concrete. You can draw it. You can hold a model of it. You can write a POSCAR for it.

But the moment we try to ask the questions a materials scientist actually cares about — what colour is this crystal? does it conduct? what diffraction pattern does it produce? what are its phonon frequencies? — the real-space picture starts to feel oddly cumbersome. Try this thought experiment.

A thought experiment

You are an electron sitting somewhere in an infinite, perfect crystal. The potential $V(\mathbf{r})$ you feel from the ions is identical at every lattice point: shifting your position by a lattice vector $\mathbf{R}$ gives you the exact same environment. So how should you label your own quantum state? "The electron at position $\mathbf{r}$ " is a useless label, because every site looks the same. There must be a different, more economical label — one that respects the periodicity instead of fighting it.

Real space is the language of positions. Crystals, however, are objects defined by translation symmetry. To describe a crystal in its native language we need variables that talk about repeating patterns: spatial frequencies, wavelengths, and periodicities. That language is reciprocal space.

Periodicity Wants to Live in Frequency

Here is an analogy that almost every physicist uses privately, even if textbooks rarely state it: real space is to reciprocal space what time is to frequency in music.

A piano chord, played and held, is a function of time — $p(t)$ , the air pressure at your eardrum. You could describe that chord by its waveform: a wiggly curve. But that curve looks like chaos. The natural description, the one your brain actually uses, is which notes are sounding and how loud each one is. That second description lives in frequency space — the Fourier transform of $p(t)$ .

Now read that sentence again with one substitution. Replace "time" with "position", "chord" with "crystal", and "frequency" with "wavevector":

The natural description of a crystal is not the wiggly arrangement of atoms in space, but which spatial frequencies are present.

A spatial frequency has units of $\text{length}^{-1}$ — it counts how many oscillations of some quantity occur per metre. It is exactly the wavevector $k = 2\pi / \lambda$ from your introductory wave physics, promoted to a 3-vector $\mathbf{k}$ in three dimensions. Reciprocal space is the space whose points are wavevectors.

A first piece of vocabulary

Real space is measured in metres (or angstroms). Reciprocal space is measured in inverse metres (or $\text{Å}^{-1}$ ). The two are mathematical duals: a long wavelength in real space corresponds to a small $|\mathbf{k}|$ in reciprocal space, and a tightly packed lattice in real space corresponds to a widely spread one in reciprocal space. Hold on to that inverse — we will see it visually in a moment.

Plane Waves: The Crystal's Native Language

Why do wavevectors matter especially in crystals, and not, say, in a glass of water? Because of one beautiful mathematical fact about plane waves and translations. Take a plane wave

\psi_{\mathbf{k}}(\mathbf{r}) \;=\; e^{i\,\mathbf{k}\cdot\mathbf{r}}

and translate it by an arbitrary lattice vector $\mathbf{R}$ . What do you get?

\psi_{\mathbf{k}}(\mathbf{r}+\mathbf{R}) \;=\; e^{i\,\mathbf{k}\cdot(\mathbf{r}+\mathbf{R})} \;=\; e^{i\,\mathbf{k}\cdot\mathbf{R}}\,\psi_{\mathbf{k}}(\mathbf{r})

The wave comes back to itself, multiplied by a single complex phase $e^{i\mathbf{k}\cdot\mathbf{R}}$ . Its shape is unchanged. Plane waves are the only functions on the whole of three-dimensional space with that property — they are the eigenfunctions of every translation. Said differently, if a problem has translation symmetry, plane waves are the basis the problem secretly wants you to use.

That sentence is short, but it does a lot of work. Try it with your hands below. Slide $k$ and watch the wavelength change. Slide $r$ to move the cyan probe along the wave. Slide $R$ to translate the amber probe away from $r$ by exactly that lattice vector. The two phasors on the right are the values $\psi(\mathbf{r})$ and $\psi(\mathbf{r}+\mathbf{R})$ drawn as unit complex numbers — same length, rotated apart by the angle $k\cdot R$ .

Translate $\psi$ by $R$ : shape stays, phase rotates

drag

k

r

R

Cyan curve is the real part $\cos(2\pi k x)$ ; dashed amber is the imaginary part $\sin(2\pi k x)$ . The cyan dot sits at $x = r$ ; the amber dot at $x = r + R$ . On the right, the two phasors plot $\psi(r)$ and $\psi(r + R)$ as unit complex numbers — same length, rotated by exactly $k \cdot R$ .

k

k = 0.60 \cdot (2\pi / a)

r

r = 1.40\,a

position of cyan dot

R

R = 2.00\,a

\psi(r)

0.54 − 0.84 i

|\psi(r)| = 1

\psi(r + R)

0.97 + 0.25 i

|\psi(r + R)| = 1

phase factor

e^{i k R}

0.31 + 0.95 i

|e^{i k R}| = 1

Same shape, rotated phase.

\psi(r + R) = e^{i k R} \cdot \psi(r)

: both phasors stay on the unit circle — only their orientation differs by

k \cdot R

. The wave's shape never changes under any translation, no matter what

k

R

you pick. That is why plane waves are the eigenfunctions of every translation.

What to notice

Whatever you do with the sliders, both phasors stay on the unit circle. The translation can never stretch or shrink the wave — only rotate its phase. That is what "shape unchanged" means. And when $k\cdot R$ happens to be a multiple of $2\pi$ (try the snap buttons $k = G,\,2G,\,3G$ together with $R = 1a,\,2a,\,3a$ ), the two phasors lock on top of each other. Those are the special wavevectors $\mathbf{G}$ we are about to name.

Why this matters for crystals

A crystal's Hamiltonian commutes with every lattice translation. By the same logic that makes $\sin$ and $\cos$ the right basis for translation-invariant problems on a circle, plane waves $e^{i\mathbf{k}\cdot\mathbf{r}}$ are the right basis for crystals. The wavevector $\mathbf{k}$ becomes the natural quantum number — and the natural variable on which every observable depends.

That sentence is the entire reason reciprocal space exists. It is a 3-D space whose every point $\mathbf{k}$ labels one plane wave — and therefore, in a crystal, one allowed quantum state per band.

Interactive — When Does a Wave Fit a Lattice?

Time to feel this with your hands. The widget below shows nine atoms on a 1D lattice, spaced by lattice constant $a$ , and a plane wave $\cos(k\,x)$ drawn in cyan. The dots on the wave are its values sampled at the atoms. That is what the lattice "sees": not the whole continuous wave, but only its values on the periodic grid.

Drag the slider through k. Watch the bottom plot — it shows the phase coherence $C(k) = \tfrac{1}{N}\sum_n \cos(k\,n\,a)$ , which averages the wave's value across the atoms. Notice the dramatic peaks at $k = 0,\; 2\pi/a,\; 4\pi/a,\; \dots$ . These are the special wavevectors at which the wave returns to exactly the same value at every atom. To the lattice, the wave at $k = 2\pi/a$ is indistinguishable from the wave at $k = 0$ .

When does a plane wave fit a 1D lattice?

drag the slider

Blue dots are atoms spaced by lattice constant a. The cyan curve is the plane wave cos(k·x). The little dots show the wave's value sampled at each atom — that's what the lattice "sees."

Phase coherence C(k) = (1/N) Σ cos(k·xₙ)C = -0.009

Wavevector k = 0.320 · (2π/a)

snap to:

For this k the wave samples a different value at each atom. Push the slider until the cyan dots line up at the same height — those special k's are integer multiples of 2π/a, and they form the reciprocal lattice you'll meet next section.

Try the snap buttons. At every $G_m = m\,(2\pi/a)$ the cyan dots line up at the same height; the wave is, from the atoms' point of view, a constant. Between those values the wave is genuinely "new" — it modulates the atomic environment differently from any of the matched waves.

Try this experiment

Compare $k = 0.1\,(2\pi/a)$ with $k = 1.1\,(2\pi/a)$ . The phase coherence is identical. Why? Because $\cos((1+\delta)(2\pi/a)\,n a) = \cos(2\pi n + \delta\,2\pi n) = \cos(\delta\,2\pi n)$ — shifting $k$ by a reciprocal-lattice vector $G = 2\pi/a$ leaves the lattice-sampled values unchanged. This is the seed of the Brillouin zone: only the remainder of $k$ modulo $G$ matters physically.

What Just Happened? The Reciprocal Lattice in Disguise

You just discovered the reciprocal lattice without anyone telling you the definition. The set of wavevectors $\{G_m = m\,(2\pi/a)\}$ on which the plane wave is invariant under lattice translations — that set is the 1-D reciprocal lattice. In 3-D the same idea will give us three reciprocal basis vectors $\mathbf{b}_1, \mathbf{b}_2, \mathbf{b}_3$ (next section), and the reciprocal lattice will be $\mathbf{G} = h\mathbf{b}_1 + k\mathbf{b}_2 + \ell\mathbf{b}_3$ for integers $h, k, \ell$ .

Three statements that all say the same thing — the language a working physicist uses interchangeably:

Statement	What it really means
G·R is an integer multiple of 2π for every R	G is a reciprocal lattice vector
The plane wave eᶦᴳ·ʳ has the lattice's full translation symmetry	G is a reciprocal lattice vector
Adding G to k changes nothing measurable about the wave at the atoms	G is a reciprocal lattice vector
Any periodic function f(r+R) = f(r) can be Fourier-expanded using only G's	G is a reciprocal lattice vector

The reciprocal lattice is, in this sense, just the bookkeeping that says "these are the spatial frequencies the crystal admits." Every other spatial frequency is forbidden by the lattice symmetry — or rather, every other frequency can be reduced, modulo a $\mathbf{G}$ , into the small region called the first Brillouin zone. We will build that zone explicitly in Section 3.

Diffraction Photographs Reciprocal Space

Here is the moment where reciprocal space stops feeling abstract. When you fire a beam of X-rays, electrons, or neutrons at a crystal and put a detector on the far side, the spotty pattern that lights up is — almost literally — a photograph of the reciprocal lattice.

This is not a coincidence. The Laue condition for constructive interference between scattered waves is

\Delta\mathbf{k} \;=\; \mathbf{k}_{\text{out}} - \mathbf{k}_{\text{in}} \;=\; \mathbf{G}

which says: the only momentum changes a perfect crystal can give to a scattered wave are exactly the reciprocal lattice vectors. Each bright spot on the detector corresponds to one $\mathbf{G}$ . Move the spots, and you have moved the reciprocal lattice. We derive this rigorously in Section 6, but the picture below already lets you feel it.

Real space ↔ diffraction pattern

slide the spacing

Left: a 2D crystal with lattice constant a. Right: what an X-ray diffractometer would record. The diffraction pattern is a picture of the reciprocal lattice. Notice how shrinking a spreads the diffraction spots apart — and vice versa.

Real space (atoms)a = 36 px

Diffraction pattern (≈ reciprocal lattice)spacing ∝ 1/a

Lattice constant a = 36 px

Big atoms-far-apart in real space ⟶ tight bright-spot pattern in diffraction. Tiny atoms-close-together ⟶ widely separated spots. The two pictures are reciprocally linked — and reciprocal space is exactly where those bright spots live.

Slide the lattice spacing in the left panel and watch the right panel. As real space contracts, reciprocal space dilates — and vice versa. The two pictures are not different objects; they are the same object viewed through two opposite lenses. This inverse relationship is the visual signature of reciprocal space.

A historical aside

This insight, formulated by Max von Laue in 1912, won him the 1914 Nobel Prize in Physics. By 1915 William Henry Bragg and his son William Lawrence Bragg had turned diffraction patterns into the first technique for solving crystal structures atom-by-atom — and reciprocal space, until then a mathematical curiosity, became the working language of an entire field. Every protein structure in the PDB and every mineral identification by powder X-ray diffraction is a measurement in reciprocal space.

Bloch's Theorem in One Sentence

We will give Bloch's theorem its full proof in Section 5. For now, here is the one-sentence version that motivates everything we will compute in VASP:

Bloch's theorem. In any periodic potential, every electronic eigenstate can be written as a plane wave $e^{i\mathbf{k}\cdot\mathbf{r}}$ times a function with the same period as the lattice: $\psi_{n,\mathbf{k}}(\mathbf{r}) = e^{i\mathbf{k}\cdot\mathbf{r}}\,u_{n,\mathbf{k}}(\mathbf{r})$ with $u_{n,\mathbf{k}}(\mathbf{r}+\mathbf{R}) = u_{n,\mathbf{k}}(\mathbf{r})$ .

Read that and pause. It says: in a crystal, every electron state is labelled by two ingredients. First, a wavevector $\mathbf{k}$ — a point in reciprocal space, drawn from the first Brillouin zone. Second, a band index $n$ — an integer that counts which solution we are looking at. The whole infinite-position labyrinth of an infinite crystal collapses to a finite, computable space: $(n, \mathbf{k})$ .

This is the technical reason solid-state physics is computable. Without translation symmetry, an infinite system has infinitely many degrees of freedom. With it — thanks to plane waves and reciprocal space — the problem reduces to a finite calculation per $\mathbf{k}$ -point in the Brillouin zone, multiplied by however many bands you care to keep.

Why VASP Lives in Reciprocal Space

Open any VASP INCAR/KPOINTS pair and you can read this section in shorthand. The KPOINTS file does nothing in real space — it is, end to end, a description of how reciprocal space is sampled.

📝text

1# KPOINTS — sample the Brillouin zone with an 8x8x8 Monkhorst-Pack grid
2Automatic mesh
30
4Gamma
58 8 8
60 0 0

Three numbers, $8 \times 8 \times 8$ , completely specify how finely VASP integrates over reciprocal space when computing observables like energy and density of states. Underneath the hood:

Plane-wave basis. VASP expands each Bloch function as $u_{n,\mathbf{k}}(\mathbf{r}) = \sum_{\mathbf{G}} c_{n,\mathbf{k},\mathbf{G}}\, e^{i\mathbf{G}\cdot\mathbf{r}}$ — a sum over reciprocal lattice vectors. The cutoff ENCUT caps the magnitude $|\mathbf{k}+\mathbf{G}|^2$ , which is literally a radius in reciprocal space.
Brillouin-zone integration. Every total energy, force, and density of states is an integral over $\mathbf{k}$ . The KPOINTS grid is the discretisation of that integral, and convergence with respect to k-points is the first thing you check before trusting any VASP result (Section 8).
Band structures and DOS. A band structure plot is a slice of $E_n(\mathbf{k})$ along a path through the Brillouin zone. The path is chosen using high-symmetry points $\Gamma, X, L, K, \dots$ defined entirely in reciprocal space (Section 4).
Diffraction comparison. When you compute structure factors $F(\mathbf{G})$ from a relaxed VASP cell to compare with experiment, you are working entirely with reciprocal lattice vectors (Sections 6 and 7).

The take-away

Every input you set in VASP that has units of inverse length — KPOINTS, ENCUT, NGX/NGY/NGZ, the high-symmetry path in the band-structure plot — is a parameter in reciprocal space. If you understand reciprocal space, you have already understood half of the VASP user manual.

Looking Ahead — The Map of Chapter 3

Now that you know why reciprocal space is forced upon us by the physics of periodicity, the rest of the chapter builds the machinery:

Section	Topic	What it gives you
3.2	Reciprocal Lattice Vectors	The 3-D analogue of 2π/a: explicit b₁, b₂, b₃ from a₁, a₂, a₃
3.3	The First Brillouin Zone	The Wigner–Seitz cell of reciprocal space — the home of every k-point
3.4	High-Symmetry Points and Paths	Γ, X, L, K, M and the standard band-structure paths
3.5	Bloch's Theorem	The full proof that turned today's one-sentence preview into rigorous physics
3.6	Structure Factor and Diffraction	From atom positions to Bragg intensities — the formal Laue derivation
3.7	Systematic Absences	Why some allowed Bragg peaks have zero intensity — symmetry in reciprocal space
3.8	k-Point Grids and Convergence	How dense your KPOINTS file actually needs to be
3.9	Monkhorst–Pack Grids	The standard recipe VASP uses by default — and when to deviate from it
3.10	Reciprocal Space in VASP	Putting it all together: KPOINTS, ENCUT, IBZKPT, EIGENVAL — read with new eyes

Summary

A perfect crystal has translation symmetry, and translation symmetry forces us to label states by wavevectors $\mathbf{k}$ rather than positions $\mathbf{r}$ .
Plane waves $e^{i\mathbf{k}\cdot\mathbf{r}}$ are the unique functions whose shape survives every lattice translation. They are the natural basis for periodic problems, just as sines and cosines are the natural basis for time-periodic problems.
Among all $\mathbf{k}$ , a discrete set — the reciprocal lattice vectors $\mathbf{G}$ — leave the lattice-sampled wave unchanged. Two wavevectors that differ by a $\mathbf{G}$ describe the same physical state.
The diffraction pattern of a crystal is, almost literally, a photograph of its reciprocal lattice. Real-space contraction maps to reciprocal-space expansion; the two pictures are inverse views of the same object.
Bloch's theorem turns the infinite-atom problem into a finite calculation parametrised by $(n, \mathbf{k})$ over the first Brillouin zone. This is what makes VASP — and all of band-theory — possible.
Every reciprocal-space concept you meet from here on (KPOINTS grids, ENCUT, band-structure paths, structure factors, Brillouin zones) is a different facet of the same idea: periodicity wants to live in frequency.

Section 3.1 Core Insight

"Real space tells you where the atoms are. Reciprocal space tells you which waves the lattice allows. Every electronic, vibrational, and diffraction property of the crystal is a statement about that second space."

Coming next: Section 3.2 — Reciprocal Lattice Vectors — where we promote today's 1-D $G_m = 2\pi m / a$ to three dimensions and derive the explicit formulae $\mathbf{b}_i \cdot \mathbf{a}_j = 2\pi\,\delta_{ij}$ that every textbook starts with — but now you will know exactly why they had to look that way.