Density of Final States

When a quantum system is perturbed, time-dependent perturbation theory hands you a transition amplitude to each individual final state. If the final states are discrete, the probability into any one of them oscillates in time and never settles into a rate. The physics changes qualitatively when the final states form a continuum — a particle escaping to infinity, an electron ejected into free space, a photon emitted into the radiation field. Then you must sum over a dense band, and the bookkeeping of how many states lie in each energy interval is the density of final states $\rho(E_f)=\mathrm{d}N/\mathrm{d}E$ . It is the factor that turns an oscillating probability into a steady decay rate, and it is the second ingredient of Fermi's Golden Rule. To make a continuum countable you place the system in a large box of side $L$ with periodic boundary conditions, so allowed wavevectors are quantised, $\mathbf k=(2\pi/L)(n_x,n_y,n_z)$ . Each state then fills a cell of volume $(2\pi/L)^d$ in $\mathbf k$ -space, and $\rho(E)$ follows by dividing the $\mathbf k$ -space shell volume by that cell and converting to energy through the dispersion relation. The one-dimensional version of this counting — $k_n=2\pi n/L$ , $L/2\pi$ states per unit $k$ — is owned by Free Particle and Plane Waves; this article extends it to two and three dimensions and to the radiation field, and proves the striking dimensionality dependence that results. That dimensionality dependence is the article's owned result. From the same counting, $\rho(E)\propto E^{(d-2)/2}$ : in 3D it rises as $\sqrt E$ from threshold, in 2D it is flat — completely energy-independent — and in 1D it diverges as $E^{-1/2}$ at the band edge (a van Hove singularity). Each behaviour is derived and numerically confirmed by brute state-counting. Finally, the fictitious box volume $V$ that appears in $\rho$ cancels exactly against the $1/V$ in box-normalised matrix elements, so every physical rate is $L$ -independent — a built-in consistency check that this article proves rather than merely asserts.

Key takeaways

The density of final states $\rho(E)=\mathrm{d}N/\mathrm{d}E$ counts quantum states per unit energy near $E$ ; it bridges a discrete-looking amplitude and a true continuum.
Count by box quantisation: each state fills $(2\pi/L)^d$ of $\mathbf k$ -space, so $\rho(E)$ follows from the dispersion relation. The 1D counting is owned by Free Particle and Plane Waves; this article extends it.
For a 3D free particle, $\rho(E)=\dfrac{V m k}{2\pi^2\hbar^2}\propto\sqrt E$ per spin — derived from $N(k)=Vk^3/6\pi^2$ .
Owned result: $\rho(E)\propto E^{(d-2)/2}$ — 1D diverges as $E^{-1/2}$ , 2D is constant, 3D rises as $\sqrt E$ ; each proven and checked by brute state-counting.
The box volume $V$ in $\rho$ cancels the $1/V$ from normalised plane-wave states, so the golden-rule rate is $L$ -independent — proven, not assumed.
Photons ( $E=\hbar ck$ ) give $\rho\propto\omega^2$ , the origin of the $\omega^3$ scaling of spontaneous emission.
Key idea: quantise momenta in a box — each state fills $(2\pi/L)^d$ of $\mathbf k$ -space — so $\rho(E)=\mathrm{d}N/\mathrm{d}E$ follows from the dispersion, scaling as $E^{(d-2)/2}$ ; this $\rho(E_f)$ converts an oscillating transition probability into a steady rate.

The core insight

Count states by quantising momenta in a box: each state fills $(2\pi/L)^d$ of $\mathbf k$ -space, so $\rho(E)=\mathrm{d}N/\mathrm{d}E$ follows directly from the dispersion relation $E(\mathbf k)$ . The result scales as $E^{(d-2)/2}$ — flat in 2D, $\sqrt E$ in 3D, divergent in 1D — and it is exactly the factor $\rho(E_f)$ that turns a first-order transition probability into a constant decay rate.

Where this lives today

The density of states is the currency in which transition and thermodynamic properties are quoted across physics. In condensed matter it sets electronic specific heat, magnetic susceptibility, and — through the golden rule — every scattering and recombination rate; its singular features (van Hove singularities, the flat 2D density, the 1D $E^{-1/2}$ spike) are read directly off tunnelling and photoemission spectra. In device engineering, band-structure design is density-of-states design: quantum wells (2D, step-edged) power semiconductor lasers, quantum wires and dots sharpen the density into spikes and delta functions for lower-threshold, narrower-linewidth emitters. In atomic physics the photon mode density gives the $\omega^3$ spontaneous-emission law and — when the mode density is reshaped by a cavity or photonic crystal — the Purcell effect, the deliberate speeding-up or suppression of emission by engineering $\rho(\omega)$ . Everywhere a quantum system decays into a continuum, the density of that continuum is half the answer.

The physical picture

Imagine throwing a quantum system at a target of final states. The transition rate depends not only on how strongly the perturbation couples initial and final states — the matrix element — but on how many final states are there to receive the system. A wide, dense band of receiving states drains probability quickly; a single isolated level just sloshes it back and forth. The density of states is the size of the catcher's mitt in energy. Picture $\mathbf k$ -space tiled by a uniform lattice of allowed momenta, one per cell of volume $(2\pi/L)^d$ . Counting the states in an energy shell is pure geometry: find the $\mathbf k$ -space volume between two constant-energy surfaces and divide by the cell volume. The dispersion relation $E(\mathbf k)$ sets how thick that shell is for a given $\mathrm{d}E$ — and that is where the dimensionality dependence is born. In $d$ dimensions the shell's surface area scales as $k^{d-1}$ , while the thickness $\mathrm{d}k$ for fixed $\mathrm{d}E$ scales as $1/k$ ; the product $\rho\propto k^{d-2}\propto E^{(d-2)/2}$ .

Figure

The density of states depends on dimensionality: ρ ∝ E^((d−2)/2) — 1D diverges, 2D is flat, 3D rises as √E

The box of side $L$ is a scaffold, not physics. It discretises the continuum so you can count, but every physical answer must be independent of $L$ . The volume $V$ that appears in $\rho$ is cancelled exactly by the $1/V$ from normalising the plane-wave final states $\psi=V^{-1/2}e^{i\mathbf k\cdot\mathbf r}$ — a consistency check that any golden-rule calculation carries built in. Different continua then give dramatically different physics from the same matrix element: massive particles give the $\sqrt E$ 3D growth, photons give $\omega^2$ , and reduced dimensionality sharpens thresholds into the step edges and spikes that engineers exploit in quantum wells and wires.

Figure

Box quantisation: allowed k-states tile k-space one per cell (2π/L)^d; the energy shell between k and k+dk is counted by geometry

The counting and its result

The density of states is built from one idea — each allowed momentum occupies a fixed cell of $\mathbf k$ -space — applied through the dispersion relation. We state the counting rule, then the 3D free-particle result, keeping every factor traceable. Box quantisation with periodic boundary conditions forces $\mathbf k=(2\pi/L)(n_x,n_y,n_z)$ , so adjacent allowed components differ by $\Delta k=2\pi/L$ and each state fills a cell of volume $(2\pi/L)^d$ .

\mathrm{d}N=\frac{V}{(2\pi)^3}\,\mathrm{d}^3k \qquad (1)

where: $\mathrm{d}N$ — number of states in the $\mathbf k$ -space element $\mathrm{d}^3k$ ; $V=L^3$ — box volume; $(2\pi/L)^3=(2\pi)^3/V$ — the $\mathbf k$ -space cell per state. The number of states is the available volume divided by the cell volume. Converting to a density in energy uses the spherical shell $4\pi k^2\,\mathrm{d}k$ and the dispersion relation $E=\hbar^2k^2/2m$ , which supplies the inverse group velocity $\mathrm{d}k/\mathrm{d}E$ .

\rho(E)=\frac{V}{(2\pi)^3}\,4\pi k^2\,\frac{\mathrm{d}k}{\mathrm{d}E}=\frac{V\,m\,k}{2\pi^2\hbar^2},\qquad k=\frac{\sqrt{2mE}}{\hbar} \qquad (2)

where: $4\pi k^2$ — the area of the constant-energy sphere in 3D; $\mathrm{d}k/\mathrm{d}E=m/\hbar^2k$ — from differentiating the dispersion; $\rho(E)$ — states per unit energy (per spin). The $\sqrt E$ growth is visible once $k=\sqrt{2mE}/\hbar$ is substituted. This $\rho(E_f)$ is precisely the factor in Fermi's Golden Rule; we write the rate to fix where $\rho$ appears, without re-deriving the rule (owned by Fermi's Golden Rule).

\Gamma_{i\to f}=\frac{2\pi}{\hbar}\,|\langle f|\hat V|i\rangle|^2\,\rho(E_f) \qquad (3)

where: $\rho(E_f)$ — the density of final states on the energy shell $E_f=E_i$ (or $E_i\pm\hbar\omega$ ); $\langle f|\hat V|i\rangle$ — matrix element between box-normalised states, carrying the compensating $1/V$ .

Derivations

Derivation AThe 3D free-particle density of states from state counting

The goal of this card is to derive $\rho(E)=Vmk/2\pi^2\hbar^2$ from first principles — counting allowed wavevectors in a spherical $\mathbf k$ -space shell and converting to energy — so that every factor of $2\pi$ , $V$ , and $\hbar$ is accounted for and can be checked. This is the template the dimensionality theorem then generalises.

Step 1. Count all states with wavevector magnitude below $k$ : the sphere of radius $k$ has volume $\tfrac43\pi k^3$ , and each state occupies $(2\pi/L)^3=(2\pi)^3/V$ , so

N(k)=\frac{\tfrac43\pi k^3}{(2\pi)^3/V}=\frac{V k^3}{6\pi^2}.

Why: the cumulative count is just the sphere volume divided by the per-state cell — pure geometry.

Step 2. Differentiate to get states per unit $k$ :

\frac{\mathrm{d}N}{\mathrm{d}k}=\frac{V k^2}{2\pi^2}.

Why: this is the surface density of states on the sphere of radius $k$ (the $4\pi k^2$ area with the $(2\pi)^3/V$ cell).

Step 3. Convert to energy with the dispersion $E=\hbar^2k^2/2m\Rightarrow \mathrm{d}k/\mathrm{d}E=m/\hbar^2k$ :

\rho(E)=\frac{\mathrm{d}N}{\mathrm{d}k}\,\frac{\mathrm{d}k}{\mathrm{d}E}=\frac{V k^2}{2\pi^2}\cdot\frac{m}{\hbar^2k}=\frac{V m k}{2\pi^2\hbar^2}.

Why: the chain rule turns a density in $k$ into a density in $E$ ; the $1/k$ from $\mathrm{d}k/\mathrm{d}E$ is the inverse group velocity.

Step 4. Express in energy: substituting $k=\sqrt{2mE}/\hbar$ ,

\rho(E)=\frac{V}{4\pi^2}\left(\frac{2m}{\hbar^2}\right)^{3/2}\sqrt E.

Why: this exhibits the characteristic $\sqrt E$ growth from threshold; multiply by 2 if both spin states are accessible. Verified: verify_aq-qm-density-of-final-states.py — the symbolic chain $N(k)\to\mathrm{d}N/\mathrm{d}k\to\rho(E)$ reproduces $Vmk/2\pi^2\hbar^2$ and its $\sqrt E$ form exactly; a brute count of periodic-BC states in a 20 nm box matches $V(2mE/\hbar^2)^{3/2}/6\pi^2$ to better than 1% at 0.5–2 eV.

Go deeper · Derivation B · 1 minThe dimensionality theorem: ρ(E) ∝ E^((d−2)/2)

Derivation BThe dimensionality theorem: ρ(E) ∝ E^((d−2)/2)

This is the article's owned structural result. The same counting run in $d$ dimensions produces a strikingly different energy dependence in each: divergent in 1D, flat in 2D, rising in 3D. The goal here is to derive the single formula $\rho\propto E^{(d-2)/2}$ that unifies all three, and to confirm it by direct state-counting — the 2D constant in particular is the fingerprint that engineers exploit in quantum wells.

Step 1. In $d$ dimensions the cumulative count scales with the volume of a $d$ -ball, $N(k)\propto k^d$ , so the states per unit $k$ scale as

\frac{\mathrm{d}N}{\mathrm{d}k}\propto k^{d-1}.

Why: the surface of a $d$ -ball has "area" $\propto k^{d-1}$ — a length in 1D (the two points $\pm k$ ), a circumference in 2D, a sphere in 3D.

Step 2. The dispersion contributes the same inverse group velocity in every dimension,

\frac{\mathrm{d}k}{\mathrm{d}E}=\frac{m}{\hbar^2 k}\propto k^{-1}.

Why: $E=\hbar^2k^2/2m$ is isotropic; only the shell geometry knows about $d$ .

Step 3. Multiply:

\rho(E)\propto k^{d-1}\cdot k^{-1}=k^{d-2}\propto E^{(d-2)/2}.

Why: the exponents combine; substituting $k\propto\sqrt E$ gives the energy power directly.

Step 4. Read off the three cases, and note the 2D constant explicitly:

\rho_{1D}\propto E^{-1/2}\ (\text{diverges}),\quad \rho_{2D}=\frac{A m}{2\pi\hbar^2}\ (\text{constant}),\quad \rho_{3D}\propto E^{+1/2}\ (\text{rises}).

Why: $d=2$ makes the exponent vanish — the shell area grows as $k$ exactly as fast as the thickness shrinks, leaving a flat, energy-independent density. This is the van Hove-free, step-edged density of a 2D electron gas. Verified: verify_aq-qm-density-of-final-states.py — for each $d$ the ratio $\rho(4E)/\rho(E)$ equals $4^{(d-2)/2}$ symbolically; the 2D density is confirmed $E$ -independent ( $\partial\rho_{2D}/\partial E=0$ ) and equal to $Am/2\pi\hbar^2$ ; the 3D $N(E)\propto E^{3/2}$ exponent recovered by a log-log fit to a brute state-count (slope $1.500$ ).

Limiting cases

The energy dependence of $\rho(E)$ follows entirely from dimensionality and the dispersion relation.

Case	Result	Meaning
3D massive particle	$\rho(E)=\dfrac{V m k}{2\pi^2\hbar^2}\propto\sqrt E$	Free electron in a box; grows from zero at threshold. Multiply by 2 for spin.
2D massive particle	$\rho(E)=\dfrac{A m}{2\pi\hbar^2}$ (constant)	Energy-independent — the hallmark of 2D systems (quantum wells, surfaces); step-edged absorption.
1D massive particle	$\rho(E)=\dfrac{L}{\pi\hbar}\sqrt{\dfrac{m}{2E}}\propto E^{-1/2}$	Diverges at the band edge — a van Hove singularity; sharp threshold peaks in quantum wires.
Photons (per polarisation)	$\rho(E)=\dfrac{V\omega^2}{2\pi^2\hbar c^3}\propto\omega^2$	Linear dispersion $E=\hbar\omega=\hbar ck$ ; the $\omega^2$ (times $\omega$ from the coupling) gives the $\omega^3$ spontaneous-emission law. Times 2 for two polarisations.

Common trapDo not quote ρ without saying "per what," and always evaluate it on the final energy shell

Two mistakes recur. First, $\rho(E)$ carries the fictitious box volume $V$ and is only meaningful because box-normalised matrix elements carry a compensating $1/V$ ; quoting $\rho$ without specifying the normalisation (per spin? per solid angle? total?) invites spurious factors — and forgetting the $V$ -cancellation makes the "rate" depend on an arbitrary box size, which is nonsense. Second, the relevant density is that of the final states, evaluated on the energy shell $E_f=E_i$ (constant perturbation) or $E_f=E_i\pm\hbar\omega$ (harmonic), not the initial states and not at some other energy. Getting either wrong corrupts every golden-rule rate downstream.

Worked examples

Example 1 — an electron in 1 cm³ at 1 eV: compute the actual density of states.

Step 1. Wavevector on shell: $k=\sqrt{2m_e E}/\hbar$ with $E=1\,\text{eV}=1.602\times10^{-19}$ J gives $k=\sqrt{2(9.109\times10^{-31})(1.602\times10^{-19})}/(1.055\times10^{-34})=5.12\times10^{9}$ m $^{-1}$ .

Step 2. Box volume $V=(10^{-2}\,\text{m})^3=10^{-6}$ m³.

Step 3. Apply the 3D formula: $\rho(E)=\dfrac{V m_e k}{2\pi^2\hbar^2}=\dfrac{(10^{-6})(9.109\times10^{-31})(5.12\times10^{9})}{2\pi^2(1.055\times10^{-34})^2}=\mathbf{2.13\times10^{40}}$ states/J.

Step 4. Convert to per-eV: multiply by $1.602\times10^{-19}$ J/eV to get $\rho=\mathbf{3.41\times10^{21}}$ states/eV — an enormous number, which is exactly why a macroscopic box behaves as a true continuum.

Step 5. Check the $L^3$ scaling: doubling the box side to 2 cm multiplies $V$ by 8 and hence $\rho$ by exactly 8 — the density scales with volume, as it must, and this $V$ later cancels in any physical rate.

Example 2 — the photon density of states and the $\omega^3$ emission law.

Step 1. Photons obey the linear dispersion $\omega=ck$ , so $k=\omega/c$ and $\mathrm{d}k=\mathrm{d}\omega/c$ .

Step 2. Count modes including the factor 2 for two transverse polarisations: $\mathrm{d}N=2\cdot\dfrac{V}{(2\pi)^3}\,4\pi k^2\,\mathrm{d}k$ .

Step 3. Substitute $k=\omega/c$ : $\rho(\omega)=\dfrac{V\omega^2}{\pi^2c^3}$ — the density grows as $\omega^2$ .

Step 4. Feed into the golden rule for spontaneous emission: the rate $A\propto|\mathbf d_{fi}|^2\rho(\omega)$ carries $\omega^2$ from the mode density and one more $\omega$ from the dipole coupling squared, giving $A\propto\omega^3|\mathbf d_{fi}|^2$ — the reason a 10 eV optical transition decays in nanoseconds while the 5.9 μeV hydrogen 21-cm transition takes 11 million years. The $\omega^3$ is entirely a density-of-states statement about the photon continuum.

Who built the counting

Origins · 1 minDensity of States and how this was discovered

Density of States

1912–1928 · Göttingen and Munich

The counting of states in a box is one of the quiet workhorses of modern physics, and its lineage runs through the founders of solid-state theory. The periodic boundary condition that makes the counting clean — treat a finite box as if it repeated forever, so allowed wavevectors are $\mathbf k=(2\pi/L)\mathbf n$ — is the Born–von Kármán condition, introduced by Max Born and Theodore von Kármán in 1912 in their theory of the specific heat of crystal lattices. Applying the same counting to the free-electron gas, Arnold Sommerfeld in 1927–28 produced the $\sqrt E$ density of states that rescued the Drude model and explained why only electrons near the Fermi surface contribute to heat capacity. The density of states is thus older than the golden rule that made it famous in transition theory: it began as a way to count lattice vibrations and electron states, and only later became the universal "how many destinations" factor in every rate calculation. Its dimensionality dependence — the flat 2D density in particular — became a design principle when semiconductor heterostructures made two- and one-dimensional electron gases real.

History

1912

— Born and von Kármán introduce periodic boundary conditions for lattice modes

1927

— Sommerfeld applies free-electron state counting to metals, deriving the √E density

1928

— the Sommerfeld model explains the small electronic specific heat via the Fermi surface

1970

— semiconductor heterostructures realise 2D electron gases with the flat density of states

1974

— quantum-well lasers exploit the step-edged 2D density in device design

Going further

References. Sakurai & Napolitano, Modern Quantum Mechanics, 2nd ed., §5.7 · Griffiths & Schroeter, Introduction to Quantum Mechanics, 3rd ed., Ch. 11 · Cohen-Tannoudji, Diu & Laloë, Quantum Mechanics, Vol. II, Ch. XIII · Shankar, Principles of Quantum Mechanics, 2nd ed., Ch. 18 · Ashcroft & Mermin, Solid State Physics, Ch. 2 (free-electron density of states).

Provenance. All counting and density-of-states formulas verified against Sakurai & Napolitano §5.7, Griffiths Ch. 11, Cohen-Tannoudji Vol. II Ch. XIII, and Ashcroft-Mermin for the 3D DOS. A verifier should recompute: (1) $\rho_{3D}=Vmk/2\pi^2\hbar^2$ from $N(k)=Vk^3/6\pi^2$ ; (2) the dimensional cases (2D constant, 1D $E^{-1/2}$ ); (3) the photon DOS $\rho(\omega)=V\omega^2/\pi^2c^3$ including the factor of 2 for polarization; (4) the Golden Rule prefactor $2\pi/\hbar$ via the sinc-squared-to-delta limit. All numerical structure is standard textbook content; no fabricated constants.

Builds on: Time-Dependent Perturbation Theory: General Formalism · First-Order Transition Amplitudes and Probabilities · Constant Perturbation Switched On Leads to: Fermi's Golden Rule · Spontaneous Emission and the Einstein A Coefficient · Absorption and Stimulated Emission · Semiclassical Theory of Atom–Radiation Interaction

Why is the density of states "per unit energy" and not "per unit k"?

Because transition rates and thermodynamic sums are naturally organised by energy — the golden rule's delta function conserves energy, and the Fermi function depends on energy. You can work with $\mathrm{d}N/\mathrm{d}k$ , but you always end up converting to $\mathrm{d}N/\mathrm{d}E=(\mathrm{d}N/\mathrm{d}k)(\mathrm{d}k/\mathrm{d}E)$ . The $\mathrm{d}k/\mathrm{d}E=1/(\hbar v_g)$ factor is the inverse group velocity, which is exactly where the dispersion relation enters and where the dimensionality dependence gets its extra $1/k$ .

Why does the 2D density of states come out constant — that seems too clean?

It is genuinely clean, and it is a consequence of two effects cancelling. In 2D the shell "area" is a circumference $2\pi k$ , which grows linearly with $k$ ; the shell thickness for fixed $\mathrm{d}E$ is $\mathrm{d}k=\mathrm{d}E\,m/\hbar^2k$ , which shrinks as $1/k$ . The product is $k$ -independent. The exponent $(d-2)/2$ is exactly zero at $d=2$ . This flatness is why 2D electron-gas absorption spectra have sharp steps rather than the smooth $\sqrt E$ onsets of 3D — the design basis of quantum-well lasers.

Does the density of states depend on the shape of the box, or just its volume?

Only the volume, in the limit of a large box. Different boundary conditions (periodic vs hard-wall) and different box shapes change the discrete level positions and small finite-size corrections, but the leading density $\rho(E)$ depends only on $V$ (and the dispersion and dimensionality). This is why the fictitious $V$ can appear at all and still cancel: it is a genuine physical volume, and any shape-dependence is a subleading $1/L$ correction that vanishes for macroscopic samples.

How is this different from the density of states in statistical mechanics?

It is the same object. The $\rho(E)$ that counts final states in a transition rate is identical to the $g(E)$ that appears in the partition function and the Fermi/Bose occupation sums — both count quantum states per unit energy. The name "density of final states" just emphasises the role it plays in a transition: it is the density of the states being decayed into. In equilibrium statistical mechanics you weight the same $\rho(E)$ by a Boltzmann or Fermi factor instead of a golden-rule matrix element.

Why must I include the factor of 2 for photon polarisations but sometimes not for electron spin?

Include a degeneracy factor for every internal degree of freedom the final states actually access. Photons always come in two transverse polarisations, so the photon mode density carries a factor 2. For electrons, whether you include the spin factor 2 depends on whether the perturbation flips or preserves spin and whether both spin final states are available: a spin-independent perturbation into a spin-degenerate band gets the 2; a spin-selective transition does not. The rule is: count exactly the final states the transition can reach.

Where does the density of states feed into scattering cross sections?

Keeping the angular factor, $\rho(E,\Omega)=(V/(2\pi)^3)k^2(\mathrm{d}k/\mathrm{d}E)\,\mathrm{d}\Omega$ , the golden-rule differential rate divided by the incident flux gives the first Born differential cross section $\mathrm{d}\sigma/\mathrm{d}\Omega=(m/2\pi\hbar^2)^2|\tilde V(\mathbf q)|^2$ . The density of states supplies exactly the angular and magnitude factors; the full scattering treatment (partial waves, the optical theorem) is developed in the scattering chapter, which consumes this angle-resolved density by name.