The Klein-Gordon equation

The Klein–Gordon equation is the first honest attempt to write a quantum wave equation that obeys special relativity, and its failures are as instructive as its successes. Schrödinger's equation is built on the non-relativistic energy $E = p^2/2m$ , which is first order in time and second order in space — an asymmetry that a Lorentz boost, mixing time and space, cannot preserve. The fix looks obvious: start instead from Einstein's relation $E^2 = p^2 + m^2$ and promote energy and momentum to the same operators, $\hat E = i\partial_t$ and $\hat p = -i\nabla$ . What drops out is $(\partial_\mu\partial^\mu + m^2)\psi = 0$ , an equation second order in both space and time and therefore manifestly invariant under boosts — the wave equation every relativistic field theory begins with. The price is paid immediately, in two currencies. Squaring the energy relation admits both roots, so plane waves carry energies $E = \pm\sqrt{p^2 + m^2}$ — a sea of negative-energy states that quantum completeness forbids you to discard. And because the equation is second order in time, the natural conserved density $\rho = i(\psi^*\partial_t\psi - \psi\partial_t\psi^*)$ evaluates to $2E|N|^2$ for a plane wave, negative whenever $E$ is, wrecking Born's reading of $\rho$ as a probability. These twin defects drove Dirac to hunt for a first-order equation in 1928. Yet Klein–Gordon was not wrong, only misread: Pauli and Weisskopf showed in 1934 that its "bad" density is a perfectly good charge density, and the equation took its permanent place as the law of every spin-0 field — from the pion to the Higgs boson.

Key takeaways

Built by feeding the relativistic relation $E^2 = p^2 + m^2$ into $\hat E = i\partial_t$ , $\hat p = -i\nabla$ , giving the covariant wave equation $(\partial_\mu\partial^\mu + m^2)\psi = 0$ .
Unlike Schrödinger's equation it is second order in both space and time, hence manifestly Lorentz invariant — the whole point of the exercise.
Plane waves $\psi = Ne^{-i(Et - \mathbf p\cdot\mathbf x)}$ obey $E = \pm\sqrt{p^2 + m^2}$ : a positive branch, a negative branch, and a forbidden mass gap of width $2m$ between them.
Its conserved density $\rho = i(\psi^*\partial_t\psi - \psi\partial_t\psi^*) = 2E|N|^2$ is not positive-definite, so it cannot be a probability — but it is a fine charge density (Pauli–Weisskopf, 1934).
A static massive source turns it into the Yukawa potential $\phi \propto e^{-mr}/r$ , whose range is the Compton wavelength $1/m$ — the physics that predicted the pion.
Key idea: forcing the relativistic $E^2 = p^2 + m^2$ onto a quantum wavefunction yields an equation second order in time, buying Lorentz invariance at the cost of negative-energy solutions and a sign-indefinite density — flaws that launched the Dirac equation but that quantum field theory later reads as the signature of antiparticles.

The core insight

Forcing the relativistic $E^2 = p^2 + m^2$ onto a quantum wavefunction yields an equation second order in time — manifestly Lorentz invariant, but carrying negative-energy solutions and a sign-indefinite density. Those "defects" are the first fingerprints of antiparticles, and the equation is the correct law for every spin-0 field.

Where this lives today

The Klein–Gordon equation stopped being a candidate single-particle theory and became infrastructure. Every scalar field in the Standard Model obeys it in the free limit, and its Fourier transform gives the scalar Feynman propagator $\tilde G(p) = i/(p^2 - m^2 + i\epsilon)$ — the pole at $p^2 = m^2$ is the mass-shell condition (1), drawn as an internal line in countless Feynman diagrams. The most consequential modern instance is the Higgs boson, discovered at the Large Hadron Collider in 2012: it is the Standard Model's only fundamental spin-0 particle, and its excitations are quanta of a Klein–Gordon field (with a self-interaction potential added). The pion, kaon, and every spin-0 meson are likewise Klein–Gordon fields, and Yukawa's $e^{-mr}/r$ remains the textbook model of the nuclear force. In cosmology the inflaton driving the early universe's expansion obeys a Klein–Gordon equation in curved spacetime, $\Box\phi + V'(\phi) = 0$ . What began as a rejected draft is now the default starting point for anything with spin zero.

The physical picture

The Klein–Gordon equation lives in two pictures at once. As a statement about a single free particle it is a spectrum: the energy–momentum relation $E^2 = p^2 + m^2$ is a hyperbola in the $(p, E)$ plane with two disconnected branches, an upper sheet at $E \ge +m$ and a lower sheet at $E \le -m$ , separated by a forbidden gap of width $2m$ that no continuous change of momentum can cross. As a statement about a field it is a wave equation with a mass term, and that mass term does something concrete and measurable: it gives the field a finite range. A massless field ( $m = 0$ ) reduces to the ordinary wave equation and spreads a $1/r$ Coulomb tail to infinity; a massive field screens itself into a short-ranged $e^{-mr}/r$ cloud. The first figure shows the spectrum with its mass gap; the second shows what the mass does to the reach of the force.

Figure

The Klein-Gordon spectrum E = ±√(p²+m²) plotted for a pion, m = 139.6 MeV. Two hyperbola branches — positive energy above +m and negative energy below −m — sandwich a forbidden mass gap of width 2m = 279 MeV that no free particle can occupy

The lower branch is not a drawing error and not something a boundary condition can throw away: a complete set of solutions to a second-order equation needs both roots, and the negative-energy waves refuse to vanish. For twenty years they looked like a catastrophe; the resolution, from quantum field theory, is that a negative-energy solution running one way is physically an antiparticle of positive energy running the other way — the mass gap $2m$ is exactly the energy to create a particle–antiparticle pair. The second figure turns to the field picture, where the mass term sets the distance over which a spin-0 field can carry a force.

Figure

A massive Klein-Gordon field screens itself. The massless case (m = 0) gives an infinite-range Coulomb tail ∝ 1/r; a pion-mass field gives the Yukawa potential ∝ e^(−mr)/r, cut off at the range ℏ/mc = 1.41 fm — the reach of the nuclear force

The core equations

Everything starts from the relativistic energy–momentum relation, written in natural units $\hbar = c = 1$ where energy, momentum, and mass share one unit. This is the relation the Klein–Gordon equation is engineered to satisfy operator-by-operator.

E^2 = \mathbf p^2 + m^2 \qquad (1)

where: $E$ — total relativistic energy; $\mathbf p$ — three-momentum; $m$ — rest mass. Reading it: unlike the non-relativistic $E = p^2/2m$ , energy appears squared, and that single fact — the presence of $E^2$ rather than $E$ — is the origin of both the equation's Lorentz invariance and its two pathologies. Promoting $E \to i\partial_t$ and $\mathbf p \to -i\nabla$ and letting both sides act on a wavefunction gives the Klein–Gordon equation. Collecting the four-dimensional derivatives into $\partial_\mu\partial^\mu \equiv \partial_t^2 - \nabla^2$ (the d'Alembertian $\Box$ , with the metric signature $+{-}{-}{-}$ ) makes its Lorentz invariance manifest.

\left(\partial_\mu\partial^\mu + m^2\right)\psi = 0 \quad\Longleftrightarrow\quad \left(\partial_t^2 - \nabla^2 + m^2\right)\psi = 0 \qquad (2)

where: $\partial_\mu\partial^\mu = \Box$ — the d'Alembertian, a Lorentz scalar; $m^2$ — a scalar; $\psi$ — a single complex scalar function. Reading it: the operator $\Box + m^2$ is built entirely from Lorentz scalars, so an observer in any inertial frame writes the same equation. It is second order in time — which will cost us a positive density. Feeding in a plane wave $\psi = N e^{-i(Et - \mathbf p\cdot\mathbf x)} = Ne^{-ip\cdot x}$ turns the differential equation back into the algebraic relation (1), but now with both signs of the square root surviving.

E = \pm\sqrt{\mathbf p^2 + m^2} \qquad (3)

where: the $+$ root is the ordinary positive-energy branch; the $-$ root is the negative-energy branch that cannot be discarded without spoiling completeness. Reading it: the two branches never touch — they are separated by the gap $|E| \ge m$ , and in field theory the lower branch is reinterpreted as antiparticles. Building a continuity equation the way one does for Schrödinger — multiply the equation by $\psi^*$ , its conjugate by $\psi$ , and subtract — yields a conserved four-current $\partial_\mu j^\mu = 0$ , whose density is the quantity that fails to be positive.

j^\mu = i\left(\psi^*\partial^\mu\psi - \psi\,\partial^\mu\psi^*\right), \qquad \rho = j^0 = i\left(\psi^*\partial_t\psi - \psi\,\partial_t\psi^*\right) = 2E|N|^2 \qquad (4)

where: $j^\mu = (\rho, \mathbf j)$ — the conserved four-current; $\rho$ — its time component. Reading it: for a plane wave $\rho = 2E|N|^2$ carries the sign of $E$ , so the negative-energy branch has $\rho < 0$ . It is not a probability — but it is (up to the charge $e$ ) a charge density, which is allowed to be negative. Finally, a static massive field sourced by a point charge $g$ obeys the time-independent Klein–Gordon equation, whose solution is the Yukawa potential — the massive cousin of the Coulomb potential.

(\nabla^2 - m^2)\phi = -g\,\delta^3(\mathbf r) \;\Longrightarrow\; \phi(r) = \frac{g}{4\pi}\,\frac{e^{-mr}}{r} \qquad (5)

where: $\phi$ — the static scalar potential; $g$ — the source strength (coupling); the exponential $e^{-mr}$ cuts the potential off beyond the range $1/m$ . Reading it: setting $m \to 0$ recovers the Coulomb $g/4\pi r$ ; keeping $m \ne 0$ gives a force that dies at the Compton wavelength — Yukawa's route to predicting the pion.

Derivation ABuilding the equation from the relativistic energy relation

The goal is to construct the Klein–Gordon equation from scratch, starting only from Einstein's $E^2 = p^2 + m^2$ and the same operator substitutions that produce the Schrödinger equation, and to show that the result is second order in time, manifestly Lorentz invariant, and free of the square root that would otherwise make a relativistic wave equation non-local. We will see exactly where the choice to square the energy relation enters, because that choice is the source of both the equation's virtues and its later troubles.

Step 1. Recall the operator dictionary that converts classical observables into quantum operators acting on a wavefunction: energy becomes $\hat E = i\partial_t$ and momentum becomes $\hat{\mathbf p} = -i\nabla$ (natural units $\hbar = 1$ ). These are the identical substitutions used to derive the Schrödinger equation. Why: the whole content of "quantizing" a classical relation is to reinterpret its symbols as these differential operators acting on $\psi$ .

Step 2. One might try to substitute directly into $E = \sqrt{p^2 + m^2}$ , giving $i\partial_t\psi = \sqrt{-\nabla^2 + m^2}\,\psi$ . But the square root of a differential operator is a power series in $\nabla^2$ of infinite order — it couples $\psi$ at a point to $\psi$ everywhere, a non-local monstrosity with no manifest Lorentz invariance. Why: this dead end is exactly why we go back and square first — locality forces us to keep $E^2$ , not $E$ .

Step 3. Instead substitute into the squared relation $E^2 = p^2 + m^2$ . The left side gives $\hat E^2\psi = (i\partial_t)^2\psi = -\partial_t^2\psi$ ; the right gives $(\hat{\mathbf p}^2 + m^2)\psi = (-\nabla^2 + m^2)\psi$ . Equating, $-\partial_t^2\psi = -\nabla^2\psi + m^2\psi$ , that is $(\partial_t^2 - \nabla^2 + m^2)\psi = 0$ . Why: squaring keeps every operator to second order, so the equation is a local partial differential equation — the price is that we have quietly admitted both signs of $E$ .

Step 4. Package the derivatives covariantly. With $\partial^\mu = (\partial_t, -\nabla)$ and $\partial_\mu = (\partial_t, +\nabla)$ , the contraction is $\partial_\mu\partial^\mu = \partial_t^2 - \nabla^2 \equiv \Box$ , a Lorentz scalar. The equation becomes $(\Box + m^2)\psi = 0$ , or $(\partial_\mu\partial^\mu + m^2)\psi = 0$ . Why: written this way, every ingredient is a scalar, so the equation is form-invariant under any Lorentz boost or rotation — precisely the property the Schrödinger equation lacks.

Step 5. Contrast the structure with Schrödinger's $i\partial_t\psi = -\nabla^2\psi/2m$ , which is first order in $\partial_t$ but second order in $\nabla$ . A boost rotates $t$ into $x$ ; an equation that treats them with different numbers of derivatives cannot survive that rotation. Klein–Gordon treats both to second order, so it does. Why: this asymmetry, not any small correction, is the fundamental reason the Schrödinger equation is non-relativistic — and fixing it is the entire motivation for equation (2).

Go deeper · Derivation B · 2 minThe conserved current that is a charge and not a probability

Derivation BThe conserved current that is a charge and not a probability

The aim is to derive the conserved four-current of the Klein–Gordon equation, evaluate its density on a plane wave, and show explicitly that this density is not positive-definite — the technical failure that seems to bar a probability interpretation. We follow the same algebra that gives Schrödinger's probability current, watch where the second time derivative changes the outcome, and arrive at the Pauli–Weisskopf reading of the density as electric charge.

Step 1. Write the Klein–Gordon equation and its complex conjugate: $(\Box + m^2)\psi = 0$ and $(\Box + m^2)\psi^* = 0$ . Multiply the first by $\psi^*$ and the second by $\psi$ : $\psi^*\Box\psi + m^2\psi^*\psi = 0$ and $\psi\,\Box\psi^* + m^2\psi\psi^* = 0$ . Why: this is the standard trick for extracting a conservation law — pairing the equation with its conjugate lets the mass terms cancel on subtraction.

Step 2. Subtract the second from the first. The identical $m^2\psi^*\psi$ terms cancel, leaving $\psi^*\Box\psi - \psi\,\Box\psi^* = 0$ . Why: the mass drops out cleanly, which already signals that whatever is conserved does not depend on $m$ — a hint it will be a generic charge, not something particle-specific.

Step 3. Recognize the left side as a total four-divergence: $\psi^*\Box\psi - \psi\Box\psi^* = \partial_\mu(\psi^*\partial^\mu\psi - \psi\,\partial^\mu\psi^*)$ , because the cross terms $\partial_\mu\psi^*\,\partial^\mu\psi$ cancel. Multiplying by $i$ to make the current real defines $j^\mu = i(\psi^*\partial^\mu\psi - \psi\,\partial^\mu\psi^*)$ with $\partial_\mu j^\mu = 0$ . Why: a vanishing four-divergence is exactly a continuity equation $\partial_t\rho + \nabla\cdot\mathbf j = 0$ , guaranteeing $\int\rho\,d^3x$ is conserved in time.

Step 4. Read off the density $\rho = j^0 = i(\psi^*\partial_t\psi - \psi\,\partial_t\psi^*)$ . For the plane wave $\psi = Ne^{-i(Et - \mathbf p\cdot\mathbf x)}$ we have $\partial_t\psi = -iE\psi$ and $\partial_t\psi^* = +iE\psi^*$ , so $\rho = i(-iE|N|^2 - iE|N|^2) = 2E|N|^2$ . Why: the density carries the sign of $E$ — and since the negative-energy branch of equation (3) is unavoidable, $\rho$ is genuinely negative for those states.

Step 5. Confront the interpretation. A probability density must be $\ge 0$ everywhere; $\rho = 2E|N|^2$ is not, so $\rho$ cannot be $|\psi|^2$ -style probability. Pauli and Weisskopf's resolution: multiply by the charge $e$ and read $e\rho$ as an electric charge density, which is allowed to take either sign. Positive-energy solutions carry one sign of charge, negative-energy solutions the opposite — particles and antiparticles. Why: the "defect" is not a defect at all once the single-particle wavefunction is promoted to a quantum field; it is the mathematical announcement that the theory describes charged particles and their antiparticles together.

Go deeper · Derivation C · 1 minRecovering the Schrodinger equation in the slow limit

Derivation CRecovering the Schrodinger equation in the slow limit

The purpose here is to show that the Klein–Gordon equation is not a rival to the Schrödinger equation but its parent: in the non-relativistic limit, where kinetic energy is tiny beside rest energy, Klein–Gordon reduces exactly to Schrödinger. The maneuver is to peel off the fast rest-mass oscillation and discard a term negligibly small precisely when the particle is slow — which both validates the equation and pins down what "non-relativistic" means.

Step 1. Factor out the rest-mass phase by writing $\psi(\mathbf x, t) = \phi(\mathbf x, t)\,e^{-imt}$ , where $\phi$ is a slowly varying amplitude. The factor $e^{-imt}$ carries the enormous frequency $m$ (the rest energy $mc^2/\hbar$ ); $\phi$ carries only the small kinetic part. Why: almost all of a slow particle's energy is rest energy, so isolating that dominant oscillation leaves a remainder $\phi$ that varies on the gentle non-relativistic timescale.

Step 2. Compute the time derivatives: $\partial_t\psi = (\dot\phi - im\phi)e^{-imt}$ and $\partial_t^2\psi = (\ddot\phi - 2im\dot\phi - m^2\phi)e^{-imt}$ . Insert into $\partial_t^2\psi - \nabla^2\psi + m^2\psi = 0$ ; the $\pm m^2\phi$ terms cancel, leaving $\ddot\phi - 2im\dot\phi - \nabla^2\phi = 0$ . Why: the cancellation of $m^2\phi$ is the point of the substitution — it removes the rest energy and exposes the kinetic dynamics underneath.

Step 3. Apply the non-relativistic approximation $|\ddot\phi| \ll m|\dot\phi|$ : the amplitude changes slowly compared with the rest-mass frequency, so the second derivative is negligible beside the first. Dropping $\ddot\phi$ gives $-2im\dot\phi - \nabla^2\phi = 0$ , i.e. $i\partial_t\phi = -\dfrac{1}{2m}\nabla^2\phi$ . Why: this is the free Schrödinger equation exactly; the condition $|\ddot\phi| \ll m|\dot\phi|$ is a precise statement of "kinetic energy much less than rest energy," the boundary of the non-relativistic regime.

Step 4. Note what became of the negative-energy branch: it was carried away in the discarded $\ddot\phi$ term, which is why the Schrödinger equation, first order in time, has no negative-energy solutions and a positive density. The pathologies of Klein–Gordon are exactly the price of retaining the relativistic $\ddot\phi$ that Schrödinger throws away. Why: this makes the trade concrete — Lorentz invariance and negative-energy states are the same coin, and a non-relativistic theory forfeits both.

Limiting cases

Case	Result	Meaning
Massless, $m = 0$	$\partial_\mu\partial^\mu\psi = 0$ ; $E = \\|\mathbf p\\|$	ordinary wave equation; photon-like dispersion, infinite-range $1/r$ field
Non-relativistic, $p \ll m$	$i\partial_t\phi = -\nabla^2\phi/2m$	Schrödinger equation recovered after removing the $e^{-imt}$ rest-mass phase
Ultra-relativistic, $p \gg m$	$E \approx \\|\mathbf p\\|$	mass negligible; the two branches approach the light-cone $E = \pm p$
Static source	$(\nabla^2 - m^2)\phi = -g\delta^3$ , $\phi \propto e^{-mr}/r$	Yukawa potential; finite range $1/m$ , the Compton wavelength
Second-quantized	field operator $\hat\phi$ ; $\rho \to$ charge density	Pauli–Weisskopf: negative $\rho$ is antiparticle charge, negative $E$ is a positive-energy antiparticle

Common trapThe Klein-Gordon equation is wrong and should be discarded

The most common misconception is that the negative-energy solutions and the sign-indefinite density prove the Klein–Gordon equation is a failed theory, to be thrown away for Dirac's. Every clause is misleading. The negative-energy solutions cannot be discarded even in principle: a second-order equation needs both roots for a complete set of states. The density $\rho$ is not a broken probability but a perfectly good charge density, sign and all (Pauli–Weisskopf, 1934). And Dirac's equation does not cure the negative-energy states — it has exactly the same negative-energy branch, merely dressed with spin. The correct statement is narrow: the Klein–Gordon equation cannot be read as a single-particle probability-conserving wave equation. Promoted to a quantum field, it is the exact and complete relativistic law for spin-0 particles — pions, kaons, and the Higgs boson all obey it.

Worked examples

1 · Yukawa's prediction of the pion (fully worked)

A short-ranged nuclear force is carried by a massive spin-0 field obeying the static Klein–Gordon equation, giving the Yukawa potential $\phi \propto e^{-mr}/r$ . If the observed range of the nuclear force is about $1.4\ \mathrm{fm}$ , what is the mass of the mediating particle? Then invert: what range does the measured pion mass $m_\pi = 139.6\ \mathrm{MeV}/c^2$ predict? Use $\hbar c = 197.3\ \mathrm{MeV\cdot fm}$ .

Step 1. Identify the range. In the Yukawa potential $e^{-mr}/r$ , the exponential falls by $1/e$ over the distance $R = 1/m$ (natural units), which restoring factors is the reduced Compton wavelength $R = \hbar/(mc) = \hbar c/(mc^2)$ . So range and mediator mass are reciprocals: a heavier mediator means a shorter force. The mass term in the Klein–Gordon equation is the inverse range of the force it carries.

Step 2. Solve for the mass from the observed range $R \approx 1.4\ \mathrm{fm}$ : $mc^2 = \hbar c / R = (197.3\ \mathrm{MeV\cdot fm})/(1.4\ \mathrm{fm}) = \mathbf{141\ MeV}$ . This is Yukawa's 1935 estimate — a new particle a few hundred times the electron mass, intermediate between electron and proton, later the meson.

Step 3. Invert with the modern pion mass to check consistency: $R = \hbar c/(m_\pi c^2) = (197.3\ \mathrm{MeV\cdot fm})/(139.6\ \mathrm{MeV}) = \mathbf{1.41\ fm}$ . The measured pion mass reproduces the nuclear-force range exactly.

Step 4. Interpret. The pion was discovered in cosmic rays in 1947, twelve years after Yukawa read its mass off the Klein–Gordon range formula. A force mediated by a mass- $m$ scalar field has range $\hbar/mc$ ; the $\sim 1.4\ \mathrm{fm}$ nuclear range demands $m \approx 140\ \mathrm{MeV}$ , and the pion delivers. This is the Klein–Gordon equation making a correct, quantitative prediction of a particle's mass.

2 · The mass gap and the negative density of a plane wave

Take a free pion plane wave $\psi = Ne^{-i(Et - \mathbf p\cdot\mathbf x)}$ with three-momentum $p = 100\ \mathrm{MeV}$ and mass $m = 139.6\ \mathrm{MeV}$ . (a) Find the two allowed energies. (b) Evaluate the Klein–Gordon density $\rho$ on each branch and comment on its sign.

Step 1. Allowed energies from equation (3): $E = \pm\sqrt{p^2 + m^2} = \pm\sqrt{100^2 + 139.6^2}\ \mathrm{MeV} = \pm\sqrt{10000 + 19488}\ \mathrm{MeV} = \pm\sqrt{29488}\ \mathrm{MeV} = \mathbf{\pm 171.7\ MeV}$ . The two solutions straddle a forbidden gap; the smallest possible $|E|$ (at $p = 0$ ) is $m = 139.6\ \mathrm{MeV}$ , so the full gap between branches is $2m = \mathbf{279.1\ MeV}$ .

Step 2. Density on the positive branch, from equation (4): $\rho_+ = 2E_+|N|^2 = 2(+171.7\ \mathrm{MeV})|N|^2 = \mathbf{+343.4\,|N|^2\ MeV}$ . Positive — this looks like a sensible (unnormalized) probability density.

Step 3. Density on the negative branch: $\rho_- = 2E_-|N|^2 = 2(-171.7\ \mathrm{MeV})|N|^2 = \mathbf{-343.4\,|N|^2\ MeV}$ . Negative. No probability density can be negative, which is the whole difficulty — but as a charge density it is fine: the two branches simply carry opposite charge.

Step 4. Interpret. The two branches carry equal and opposite $\rho$ , which under Pauli–Weisskopf is exactly a particle and its oppositely charged antiparticle. The number $\rho = 2E|N|^2$ shows the sign problem is not an accident of a bad wavefunction but a rigid consequence of the negative-energy branch that equation (3) forces on us.

Physics of the portrait

The Klein–Gordon equation is a rare case where a wrong turn became a foundation: born as a discarded draft of the Schrödinger equation, blamed for a decade as a diseased theory, then rehabilitated as the cornerstone of every scalar field theory — its "diseases" turning out to be the first sightings of antimatter.

Origins · 1 minKlein-Gordon equation and how this was discovered

Klein-Gordon equation

the relativistic scalar wave equation (□ + m²)ψ = 0 · relativistic quantum mechanics of spin-0 fields

The equation carries two names but was written by at least half a dozen people in a single remarkable year. Erwin Schrödinger derived the relativistic wave equation first, in late 1925 — but when he applied it to hydrogen it gave the wrong fine structure (it lacks electron spin, unknown at the time), so he set it aside and published the non-relativistic version that bears his name. In 1926 the relativistic equation was rediscovered independently by Oskar Klein, Walter Gordon, Vladimir Fock, Théophile de Donder, and Johann Kudar; history awarded it to Klein and Gordon. For over a decade it was regarded as defective: the negative-energy solutions seemed absurd, and the sign-indefinite density $\rho$ appeared to contradict the meaning of a wavefunction. Paul Dirac's response in 1928 — seeking a first-order equation for a positive density — produced the Dirac equation and, with it, spin and the positron, but did nothing to remove negative energies. The decisive rehabilitation came in 1934, when Wolfgang Pauli and Victor Weisskopf quantized the Klein–Gordon field and showed its density is a charge density and its negative-energy solutions are antiparticles — a result contemporaries jokingly called the "anti-Dirac" paper, since it proved a scalar theory was perfectly consistent without any spin machinery. Yukawa's 1935 use of the equation to predict the pion, and the discovery of the Higgs, confirmed the verdict: $(\Box + m^2)\psi = 0$ is the universal law of spin-0 fields.

History

1925

— Schrödinger derives the relativistic equation, then rejects it for giving the wrong hydrogen fine structure

1926

— Klein, Gordon, Fock, de Donder, and Kudar publish the equation independently

1928

— Dirac seeks a first-order equation for a positive density, producing the Dirac equation and spin

1934

— Pauli and Weisskopf quantize the field, reading ρ as charge and negative energies as antiparticles

1935

— Yukawa applies the static equation to the nuclear force and predicts the pion

1947

— the pion, the first spin-0 Klein-Gordon particle, is discovered in cosmic rays

2012

— the Higgs boson, a fundamental scalar field, is observed at the LHC

Going further

Texts. Griffiths, Introduction to Elementary Particles, 2nd ed., §7.1 (Eqs. 7.4, 7.9) · Thomson, Modern Particle Physics, §4.1 · Peskin and Schroeder, An Introduction to Quantum Field Theory, §2.1–2.3 (the scalar field and its propagator) · Schwartz, Quantum Field Theory and the Standard Model, Ch. 8 · Halzen and Martin, Quarks and Leptons, §3.1–3.3. The key results of this article are self-consistent: the operator substitutions $\hat E = i\partial_t$ , $\hat{\mathbf p} = -i\nabla$ applied to $E^2 = p^2 + m^2$ yield $(\partial_t^2 - \nabla^2 + m^2)\psi = 0$ ; plane-wave solutions give the dispersion $E = \pm\sqrt{p^2+m^2}$ and density $\rho = 2E|N|^2$ ; the conserved current satisfies $\partial_\mu j^\mu = 0$ with $j^\mu = i(\psi^*\partial^\mu\psi - \psi\partial^\mu\psi^*)$ ; the substitution $\psi = \phi e^{-imt}$ with $|\ddot\phi| \ll m|\dot\phi|$ reduces the equation to $i\partial_t\phi = -\nabla^2\phi/2m$ ; and the static Yukawa solution $(\nabla^2 - m^2)\phi = -g\delta^3 \Rightarrow \phi = (g/4\pi)e^{-mr}/r$ gives range $1/m = \hbar/mc = 1.41\ \mathrm{fm}$ for $m_\pi = 139.6\ \mathrm{MeV}$ , using the standard value $\hbar c = 197.3\ \mathrm{MeV\cdot fm}$ .

Prerequisites. Schrödinger Equation and Operator Substitutions · Four-Vector Notation and Lorentz Invariance. Leads to. The Dirac Equation · Antiparticles · The Dirac Probability Current.

Ask the tutor: Ask me to walk through why factoring $E^2 = p^2 + m^2$ into a first-order equation forces Dirac's gamma matrices — and how the same negative-energy branch survives there too.

Is the Klein-Gordon equation actually wrong or not

It is not wrong — it is misread when treated as a single-particle probability equation. As a relativistic wave equation for the amplitude of one particle, it fails: its density can be negative, so there is no consistent Born interpretation. But that is the only thing it fails at. Promoted to a quantized field, it is the exact and universally used description of spin-0 particles — negative-energy solutions become antiparticles, the "bad" density becomes electric charge. The honest statement is narrow: it is not a valid single-particle quantum mechanics, but it is a completely valid quantum field theory, which is what nature actually uses.

Why does the equation have to be second order in time

Because it descends from $E^2 = p^2 + m^2$ , in which energy appears squared. Under $\hat E = i\partial_t$ , the $E^2$ becomes $(i\partial_t)^2 = -\partial_t^2$ , a second time derivative. You cannot avoid this while keeping Lorentz invariance and locality: taking a single power of $E$ would require the square root $\sqrt{p^2 + m^2}$ of a differential operator, which is non-local and not manifestly covariant. Dirac's escape was to factor the square root by introducing matrices (and hence spin), producing a first-order equation — but only for spin-½. For a genuine scalar there is no such factorization, so second order in time is unavoidable, and with it comes a second time derivative in the density, which is why $\rho$ is not sign-definite.

What are the negative-energy solutions physically

In the single-particle reading they are an embarrassment: states with $E < 0$ into which any particle could cascade without limit. Dirac's stopgap was the "sea" — imagine all negative-energy states already filled, so exclusion blocks the fall and a hole is a positive-energy antiparticle. But that relies on Fermi statistics and fails for spin-0 bosons, which do not obey exclusion. The correct, general resolution is field-theoretic: a negative-energy solution of momentum $\mathbf p$ propagating "backward" is physically a positive-energy antiparticle of momentum $-\mathbf p$ propagating forward. The mass gap $2m$ is then the threshold to create a particle–antiparticle pair — no sea needed, and it works identically for bosons and fermions.

How does the Klein-Gordon equation relate to the Dirac equation

They are the spin-0 and spin-½ relativistic wave equations, and more closely related than they appear. Seeking a positive-definite density, Dirac factored $E^2 = p^2 + m^2$ into first-order form, which forced $4\times4$ gamma matrices — encoding spin-½ and predicting the positron. Crucially, every component of a Dirac spinor also satisfies the Klein–Gordon equation: apply the Dirac operator twice and you recover $(\Box + m^2)\psi = 0$ , so Klein–Gordon is the "square" of Dirac. Dirac gained a positive density and spin, but did not eliminate negative-energy solutions; they survive there too and need the same antiparticle interpretation. Klein–Gordon governs spin-0 (pions, Higgs); Dirac governs spin-½ (electrons, quarks).

What real particles obey the Klein-Gordon equation

Every spin-0 particle. The pion ( $m \approx 140\ \mathrm{MeV}$ ), the kaon, the eta, and the rest of the pseudoscalar and scalar mesons are Klein–Gordon fields, as are hypothetical scalars like the axion. The crown example is the Higgs boson, discovered in 2012: it is the Standard Model's only fundamental spin-0 particle, and its free propagation is Klein–Gordon dynamics (with an added self-interaction that gives it mass through spontaneous symmetry breaking). In cosmology the inflaton and quintessence fields are scalars obeying the equation in curved spacetime. So "which equation" is really a question of spin: zero means Klein–Gordon, spin-½ means Dirac.

Why does the Schrodinger equation avoid these problems

Because it is non-relativistic, and it pays for that with the very symmetry Klein–Gordon was built to have. Schrödinger's equation comes from $E = p^2/2m$ , linear in $E$ , so it is first order in time — one time derivative means the density is simply $|\psi|^2$ , manifestly positive, and there is no square root to admit a negative-energy branch. But being first order in time and second order in space makes it treat time and space asymmetrically, which a Lorentz boost cannot preserve. In fact the Klein–Gordon equation contains the Schrödinger equation: peel off the rest-mass phase $e^{-imt}$ and drop the small second time derivative, and Schrödinger's equation falls out. The positive density and the absence of negative energies are exactly what you gain by throwing that term away — and the loss is relativity.