Klein-Gordon Probability Current and Negative Densities

Every wave equation worth its salt comes with a conservation law. The Schrödinger equation hands you a positive probability density $\rho=|\Psi|^2$ and a current that shuttles it around without loss — the article "Probability Current and the Continuity Equation" builds exactly that. When you write down the first relativistic wave equation, the Klein-Gordon equation, and ask the same question — what is locally conserved? — you get a Lorentz-covariant four-current $j^\mu$ that looks entirely reasonable. Then you inspect its time component and the floor gives way. The trouble is structural, not a slip of algebra. The Klein-Gordon equation is second order in time, so specifying a state means giving both $\phi$ and $\dot\phi$ as independent initial data. The conserved density is forced to reach down and grab $\dot\phi$ , and the instant a density depends on a freely chosen time derivative, it loses positivity. Concretely, for a plane wave $\rho\propto E$ — and since $E^2=p^2c^2+m^2c^4$ has two roots, the negative-energy branch carries negative density. You cannot quietly discard those solutions: they are needed for completeness, and interactions scatter into them. Historically this "negative probability" was the pressure that produced the Dirac equation, and eventually the deeper answer: the Klein-Gordon $\phi$ is not a single-particle wavefunction at all but a field. Once quantized, $e\,j^\mu$ is an electric-charge current, and a negative $\rho$ simply means an excess of antiparticles of opposite charge. The negative density was never a defect — it was a prophecy of antimatter.

Key takeaways

The Klein-Gordon equation admits a conserved four-current $j^\mu=\frac{i\hbar}{2m}(\phi^*\partial^\mu\phi-\phi\,\partial^\mu\phi^*)$ ; its divergence vanishes because $\phi$ solves the field equation.
Being second order in time forces the density $\rho=j^0/c$ to depend on $\partial_t\phi$ , so it is not positive-definite.
For a plane wave $\rho=\dfrac{|N|^2E}{mc^2}$ : the sign of the density is the sign of the energy. Negative-energy modes carry $\rho<0$ .
The spatial current $\mathbf j$ is identical in form to the Schrödinger current — the pathology lives entirely in the time component.
The cure is a change of framework: $\phi$ becomes a quantized field and $j^\mu$ a charge current, so $\rho<0$ means net antiparticle charge.
Key idea: a second-order-in-time equation forces the conserved density to depend on $\partial_t\phi$ , so $\rho$ cannot be positive-definite — the honest reading is charge density, not probability density.

The core insight

A relativistic scalar wave equation that is second order in time cannot support a positive-definite probability density; its conserved $j^0$ tracks the sign of the energy and is properly a charge density.

Where this lives today

The negative-density "problem" is now a feature at the heart of every quantum field theory of charged bosons. The conserved current $j^\mu$ is exactly the object that couples a complex scalar to electromagnetism through minimal coupling, giving the scalar-QED vertex $\sim e(p+p')^\mu$ used to compute production and scattering of charged spin-0 particles — charged pions, kaons, and, in the Standard Model, the Higgs and $W$ boson interactions built on the same template. The indefinite sign of $\rho$ is precisely what allows a single field to describe both a particle and its oppositely charged antiparticle. The current also drives the Klein paradox: when a Klein-Gordon particle meets a supercritical electrostatic step, tracking $j^\mu$ across the barrier reveals a transmitted flux carrying the "wrong" density sign — the signature of pair creation at the barrier. Modern condensed-matter analogues in graphene, where low-energy electrons obey an effective relativistic equation, have made Klein tunneling an experimentally accessible phenomenon rather than a paradox to be explained away.

The physical picture

A continuity equation $\partial_\mu j^\mu=0$ is pure bookkeeping: whatever flows out of a small box must drain what sits inside. In the nonrelativistic world the inside-quantity is $|\Psi|^2\ge 0$ , and the Schrödinger equation — first order in $\partial_t$ — lets that positive quantity be conserved. The Klein-Gordon equation is second order in time, and that single fact reshapes everything downstream: its conserved density must involve $\partial_t\phi$ , and a density built from a freely specifiable time derivative can point either way. The cleanest way to feel the problem is to look at a stationary state, where $\rho\propto E$ . Relativistic energy comes in two signs because $E^2=p^2c^2+m^2c^4$ has two roots $E=\pm\sqrt{p^2c^2+m^2c^4}$ . A positive-energy mode deposits positive density; a negative-energy mode deposits negative density. The formalism has no way to declare one of them "the particle" and throw the other out — they are both needed to expand a general field. So the single-particle probability story collapses, while the field story happily keeps both and names them particle and antiparticle.

Figure

Density tracks the energy branch

The same manipulation that produces the poisonous $j^0$ produces a perfectly innocent spatial current $\mathbf j=\frac{\hbar}{2mi}(\phi^*\nabla\phi-\phi\nabla\phi^*)$ , letter-for-letter the Schrödinger probability current. The pathology is confined to the time slot — a direct fingerprint of the second time derivative and nowhere else.

Figure

First order versus second order in time

The core equations

The starting point is the field equation itself. The article "The Klein-Gordon Equation" derives it by promoting the relativistic dispersion $E^2=p^2c^2+m^2c^4$ to operators, and it pins the conventions we inherit throughout this chapter: metric $g_{\mu\nu}=\mathrm{diag}(+,-,-,-)$ , and (in the dimensional form kept here) explicit $\hbar$ and $c$ . We restate the equation only to have the object whose current we build.

\left(\Box+\frac{m^2c^2}{\hbar^2}\right)\phi=0,\qquad \Box\equiv\partial_\mu\partial^\mu=\frac{1}{c^2}\partial_t^2-\nabla^2 \qquad (1)

where: $\phi$ is a complex scalar field (the candidate wavefunction); $\Box$ is the d'Alembertian; $m$ is the rest mass; $\hbar,c$ the reduced Planck constant and light speed. Reading it physically: the operator is real, so $\phi^*$ obeys the same equation — the fact we will lean on twice. To find what is conserved, one mimics the Schrödinger derivation. The Klein-Gordon four-current that emerges is

j^\mu=\frac{i\hbar}{2m}\left(\phi^*\partial^\mu\phi-\phi\,\partial^\mu\phi^*\right) \qquad (2)

where: $j^\mu=(c\rho,\mathbf j)$ is the conserved four-current, and $\partial^\mu=(\tfrac1c\partial_t,-\nabla)$ in the $(+,-,-,-)$ metric. The prefactor $i\hbar/2m$ is chosen to make $j^\mu$ real and to make $\rho$ reduce to $|\psi|^2$ in the nonrelativistic limit. This object is genuinely conserved, but the payoff is entirely in how its pieces split into a density and a flux. Split into time and space parts. The density is the time component,

\rho=\frac{j^0}{c}=\frac{i\hbar}{2mc^2}\left(\phi^*\partial_t\phi-\phi\,\partial_t\phi^*\right) \qquad (3)

where: $\rho$ carries an independent dependence on $\partial_t\phi$ . This is the whole story in one line: because $\phi$ and $\partial_t\phi$ can be chosen independently, $\rho$ is not sign-definite. Contrast the Schrödinger case, where "Probability Current and the Continuity Equation" gives $\rho=|\Psi|^2\ge 0$ with no time derivative in sight. The spatial current, by contrast, is the familiar object,

\mathbf j=\frac{\hbar}{2mi}\left(\phi^*\nabla\phi-\phi\,\nabla\phi^*\right) \qquad (4)

where: $\mathbf j$ is identical in form to the Schrödinger probability flux $\frac{\hbar}{m}\,\mathrm{Im}(\Psi^*\nabla\Psi)$ that the current article tabulates. The relativistic disease is nowhere here; it is quarantined in $\rho$ . Together they satisfy local conservation,

\partial_t\rho+\nabla\cdot\mathbf j=0 \qquad (5)

where: this continuity equation is not an added postulate — it follows from $(1)$ , as the first derivation shows. Evaluating $(3)$ on a plane wave exposes the sign problem numerically. For $\phi=Ne^{i(\mathbf p\cdot\mathbf x-Et)/\hbar}$ one finds

\rho_{\text{plane}}=\frac{|N|^2E}{mc^2} \qquad (6)

where: $E=\pm\sqrt{p^2c^2+m^2c^4}$ is the energy eigenvalue. The density is $E/mc^2$ times $|N|^2$ : positive for the upper branch, negative for the lower. There is no algebraic escape — the sign of $\rho$ is the sign of $E$ . The resolution lives one framework up. Promoting $\phi$ to a quantum field turns $\int\rho\,\mathrm d^3x$ into a conserved charge,

Q=\int\rho\,\mathrm d^3x=e\,(N_{+}-N_{-}) \qquad (7)

where: $Q$ is the total electric charge, $e$ the quantum of charge, and $N_\pm$ the particle/antiparticle numbers. Negative density now reads as net negative charge — physically sensible and experimentally confirmed. This equation is a signpost, not something we derive here; the construction belongs to "Charge Conjugation and the Positron."

Derivation AConstruction and conservation of the Klein-Gordon current

We build $j^\mu$ and prove $\partial_\mu j^\mu=0$ directly from the field equation, following the exact recipe that yields the Schrödinger current but now carrying two time derivatives. The strategy is to write the equation for $\phi$ and for $\phi^*$ , cross-multiply so the mass terms cancel, and recognize what is left as a total four-divergence. The one thing to watch is that the derivative cross-terms cancel identically — that is what collapses the expression into a pure divergence and identifies the conserved current.

Step 1. Write the two field equations. Because the Klein-Gordon operator is real, the conjugate field obeys the same equation:

\Box\phi+\frac{m^2c^2}{\hbar^2}\phi=0,\qquad \Box\phi^*+\frac{m^2c^2}{\hbar^2}\phi^*=0.

Step 2. Multiply the first by $\phi^*$ , the second by $\phi$ , and subtract:

\phi^*\Box\phi-\phi\,\Box\phi^*+\frac{m^2c^2}{\hbar^2}\big(\phi^*\phi-\phi\phi^*\big)=0.

Why the mass term dies: $\phi^*\phi-\phi\phi^*=0$ for ordinary (commuting) fields, so the entire mass contribution vanishes and only derivative terms survive.

Step 3. Recognize the survivor as a divergence. Using the product rule,

\phi^*\Box\phi-\phi\,\Box\phi^*=\partial_\mu\big(\phi^*\partial^\mu\phi-\phi\,\partial^\mu\phi^*\big),

because the cross-terms $\partial_\mu\phi^*\,\partial^\mu\phi-\partial_\mu\phi\,\partial^\mu\phi^*$ cancel identically. Why: both cross-terms are the same scalar contraction, so their difference is exactly zero.

Step 4. Normalize. Defining

j^\mu\equiv\frac{i\hbar}{2m}\big(\phi^*\partial^\mu\phi-\phi\,\partial^\mu\phi^*\big),

Steps 2–3 give $\partial_\mu j^\mu=0$ . The prefactor $i\hbar/2m$ makes $j^\mu$ real and fixes the nonrelativistic normalization. Verified: symbolic check confirms $\phi^*\Box\phi-\phi\,\Box\phi^*=\partial_\mu(\phi^*\partial^\mu\phi-\phi\,\partial^\mu\phi^*)$ as an identity, and that substituting the Klein-Gordon equation sends it to zero — mass terms cancel exactly.

Go deeper · Derivation B · 1 minSign of the density on a plane wave

Derivation BSign of the density on a plane wave

Here we pin the crisis to a number: the density of a plane wave is $|N|^2E/mc^2$ , so a negative-energy solution carries strictly negative density. The route is a direct substitution of the plane-wave ansatz into $\rho$ , tracking every factor of $i$ and $E$ . The subtle point is that both roots $E=\pm\sqrt{p^2c^2+m^2c^4}$ are legitimate solutions of the same equation, so the negative-density case is not exotic — it is half of the solution space.

Step 1. Take the positive-frequency ansatz $\phi=Ne^{i(\mathbf p\cdot\mathbf x-Et)/\hbar}$ , so that $\partial_t\phi=-\tfrac{iE}{\hbar}\phi$ and $\partial_t\phi^*=+\tfrac{iE}{\hbar}\phi^*$ .

Step 2. Insert into $\rho=\frac{i\hbar}{2mc^2}(\phi^*\partial_t\phi-\phi\,\partial_t\phi^*)$ :

\rho=\frac{i\hbar}{2mc^2}\Big(\phi^*\big(-\tfrac{iE}{\hbar}\big)\phi-\phi\big(+\tfrac{iE}{\hbar}\big)\phi^*\Big).

Step 3. Collect. With $\phi^*\phi=|N|^2$ ,

\rho=\frac{i\hbar}{2mc^2}\Big(-\tfrac{2iE}{\hbar}\Big)|N|^2=\frac{|N|^2E}{mc^2}.

Why this is the crisis: choosing the negative root $E=-\sqrt{p^2c^2+m^2c^4}<0$ gives $\rho<0$ with no way to blame normalization — $|N|^2>0$ always. Verified: sympy substitution gives $\rho=E\,|N|^2/mc^2$ for the $-Et$ form and $\rho=-E\,|N|^2/mc^2$ for the $+Et$ (negative-energy) form; the density sign equals the energy sign in both.

Go deeper · Derivation C · 1 minNonrelativistic limit restores Schrödinger's positive density

Derivation CNonrelativistic limit restores Schrödinger's positive density

Finally we show the disease is purely relativistic: strip off the rest-mass oscillation and the density collapses to $|\psi|^2\ge 0$ . The strategy is to write $\phi=e^{-imc^2t/\hbar}\psi$ with a slowly varying envelope $\psi$ , then keep only the leading rest-mass term in $\partial_t\phi$ . The point to watch is which term dominates: in the nonrelativistic regime $\tfrac{mc^2}{\hbar}|\psi|\gg|\partial_t\psi|$ , so the fast phase wins and everything else is a $v^2/c^2$ correction.

Step 1. Separate the rest-mass phase: $\phi(\mathbf x,t)=e^{-imc^2t/\hbar}\psi(\mathbf x,t)$ , where $\psi$ is the nonrelativistic wavefunction.

Step 2. Differentiate: $\partial_t\phi=e^{-imc^2t/\hbar}\big(-\tfrac{imc^2}{\hbar}\psi+\partial_t\psi\big)$ . Why keep only the first term: $|\partial_t\psi|\ll\tfrac{mc^2}{\hbar}|\psi|$ in the slow regime, so the rest-energy term dominates by a factor $\sim c^2/v^2$ .

Step 3. Insert into $\rho$ and keep the leading pieces:

\rho=\frac{i\hbar}{2mc^2}\Big(\psi^*\big(-\tfrac{imc^2}{\hbar}\big)\psi-\psi\big(+\tfrac{imc^2}{\hbar}\big)\psi^*+\cdots\Big)=|\psi|^2\ge 0.

Verified: the full symbolic $\rho$ equals $|\psi|^2$ plus $O(\hbar/mc^2)$ correction terms carrying $\partial_t\psi$ ; dropping those (the $v^2/c^2$ pieces) leaves exactly $\psi^*\psi$ .

Limiting cases

The behavior of $\rho$ across the regimes that matter, each recomputed from $(3)$ and $(6)$ :

case	result	meaning
Positive-energy plane wave	$\rho=\dfrac{\|N\|^2E}{mc^2}>0$	Positive density, proportional to energy — a "normal" probability up to the factor $E/mc^2$ .
Negative-energy plane wave	$E<0\Rightarrow\rho=\dfrac{\|N\|^2E}{mc^2}<0$	The lower root gives negative density — the heart of the crisis.
Nonrelativistic limit	$E\to mc^2,\;\rho\to\|\psi\|^2$	With $\phi=e^{-imc^2t/\hbar}\psi$ , Schrödinger's positive density is recovered.
Real (neutral) field	$\phi=\phi^*\Rightarrow j^\mu\equiv 0$	A real field carries no current — an uncharged particle that is its own antiparticle.

Worked examples

Example 1 — The density of a negative-energy pion mode. Take a charged pion mode ( $m_\pi c^2=139.6\ \text{MeV}$ ) with momentum $pc=100\ \text{MeV}$ , and compare the densities of its two energy roots at fixed normalization $|N|^2=1$ .

Step 1. Compute the energy magnitude. $E=\sqrt{(pc)^2+(mc^2)^2}=\sqrt{100^2+139.6^2}=\sqrt{10000+19488}=\sqrt{29488}=171.7\ \text{MeV}$ .

Step 2. Positive-energy branch, $E=+171.7\ \text{MeV}$ . From $\rho=|N|^2E/mc^2$ : $\rho_+=+171.7/139.6=+1.230$ (in units where $|N|^2=1$ ).

Step 3. Negative-energy branch, $E=-171.7\ \text{MeV}$ . The same formula gives $\rho_-=-171.7/139.6=-1.230$ .

Step 4. Read the result. The two modes carry densities of equal magnitude and opposite sign, $\rho_\pm=\pm1.230$ . No single-particle probability can be negative, so the lower mode has no probabilistic reading — but as a charge density $\rho_\pm=\pm1.230\,e/mc^2$ it describes $\pi^-$ versus $\pi^+$ , the physically observed pair.

Example 2 — Size of the relativistic correction to the density. Estimate how large the "extra" density term is for a mildly relativistic electron, $v=0.1c$ , relative to the Schrödinger value $|\psi|^2$ .

Step 1. From the reduction, the leading density is $|\psi|^2$ and the first correction scales as the ratio of the envelope's frequency to the rest-mass frequency, $\sim(E_{\text{kin}}/mc^2)$ .

Step 2. Kinetic energy $E_{\text{kin}}\approx\tfrac12 mv^2=\tfrac12 mc^2(v/c)^2=\tfrac12 mc^2(0.1)^2=0.005\,mc^2$ .

Step 3. So the fractional correction is $E_{\text{kin}}/mc^2\approx0.005$ , i.e. about 0.5%.

Step 4. Read the result. At $v=0.1c$ the Klein-Gordon density differs from $|\psi|^2$ by only half a percent — confirming that the negative-density pathology is invisible in the nonrelativistic world and switches on only as $v\to c$ , precisely where the negative-energy branch becomes accessible.

Origins · 1 minwalter gordon and how this was discovered

walter gordon

1893–1939 · German physicist who, in 1926, independently wrote down the relativistic scalar wave equation and constructed its conserved current — the "Gordon" half of the Klein-Gordon equation.

Walter Gordon published the relativistic scalar equation and its four-current in 1926, the same year as Oskar Klein, Vladimir Fock, and others; the equation carries his and Klein's names. Erwin Schrödinger had in fact derived the same equation first, before his celebrated nonrelativistic equation, but discarded it because its predicted hydrogen fine structure disagreed with experiment — a discrepancy later traced to the missing electron spin. Gordon's construction of $j^\mu$ is the object at the center of this article; his name also survives in the "Gordon decomposition" of the Dirac current.

History

1926

— Schrödinger derives the relativistic scalar equation first, discards it over the hydrogen fine-structure mismatch, then publishes the nonrelativistic equation

1926

— Klein and Gordon independently publish the relativistic scalar equation and its conserved current

1927

— Explicit recognition that the Klein-Gordon time density

j^0

is not positive-definite, undermining the single-particle probability interpretation

1928

— Dirac, seeking a first-order equation with a positive density, writes the Dirac equation

1934

— Pauli and Weisskopf reinterpret

e\,j^\mu

as the charge current of a quantized scalar field, resolving the crisis

Going further

References. Griffiths, Introduction to Elementary Particles, 2nd ed., §7.1 · Bjorken & Drell, Relativistic Quantum Mechanics, ch. 9 · Sakurai & Napolitano, Modern Quantum Mechanics, 2nd ed., ch. 8 · Shankar, Principles of Quantum Mechanics, 2nd ed., ch. 20 · Cohen-Tannoudji, Diu & Laloë, Quantum Mechanics · Pauli & Weisskopf, Helv. Phys. Acta 7, 709 (1934).

Provenance. Equations standard from Griffiths (Elementary Particles §7.1), Bjorken-Drell ch. 9, Sakurai-Napolitano ch. 8, Shankar ch. 20, Cohen-Tannoudji. A verifier should: (1) re-derive $\partial_\mu j^\mu=0$ by substituting the Klein-Gordon equation and confirming the mass-term cancellation; (2) check the plane-wave result $\rho=|N|^2E/mc^2$ via $\partial_t\phi=-iE\phi/\hbar$ ; (3) confirm the nonrelativistic reduction $\rho\to|\psi|^2$ using $\phi=e^{-imc^2t/\hbar}\psi$ . Metric convention $(+,-,-,-)$ , $\hbar,c$ explicit. Pauli-Weisskopf reinterpretation: Helv. Phys. Acta 7, 709 (1934). All three checks reproduced symbolically in the accompanying verifier; no base error found.

Builds on: The Klein-Gordon Equation · Negative-Energy Solutions and the Interpretive Crisis · Probability Current and the Continuity Equation. Leads to: The Dirac Equation · Charge Conjugation and the Positron · The Dirac Sea and Antiparticles.

If the density can be negative, how can the current be "conserved" in any useful sense?

Conservation means $\partial_t\rho+\nabla\cdot\mathbf j=0$ holds locally, so $\int\rho\,\mathrm d^3x$ is time-independent — that is true regardless of the sign of $\rho$ . What fails is not conservation but the probabilistic interpretation: a conserved quantity that can be negative cannot be a probability. The fix is to reinterpret the conserved integral as electric charge, which is perfectly happy being positive, negative, or zero.

Why can't we just keep the positive-energy solutions and throw the negative ones away?

Two reasons. First, completeness: an arbitrary field configuration requires both branches in its expansion, so restricting to positive energy is not a closed subspace under time evolution once interactions are present. Second, interactions scatter positive-energy states into negative-energy ones — the "hole" cannot be plugged by fiat. The honest response is to change the interpretation of the whole theory, which is what field quantization does.

Is the spatial current

\mathbf j

also pathological?

No — and that is a crucial diagnostic. The spatial current $\mathbf j=\frac{\hbar}{2mi}(\phi^*\nabla\phi-\phi\nabla\phi^*)$ is identical in form to the Schrödinger probability flux and has no sign problem. The entire pathology lives in the time component $\rho=j^0/c$ , which is exactly the piece forced to carry a time derivative by the second-order-in-time structure. This localization is the strongest hint that the issue is structural, not a computational slip.

How does this relate to the negative-energy problem Dirac faced?

They are historical siblings. The Klein-Gordon negative density and the negative-energy solutions both stem from the equation being quadratic in $E$ . Dirac's motivation was to build a first-order equation whose density would look like $\psi^\dagger\psi\ge 0$ — and he succeeded on the density front. But the negative-energy solutions came right back, forcing the Dirac sea and, ultimately, antiparticles. So Dirac cured the density but not the spectrum; field theory cured both.

What does the prefactor

i\hbar/2m

actually do?

Three things. The $i$ makes the combination $\phi^*\partial^\mu\phi-\phi\,\partial^\mu\phi^*$ (which is pure imaginary) into a real current. The $1/2$ symmetrizes the two terms. The $1/m$ (with the $c^2$ in $\rho$ ) sets the normalization so that in the nonrelativistic limit $\rho\to|\psi|^2$ exactly, matching the Schrödinger convention. Different textbooks absorb factors of $\hbar$ and $c$ differently, so always check the convention before comparing formulas.

Does a neutral scalar like the Higgs have a Klein-Gordon current?

Not a conserved $U(1)$ current of this form. If $\phi=\phi^*$ (a real field), then $j^\mu\equiv 0$ identically — a neutral particle carries no conserved charge and is its own antiparticle. The physical Higgs boson is neutral and indeed has no particle–antiparticle distinction. Charged scalars (like $\pi^\pm$ , or the charged components of the Higgs doublet before symmetry breaking) are complex fields and do carry the current $j^\mu$ .

Why does

\rho\propto E

specifically, and not

\rho\propto E^2

\rho\propto|E|

Because $\rho$ is linear in the time derivative $\partial_t\phi$ , and for a plane wave $\partial_t\phi=-\tfrac{iE}{\hbar}\phi$ brings down exactly one factor of $E$ . The combination $\phi^*\partial_t\phi-\phi\,\partial_t\phi^*$ doubles it and cancels the imaginary unit, leaving $\rho\propto E$ . If the density had been quadratic in $\partial_t\phi$ (as the energy density is), it would have been sign-definite — but then it would not have been the conserved charge that the continuity equation demands.