Negative-Energy Solutions and the Interpretive Crisis

Take the one equation everyone agrees relativity demands — the energy–momentum relation $E^2=\mathbf p^2c^2+m^2c^4$ — and quantize it. The moment you solve for $E$ you are forced to take a square root, and a square root has two signs. Alongside the sensible positive-energy electron sits a twin: a solution with $E=-\sqrt{\mathbf p^2c^2+m^2c^4}\le -mc^2$ , an energy that can be made arbitrarily large and negative. Every relativistic wave equation built to reproduce the correct dispersion — the Klein–Gordon equation for spin-0, the Dirac equation for spin-½ — inherits these negative-energy solutions. They are not an artifact of a clumsy equation; they are baked into relativistic kinematics. The tempting move is to simply throw them away. You cannot. The full set of plane waves — positive and negative energy together — is a complete basis, and the Fourier decomposition of any localized state necessarily contains negative-energy pieces. Worse, the moment you switch on a potential, or try to pin the particle to a region smaller than its Compton wavelength, the two sectors mix. There is no Lorentz-invariant, dynamically stable projector onto positive energies alone within a single-particle theory. The trouble has two faces. First, stability: a Hamiltonian whose spectrum runs down to $-\infty$ has no ground state, so a free electron could cascade downward forever, radiating without limit. Second, for Klein–Gordon, the conserved density goes negative exactly where the energy does, so it cannot be a probability. The way out is not to delete the unwanted solutions but to reinterpret them — as positive-energy antiparticles. That reinterpretation, Dirac's sea and its modern field-theory form, turned a crisis into the prediction of the positron.

Key takeaways

The relativistic dispersion $E^2=\mathbf p^2c^2+m^2c^4$ has two roots, $E=\pm\sqrt{\mathbf p^2c^2+m^2c^4}$ , so negative-energy solutions are mathematically unavoidable in every relativistic wave equation.
You cannot discard them by hand: they are half of a complete set of modes, and localization or any interaction mixes the positive- and negative-energy sectors.
A spectrum unbounded below means no stable ground state — a free particle could fall forever down the negative continuum radiating unbounded energy, contradicting the stability of matter.
The two continua are separated by a gap of width $2mc^2$ ( $=1.022$ MeV for an electron); interference across that gap is the origin of Zitterbewegung at frequency $\omega_Z=2mc^2/\hbar\approx1.55\times10^{21}\,\mathrm{s^{-1}}$ .
Klein–Gordon compounds the problem with an indefinite density; Dirac cures the density but not the spectrum.
Key idea: negative-energy solutions are forced on you by relativistic kinematics and completeness — you cannot delete them, you must reinterpret them as positive-energy antiparticles, a step completed only in quantum field theory.

The core insight

Negative-energy solutions are forced on you by relativistic kinematics and completeness; they make the spectrum unbounded below and (for Klein–Gordon) the density indefinite — so they must be reinterpreted as positive-energy antiparticles, not discarded.

Where this lives today

The negative-energy problem is the historical hinge between single-particle relativistic quantum mechanics and quantum field theory, and its descendants are everywhere in modern physics. In strong-field QED, the Klein paradox becomes real: an electric field approaching the Schwinger critical value $E_c=m^2c^3/e\hbar\approx1.3\times10^{18}\,\mathrm{V/m}$ pulls particle–antiparticle pairs out of the vacuum, a process that next-generation laser facilities and heavy-ion collisions are actively probing. Zitterbewegung, long thought purely theoretical, has been observed in analog systems — trapped ions, ultracold atoms in engineered lattices, and graphene, where the low-energy electrons obey a two-dimensional Dirac equation with an effective " $c$ " of the Fermi velocity, so the two-sector interference happens at accessible frequencies. More broadly, the reinterpretation of negative-energy solutions as antiparticles is the reason particle number is not conserved in relativistic quantum theory: enough energy localized in a small enough region creates pairs. Every collider event, every calculation of the electron's anomalous magnetic moment, and every treatment of the vacuum as a dynamical medium rests on the resolution that Dirac's crisis forced. What began as an embarrassing extra square-root sign is now a load-bearing pillar of the Standard Model.

The physical picture

Square roots come in pairs, and that single fact reshapes quantum mechanics when it meets relativity. Non-relativistic quantum mechanics rests on $E=\mathbf p^2/2m$ : a single-valued spectrum, bottomless on one side but with a clean floor at $E=0$ . Relativity replaces it with $E^2=\mathbf p^2c^2+m^2c^4$ . Squaring an energy destroys the sign information — the wave equation only ever sees $E^2$ , and both signs of $E$ solve it equally well. This is not a defect you can engineer away; it is a property of the relativistic dispersion relation itself, and it therefore afflicts the Klein–Gordon equation, the Dirac equation, and everything built to respect Lorentz invariance. Picture the spectrum as two disconnected continua separated by a gap of width $2mc^2$ : one running from $+mc^2$ upward, the other from $-mc^2$ downward. In empty space the gap is your friend — a positive-energy state stays positive, and you can pretend the lower continuum does not exist. But the gap is not a wall. Confine a particle to a region smaller than its Compton wavelength $\lambda_C=\hbar/mc$ and Heisenberg forces in momentum components of order $mc$ or larger — precisely the relativistic regime where the positive- and negative-energy amplitudes are strongly mixed. Interactions bridge the gap dynamically; localization bridges it kinematically.

Figure

The relativistic spectrum: two continua and a mass gap

A spectrum unbounded below is intolerable for the same reason a classical potential with no minimum is intolerable: nothing stops the particle from falling forever, dumping arbitrarily much energy into radiation as it goes. Real electrons manifestly do not do this. The mathematics is correct, so the interpretation — "a single particle whose energy can be negative" — must be wrong.

Figure

Localization forces you across the gap

The escape is a change of reading, not of mathematics. Stop hearing the minus sign as "a particle with negative energy moving forward in time" and start hearing it as "an antiparticle with positive energy." The very solution that horrified Dirac becomes the positron — a triumph born from refusing to discard the unwanted half of the theory.

The core equations

Everything begins with the relativistic invariant. The squared four-momentum of a particle of mass $m$ is fixed, and that single constraint already contains both energy signs.

p^\mu p_\mu = m^2c^2 \;\;\Longleftrightarrow\;\; \frac{E^2}{c^2}-\mathbf p^2 = m^2c^2 \;\;\Longrightarrow\;\; E = \pm\sqrt{\mathbf p^2c^2+m^2c^4} \qquad (1)

where: $E$ — total relativistic energy of the mode; $\mathbf p$ — three-momentum; $m$ — rest mass; $c$ — speed of light. We use the mostly-minus metric $g_{\mu\nu}=\mathrm{diag}(+,-,-,-)$ (the convention pinned in The Klein-Gordon Equation), and we keep $\hbar$ and $c$ explicit wherever a physical number is quoted. Reading it physically: no additional input in the free equation selects a sign, so the two roots are solutions of equal mathematical standing. The positive root is the particle you expected; the negative root is the one you must explain. That both signs must be kept is not a stylistic choice — it is forced by the order of the wave equation, and the cleanest place to see it is the Klein–Gordon operator, second order in time. The Klein-Gordon Equation establishes that its plane-wave solutions require $p^0=\pm\sqrt{\mathbf p^2c^2+m^2c^4}/c$ ; here we make precise why the negative root cannot be dropped.

Derivation AThe negative-energy branch is required for completeness

The goal is to show that the negative-energy solutions are not optional decorations you may throw away, but a mandatory half of a complete solution set — because the Klein–Gordon (and Dirac) equation is second order in time. The strategy is to reduce each Fourier mode to an ordinary differential equation in time, count its independent solutions, and identify them with the two energy signs. The one thing to watch: "completeness" is a statement about spanning all initial data $(\phi,\dot\phi)$ , not just the ones you happen to like.

Step 1. Fourier-transform in space. Each momentum mode $\phi_{\mathbf p}(t)$ of a Klein–Gordon field obeys $\ddot\phi_{\mathbf p}+\omega_{\mathbf p}^2\,\phi_{\mathbf p}=0$ with $\omega_{\mathbf p}=E_{\mathbf p}/\hbar=\sqrt{\mathbf p^2c^2+m^2c^4}/\hbar$ . This is a linear second-order ODE in time. Why second order matters: the Schrödinger equation is first order in time, so one function $\phi(\mathbf x)$ fixes the future. The Klein–Gordon equation is second order, so you must supply both $\phi$ and $\dot\phi$ — twice as much initial data per mode.

Step 2. A second-order linear ODE has a two-dimensional solution space. Its independent solutions are $e^{-i\omega_{\mathbf p}t}$ and $e^{+i\omega_{\mathbf p}t}$ , i.e. the two energy signs $E=+E_{\mathbf p}$ and $E=-E_{\mathbf p}$ .

Step 3. Therefore the general mode is $\phi_{\mathbf p}(t)=a_{\mathbf p}\,e^{-iE_{\mathbf p}t/\hbar}+b_{\mathbf p}\,e^{+iE_{\mathbf p}t/\hbar}$ . Fixing arbitrary $(\phi_{\mathbf p},\dot\phi_{\mathbf p})$ at $t=0$ requires both coefficients $a_{\mathbf p}$ and $b_{\mathbf p}$ . Setting $b_{\mathbf p}\equiv0$ (deleting negative energies) leaves you unable to match a generic $\dot\phi_{\mathbf p}$ . Why this closes the escape route: "keep only $E>0$ " would halve the solution space and fail to represent even a freely evolving field with the wrong initial velocity — the positive-energy modes alone are not complete. Verified: the second-order time ODE yields a 2-parameter solution space and both exponentials $e^{\mp iE_{\mathbf p}t/\hbar}$ are independent — checked symbolically (root count equals equation order) in verify_aq-qm-negative-energy-solutions.py.

The negative root is thus unavoidable — and it immediately creates two distinct crises. The first is about energy: the spectrum of the Hamiltonian has no floor. The Dirac Hamiltonian makes this brutally explicit.

H_{\text{Dirac}} = c\,\boldsymbol\alpha\cdot\mathbf p + \beta mc^2,\qquad \mathrm{spec}(H)=\{E:\,|E|\ge mc^2\}=(-\infty,-mc^2]\cup[mc^2,+\infty) \qquad (2)

where: $\boldsymbol\alpha,\beta$ — Hermitian $4\times4$ matrices with $\alpha^i=\gamma^0\gamma^i$ , $\beta=\gamma^0$ ; the spectrum splits into two continua with a gap $2mc^2$ . Reading it physically: the lower continuum reaches to $-\infty$ , so there is no lowest state — a positive-energy electron coupled to the electromagnetic field could tumble down through it forever.

Go deeper · Derivation B · 1 minThe free Dirac Hamiltonian squares to a scalar

Derivation BThe free Dirac Hamiltonian squares to a scalar

The goal is to prove that the Dirac Hamiltonian, despite being a $4\times4$ matrix of first-order operators, has eigenvalues exactly $\pm\sqrt{\mathbf p^2c^2+m^2c^4}$ — the same two signs as the scalar dispersion — so that the spectrum is genuinely unbounded below. The strategy is to square $H$ and use only the anticommutation relations of $\boldsymbol\alpha$ and $\beta$ ; nothing about the explicit form of the matrices is needed. Watch the cross terms: they must vanish for the result to collapse to a multiple of the identity.

Step 1. Write $H=c\,\alpha^i p_i+\beta mc^2$ with $p_i$ c-numbers on a plane wave.

Step 2. Square it: $H^2 = c^2\alpha^i\alpha^j p_ip_j + mc^3(\alpha^i\beta+\beta\alpha^i)p_i + \beta^2 m^2c^4$ .

Step 3. Apply the Clifford relations $\{\alpha^i,\alpha^j\}=2\delta^{ij}$ , $\{\alpha^i,\beta\}=0$ , $\beta^2=\mathbb 1$ . The cross term dies because $\{\alpha^i,\beta\}=0$ ; the momentum sum symmetrizes to $\alpha^i\alpha^j p_ip_j=\tfrac12\{\alpha^i,\alpha^j\}p_ip_j=\mathbf p^2$ . Why the symmetrization is legitimate: $p_ip_j$ is symmetric in $(i,j)$ , so only the symmetric (anticommutator) part of $\alpha^i\alpha^j$ survives the contraction.

Step 4. Collapse: $H^2=(\mathbf p^2c^2+m^2c^4)\,\mathbb 1$ , a multiple of the $4\times4$ identity. Taking eigenvalues, $E=\pm\sqrt{\mathbf p^2c^2+m^2c^4}$ — each sign doubly degenerate (the two spin states). Verified: with the explicit Dirac-representation $\alpha^i,\beta$ , $H^2=(\mathbf p^2c^2+m^2c^4)\mathbb 1$ and the four eigenvalues are $\{-E,-E,+E,+E\}$ — checked numerically in verify_aq-qm-negative-energy-solutions.py.

The second crisis is about probability, and it is specific to Klein–Gordon. Klein-Gordon Probability Current and Negative Densities owns the full story; the one fact we borrow is that its conserved density evaluated on a plane wave is proportional to the energy.

\rho = \frac{E}{mc^2}\,|N|^2 \;\;\xrightarrow{\;E<0\;}\;\; \rho<0 \qquad (3)

where: $\rho$ — the conserved Klein–Gordon density; $N$ — the plane-wave normalization. Reading it physically: because $\rho$ tracks the sign of $E$ , negative-energy solutions carry negative "probability" — a contradiction that tells you $\phi$ is a field amplitude, not a single-particle probability amplitude. Contrast the Schrödinger current, which Probability Current and the Continuity Equation shows is positive-definite; that positivity is exactly what relativistic covariance costs you. The resolution to both crises is a single move: stop treating the negative-energy solution as a particle. Charge conjugation maps it to a positive-energy antiparticle, and second quantization builds a Hamiltonian bounded below from the ground up. We state these resolutions here; constructing quantum field theory is the job of later articles.

\hat H = \sum_{\mathbf p,s}E_{\mathbf p}\!\left(a^\dagger_{\mathbf p,s}a_{\mathbf p,s}+b^\dagger_{\mathbf p,s}b_{\mathbf p,s}\right)+\text{const},\qquad E_{\mathbf p}=+\sqrt{\mathbf p^2c^2+m^2c^4} \qquad (4)

where: $a^\dagger_{\mathbf p,s}$ — creates a particle; $b^\dagger_{\mathbf p,s}$ — creates an antiparticle; every term carries $E_{\mathbf p}>0$ . Reading it physically: the negative-frequency mode's coefficient is reinterpreted as an antiparticle creation operator, so the field energy is a sum of non-negative pieces — the spectrum is bounded below by the vacuum, and the crisis evaporates. This is the modern answer; the historical bridge to it, The Dirac Sea and Antiparticles, keeps the single-particle language one step longer.

Limiting cases

How the negative-energy problem shows its face across regimes:

Case	Result	Meaning
Non-relativistic ( $pc\ll mc^2$ )	$E_+\approx mc^2+\dfrac{\mathbf p^2}{2m}$	The positive root reduces to rest energy plus ordinary kinetic energy; the negative branch sits a finite $2mc^2$ away and decouples — which is exactly why the Schrödinger equation never met the problem.
Massless ( $m\to0$ )	$E=\pm\|\mathbf p\|c$	The gap closes: the two continua touch at the origin, so negative modes lie arbitrarily close to positive ones and no single-particle interpretation survives.
Sharp/strong step (Klein paradox)	$V_0>E+mc^2$	A potential step taller than $E+mc^2$ gives "transmitted" waves moving into the barrier and reflection exceeding unity — sensible only once the interior states are read as antiparticles (spontaneous pair creation).
Field-theory limit	$b^\dagger_{\mathbf p}$ replaces "negative-energy particle"	Second quantization assigns the negative-frequency mode an antiparticle creation operator; $\hat H=\sum E_{\mathbf p}(a^\dagger a+b^\dagger b)$ is manifestly bounded below.

Common trapThe two crises are not the same crisis — and Dirac fixed only one

The single most common error is to "just keep the positive-energy solutions." That is illegal: the positive-energy subspace is not preserved by localization, by a Lorentz boost in the presence of interactions, or by any potential; and the negative modes are needed for completeness. The second, subtler error is conflating the two distinct problems. Negative energy is a spectrum/stability problem (no ground state) and afflicts both Klein–Gordon and Dirac. Negative density is a probability problem specific to Klein–Gordon. The Dirac equation's positive-definite density $\rho=\psi^\dagger\psi\ge0$ cures the density — but its Hamiltonian still has a negative continuum, so the spectrum crisis is untouched. Fixing one is not fixing the other.

Worked examples

Example 1 — The Zitterbewegung frequency: how fast do the two sectors interfere? A localized Dirac wave packet contains both positive- and negative-energy amplitudes, and their interference beats at a frequency set by the energy gap. Let us compute it for an electron and see why it is unobservably fast.

Step 1. The energy separation between the positive- and negative-energy branches at rest is the gap $\Delta E = E_+-E_- = 2mc^2$ . A superposition of two energy eigenstates separated by $\Delta E$ oscillates at angular frequency $\omega=\Delta E/\hbar$ .

Step 2. Hence the Zitterbewegung frequency is $\omega_Z=2mc^2/\hbar$ . Insert the electron rest energy $mc^2=0.511\,\mathrm{MeV}=8.187\times10^{-14}\,\mathrm J$ and $\hbar=1.055\times10^{-34}\,\mathrm{J\,s}$ :

\omega_Z=\frac{2\times8.187\times10^{-14}}{1.055\times10^{-34}}\ \mathrm{s^{-1}}.

Step 3. Evaluate: $\omega_Z=\mathbf{1.55\times10^{21}\ s^{-1}}$ , a period $T_Z=2\pi/\omega_Z=\mathbf{4.0\times10^{-21}\ s}$ . This is roughly a thousand times faster than the fastest attosecond pulse, which is why the trembling motion has never been directly resolved for a real electron — though it has been simulated in trapped-ion and cold-atom analog systems.

Example 2 — Why localizing an electron forces in the negative-energy tail. Attempt to confine an electron to a box the size of its Compton wavelength and see what momentum, hence what energy sector, you are forced into.

Step 1. The Compton wavelength is $\lambda_C=\hbar/mc$ . Numerically, $\lambda_C=(1.055\times10^{-34})/(9.109\times10^{-31}\times2.998\times10^8)=\mathbf{3.86\times10^{-13}\ m}$ .

Step 2. By the uncertainty principle, localizing to $\Delta x\sim\lambda_C$ requires momentum spread $\Delta p\sim\hbar/\Delta x=\hbar/\lambda_C=mc$ . The corresponding energy scale is $pc\sim mc\cdot c=mc^2$ .

Step 3. So the packet necessarily contains momentum components with $pc\sim mc^2=\mathbf{0.511\ MeV}$ — squarely relativistic, where the $E>0$ and $E<0$ amplitudes are strongly mixed. There is no way to keep the state both this localized and purely positive-energy: enforcing positive energy spreads the packet back out to at least a Compton wavelength. This is the kinematic origin of both Zitterbewegung and the impossibility of a sharp single-particle position for a relativistic electron.

Origins · 1 mindirac sea and how this was discovered

dirac sea

concept introduced 1930 by P.A.M. Dirac · A picture that turned a crisis into a prediction

The Dirac sea is Dirac's 1930 response to the negative-energy catastrophe: imagine every negative-energy state already filled by electrons, so the Pauli exclusion principle forbids a positive-energy electron from cascading downward. A hole in this otherwise-full sea then behaves as a particle of positive energy and opposite charge. Dirac first guessed the hole was the proton, but the mass had to match the electron's — forcing the prediction of a new particle. Ernst Stückelberg (1941) and Richard Feynman (1949) later replaced the sea entirely with a cleaner idea: a negative-energy solution is a positive-energy antiparticle propagating backward in time, an interpretation that works for bosons too, where no exclusion principle and hence no sea exists.

History

1926

— Schrödinger derives the relativistic (later "Klein–Gordon") equation first, but discards it because it gives the wrong hydrogen fine structure, and publishes his non-relativistic equation instead

1926

— Klein and Gordon independently publish the spin-0 relativistic wave equation that now bears their names

1928

— Dirac finds a first-order equation with positive-definite density, but the negative-energy solutions remain

1930

— Dirac proposes hole theory: the filled negative-energy sea, with holes as positively charged particles

1932

— Anderson discovers the positron in cosmic-ray cloud-chamber tracks, confirming Dirac's prediction

1941

— Stückelberg reinterprets negative-energy solutions as antiparticles, later sharpened by Feynman in 1949

Going further

References. Bjorken & Drell, Relativistic Quantum Mechanics, Ch. 3–5 · Sakurai & Napolitano, Modern Quantum Mechanics, 2nd ed., Ch. 8 (Relativistic QM) · Griffiths, Introduction to Elementary Particles, 2nd ed., Ch. 7 · Shankar, Principles of Quantum Mechanics, 2nd ed., Ch. 20 · Cohen-Tannoudji, Diu & Laloë, Quantum Mechanics, Vol. 3 (relativistic appendices) · Dirac, Proc. R. Soc. Lond. A 117, 610 (1928); A 126, 360 (1930) · Anderson, Phys. Rev. 43, 491 (1933).

Provenance. All equations are standard relativistic QM results; a verifier should recompute: (1) the two roots of $E^2=\mathbf p^2c^2+m^2c^4$ and the mass gap $2mc^2=1.022$ MeV for the electron; (2) the root-count-equals-order argument establishing completeness (the second-order KG time ODE has a two-parameter solution space); (3) $H_{\text{Dirac}}^2=(\mathbf p^2c^2+m^2c^4)\mathbb 1$ from the $\alpha,\beta$ algebra, hence eigenvalues $\pm\sqrt{\mathbf p^2c^2+m^2c^4}$ ; (4) the Zitterbewegung frequency $\omega_Z=2mc^2/\hbar\approx1.55\times10^{21}\,\mathrm{s^{-1}}$ and Compton wavelength $\lambda_C=\hbar/mc=3.86\times10^{-13}$ m. These are checked in verify_aq-qm-negative-energy-solutions.py. No non-standard numbers are introduced. Historical claims (Schrödinger's discarded 1926 relativistic equation, Dirac 1928/1930, Anderson 1932 positron) are textbook-standard.

Builds on: The Klein-Gordon Equation · Klein-Gordon Probability Current and Negative Densities · The Dirac Equation · Probability Current and the Continuity Equation Leads to: The Dirac Sea and Antiparticles · Charge Conjugation and the Positron · Dirac Spinors and Plane-Wave Solutions

If the negative-energy solutions are unphysical, why do they appear in the first place?

They appear because the relativistic dispersion relation $E^2=\mathbf p^2c^2+m^2c^4$ is quadratic in $E$ , and any wave equation reproducing it must be at least second order in time (Klein–Gordon) or share its squared content (Dirac). A quadratic has two roots; the equation cannot tell them apart because it only "knows" $E^2$ . Calling them unphysical is premature: they are perfectly good mathematical solutions and, once reinterpreted as antiparticles, perfectly physical ones. The lesson is that the naive single-particle reading is what fails, not the solutions themselves.

Doesn't the Dirac equation solve the negative-energy problem? It's supposed to be the fix.

It solves half of it. Dirac's motivation was the negative density of Klein–Gordon, and a first-order equation does give a positive-definite density $\rho=\psi^\dagger\psi\ge0$ . But squaring the Dirac Hamiltonian shows its eigenvalues are still $\pm\sqrt{\mathbf p^2c^2+m^2c^4}$ , so the spectrum is just as unbounded below as before. Dirac himself recognized this and invented the filled negative-energy sea to stabilize it. The genuine, universal fix is second quantization, which works for bosons (no sea possible) as well as fermions.

Why can't I just project onto the positive-energy subspace once and be done?

Because that projector is not preserved by the dynamics. In free space it is fine — a positive-energy state stays positive — but a Lorentz boost in the presence of a potential, or localization below the Compton wavelength, mixes the two sectors. The positive-energy projector is also nonlocal on the scale $\hbar/mc$ : enforcing pure positive energy smears the particle over at least a Compton wavelength. The Foldy–Wouthuysen transformation does decouple the sectors, but only approximately, order by order in $1/c$ — its very existence as an expansion, rather than an exact map, quantifies the impossibility of a clean split.

What exactly is the difference between the "energy crisis" and the "density crisis"?

The energy crisis is that the Hamiltonian's spectrum runs to $-\infty$ , so there is no ground state and matter is nominally unstable against radiative cascade. It afflicts both Klein–Gordon and Dirac. The density crisis is that the Klein–Gordon conserved density $\rho\propto E$ goes negative for $E<0$ , so it cannot be a probability density — this one is specific to Klein–Gordon, whose second-order-in-time structure makes $\rho$ depend on $\dot\phi$ independently. Dirac's first-order equation cures the density but not the spectrum. Keeping these two straight is the single most useful thing to remember about this topic.

Is Zitterbewegung a real, observable motion of the electron?

Not for a free electron in the lab: the frequency $\omega_Z=2mc^2/\hbar\approx1.55\times10^{21}\,\mathrm{s^{-1}}$ (period $\sim4\times10^{-21}$ s) and amplitude $\sim\lambda_C\approx4\times10^{-13}$ m are far beyond current time resolution, and in a full field theory the effect is largely absorbed into renormalization. But it has been observed in engineered analog systems whose low-energy excitations obey a Dirac-like equation — trapped ions, ultracold atoms in optical lattices, and electrons in graphene — where the effective mass gap and "speed of light" bring the trembling into an accessible regime. So it is real physics, just not directly seen in the electron it was invented for.

How does quantum field theory actually make the energy bounded below?

When you expand the field operator, the negative-frequency modes come multiplied by coefficients you are free to interpret. In QFT you call those coefficients antiparticle creation operators $b^\dagger$ rather than "negative-energy particle" operators. After normal ordering, the Hamiltonian reads $\hat H=\sum_{\mathbf p,s}E_{\mathbf p}(a^\dagger a+b^\dagger b)$ with every $E_{\mathbf p}>0$ , so it is bounded below by the vacuum. A would-be negative-energy electron is now a genuine positive-energy positron. The price is that particle number is no longer conserved — which is exactly the physics the Klein paradox was pointing at.

Why does localizing a particle force in negative-energy components — where does that come from mathematically?

A state localized to width $\Delta x$ has, by Fourier analysis, momentum content up to $\Delta p\sim\hbar/\Delta x$ . When $\Delta x\lesssim\hbar/mc$ (the Compton wavelength), this means $\Delta p\gtrsim mc$ , i.e. relativistic momenta. Expand the localized state in the complete set of plane-wave solutions $\{u(\mathbf p)e^{i\mathbf p\cdot\mathbf x/\hbar}, v(\mathbf p)e^{-i\mathbf p\cdot\mathbf x/\hbar}\}$ ; demanding the negative-energy weights $b(\mathbf p)$ all vanish while achieving that sharp a profile is inconsistent, because the positive-energy spinors/amplitudes alone cannot build a function that narrow. So $b(\mathbf p)\neq0$ is forced — the sectors mix, and that mixing is Zitterbewegung in disguise.