Four-Force (Minkowski Force)

Here is a four-vector whose Minkowski length is negative, and it is not a pathology — it is a theorem. Take a particle of constant rest mass. Its four-velocity $U^\mu$ is timelike, with the fixed length $U\cdot U = +c^2$ . Its four-force $K^\mu = \mathrm{d}p^\mu/\mathrm{d}\tau$ is Minkowski-orthogonal to it, $K\cdot U = 0$ . Go to the particle's instantaneous rest frame, where $U^\mu = (c, \vec 0\,)$ . Orthogonality there reads $K^0c = 0$ , so $K^0 = 0$ : the four-force has no time component in the rest frame. It is a purely spatial four-vector — and in the signature $(+,-,-,-)$ a purely spatial four-vector has a negative norm, automatically, with nothing left to argue about. So $K\cdot K \le 0$ . Always. The four-force of a mass-preserving force is spacelike, for a reason that never mentions the force: being orthogonal to a timelike vector leaves you nowhere to go but outside the light cone. In components that reads $K\cdot K = \gamma^2\!\left(\dfrac{(\vec F\cdot\vec v)^2}{c^2} - \vec F^{\,2}\right)$ , and since $|\vec F\cdot\vec v| \le Fv < Fc$ the bracket is negative for any real force on any massive particle. Worth stating bluntly, because the AllQFT base article for this topic prints the opposite sign — it publishes $K\cdot K = \gamma^2(\vec F^{\,2} - (\vec F\cdot\vec v)^2/c^2)$ , a positive quantity, while its own derivation note underneath says "the scalar is negative (spacelike)". The card contradicts itself. That is not a physics error but a transcription error, the commonest defect in this corpus, and correcting it in public is part of why this article exists. The corrected sign also hides a payoff: $-K\cdot K$ is the square of the force the particle actually feels — the proper force $F_0$ , measured in its own rest frame. So $K\cdot K = -F_0^2$ , an invariant with a physical name.

Key takeaways

Signature $(+,-,-,-)$ , $\eta = \mathrm{diag}(+1,-1,-1,-1)$ , $\;p\cdot p = +m^2c^2$ , $\;U\cdot U = +c^2$ . Plus.
The four-force is the proper-time derivative of the four-momentum: $\boxed{K^\mu = \dfrac{\mathrm{d}p^\mu}{\mathrm{d}\tau} = \left(\dfrac{\gamma}{c}\,\vec F\cdot\vec v,\;\; \gamma\vec F\right)}$ . Its spatial part is $\gamma\vec F$ , not $\vec F$ .
For constant rest mass, $\boxed{K^\mu = mA^\mu}$ — the covariant Newton's second law, with $A^\mu$ the four-acceleration and $m$ the invariant rest mass.
Owned identity: $\boxed{K\cdot U = c^2\,\mathrm{d}m/\mathrm{d}\tau}$ , always. So the famous orthogonality $K\cdot U = 0$ is not an extra assumption — it is exactly equivalent to the rest mass being constant. A pure force cannot change a particle's rest mass, and if something does change it, the four-force is not orthogonal to $U$ .
Owned result, and the base's error corrected: $\boxed{K\cdot K = \gamma^2\!\left(\dfrac{(\vec F\cdot\vec v)^2}{c^2} - \vec F^{\,2}\right) \le 0}$ — SPACELIKE. The base prints this with the sign reversed while its own note calls it negative.
The invariant has a name: $\boxed{K\cdot K = -F_0^{\,2}}$ , where $F_0 = \gamma\sqrt{\vec F^{\,2} - (\vec F\cdot\vec v)^2/c^2}$ is the proper force — the three-force in the particle's own rest frame.
Key idea: a purely longitudinal force has $F_0 = F$ — the particle feels exactly what the lab applies. A purely transverse force has $F_0 = \gamma F$ — the particle feels $\gamma$ times more than the lab applies. That asymmetry, not any "mass", is what a fast particle actually experiences.

The core insight

Orthogonality to a timelike vector is a severe constraint: it confines you to the three-dimensional spacelike hyperplane perpendicular to it, and every non-zero vector in that hyperplane has negative norm. So the moment you know $K\cdot U = 0$ and $U\cdot U = +c^2$ , you know $K\cdot K \le 0$ without computing anything. The sign of the four-force invariant is fixed by geometry before any force is specified — which is why a printed positive value is detectably wrong on sight.

Where this lives today

The four-force is how every covariant equation of motion is written, and its orthogonality is a working constraint rather than a decoration. The electromagnetic case is the archetype: the Lorentz force in four-vector form is $K^\mu = (q/c)F^{\mu\nu}U_\nu$ , and because the field tensor $F^{\mu\nu}$ is antisymmetric, the contraction $K^\mu U_\mu = (q/c)F^{\mu\nu}U_\nu U_\mu$ vanishes identically — an antisymmetric tensor contracted with a symmetric pair of indices always does. So electromagnetism cannot change a particle's rest mass, and it cannot do so by construction, not by accident. That derivation belongs to the relativistic-electrodynamics band; the orthogonality it leans on is on this page. The constraint bites hardest where new terms are added to the equation of motion. Radiation reaction is the classic case: an accelerating charge radiates, and the recoil needs some extra four-force — but any such term must keep $K\cdot U = 0$ , or the electron's rest mass would slowly drift, which it does not. Dirac's 1938 solution is built to satisfy exactly that, and the same requirement constrains the models used today to simulate electrons in the focus of a petawatt laser, where radiation reaction is no longer a small correction. And the proper force $F_0 = \sqrt{-K\cdot K}$ is what matters physically in accelerator design: a relativistic proton in a bending magnet feels $\gamma$ times the lab force, and it is that amplified force which sets how hard the particle is shaken and therefore how much synchrotron radiation it throws away. The sign of $K\cdot K$ is no bookkeeping curiosity — get it wrong and the quantity you extract from it is imaginary.

The physical picture

The three-force $\vec F = \mathrm{d}\vec p/\mathrm{d}t$ is a perfectly good object, but it is not the spatial part of any four-vector, because $\mathrm{d}t$ is a coordinate and not a scalar. The repair is the one that built the four-velocity: divide by the invariant proper time instead. $K^\mu \equiv \mathrm{d}p^\mu/\mathrm{d}\tau$ is then manifestly a four-vector — a four-vector differentiated by a scalar — and its components are the old ones with a $\gamma = \mathrm{d}t/\mathrm{d}\tau$ attached. That single $\gamma$ is the whole difference, and it is why the spatial part of $K^\mu$ is $\gamma\vec F$ rather than $\vec F$ . The time part is new: $\gamma/c$ times the power $\vec F\cdot\vec v$ , so the four-force packages the force law and the work–energy theorem into one object, exactly as the four-momentum packages momentum and energy. Then geometry takes over. $U^\mu$ points into the future along the worldline and has fixed length; $K^\mu$ , for a mass-preserving force, is perpendicular to it. Perpendicular to a timelike direction means entirely within the particle's space of simultaneity — and that space is spacelike. There is no room for $K^\mu$ to have a positive norm.

Figure

PUNCHLINE — the four-force is SPACELIKE, and geometry forces it. U is timelike (U·U = +c²) and points up the worldline. K is Minkowski-orthogonal to it, so in the rest frame K has NO time component: K = (0, F₀), purely spatial — and a purely spatial four-vector has NEGATIVE norm in (+,−,−,−). Hence K·K = −F₀² ≤ 0 before any force is even named. Worked case F = (3,4,0) N at v = 0.6c gives K·K = −34 N², not +34

The core equations

The definition is forced: differentiate the four-momentum by the only scalar available.

\boxed{\;K^\mu \;\equiv\; \frac{\mathrm{d}p^\mu}{\mathrm{d}\tau} \;=\; \gamma\,\frac{\mathrm{d}p^\mu}{\mathrm{d}t} \;=\; \left(\frac{\gamma}{c}\,\vec F\cdot\vec v,\;\; \gamma\vec F\right)\;} \qquad (1)

where: $\tau$ — proper time; $\gamma = \mathrm{d}t/\mathrm{d}\tau$ ; $\vec F = \mathrm{d}\vec p/\mathrm{d}t$ is the ordinary three-force. Physical reading: $p^\mu$ is a four-vector and $\mathrm{d}\tau$ a scalar, so $K^\mu$ is a four-vector — that is the entire justification, and it is why $\mathrm{d}p^\mu/\mathrm{d}t$ would not be one. The time component is $\gamma/c$ times the power $\vec F\cdot\vec v = \mathrm{d}E/\mathrm{d}t$ , so the four-force carries the work–energy theorem in its $\mu = 0$ slot exactly as $p^\mu$ carries the energy. For constant rest mass, $p^\mu = mU^\mu$ and $m$ passes through the derivative:

K^\mu = m\,\frac{\mathrm{d}U^\mu}{\mathrm{d}\tau} = m A^\mu \qquad (m = \text{const}) \qquad (2)

where: $A^\mu = \mathrm{d}U^\mu/\mathrm{d}\tau$ — the four-acceleration. Physical reading: Newton's second law, made manifestly covariant. It looks like $\vec F = m\vec a$ and, unlike $\vec F = m\vec a$ , it is true, because $K^\mu$ and $A^\mu$ are genuine four-vectors and the equation holds in every frame at once. The three-vector version failed precisely because $\vec F$ and $\vec a$ are not parallel; the four-vector version succeeds because $K^\mu$ and $A^\mu$ are. The orthogonality is usually stated as a condition. It is better stated as an identity:

\boxed{\;K\cdot U \;=\; c^2\,\frac{\mathrm{d}m}{\mathrm{d}\tau}\;} \qquad\Longrightarrow\qquad K\cdot U = 0 \iff m = \text{const} \qquad (3)

where: $m$ is the invariant rest mass, permitted here to vary so we can see what its variation costs. Physical reading: orthogonality is not an assumption bolted onto the four-force — it is equivalent to rest-mass conservation. A pure force cannot change rest mass; anything that does (a rocket burning fuel, a body being heated) is not describable by a mass-preserving $K^\mu$ alone. And the invariant — the equation the base gets wrong:

\boxed{\;K\cdot K \;=\; \gamma^2\!\left(\frac{(\vec F\cdot\vec v)^2}{c^2} - \vec F^{\,2}\right) \;=\; -\,F_0^{\,2} \;\le\; 0\;}, \qquad F_0 = \gamma\sqrt{\vec F^{\,2} - \frac{(\vec F\cdot\vec v)^2}{c^2}} \qquad (4)

where: $F_0$ — the proper force, the magnitude of the three-force in the particle's instantaneous rest frame. Physical reading: spacelike, always. Since $|\vec F\cdot\vec v| \le Fv < Fc$ for any massive particle, the bracket is strictly negative unless $\vec F = \vec 0$ . And $-K\cdot K$ is no abstraction: it is the squared force the particle feels. Note the sign — in $(-,+,+,+)$ this flips to $+F_0^2$ , which is one of the ways the error gets in.

Why the four-force is spacelike

Derivation ABuilding the four-force, and finding the work-energy theorem in its time slot

One rule forces the construction: to make a four-vector, divide a four-vector by a Lorentz scalar. The price is a single $\gamma$ , which then shows up in every component.

Step 1. Convert the derivative. $\;\dfrac{\mathrm{d}}{\mathrm{d}\tau} = \dfrac{\mathrm{d}t}{\mathrm{d}\tau}\dfrac{\mathrm{d}}{\mathrm{d}t} = \gamma\dfrac{\mathrm{d}}{\mathrm{d}t}$ , since $\mathrm{d}t = \gamma\,\mathrm{d}\tau$ (time dilation). Why: this is the only input. Everything below is bookkeeping on that factor.

Step 2. Do the spatial components. $\;K^i = \gamma\,\dfrac{\mathrm{d}p^i}{\mathrm{d}t} = \gamma F^i$ , so $\vec K = \gamma\vec F$ . Why: the spatial part of the four-force is $\gamma\vec F$ , not $\vec F$ . This is the single most common slip in the subject, and it is why $K^\mu = mA^\mu$ does not contradict $\vec F \ne m\vec a$ .

Step 3. Do the time component. $\;K^0 = \gamma\,\dfrac{\mathrm{d}p^0}{\mathrm{d}t} = \gamma\,\dfrac{\mathrm{d}}{\mathrm{d}t}\!\left(\dfrac{E}{c}\right) = \dfrac{\gamma}{c}\dfrac{\mathrm{d}E}{\mathrm{d}t} = \dfrac{\gamma}{c}\,\vec F\cdot\vec v$ , using the work–energy theorem $\mathrm{d}E/\mathrm{d}t = \vec F\cdot\vec v$ from Relativistic Force. Why: the $\mu=0$ component is the work–energy theorem. Energy and momentum were unified by $p^\mu$ ; the force law and the power law are unified by $K^\mu$ .

Step 4. Get the covariant second law. If $m$ is constant, $p^\mu = mU^\mu$ gives $K^\mu = \mathrm{d}(mU^\mu)/\mathrm{d}\tau = m\,\mathrm{d}U^\mu/\mathrm{d}\tau = mA^\mu$ . Why: $m$ is a scalar and does not get differentiated. This restores the shape of $\vec F = m\vec a$ that the three-vector law destroyed, because $K^\mu$ and $A^\mu$ carry the very $\gamma$ 's that made $\vec F$ and $\vec a$ disagree. *Verified in verify_aq-sr-four-force.py: $K^\mu = \gamma\,\mathrm{d}p^\mu/\mathrm{d}t$ from $\mathrm{d}t = \gamma\mathrm{d}\tau$ ; $K^0 = (\gamma/c)\vec F\cdot\vec v$ with $E = \gamma mc^2$ ; $\vec K = \gamma\vec F$ ; and $K^\mu = mA^\mu$ symbolically for arbitrary $\vec v(\tau)$ at constant $m$ .*

Go deeper · Derivation B · 1 minOrthogonality is not an assumption. It is exactly rest-mass conservation.

Derivation BOrthogonality is not an assumption. It is exactly rest-mass conservation.

Textbooks introduce $K\cdot U = 0$ with the caveat "for constant rest mass" and leave it there. That caveat does more work than it looks: the orthogonality and the constancy are the same statement, and the identity behind them says exactly what a non-orthogonal four-force is doing.

Step 1. Start from the mass shell, allowing $m$ to vary. $\;p\cdot p = m^2c^2$ , with $m = m(\tau)$ for now. Why: we are not assuming constancy — we are asking what constancy costs.

Step 2. Differentiate with respect to proper time. The metric is constant, so it passes through: $\;\dfrac{\mathrm{d}}{\mathrm{d}\tau}(p\cdot p) = 2\,p\cdot\dfrac{\mathrm{d}p}{\mathrm{d}\tau} = 2\,p\cdot K$ . And the right side gives $\dfrac{\mathrm{d}}{\mathrm{d}\tau}(m^2c^2) = 2mc^2\dfrac{\mathrm{d}m}{\mathrm{d}\tau}$ . Hence $\;p\cdot K = mc^2\,\mathrm{d}m/\mathrm{d}\tau$ . Why: the product rule on a Minkowski inner product works exactly as in Euclidean space, because $\eta$ has constant components.

Step 3. Replace $p$ by $mU$ . $\;p\cdot K = m\,(U\cdot K)$ , so $m(U\cdot K) = mc^2\,\mathrm{d}m/\mathrm{d}\tau$ , and dividing by $m \ne 0$ : $\;\boxed{K\cdot U = c^2\,\mathrm{d}m/\mathrm{d}\tau}$ . Why: an identity, not a condition. It holds whether or not the mass is constant, and it says what the non-orthogonal part of the four-force is for: it is changing the rest mass.

Step 4. Read off both directions. If $m$ is constant, $\mathrm{d}m/\mathrm{d}\tau = 0$ and $K\cdot U = 0$ . Conversely, if $K\cdot U = 0$ then the rest mass cannot change. A mass-preserving force is orthogonal to the four-velocity, and an orthogonal four-force preserves mass. Why: this is the precise sense in which "a pure force cannot change rest mass". A rocket burning fuel, or a body being heated, is not a mass-preserving four-force — and $K^\mu = mA^\mu$ fails there too, acquiring an extra term $(\mathrm{d}m/\mathrm{d}\tau)U^\mu$ .

Step 5. Sanity-check in components. $\;K\cdot U = \dfrac{\gamma}{c}(\vec F\cdot\vec v)\cdot\gamma c - (\gamma\vec F)\cdot(\gamma\vec v) = \gamma^2(\vec F\cdot\vec v) - \gamma^2(\vec F\cdot\vec v) = 0$ for constant $m$ . Why: the cancellation is the work–energy theorem cancelling against itself — $K^0$ was built from $\vec F\cdot\vec v$ , and the spatial contraction hands the same quantity back with the metric's minus sign. *Verified in verify_aq-sr-four-force.py: $K\cdot U = c^2\,\mathrm{d}m/\mathrm{d}\tau$ symbolically with $m(\tau)$ and $\vec v(\tau)$ arbitrary in 3D; $K\cdot U = 0$ identically at constant $m$ ; and the componentwise cancellation.*

Go deeper · Derivation C · 2 minThe invariant is minus the proper force squared, and it cannot be positive

Derivation CThe invariant is minus the proper force squared, and it cannot be positive

The article's owned result, and the repair of the base's defect. Two independent routes to the sign, both worth having: a geometric one that needs no algebra and settles the matter in two lines, and a componentwise one that produces the formula and its meaning. When they agree, a printed positive sign is not a matter of convention — it is simply wrong.

Step 1. The geometric route. Go to the instantaneous rest frame, where $U^\mu = (c, \vec 0\,)$ . Orthogonality gives $K\cdot U = K^0c - \vec K\cdot\vec 0 = K^0 c = 0$ , so $\;K^0 = 0$ . The four-force is purely spatial there: $K^\mu = (0,\ \vec F_0)$ , where $\vec F_0$ is by definition the three-force in the rest frame — the proper force. Why: no force has been specified. This holds for any mass-preserving four-force whatsoever.

Step 2. Take its norm in that frame. $\;K\cdot K = (K^0)^2 - |\vec K|^2 = 0 - F_0^{\,2} = \boxed{-F_0^{\,2}}$ . And $K\cdot K$ is an invariant, so this is its value in every frame. Why: in $(+,-,-,-)$ , a purely spatial four-vector has negative norm — that is what the signature means. So $K\cdot K \le 0$ is forced, with equality only if the proper force vanishes. The sign is a theorem about orthogonality to a timelike vector, not a computation about a particular force.

Step 3. The componentwise route, for the formula. With $K^\mu = \big((\gamma/c)\vec F\cdot\vec v,\ \gamma\vec F\big)$ : $\;K\cdot K = \dfrac{\gamma^2(\vec F\cdot\vec v)^2}{c^2} - \gamma^2\vec F^{\,2} = \boxed{\gamma^2\!\left(\dfrac{(\vec F\cdot\vec v)^2}{c^2} - \vec F^{\,2}\right)}$ . Why: note the order of the terms. The power term is positive and the force term is negative — the opposite of what the base article prints.

Step 4. Confirm the sign from the Cauchy–Schwarz bound. $\;|\vec F\cdot\vec v| \le F v < F c$ for any massive particle ( $v < c$ ), so $(\vec F\cdot\vec v)^2/c^2 < \vec F^{\,2}$ and the bracket is strictly negative whenever $\vec F \ne \vec 0$ . Why: there is no force, and no speed short of $c$ , that makes this positive. A positive value would require $v \ge c$ and $\vec F \parallel \vec v$ — i.e. it is not merely unusual, it is unreachable.

Step 5. Match the two routes and extract the proper force. Equating: $\;F_0^{\,2} = \gamma^2\!\left(\vec F^{\,2} - \dfrac{(\vec F\cdot\vec v)^2}{c^2}\right)$ , so $\;F_0 = \gamma\sqrt{\vec F^{\,2} - (\vec F\cdot\vec v)^2/c^2}$ . Why: the invariant now has a physical name — it is (minus) the square of the force the particle feels in its own frame, which is what an accelerometer riding along would read.

Step 6. Evaluate the two aligned cases, which differ strikingly. Longitudinal ( $\vec F \parallel \vec v$ ): $\vec F\cdot\vec v = Fv$ , so $F_0 = \gamma F\sqrt{1 - \beta^2} = \gamma F/\gamma = \boxed{F}$ — the proper force equals the lab force. Transverse ( $\vec F \perp \vec v$ ): $\vec F\cdot\vec v = 0$ , so $F_0 = \boxed{\gamma F}$ . Why: push a fast particle forwards and it feels precisely your push; push it sideways and it feels $\gamma$ times your push. The lab's transverse force is amplified in the particle's own frame — which is why a bending magnet is felt so violently by a relativistic proton, and why synchrotron radiation is so fierce. *Verified in verify_aq-sr-four-force.py: $K\cdot K = \gamma^2((\vec F\cdot\vec v)^2/c^2 - \vec F^{\,2})$ symbolically in 3D; $K\cdot K < 0$ numerically at 500 random $(\vec F, \vec v)$ pairs with $v < c$ (zero violations); $K\cdot K = -F_0^2$ via an explicit boost to the rest frame; $F_0 = F$ longitudinally and $F_0 = \gamma F$ transversely; and the printed numbers $\vec F = (3,4,0)$ N at $v = 0.6c$ giving $K = (2.25, 3.75, 5.00, 0)$ N, $K\cdot K = -34\ \mathrm{N^2}$ , $F_0 = \sqrt{34} = 5.831$ N. The base's printed $+34$ is asserted to be wrong.*

Figure

The proper force — what the particle actually feels. K·K = −F₀², and F₀ = γ√(F² − (F·v)²/c²). A purely LONGITUDINAL push is felt exactly as applied (F₀ = F); a purely TRANSVERSE push is felt γ times stronger (F₀ = γF). Numbers at v = 0.6c, γ = 1.25, with a 5 N lab force: longitudinal F₀ = 5.00 N, transverse F₀ = 6.25 N, and the general 3-4 case F₀ = 5.831 N

Limiting cases

Case	Result	Meaning
$\beta \to 0$ , $\gamma\to 1$	$K^\mu \to (0,\ \vec F\,)$	Newton. The time slot dies, the spatial slot becomes the ordinary force
$m$ constant	$K\cdot U = 0$ , $\;K^\mu = mA^\mu$	the covariant second law; orthogonality holds
$m$ varying	$K\cdot U = c^2\,\mathrm{d}m/\mathrm{d}\tau \ne 0$	not orthogonal; $K^\mu = mA^\mu + (\mathrm{d}m/\mathrm{d}\tau)U^\mu$
any $\vec F \ne 0$ , $v < c$	$K\cdot K < 0$	spacelike, strictly. There is no exception
$\vec F \parallel \vec v$	$F_0 = F$ , $\;K\cdot K = -F^2$	the particle feels exactly the lab force
$\vec F \perp \vec v$	$F_0 = \gamma F$ , $\;K\cdot K = -\gamma^2F^2$	the particle feels $\gamma$ times the lab force
$\vec F = \vec 0$	$K^\mu = 0$ , $\;K\cdot K = 0$	free particle; the only case with a null (zero) four-force
signature $(-,+,+,+)$	$K\cdot K = +F_0^2$	the same physics, opposite sign — and one route by which the error enters

Common trapThe sign of K·K. The base article for this topic prints it backwards.

In $(+,-,-,-)$ the correct invariant is

K\cdot K = \gamma^2\!\left(\frac{(\vec F\cdot\vec v)^2}{c^2} - \vec F^{\,2}\right) = -F_0^{\,2} \;\le\; 0 .

The AllQFT base article prints $K\cdot K = \gamma^2\big(\vec F^{\,2} - (\vec F\cdot\vec v)^2/c^2\big)$ — the negation — while the derivation note directly beneath it states "the scalar is negative (spacelike)". The card contradicts its own working. This is the corpus's dominant defect class: the physics was done correctly and then mis-typed into the headline, and nothing downstream caught it. Two ways to catch it yourself, neither of which requires trusting anyone:

Geometry. $K$ is orthogonal to the timelike $U$ , so in the rest frame $K^0 = 0$ and $K$ is purely spatial. A purely spatial four-vector has negative norm in this signature. A positive $K\cdot K$ is therefore impossible before any force is named.

Bounds. $|\vec F\cdot\vec v| \le Fv < Fc$ , so $(\vec F\cdot\vec v)^2/c^2 < \vec F^{\,2}$ always. The bracket cannot come out positive for any real force on any massive particle. Concretely: $\vec F = (3,4,0)\ \mathrm{N}$ at $v = 0.6c$ gives $K\cdot K = \mathbf{-34\ N^2}$ , not $+34\ \mathrm{N^2}$ . If your algebra hands you a positive number here, you have either flipped a sign or silently changed signature.

Common trapThe spatial part of the four-force is γF, not F.

$K^\mu = \big((\gamma/c)\,\vec F\cdot\vec v,\ \gamma\vec F\big)$ . The $\gamma$ on the spatial part is not decoration and is not optional: it is $\mathrm{d}t/\mathrm{d}\tau$ , and dropping it destroys the transformation law, leaving an object that is not a four-vector at all. The confusion is easy to fall into because $K^\mu = mA^\mu$ looks so much like $\vec F = m\vec a$ — but $\vec F = m\vec a$ is false (Relativistic Force shows $\vec F$ and $\vec a$ are not even parallel) while $K^\mu = mA^\mu$ is true, and the reconciliation is exactly the $\gamma$ 's hiding inside $K^\mu$ and $A^\mu$ . If you ever find yourself writing $K^i = F^i$ , you have lost a $\gamma$ and the equation is no longer covariant.

Worked examples

1 · What a 0.9c proton actually feels in a 1 tesla bending magnet

A proton with $\beta = 0.9$ crosses a uniform magnetic field $B = 1.00\ \mathrm{T}$ perpendicular to its motion. Find the four-force, its invariant, and the proper force — the force the proton feels in its own frame. Cross-check against the field the proton sees.

Step 1 — identify. The magnetic force is transverse, so $\vec F\cdot\vec v = 0$ . That kills the time component of $K^\mu$ immediately, and puts us in the transverse case of equation (4): $F_0 = \gamma F$ .

Step 2 — the lab force. $\gamma = 1/\sqrt{1-0.81} = \mathbf{2.2942}$ ; $v = 0.9c = 2.6981\times10^{8}\ \mathrm{m\,s^{-1}}$ . Then $F = qvB = (1.6022\times10^{-19})(2.6981\times10^{8})(1.00) = \mathbf{4.3229\times10^{-11}\ \mathrm{N}}$ .

Step 3 — the four-force. $K^0 = (\gamma/c)(\vec F\cdot\vec v) = \mathbf{0}$ , and $\vec K = \gamma\vec F$ with $|\vec K| = 2.2942 \times 4.3229\times10^{-11} = \mathbf{9.9174\times10^{-11}\ \mathrm{N}}$ . So $K^\mu = (0,\ \gamma\vec F\,)$ — purely spatial in the lab too, because a transverse force does no work in any frame.

Step 4 — the invariant, with its sign. $K\cdot K = (K^0)^2 - |\vec K|^2 = 0 - (9.9174\times10^{-11})^2 = \mathbf{-9.8354\times10^{-21}\ \mathrm{N^2}}$ . Negative — spacelike, as it must be. Hence $F_0 = \sqrt{-K\cdot K} = \mathbf{9.9174\times10^{-11}\ \mathrm{N}} = \gamma F$ .

Step 5 — the boundary number, and an independent check. The proton feels $\gamma = 2.29$ times the force the lab applies. Why? In the proton's rest frame the lab's pure magnetic field appears partly electric, with $E' = \gamma vB = 2.2942 \times 2.6981\times10^{8} \times 1.00 = 6.1899\times10^{8}\ \mathrm{V\,m^{-1}}$ , and the proton feels $qE' = (1.6022\times10^{-19})(6.1899\times10^{8}) = \mathbf{9.9174\times10^{-11}\ \mathrm{N}}$ — identical, to five figures, to $\sqrt{-K\cdot K}$ . Two completely different routes, one number. (The field transformation is the relativistic-electrodynamics band's business; here it serves only as a check that the proper force is real and measurable.)

2 · A general force, and why the base's sign cannot be right

A particle moves at $v = 0.6c$ along $x$ and is pushed with $\vec F = (3, 4, 0)\ \mathrm{N}$ — magnitude $5\ \mathrm{N}$ , at an angle to the motion. Compute $K^\mu$ , $K\cdot K$ and $F_0$ , and compare with what the base article's formula would give.

Step 1 — identify. A general force: neither parallel nor perpendicular to $\vec v$ , so both slots of $K^\mu$ are alive and equation (4) must be used in full.

Step 2 — the ingredients. $\gamma = 1/\sqrt{1 - 0.36} = 1/0.8 = \mathbf{1.25}$ . The power term: $\vec F\cdot\vec v = (3)(0.6c) = 1.8c$ , so $\vec F\cdot\vec v/c = 1.8\ \mathrm{N}$ .

Step 3 — the four-force. $K^0 = (\gamma/c)(\vec F\cdot\vec v) = 1.25 \times 1.8 = \mathbf{2.25\ \mathrm{N}}$ , and $\vec K = \gamma\vec F = 1.25 \times (3,4,0) = \mathbf{(3.75,\ 5.00,\ 0)\ \mathrm{N}}$ .

Step 4 — the invariant. $K\cdot K = (2.25)^2 - (3.75)^2 - (5.00)^2 = 5.0625 - 14.0625 - 25.0000 = \mathbf{-34.00\ \mathrm{N^2}}$ . Negative, as the geometry demanded. The proper force is $F_0 = \sqrt{34} = \mathbf{5.831\ \mathrm{N}}$ — larger than the $5\ \mathrm{N}$ the lab applied, because part of this push is transverse and transverse pushes are amplified by $\gamma$ in the rest frame.

Step 5 — the boundary number, and the defect. The base article's printed formula, $\gamma^2\big(\vec F^{\,2} - (\vec F\cdot\vec v)^2/c^2\big)$ , gives $1.5625 \times (25 - 3.24) = \mathbf{+34.00\ \mathrm{N^2}}$ — the right magnitude with the wrong sign. A positive $K\cdot K$ would make the four-force timelike, which would mean there exists a frame in which it is purely temporal, i.e. a frame in which the particle is being "pushed in the time direction" with no spatial force at all. No such frame exists, and no such force exists. The magnitude was computed correctly and then typed in backwards — which is precisely why the verification script for this article reads the numbers back out of the prose and checks them, rather than trusting the algebra to have survived the journey to the page.

Physics of the portrait

The four-force carries Minkowski's name because he built it, in the same 1908 paper that built the whole four-dimensional apparatus. It is the object that made relativistic dynamics look like dynamics again — a single covariant equation of motion, after three years in which relativity had seemed to demand two different masses and a force that pointed the wrong way.

Origins · 2 minFour-force and how this was discovered

Four-force

1907–present · Minkowski's Grundgleichungen · von Laue's monograph · the Lorentz–Dirac equation

Hermann Minkowski introduced the force four-vector in "Die Grundgleichungen für die elektromagnetischen Vorgänge in bewegten Körpern", presented to the Göttingen scientific society on 21 December 1907 and published in the Göttinger Nachrichten in 1908 — the paper in which four-velocity, four-momentum and the whole four-dimensional calculus appear together for the first time. The mass-preserving force satisfying $K^\mu U_\mu = 0$ has been called the Minkowski force ever since. Its importance was structural rather than computational: Planck had already got relativistic dynamics right in 1906 by keeping $\vec F = \mathrm{d}\vec p/\mathrm{d}t$ , but the result was a law whose form changed from frame to frame, with a force not parallel to the acceleration and two direction-dependent response coefficients. Minkowski's $K^\mu = mA^\mu$ restored a single equation valid in every inertial frame at once, at the modest price of dividing by proper time instead of coordinate time. Max von Laue's Das Relativitätsprinzip (1911), the first monograph on relativity, made the four-force standard pedagogy, and Pauli's Relativitätstheorie (1921) — written at twenty-one, commissioned by Sommerfeld — codified the orthogonality condition in the form still taught. The constraint has teeth to this day: when Dirac derived the classical equation of motion for a radiating electron in 1938, the radiation-reaction term had to be built so the total four-force stayed orthogonal to $U^\mu$ , or the electron's rest mass would drift.

History

1906

— Planck founds relativistic dynamics by keeping Newton's F = dp/dt with p = γmv, but the resulting law changes form between frames.

1907

— Minkowski presents "Die Grundgleichungen für die elektromagnetischen Vorgänge in bewegten Körpern" to the Göttingen scientific society on 21 December; published in the Göttinger Nachrichten in 1908, it introduces the four-velocity, the four-momentum and the force four-vector; the mass-preserving force with K·U = 0 is still called the Minkowski force.

1908

— Minkowski's Cologne lecture "Raum und Zeit", delivered on 21 September, makes the four-dimensional formalism public and famous.

1911

— Max von Laue's Das Relativitätsprinzip, the first monograph on relativity, adopts the four-force throughout and makes the covariant equation of motion standard pedagogy.

1921

— Pauli's encyclopaedia article "Relativitätstheorie", written at twenty-one and commissioned by Sommerfeld, codifies K = dp/dτ and the orthogonality K·U = 0 for constant rest mass.

1938

— Dirac's "Classical theory of radiating electrons" (Proc. Roy. Soc. A 167, 148) adds a radiation-reaction four-force, which must itself respect K·U = 0 or the electron's rest mass would drift.

Going further

Texts. Rindler, Introduction to Special Relativity, 2nd ed., §§35–38 · Griffiths, Introduction to Electrodynamics, 4th ed., §12.2.4 · Jackson, Classical Electrodynamics, 3rd ed., §11.5 · Minkowski, "Die Grundgleichungen für die elektromagnetischen Vorgänge in bewegten Körpern", Gött. Nachr. (1908) · Dirac, "Classical theory of radiating electrons", Proc. Roy. Soc. A 167, 148 (1938).

Provenance. Signature $(+,-,-,-)$ : $\;p\cdot p = +m^2c^2$ , $\;U\cdot U = +c^2$ , and $m$ is the invariant rest mass — no relativistic mass appears anywhere. Base-defect repair (logged): the AllQFT base's eqCard[5] prints $K\cdot K = \gamma^2(\vec F^{\,2} - (\vec F\cdot\vec v)^2/c^2)$ , a positive quantity, while its own derivation note states "the scalar is negative (spacelike)". The printed sign is wrong; the note is right. Corrected here to $K\cdot K = \gamma^2((\vec F\cdot\vec v)^2/c^2 - \vec F^{\,2}) = -F_0^2 \le 0$ , proved twice — geometrically from orthogonality to a timelike vector, and componentwise with a Cauchy–Schwarz bound. Everything above is recomputed in verify_aq-sr-four-force.py, which parses this article's own printed digits.

Scope. The electromagnetic four-force $K^\mu = (q/c)F^{\mu\nu}U_\nu$ and the field transformations are not derived here — they belong to the relativistic-electrodynamics band. The three-force law, the response coefficients $\gamma^3m$ and $\gamma m$ , and the angle between $\vec F$ and $\vec a$ belong to Relativistic Force. Hyperbolic motion under constant proper force belongs to Relativistic Uniform Acceleration.

Builds on. Four-Momentum · Four-Velocity · Four-Acceleration · Relativistic Force. Leads to. Conservation of Four-Momentum · Energy–Momentum Relation · Mass–Energy Equivalence.

How can a four-vector have a negative length? Is that not a contradiction?

Only in Euclidean geometry, where lengths are sums of squares. Minkowski geometry has an indefinite metric: $A\cdot A = (A^0)^2 - |\vec A|^2$ can be positive (timelike), zero (null) or negative (spacelike), and all three occur physically. A spacelike four-vector "points more in space than in time" — and the four-force does, having no time component at all in the particle's own frame.

If K·U = 0 always, how do rockets work?

$K\cdot U = 0$ holds only for a mass-preserving force. A rocket's rest mass falls as it ejects fuel, so $\mathrm{d}m/\mathrm{d}\tau \ne 0$ and by identity (3) the total four-force is not orthogonal to $U^\mu$ ; the covariant law picks up the extra term $K^\mu = mA^\mu + (\mathrm{d}m/\mathrm{d}\tau)U^\mu$ . Nothing is violated — the rocket is simply not the case orthogonality describes.

Why is the spatial part of K equal to γF rather than F?

Because $K^\mu$ divides by proper time and $\vec F$ divides by coordinate time, and the two differ by $\mathrm{d}t/\mathrm{d}\tau = \gamma$ . Dropping the $\gamma$ does not merely rescale the answer — it yields an object that no longer transforms with $\Lambda$ , so it is not a four-vector and none of its invariants can be trusted.

What does the proper force actually mean?

It is the three-force the particle measures in its own instantaneous rest frame — what an accelerometer riding along would read, in newtons. It is an invariant, $F_0 = \sqrt{-K\cdot K}$ , so every observer agrees on it even while disagreeing about $\vec F$ . The practical content is the asymmetry: a longitudinal push is felt exactly as applied ( $F_0 = F$ ), a transverse push $\gamma$ times more strongly ( $F_0 = \gamma F$ ).

The four-force is orthogonal to the four-velocity. Do they look perpendicular on a spacetime diagram?

No, and generally they cannot. Minkowski orthogonality means the contraction $\eta_{\mu\nu}K^\mu U^\nu$ vanishes, not that the arrows make a right angle on Euclidean paper — orthogonal pairs appear to make equal angles with the light cone, closing toward it as the speed rises. The physical content is unambiguous regardless: in the particle's own frame $U = (c,\vec 0\,)$ and $K = (0, \vec F_0)$ , and there the perpendicularity is literal.

Could I just use the (−,+,+,+) signature and get a positive K·K?

Yes, and the physics would be identical: in $(-,+,+,+)$ timelike vectors have negative norm and spacelike ones positive, so $K\cdot K = +F_0^2$ there. What you may not do is mix the two — and mixing is exactly how the base's error is most easily made, since a formula copied from an MTW-convention text into a $(+,-,-,-)$ calculation inverts every sign silently while staying internally consistent and wrong. State your signature, use it everywhere, and test it against a sign you already know: $U\cdot U = +c^2$ is the fastest.