Reinterpretation Principle and Causal Paradoxes

In 1949 Richard Feynman took an embarrassment and turned it into a particle. Relativistic wave equations insist on solutions with negative energy — states of arbitrarily large negative energy, into which every electron in the universe ought to cascade. The fix, which Ernst Stückelberg had already published in 1941, was not to delete them but to relabel them: a negative-energy electron running backwards in time is exactly a positive-energy positron running forwards. The bookkeeping is identical, the disaster evaporates, and the antiparticle — a real object, found in a cloud chamber seventeen years earlier — falls out of the relabelling. It is one of the great moves in physics, and so cheap that people naturally tried it again. If a tachyon could carry a signal, some observer would see it running backwards in time carrying negative energy, and two such signals would let a reply arrive before its question. So relabel: what one observer calls "emitting a negative-energy tachyon into the past" another calls "absorbing a positive-energy tachyon from the past". No negative energies anywhere. This is the reinterpretation principle of Bilaniuk, Deshpande and Sudarshan, and at the level of the energy ledger it works perfectly.

And it does not save the telephone. This article shows exactly why, and the reason is a single sign. The relabelling is legitimate when the four-momentum is timelike and illegitimate when it is spacelike, because that one fact decides whether the sign of the energy is something all observers agree about. Same move, two regimes, opposite verdicts.

Key takeaways

Convention lock. $(+,-,-,-)$ , $\eta = \mathrm{diag}(+1,-1,-1,-1)$ . Timelike is POSITIVE: $p\cdot p = +m^2c^2 > 0$ . A tachyon is spacelike: $p\cdot p = -\mu^2c^2 < 0$ .
The antitelephone is not re-derived here. `aq-sr-causal-structure` owns it, including the correction that a closed loop needs $v > 2wc^2/(w^2+c^2)$ (in rapidities $\varphi > 2\chi$ ), not the folklore $vw > c^2$ , which reverses only one leg. At $w = 2c$ the loop needs $v > 0.8c$ ; the often-quoted $v = 0.6c$ leaves the reply arriving $30.6\%$ late.
The owned result — a one-line criterion for when relabelling is allowed. Under a boost $E' = \gamma(E - Vp_x)$ . If $p$ is timelike then $|p_x|c < E$ , so $E' > 0$ for every $|V| < c$ : the sign of the energy is a Lorentz invariant. If $p$ is spacelike then $|p|c > E$ , and the boost $V^{*}/c = E/(p_xc) < 1$ flips it.
The topology says it too. The timelike shell has two disconnected sheets ( $E \geq +mc^2$ , $E \leq -mc^2$ ) separated by a gap no boost can jump. The spacelike shell is one connected surface crossing $E = 0$ . That is the whole difference.
So Feynman–Stückelberg is a frame-independent relabelling and the tachyonic reinterpretation is not. Every observer agrees which solutions are the positrons; no two observers need agree which end of a tachyon's worldline was the emitter.
Where it breaks. Reinterpretation rescues the energetics — nobody ever sees a negative-energy tachyon. It cannot rescue signalling: relabelling emission as absorption is a change of description, and a description cannot alter a correlation.
Key idea: the reinterpretation principle is exactly as good as the sign of $p\cdot p$ . Timelike, and it is a triumph that gives you the positron. Spacelike, and it buys a tidy energy ledger while the causal loop closes behind your back. The move did not fail because it was applied carelessly. It failed because a controllable choice is a fact about the world, and no amount of relabelling can permute a fact.

The core insight

Both applications of the relabelling do the same formal thing: they take a process running one way in time with negative energy and rewrite it as a process running the other way with positive energy. Whether that is a reinterpretation or a fiction turns on one question — do all observers agree on the sign of the energy? For a timelike four-momentum they do, necessarily: $E^2 = |\vec p\,|^2c^2 + m^2c^4$ forces $|E| \geq mc^2 > 0$ , and the two branches are separated by a gap no boost can cross. The relabelling is then a global dictionary — everyone applies it to the same solutions and gets the same answer, and what comes out is the antiparticle. For a spacelike four-momentum $E$ can be anything, including zero; the mass shell is a single connected surface, and boosting slides you continuously across $E = 0$ . Now the dictionary is observer-dependent, and while the energy ledger balances in every frame separately, the thing it was supposed to protect — the causal order of a controllable signal — is untouched by any of it.

Where this lives today

No tachyon has ever been found. Searches — cosmic-ray showers, time-of-flight, collider kinematics — return null results, and the one famous scare, OPERA's 2011 report of neutrinos beating light from CERN to Gran Sasso, was traced in 2012 to a loose fibre-optic connector and a fast clock oscillator. But the word survives, meaning something else entirely — the reinterpretation story told a third time. In quantum field theory a field with $m^2 < 0$ is called tachyonic, and it does not propagate faster than light. A negative $m^2$ signals that you are expanding about the wrong vacuum: the potential has a maximum where you assumed a minimum, and the "tachyonic mass" is the growth rate of the instability rolling off it. Let it roll, expand about the true minimum, and the quanta come back with $m^2 > 0$ . This is the Higgs mechanism — the Standard Model's scalar is tachyonic in the unbroken phase and perfectly ordinary in the broken one. Nothing exceeds $c$ ; microcausality holds throughout. What that leaves is narrower and sharper than the slogan. Not "nothing is faster than light" — phase velocities, sweeping laser spots and cosmological recession all are, and entangled pairs are correlated across spacelike separation. Not "no negative energies" — reinterpretation disposes of those. What is forbidden is the transfer of a controllable bit outside the light cone, and quantum field theory encodes exactly that as microcausality; `aq-sr-causality-light-cone-constraint` shows it is the same statement as the support of the retarded propagator. The ban was never on speed. It was always on signalling.

The physical picture

Take the antitelephone as given. `aq-sr-causal-structure` builds it and settles its arithmetic: a superluminal signal of speed $w > c$ is seen running backwards in time by any observer with $vw > c^2$ — that is one leg — while the closed loop, in which Alice's reply reaches her before she asked, needs the strictly stronger

v > \frac{2wc^2}{w^2+c^2}, \qquad\text{equivalently}\qquad \varphi > 2\chi \quad (\tanh\chi = c/w,\ \tanh\varphi = v/c).

At $w = 2c$ that is $v > 0.8c$ , and the folklore case $v = 0.6c$ leaves the reply arriving $30.6\%$ late. All of that is shipped, and this article does not re-derive it. What it does is take the loop seriously and ask: is there any way out? There is one candidate, and it has an excellent pedigree. The trouble with a tachyon is not really that it is fast. It is that in some frames it runs backwards in time carrying negative energy — and physics already knows how to handle both, because both happen to ordinary antiparticles in the Feynman–Stückelberg reading. So apply the same medicine. Whenever a frame reports "a negative-energy tachyon was emitted at $B$ and absorbed at $A$ , travelling backwards", rewrite it as "a positive-energy tachyon was emitted at $A$ and absorbed at $B$ , travelling forwards". Emission and absorption swap labels; the energy comes out positive; nobody ever writes down a negative-energy particle. Do the bookkeeping and it works. Every observer describes a world in which only positive-energy tachyons are emitted, always forwards in time. The ledger balances. This is the reinterpretation principle, and it is not a swindle — it is exactly the move that produced the positron. And now the trap. The two cases are not the same, and the difference is not a matter of taste. In the antiparticle case the relabelling is forced on everybody identically, because all observers agree about the sign of the energy. In the tachyon case they do not. The next section proves that in one line, and everything else follows.

The core equations

The mass shell, in the locked signature — re-derived from $\eta$ , not transcribed:

p\cdot p \;=\; \eta_{\mu\nu}p^\mu p^\nu \;=\; \frac{E^2}{c^2} - |\vec p\,|^2 \;=\; \begin{cases} +m^2c^2 & \text{timelike (ordinary matter)}\\[2pt] -\mu^2c^2 & \text{spacelike (a tachyon)} \end{cases} \qquad (1)

where: $\mu$ is real and positive; $m = i\mu$ is what is meant by a tachyon's "imaginary mass". Physical reading: the sign of $p\cdot p$ is a Lorentz invariant, and it is the only thing that matters below. The energy under a boost of speed $V$ along $x$ :

E' = \gamma\,(E - V p_x), \qquad \gamma = \frac{1}{\sqrt{1 - V^2/c^2}} > 0 \qquad (2)

where: since $\gamma > 0$ always, the sign of $E'$ is the sign of the bracket — and the whole question is whether that bracket can change sign for some legal $|V| < c$ . Solving $E' = 0$ gives the boost that would zero the energy, and hence the threshold beyond which it goes negative:

\boxed{\;\frac{V^{*}}{c} \;=\; \frac{E}{p_x c}\;} \qquad (3)

where: $V^{*}$ is a real, legal boost if and only if $|E| < |p_x| c$ , which by equation (1) is exactly $p\cdot p < 0$ : spacelike. Physical reading: this is the hinge of the article. The boost that would flip the energy exists precisely when the four-momentum is spacelike — precisely for tachyons, and never for anything else. The two mass shells, read as surfaces:

\text{timelike: } |E| \geq mc^2 > 0 \quad\text{(two disconnected sheets)}; \qquad \text{spacelike: } E \in (-\infty, \infty) \quad\text{(one connected surface)} \qquad (4)

where: for a timelike shell $E^2 \geq m^2c^4$ , so there is a gap of width $2mc^2$ around $E = 0$ containing no states. Physical reading: a continuous boost cannot jump a gap. The orthochronous Lorentz group preserves each sheet, so the sign of $E$ is invariant — equation (3), stated topologically. For a spacelike shell $E^2 = |\vec p\,|^2c^2 - \mu^2c^4$ passes smoothly through zero at $|\vec p\,|c = \mu c^2$ : no gap, one surface, and a boost slides you across it.

Figure

THE WHOLE ARTICLE IN ONE PICTURE. LEFT — the TIMELIKE mass shell E² = p²c² + m²c⁴. It has TWO DISCONNECTED SHEETS, separated by a gap of width 2mc² that contains no states. A boost slides a point ALONG a sheet and can never jump the gap, so the sign of E is a LORENTZ INVARIANT. Every observer agrees which solutions are the negative-energy ones — so relabelling them as positrons is a GLOBAL dictionary that everybody applies identically. That is Feynman–Stueckelberg, and it works. RIGHT — the SPACELIKE shell E² = p²c² − μ²c⁴ of a tachyon. There is NO GAP: it is ONE CONNECTED surface crossing E = 0 at pc = μc². A boost slides continuously across the crossing, so the sign of E is FRAME-DEPENDENT. Observers disagree about which end of the worldline was the emitter, the dictionary is different in every frame — and the energy ledger balances while the causal loop closes behind your back

Two theorems, and where the second one runs out

Derivation AThe sign of the energy is boost-invariant if and only if the four-momentum is timelike

The hinge of the article, in four lines. For a timelike four-momentum no legal boost can make the energy negative — so "positive energy" is an objective, frame-independent property of a solution — while for a spacelike four-momentum there is always a legal boost that does, and we write it down. Everything afterwards, including the exact point at which the reinterpretation principle stops working, is a corollary.

Step 1. Start from equation (2): $E' = \gamma(E - V p_x)$ with $\gamma > 0$ , so $\mathrm{sign}(E') = \mathrm{sign}(E - Vp_x)$ . The question is whether the bracket can change sign for some $|V| < c$ . Take $E > 0$ without loss of generality. Why: the whole problem is now one bracket.

|V p_x| \;<\; c\,|p_x| \;\leq\; |\vec p\,|c \;<\; E \qquad\Longrightarrow\qquad E - Vp_x > 0 \qquad\Longrightarrow\qquad E' > 0 .

Why: two strict inequalities multiply, and the bracket cannot reach zero. The observer is subluminal and the momentum is sub-energetic; together those forbid the flip. The sign of a massive particle's energy is not a frame choice — it is a Lorentz invariant, as much as the mass is.

V^{*} = \frac{E}{p_x} \qquad\Longrightarrow\qquad \frac{V^{*}}{c} = \frac{E}{p_xc} < 1 .

The zeroing boost is subluminal, so it exists, and any $V > V^{*}$ gives $E' < 0$ . Why: the same inequality read the other way. For a tachyon the momentum out-runs the energy. There is no gap to jump — equation (4).

Step 4. State it as the criterion the rest of the article will use.

\boxed{\;\mathrm{sign}(E) \text{ is Lorentz invariant} \iff p\cdot p > 0 \;}

Why: this is the whole difference between the positron and the tachyon telephone, and it is one sign. A relabelling of "negative-energy, backwards" into "positive-energy, forwards" is a well-defined operation on solutions only if all observers agree which solutions need relabelling. Step 2 says they do, for matter. Step 3 says they do not, for tachyons. *Verified in verify_aq-sr-reinterpretation-principle-causal-paradoxes.py: for timelike four-momenta, $E' > 0$ is asserted symbolically and over randomly sampled masses, momenta and boosts; for spacelike four-momenta the zeroing boost $V^{*}/c = E/(p_xc)$ is shown to be strictly subluminal and to give $E' < 0$ beyond it; and the two mass shells are shown to have, respectively, a forbidden gap $|E| \geq mc^2$ and a zero crossing at $|\vec p\,|c = \mu c^2$ .*

Go deeper · Derivation B · 2 minFeynman–Stueckelberg: why relabelling a negative-energy solution is legitimate, and what it produces

Derivation BFeynman–Stueckelberg: why relabelling a negative-energy solution is legitimate, and what it produces

With the sign of the energy proved objective for timelike momenta, we can say precisely what the relabelling is and why it is not a swindle: a negative-energy solution read as running backwards in time is identically a positive-energy solution of the charge-conjugate problem running forwards — the same physics, written twice. This is the benchmark against which the tachyonic version will be judged.

Step 1. Write a plane-wave solution of the Klein–Gordon equation for a particle of charge $q$ : $\psi_+ \propto \exp[-\tfrac{i}{\hbar}(Et - \vec p\cdot\vec x)]$ with $E = +\sqrt{|\vec p\,|^2c^2 + m^2c^4}$ . The equation is quadratic in $\partial_t$ , so it also admits $E \to -E$ , and that negative-energy mode is the embarrassment. Why: it cannot simply be discarded — the two together make the mode expansion complete.

Step 2. Look at the negative-energy mode honestly. With $E = -|E|$ ,

\psi_- \;\propto\; \exp\!\left[+\frac{i}{\hbar}\big(|E|\,t + \vec p\cdot\vec x\big)\right] \;=\; \left(\exp\!\left[-\frac{i}{\hbar}\big(|E|\,t - (-\vec p)\cdot\vec x\big)\right]\right)^{\!*}.

A negative-energy mode of momentum $\vec p$ is the complex conjugate of a positive-energy mode of momentum $-\vec p$ . Why: this is the entire technical content of the Feynman–Stückelberg move. The relabelling is not a story told over the mathematics; it is an identity in it.

Step 3. Ask what complex conjugation does physically. For a charged field it is charge conjugation: the conserved current built from $\psi^*$ carries charge $-q$ where the one built from $\psi$ carried $+q$ . So the object in Step 2 is a particle of charge $-q$ , energy $+|E|$ , momentum $-\vec p$ , moving forwards in time. Why: that is the antiparticle, and nothing was added to get it. One object, two descriptions — and the second is what a detector sees.

Step 4. Now check the thing that makes it legitimate — the step everyone skips. The relabelling is applied to exactly those solutions with $E < 0$ , and by Step 2 of the previous derivation every observer agrees which those are. The dictionary is global: two physicists in relative motion, handed the same solution, relabel it in exactly the same cases. Why: this is the load-bearing fact, and precisely what will fail for tachyons. A relabelling that different observers apply to different events is not a change of description — it is a change of physics. *Verified in verify_aq-sr-reinterpretation-principle-causal-paradoxes.py: the identity of Step 2 is checked symbolically, and both modes are confirmed to satisfy the Klein–Gordon equation in the locked signature.*

Figure

THE LEDGER THAT BALANCES WHILE THE LOOP STAYS CLOSED. Alice's worldline is vertical; Bob moves at v = 0.9c; the tachyon legs run at w = 2c. LEFT — the description BEFORE reinterpretation: leg 2 carries NEGATIVE energy and runs backwards in Bob's frame. The reply lands at t_C = 76/121 = 0.628 t_A, i.e. 0.372 t_A BEFORE Alice asked — the loop is closed. RIGHT — the SAME diagram after the reinterpretation principle is applied: every arrow flipped, every energy now POSITIVE, the ledger clean in every frame. But look at the event marked ✗: it has not moved. The reply is still recorded at 0.628 t_A. Reinterpretation renames the arrows; it does not move the events — and the paradox lives in the events, not in the arrows

Go deeper · Derivation C · 2 minWhere reinterpretation runs out: a relabelling cannot permute a correlation

Derivation CWhere reinterpretation runs out: a relabelling cannot permute a correlation

The principle does something real: it removes negative energies from the description of tachyon exchange in every frame. The claim under test is much stronger — that having fixed the energetics, the causal paradox is dissolved too. It is not, and the argument needs no new relativity: reinterpretation is a change of description, while the paradox is a statement about a correlation.

Step 1. Say what the paradox actually asserts. Alice has a free bit $A \in \{0,1\}$ — she presses "send" or she does not — and the antitelephone claims there is an event in Alice's own past at which Bob's reply is registered, with detector reading $B$ . The paradox is that

P(B \mid A = 1) \;\neq\; P(B \mid A = 0),

so Alice can read off her own future choice before making it. Why: notice what the paradox is not. Not "some quantity came out negative", not "an observer used odd coordinates". It is a statistical dependence between a free variable and an earlier record.

Step 2. Say what reinterpretation actually does. At each vertex it exchanges the labels emitter and absorber and flips the sign of the tachyon's energy and its worldline's arrow, so the energy comes out positive. Crucially it does not move, add or delete any event. Why: it re-describes the same worldline diagram with the arrows reversed. Every vertex, every coincidence of two worldlines, is exactly where it was.

Step 3. Put the two together. $P(B \mid A)$ is built from coincidences: whether Alice's switch was closed, and what Bob's detector recorded, are facts about events, and the correlation is a fact about the joint list of events over many runs. Step 2 says that list is invariant under reinterpretation. Therefore

\boxed{\;P(B \mid A) \text{ is unchanged by the reinterpretation principle}\;}

Why: and that is the end of it. If the loop was closed before the relabelling, it is closed after. Reinterpretation is a dictionary between two ways of narrating the same diagram, and you cannot destroy a correlation by renaming its endpoints.

Step 4. Now find the case reinterpretation genuinely does save, because it is not nothing. Suppose the emissions are spontaneous — thermal, chosen by nobody. Then there is no free bit $A$ , nothing to be correlated with, and the two descriptions ("Alice emitted a negative-energy tachyon into the past" and "Alice absorbed a positive-energy tachyon from the past") are genuinely equivalent: same events, same probabilities, neither preferred. Reinterpretation is exactly right here. Why: this is the honest boundary. The principle rescues the energetics of uncontrolled tachyons and nothing else. It buys a consistent energy ledger; it does not buy back the light cone.

Step 5. So state what the paradox was always driving at. The residual demand is not "no faster-than-light particles" and not "no negative energies". It is: no controllable information may propagate outside the light cone. Quantum field theory imposes exactly that, as a condition on operators rather than particles — microcausality, $[\hat\phi(x), \hat\phi(y)] = 0$ at spacelike separation. Why: a controllable bit is precisely a correlation with a free variable. Everything else — phase velocities, spontaneous emissions, spacelike correlations of entangled pairs — is permitted, and permitted for exactly this reason. *Verified in verify_aq-sr-reinterpretation-principle-causal-paradoxes.py: the invariance of $P(B\mid A)$ under relabelling is checked on a toy joint distribution — the emitter/absorber swap is applied as a permutation of event labels and the conditional asserted unchanged; the stochastic case is shown to carry no signal; the loop numbers recalled from aq-sr-causal-structure ( $64/49$ at $v=0.6c$ , $76/121$ at $v=0.9c$ , $w=2c$ ) are recomputed exactly.*

Limiting cases

Case	Result	Meaning
$p\cdot p > 0$ (timelike)	$\mathrm{sign}(E)$ invariant	the shell has a gap; relabelling is global; the positron
$p\cdot p < 0$ (spacelike)	$\mathrm{sign}(E)$ frame-dependent	no gap; each frame relabels differently; the telephone survives
$\vert\vec p\,\vert c \to \mu c^2$	$E \to 0$ , $V^{*} \to 0$	the tachyon's energy vanishes; any observer, however slow, sees it negative
$\vert\vec p\,\vert c \gg \mu c^2$	$E \to \vert\vec p\,\vert c$ , $V^{*} \to c$	an ultra-relativistic tachyon looks almost null; only a fast observer flips it
emissions spontaneous	reinterpretation succeeds	no free bit to correlate with; the two descriptions are equivalent
emissions controllable	reinterpretation fails	$P(B\vert A)$ is a fact; relabelling cannot permute it

Common trapReinterpretation is a real result. Do not overstate what it does — or what it fails to do

Two opposite errors are common. The first is to dismiss the principle as a dodge. It is not. Applied to timelike four-momenta it is the Feynman–Stückelberg interpretation, exactly correct, and it is where the positron comes from. Applied to uncontrolled tachyons it is also correct: it genuinely removes negative energies from every frame's description, and the two narrations are equivalent because there is no free variable to distinguish them.

The second is to believe it dissolves the antitelephone. It does not. Reinterpretation relabels the arrows on a worldline diagram without moving a single event, and the paradox lives in the correlation between Alice's free choice and a record in her own past. Rename the endpoints all you like; the coincidence table is the same table.

The one-line test: ask whether anything an experimenter can choose ends up correlated with anything recorded earlier. If yes, the loop is closed, and no dictionary will open it.

Worked examples

1 · The boost that flips a tachyon's energy — and the one that cannot flip a proton's

Take a hypothetical tachyon with $\mu c^2 = 1.00\ \mathrm{GeV}$ carrying momentum $|\vec p\,|c = 2.00\ \mathrm{GeV}$ . Find the observer who sees its energy vanish. Then do the same sum for a proton of the same momentum, and see what stops you.

Step 1 — identify. Equation (1) for the energies, equation (3) for the flipping boost. The whole exercise is one inequality: is $E/(p_xc)$ below one, or not?

Step 2 — the tachyon. Spacelike, so $E = \sqrt{|\vec p\,|^2c^2 - \mu^2c^4} = \sqrt{(2.00)^2 - (1.00)^2} = \sqrt3 = \mathbf{1.7321\ GeV}$ , and from equation (3)

\frac{V^{*}}{c} = \frac{E}{p_xc} = \frac{1.7321}{2.00} = \frac{\sqrt3}{2} = \mathbf{0.8660}, \qquad \gamma(V^{*}) = \mathbf{2.000}\ \text{exactly}.

An observer moving at $0.866c$ along the tachyon's momentum sees it carrying exactly zero energy — and anyone faster sees negative energy, running backwards in time. That observer is not exotic.

Step 3 — the proton, same momentum. Timelike, with $m_pc^2 = 0.938272\ \mathrm{GeV}$ :

E = \sqrt{(2.00)^2 + (0.938272)^2}\ \mathrm{GeV} = \mathbf{2.2092\ GeV}, \qquad \frac{V^{*}}{c} = \frac{2.2092}{2.00} = \mathbf{1.1046}.

Step 4 — the boundary number, and it is the point. $V^{*}/c = 1.1046 > 1$ . The boost that would flip the proton's energy is superluminal, so it does not exist. No observer, at any speed, ever sees this proton with negative energy — and that is no accident of the numbers. By Step 2 of the first derivation it is a theorem: $E > |\vec p\,|c$ for every timelike momentum, so $V^{*}/c > 1$ always.

Step 5 — the moral. The same formula $V^{*}/c = E/(p_xc)$ is a legal boost for the tachyon and an impossible one for the proton; the only thing that changed was the sign of $p\cdot p$ . Feynman–Stückelberg relabels a class of solutions every observer can identify. The tachyonic reinterpretation does not.

2 · The loop that reinterpretation cannot open

Take the antitelephone at $w = 2c$ and a replying observer at $v = 0.9c$ . Confirm the loop closes, then apply the reinterpretation principle and watch nothing happen to it.

Step 1 — recall, do not re-derive. `aq-sr-causal-structure` gives the loop condition $v > 2wc^2/(w^2+c^2)$ — at $w = 2c$ , $v > 0.8c$ — and the arrival ratio $t_C/t_A = w^2(1-v^2/c^2)/(w-v)^2$ . In rapidities, with $\tanh\chi = c/w$ and $\tanh\varphi = v/c$ :

\chi = \mathrm{artanh}(0.5) = 0.5493, \qquad \varphi = \mathrm{artanh}(0.9) = \mathbf{1.4722}, \qquad 2\chi = 1.0986.

$\varphi > 2\chi$ , so the loop closes — with a rapidity margin of $\mathbf{0.3736}$ .

Step 2 — the number. $t_C/t_A = 4(1 - 0.81)/(2-0.9)^2 = 76/121 = \mathbf{0.6281}$ . If Alice transmits one year after passing Bob, the reply reaches her at $0.6281$ yr — that is $\mathbf{0.3719}$ yr, about 136 days, before she asked.

Step 3 — now apply reinterpretation. In Bob's frame the outbound tachyon has negative energy and runs backwards; so reinterpret — call it a positive-energy tachyon emitted by Bob and absorbed by Alice. Do the same on the return leg. Every frame now describes only positive-energy tachyons moving forwards in time. The energy ledger is clean in every frame — exactly what the principle promised.

Step 4 — the boundary number, and it does not move. Ask the only question that matters: does Alice's reply still arrive at $0.6281\,t_A$ ? Reinterpretation moved no events. The reception event sits at the same spacetime point; her logbook holds the same record. So if her decision to transmit depended on that record — and nothing stops her wiring it that way — then $P(B \mid A)$ is still what it was, in every frame, under every relabelling.

Step 5. Reinterpretation converted a paradox with negative energies into a paradox without them. The negative energies were never the problem.

Physics of the portrait

Origins · 2 minFeynman-Stueckelberg interpretation and how this was discovered

Feynman-Stueckelberg interpretation

1941–1967 · Stueckelberg's backwards-running particle · Feynman's positron · and the move that was borrowed once too often

The interpretation has two fathers and an awkward afterlife. Ernst Stückelberg — Swiss, publishing as Stueckelberg — proposed in Helvetica Physica Acta 14, 588 (1941) that the unwanted negative-energy solutions be read as particles propagating backwards in time. Richard Feynman arrived at the same reading independently and made it work, in "The Theory of Positrons", Physical Review 76, 749 (1949), where he credits Stückelberg explicitly: the negative-energy electron running backwards is the positron running forwards, and with that one relabelling the whole of positron theory falls out of electron theory with no new machinery. It is unimprovable physics — and it is unimprovable because the energy sign of a timelike four-momentum is an invariant, so the relabelling is a dictionary every observer shares. Then it was borrowed. In 1962 Bilaniuk, Deshpande and Sudarshan published "'Meta' Relativity" (American Journal of Physics 30, 718), applying the same swap to hypothetical faster-than-light particles: the reinterpretation principle. Gerald Feinberg gave the objects their name — "Possibility of Faster-Than-Light Particles", Physical Review 159, 1089 (1967) — and the word tachyon entered the language. The borrowing was honest and the algebra was right, and it still did not work, for the reason this article exists to give: a spacelike four-momentum has no invariant energy sign, so the dictionary differs from frame to frame — and a paradox that lives in a correlation is untouched by a change of dictionary.

History

1917

— Tolman, in The Theory of the Relativity of Motion, notes that an influence faster than light would let an effect precede its cause in some frame. He gives only the one-way form, and hedges it.

1941

— Stueckelberg, in Helvetica Physica Acta 14, 588, proposes reading negative-energy solutions as particles propagating backwards in time.

1949

— Feynman's "The Theory of Positrons", Physical Review 76, 749, develops the same reading and credits Stueckelberg. The negative-energy electron becomes the positron.

1962

— Bilaniuk, Deshpande and Sudarshan publish "'Meta' Relativity", American Journal of Physics 30, 718, applying the relabelling to faster-than-light particles: the reinterpretation principle.

1967

— Feinberg's "Possibility of Faster-Than-Light Particles", Physical Review 159, 1089, coins the word tachyon.

1970

— Benford, Book and Newcomb publish "The Tachyonic Antitelephone", Physical Review D 2, 263, giving the closed-loop condition for a superluminal signal to deliver a reply before its question.

2012

— OPERA traces its 2011 faster-than-light neutrino result to a loose fibre-optic connector and a miscalibrated clock oscillator. Corrected, the neutrinos travel at c.

Going further

Texts. Rindler, Introduction to Special Relativity, 2nd ed., §7 · Peskin & Schroeder, §2.4 (microcausality) and §11.1 (tachyonic masses). Primary: E. C. G. Stueckelberg, Helv. Phys. Acta 14, 588 (1941); R. P. Feynman, "The Theory of Positrons", Phys. Rev. 76, 749 (1949); O. M. P. Bilaniuk, V. K. Deshpande and E. C. G. Sudarshan, "'Meta' Relativity", Am. J. Phys. 30, 718 (1962); G. Feinberg, "Possibility of Faster-Than-Light Particles", Phys. Rev. 159, 1089 (1967); G. A. Benford, D. L. Book and W. A. Newcomb, "The Tachyonic Antitelephone", Phys. Rev. D 2, 263 (1970); Y. Aharonov, A. Komar and L. Susskind, Phys. Rev. 182, 1400 (1969).

Provenance. Signature $(+,-,-,-)$ ; timelike positive, $p\cdot p = +m^2c^2$ ; a tachyon is spacelike. Base defect corrected: the base's eq[2] prints the antitelephone condition as " $u\,v > c^2 \Longrightarrow$ closed causal loop". False — $uv > c^2$ reverses one leg; the closed loop needs $v > 2wc^2/(w^2+c^2)$ , which the base's own DERIV[0] correctly boxes, so the base contradicts itself between its equation card and its derivation. `aq-sr-causal-structure` owns the correction, recalled here rather than re-derived. Second base defect: DERIV[0] step 3 writes the round-trip elapsed time as $\Delta t_{\text{Alice}} = \gamma^2 t_1(1 - Vu/c^2)(1 - Vu/c^2) < 0$ — a perfect square asserted to be negative, impossible for real $V$ , $u$ . The correct arrival is $t_{\mathrm{arr}} = t_1[\,2u - V(1 + u^2/c^2)\,]/(u - V)$ . verify_aq-sr-reinterpretation-principle-causal-paradoxes.py asserts every number printed above.

Builds on. `aq-sr-causality-light-cone-constraint` · `aq-sr-tachyons-superluminal-particles` · `aq-sr-causal-structure` · `aq-sr-center-of-momentum-invariant-mass`. Leads to. `aq-sr-relativistic-collision-kinematics` · `aq-sr-mandelstam-variables`.

If the reinterpretation principle really removes all the negative energies, what exactly is left to complain about?

The causal loop, which was never made of negative energies. The two arrive together but are separate complaints. The energy complaint — "in Bob's frame this tachyon carries $E < 0$ , which is unphysical" — reinterpretation answers completely. The causal complaint — "there is a record in Alice's past whose content depends on a choice she has not yet made" — it does not touch, because the relabelling changes only which end of a worldline is called the emitter. It does not move the worldline, the record, or what the record says. You have converted a paradox with negative energies into a paradox without them.

Why does the same relabelling work perfectly for the positron?

Because for a timelike four-momentum the sign of the energy is a Lorentz invariant — proved in the first derivation, and visible in the mass shell's shape: the two branches are separated by a gap of width $2mc^2$ containing no states, and a continuous boost cannot jump a gap. So every observer picks out the same solutions as the negative-energy ones. The relabelling is a single global dictionary everybody applies identically, and what it produces — opposite charge, positive energy, moving forwards — is a frame-independent object. It is a positron, and you can put it in a detector. For a tachyon the shell has no gap, the sign of $E$ is frame-dependent, and there is no global dictionary to be had.

Isn't the "free choice" of the experimenter a philosophically loaded assumption?

It is an assumption, and it does work — but it is a cheap one, and all of experimental physics already runs on it. The claim is only that an experimenter's setting is statistically independent of the system being measured before the experiment begins: whether Alice presses the button is not pre-correlated with what Bob's detector was always going to read. Deny that and you deny the possibility of a controlled experiment in general, not merely rescue tachyons. The escape hatch exists — superdeterminism, the same hatch available to evade Bell's theorem — but nobody who pays that price gets to keep ordinary laboratory practice.

Does the tachyonic antitelephone rule tachyons out completely?

It rules out tachyons that can carry a controllable signal, which is the only kind anybody cared about. It does not, by itself, rule out spacelike four-momenta appearing in a theory. Field-theoretic "tachyons" — modes with $m^2 < 0$ — are respectable and common; they signal an unstable vacuum rather than a superluminal particle, they carry no signal, and the Higgs field is one of them in its unbroken phase. The statement relativity forces is narrow: no controllable information outside the light cone. A spacelike mode nobody can modulate does not conflict with it.

The base article says the antitelephone condition is uv greater than c squared. Why does this article not use it?

Because it is the wrong condition, and the base contradicts itself about it — its equation card prints $uv > c^2$ as the loop condition while its own derivation correctly boxes $v > 2uc^2/(u^2+c^2)$ . The inequality $uv > c^2$ is the condition for one leg to run backwards in some legal frame — already fatal for a Lorentz-invariant theory, and all the inequality genuinely buys. But the loop needs both legs dragged past their reversal thresholds, and in rapidities that is $\varphi > 2\chi$ : exactly twice. At $w = 2c$ the loop needs $v > 0.8c$ , and the frequently quoted $v = 0.6c$ leaves the reply arriving $30.6\%$ late. `aq-sr-causal-structure` owns the correction.

What is the connection between reinterpretation and microcausality?

They are the two ends of the same argument. Reinterpretation is the last attempt to have superluminal particles and keep a consistent world by adjusting the description; microcausality is what you are forced to when that fails, and it adjusts the dynamics. The condition $[\hat\phi(x),\hat\phi(y)] = 0$ for $(x-y)^2 < 0$ says no local observable at $x$ can be influenced by an operation at spacelike-separated $y$ . Notice what it constrains: not the existence of spacelike momenta, not the sign of any energy, but the influence of a choice — precisely the residue the reinterpretation argument leaves behind.