THE INVARIANT INTERVAL

Euclid's theorem says that the squared distance between two points in the plane, $d^2 = \Delta x^2 + \Delta y^2$, is invariant under rotations. Rotate your coordinate axes by any angle and $d^2$ does not change. The individual components $\Delta x$ and $\Delta y$ mix together as the axes turn, but their sum of squares is a fixed number — a property of the two points themselves, independent of the coordinate frame you choose to describe them in.

Special relativity pulls off the same trick in four dimensions, but with one crucial sign flip. When you apply a Lorentz boost — the SR analog of a rotation — the time coordinate $\Delta t$ and the spatial coordinate $\Delta x$ mix with each other exactly as $\Delta x$ and $\Delta y$ do under an ordinary rotation. Two observers in relative motion will each measure a different $\Delta t$ and a different $\Delta x$ between the same pair of events. Yet when they compute the combination $c^2 \Delta t^2 - \Delta x^2$, they get the same number. That combination is the Invariant interval, and the minus sign is not a typo — it is the geometrical signature of Spacetime.

understood this clearly in 1908 when he showed that the Lorentz transformation is a hyperbolic rotation in a 4-dimensional pseudo-Euclidean space. The prefix "pseudo" is the minus sign: the Minkowski metric is $\text{diag}(+1, -1, -1, -1)$ rather than Euclid's $\text{diag}(+1, +1, +1, +1)$. had already written the invariant in 1905, calling it the "Lorentz group invariant," but it was Minkowski who gave it geometric status and showed that the algebra of §02 was secretly a geometry.

For two events with spacetime separations $\Delta t$, $\Delta x$, $\Delta y$, $\Delta z$, the invariant interval squared is:

Every inertial observer — no matter how fast they move relative to the events — will compute the same value of $s^2$. It is a Lorentz scalar: a single number that belongs to the pair of events themselves, not to any particular frame. The individual pieces $c^2\Delta t^2$ and $\Delta x^2$ are frame-dependent; their difference is not.

To see this working in real time, drag the $\beta$ slider below. The two events A and B sit at fixed spacetime positions in the lab frame. As you boost the frame, the coordinates $\Delta t$ and $\Delta x$ shift — but the HUD quantity $s^2$ stays frozen. That frozen number is the geometric fact §03.2 is built on. Try all three presets — timelike, null, spacelike — and confirm that the invariance holds regardless of the sign of $s^2$.

The four-dimensional analog of Pythagoras's theorem is complete. Where Euclid's distance squared is a sum, Minkowski's interval squared is a difference. One minus sign separates Euclidean geometry from the geometry of the universe we live in.

The sign of $s^2$ divides all event pairs into three mutually exclusive classes, each with a distinct causal meaning:

A Timelike interval means $c^2\Delta t^2 > \Delta x^2$: the time separation dominates. A massive particle can travel from one event to the other at some subluminal speed. Crucially, every observer agrees on which event happened first — the time ordering of timelike-separated events is Lorentz-invariant. A cause and its effect are always timelike-separated.

A Null interval means $s^2 = 0$: the spatial and temporal separations are exactly balanced as $\Delta x = c\Delta t$. Only a photon connects them. The Light-cone is precisely the surface $s^2 = 0$ in 4D spacetime.

A Spacelike interval means $\Delta x^2 > c^2\Delta t^2$: space dominates. No massive particle and no signal traveling at or below $c$ can connect the two events. Different observers can disagree on which event came first — the time ordering of spacelike-separated events is frame-dependent. This is why spacelike separation is the mathematical definition of "causally disconnected."

For a timelike interval the invariant interval has a direct physical meaning. There exists a frame — the rest frame of a clock that travels along the straight worldline between the two events — in which $\Delta x = 0$. In that frame $s^2 = c^2\Delta\tau^2$, where $\Delta\tau$ is the time elapsed on the clock. Solving:

This is the Proper time: the time recorded by a clock that rides along the worldline. Because $s^2$ is Lorentz-invariant, $\Delta\tau$ is Lorentz-invariant. Every observer, regardless of their velocity, agrees on how much the traveling clock ticked between the two events. They disagree on $\Delta t$ — the coordinate time in their own frame — but that disagreement is exactly absorbed by the Time dilation factor $\gamma$: $\Delta t = \gamma\,\Delta\tau$.

The bar chart below makes the bookkeeping visual. A clock rides between two fixed events at velocity $\beta c$ relative to the lab. As you slide $\beta$ upward:

The ordinary rotation $x' = x\cos\theta + y\sin\theta$, $y' = -x\sin\theta + y\cos\theta$ preserves $x^2 + y^2$ because $\cos^2\theta + \sin^2\theta = 1$. A Lorentz boost is structurally identical but uses hyperbolic functions. Define the rapidity $\phi$ by $\tanh\phi = \beta$; then the boost becomes:

The invariance of $s^2$ follows immediately: $\cosh^2\phi - \sinh^2\phi = 1$ — the hyperbolic identity replaces the Pythagorean identity, and the minus sign in $s^2$ is exactly what makes the identity fire. A Lorentz boost is a rotation by imaginary angle $i\phi$ in the $(ct, x)$ plane; the geometry is hyperbolic, not circular.

This is the reason the algebra of §02 is secretly a geometry. The Lorentz transformation is not an ad hoc patch to electromagnetism — it is the statement that spacetime has a definite metric structure, and inertial observers are related by the symmetry group of that metric. The group is $\text{SO}(1,3)$, the "1" in the signature coming from the single timelike direction and the "3" from the three spacelike ones. Every Lorentz boost, every spatial rotation, every combination of both, preserves $s^2$ for every pair of events.

The consequences cascade. §03.3 shows that the Light-cone is the surface $s^2 = 0$, and that causal structure is therefore a geometric property of spacetime itself — not a dynamical law imposed from outside. §03.4 shows that the four-velocity, four-momentum, and the EM application of the same four-vector machinery all emerge naturally from differentiating along a proper-time parameter — the very $\Delta\tau$ we found above. §03.5 shows that the twin paradox is not a paradox at all: it is the statement that a kinked worldline has shorter proper length in Minkowski metric than the straight worldline between the same endpoints, exactly as a bent path in Euclidean space has greater length than the straight one.

The Cologne lecture of 1908 is where announced this to the world. The §03 module is the geometry he described. The §02 algebra was the geometry all along.