The Universe Is a ComputerEpisode 8 (extended) ／ The imaginary dimension

After the finale ── from a single question, "what happens if you make \(D\) imaginary?"

The imaginary dimension ── when zoom gets notched In Episode 1 we found \(D\) is a "reading." So what happens if you open that reading to complex numbers?
The imaginary part does not measure "size." It measures how it repeats under zoom ── an orthogonal, separate axis.

Core: imaginary part = your favorite zoom factor / because we are on the real side, the flow never closes Hook equation: \(D=a+i\beta\)

Episode 1 concluded that the dimension \(D\) is not a number the world possesses, but a gauge reading on an instrument, read off via \(D=-d\ln F/d\ln n\). If it is a scale reading ── then it need not be real. This episode opens \(D\) straight to complex numbers. The answer is astonishingly clean: the real part is decay (the ordinary dimension), the imaginary part is oscillation (the notch of zoom). And finally, why our world sits on the imaginary-part-zero side, and why that makes the flow of mass "never close" ── it all links in one line to the finale of the sister series "Mass That Clicks."

01Recap ── a purely real \(D\) is a "straight slope"

Setting \(x=Cn\) (size measured in units of resolution), the ruler formula becomes a single exponential:

Opening it up in logs $$F=\frac{1}{(Cn)^{D}}=(Cn)^{-D}=e^{-D\,\ln(Cn)}$$

If \(D\) is real, the log-log plot (\(\ln F\) versus \(\ln n\)) is a straight line, with slope \(-D\). That is what you saw with the Episode 1 slider. A surface gives \(D=2\), a wave (line) \(D=1\), a solid \(3\). Integers are just special readings, and non-integer values could be read off too ── but they were all real.

Now let us push one step further. If \(D\) is a "reading," then feeding in a complex number and computing doesn't make the formula complain. Let's try it.

02Opening \(D\) to complex numbers ── the real part decays, the imaginary part rotates

Feed \(D=a+i\beta\) (real part \(a\), imaginary part \(\beta\)) into the exponent. By the law of exponents it splits in two:

The reveal ── one line splits the roles $$F=(Cn)^{-a}\cdot e^{-i\beta\ln(Cn)} =\underbrace{(Cn)^{-a}}_{\text{decay}}\cdot\underbrace{\big[\cos(\beta\ln Cn)-i\,\sin(\beta\ln Cn)\big]}_{\text{rotation (oscillation)}}$$

Just Euler's formula \(e^{-i\theta}=\cos\theta-i\sin\theta\), with \(\theta=\beta\ln(Cn)\).

The purest case ── the pure imaginary \(D=i\beta\) with \(a=0\) ── lays the truth bare. \(|F|=1\) always holds. Nothing grows or shrinks. Scale \(n\) and \(F\) just spins around the unit circle ── pure log-periodic oscillation.

Figure 1: the complex dimension \(D=a+i\beta\). The horizontal axis is \(n\) (log). The real part \(a\) slider sets the decay of the envelope \(\pm(Cn)^{-a}\) = the ordinary dimension, and the imaginary part \(\beta\) slider sets the fineness of the ripples running inside it. Set \(\beta=0\) and the ripples vanish, returning to Episode 1's plain straight line (power law).
Move the sliders and the envelope (decay = real part) and the ripples (oscillation = imaginary part) change separately.
\(\mathrm{Re}\,F=(Cn)^{-a}\cos(\beta\ln Cn)\) envelope \(\pm(Cn)^{-a}\) (real part a) notch of favorite zoom λ

03Meaning ── the imaginary dimension is "a mark that zoom has been notched"

Let's derive the condition for one full turn. Each time the phase \(\beta\ln(Cn)\) increases by \(2\pi\), \(F\) returns to its original value. This happens when \(n\) changes by a fixed factor:

The favorite zoom factor $$n\to n\cdot\lambda \ \text{makes } F \text{ return},\qquad \lambda=e^{2\pi/\beta},\qquad \beta=\frac{2\pi}{\ln\lambda}$$

Here is the crux. A pure power law \(n^{-D}\) allows \(n\to\lambda n\) for any \(\lambda\) ── it has a continuous scale symmetry. But when the world has a special magnification (a branching ratio, one renormalization step, a nesting ratio), that symmetry drops to a discrete scale symmetry allowing only \(n\to\lambda^k n\) (integer \(k\)). That favorite factor is exactly what shows up as the imaginary part of \(D\):

This episode in one line

The imaginary dimension = a mark that "continuous zoom" has broken into "notched zoom." The imaginary part \(\beta\) is that notch width itself (the log of the favorite factor \(\lambda\)). Where the real part measures size, the imaginary part measures how it repeats under zoom ── an orthogonal, separate axis.

Connects to Episode 3 In Episode 3, "the zoo of representations," we saw that continuous and discrete are one invariant read in different directions. The imaginary dimension is the equation-form version of that bridge ── the discreteness of scale (a favorite zoom) appears as the imaginary part of a dimension that looks continuous. Discrete or continuous is representation, but "is there a notch" becomes the readable difference of whether \(\beta\ne0\) or \(\beta=0\).
◇ ◇ ◇

04This is not a game ── the complex dimensions of the Cantor set

The imaginary dimension is not a mathematical prank but real mathematics. Self-similar fractals in fact carry complex dimensions. Take the middle-thirds Cantor set (divide into three, discard the middle, repeat forever): its naive dimension is \(\ln2/\ln3\approx0.63\), but the complete set of dimensions is:

Complex dimensions of the Cantor set (Lapidus) $$D_k=\underbrace{\frac{\ln 2}{\ln 3}}_{\text{real part = dimension of size}}+\;i\,\underbrace{\frac{2\pi k}{\ln 3}}_{\text{imaginary part = imprint of zoom ratio 3}},\qquad k=0,\pm1,\pm2,\dots$$

The real part is the familiar fractal dimension. The imaginary spacing \(2\pi/\ln3\) is the direct imprint that this set returns to itself under a 3× zoom (\(\lambda=3\)). So it is exactly the general formula \(\beta=2\pi/\ln\lambda\).

As a result, measuring the "volume" of the Cantor set (properly, the Minkowski content) shows, on top of the power law, a log-periodic pulsation with period \(\ln3\). This is a real, observable phenomenon and even has a name ── discrete scale invariance (DSI). It has been observed and studied as a log-periodic precursor oscillation in rock fracture before earthquakes, prices before financial crashes, aftershock sequences, and more (Sornette et al.).

Figure 2: dimensions arrayed on the complex plane. Horizontal axis = real part (dimension of size), vertical axis = imaginary part. The number of divisions b and number kept k sliders array the complex dimensions of a self-similar fractal as points. The real part is fixed at the single value \(\ln k/\ln b\), and the imaginary part is notched like a ladder at spacing \(2\pi/\ln b\) ── that notch is the fingerprint of the zoom ratio \(b\). (\(b=3,\,k=2\) is the Cantor set.)
With the sliders, the real part (dimension of size) and the imaginary notch (zoom ratio) move separately.
complex dimension \(D_k\) real part \(=\ln k/\ln b\) (dimension of size) imaginary notch \(2\pi/\ln b\)
◇ ◇ ◇

05So why are we on the "real side"? ── and why mass never ends

Here we land on the topic of mass. The dimension we observe (3 for space, 4 for spacetime) is purely real, imaginary part zero. It is the state with \(\beta=0\) in Figure 1. Which means:

The consequence of being on the real side

There is no closing period. Raise the scale and \(F\) never returns to the same value again. It just flows on smoothly ── this is the true nature of "the flow of mass never ends."

This is exactly Episode 4 of the sister series "Mass That Clicks." Spacetime \(D=4\) is the special dimension where couplings run only logarithmically (marginal). Log flow is the slowest and never stops on its own. Into that never-stopping flow, one scale sneaks in as an integration constant ── this is dimensional transmutation:

Mass wells up from a flow that never ends $$\Lambda = E\,e^{-1/(b\,\alpha)}\qquad(\text{the integration constant of log running gives birth to a mass scale})$$

If there were an imaginary part = a notch of your own, the flow would close and complete itself every factor of \(\lambda\). But the real world has none. So the flow runs to the very end, and the stop (cutoff) must be given from outside ── above by \(M_{\text{Pl}}\) (Planck), below by \(\hbar H_0/c^2\) (the floor of minimum mass set by the size of the universe).

The payout ── because it doesn't close, the "geometric mean" appears

And finally it all connects cleanly. When the flow has no favorite scale (purely real, log-symmetric), the only natural scale you can build between the two ends \(m_{\text{IR}}\) and \(M_{\text{Pl}}\) is the midpoint on the log axis ── that is, the geometric mean:

The minimum always points to the maximum
$$m\sim\sqrt{m_{\text{IR}}\cdot M_{\text{Pl}}}\quad(\approx\mathrm{meV})$$

"No special factor = symmetric about the log axis = midpoint," so it becomes the geometric mean. The imaginary part being zero, so the flow never closes, is conversely the reason the two ends are joined by a geometric mean. Precisely because it doesn't close, the two ends point at each other.

Put another way ── we observe with a \(D\) of imaginary part zero. So mass rides not a closed period but a flow that keeps running logarithmically, and its scale lands on the geometric mean of the two ends. The story of the imaginary dimension connects in one line all the way to the finale of mass.

The honest line

Introduce an imaginary dimension and \(F\) becomes complex. If \(F\) is a real observable like a force, a count, or a probability, it cannot mean anything as it stands ── what you can read is \(\mathrm{Re}\,F\) or \(|F|\), or \(F\) reinterpreted as an "amplitude." What is physically observed is the log-periodic ripple riding on the power law, not the imaginary part of \(D\) itself.

Discrete scale invariance, complex dimensions, Lapidus's theory, and log-periodic precursors are all established mathematics / phenomenology. However, log periodicity has not been observed in the running of the fundamental laws of particle physics (were it there, it would be evidence of a discrete scale symmetry = a notched spacetime). And "why is space exactly real 3-dimensional, with no imaginary part" is not a puzzle specific to this framework but an open problem of physics as a whole. We do not fill it here ── "do not say solved" is the rule of the whole series. The imaginary dimension is not "something between a surface and a wave"; the one thing that is certain is only this: it is a separate axis, orthogonal to size.

Practice problems (solvable from this sheet)
  1. For a pure imaginary \(D=i\beta\), what happens to \(|F|\)?
    See the answer
    \(|F|=|(Cn)^{-i\beta}|=|e^{-i\beta\ln Cn}|=1\). Always 1. Nothing grows or shrinks; scale \(n\) and it just spins around the unit circle = pure log-periodic oscillation. Decay (size) is carried by the real part alone; the imaginary part moves only the phase = the oscillation.
  2. What is the relation between the imaginary part \(\beta\) and the "favorite zoom factor \(\lambda\)"? What is it for the Cantor set (3× zoom)?
    See the answer
    \(\lambda=e^{2\pi/\beta}\), inversely \(\beta=2\pi/\ln\lambda\), since it returns after the phase advances by \(2\pi\). For a 3× zoom (\(\lambda=3\)), \(\beta=2\pi/\ln3\approx5.71\). It is the mark of a continuous scale symmetry having dropped to a discrete scale symmetry of factor 3.
  3. "We observe with a real dimension, so the flow of mass never ends" ── what does this mean, and what does it not mean?
    See the answer
    It means: with no imaginary part = no closing period, the running of couplings across scale does not stop on its own and keeps flowing logarithmically. So the stops (\(M_{\text{Pl}}\), \(\hbar H_0/c^2\)) are given from outside, and the scale between lands on the log midpoint = the geometric mean \(\sqrt{m_{\text{IR}}M_{\text{Pl}}}\). It does not mean: that the value of mass is infinite (the value is finite and definite). What never ends is the flow, not the value. Nor does it link directly to "because space is 3-dimensional" ── what matters is the separate thread that spacetime's 4 dimensions are the critical dimension of log running.

SummaryThe imaginary part is "the notch of zoom"; because the real part is nonzero, the flow never closes

Because the dimension \(D\) is a reading, it can be opened to complex numbers. Feed in \(D=a+i\beta\) and \(F=(Cn)^{-a}[\cos(\beta\ln Cn)-i\sin(\beta\ln Cn)]\) ── the real part \(a\) is decay (the ordinary dimension of size), the imaginary part \(\beta\) is log-periodic oscillation. The imaginary part is not "between a surface and a wave" but an orthogonal axis measuring how it repeats under zoom, with \(\beta=2\pi/\ln\lambda\) = the notch of a favorite factor \(\lambda\). It is the mark of a continuous scale symmetry broken into a discrete one, and it exists in reality as the complex dimensions of the Cantor set and as log-periodic precursors (DSI).

We are on the real side with imaginary part zero ── so there is no closing period, mass keeps running logarithmically (dimensional transmutation \(\Lambda=Ee^{-1/b\alpha}\)), the stops come from outside (\(M_{\text{Pl}}\), \(\hbar H_0/c^2\)), and the scale between lands on the log midpoint = the geometric mean \(\sqrt{m_{\text{IR}}M_{\text{Pl}}}\approx\mathrm{meV}\). Precisely because it doesn't close, the minimum points to the maximum. The remaining "why real 3 dimensions" is a hole of physics as a whole ── we leave it honestly open.

← Finale: not a flag, but a marker Episode 9: The equations of information and the equations of physics →
This document is Episode 8 (extended, a summary of a conversation, honest version) of the series "The Universe Is a Computer." Euler's formula \(e^{-i\theta}=\cos\theta-i\sin\theta\), the decomposition of the complex exponent \((Cn)^{-(a+i\beta)}\), discrete scale invariance (DSI) and log-periodic oscillation, the complex dimensions of fractals (M. Lapidus's theory of complex dimensions, with \(D_k=\ln2/\ln3+i\,2\pi k/\ln3\) as poles of the geometric zeta function), the phenomenology of log-periodic precursors (D. Sornette et al.'s work on fracture, earthquakes, and finance), logarithmic running of couplings in 4-dimensional spacetime and dimensional transmutation \(\Lambda=Ee^{-1/b\alpha}\), and the geometric mean \(\mathrm{meV}\sim\sqrt{m_{\text{IR}}M_{\text{Pl}}}\) based on the CKN bound are all established mathematics / physics, or content honestly delineated in the sister series "Mass That Clicks." \(F=1/(Cn)^D\) is a rewriting of the defining relation of dimension, \(D=-d\ln F/d\ln n\), and opening \(D\) to complex numbers is mathematically consistent, but log periodicity (an imaginary part) has not been observed in the running of the fundamental laws. "Why is space real 3-dimensional with no imaginary part" is an open problem today. ── To print, use your browser's "Print" and "Save as PDF" (in the print version the sliders and answers are static and hidden).

Print / PDF: Ctrl+P (⌘+P on Mac). On screen, in Figure 1 the real part = decay and imaginary part = ripples move separately, and in Figure 2 the complex dimensions array like a ladder. Click "See the answer" to reveal each solution.