<span class="mw-page-title-main">November (2026)</span>
Fabrice P. Laussđ•Ș’s â„€ymocidal Web

22 November (2026)

Today we visit bridges.


More "supplementary material" of CSIT, which I don't want to put in the paper to optimise size and stick to the main points, but that I like (or need) to have in mind: the maximum value of local violation (as opposed to—in the text—its angular average):


cf. CSIT's Fig. 6

See Fig. 7 on â„€ ⟶

21 November (2026)

This figure will not go in the CSIT paper[1] but I want to keep it here on â„€ because it shows the loss of nonlocal violation as local violations get lost (Fig. 6 of the paper) through distortion of the dipoles. I believe there's a link between the two, distant violations being linked or bounded by local ones, but I'll probably won't pursue this statement (at least not today), so I'll leave it as a lossy—or lousy—(laussy) conjecture. It's probably not even difficult to show, more triangle inequalities... The violation is too small to see, but it is there in between the two narrow bands along $\theta_1+\theta_2=\pi/2$ and $3\pi/2$ (very faint greenish region there).

The ellipses are to show the difference between the two bottom cases: a very small change in their superposition leads to massive change of CSI landscapes. No changes in any case for the one-photon density (a donut), as usual. $\sqrt2-1$ is $\tan(\pi/8)$ the threshold of no violation. Also on â„€ ⟶


Ego: Dimittesne mihi, Angele?
Clamor in capite meo surdum me faciebat.
Strepitus clamorem silere faciet.
Vos omnes silere sufficit.


I'm playing with a way to backup some of my 𝕐 posts to 𝕏, with some 𝕐2𝕏 gizmo. This is the first one of them.


Meanwhile, week-end in Soto:


So the guy who solved (allegedly) the Riemann hypothesis is Levent Alpöge, the same guy who settled the Jacobian conjecture, for which he took no further credit than a tweet... He was after much bigger things.

Now that he might have solved—with the assistance of Claude—the most important problem of Mathematics (will he get the Fields medal?) what will he be working on? I mean apart from the destruction of humanity, by catalysing the transition from Sapiens Sapiens to Machina Sapiens? (There will be no need to transit through the Homo Digitalis, probably.) Consciousness, surely. I think there's no more urgent and important problem now. And when/if he gets there, God, then. And when/if he also gets there, I don't know... we'll run out of problems. Happiness? Women? There are still a few, but they're probably particular cases of the previous ones.

The guy is funny, he's playful, clearly a highly creative mind. He asked the machine to produce a paper version of the proof of what is now the Riemann theorem "just for fun", and it came out as a 250,000-page document. That's my type of fun. He went from someone I followed on Twitter to one of the most significant and influential—I don't want to write "thinker", he's more of a "problem solver"—of our time.


Computer computes faster than I can think. It finished its round of #jc-bubbles calculations and then went idle. I have to analyse the linewidth structure, which I didn't do yet because it's much messier and complicated than the energy one. I didn't even show any of it in our earlier 2009-2011 works in presence of pumping. I remember that it still has a neat branching structure, though, which is already beautiful in the unpumped dissipative Jaynes-Cumming ladder:[2]

But with pumping, it all goes berserk, and the range of variations is such that it becomes very difficult to navigate through. Will do, will do... Now it's inescapable. But until I can give it utmost attention, I thought I'd acquire a collection of sightings, a photographic record of the phenomenology. But I take so much time for everything else that the computer dutifully reports task after task before I get the chance to think of where to bring it next, and it's not entirely trivial to stay away from both useless and unending computations.

Animations are nice and help develop an intuition of the structures. And they are kh-hungry, so good candidates to keep the machine busier than me. The $\gamma_\sigma$ animation is nice but too few frames, one can only get the roughest idea of how the skeleton distorts. This could be expanded, although those are so heavy that it'd be prohibitive to make it marginally better.

One that I think will be a killer is the usual fractal dive-in: zooming into $\omega_p$, and see how new lines and new structures keep popping out. With an horizontal drift that should be a quite nice and dizzying jump into the quantum infinite, a sort of Mandelbrot counterpart but with eigenvalues. I'll try to realise this artistic vision, probably around Fig. (0QbAG), although it's not clear that more strongly-coupled systems in deeper lasing regimes are the most fractal ones. Thermal components could make things much more intermingled due to fluctuations over much more rungs. So, still much to do when you give it a couple of minutes attention.

20 November (2026)

Fucking black Friday really was black to me... It was facepalm Friday for our WP4 meeting

Daniel was obviously thinking about something else... All our attempts to close something get thwarted by the universe throwing more stuff at us so that we'll never converge to the final result. Now we're into synthetic fluctuations...

At least, submitting Ergoli to our second option after failing to interest the first one gave me the small satisfaction of having had to find "sub-headings" for some sections, which I liked. That's what I confided to Emacs:

laussy@yzao:~/publi/20/25/5--NataliaCrystal/ergoli12$ cat ergoli12.tex | grep subsection
\subsection*{The canonical procedure}
\subsection*{One frame with many photons vs many frames of few photons}
\subsection*{Photon bunching is Simpson's paradox}
\subsection*{Removing confounding factors}
\subsection*{The physical meaning of multiphoton correlations}
\subsection*{Multiphoton observables}
\subsection*{Conclusions}

Although I'll probably go "as" rather than "is" for Simpson.

My other Chinese friends didn't do anything interesting, but I'm convinced that a huuuuge result is shouting at us despite all their efforts not to find it.

The Zeno speed at which they monitor their single trajectories is pumping quantum coherence! which shows up in this broken observable of relative phase that they like so much, as a higher contrast and emergence of this $\delta$ peak at $\Delta\theta=0$. Next meeting for next Friday. Let's hope it'll be a less black one.

Then the Internet rumours that someone solved the Riemann hypothesis! Not today, please.... I won't enjoy it today...

And I've been struggling for most of the day to decide if to keep the $C_{02}C_{20}\neq0$ in CSIT or just flush it into â„€. What do you think?

Patricia wrote to me that she already solved two of the six problems I put on her desk yesterday. I read her web page where she says that everybody likes her! It's funny because it looks true. I like people who solve problems. When she dragged me from office to office to demonstrate her innate talent for doing so, she said to someone «a ver si tomamos un café después.» I asked them where do they have coffee, and she said «oh aquí en la oficina.» When I summed up my wish list later by email, I ventured that next time I'd also invite her for a coffee, arguing lamentably that «conozco sitios al lado mejores que la cafetera de la oficina», and she replied «un café no se le deniega a nadie». It's also funny because, this time, it's not true. Well, there are some problems that she cannot solve, of course, but only because nobody can.


What the one-atom lasing transition is, is a moiré effect with eigenvalues. Crossing of ladders are tinyly displayed, with $\sqrt{n}$ everywhere, and they slide through each others as anharmonic grids that produce those onion-looking shapes making necklaces of beads that interwoven intricately. If it wouldn't be for the interferences, it'd definitely be a simple analytical superpositions of square roots. But there is branch coupling, avoided crossings, exceptional points everywhere.


I feel like those lyrics written by Mike Oldfield for his beautiful vocalist Maggie Reilly:

The last that ever she saw him
Carried away by a moonlight shadow
[...]
Will you come to talk to me this night?
But she couldn't find how to push through
[...]
I stay, I pray
See you in Heaven one day.

Songs of beautiful melodies in which sentences are tangled, woven, trapped, an impossible escape, songs of being stuck away from yourself:

Don't you know you're never going to get to France.

19 November (2026)

Resolution is very bad (too much compression) but we have to start somewhere.


We're entering the Y triduum of AF, until 0YU6G (af).

Where will you be on 0Y000 (af)? And on 0Ywhy (af)? And on 0Yyou (af)? And on 0Ynot (af)?

Where are you now? What are you doing?


Our attempts to timelapse the office board got one step further today. Jacob still has to fiddle with the focus and zoom as images are currently not good quality, but he could automatise and stabilise everything! First board-lapses coming soon...


Refinement of one of the structures of 0XRnQ in a window where numerical precision goes blind. Interestingly, the red lines are gone. I would have thought they are more robust because their sign is settled at lower truncation, and they seem to arise from stronger features (a higher rung closing), but they turn out to be indeed artifacts that better precision removes.

18 November (2026)

A low-pump result, without and without interpolation, at two levels of zooming in:

I'm not so confident about analytical results even for the first bubbles...

Also on â„€ ⟶


The colours of November.

Lower numerical accuracy looses the structure. The point is, how to get a consistent, overall grasp of it. It's infinite in complexity.

I've started a new ten-go campaign, futatabi, on $k=5$, with 70 pumps of order 400 at 80 digits, from 11.51 to 13.718$g$, so that it can be checked as 300 against 400. About 198 kh. We'll know tonight. See the colour code on â„€ ⟶


At last my fermions show up in a way I like: they provide the maximum coincidence rate by which that of bosons can be found in excess of the classical rate. Interestingly, this is weighted by the concurrence $\mathcal{K}$ of the bosons. I think I could (or at least, should) have seen that before, but it helps to work with $D$ rather than $R$, even for pure states. It also shows why the phase is so important and brings us back to the geometry of the wavefunction.


In the CSI problem beyond Fock states, when it turns to mixtures, the ratio $R$ is not so convenient and the difference $D$ becomes preferable: $$D\equiv G^{(2)}_{12}-\sqrt{G^{(2)}_{11}G^{(2)}_{22}}\,.$$ This is because this measures how many coincidences are in excess as compared to the classical maximum (saturation), and this adds up over the components of a mixture (I have proven already that a mixture does not violate more than its best-violating pure-state component).

Besides, $D$ is convex over the pair state ($R$ is not), so that the excess of a mixture is at most the sum of the excesses of its components, $D\big(\sum_kq_k\rho_k\big)\le\sum_kq_kD(\rho_k)$.

$D$ and $R$ are equivalent anyway, in the sense that $D>0\iff R>1$, with \begin{align} \label{eq:0XLrx} D&=\sqrt{G^{(2)}_{11}G^{(2)}_{22}}\,\big(\sqrt R-1\big)=G^{(2)}_{12}\,\big(1-R^{-1/2}\big)\,,\\ R&=\bigg(1+\frac{D}{\sqrt{G^{(2)}_{11}G^{(2)}_{22}}}\bigg)^{2}=\bigg(1-\frac{D}{G^{(2)}_{12}}\bigg)^{-2}\,. \end{align}

Also on â„€ ⟶


We don't lament so much about lost time anymore than about lost kernel-hours (kh). A flaw in my attempt to organise a constraint of deadlines in ten-go led to 235.8 kh essentially thrown away (essentially because this still provides the minor sanity proof that different machines compute the same thing). At least I noticed "relatively" quickly (it's about 8h lost, a human's day, the machines live in a parallel world, albeit a classical one).

17 November (2026)

We welcome a new friend—Matilde—with whom to unravel the secrets of the universe. We start on Monday.

16 November (2026)

Najmeh's work on synchronization[3] is a nice read in parallel of our CSI violations:

This is the first-order one, with actual vectors of the Hilbert space, so no possible violation, only saturation. Although they claim they are in a quantum regime, they don't study CSI violations (of what can be violated, i.e., going $g^{(2)}$), which would certainly occur as they are in the qubit limit. They don't discuss exceptional points, that are certainly also central to their phenomenology. They find a huge variance in entanglement when oscillators synchronize, but that doesn't mean that they get entangled, only that it happens more frequently, or indifferently (something that was preventing them to now lets them). It's a nice work but with more questions left hanging than are actually answered. I believe their dynamics is ruled by a dark state, as we ourselves found a long time ago.[4] Also on â„€ ⟶


Five-minute discussion with Jacob in the office.


Three-hundred years ago, Mac Laurin's derivation of the inequality between the arithmetic and geometric means, in his Letter to Martin Folkes, Esq; V. Pr. R. S., «concerning Êquations with impossible root»:[5]

This inspired Cauchy for his CSI.[6] It is delightful to read—not their blog, poor chaps didn't have Internet—but the similarly spirited prose with which they describe their intentions, their frustrated attempts, lavish the reader with excuses for not having written in a long time, etc. See how, in a second letter,[7] MacLaurin explains why the «To be continued» of the previous one (three years earlier) remained hanging, until now, that «some reasons have now determined [him] to send [...] the continuation». We don't know what those reasons are (probably his priority dispute with Campbell[8]), but we're given plenty of details on his plans to write a Treati∫e of Algebra

He died before writing the book, of course. But the treatise was posthumously published anyway, so it's worth loosing time.


I need something on laussy.org to visualise series of figures, this will be necessary to keep track of the Jaynes-Cummings skeleta. Probably an extension of pan. The feature I need, which exists in pan in a rudimentary form, is to scroll between tiles, holding the viewing window. It must scroll with the mouse, and be efficient. Gwenview does something close enough on linux but is far from ideal anyway (slow). I'm inspecting the low-pumping series (bijaku, ćŸźćŒ±, "weak" ) and either they get stored here, or they'll be lost forever on đ”» after I perused them once. So far they look kinky although the √ closing of the lines when they meet their exceptional point allows some rounding of the curves:

I'm running an accelerating van der Corput order refinement to get a smooth structure. Also this gives hopes that such low-$P_\sigma/g$ limit could be described exactly analytically, at least the first bubble (that closes down to the axis, apparently with alternating colours like a Zeno bubble). Also on â„€ ⟶

15 November (2026)

That's how my AF tag completion works:

And then from Emacs, if I type:

In Sec. C-)

where C-) is the Emacs key-mapping to the function, it expands this to:

In Sec.~\ref{sec:0

filling up the 0 because everything to date is in the first Fabriad. Then TAB offers completions:

So with my lossy memory, if I type 0Hbi plus another tab, then it expands fully. If ambiguity persists, it narrows it down. For some of them, like 0HAGAR (af), I don't even need any tab.

It separates Eqs. from Figs. from Sec., etc. It's also resilient to case errors, e.g., if I type 0F it will expand to 0fsIq, because nothing has 0F but there's one and only one 0f.

«You're so cute, honey, gee!»
–:––

I'm implementing something cool that expands my AF tags in Emacs, from those defined in the document. So I don't need to remember the exact (in particular, case-sensitive) variation, but just something that closely resemble it. The machine does the rest. And AF tags are perfect for that. They're all memorable! 0WhoV (af). See? That wins a lot of time, and, I believe, that also weaves new patterns, in addition to sensory or cognitive ones (which is why you should read real physical books, not screen versions).

I was somehow uneased that such shortcuts are not challenging my memory and letting the machine spoonfeed my brain. But the opposite is true. New links, actually more precise ones, emerge. Instead of "the equation about two modes below the cute metaphor" becomes "the 0WTF-looking one.

I'm telling you, transcendence in the universe is lossy. You don't want exactitude: exactitude is for the machine, for matter, for the basic laws of physics. Emergence is lossy. Intelligence is lossy. It's approximate, it's close-enough, it's quite-good as opposed to perfect. Lossy is for the intuition, for creativity, for insights.

That's what makes intelligence, intelligent. If it'd be exact, it'd be mathematics, and every intelligent entity would be the same as any other: a theorem.

But because it's lossy, each agent is unique, each brings its own almost-perfect variation, and everything and everybody has an approximate, partial, but building-up to something contribution.


One-atom lasing updates. Since last week, computers have been computing the skeleton for all cases of Fig. (0NNbp), to see how this evolves. That's how: I don't show the thermal case because there's only one line then (all its certified lines are on the axis), the skeleton has completely collapsed.

Although it took a week to compute (nonstop, on 40 kernels at peak), this is still very low-res. Still the overall structure and how it moves is clear. I mean, it's clear from the observational point of view. From the interpretational point of view, it has several unexpected attributes:

  1. Going away from lasing doesn't make the structure simpler (until it collapses in the non-lasing case). Instead the bubbles pack up and interfere more strongly. What lasing brings us is room. I'm not too sure then that the bubbles correspond to photons falling into the condensate. If that were the case, the trend would be opposite.
  2. The above comment is for the bubbles. The outer rungs do behave as expected, and recede to the right as the system gets away from lasing. It thus seems that various nested structures interfere. It would be worth doing a log-$P_\sigma$ scan, say up to $0.2$, to see how such first few rungs do interact with each other. It's probably very clean, as they get neatly separated. It's unexpected that systems less in the strong-coupling see more of the JC ladder, in one-photon observables.
  3. Relatedly at higher pumps, it gets very busy, and thus less well defined, as $\gamma_\sigma$ increases, again away from my expectations. The transition from this to the thermal case is unclear. Maybe thermal is just so intricately connected that nothing resolves, just the main line.

That was for the zoomed-in view around the condensate. The global view is shown on â„€ (follow-up link below). It shows the continuity of the outer-rungs patterns, establishing their existence, although it also exhibits some noisy low $P_\sigma/g$, outer $\omega_p/g$ structures, which look like numerical noise. The computation guarantees the positions, but not always their colour. If it is genuine, I have no idea what it is.

I have, however, confirmed that the black lines can be certified and get their proper (blue/positive, red/negative) colour at higher precision, and also that the clouds of scattered points of Fig. (0PgZn) are artefacts, as opposed to snapshots of a greater, very steep structure. This is the block-recursion confirmed sign of the eigenvalues. Positions are good, but sign is easily unresolved:

That most probably explains the numerical noise at small $P_\sigma$ and large $|\omega_p|$.

I also confirmed, furthermore, that the number of smooth features is infinite, and will never stop, so it has to be followed up to some order, possibly with colour-coding (white→0).

Computation of this set is completing up to $P_\sigma/g\le 15$ (getting more of the pac-man) which will bring us towards early next week. I'll then detail the all-coloured-lines and infinite-structure stance and act accordingly, either nursing the existing computations or computing log-zoom of small pumps, leaving this problem back to computers for the coming days.

See the corresponding global view on â„€ ⟶


I'm briefly looking at (classical) chaos as a first step to describe in the one-atom lasing problem. I found this delightful (novel-level) account[9] of PoincarĂ©'s blunder of publishing the opposite result of what would make him most famous (and, anecdotally, what earned him the prize!) He paid of his own pocket to have all printed copies recalled and destroyed. This cost him more than what the prize earned him. This didn't go unnoticed as several articles have been written about it (see also Ref. [10]). There's even a book about it!

14 November (2026)


Otra Salazar-Laussy colaboraciĂłn.




Hoy pizzas caseras...

13 November (2026)

Satyanad Kichenassamy gives what I think is the best historical account of the Cauchy-Schwarz inequality in Ref. [6], through Bunyakovsky's derivation of integral version in 1859, which he regarded as a particular case of Cauchy's finite sums version (that itself follows from Lagrange's equality). Schwarz's proof of the integral version came much later (1888).

It comes with a lot of insights:


Typical Friday evening...


AHBT is alive. Alexandros did a first "quick" (waveplate) test for three modes and Daniel confirmed in the night that it works as it should. Jacob is now computing the dividends. I'm looking at whether we should include the population imbalance, which is interesting for the reasons below.

Intensity is a dummy player for balanced modes. For non-balanced modes $n_1\neq n_2$ (mean populations), the sum of two exponential distributions fluctuates as a random variable $W\equiv I_1+I_2$ that follows a $$ P(W) = \frac{e^{−W/n_1} − e^{−W/n_2}}{n_1 − n_2}$$ distribution with mean $\langle W\rangle=n_1+n_2$. This becomes a $\Gamma(2)$ distribution when $n_1=n_2$. The second moments of $P$ is $\langle W^2\rangle=(n_1+n_2)^2+n_1^2+n_2^2$ so that

\begin{equation} \label{eq:0TRCK} R\equiv\frac{\langle W^2\rangle}{\langle W\rangle^2}=1+p_1^2+p_2^2=\frac32+\frac{(p_1-p_2)^2}{2}\,. \end{equation}

Furthermore, $W$ is no longer independent from the shape $a\equiv\sqrt{I_1I_2}/W$ and two new dividends thus appear: \begin{equation} \label{eq:0TRCK2} \mathrm{IS}=\frac{g^{(2)}_{\mathrm{IS}}}{g^{(2)}_{\mathrm I}\,g^{(2)}_{\mathrm S}}\,,\qquad % \mathrm{IPS}=\frac{g^{(2)}_{\mathrm{IPS}}}{g^{(2)}_{\mathrm I}\,g^{(2)}_{\mathrm{PS}}\,\mathrm{IS}}\,, \mathrm{IPS}=\frac{g^{(2)}_{\mathrm{IPS}}g^{(2)}_{\mathrm S}}{\,g^{(2)}_{\mathrm{PS}}\,g^{(2)}_\mathrm{IS}}\,, \end{equation} so that \begin{equation}\label{eq:0Vf3L}g^{(2)}_{\mathrm{IPS}}=\mathrm{I}\cdot\mathrm{P}\cdot\mathrm{S}\cdot\mathrm{PS}\cdot\mathrm{IS}\cdot\mathrm{IPS}\,,\end{equation}

which has all the coalitions except IP which is one (and $\emptyset$ of course, which also contributes 1), so $g^{(2)}$ is in this case a product of six terms.

Defining $\mu_2\equiv p_1p_2$, $\nu_1\equiv\tfrac\pi4\sqrt{p_1p_2}$ and $\mu_1\equiv\tfrac{3\pi}8\sqrt{p_1p_2}$, the corresponding correlations are found as:

$$\begin{align}g^{(2)}_{\mathrm{IS}}&=[R+2\mu_1(c_1+c_2)+4\mu_2c_1c_2]/[(1+2\nu_1c_1)(1+2\nu_1c_2)]\,,\\ g^{(2)}_{\mathrm{IPS}}&=R+2\mu_2\cos2\Delta\theta\,,\end{align}$$

with all other variables previously defined. At coincidence $g^{(2)}_{\mathrm{IPS}}=1+p_1^2+p_2^2+2p_1p_2=2$, so the six dividends equilibrate to the $g^{(2)}$ of thermal light, i.e., 2. The redistribution between dividends can be confirmed again by freezing the confounding factors.

This is what is down below with Monte Carlo sampling and Claude's freezing:

If you multiply all the curves of the bottom row (the dividends), you get 2, the thermal value. This is according to Eq. \eqref{eq:0Vf3L}. If you remove a confounding factor from those products, in particular, here, P (the phase), i.e., if you align the frames, then the $g^{(2)}$ is the product of all the dividends who don't have the letter P, i.e., $\mathrm{I}\cdot \mathrm{S}\cdot \mathrm{IS}$. Since $\mathrm{IS}=1$ for balanced case, the superbunching peak is the product of the red and grey lines, that is, these strong correlations at the waist are the result of shape fluctuations, not the intensity, as people would assume given the little signal there. The intensity fluctuations are the same here than when there is a lot of signal: $3/2$. Besides, if the populations are not balanced, then $\mathrm{IS}\neq1$ and it even becomes the leading factor to explain the final correlations: now a coalition between shape and intensity emerged to account for the physics of azimuthal photon correlations.

Below is the IS coalition as a function of the population imbalance (black dot the optimum case above):

Also on â„€ ⟶


Working 300% (3PM) for our WP4 Heisingberg meeting.

12 November (2026)

Meetings, meetings and still more meetings. Videoconferencing is the pill to swallow before the $\omega$ point. You get to see very nice things, though.


We really live in science fiction. People born after 2010 will never understand how computers used to be brute, simple algorithmic machines following instructions more stubbornly than a tree. They'll see it as normal that a machine works out their problems, advises them, humours them...

The solitude of the soul—which is what interests me in the last stretch of my interacting with molecules and light—can only be greater then. It is lulled into the illusion that we are watched and tended, not by a cosmical abyss surrounding us in spacetime, but by a parallelepiped of plastic that plays a piano chord when you open its lid.

11 November (2026)

The MULTIPHOTONICS (2025) people. Who do you think is everybody?


Alexandros showing us his beautiful qCMOS toy with which we'll measure the Harsanyi dividends...


Okay Fable 5.2, let us see what you got...

10 November (2026)

La question que je me pose, c'est ce qu'il y a d'intemporel chez l'homme, et ce qui est passager, le goĂ»t d'une Ă©poque, le Zeitgeist... Il me semble que les jeunes gĂ©nĂ©rations sont dĂ©jĂ  bien dĂ©tachĂ©es de ce qui constituait le corpus de l'humanitĂ© tel qu'il a Ă©tĂ© ces quelque trois millĂ©naires passĂ©s. Les choses Ă©voluent vite, et lĂ  oĂč nos grands-pĂšres nous Ă©taient semblables, en particulier parce qu'ils ne devenaient pas bien vieux, nos petits-enfants n'auront plus rien Ă  voir avec nous. Je crois que cette solitude qui me broie le cƓur sera bien plus cruelle encore pour nos cadets, parce qu'ils ne se reconnaĂźtront ni chez ceux qui sont un peu plus vieux, ni chez ceux qui sont Ă  peine plus jeunes. Et ceux qui auront leur Ăąge partageront avec eux, avant tout, le mĂȘme modĂšle de tĂ©lĂ©phone. Ils n'auront, par-dessus tout, plus la compagnie de la littĂ©rature, qui leur sera devenue Ă©trangĂšre: des siĂšcles d'aspirations qui ne leur diront rien. Au moins, nous avons la compagnie des cimetiĂšres. À ceux qui nous suivent, l'Ă©merveillement des rĂ©volutions technologiques, mais la ruine de l'Ăąme.


What strikes me most about Russell Brand—beyond this travesty of justice, which strikes everybody—is that this is someone who, after his ups and downs in life, after his successes and his excesses, his deeds and his wrongdoings, had everything to be happy, live a quiet and retired life in the comfort of the autumn of his life.

Yet he decided to defy something and someone, to reject contemplation and equanimity, for a call to action, a will to change and transform, a show of bravado against the status quo.

He voiced his belief that he could heal himself and the world, that he could inspire others, rise as a model, a virtue, a moral compass, a figure of compassion, fight injustice, combat the bad and evil, he invoked the name of God and trumpeted that he was doing His work, he proclaimed the strength of his soul and the might of his conviction.

And it's as if something or someone took him up on the challenge. As if the reply came: "You think you can do it? Well, let's see then, luv..." That's in his Estuary twang; maybe the address was a bit more ceremonial, something like:

Can you draw out Leviathan with a fishhook
or press down his tongue with a cord?
Can you put a rope in his nose
or pierce his jaw with a hook?

It impacts me because something similar happened to me. I had everything to be relaxed and slow-paced about, and vanity challenged the outside inside me. Not in a prayer but in a loud internal voice, I spat at the ugliness of the unbearable world, I said to myself I would not accept the trauma in the far corners of the world, the conformism of the masses, the imbecility of everybody, the crassness of remaining silent when the world is weeping, I was looking for a way to raise my voice, I was begging for an occasion to stand up, I was writing about the children of Gaza and the American wars, and the bloodshed in Ukraine... Big topics, for someone sitting in his armchair.

And then all by itself, things unravelled. In my case, it might have sounded like: "Oh vraiment? Eh bien, d'accord..." Or maybe it was more ceremonial too:

Can you fill his skin with harpoons
or his head with fishing spears?
Lay your hands on him;
remember the battle—you will not do it again!

It doesn't come as a poem. It comes as a hurricane. From one day to the next, things that I never thought would or even could happen, did. I was put in front of myself, tossed with a gentle push in the back, towards an abyss of solitude and despair, everything around me that had been peace, joy and blessings now turned to torments, accusations and curses. Out of nowhere. No reason... only an echo from within that got heard and answered.

I hope Russell will be up to the considerably more trying challenge he's facing. Everybody is judged by the measure of their pretensions, and their strength too. «God is faithful, and he will not let you be tempted beyond your ability.» I only quoted out of academic erudition (or pedantry), I didn't have the Messiah complex, which in the case of Brand, turned out to become more than a gig, so I'm only in the jail of my mind, rather than in a literal jail with literal rapists.

My recommendation so far is to enjoy what you can, while you can, with no remorse, no guilt, no feeling of selfishness, of being privileged, no time lost suffering for the world in the attempt to ease its load of cruelty and pain. There's so much of it, yours will not make the slightest difference. The time will come sure enough when your soul will have to clean itself and be used to wipe the blood and the excrement of existence. It's coming for everybody. I'm sure of it. I hope Russell knows it too, and that this will help him walk up the rest of the mountain he once claimed he would reach the top of, as if it were a quick hop. He has just started his ascent. It starts in the mud, the infamy, the shame, the scorn, the lies, the abject... It gets worse as you go on. You think the journey could even bring you somewhere, but it only pushes you farther away still; even a flower you may find on the way only has thorns to scorch you away. Yet, I am also sure of it, at some point the path gets clearer. And it becomes light as it reaches the sky.

9 November (2026)

My margin of progression.


Matutinal writings...

in the rain and a very windy still-nightly autumnal morning.

8 November (2026)

Today, I decided how to present our AHBT effect, starting with the "simple" formula:

which shows where all the physics that one has in Eq. (4) originate from, but in a way that is almost purely academic, in the sense that you have to understand precisely what you're doing to keep it true and general. Instead, if we replace it with this more robust and meaningful, yet less pretty, version:

then we can very happily and almost blindly remove confounding factors in any combination (game-theorists speak of "coalitions") and get the experimentally testable result without thinking about it.

However, I could not rescue this beauty in our narrative: \begin{multline}\label{eq:0SMv8}g^{(n)}(\theta,\dots,\theta)-1=\underbrace{\frac{(n+1)!}{2^n}-1}_{\text{intensity}}+\underbrace{g^{(n)}_N\sum_{p\ge1}\binom{n}{2p}\binom{2p}{p}\Bigl(\frac{\pi}{8}\Bigr)^{2p}}_{\text{phase}}\\+\underbrace{g^{(n)}_N\sum_{p\ge1}\binom{n}{2p}\binom{2p}{p}\Bigl[m_{2p}-\Bigl(\frac{\pi}{8}\Bigr)^{2p}\Bigr]}_{\text{shape}}=n!-1\,.\end{multline} which is also true (it's a particular case of the above) but even more so conditioned to the order in which you remove the confounding factors, as it is now very explicit and leaves little room for interpretations of the averages. This is this type of decomposition that greatly excited me earlier in the week until I realized they are at best misleading, which forced me to turn to game theory, and the Shapley coefficients and the Harsanyi dividends. And I think I'm finally getting to the end of it!

Almost... in principle it'd be nice to add two variations:

  • Unbalanced thermal states make the intensity fluctuation cooperate in the bunching.
  • There are various qualitative changes of which confounding factor dominates as a function of $n$, and it'd be nice to show at least one example.

I don't think I have the force today, though, as I also want to add the three-mode theory...


Last Sunday, I put to run my second case of one-atom lasing, nijin (äșŒé™Ł), to have an idea of how those things evolve in the parameter space. I let it run in the background without intermediate peeks. It completed yesterday at 0Tvlf (af), after 5623.2kh of computation (7.5 months worth on a single core), at which point the 3rd campaign, was launched, that would have finished next Friday...

This morning I looked at the results from nijin (for which I made a browsable/zoomable version of {{pan}}, but I won't import it there on laussy.org yet because it's heavy and I don't have time. Here's a snapshot:

The vaginal rugae shape and fractal structure are still there, morphing indeed, and better reproduced for the consistent, linear grid, as attempts at local refinements actually result in noise. I ordered a retreat of the Ten-Go tasks because there's no need to go to 0.004g resolution with the interpolation, and 0.032g is enough. So until I get to resolve the numerical truncation problem and nature of the black lines, I've replaced sanjin with a teisatsu at same cost (finish end of next week), to make a similarly, low-res, but full coverage of Fig. (0NNbp), so as to produce the anticipated animation that will visualize how the skeleton evolves and morphs.

By this time I hope I'll have gotten the chance to sort out the truncation problem to know how to best meaningfully capture the phenomenon. It's a wild one. Maybe the wildest from the simplest possible case.

7 November (2026)

Clafoutis in November, by the chimney... what a time to live in!


I'm publishing on my blog a piece which was invited, but never published, by the Royal Society:

Standing on the shoulders of photons.

They had asked Elena, their fellow, to write a perspective of the human adventure that was behind her Newton fellowship and our work. I wrote that beautiful text, which provided a genuinely original and important observation on the creative scientific process. But the point got lost by the AI of the time, Alice Power (she/her), who presents herself with "I like science" on her social media, but who didn't like what I wrote:

I told Elena that then we don't publish anything. There are conventions and a minimum amount of rules for scientific texts, which I always try to bend, but that I always eventually submit to. But a blog, a personal piece, I hate it when editors of any type try to curb you into their own vision of how one should express things.

The text then stayed in Elena, Alice and I's email inboxes. We probably had sent it to Eduardo to, who understands everything so definitely was fine with it.

But now I decided to publish it anyway, here, if only because the AI crawlers will find it and digest it, and maybe the collective intelligence—neither Alice nor any actual human reader—will get something out of it.


Well, after all, there is some beauty to Fig. (0TfSh). It looks like $\circledcirc$ is a little black hole lost in the height of algebra which sucked-in all the rest of the equation, including material from the margin. Almost it got bits from the following paper. I wonder if we shouldn't leave it like that. Correct everything except equation (2). We might be the first to have the equation number in the superscript...


Here are the results of the numerical experiment:

So the procedure is sound and works pretty well, everything based on raw data from Monte Carlo simulations, removing all possible combinations of coalitions, and patching them up again to take into account their synergy.

The theory is right. Also on â„€ ⟶

6 November (2026)

Speak of other people having a rotten week-end... Russel Brand, who is someone I admire, will likely go to jail, for having allegedly raped someone in 1999. The details are not particularly edifying, but look suspiciously similar to those held against Julian Assange (unwanted practise as part of an otherwise consensual, although casual, encounter). For Assange, I remember Stallman commenting that being woken up by someone you love making love to you is the most blissful experience. I'm not too sure about that. I never thought too much much about it either. I just remember it. Funny that you can remember things you don't think about. In Brand's case, I'm even less sure this would apply. Anyway, the allegations are obviously grotesque and he is clearly innocent. Legally innocent, that is. He is, of course, guilty as much as can be in the eye of God—and it seems to matter to him now—for all the casualness, the addictions, for treating someone—anyone—as if they were one night when nobody should be less than an eternity. Still, when he says that he is not guilty and that «I pray that you can see that, by looking at my eyes», I believe him. I believe that he has something inside him which was already alive when he was given to debauchery and corruption, which is why I admire him. He had the innocence of the child to fall for temptation, the weakness of youth to let it spiral out of control, and the strength of man to recognize it and transform himself, from a fallen angel to one back on his feet. He looks sincere, down to his vice and depravity. He's now wiping this out with his soul. He's not hiding what is ugly as if to make it non-existent. He's facing it, he's putting Russel Brand in front of what was devouring him. This is rare. For people are not sincere. They are fake, or pausing, or idiots. All of which exclude sincerity. When you find someone who is—sincere—even if he is damned and confessing his turpitude, this matters little. The need, the courage, the lucidity to tell oneself absolves everything. Especially if what you have to confess is ugly, is sordid, is disgusting. There is this Hebraic story of Moses being wicked, which is why God chose him, for it is only with the weight of sin and hardship, once you get rid of it, that you can ascend towards truth and beauty, like a bubble at the bottom of the sea.


What's up for your Friday evening? Here the plans for mine: back to formulating an experimental protocol for coalitions in thermal bunching.

I'll generate Monte Carlo thermal frames of bi-modal vortices.[11] I'll ask Claude to extract from this I-free, stratified, aligned-frames, and from those, to compute all the $g^{(n)}_T$, even if they are not all needed. I want to see everything. I'm like that. And then from those, we'll finally compute the Harsanyi dividends:

\begin{gathered} \begin{aligned} \mathrm{I} &= g^{(n)}_\mathrm{I}\,, & \mathrm{IP} &= \frac{g^{(n)}_\mathrm{IP}}{g^{(n)}_\mathrm{I}\,g^{(n)}_\mathrm{P}}\,,\\ \mathrm{P} &= g^{(n)}_\mathrm{P}\,, & \mathrm{IS} &= \frac{g^{(n)}_\mathrm{IS}}{g^{(n)}_\mathrm{I}\,g^{(n)}_\mathrm{S}}\,,\\ \mathrm{S} &= g^{(n)}_\mathrm{S}\,, & \mathrm{PS} &= \frac{g^{(n)}_\mathrm{PS}}{g^{(n)}_\mathrm{P}\,g^{(n)}_\mathrm{S}}\,, \end{aligned}\\[1ex] \mathrm{IPS} = \frac{g^{(n)}_\mathrm{IPS}\,g^{(n)}_\mathrm{I}\,g^{(n)}_\mathrm{P}\,g^{(n)}_\mathrm{S}}{g^{(n)}_\mathrm{IP}\,g^{(n)}_\mathrm{IS}\,g^{(n)}_\mathrm{PS}}\,. \end{gathered}

And plot them. If it works, I'll ask Alexandros to do this in his lab. Looks like fun? Did you have a better Friday evening? Probably yes... But if you don't tell the rest of the world about it, who knows, really?

I'll be looking at what other people did as the computer Monte Carles (I just made up this new denominal verb, you see, my week-end is getting better already). Also on â„€ ⟶


Visiting the neighbour. Carlos explaining his beautiful SPDC source.


Proofs for bosoc are giving me stomach ache...

It looks like some AI handled the transcription of our beautiful $\mathrm{\TeX}$ into their house-style version because the systematic chunking of the equation into the superscript of its variable looks like a stubborn machine hallucinating its version of things. The equation number that joined the wreck in the superscript could be someone's fault, though, if only to fail noticing this just does not compile.


All are nice to reproduce experimentally, but Fig. (0SuXX) (what an interesting misnomer) should be particularly convenient because the stratification can produce all the required states for the $x$ axis. The procedure would go something like this:

  • All frames are aligned by rotating them so that the lobe of each frame sits at $\theta=0$ (by a fit of the frame's own photons in an experiment, in absence of an oracle).
  • Stratification are then made by shapes, e.g, by selecting the contrast to lie in a given window. There are more dipoles than donuts, so more signal on one side.
  • Then pooling different strata from the ensemble, we reconstruct the various $g^{(2)}_T$ curves ($T$ the "coalition"), which, from Eq. \eqref{eq:0SXYY} and the definitions above, finally give us the Harsanyi dividends. Then we have fun multiplying them and seeing them reconstruct collectively the thermal (HBT) fluctuations.

Removing the intensity fluctuation is a bit tricky. It doesn't consist in fixing the number of particles: this complete absence of fluctuation is in itself a strong correlation—that of Fock states—so although it becomes a good one as $N\gg1$, it is not exact. No-intensity-fluctuation in a random process doesn't mean a literal absence of fluctuations, it means Poisson fluctuations instead. Correlations computed on such frames could be corrected by dividing by appropriate factors, but although it's the most efficient (no data loss), somehow I don't like doing this. I'd see it a more robust and foolproof protocol to strip down the intensity fluctuations, which can be done my making a genuine random selection.

Let us say we want $\nu$ as the new mean. Ideally it should be that $\nu\ll\mu$ the thermal average, in fact $\nu$ should be well below the smallest number of photons, which for thermal is zero, so some careful bias analysis will be needed there. Now let us make a Poisson sampling of $k$ random photons from each frame, with $k$ drawn afresh for every frame with Poisson mean $\nu$ and the $k$ photons picked at random without replacement (discard the rest), then we have removed our intensity confounder: the frame is exactly what a source of fixed mean $\nu$ with the same shape and phase would have given. I call such frames I-free frames below. In the table, I label the I-free "resampled" to focus on the procedure to get them "my way."

Then:

  1. Fluctuations of the bucket detector (whole camera) should give us $g^{(2)}_I=3/2$, constant for all strata.
  2. Fluctuations of the shape-frozen I-free frames, with only the phase fluctuation, should give us $g^{(2)}_P=1+2a_0^2\cos2\theta$, going up to the 3/2 of RPCS when doing it with dipoles. Doing it with donuts of course gives us 1 as they are themselves phase invariant.
  3. Fluctuations of the intensity-fluctuating shape-frozen frames should give us the intensity-phase fluctuating $g^{(2)}_{IP}=\tfrac32(1+2a_0^2)$, the product of the two previous ones, since the intensity is independent of the phase (dummy player).
  4. Fluctuations of the I-free phase-aligned all (no-shape frozen) frames should give us $g^{(2)}_S=\langle(1+2ac)^2\rangle/(1+2m_1c)^2$ with $c=\cos2\theta$, the aligned ensemble over its own singles. This is Fig. 4(a) of Ergoli's submitted version: $1.016$ at the lobe, $2.08$ at the waist, $0.87$ for the lobe-waist pair.
  5. Fluctuations of the intensity fluctuating, phase-aligned, all (no-shape frozen) frames should give us $g^{(2)}_{IS}=\tfrac32\,g^{(2)}_S$ for balanced modes, $1.52$ at the lobe and $3.12$ at the waist.
  6. Fluctuations of I-free, all (no-shape frozen), phase-fluctuating frames should give us $g^{(2)}_{PS}=\langle(1+2ac)^2\rangle_{a,\phi}=1+2m_2=\tfrac43$ at any $\theta$ ($1+\tfrac13\cos2\theta$ for the lobe–$\theta$ pair), the intensity-free thermal correlation
  7. Fluctuations of everything is the full thermal ensemble, and should give us $g^{(2)}=2$ (thermal light) at coincidence, $1+\cos^2\theta$ for the azimuthal pairs.
  8. Last but not least—the first in the table below—fluctuations of I-free, shape-frozen, phase-aligned frames should give us... $g^{(2)}=1$, uncorrelated light! Simpson's paradox.[12]
Glauber correlators for all the combinations of the confounding factors.
correlator frames aligned any $n$ $n=2$, lobe / waist
$g^{(n)}_\varnothing$ stratum $a_0$, resampled yes $1$ $1$ / $1$
$g^{(n)}_{\mathrm I}$ stratum $a_0$, as recorded yes $(n+1)!/2^n$ $\tfrac32$ / $\tfrac32$
$g^{(n)}_{\mathrm P}$ stratum $a_0$, resampled no $\sum_p\binom{n}{2p}\binom{2p}{p}a_0^{2p}$ $1+2a_0^2$ / same
$g^{(n)}_{\mathrm{IP}}$ stratum $a_0$, as recorded no $\frac{(n+1)!}{2^n}\sum_p\binom{n}{2p}\binom{2p}{p}a_0^{2p}$ $\tfrac32(1+2a_0^2)$ / same
$g^{(n)}_{\mathrm S}$ all frames, resampled yes $\langle(1+2ac)^n\rangle/(1+2m_1c)^n$ $1.016$ / $2.08$
$g^{(n)}_{\mathrm{IS}}$ all frames, as recorded yes $\frac{(n+1)!}{2^n}\,\langle(1+2ac)^n\rangle/(1+2m_1c)^n$ $1.52$ / $3.12$
$g^{(n)}_{\mathrm{PS}}$ all frames, resampled no $2^n/(n+1)$ $\tfrac43$ / $\tfrac43$
$g^{(n)}_{\mathrm{IPS}}$ all frames, as recorded no $n!$ $2$ / $2$

This gives us the following results:

The correlators that do not depend on $a_0$ being defined (or frozen) are shown outside of the figure, but aligned with its $y$-axis anyway. Those that are constant although a function of $a_0$, are shown as horizontal line over the entire span of the $x$-axis. Also on â„€ ⟶

5 November (2026)

Let's put some numbers on 0SpeK and draw some figures (and some conclusions).

Let's say we freeze our shape to that of the mean collapse. Then the decomposition $g^{(2)}_{\mathrm{IPS}}=\mathrm{I}\cdot\mathrm{P}\cdot\mathrm{S}\cdot\mathrm{PS}$ evaluates numerically to the table below, which is the same whether $\mathrm S$ is normalized by the frozen fringe or by its own singles, the mean fringe $1+2m_1c$ being both; the two conventions part company at any other reference:

Factors of $g^{(2)}$ for two-mode thermal light along the pinned fringe, shape frozen at the mean collapse, $a_0=\pi/8$.
$\theta$ coincidence, $g^{(2)}(\theta,\theta)$ cross-correlations, $g^{(2)}(0,\theta)$
I P S PS $\mathrm{I}\cdot\mathrm{P}\cdot\mathrm{S}\cdot\mathrm{PS}$ I P S PS $\mathrm{I}\cdot\mathrm{P}\cdot\mathrm{S}\cdot\mathrm{PS}$
$0$ 1.5000 1.3084 1.0156 1.0034 2.0000 1.5000 1.3084 1.0156 1.0034 2.0000
$\pi/8$ 1.5000 1.3084 1.0103 1.0087 2.0000 1.5000 1.2181 1.0127 1.0018 1.8536
$\pi/4$ 1.5000 1.3084 1.0000 1.0190 2.0000 1.5000 1.0000 1.0000 1.0000 1.5000
$3\pi/8$ 1.5000 1.3084 1.1260 0.9050 2.0000 1.5000 0.7819 0.9556 1.0229 1.1464
$\pi/2$ 1.5000 1.3084 2.0817 0.4895 2.0000 1.5000 0.6916 0.8700 1.1080 1.0000

If we freeze for the dipole instead, with every correlator normalized by its own singles, so that the shape factor $\mathrm S$ is the aligned ensemble over its own mean fringe and does not refer to the frozen shape, which enters $\mathrm P$ alone:

The same with the shape frozen at the dipole, $a_0=\tfrac12$, so that P is the RPCS; S is the same as above, every correlator being normalized by its own singles, and PS is the remainder.
$\theta$ coincidence, $g^{(2)}(\theta,\theta)$ cross-correlations, $g^{(2)}(0,\theta)$
I P S PS $\mathrm{I}\cdot\mathrm{P}\cdot\mathrm{S}\cdot\mathrm{PS}$ I P S PS $\mathrm{I}\cdot\mathrm{P}\cdot\mathrm{S}\cdot\mathrm{PS}$
$0$ 1.5000 1.5000 1.0156 0.8752 2.0000 1.5000 1.5000 1.0156 0.8752 2.0000
$\pi/8$ 1.5000 1.5000 1.0103 0.8798 2.0000 1.5000 1.3536 1.0127 0.9015 1.8536
$\pi/4$ 1.5000 1.5000 1.0000 0.8889 2.0000 1.5000 1.0000 1.0000 1.0000 1.5000
$3\pi/8$ 1.5000 1.5000 1.1260 0.7894 2.0000 1.5000 0.6464 0.9556 1.2372 1.1464
$\pi/2$ 1.5000 1.5000 2.0817 0.4270 2.0000 1.5000 0.5000 0.8700 1.5326 1.0000

The S columns are those of the mean collapse; P is the RPCS, flat at coincidence and $1+\tfrac12\cos2\theta$ for the pair; PS takes what the reference moves, and the waist is regular.

This is the evolution of those quantities as function of the frozen shape (the two tables are on the $x$-axis):

S is flat in $a_0$ (1.016 at the lobe, 2.08 at the waist, 0.87 for the lobe–waist pair), P rises from 1 to the RPCS, and PS carries the rest, at fixed product; nothing diverges.

And this is as a function of $\theta$ (now plotting the column of the tables as $x$-axis):

Also on â„€ ⟶


This cooperativity game theory was a bit of a downer. I liked a lot the Shapley values for the additive game, first because the concept itself is quite interesting. Wikipedia page gives a good account. We're trying to distribute the amount of correlations between "players" (confounding factors) who contribute to it, in a cooperative (interacting) way. Second, I liked it because they looked qualitatively like my initial approach, just adding cooperativity. We're adding! Neat.

But then it turns out that $g^{(n)}$ behaves as a multiplicative game, and cannot be fully defined otherwise (coincidences can, cross-correlations cannot). This comes with strong and wanted properties like anticorrelations being equally valuable than correlations (through the use of the log), which is physical. In the additive game, anticorrelations are a loss. So multiplicative game it is!

Problem is, then, that it behaves qualitative very differently than the additive and the initial decomposition. While I had before interplay of phase and shape even overcoming intensity contributions, in the multiplicative game, intensity fluctuations always dominate overwhelmingly and the behaviour is boring. That's a typical one:

I'e got an infinite amount of variations. In thick black is the "gain" (total thermal correlations) the "players" (confounding factors) are getting a contribution to, a part of. The intensity, in grey, is a dummy player: it doesn't interact or collaborate with the others. In this picture, you see it's the bulk of the total, one of the things I don't like. There's lobe & waist (and also ring, elsewhere) because those are very different when the phase is pinned, and I want to look at both. The phase-waist indeed is very different, starting with the fact it's negative! All those quantities, when belonging together (waist with waist, lobe with lobe, etc.) add to $\ln n!$ (additions of ln is a product, we're playing multiplications, remember?)

What's the physics in that? Buahhh. Disgusting. At best, anecdotal. I don't like it.

I was happy earlier in the day, when there were things I was starting to describe as "not surprisingly" or "expectedly"... So I try to understand why various ways to look at it behave so differently, and it boils down to, $g^{(n)}$ being really a multiplicative game and describing it as additive gives it more structure but not more substance. It's basically meaningless (or subtle to a point of being useless).

What a bummer.

So I have to play with multiplicative games. If they're boring, it's because the universe is, I don't decide the rules of the game, I just play it.

Let us write our thermal correlations in this form: \begin{equation} \label{eq:0Sx9q} g^{(n)}(\theta_1,\dots,\theta_n)=\bigl[g^{(n)}_N\bigr]^{\iota}\times \begin{cases} \displaystyle\sum_{k=0}^{n}w_k\,|e_k(z)|^2 & \text{phase on ($\phi$ uniform)}\,,\\[3ex] \displaystyle\frac{\sum_{k=0}^{n}2^k\mu_k\,e_k(c)}{\prod_{j=1}^n(1+2\mu_1c_j)} & \text{phase off ($\phi=\phi_0$)}\,, \end{cases} \end{equation} where $\iota=1$ or $0$ for the intensity on or off, $z_j\equiv e^{i\Delta\ell\theta_j}$, $c_j\equiv\cos(\Delta\ell\,\theta_j-\phi_0)$ for the pinned phase $\phi_0$, $e_k$ are the elementary symmetric polynomials ($e_0\equiv1$, $e_1\equiv\sum_jz_j$, $\dots$, $e_n\equiv\prod_jz_j$), and $(w_k,\mu_k)\equiv(\tau_{n,k},m_k)$ with the shape on, and $(t_0^{\,n-k}(1-t_0)^k,a_0^k)$ with it off, in terms of the bucket (full-camera) correlator $g^{(n)}_N\equiv\langle W^n\rangle/\langle W\rangle^n$, the moments $m_k\equiv\langle a^k\rangle$ of the contrast $a\equiv\sqrt{t(1-t)}$ of a collapse of mode balance $t$, the shape law $\tau_{n,k}\equiv\langle t^{n-k}(1-t)^k\rangle$ and the frozen shape $a_0^2\equiv t_0(1-t_0)$. For two-mode thermal light, $g^{(n)}_N=(n+1)!/2^n$, $\tau_{n,k}=(n-k)!\,k!/(n+1)!$ and $m_k=\Gamma(\tfrac k2+1)^2/\Gamma(k+2)$, so $m_1=\tfrac{\pi}8$ and $m_2=\tfrac16$.

Here we see i) products, we see ii) intensity fluctuations as a dummy player and we see iii) the collaboration: phase and shape intermingle.

I won't bore you with how I spent my day circling around those Shapley values, and jump directly to what I think is the best picture of it. Meet the Harsanyi dividends: for a coalition $T$ of confounding factors, worth $u(T)\equiv\ln g^{(n)}_T$ with $g^{(n)}_T$ the correlator when only the factors in $T$ fluctuate ($u(\varnothing)=0$), the dividend is what $T$ is worth beyond the dividends of all its sub-coalitions, $m(T)\equiv\sum_{U\subseteq T}(-1)^{|T|-|U|}u(U)$, i.e., $m(\mathrm I)=\ln g^{(n)}_{\mathrm I}$, $m(\mathrm{IP})=\ln g^{(n)}_{\mathrm{IP}}-\ln g^{(n)}_{\mathrm I}-\ln g^{(n)}_{\mathrm P}$, $m(\mathrm{PS})=\ln g^{(n)}_{\mathrm{PS}}-\ln g^{(n)}_{\mathrm P}-\ln g^{(n)}_{\mathrm S}$, etc., the seven adding up to $\ln g^{(n)}_{\mathrm{IPS}}=\ln n!$.

In our case, we have this holy family: \begin{equation} \label{eq:0SXYY} \begin{aligned} \mathrm{I}&\equiv e^{m(\mathrm I)}=g^{(n)}_N=\frac{\langle W^n\rangle}{\langle W\rangle^n}\,,\\ \mathrm{P}&\equiv e^{m(\mathrm P)}=g^{(n)}_{\mathrm P}=\mathbb{E}_\phi\prod_{j=1}^n(1+2a_0c_j)\,,\\ \mathrm{S}&\equiv e^{m(\mathrm S)}=g^{(n)}_{\mathrm S}=\frac{\langle\prod_j(1+2ac_j)\rangle_a}{\prod_j(1+2m_1c_j)}\,,\\ \mathrm{PS}&\equiv e^{m(\mathrm{PS})}=\frac{g^{(n)}_{\mathrm{PS}}}{\mathrm{P}\cdot\mathrm{S}}\,,\quad g^{(n)}_{\mathrm{PS}}=\Bigl\langle\prod_{j=1}^n(1+2ac_j)\Bigr\rangle_{a,\phi}\,, \end{aligned} \end{equation}

with $c_j\equiv\cos(\Delta\ell\,\theta_j-\phi_0)$ for the pinned phase $\phi_0$ and $a_0$ the frozen contrast.

$\mathrm{I}$ is dummy, so there is no $\mathrm{IP}$, $\mathrm{IS}$, etc. (etc. in this case being $\mathrm{IPS}$, there's no more). So we only have $\mathrm{PS}$ as the "dividend", the collaborative, synergistic—you name it—action of the phase and the shape together beyond what each does alone. The so-called Möbius property states that a "coalition" is worth the sum of the dividends of its sub-coalitions, which, reads as a product, gives: \begin{gather} g^{(n)}_{\mathrm{IPS}}=\mathrm{I}\cdot\mathrm{P}\cdot\mathrm{S}\cdot\mathrm{PS}\,,\quad g^{(n)}_{\mathrm{PS}}=\mathrm{P}\cdot\mathrm{S}\cdot\mathrm{PS}\,,\quad g^{(n)}_{\mathrm{IP}}=\mathrm{I}\cdot\mathrm{P}\,,\quad g^{(n)}_{\mathrm{IS}}=\mathrm{I}\cdot\mathrm{S}\,,\label{eq:0T1kQ}\\ g^{(n)}_{\mathrm{I}}=\mathrm{I}\,,\qquad g^{(n)}_{\mathrm{P}}=\mathrm{P}\,,\qquad g^{(n)}_{\mathrm{S}}=\mathrm{S}\,,\qquad g^{(n)}_{\varnothing}=1\,. \end{gather}

That is, in essence, the HBT effect as an interplay of confounding factors. Removing a confounder removes every factor that carries its letter. This is Eq. (\ref{eq:0Sx9q}) rewritten as eight combinations, that reduce to four independent cases once the intensity is a factor.

For two-mode thermal light at coincidence, Eq. \eqref{eq:0SXYY} above evaluates to: \begin{align} \mathrm{I}&=(n+1)!/2^n\,.\\ \mathrm{P}&=\sum_p\binom{n}{2p}\binom{2p}{p}a_0^{2p}\,.\label{eq:0TUcX}\\ \mathrm{S}&=\sum_k\binom nk(2c)^km_k/(1+2m_1c)^n\,.\\ \mathrm{PS}&=[2^n/(n+1)]/(\mathrm{P}\cdot\mathrm{S})\,. \end{align}

Next I give some examples and plots. Also on â„€ ⟶


I need people who like what's different. Where are they?

–:––

Je connais un voleur
Si gentil
Quâ€Čil a volĂ© mon coeur
Et depuis
J'aime bien les voleurs


My dilemma is the following. All the combinations of how the confounding factors contributes to the correlations can be listed as follows:

$g^{(n)}(\theta,\dots,\theta)$ for two-mode thermal light. The confounding factors are those fluctuating.
$n$ Same value everywhere S, phase pinned IS, phase pinned
none I P IP PS IPS lobe waist lobe waist
2 1 1.5 1.308 1.963 1.333 2 1.016 2.082 1.523 3.122
3 1 3 1.925 5.776 2 6 1.045 5.540 3.134 16.62
4 1 7.5 2.993 22.45 3.2 24 1.086 16.76 8.142 125.7
5 1 22.5 4.798 107.9 5.333 120 1.138 54.89 25.60 1235
6 1 78.75 7.840 617.4 9.143 720 1.201 189.8 94.57 14946

The letters name the confounding factors present, all=IPS. The shape here is frozen at the mean collapse $a_0=\pi/8$ when off. With the phase on, or with no fringe at all (none, I), the value is the same at every point of the ring; with the phase pinned and the shape on (S, IS) it is given at the lobe, $c=1$, and at the waist, $c=-1$: $\sum_k(\pm2)^km_k\binom nk/(1\pm\tfrac\pi4)^n$, times $g^{(n)}_N$ for IS.

So this is clearly not additive, i.e., one confounding factors affect the others. I was a bit naive (and enthusiastic) of me to find in a mathematical decomposition a physical meaning to something that was just unravelling nicely on paper (or constraining to some nesting of the effects, which is what I don't like).

Now, looking around, the way such interconnected things are described is with the Shapley value, to describe situations beyond those when «games do not depend on other games»:[13]

More related to us, in regression analysis, this is known as LMG or dominance analysis. I am now looking into that. Maybe another dead end.

Shapley values. With the cube above as a cooperative game, the Shapley value of a confounding factor is its increment averaged over the six orders in which the three can be switched on, and the three values add up to the whole whatever the order. In the additive game, where a coalition is worth $g-1$, pinned coalitions at the lobe:

\begin{align} g^{(2)}(\theta,\theta)-1&=1=0.583+0.392+0.026\,,\nonumber\\ g^{(3)}(\theta,\theta,\theta)-1&=5=2.990+1.886+0.124\,,\nonumber\\ g^{(4)}(\theta,\dots,\theta)-1&=23=13.519+8.794+0.687\,,\nonumber\\ g^{(5)}(\theta,\dots,\theta)-1&=119=66.658+47.673+4.669\,,\label{eq:0SOeU}\\ g^{(6)}(\theta,\dots,\theta)-1&=719=\underbrace{380.024}_{\text{intensity}}+\underbrace{301.857}_{\text{phase}}+\underbrace{37.119}_{\text{shape}}\,,\nonumber \end{align}

against the chain of Eq. \eqref{eq:0ShpF}: the intensity–phase interaction goes half to each instead of wholly to the phase, which arrived second, and the shape keeps a third of what it only does together with the other two. In the multiplicative game, where a coalition is worth $\ln g$, the values are logarithms of factors, and the intensity's factor is exact (it multiplies everything and interacts with nothing):

\begin{align} g^{(2)}(\theta,\theta)&=2=1.500\times1.311\times1.017\,,\nonumber\\ g^{(3)}(\theta,\theta,\theta)&=6=3.000\times1.920\times1.042\,,\nonumber\\ g^{(4)}(\theta,\dots,\theta)&=24=7.500\times2.970\times1.077\,,\nonumber\\ g^{(5)}(\theta,\dots,\theta)&=120=22.500\times4.742\times1.125\,,\label{eq:0SUzo}\\ g^{(6)}(\theta,\dots,\theta)&=720=\underbrace{78.750}_{\text{intensity}}\times\underbrace{7.726}_{\text{phase}}\times\underbrace{1.183}_{\text{shape}}\,,\nonumber \end{align}

or in logarithms for $n=5$:

\begin{equation} \label{eq:0Srlz} \ln g^{(5)}(\theta,\dots,\theta)=\ln120=4.787=3.114+1.557+0.117\,. \end{equation}

The chain and the Shapley value nearly coincide in this game at the lobe (the chain's factors at $n=5$ are $22.5\times4.798\times1.112$). The product of the phase and shape factors is exactly the PS coalition, $1.311\times1.017=4/3$, but the products with the intensity are not the IP and IS coalitions ($1.5\times1.311=1.967$ against $1.963$): three numbers reproduce the seven coalitions only to the extent that the phase–shape dividend vanishes.

Two points. For $g^{(2)}(0,\theta)$, $\theta_1=0$ the lobe of the pinned fringe and $\theta$ the second detector, the three intensity-free coalitions are in closed form, $g^{(2)}_{\mathrm P}=1+2a_0^2\cos2\theta$ (phase alone), $g^{(2)}_{\mathrm S}=\frac{1+2m_1(1+\cos2\theta)+\frac23\cos2\theta}{(1+2m_1)(1+2m_1\cos2\theta)}$ (shape alone), $g^{(2)}_{\mathrm{PS}}=\frac23(1+\cos^2\theta)$ (both), each coalition being $[g^{(2)}_N]^{\iota}F_T$ with $g^{(2)}_N=\frac32$. The additive Shapley values are

\begin{align} \varphi_{\mathrm I}&=(g^{(2)}_N-1)\Bigl[\tfrac13+\tfrac16(g^{(2)}_{\mathrm P}+g^{(2)}_{\mathrm S})+\tfrac13g^{(2)}_{\mathrm{PS}}\Bigr]\,,\nonumber\\ \varphi_{\mathrm P}&=(g^{(2)}_{\mathrm P}-1)\bigl(\tfrac13+\tfrac16g^{(2)}_N\bigr)+(g^{(2)}_{\mathrm{PS}}-g^{(2)}_{\mathrm S})\bigl(\tfrac16+\tfrac13g^{(2)}_N\bigr)\,,\qquad\varphi_{\mathrm S}=\text{the same with P}\leftrightarrow\text{S}\,,\label{eq:0Sn87} \end{align}

adding up to $g^{(2)}(0,\theta)-1=\cos^2\theta$ (the intensity's value is not constant: in the additive game it is not a dummy player, its increment $(g^{(2)}_N-1)F_T$ depends on the coalition it joins), and the multiplicative ones, as factors,

\begin{equation} \label{eq:0SFYJ} e^{\varphi_{\mathrm I}}=g^{(2)}_N=\tfrac32\,,\qquad e^{\varphi_{\mathrm P}}=\sqrt{g^{(2)}_{\mathrm P}\,\frac{g^{(2)}_{\mathrm{PS}}}{g^{(2)}_{\mathrm S}}}\,,\qquad e^{\varphi_{\mathrm S}}=\sqrt{g^{(2)}_{\mathrm S}\,\frac{g^{(2)}_{\mathrm{PS}}}{g^{(2)}_{\mathrm P}}}\,,\qquad e^{\varphi_{\mathrm I}}e^{\varphi_{\mathrm P}}e^{\varphi_{\mathrm S}}=1+\cos^2\theta\,, \end{equation}

each the geometric mean of the player's two factors, alone and on top of the other. Along the fringe:

Shapley decomposition of $g^{(2)}(0,\theta)$, shape frozen at the mean collapse when off
$\theta$ coalitions additive: $g^{(2)}-1=\varphi_{\mathrm I}+\varphi_{\mathrm P}+\varphi_{\mathrm S}$ multiplicative: $g^{(2)}=\tfrac32\,e^{\varphi_{\mathrm P}}e^{\varphi_{\mathrm S}}$
$g^{(2)}_{\mathrm P}$ $g^{(2)}_{\mathrm S}$ $g^{(2)}_{\mathrm{PS}}$ $\cos^2\theta$ $\varphi_{\mathrm I}$ $\varphi_{\mathrm P}$ $\varphi_{\mathrm S}$ $1+\cos^2\theta$ $e^{\varphi_{\mathrm P}}$ $e^{\varphi_{\mathrm S}}$
$0$ 1.308 1.016 1.333 1.000 0.583 +0.392 +0.026 2.000 1.311 1.017
$\pi/8$ 1.218 1.013 1.236 0.854 0.559 +0.276 +0.019 1.854 1.219 1.014
$\pi/4$ 1.000 1.000 1.000 0.500 0.500 +0.000 +0.000 1.500 1.000 1.000
$3\pi/8$ 0.782 0.956 0.764 0.146 0.439 -0.255 -0.038 1.146 0.791 0.966
$\pi/2$ 0.692 0.870 0.667 0.000 0.408 -0.315 -0.092 1.000 0.728 0.916

At the quarter fringe the three intensity-free coalitions are all $1$ and both games give the whole of $g^{(2)}(0,\pi/4)=\tfrac32$ to the intensity; at the waist the phase and the shape anticorrelate the lobe with the waist, by the factors $0.73$ and $0.92$, and the intensity's bunching by $\tfrac32$ cancels them exactly: $g^{(2)}=1$ there is a cancellation, not an absence of correlation.

Dividends. The Harsanyi dividends (Möbius transform of the worth: what a coalition is worth beyond the sum of the dividends of its sub-coalitions; the Shapley value gives each member of a coalition an equal part of its dividend) are, for any point and in closed form, with $m(T)$ the dividend of $T$,

\begin{align} \text{additive:}&\quad m(\mathrm I)=g^{(n)}_N-1\,,\qquad m(\mathrm P)=g^{(n)}_{\mathrm P}-1\,,\qquad m(\mathrm S)=g^{(n)}_{\mathrm S}-1\,,\nonumber\\ &\quad m(\mathrm{IP})=(g^{(n)}_N-1)(g^{(n)}_{\mathrm P}-1)\,,\qquad m(\mathrm{IS})=(g^{(n)}_N-1)(g^{(n)}_{\mathrm S}-1)\,,\nonumber\\ &\quad m(\mathrm{PS})=g^{(n)}_{\mathrm{PS}}-g^{(n)}_{\mathrm P}-g^{(n)}_{\mathrm S}+1\,,\qquad m(\mathrm{IPS})=(g^{(n)}_N-1)\,m(\mathrm{PS})\,;\label{eq:0SCwt}\\ \text{multiplicative:}&\quad m(\mathrm I)=\ln g^{(n)}_N\,,\qquad m(\mathrm P)=\ln g^{(n)}_{\mathrm P}\,,\qquad m(\mathrm S)=\ln g^{(n)}_{\mathrm S}\,,\nonumber\\ &\quad m(\mathrm{PS})=\ln\frac{g^{(n)}_{\mathrm{PS}}}{g^{(n)}_{\mathrm P}g^{(n)}_{\mathrm S}}\,,\qquad m(\mathrm{IP})=m(\mathrm{IS})=m(\mathrm{IPS})=0\,,\nonumber \end{align}

the seven adding up to $g-1$ and to $\ln g$ respectively. In the multiplicative game the intensity has no dividend with anyone (a dummy player); in the additive game every dividend with the intensity is $(g_N-1)$ times the dividend without it. For $g^{(2)}(0,\theta)$:

Harsanyi dividends of $g^{(2)}(0,\theta)$
$\theta$ additive, sum $=\cos^2\theta$ multiplicative, sum $=\ln(1+\cos^2\theta)$
I P S IP IS PS IPS I P S PS
$0$ +0.500 +0.308 +0.016 +0.154 +0.008 +0.009 +0.005 +0.405 +0.269 +0.016 +0.003
$\pi/8$ +0.500 +0.218 +0.013 +0.109 +0.006 +0.005 +0.002 +0.405 +0.197 +0.013 +0.002
$\pi/4$ +0.500 +0.000 +0.000 +0.000 +0.000 +0.000 +0.000 +0.405 +0.000 +0.000 +0.000
$3\pi/8$ +0.500 -0.218 -0.044 -0.109 -0.022 +0.027 +0.013 +0.405 -0.246 -0.045 +0.023
$\pi/2$ +0.500 -0.308 -0.130 -0.154 -0.065 +0.105 +0.053 +0.405 -0.369 -0.139 +0.103

At the lobe the phase–shape dividend is small in both games ($+0.0034$ nats: the two nearly separate, $g^{(2)}_{\mathrm{PS}}\simeq g^{(2)}_{\mathrm P}g^{(2)}_{\mathrm S}$) and the additive value is mostly in the singletons and the IP interaction; at the waist the PS dividend is what the Shapley value halves between the two. (Report 0Soyr, Secs. 11–17.) Also on â„€ ⟶


Eq. \eqref{eq:0SCpX} is general for any quantum state of bi-modal vortices (phase invariant, probably, to check). The explicit case for thermal states is: \begin{multline} \label{eq:0Sanf} g^{(n)}(\theta_1,\dots,\theta_n)-1 =\underbrace{g^{(n)}_N-1}_{\text{intensity}} +\underbrace{g^{(n)}_N\Bigl[\sum_{k=0}^nt_0^{\,n-k}(1-t_0)^k|e_k(z)|^2-1\Bigr]}_{\text{phase}}\\ +\underbrace{g^{(n)}_N\sum_{k=0}^n\bigl[\tau_{n,k}-t_0^{\,n-k}(1-t_0)^k\bigr]|e_k(z)|^2}_{\text{shape}}\,, \end{multline} where $z_j\equiv e^{i\Delta\ell\theta_j}$, $e_k(z)$ are the elementary symmetric polynomials of the $z_j$ ($e_0=1$, $e_1=\sum_jz_j$, $\dots$, $e_n=\prod_jz_j$), $t_0$ is the mode balance of the frozen shape, $a_0^2=t_0(1-t_0)$ ($t_0=\tfrac12$ for the dipole, $0.19$ for the mean collapse $a_0=\pi/8$), $g^{(n)}_N\equiv\langle N(N-1)\cdots(N-n+1)\rangle/\langle N\rangle^n$ is the whole-ring (bucket) correlator, $(n+1)!/2^n$ for two thermal modes, and $\tau_{n,k}\equiv\langle W^nt^{n-k}(1-t)^k\rangle/\langle W^n\rangle$ is the shape law of the state at order $n$, the law of the mode balance $t$ weighted by $W^n$, which is $(n-k)!\,k!/(n+1)!$ for two thermal modes ($t$ uniform).

Then the autocorrelation case reads nicely: \begin{multline} \label{eq:0S4Qt} g^{(n)}(\theta,\dots,\theta)-1= \underbrace{\frac{(n+1)!}{2^n}-1}_{\text{intensity}} +\underbrace{g^{(n)}_N\sum_{p\ge1}\binom{n}{2p}\binom{2p}{p}a_0^{2p}}_{\text{phase}}\\ +\underbrace{g^{(n)}_N\sum_{p\ge1}\binom{n}{2p}\binom{2p}{p}\bigl[m_{2p}-a_0^{2p}\bigr]}_{\text{shape}} =n!-1\,. \end{multline}

as a function of the shape $a_0$ one wants to refer to, with $m_{2p}\equiv\langle a^{2p}\rangle=(p!)^2/(2p+1)!$ the even moments of the thermal contrast $a=\sqrt{t(1-t)}$. This could be the mean sample (my choice so far, for the reasons previously explained, the atom (meaningful), or the dipole (interesting).

In this case, the decompositions in Eq. \eqref{eq:0Sc1u} are state dependent. They involve negative contributions for the dipoles. And this is what made me realize that this decomposition is not truly additive, it requires one respects a nesting order:

PSI $\to$ PS $\to$ S $\to$ none

where PSI is removing all confounding factors, I (intensity), P (phase) and S (shape), in the order that unests them. Removing all gives $g^{(n)}(\theta_1,\dots,\theta_n)=1$. Removing none gives what everybody is doing.

I don't like it, though. It's too complicated. I'm looking for something more "modular": remove this confounding factor, and then you get "this", but in any order. And how does this contribute to the whole? Also on â„€ ⟶


It's hardly six in the morning and I already got my interesting result for the day. The shape fluctuations, that are essentially negligible for $g^{(2)}$, "quickly" become the dominant contribution to multiphoton correlations. I put quickly between quotes because that happens for $g^{(14)}$ and who does fourteen-photon correlations? Still, the evolution is remarkable:

People who think thermal fluctuations (or correlations) are (due to) intensity fluctuations are not multiphotonics people.

Let us formulate an experiment to measure that. Also on â„€ ⟶


I spent the night on the $n$th-order correlator $g^{(n)}$ of the AHBT, whose generalization reads: \begin{equation} \label{eq:0ShQq} g^{(n)}(\theta_1,\dots,\theta_n) =g^{(n)}_N\,\frac{\sum_{k=0}^{n}2^km_k\,\mathbb{E}_\phi[e_k]} {\prod_{j=1}^{n}\mathbb{E}_\phi[1+2m_1c_j]} \end{equation} and its confounding factor decomposition becomes: \begin{multline} \label{eq:0SCpX} g^{(n)}=1+\underbrace{(g^{(n)}_N-1)}_{\text{intensity}}\\ +g^{(n)}_N\,\frac{\overbrace{\mathbb{E}_\phi\Bigl[\prod_{j=1}^{n}(1+2m_1c_j)\Bigr]-\prod_{j=1}^{n}\mathbb{E}_\phi[1+2m_1c_j]}^{\text{phase}}} {\prod_{j=1}^{n}\mathbb{E}_\phi[1+2m_1c_j]}\\ +g^{(n)}_N\,\frac{\overbrace{\sum_{k=2}^{n}2^k(m_k-m_1^k)\,\mathbb{E}_\phi[e_k]}^{\text{shape}}} {\prod_{j=1}^{n}\mathbb{E}_\phi[1+2m_1c_j]} \end{multline}

The autocorrelation—with all photons at the same angle—goes with $c_j=c\equiv\cos(\Delta\ell\,\theta-\phi)$, $e_k=\binom nk c^k$ and makes Eq. (\ref{eq:0SCpX}) the ratio of the $n$-th moment of the local intensity $1+2ac$ to the $n$-th power of its mean for $\phi$ random:

$$g^{(n)}(\theta,\dots,\theta)=g^{(n)}_N\sum_{p=0}^{\lfloor n/2\rfloor}\binom n{2p}\binom{2p}{p}\,m_{2p}$$ which for thermal light evaluates to: $$g^{(n)}(\theta,\dots,\theta)=n!$$ i.e., thermal light with its three confounders is therefore locally single-mode thermal light at every order. This is the case while $g^{(n)}_N=(n+1)!/2^n$ is the $n$th-order correlator for bimodal thermal field, and is below $n!$ by the factor $2^n/(n+1)$. For $M$ thermal modes of equal mean, instead, it would be:

$$g^{(n)}_N=\frac{\langle\hat N(\hat N-1)\cdots(\hat N-n+1)\rangle}{\langle\hat N\rangle^n}=\frac{\mathbb{E}[W^n]}{\mathbb{E}[W]^n}=\frac{\Gamma(M+n)}{\Gamma(M)\,M^n}=\prod_{k=0}^{n-1}\Bigl(1+\frac{k}{M}\Bigr)$$

So the autocorrelation $g^{(n)}_N$ when only the intensity fluctuates teams up with $2^n/(n+1)$ the autocorrelation when the intensity does not, so that together they multiply to n!.

With the phase pinned (and intensity and shape kept fluctuating)—i.e., setting the phase term of Eq. \eqref{eq:0SCpX} to zero—the autocorrelation becomes a function of the angle:

$$g^{(n)}(\theta,\dots,\theta)=g^{(n)}_N\,\frac{\mathbb{E}_t[(1+2ac)^n]}{(1+2m_1c)^n}\,,\qquad \text{with }c\equiv\cos(\Delta\ell\,\theta)$$

which is the ratio of the $n$th moment of the local intensity of the collapses at that angle, to the $n$-th power of the mean collapse.

At the lobes $c=1$ and waists $c=-1$, we find: $$g^{(n)}(\theta,\dots,\theta)\Big|_{c=\pm1} =g^{(n)}_N\,\frac{A_n\pm\pi B_n}{(1\pm\pi/4)^n}\,,$$ where $$ A_n\equiv\sum_{p=0}^{\lfloor n/2\rfloor}\binom{n}{2p}\frac{4^p\,(p!)^2}{(2p+1)!}\,,\qquad B_n\equiv\sum_{p=0}^{\lfloor (n-1)/2\rfloor}\binom{n}{2p+1}\frac{1}{2^{2p+3}}\binom{2p+2}{p+1}\,. $$

  • At the in-betweener points, $c=0$ ($\theta=\pi/4$ for $\Delta\ell=2$): the local intensity is $W$ and the value is exactly $g^{(n)}_N$, as if the shape were frozen.
  • At the lobes, $c=1$: the local intensity runs from $W$ to $2W$ with the contrast, a mild spread, so the value sits just above $g^{(n)}_N$: 1.52, 3.13, 8.1, 25.6, 94.6 for $n=2$ to $6$, against 1.5, 3, 7.5, 22.5, 78.75.
  • At the waist, $c=-1$: the local intensity is $W(1-2a)$, zero for a balanced collapse and $W$ for a one-mode collapse, with mean $0.215\,W$. The rare low-contrast frames, which stay bright where the mean collapse is nearly dark, dominate the $n$-th moment, and the value explodes: 3.12, 16.6, 126, 1235, 14945, etc.

So pinning the phase trades the angle-independent $n!$ of random-phase thermal light for a profile locked to the lobes: thermal-like at the quarter points, barely more at the lobes, super-bunched at the waists. Pinning the phase and freezing the shape too leaves the flat $g^{(n)}_N$.

The share of each confounding factor increases with $n$ away from intensity towards shape and, mainly, phase:

\begin{align} g^{(2)}(\theta,\theta)-1&=1=0.500+0.463+0.037\,,\\ g^{(3)}(\theta,\theta,\theta)-1&=5=2.000+2.776+0.224\,,\\ g^{(4)}(\theta,\dots,\theta)-1&=23=6.500+14.949+1.551\,,\label{eq:0Sc1u}\\ g^{(5)}(\theta,\dots,\theta)-1&=119=21.500+85.448+12.052\,,\label{eq:0ShpF}\\ g^{(6)}(\theta,\dots,\theta)-1&=719=\underbrace{77.750}_{\text{intensity}}+\underbrace{538.655}_{\text{phase}}+\underbrace{102.595}_{\text{shape}}\,. \end{align}

In closed form the intensity term is $(n+1)!/2^n − 1$, the phase term is $g^{(n)}_N\sum_{p\ge1}\binom{n}{2p}\binom{2p}{p}\bigl(\frac{\pi}{8}\bigr)^{2p}$ (a polynomial in $\pi^2$), and the shape term takes the rest to get to $n!-1$. E.g., for the first two orders:

$$g^{(2)}-1=\tfrac12+\tfrac{3\pi^2}{64}+\Bigl(\tfrac12-\tfrac{3\pi^2}{64}\Bigr),\qquad g^{(3)}-1=2+\tfrac{9\pi^2}{32}+\Bigl(3-\tfrac{9\pi^2}{32}\Bigr), etc.$$

More breakdown:

The $n$-photon autocorrelation $g^{(n)}(\theta,\dots,\theta)$ of two-mode thermal light ($\Delta\ell=2$).
$n$ 2 3 4 5 6
$g^{(n)}_N=(n+1)!/2^n$ (ring value, intensity only) $3/2$ $3$ $15/2$ $45/2$ $315/4$
$g^{(n)}(\theta,\dots,\theta)$, all confounders, $\phi$ random: $n!$ $2$ $6$ $24$ $120$ $720$
intensity term $g^{(n)}_N-1$ $0.500$ $2.000$ $6.500$ $21.50$ $77.75$
phase term $0.463$ $2.776$ $14.95$ $85.45$ $538.7$
shape term $0.037$ $0.224$ $1.551$ $12.05$ $102.6$
intensity removed, $\phi$ random: $2^n/(n+1)$ $4/3$ $2$ $16/5$ $16/3$ $64/7$
shape fixed at the mean collapse, $\phi$ random $1.963$ $5.776$ $22.45$ $107.9$ $617.4$
shape fixed at the atom, $\phi$ random $2$ $6$ $23.75$ $116.3$ $676.7$
shape fixed at the dipole, $\phi$ random: $g^{(n)}_N\binom{2n}{n}/2^n$ $9/4$ $15/2$ $32.81$ $177.2$ $1137$
$\phi$ pinned, antinode $c=1$ $1.523$ $3.134$ $8.142$ $25.60$ $94.57$
$\phi$ pinned, node $c=-1$ $3.122$ $16.62$ $125.7$ $1235$ $14946$

The rows with the shape fixed correspond to $a_0=\pi/8$, $1/\sqrt6$, $\tfrac12$ with $m_{2p}\to a_0^{2p}$. Also on â„€ ⟶

4 November (2026)

The intensity and shape confounding factors for the unbalanced case do not fluctuate independently, which is why we had the $Q$ instead of $P$ for them.[12]

As a result of those correlations, the phase-pinned unbalanced thermal case can be more bunched (at the waist) than the balanced one:

Namely, its $g^{(2)}_\mathrm{max}\approx 3.32$ as opposed to 1.12 for the balanced case.

Maybe not the most important discovery of today but worth a mention here on â„€. Also on â„€ ⟶


This is the central theoretical result of AHBT: \begin{equation} \label{eq:0SxL1} g^{(2)}(\theta_1,\theta_2)=1+\underbrace{(g^{(2)}_N-1)}_{\text{intensity}} +g^{(2)}_N\,\frac{\overbrace{4m_1^2\operatorname{Cov}_\phi(c_1,c_2)}^{\text{phase}} +\overbrace{4\operatorname{Var}a\;\mathbb{E}_\phi[c_1c_2]}^{\text{shape}}} {\mathbb{E}_\phi[1+2m_1c_1]\,\mathbb{E}_\phi[1+2m_1c_2]}\,, \end{equation} with definitions of all the quantities in â„€ below. What this shows is the detailed contribution of each confounding factor to the photon correlations. For thermal vortices, there are three: intensity (I), phase (P) and shape (S). Remove one of those confounding factors, and the correlations adapt accordingly. This is a layout of all the possibilities:

The phase makes it rotationally invariant, and spread them so that a single relative $g^{(2)}(\Delta\theta)$ is enough. If you pin the phase, a new structure arises, and the other confounding factors make it exhibit correlations, some of which are well known ($g^{(2)}=3/2$ for I [panel (d)], i.e., for two modes), others are our own contribution to the field (the Ergoli anti/super-bunching $g^{(2)}=4\times(10\pm3\pi)/(4\pm\pi)^2$ [panel (c)]). If you remove all confounding factors, of course you get no correlation [panel (h)].

The notable thing here, which I didn't fully appreciate before, is that with only intensity and phase (IP), the correlations you get is not the textbook $g^{(2)}=2$ but anything between $1.5$ and $2.25$, depending on which shape you fix (former is the donut, so bringing no correlations and recovering 1.5 of single-mode thermal light only, the latter is the dipole, exceeding the RPCS's 1.5 because it also has intensity fluctuations). Indeed, in such a case, Eq. \eqref{eq:0SxL1} reduces to

$$g^{(2)}(\Delta\theta)=g^{(2)}_N\big[1+2a_0^2\cos(\Delta\ell\,\Delta\theta)\big]\,,$$

with $0\le a_0\le\tfrac12$ and then we spread this into a donut. So which shape to choose from? RPCS would be nice but we need to tweak Eq. \eqref{eq:0SxL1}. The atom would give us the thermal correlations, so nothing new (misleadingly). By itself, what the formula offers is the mean collapse, i.e., the average pattern of the single shots pinned to the same orientation: fix $\phi$, average the profile $1+2a\cos(\Delta\ell\,\theta-\phi)$ over the shapes that occur, and since only $a$ varies: \begin{align} \label{eq:0S0Rl} \mathbb{E}_t\big[1+2a\cos(\Delta\ell\,\theta-\phi)\big]&=1+2m_1\cos(\Delta\ell\,\theta-\phi)\,,\\ m_1=\mathbb{E}_t[a]&=\int_0^1\sqrt{t(1-t)}\,dt=\frac{\pi}{8}\,, \end{align} but this is very close to the thermal atom anyway:

while different enough to make the point. So we'll go with this, it's the best possible option (simplest decomposition, not thermal & with some justification). The only drawback is that it is a bit contrived, fixing the shape fluctuations doesn't obviously mean fixing them to that of the mean. The dipole would be more illustrative, but it would be forced by hand in a non-natural way.

In the mean-collapse case, $g^{(2)}=\tfrac32+\frac{3\pi^2}{64}=1.96264$ and same with $-$ for the minimum correlations, which brings them to $1.03736$ instead of 1. Very small deviations but enough to put a finger on them. See the definitions on â„€ ⟶


I wonder if I shouldn't welcome Jack and Sally in my list of limerent couples... There is the most beautiful exchange, which I kept for my Halloween dialogue between the pumpkin and the moon. But there are others quite telling:

Jack: Oh, there's an empty place in my bones. That calls out for something unknown.
Sally: I sense there's something in the wind. That feels like tragedy's at hand.

This one was particularly infinite:

Jack: And for a moment, why, I even touched the sky!
Sally: Try as I may, it doesn't last.

But it is an odd case of crossed-limerence, and it ends up working out well for them—«And sit together, now and forever. For it is plain as anyone can see. We're simply meant to be.»—which maybe washes out the limerence with love, and repairs the wounds with triteness, denies fate with a miracle. I'll probably not include them. I should watch the movie again. Some future Halloween or some future Christmas.

3 November (2026)

You'll never know the good things of me and I'll never know the bad things of you, and I thought this was terribly sad...


We're also back on db, which went staled since 29 July (2026), where I was already in the CSI business. And then I should also get the seminal, founding paper that was tossed by this Meera Parish, finally be published somewhere.


The AHBT as function of $\Delta\ell$:

Also on â„€ ⟶

2 November (2026)

We left Alex at the airport... it was a short but intense program: Spanish food and drinks (Paella de Bogavante, paletilla asada, my brandade de morue...), the nature of time and literature (Vonnegut's «If this isn't nice, then I don't know what is.»), Wittgenstein's lion opinion of women, cultural appropriations, metrosexuality, China and Copenhagen, geopolitics and interpretations... We also submitted Ergoli, and we laid down the grounds for two further immediate collaborations, of which we are sending drafts from us three to us all, so that he's got something to welcome him and keep him busy when he lands in Hangzhou.

Daniel simulated this beautiful violation of inequalities based on their experimental setup (line is theory, pre Anno Fabri!):

Tipping your head into what's forbidden time to time, just to see... The data are Monte Carlo points cross-correlating angles and checking whether what happens between fluctuations is possible or not, according to classical physics. I hope to show the same figure in December with experimental points.


Soto sunset.


Last hours of the collaboration.


In between discussions... more discussions.

Or explaining how I plan to make this place a mixture between the international center of multiphotonics and a private club where the unlikely meets the impossible.


Just found this: On the application of a monolithic array for detecting intensity-correlated photons emitted by different source types. D. L. Boiko, N. J. Gunther, N. Brauer, M. Sergio, C. Niclass, G. B. Beretta and E. Charbon in Opt. Express 17:15087 (2009). Which I need to gloss!

1 November (2026)

Showing my mountains to Alexandros.


Above the trees.


I need to study that business of local-density of states,[14] as well as Refs. [[15],[16],[17],[18]] among others. Also on â„€ ⟶


Beautiful people.


Spanish breakfast, post Ergoli-submission. Chocolate is coming and so is the editor decision. Sweet things.


My Fig. (0RiBM) looks like Karabekian's "The Temptation of Saint Anthony," the «unwavering band of light». After scorching me, I feel like the Universe is speaking to me again...

My eyes are crying, Oh no, my eyes are crying...
–:––

I'm not feeling free, Oh no.
Oh, I'm falling. Oh, I'm crying. Oh, I'm yelling.
Oh, I love it.

It'll never stop, of course. As a high-end primate of the sapiens sapiens species, evolution has wired limerence into all the tips of all my intermingled neurons, and they always flare up the same intuition of the absolute. It's such a fascinating mental process. I'm sure spiritual aspirations feed on the same shortcuts in the same circuitry.

uijin will complete today and I'll be launching the second wave—nijin (äșŒé™Ł)—for another point of the Fig. (0NNbp) series, to get an idea of how the structure morphs through the parameter space. This'll take most of next week. I expect ultimately some nice animation of something like Fig. (0OG5J) popping out bubbles as it spread over the lasing plateau. Quite a sight!


Validating pdfs on the editor server in the earliest hours of the day; my kind of Halloween.

  1. ↑ Violation of Cauchy-Schwarz Inequalities in Space and Time. D. E. Salazar and F. P. Laussy in Unpublished (2026).
  2. ↑ Luminescence spectra of quantum dots in microcavities. II. Fermions. E. del Valle, F. P. Laussy and C. Tejedor in Phys. Rev. B 79:235326 (2009). 
  3. ↑ Synchronization along quantum trajectories. N. Es'haqi-Sani, G. Manzano, R. Zambrini and R. Fazio in Phys. Rev. Res. 2:023101 (2020).
  4. ↑ Entanglement and lasing with two quantum dots in a microcavity. E. del Valle, F. P. Laussy, F. Troiani and C. Tejedor in Phys. Rev. B 76:235317 (2007).
  5. ↑ IV. A letter from Mr. Colin Mac Laurin, Professor of Mathematicks at Edinburgh, and F. R. S. to martin folkes, Esq; V. Pr. R. S. concerning équations with impossible roots. C. MacLaurin in Philosophical Transactions of the Royal Society of London 34:104 (1726).
  6. ↑ 6.0 6.1 Apodictic discourse and the Cauchy-Bunyakovsky-Schwarz inequality. S. Kichenassamy in Ganita Bharati 42:129 (2020).
  7. ↑ IV. A second letter from Mr. Colin McLaurin, Professor of Mathematicks in the University of Edinburgh and F. R. S. to Martin Folkes, Esq; concerning the roots of equations, with the demonstration of other rules in algebra; being the continuation of the letter published in the Philosophical Transactions, N° 394. C. MacLaurin in Philosophical Transactions of the Royal Society of London 36:59 (1729).
  8. ↑ II. A method for determining the number of impossible roots in adfected équations. G. Campbell in Philosophical Transactions of the Royal Society of London 35:515 (1728).
  9. ↑ Oscar II's prize competition and the error in PoincarĂ©'s memoir on the three body problem. J. Barrow-Green in Arch. Hist. Exact Sci. 48:107 (1994).
  10. ↑ Observation of Bose-Einstein condensation in a dilute atomic vapor. M. H. Anderson, J. R. Ensher, M. R. Matthews, C. E. Wieman and E. A. Cornell in Science 269:198 (1995).
  11. ↑ Spatial correlations of vortex quantum states. Eduardo Zubizarreta Casalengua and Fabrice P. Laussy in arXiv:2402.01627 (2024).
  12. ↑ 12.0 12.1 Bosonic correlations are spurious for classical states. D. E. Salazar and F. P. Laussy in Phys. Rev. Lett. [1] (2026).
  13. ↑ Template:Shapley
  14. ↑ Analogs of spontaneous emission and lasing in photonic time crystals. K. Lee, M. Kyung, Y. Kim, J. Park, H. Lee, J. Choi, C. T. Chan, J. Shin, K. W. Kim and B. Min in Phys. Rev. Lett. 136:093802 (2026).
  15. ↑ Nondissipative non-Hermitian dynamics and exceptional points in coupled optical parametric oscillators. A. Roy, S. Jahani, Q. Guo, A. Dutt, S. Fan, M.-A. Miri and A. Marandi in Optica 8:415 (2021).
  16. ↑ Exceptional-point-assisted entanglement, squeezing, and reset in a chain of three superconducting resonators. W. S. Teixeira, V. Vadimov, T. Mörstedt, S. Kundu and M. Möttönen in Phys. Rev. Res. 5:033119 (2023).
  17. ↑ Parametrically driving a quantum oscillator into exceptionality. C. A. Downing and A. Vidiella-Barranco in Sci. Rep. 13:11004 (2023).
  18. ↑ Nonmonotonic quantum-to-classical transition in multiparticle interference. Y.-S. Ra, M. C. Tichy, H.-T. Lim, O. Kwon, F. Mintert, A. Buchleitner and Y.-H. Kim in Proc. Natl. Acad. Sci. 110:1227 (2013).