Tag: topos

Last time, we’ve viewed major and minor triads (chords) as inscribed triangles in a regular $12$-gon.

If we move clockwise along the $12$-gon, starting from the endpoint of the longest edge (the root of the chord, here the $0$-vertex) the edges skip $3,2$ and $4$ vertices (for a major chord, here on the left the major $0$-chord) or $2,3$ and $4$ vertices (for a minor chord, here on the right the minor $0$-chord).

The symmetries of the $12$-gon, the dihedral group $D_{12}$, act on the $24$ major- and minor-chords transitively, preserving the type for rotations, and interchanging majors with minors for reflections.

Mathematical Music Theoreticians (MaMuTh-ers for short) call this the $T/I$-group, and view the rotations of the $12$-gon as transpositions $T_k : x \mapsto x+k~\text{mod}~12$, and the reflections as involutions $I_k : x \mapsto -x+k~\text{mod}~12$.

Note that the elements of the $T/I$-group act on the vertices of the $12$-gon, from which the action on the chord-triangles follows.

There is another action on the $24$ major and minor chords, mapping a chord-triangle to its image under a reflection in one of its three sides.

Note that in this case the reflection $I_k$ used will depend on the root of the chord, so this action on the chords does not come from an action on the vertices of the $12$-gon.

There are three such operations: (pictures are taken from Alexandre Popoff’s blog, with the ‘funny names’ removed)

The $P$-operation is reflection in the longest side of the chord-triangle. As the longest side is preserved, $P$ interchanges the major and minor chord with the same root.

The $L$-operation is refection in the shortest side. This operation interchanges a major $k$-chord with a minor $k+4~\text{mod}~12$-chord.

Finally, the $R$-operation is reflection in the middle side. This operation interchanges a major $k$-chord with a minor $k+9~\text{mod}~12$-chord.

From this it is already clear that the group generated by $P$, $L$ and $R$ acts transitively on the $24$ major and minor chords, but what is this $PLR$-group?

If we label the major chords by their root-vertex $1,2,\dots,12$ (GAP doesn’t like zeroes), and the corresponding minor chords $13,14,\dots,24$, then these operations give these permutations on the $24$ chords:

 P:=(1,13)(2,14)(3,15)(4,16)(5,17)(6,18)(7,19)(8,20)(9,21)(10,22)(11,23)(12,24) L:=(1,17)(2,18)(3,19)(4,20)(5,21)(6,22)(7,23)(8,24)(9,13)(10,14)(11,15)(12,16) R:=(1,22)(2,23)(3,24)(4,13)(5,14)(6,15)(7,16)(8,17)(9,18)(10,19)(11,20)(12,21) 

Then GAP gives us that the $PLR$-group is again isomorphic to $D_{12}$:

 gap> G:=Group(P,L,R);; gap> Size(G); 24 gap> IsDihedralGroup(G); true 

In fact, if we view both the $T/I$-group and the $PLR$-group as subgroups of the symmetric group $Sym(24)$ via their actions on the $24$ major and minor chords, these groups are each other centralizers! That is, the $T/I$-group and $PLR$-group are dual to each other.

For more on this, there’s a beautiful paper by Alissa Crans, Thomas Fiore and Ramon Satyendra: Musical Actions of Dihedral Groups.

What does this new MaMuTh info learns us more about our Elephant, the Topos of Triads, studied by Thomas Noll?

Last time we’ve seen the eight element triadic monoid $T$ of all affine maps preserving the three tones $\{ 0,4,7 \}$ of the major $0$-chord, computed the subobject classified $\Omega$ of the corresponding topos of presheaves, and determined all its six Grothendieck topologies, among which were these three:

Why did we label these Grothendieck topologies (and corresponding elements of $\Omega$) by $P$, $L$ and $R$?

We’ve seen that the sheafification of the presheaf $\{ 0,4,7 \}$ in the triadic topos under the Grothendieck topology $j_P$ gave us the sheaf $\{ 0,3,4,7 \}$, and these are the tones of the major $0$-chord together with those of the minor $0$-chord, that is the two chords in the $\langle P \rangle$-orbit of the major $0$-chord. The group $\langle P \rangle$ is the cyclic group $C_2$.

For the sheafication with respect to $j_L$ we found the $T$-set $\{ 0,3,4,7,8,11 \}$ which are the tones of the major and minor $0$-,$4$-, and $8$-chords. Again, these are exactly the six chords in the $\langle P,L \rangle$-orbit of the major $0$-chord. The group $\langle P,L \rangle$ is isomorphic to $Sym(3)$.

The $j_R$-topology gave us the $T$-set $\{ 0,1,3,4,6,7,9,10 \}$ which are the tones of the major and minor $0$-,$3$-, $6$-, and $9$-chords, and lo and behold, these are the eight chords in the $\langle P,R \rangle$-orbit of the major $0$-chord. The group $\langle P,R \rangle$ is the dihedral group $D_4$.

More on this can be found in the paper Commuting Groups and the Topos of Triads by Thomas Fiore and Thomas Noll.

The operations $P$, $L$ and $R$ on major and minor chords are reflexions in one side of the chord-triangle, so they preserve two of the three tones. There’s a distinction between the $P$ and $L$ operations and $R$ when it comes to how the third tone changes.

Under $P$ and $L$ the third tone changes by one halftone (because the corresponding sides skip an even number of vertices), whereas under $R$ the third tone changes by two halftones (a full tone), see the pictures above.

The $\langle P,L \rangle = Sym(3)$ subgroup divides the $24$ chords in four orbits of six chords each, three major chords and their corresponding minor chords. These orbits consist of the

• $0$-, $4$-, and $8$-chords (see before)
• $1$-, $5$-, and $9$-chords
• $2$-, $6$-, and $10$-chords
• $3$-, $7$-, and $11$-chords

and we can view each of these orbits as a cycle tracing six of the eight vertices of a cube with one pair of antipodal points removed.

These four ‘almost’ cubes are the NE-, SE-, SW-, and NW-regions of the Cube Dance Graph, from the paper Parsimonious Graphs by Jack Douthett and Peter Steinbach.

To translate the funny names to our numbers, use this dictionary (major chords are given by a capital letter):

The four extra chords (at the N, E, S, and P places) are augmented triads. They correspond to the triads $(0,4,8),~(1,5,9),~(2,6,10)$ and $(3,7,11)$.

That is, two triads are connected by an edge in the Cube Dance graph if they share two tones and differ by an halftone in the third tone.

This graph screams for a group or monoid acting on it. Some of the edges we’ve already identified as the action of $P$ and $L$ on the $24$ major and minor triads. Because the triangle of an augmented triad is equilateral, we see that they are preserved under $P$ and $L$.

But what about the edges connecting the regular triads to the augmented ones? If we view each edge as two directed arrows assigned to the same operation, we cannot do this with a transformation because the operation sends each augmented triad to six regular triads.

Alexandre Popoff, Moreno Andreatta and Andree Ehresmann suggest in their paper Relational poly-Klumpenhouwer networks for transformational and voice-leading analysis that one might use a monoid generated by relations, and they show that there is such a monoid with $40$ elements acting on the Cube Dance graph.

Popoff claims that usual presheaf toposes, that is contravariant functors to $\mathbf{Sets}$ are not enough to study transformational music theory. He suggest to use instead functors to $\mathbf{Rel}$, that is Sets with as the morphisms binary relations, and their compositions.

Another Elephant enters the room…

(to be continued)

These three ideas (re)surfaced over the last two decades, claiming to have potential applications to major open problems:

• (2000) $\mathbb{F}_1$-geometry tries to view $\mathbf{Spec}(\mathbb{Z})$ as a curve over the field with one element, and mimic Weil’s proof of RH for curves over finite fields to prove the Riemann hypothesis.
• (2012) IUTT, for Inter Universal Teichmuller Theory, the machinery behind Mochizuki’s claimed proof of the ABC-conjecture.
• (2014) topos theory : Connes and Consani redirected their RH-attack using arithmetic sites, while Lafforgue advocated the use of Caramello’s bridges for unification, in particular the Langlands programme.

It is difficult to voice an opinion about the (presumed) current state of such projects without being accused of being either a believer or a skeptic, resorting to group-think or being overly critical.

We lack the vocabulary to talk about the different phases a mathematical idea might be in.

Such a vocabulary exists in (information) technology, the five phases of the Gartner hype cycle to represent the maturity, adoption, and social application of a certain technology :

1. Technology Trigger
2. Peak of Inflated Expectations
3. Trough of Disillusionment
4. Slope of Enlightenment
5. Plateau of Productivity

This model can then be used to gauge in which phase several emerging technologies are, and to estimate the time it will take them to reach the stable plateau of productivity. Here’s Gartner’s recent Hype Cycle for emerging Artificial Intelligence technologies.

Picture from Gartner Hype Cycle for AI 2021

What might these phases be in the hype cycle of a mathematical idea?

1. Technology Trigger: a new idea or analogy is dreamed up, marketed to be the new approach to that problem. A small group of enthusiasts embraces the idea, and tries to supply proper definitions and the very first results.
2. Peak of Inflated Expectations: the idea spreads via talks, blogposts, mathoverflow and twitter, and now has enough visibility to justify the first conferences devoted to it. However, all this activity does not result in major breakthroughs and doubt creeps in.
3. Trough of Disillusionment: the project ran out of steam. It becomes clear that existing theories will not lead to a solution of the motivating problem. Attempts by key people to keep the idea alive (by lengthy papers, regular meetings or seminars) no longer attract new people to the field.
4. Slope of Enlightenment: the optimistic scenario. One abandons the original aim, ditches the myriad of theories leading nowhere, regroups and focusses on the better ideas the project delivered.

A negative scenario is equally possible. Apart for a few die-hards the idea is abandoned, and on its way to the graveyard of forgotten ideas.

5. Plateau of Productivity: the polished surviving theory has applications in other branches and becomes a solid tool in mathematics.

It would be fun so see more knowledgable people draw such a hype cycle graph for recent trends in mathematics.

Here’s my own (feeble) attempt to gauge where the three ideas mentioned at the start are in their cycles, and here’s why:

• IUTT: recent work of Kirti Joshi, for example this, and this, and that, draws from IUTT while using conventional language and not making exaggerated claims.
• $\mathbb{F}_1$: the preliminary programme of their seminar shows little evidence the $\mathbb{F}_1$-community learned from the past 20 years.
• Topos: Developing more general theory is not the way ahead, but concrete examples may carry surprises, even though Gabriel’s topos will remain elusive.

Clearly, you don’t agree, and that’s fine. We now have a common terminology, and you can point me to results or events I must have missed, forcing me to redraw my graph.

Brendan Fong, David Spivak and Remy Tuyeras cooked up a vast generalisation of neural networks in their paper Backprop as Functor: A compositional perspective on supervised learning.

Here’s a nice introduction to neural networks for category theorists by Bruno Gavranovic. At 1.49m he tries to explain supervised learning with neural networks in one slide. Learners show up later in the talk.

$\mathbf{Poly}$ is the category of all polynomial functors, that is, things of the form
$p = \sum_{i \in p(1)} y^{p[i]}~:~\mathbf{Sets} \rightarrow \mathbf{Sets} \qquad S \mapsto \bigsqcup_{i \in p(1)} Maps(p[i],S)$
with $p(1)$ and all $p[i]$ sets.

Last time I gave Spivak’s ‘corolla’ picture to think about such functors.

I prefer to view $p \in \mathbf{Poly}$ as an horribly discrete ‘sheaf’ $\mathcal{P}$ over the ‘space’ $p(1)$ with stalk $p[i]=\mathcal{P}_i$ at point $i \in p(1)$.

A morphism $p \rightarrow q$ in $\mathbf{Poly}$ is a map $\varphi_1 : p(1) \rightarrow q(1)$, together with for all $i \in p(1)$ a map $\varphi^{\#}_i : q[\varphi_1(i)] \rightarrow p[i]$.

In the sheaf picture, this gives a map of sheaves over the space $p(1)$ from the inverse image sheaf $\varphi_1^* \mathcal{Q}$ to $\mathcal{P}$.

But, unless you dream of sheaves in the night, by all means stick to Spivak’s corolla picture.

A learner $A \rightarrow B$ between two sets $A$ and $B$ is a complicated tuple of things $(P,I,U,R)$:

• $P$ is a set, a parameter space of some maps from $A$ to $B$.
• $I$ is the interpretation map $I : P \times A \rightarrow B$ describing the maps in $P$.
• $U$ is the update map $U : P \times A \times B \rightarrow P$, the learning procedure. The idea is that $U(p,a,b)$ is a map which sends $a$ closer to $b$ than the map $p$ did.
• $R$ is the request map $R : P \times A \times B \rightarrow A$.

Here’s a nice application of $\mathbf{Poly}$’s set-up:

Morphisms $\mathbf{P y^P \rightarrow Maps(A,B) \times Maps(A \times B,A) y^{A \times B}}$ in $\mathbf{Poly}$ coincide with learners $\mathbf{A \rightarrow B}$ with parameter space $\mathbf{P}$.

This follows from unpacking the definition of morphism in $\mathbf{Poly}$ and the process CT-ers prefer to call Currying.

The space-map $\varphi_1 : P \rightarrow Maps(A,B) \times Maps(A \times B,A)$ gives us the interpretation and request-map, whereas the sheaf-map $\varphi^{\#}$ gives us the more mysterious update-map $P \times A \times B \rightarrow P$.

$\mathbf{Learn(A,B)}$ is the category with objects all the learners $A \rightarrow B$ (for all paramater-sets $P$), and with morphisms defined naturally, that is, maps between the parameter-sets, compatible with the structural maps.

A surprising result from David Spivak’s paper Learners’ Languages is

$\mathbf{Learn(A,B)}$ is a topos. In fact, it is the topos of all set-valued representations of a (huge) directed graph $\mathbf{G_{AB}}$.

This will take some time.

Let’s bring some dynamics in. Take any polynmial functor $p \in \mathbf{Poly}$ and fix a morphism in $\mathbf{Poly}$
$\varphi = (\varphi_1,\varphi[-])~:~p(1) y^{p(1)} \rightarrow p$
with space-map $\varphi_1$ the identity map.

We form a directed graph:

• the vertices are the elements of $p(1)$,
• vertex $i \in p(1)$ is the source vertex of exactly one arrow for every $a \in p[i]$,
• the target vertex of that arrow is the vertex $\phi[i](a) \in p(1)$.

Here’s one possibility from Spivak’s paper for $p = 2y^2 + 1$, with the coefficient $2$-set $\{ \text{green dot, yellow dot} \}$, and with $1$ the singleton $\{ \text{red dot} \}$.

Start at one vertex and move after a minute along a directed edge to the next (possibly the same) vertex. The potential evolutions in time will then form a tree, with each node given a label in $p(1)$.

If we start at the green dot, we get this tree of potential time-evolutions

There are exactly $\# p[i]$ branches leaving a node labeled $i \in p(1)$, and all subtrees emanating from equal labelled nodes are isomorphic.

If we had started at the yellow dot we had obtained a labelled tree isomorphic to the subtree emanating here from any yellow dot.

We can do the same things for any morphism in $\mathbf{Poly}$ of the form
$\varphi = (\varphi_1,\varphi[-])~:~Sy^S \rightarrow p$
Now, we have a directed graph with vertices the elements $s \in S$, with as many edges leaving vertex $s$ as there are elements $a \in p[\varphi_1(s)]$, and with the target vertex of the edge labeled $a$ starting in $s$ the vertex $\varphi[\varphi_1(s)](A)$.

Once we have this directed graph on $\# S$ vertices we can label vertex $s$ with the label $\varphi_1(s)$ from $p(1)$.

In this way, the time evolutions starting at a vertex $s \in S$ will give us a $p(1)$-labelled rooted tree.

But now, it is possibly that two distinct vertices can have the same $p(1)$-labeled tree of evolutions. But also, trees corresponding to equal labeled vertices can be different.

Right, I guess we’re ready to define the graph $G_{AB}$ and prove that $\mathbf{Learn(A,B)}$ is a topos.

In the case of learners, we have the target polynomial functor $p=C y^{A \times B}$ with $C = Maps(A,B) \times Maps(A \times B,A)$, that is
$p(1) = C \quad \text{and all} \quad p[i]=A \times B$

Start with the free rooted tree $T$ having exactly $\# A \times B$ branches growing from each node.

Here’s the directed graph $G_{AB}$:

• vertices $v_{\chi}$ correspond to the different $C$-labelings of $T$, one $C$-labeled rooted tree $T_{\chi}$ for every map $\chi : vtx(T) \rightarrow C$,
• arrows $v_{\chi} \rightarrow v_{\omega}$ if and only if $T_{\omega}$ is the rooted $C$-labelled tree isomorphic to the subtree of $T_{\chi}$ rooted at one step from the root.

A learner $\mathbf{A \rightarrow B}$ gives a set-valued representation of $\mathbf{G_{AB}}$.

We saw that a learner $A \rightarrow B$ is the same thing as a morphism in $\mathbf{Poly}$
$\varphi = (\varphi_1,\varphi[-])~:~P y^P \rightarrow C y^{A \times B}$
with $P$ the parameter set of maps.

Here’s what we have to do:

1. Draw the directed graph on vertices $p \in P$ giving the dynamics of the morphism $\varphi$. This graph describes how the learner can cycle through the parameter-set.

2. Use the map $\varphi_1$ to label the vertices with elements from $C$.

3. For each vertex draw the rooted $C$-labeled tree of potential time-evolutions starting in that vertex.

In this example the time-evolutions of the two green vertices are the same, but in general they can be different.

4. Find the vertices in $G_{AB}$ determined by these $C$-labeled trees and note that they span a full subgraph of $G_{AB}$.

5. The vertex-set $P_v$ consists of all elements from $p$ whose ($C$-labeled) vertex has evolution-tree $T_v$. If $v \rightarrow w$ is a directed edge in $G_{AB}$ corresponding to an element $(a,b) \in A \times B$, then the map on the vertex-sets corresponding to this edge is
$f_{v,(a,b)}~:~P_v \rightarrow P_w \qquad p \mapsto \varphi[\varphi_1(p)](a,b)$

A set-valued representation of $\mathbf{G_{AB}}$ gives a learner $\mathbf{A \rightarrow B}$.

1. Take a set-valued representation of $G_{AB}$, that is, the finite or infinite collection of vertices $V$ in $G_{AB}$ where the vertex-set $P_v$ is non-empty. Note that these vertices span a full subgraph of $G_{AB}$.

And, for each directed arrow $v \rightarrow w$ in this subgraph, labeled by an element $(a,b) \in A \times B$ we have a map
$f_{v,(a,b)}~:~P_v \rightarrow P_w$

2. The parameter set of our learner will be $P = \sqcup_v P_v$, the disjoint union of the non-empty vertex-sets.

3. The space-map $\varphi_1 : P \rightarrow C$ will send an element in $P_v$ to the $C$-label of the root of the tree $T_v$. This gives us already the interpretation and request maps
$I : P \times A \rightarrow B \quad \text{and} \quad R : P \times A \times B \rightarrow A$

4. The update map $U : P \times A \times B \rightarrow P$ follows from the sheaf-map we can define stalk-wise
$\varphi[\varphi_1(p)](a,b) = f_{v,(a,b)}(p)$
if $p \in P_v$.

That’s all folks!

$\mathbf{Learn(A,B)}$ is equivalent to the (covariant) functors $\mathbf{G_{AB} \rightarrow Sets}$.

Changing the directions of all arrows in $G_{AB}$ any covariant functor $\mathbf{G_{AB} \rightarrow Sets}$ becomes a contravariant functor $\mathbf{G_{AB}^o \rightarrow Sets}$, making $\mathbf{Learn(A,B)}$ an honest to Groth topos!

Every topos comes with its own logic, so we have a ‘learners’ logic’. (to be continued)

Some months ago, Peter Scholze wrote a guest post on the Xena-blog: Liquid tensor experiment, proposing a challenge to formalise the proof of one of his results with Dustin Clausen on condensed mathematics.

Scholze and Clausen ran a masterclass in Copenhagen on condensed mathematics, which you can binge watch on YouTube starting here

Scholze also gave two courses on the material in Bonn of which the notes are available here and here.

Condensed mathematics claims that topological spaces are the wrong definition, and that one should replace them with the slightly different notion of condensed sets.

So, let’s find out what a condensed set is.

Definition: Condensed sets are sheaves (of sets) on the pro-étale site of a point.

(there’s no danger we’ll have to rewrite our undergraduate topology courses just yet…)

In his blogpost, Scholze motivates this paradigm shift by observing that the category of topological Abelian groups is not Abelian (if you put a finer topology on the same group then the identity map is not an isomorphism but doesn’t have a kernel nor cokernel) whereas the category of condensed Abelian groups is.

It was another Clausen-Scholze result in the blogpost that caught my eye.

But first, for something completely different.

In “Musical creativity”, Guerino Mazzola and co-authors introduce a seven steps path to creativity.

Here they are:

1. Exhibiting the open question
2. Identifying the semiotic context
3. Finding the question’s critical sign
4. Identifying the concept’s walls
5. Opening the walls
6. Displaying extended wall perspectives
7. Evaluating the extended walls

Looks like a recipe from distant flower-power pot-infused times, no?

In Towards a Categorical Theory of Creativity for Music, Discourse, and Cognition, Mazzola, Andrée Ehresmann and co-authors relate these seven steps to the Yoneda lemma.

1. Exhibiting the open question = to understand the object $A$
2. Identifying the semiotic context = to describe the category $\mathbf{C}$ of which $A$ is an object
3. Finding the question’s critical sign = $A$ (?!)
4. Identifying the concept’s walls = the uncontrolled behaviour of the Yoneda functor
$@A~:~\mathbf{C} \rightarrow \mathbf{Sets} \qquad C \mapsto Hom_{\mathbf{C}}(C,A)$
5. Opening the walls = finding an objectively creative subcategory $\mathbf{A}$ of $\mathbf{C}$
6. Displaying extended wall perspectives = calculate the colimit $C$ of a creative diagram
7. Evaluating the extended walls = try to understand $A$ via the isomorphism $C \simeq A$.

(Actually, I first read about these seven categorical steps in another paper which might put a smile on your face: The Yoneda path to the Buddhist monk blend.)

Remains to know what a ‘creative’ subcategory is.

The creative moment comes in here: could we not find a subcategory
$\mathbf{A}$ of $\mathbf{C}$ such that the functor
$Yon|_{\mathbf{A}}~:~\mathbf{C} \rightarrow \mathbf{PSh}(\mathbf{A}) \qquad A \mapsto @A|_{\mathbf{A}}$
is still fully faithful? We call such a subcategory creative, and it is a major task in category theory to find creative categories which are as small as possible.

All the ingredients are here, but I had to read Peter Scholze’s blogpost before the penny dropped.

Let’s try to view condensed sets as the result of a creative process.

1. Exhibiting the open question: you are a topologist and want to understand a particular compact Hausdorff space $X$.
2. Identifying the semiotic context: you are familiar with working in the category $\mathbf{Tops}$ of all topological spaces with continuous maps as morphisms.
3. Finding the question’s critical sign: you want to know what differentiates your space $X$ from all other topological spaces.
4. Identifying the concept’s walls: you can probe your space $X$ with continuous maps from other topological spaces. That is, you can consider the contravariant functor (or presheaf on $\mathbf{Tops}$)
$@X~:~\mathbf{Tops} \rightarrow \mathbf{Sets} \qquad Y \mapsto Cont(Y,X)$
and Yoneda tells you that this functor, up to equivalence, determines the space $X$ upto homeomorphism.
5. Opening the walls: Tychonoff tells you that among all compact Hausdorff spaces there’s a class of pretty weird examples: inverse limits of finite sets (or a bit pompous: the pro-etale site of a point). These limits form a subcategory $\mathbf{ProF}$ of $\mathbf{Tops}$.
6. Displaying extended wall perspectives: for every inverse limit $F \in \mathbf{ProF}$ (for ‘pro-finite sets’) you can look at the set $\mathbf{X}(F)=Cont(F,X)$ of all continuous maps from $F$ to $X$ (that is, all probes of $X$ by $F$) and this functor
$\mathbf{X}=@X|_{\mathbf{ProF}}~:~\mathbf{ProF} \rightarrow \mathbf{Sets} \qquad F \mapsto \mathbf{X}(F)$
is a sheaf on the pre-etale site of a point, that is, $\mathbf{X}$ is the condensed set associated to $X$.
7. Evaluating the extended walls: Clausen and Scholze observe that the assignment $X \mapsto \mathbf{X}$ embeds compact Hausdorff spaces fully faithful into condensed sets, so we can recover $X$ up to homeomorphism as a colimit from the condenset set $\mathbf{X}$. Or, in Mazzola’s terminology: $\mathbf{ProF}$ is a creative subcategory of $\mathbf{(cH)Tops}$ (all compact Hausdorff spaces).

It would be nice if someone would come up with a new notion for me to understand Mazzola’s other opus “The topos of music” (now reprinted as a four volume series).

No kidding, this is the final sentence of Le spectre d’Atacama, the second novel by Alain Connes (written with Danye Chéreau (IRL Mrs. AC) and his former Ph.D. advisor Jacques Dixmier).

The book has a promising start. Armand Lafforet (IRL AC) is summoned by his friend Rodrigo to the Chilean observatory Alma in the Altacama desert. They have observed a mysterious spectrum, and need his advice.

Armand drops everything and on the flight he lectures the lady sitting next to him on proofs by induction (breaking up chocolate bars), and recalls a recent stay at the La Trappe Abbey, where he had an encounter with (the ghost of) Alexander Grothendieck, who urged him to ‘Follow the motif!’.

“Comment était-il arrivé là? Il possédait surement quelques clés. Pourquoi pas celles des songes?” (How did he get
there? Surely he owned some keys, why not those of our dreams?)

A few pages further there’s this on the notion of topos (my attempt to translate):

“The notion of space plays a central role in mathematics. Traditionally we represent it as a set of points, together with a notion of neighborhood that we call a ‘topology’. The universe of these new spaces, ‘toposes’, unveiled by Grothendieck, is marvellous, not only for the infinite wealth of examples (it contains, apart from the ordinary topological spaces, also numerous instances of a more combinatorial nature) but because of the totally original way to perceive space: instead of appearing on the main stage from the start, it hides backstage and manifests itself as a ‘deus ex machina’, introducing a variability in the theory of sets.”

So far, so good.

We have a mystery, tidbits of mathematics, and allusions left there to put a smile on any Grothendieck-aficionado’s face.

But then, upon arrival, the story drops dead.

Rodrigo has been taken to hospital, and will remain incommunicado until well in the final quarter of the book.

As the remaining astronomers show little interest in Alain’s (sorry, Armand’s) first lecture, he decides to skip the second, and departs on a hike to the ocean. There, he takes a genuine sailing ship in true Jules Verne style to the lighthouse at he end of the world.

All this drags on for at least half a year in time, and two thirds of the book’s length. We are left in complete suspense when it comes to the mysterious Atacama spectrum.

Perhaps the three authors deliberately want to break with existing conventions of story telling?

I had a similar feeling when reading their first novel Le Theatre Quantique. Here they spend some effort to flesh out their heroine, Charlotte, in the first part of the book. But then, all of a sudden, their main character is replaced by a detective, and next by a computer.

Anyway, when Armand finally reappears at the IHES the story picks up pace.

The trio (Armand, his would-be-lover Charlotte, and Ali Ravi, Cern’s computer guru) convince CERN to sell its main computer to an American billionaire with the (fake) promise of developing a quantum computer. Incidentally, they somehow manage to do this using Charlotte’s history with that computer (for this, you have to read ‘Le Theatre Quantique’).

By their quantum-computing power (Shor and quantum-encryption pass the revue) they are able to decipher the Atacame spectrum (something to do with primes and zeroes of the zeta function), send coded messages using quantum entanglement, end up in the Oval Office and convince the president to send a message to the ‘Riemann sphere’ (another fun pun), and so on, and on.

The book ends with a twist of the classic tale of the mathematician willing to sell his soul to the devil for a (dis)proof of the Riemann hypothesis:

After spending some time in purgatory, the mathematician gets a meeting with God and asks her the question “Is the Riemann hypothesis true?”.

“Of course”, God says.

“But how can you know that all non-trivial zeroes of the zeta function have real part 1/2?”, Armand asks.

And God replies:

“Simple enough, I can see them all at once. But then, don’t forget I’m God. I can see the disappointment in your face, yes I can read in your heart that you are frustrated, that you desire an explanation…

Well, we’re going to fix this. I will call archangel Gabriel, the angel of geometry, he will make you a topos!”

If you feel like running to the nearest Kindle store to buy “Le spectre d’Atacama”, make sure to opt for a package deal. It is impossible to make heads or tails of the story without reading “Le theatre quantique” first.

But then, there are worse ways to spend an idle week than by binge reading Connes…

Edit (February 28th). A short video of Alain Connes explaining ‘Le spectre d’Atacama’ (in French)