Auslander–Reiten conjecture

Start with two states, 0 and 1, added modulo two: two copies of 1 add to 0. A permitted operation sends both states to 0. More generally, can an additive object pass every higher test for joining it with itself and with its coordinate building blocks, yet fail to be an independent piece of a free coordinate object?

The action rules come from an algebra, and the objects it acts on are called modules. The tests are Ext groups; the independent-piece condition is called projective. The conjecture asks for this implication over every Artin algebra and every finitely generated left module. The small two-state object illustrates the objects and does not claim to pass the tests.[1]

Can all higher tests for joining an algebraic object with itself and with its algebra find no nonzero extension class, while the object still cannot be an independent piece of a free coordinate object? The target asks this for every Artin algebra and every finitely generated left module.

A module is an additive group with a compatible left action of the algebra. A free module is a list of independent algebra coordinates; projective means being a direct summand of one, with another module as its complement. An Artin algebra is finite as a module over a commutative Artinian base ring; finite-dimensional algebras over fields are important examples, not the whole scope.

A small example shows what the first extension test detects. Take $A=\mathbb F_2[\varepsilon]/(\varepsilon^2)$, the four elements $0,1,\varepsilon,1+\varepsilon$, and $M=A/(\varepsilon)$, the two-element module on which $\varepsilon$ acts as zero. The sequence $0\to M\to A\to M\to0$ does not split. A putative compatible lift of 1 back to $A$ would be $1$ or $1+\varepsilon$; multiplying either by $\varepsilon$ gives $\varepsilon$, contradicting the zero action on $M$. Thus this nonprojective example has a nonzero self-extension and does not satisfy the conjecture’s hypothesis.

Ext groups record extension classes; in degree one, zero says every short extension of the specified modules splits. Higher degrees concern longer chains, not a claim that all exact sequences are direct sums. The required condition is $\operatorname{Ext}_A^i(M,M\oplus A)=0$ for every $i\ge1$, with $M\oplus A$ the direct sum. Does it force projectivity?[1]

Certain algebras have an extra shared-building-block property, called Igusa–Todorov: after a fixed number of stages in resolving a module’s relations, the remaining relations can be presented using modules from one fixed collection. The September preprint claims the implication for that family. It does not establish the implication for all Artin algebras. The missing result must show that passing both test families in every positive degree forces projectivity, without that extra condition on the algebra.[1]

The general implication to projectivity is unproved.[1]

Modules describe compatible actions and the relations they impose. Projective modules let compatible maps be lifted through surjections, rather than having the relations obstruct a lift. The problem connects that structural flexibility with information from two particular families of homological tests, linking how an object is built with how it can fit into larger objects.[1]

A projective module allows a map into a quotient to be lifted through the quotient’s surjection. This is a precise form of freedom from relations obstructing the lift, and explains why projective modules serve as building blocks when resolving more complicated modules. Extension groups measure failures of such splitting and lifting.

The finite example demonstrates one detected failure, but the general question demands more than a first-degree check. It asks whether self-extension and algebra-extension data in every degree detect the same projective structure as tests against all modules. The cited preprint relates this question to broader conjectures about projective resolutions; implications valid for all algebras cannot automatically be used as implications for one fixed algebra.[1]