Mancosu - Philosophy of Mathematical Practice (Oxford, 2008).pdf

394 colin mclartyThe proof shows this condition is also sufficient. And the function a is uniquelydetermined at all points (p) where p divides either one of m 1 and m 2 .Butthoseare the prime factors of LCM(m 1 , m 2 ) so the unique determinacy says any twosolutions a, b havea ≡ b (mod LCM(m 1 , m 2 )The Chinese remainder theorem patches part of one function r 1 withpart of another r 2 to get a single function a on Spec(Z) just as you canpatch parts of continuous functions f 1 , f 2 into a single continuous functiona on the real line R. But the Chinese remainder theorem only deals withparts given by finite numbers of primes. It follows from a more generalresult with a very similar proof: for every ring R you can patch compatiblepartial functions on the affine scheme X = Spec(R). In technical terms: everystructure sheaf O X is actually a sheaf, so it agrees with its own 0-dimensionalČech cohomology.²¹The point is simple. Many arithmetic problems are easily solved ‘locally’in some sense, while we want ‘global’ solutions. In the Chinese remaindertheorem the list of congruences is easily solved ‘locally at each prime’ and thetrick is to patch together one solution on the whole line.²² Cohomology is allabout patching.14.5 The philosophy of structure14.5.1 Understanding sheaves and schemesIntuitive ideas available in some way to some experts often become explicitand publicly available by passing to higher levels of structure, and they oftengrow further in the passage. Section 14.2.2 gave homology group morphismsand homology theories as two examples. For another, there was a seriousproblem in the early 1950s of how to understand sheaves and work with them.Several nuts-and-bolts definitions were known but none were as simple as theideas seemed to be. And the specialists seeking a Weil cohomology to provethe Weil conjectures saw that none of these definitions was vaguely suitablefor that. One thing was clear. People who worked with sheaves would drawvast commutative diagrams ‘full of arrows covering the whole blackboard’,dumbfounding the student Grothendieck (1985–87, p.19). The vertices wereAbelian sheaves and the arrows were sheaf morphisms between them. One²¹ Compare Hartshorne (1977, Proposition II.2.2) and Tamme (1994, p.41).²² Eisenbud (1995, Exercise2.6) gives a basically homological proof.