LONDON MATHEMATICAL SOCIETY LECTURE NOTE SERIES Managing Editor: Professor M. Reid, Mathematics Institute, University of Warwick, Coventry CV4 TAL, United Kingdom “The titles below are available from booksellers, or from Cambridge University Press at www.cambridge.org/mathematics 224 225 226 207 228 29 20 21 232 233, 234 235, 236 237 238 240 241 242 243, 244 245 246 287 248. 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 268 265 267 268 269 mm m 3 274 275 276 2 278 279 280 281 282 285 286 301 ‘Computability, enumerability, unsolvability, S.B. COOPER, T. A. SLAMAN & S. S. WAINER (eds) ‘A mathematical introduction to string theory, S. ALBEVERIO et al. Novikov conjectutes, index theorems and rigidity I, S.C. FERRY, A. RANICKI & J. ROSENBERG (eds) Novikov conjectures, index theorems and rigidity Il, S.C. FERRY, A. RANICKI & J. ROSENBERG (eds) Ergodic theory of Zé actions, M. POLLICOTT & K. SCHMIDT (eds) Ergodicity for infinite dimensional systems, G. DA PRATO & J. ZABCZYK Prolegomena to a middlebrow arithmetic of curves of genus 2, J. W. S, CASSELS & E. V. FLYNN Semigroup theory and its applications, K. H. HOFMANN & M. W. MISLOVE (eds) ‘The descriptive set theory of Polish group actions, H. BECKER & A. S. KECHRIS Finite fields and applications, S, COHEN & H. NIEDERREITER (eds) Introduction to subfactors, V. JONES & ¥. S. SUNDER Number theory: Séminaire de théorie des nombxes de Paris 1993-94, S. DAVID (ed) ‘The James forest, H. FETTER & B.G. DE BUEN Sieve methods, exponential sums, and their applications in number theory, G. R. H. GREAVES et af Representation theory and algebraic geometry, A. MARTSINKOVSKY & G. TODOROV (eds) Stable groups, FO. WAGNER, Surveys in combinatorics, 1997, R. A. BAILEY (ed.) Geometric Galois actions I, L. SCHNEPS & P. LOCHAK (eds) Geometric Galois actions Il, L. SCHNEPS & P. LOCHAK (eds) Model theory of groups and automorphism groups. D. M. EVANS (ed.) Geometry, combinatorial designs and related structures, J. W. P. HIRSCHFELD er al. ‘p-Automorphisms of finite p-groups, E. I. KHUKHRO ‘Analytic number theory, ¥. MOTOHASHI (ed.) ‘Tame topology and O-minimal structures, L. VAN DEN DRIES ‘The atlas of finite groups: Ten years on, R. CURTIS & R. WILSON (eds) Characters and blocks of finite groups, G, NAVARRO Griibner bases and applications. B, BUCHBERGER & F. WINKLER (eds) . Geometry and cohomology in group theory, P. KROPHOLLER, G. NIBLO, R. STOHR (eds) The q-Schur algebra, S. DONKIN Galois representations in arithmetic algebraic geometry, A. J. SCHOLL & R. L. TAYLOR (eds) ‘Symmetries and integrability of difference equations, P. A. CLARKSON & FW. NUHOFF (eds) Aspects of Galois theory. H. VOLKLEIN ef al, An introduction to noncommutative differential geometry and its physical applications (2nd edition), J. MADORE Sets and proofs, S. B. COOPER & J. TRUSS (eds) Models and computability. S, B. COOPER & J. TRUSS (eds) Groups St Andrews 1997 in Bath, I, C.M. CAMPBELL et al, Groups St Andrews 1997 in Bath, Tl, C.M, CAMPBELL er al. ‘Analysis and logic, C. W. HENSON, J. IOVINO, A. S. KECHRIS & E. ODELL Singularity theory, B. BRUCE & D. MOND (eds) New tends in algebraic geometry, K, HULEK, F. CATANESE, C. PETERS & M. REID (eds) Elliptic curves in cryptography. I. BLAKE, G. SEROUSSI & N. SMART Surveys in combinatorics, 1999, J. D. LAMB & D. A. PREECE (eds) _ ‘Spectral asymptotics in the semi-classical limit, M. DIMASSI & J. SOSTRAND Ergodic theory and topological dynamics, M.B. BEKKA & M. MAYER ‘Singular perturbations of differential operators, S. ALBEVERIO & P. KURASOV Character theory for the odd order theorem, _T. PETERFALVI. Translated by R. SANDLING Spectral theory and geometry, E. B. DAVIES & Y, SAFAROV (eds) ‘The Mandelbrot set, theme and variations, TAN LEI (ed.) Descriptive set theory and dynamical systems, _M. FOREMAN er al. ‘Singularities of plane curves, E. CASAS-ALVERO Computational and geometric aspects of modern algebra, M.D. ATKINSON et al. Global attractors in abstract parabolic problems, J. W, CHOLEWA & T. DLOTKO ‘Topics in symbolic dynamics and applications, F. BLANCHARD. A. MAASS & A. NOGUEIRA (eds) Characters and automorphism groups of compact Riemann surfaces, T. BREUER Explicit birational geometry of 3-folds, A. CORTI & M. REID (eds) Auslander-Buchweitz approximations of equivariant modules, M. HASHIMOTO. Nonlinear elasticity, Y. FU & R. OGDEN (eds) Foundations of computational mathematics, R, DEVORE, A. ISERLES & E. SULI (eds) Rational points on curves over finite fields, | H. NIEDERREITER & C. XING Clifford algebras and spinors (2nd edition), P. LOUNESTO Topics on Riemann surfaces and Fuchsian groups, £. BUJALANCE et al Surveys in combinatorics, 2001, J. HIRSCHFELD (e4.) ‘Aspects of Sobolev-type inequalities, L, SALOFF-COSTE (Quantum groups and Lie theory, A. PRESSLEY (ed) Tits buildings and the model theory of groups. K. TENT (ed) ‘A quantum groups primer, S, MAJID. ‘Second exder partial differential equations in Hilbert spaces, G,DA PRATO & J. ZABCZYK Introduction to operator space theory, G. PISIER Geometry and integrability, L, MASON & ¥. NUTKU (eds) Lectures on invariant theory, 1. DOLGACHEV ‘The homotopy category of simply connected 4-manifolds, H.-J. BAUES Higher operads, higher categories, T. LEINSTER (ed) Kleinian groups and hyperbolic 3-manifolds. Y. KOMORI, V. MARKOVIC & C. SERIES (eds) Introduction to Mabius differential geometry, U. HERTRICH-JEROMIN Stable modules and the D(2)-problem, FE.A. JOHNSON325 326 327 328, 329 330 331 332 333 334 335 336 337 338 339) 34 342, 348 345 347 348 349 350 351 352 353 354 355 356 387 358 359 361 362 363 364 365, 367 368 370 a7 372 323 304 375 376 a7 378 Discrete and continuous nonlinear Schidinger systems, MJ. ABLOWITZ, B. PRINARI & A.D. TRUBATCH Number theory and algebraic geometry, M. REID & A. SKOROBOGATOV (eds) Groups St Andrews 2001 in Oxford I, CM. CAMPBELL, E.F. ROBERTSON & G.C. SMITH (eds) Groups St Andrews 2001 in Oxford Il, _C.M. CAMPBELL, E.F. ROBERTSON & G.C. SMITH (eds) Geometric mechanics and symmetry, J. MONTALDI & T. RATTU (eds) Surveys in combinatorics 2003, C.D. WENSLEY (ed.) ‘Topology, geometry and quantum field theory, U.L. TILLMANN (ed) Corings and comodules, »T. BRZEZINSKI & R. WISBAUER Topics in dynamics and ergodic theory, S. BEZUGLY1 & S. KOLYADA (eds) Groups: topological, combinatorial and arithmetic aspects, T.W. MULLER (ed) Foundations of computational mathematics, Minneapolis 2002, _F. CUCKER et al (eds) ‘Transcendental aspects of algebraic cycles. $. MULLER-STACH & C. PETERS (eds) ‘Spectral generalizations of line graphs, D. CVETKOVIC, P, ROWLINSON & S. SIMIC Structured ring spectra, A. BAKER & B. RICHTER (eds) Linear logic in computer science, T. EHRHARD, P. RUET, JY. GIRARD & P. SCOTT (eds) ‘Advances in elliptic curve cryptography, LF. BLAKE, G. SEROUSSI & N.P. SMART (eds) Perturbation of the boundary in boundary-value problems of partial differential equations, D. HENRY Double affine Hecke algebras, 1. CHEREDNIK LL-functions and Galois representations, D. BURNS, K. BUZZARD & J. NEKOVAR (eds) ‘Surveys in modem mathematics, V. PRASOLOV & Y. [LYASHENKO (eds) Recent perspectives in random matrix theory and number theory, F. MEZZADRI & N.C. SNAITH (eds) Poisson geometry, deformation quantisation and group representations, S. GUT etal (eds) ‘Singularities and computer algebra, _C. LOSSEN & G. PFISTER (eds) Lectures on the Ricci flow, P. TOPPING Modular representations of finite groups of Lie type, J.E. HUMPHREYS. ‘Surveys in combinatorics 2005, B.S. WEBB (ed) Fundamentals of hyperbolic manifolds, R. CANARY. D. EPSTEIN & A. MARDEN (eds) ‘Spaces of Kleinian groups, Y. MINSKY, M. SAKUMA & C. SERIES (eds) ‘Noncommutative localization in algebra and topology, A. RANICKI (ed) Foundations of computational mathematics, Santander 2005, LM. PARDO, A. PINKUS. E. SULI & MJ. TODD (eds) Handbook of tilting theory, L. ANGELERI HUGEL, D. HAPPEL & H. KRAUSE (eds) ‘Synthetic differential geometry (2nd Edition), A. KOCK ‘The Navier-Stokes equations, N. RILEY & P. DRAZIN Lectures on the combinatorics of free probability, A.NICA & R. SPEICHER Integral closure of ideals, rings, and modules, I. SWANSON & C. HUNEKE Methods in Banach space theory, J.M.F. CASTILLO & WB. JOHNSON (eds) ‘Surveys in geometry and number theory, N. YOUNG (ed) Groups St Andrews 20051, CM. CAMPBELL, M.R. QUICK, EF. ROBERTSON & G.C. SMITH (eds) ‘Groups St Andrews 2005 Il, CM. CAMPBELL. M.R. QUICK, EF. ROBERTSON & G.C. SMITH (eds) Ranks of elliptic curves and random matrix theory. J.B. CONREY, D.W. FARMER, F. MEZZADRI & NC. SNAITH (eds) Elliptic cohomology, H.R. MILLER & D.C. RAVENEL (eds) Algebraic cycles and motives I, J. NAGEL & C. PETERS (eds) Algebraic cycles and motives Il, J. NAGEL & C. PETERS (eds) ‘Algebraic and analytic geometry, A. NEEMAN Surveys in combinatorics 2007, A. HILTON & J. TALBOT (eds) ‘Surveys in contemporary mathematics, N. YOUNG & Y. CHOI (eds) ‘Transcendental dynamics and complex analysis, PJ. RIPPON & G.M. STALLARD (eds) Model theory with applications to algebra and analysis , Z. CHATZIDAKIS. D. MACPHERSON, A. PILLAY & A. WILKIE (eds) Model theory with applications to algebra and analysis Il, Z. CHATZIDAKIS, D. MACPHERSON, A. PILLAY & ‘A. WILKIE (eds) Finite von Neumann algebras and masas, A.M. SINCLAIR & R.R. SMITH ‘Number theory and polynomials, J. MCKEE & C. SMYTH (eds) ‘Trends in stochastic analysis, J. BLATH, P. MORTERS & M. SCHEUTZOW (eds) Groups and analysis, _K. TENT (ed) Non-equilibrium statistical mechanics and turbulence, J. CARDY, G. FALKOVICH & K. GAWEDZKI Elliptic curves and big Galois representations, D. DELBOURGO Algebraic theory of differential equations, M.A.H. MACCALLUM & A.V. MIKHAILOV (eds) Geometric and cohomological methods in group theory, _M.R. BRIDSON, PH. KROPHOLLER & IJ. LEARY (eds) Moduli spaces and vector bundles, L. BRAMBILA-PAZ, S.B. BRADLOW, O. GARCIA-PRADA & '$. RAMANAN (eds) Zariski geometries, B. ZILBER ‘Words: Notes on verbal width in groups, _D. SEGAL Differential tensor algebras and their module categories, R. BAUTISTA. L. SALMERON & R. ZUAZUA Foundations of computational mathematics, Hong Kong 2008, F. CUCKER, A. PINKUS & MJ. TODD (eds) Partial differential equations and fuid mechanics, J.C. ROBINSON & J.L. RODRIGO (eds) Surveys in combinatorics 2009, S. HUCZYNSKA, J.D. MITCHELL & C.M. RONEY-DOUGAL (eds) Highly oscillatory problems, B. ENGQUIST, A. FOKAS. E. HAIRER & A. ISERLES (eds) Random matrices: High dimensional phenomena, G. BLOWER Geometry of Riemann surfaces, FP. GARDINER, G. GONZALEZ-DIEZ & C, KOUROUNIOTIS (eds) Epidemics and rumours in complex networks, M. DRAIEF & L. MASSOULIE ‘Theory of p-adic distributions, S. ALBEVERIO, A. YU. KHRENNIKOV & VM. SHELKOVICH Conformal fractals, F. PRZYTYCKI & M. URBANSKI Moonshine: The first quarter century and beyond, J. LEPOWSKY, ]. MCKAY & M.P. TUITE (eds) ‘Smoothness, regularity, and complete intersection, J. MAJADAS & A. RODICIO Geometric analysis of hyperbolic differential equation: An introduction, S. ALINHAC ‘Triangulated categories, T. HOLM, P. JORGENSEN & R. ROUQUIER (eds) Permutation patterns, S. LINTON, N, RUSKUC & V, VATTER (eds) ‘An introduction to Galois cohomology and its applications, _G. BERHUY Probability and mathematical genetics. N.H. BINGHAM & C.M. GOLDIE (eds)London Mathematical Society Lecture Note Series: 375 Triangulated Categories Edited by THORSTEN HOLM Leibniz Universitit Hannover, Germany PETER JORGENSEN University of Newcastle upon Tyne RAPHAEL ROUQUIER University of Oxford EEE) CAMBRIDGE [p UNIVERSITY PRESS“| TOK EABE AOS ACR CAMBRIDGE UNIVERSITY PRESS Cambridge, New York, Melbourne, Madrid, Cape Town, Singapore, Sao Paulo, Delhi, Dubai, Tokyo Cambridge University Press The Edinburgh Building, Cambridge CB2 8RU, UK Published in the United States of America by Cambridge University Press, New York www.,cambridge.org Information on this title: www.cambridge.org/9780521744317 © Cambridge University Press 2010 This publication is in copyright. Subject to statutory exception and to the provisions of relevant collective licensing agreements, no reproduction of any part may take place without the written permission of Cambridge University Press. First published 2010 Printed in the United Kingdom at the University Press, Cambridge A catalogue record for this publication is available from the British Library Library of Congress Cataloguing in Publication data Triangulated categories / edited by Thorsten Holm, Peter Jorgensen, Raphaél Rouquier. p. cm. - (London Mathematical Society lecture note series ; 375) Includes index. ISBN 978-0-521-74431-7 (pbk.) 1. Triangulated categories. I. Holm, Thorsten, 1965— Il. Jorgensen, Peter, 1970- III. Rouquier, Raphaél. IV. Title. V. Series. QA169.T685 2010 512'.62 - de22 2010012362 ISBN 978-0-521-74431-7 Paperback Cambridge University Press has no responsibility for the persistence or accuracy of URLs for external or third-party internet websites referred to in this publication, and does not guarantee that any content on such websites is, or will remain, accurate or appropriate.Contents Preface page Vii Triangulated categories: definitions, properties, and examples 1 THORSTEN HOLM AND PETER J@RGENSEN Cohomology over complete intersections via exterior algebras 52 LUCHEZAR L. AVRAMOV AND SRIKANTH B. IYENGAR Cluster algebras, quiver representations and triangulated categories 716 BERNHARD KELLER Localization theory for triangulated categories 161 HENNING KRAUSE Homological algebra in bivariant K-theory and other triangulated categories. I 236 RALF MEYER AND RYSZARD NEST Derived categories and Grothendieck duality 290 AMNON NEEMAN Derived categories and algebraic geometry 351 RAPHAEL ROUQUIER Triangulated categories for the analysts 371 PIERRE SCHAPIRA Algebraic versus topological triangulated categories 389 STEFAN SCHWEDEvi Contents Derived categories of coherent sheaves on algebraic varieties 408 YUKINOBU TODA Rigid dualizing complexes via differential graded algebras (survey) 452 AMNON YEKUTIELIPreface This volume grew out of a Workshop on Triangulated Categories held at the University of Leeds in August 2006. The meeting, a Satellite of the Interna- tional Congress of Mathematicians 2006, has been generously supported by the Leverhulme Foundation (via the network Algebras, Representations and Appli- cations), the London Mathematical Society (Conference Grant Ref. 1438) and the University of Leeds. Over the past decades, triangulated categories have made their way into many different parts of mathematics, to the extent that today, they can be viewed as a unifying theory underlying major parts of modern mathematics. The Leeds workshop has brought together researchers from many parts of mathematics who all use triangulated methods but would not usually meet at more specialized conferences, with the aim to promote cross fertilization leading to new applications of triangulated categories. The present book collects surveys by leading experts reflecting a broad range of important topics covered at the workshop. However, it is not a proceedings volume recording precisely the talks given at the conference and it does not claim to be a comprehensive coverage of all the numerous applications of triangulated categories throughout mathematics. There are contributions dealing with fundamental general aspects of trian- gulated categories as well as articles covering important applications, e.g. in algebraic geometry, algebraic topology, commutative algebra, algebraic analy- sis, K-theory or representation theory. We wish to express our sincere thanks to the authors of the contributions, as well as to the referees. We think that the interdisciplinary spirit of the successful Leeds workshop and the many fruitful discussions having taken place there are well reflected by the articles and we hope that specialists and non-specialists alike will benefit from the broad perspective on triangulated categories and their applications provided by the surveys. We are very grateful to the staff at Cambridge University Press for their help, their patience and constant support in bringing this book together. Hannover, Newcastle, and Oxford, January 2010 Thorsten Holm, Peter Jérgensen, and Raphaél Rouquier viiTriangulated categories: definitions, properties, and examples THORSTEN HOLM AND PETER JORGENSEN Triangulated categories were introduced in the mid 1960’s by J.L. Verdier in his thesis, reprinted in [16]. Axioms similar to Verdier’s were indepen- dently also suggested in [2]. Having their origins in algebraic geometry and algebraic topology, triangulated categories have by now become indispensable in many different areas of mathematics. Although the axioms might seem a bit opaque at first sight it turned out that very many different objects actu- ally do carry a triangulated structure. Nowadays there are important appli- cations of triangulated categories in areas like algebraic geometry (derived categories of coherent sheaves, theory of motives) algebraic topology (stable homotopy theory), commutative algebra, differential geometry (Fukaya cate- gories), microlocal analysis or representation theory (derived and stable module categories). It seems that the importance of triangulated categories in modern mathe- matics is growing even further in recent years, with many new applications only recently found; see B. Keller’s article in this volume for one striking example, namely the cluster categories occurring in the context of S. Fomin and A. Zelevinsky’s cluster algebras which have been introduced only around 2000. In this chapter we aim at setting the scene for the survey articles in this volume by providing the relevant basic definitions, deducing some ele- mentary general properties of triangulated categories, and providing a few examples. Certainly, this cannot be a comprehensive introduction to the subject. For more details we refer to one of the well-written textbooks on triangulated categories, e.g. [4], [5], [7], [8], [12], [17], and for further topics also to the surveys in this volume. This introductory chapter should be accessible for a reader with a good background in algebra and some basic knowledge of category theory and homo- logical algebra.2 Thorsten Holm and Peter Jorgensen 1. Additive categories In this first section we shall discuss the fundamental notion of an additive category and provide some examples. In particular, the category of complexes over an additive category is introduced which will play a fundamental role in the sequel. Definition 1.1. A category A is called an additive category if the following conditions hold: (Al) For every pair of objects X,Y the set of morphisms Hom,(X, Y) is an abelian group and the composition of morphisms Hom a(Y, Z) x Homy(X, Y) > Homy(X, Z) is bilinear over the integers. (A2) A contains a zero object 0 (i.e. for every object X in A each morphism set Hom4(X, 0) and Hom,(0, X) has precisely one element). (A3) For every pair of objects X,Y in A there exists a coproduct X ® Y in A. Remark 1.2, (i) A category satisfying (A1) and (A2) is called a preadditive category. (ii) We recall the notion of coproduct from category theory. Let C be a category and X,Y objects in C, A coproduct of X and Y in C is an object X @Y together with morphisms ty : X — X @ Yandty : ¥Y > X @ ¥ satisfying the following universal property: for every object Z in C and morphisms fx : X — Zand fy : Y > Z there is aunique morphism f : X @Y > Z making the following diagram commutative. vine X—yr X@r-a-¥ Example 1.3. (i) Let R be a ring and consider R as a category Cr with only one object. The unique morphism set is the underlying abelian group and composition of morphisms is given by ring multiplication. Then Cp satisfies (A1) and {A2), thus preadditive categories can be seen as generalizations of rings. But Cz is not additive in general; in fact the coproduct of the unique object with itself would have to be again this object together with fixed ring elements ¢;, ¢2, and the universal property would mean that for arbitraryTriangulated categories 3 ring elements f,, f2 there existed a unique element f factoring them as fi = furand fp = fo. (ii) Let R be a ring (associative, with unit element). Then the category R- Mod of all R-modules is additive. Similarly, the category R-mod of finitely generated R-modules is additive. In particular, the categories Ab of abelian groups and Vecx of vector spaces over a field K are additive. (ii) The full subcategory of Ab of free abelian groups is additive. (iv) For a ring R the full subcategory R-Proj of projective R-modules is additive; similarly for R-proj, the category of finitely generated projective R-modules. 1.1. The category of complexes Let A be an additive category. A complex over Ais a family X = (X,, dnez where X, are objects in A and dx : X, > X,_, are morphisms such that d, © dn41 =O for all n € Z. Usually, a complex is written as a sequence of objects and morphisms as follows. dy 4 we Xng1 THX, Se Xp Let X = (X,, d*) and Y = (¥,, d*) be complexes over A. A morphism of com- plexes f :X — ¥ isa family of morphisms f = (f, : Xn > Yn)acz Satisfy- ing d? o f, = fn-10 dX forall n € Z, i.e. we have the following commutative diagram. * > Xqup— Xp Xp ost [fe | fort + Yng1 TO Vn Yn The complexes over an additive category A together with the morphisms of complexes form a category C(A), the category of complexes over A. Proposition 1.4. Let A be an additive category. Then the category of complexes C(A) is again additive. Proof, (Al) Addition of morphisms is defined degreewise, i.e. for two mor- phisms f =(fa)nez and g =(n)ncz from X to Y their sum is f + g:= (fn + 8n)nez- Using the additive structure of A it is then easy to check that (A1) holds. (A2) The zero object in C(A) is the complex (0.4, d) where 0, is the zero object of the additive category A and all differentials are the unique (zero) morphism on the zero object.4 Thorsten Holm and Peter Jorgensen (A3) The coproduct of two complexes X = (X,,, dX) and Y = (¥,, 4) is defined degreewise by using the coproduct in the additive category A. More precisely X ® ¥ = (X, ® ¥n, dn)ncz where the differential is obtained by the universal property as in the following diagram. Xn-1 ® ¥y-1 tat a Nee Xn Tr Xn On Nn From uniqueness in the universal property applied to Xn-2 B Yn-2 ved Xn i Xn ®Yn ra Yn it follows that d,_, od, = 0. This complex indeed satisfies the properties of a coproduct in the category of complexes C(A), with morphisms of complexes ty = (ty, )nez 1 X > X @Y and ty = (ty, )nez | ¥Y > X ® Y. For checking the universal property let Z be an arbitrary complex and let fy : X > Z and fy : ¥ —> Z be arbitrary morphisms. The unique morphism of complexes satisfying fy = f oty and fy = f oty is f =(fa)lnez : X BY — Z, where Jn is obtained from the universal property in degree n as in the following diagram. Zn (H06/ sl Ny. Xn Ty Xn @ Yn Yn o Remark 1.5. For complexes over A = R-Mod where R is a ring with unit (and other similar examples) the coproduct of two complexes is more easily be described on elements as X @ Y = (Xn @ Yn, dn)nez Where the differential is given by da(Xn, Yn) = (d* (xn), d*(yn)) for x, € Xq and y, € ¥,, and with morphisms 1x : X > X @Y and ty : ¥Y > X @Y being the inclusion maps. The unique morphism of complexes satisfying fy = f otx and fy = f oly is then given by fn(xns Yn) = Fx(tn) + fy (Yn)Triangulated categories 5 1.2. The homotopy category of complexes Let A be an additive category. Morphisms f, ¢ : X¥ + ¥ in the category C(A) of complexes are called homotopic, denoted f ~ g, if there exists a family (Sn)nez Of morphisms s, : X,— Yn+1 in A, satisfying fy — 8. = 47415, + Sn1dX for all n € Z. In particular, setting g to be the zero morphism, we can speak of morphisms being homotopic to zero. It is easy to check that ~ is an equivalence relation. Moreover, if f ~ g : X — Y are homotopic anda : W — X is anarbitrary morphism of complexes, then also the compositions fa ~ ga are homotopic. In fact, (Span )nez are homotopy maps since (Sn — Bn )Otn = (drain + Sn tbe Mn = A (SpOtn) + (Sp 10 y" Similarly, if f, g : X — Y are homotopic and B : Y > Z is a morphism of complexes then Bf ~ Bg are homotopic. This implies that we have a well-defined composition of equivalence classes of morphisms modulo homotopy by defining the composition on representa- tives. Definition 1.6, Let A be an additive category. The homotopy category K(A) has the same objects as the category C(A) of complexes over A. The morphisms in the homotopy category are the equivalence classes of morphisms in C(A) modulo homotopy, i.e. Homx,a)(X, ¥) := Home4(X, ¥)/ ~~ Proposition 1.7. Let A be an additive category. Then the homotopy category K(A) is again an additive category. Proof. Addition of morphisms in K(A) is defined via addition on representa- tives (it is an easy observation that this is well-defined) and then the sets of morphisms Homx,)(X, Y) inherit the structure of an abelian group from the category C(A) of complexes, and also bilinearity of composition. Moreover, the zero object is the same as in C(A). It remains to be checked that the universal property of the coproduct X ® Y in C(A) (cf. Proposition 1.4) also carries over to the homotopy category. In fact, the equivalence classes of the morphisms ¢x, ¢y and f still make the relevant diagram (cf. Remark 1.2) commutative; for uniqueness we observe that if there is another morphism g making the diagram for the universal property commutative in K(A), i.e. up to homotopy, then this gives a homotopy between f and g. a6 Thorsten Holm and Peter Jorgensen 2. Abelian categories In this section we shall review the fundamental definition of an abelian category, including the necessary background on the categorical notions of kernels and cokernels. The prototype example of an abelian category will be the category R-Mod of modules over a ring R; but we will also see other examples in due course. We first recall some notions from category theory. Let A be an additive category; in particular for every pair of objects X, Y there is a zero morphism, namely the composition of the unique morphisms X — 0 — ¥ involving the zero object of A. The kernel of a morphism f : X > Y is an object K together with a mor- phism k : K — X such that @ fok=0, (ii) (universal property) for every morphism k’ : K' > X such that f ok’ = 0, there is a unique morphism g : K’ > K making the following diagram commutative. xox By the usual universal property argument, the kernel, if it exists, is unique up to isomorphism; notation: ker f. Dually, the cokernel of a morphism f : X — Y is an object C together with amorphism c : Y > C such that (i) cof =9, (ii) (universal property) for every morphism c’ : ¥Y > C’ such thatc’o f =0, there is a unique morphism g : C > C’ making the following diagram commutative.Triangulated categories 7 Again, the cokernel, if it exists, is unique up to isomorphism; notation: coker f. If the above morphism k : ker f —> X has a cokernel in A, this is called the coimage of f, and it is denoted by coim f. If the above morphism c : Y — coker f has a kernel in A, this is called the image of f and it is denoted by im f. Example 2.1. Let R be a ring. In the category R-Mod of all R-modules the categorical kernels and cokernels are the usual ones, i.e., for a morphism f : X — Y we have ker f = {x € X| f(x) =0} and coker f = Y/im f where im f = (f(x)|x € X} is the usual image of f. Remark 2.2. Suppose that for a morphism / both the coimage and the image exist. Then we claim that it follows from the universal properties that there is a natural morphism coim f > im f. In fact, the image of f is the kernel of c : ¥Y + coker f, hence there is a morphism k : im f + Y such that co k = 0 and by the universal property there exists a unique morphism 3 : X + im f making the following diagram commutative. coker f No y ~*imf RA Note thatk o 0k = f ok =0, which implies that g 0 k : ker f > im f must be zero, by using the uniqueness in the following diagram. coker f c I \o k y +*_imf 0 ker f Then we can consider the following diagram for the universal property of the coimageThorsten Holm and Peter Jorgensen X, k, é = ker f. 20, coim f 0 Nim f and deduce that there is a unique morphism coim f — im f, as desired. Definition 2.3. An additive category A is called an abelian category if the following axioms are satisfied: (A4) Every morphism in A has a kernel and a cokernel. (A5) For every morphism f : X — Y in A, the natural morphism coim f > im f is an isomorphism. Example 2.4. (i) Let R be a ring. The category R-Mod of all R-modules is an abelian category. In fact, (A5) follows directly from the isomorphism theorem for R-modules. However, the subcategory R-mod of finitely generated modules is not abelian in general since kernels of homomorphisms between finitely gen- erated modules need not be finitely generated. Indeed we have that R-mod is an abelian category if and only if R is Noetherian. In particular, the category of finite-dimensional vector spaces over a field is abelian, and the category of finitely generated abelian groups is abelian. (ii) The subcategory of Ab consisting of free abelian groups is not abelian. On the other hand, for a prime number p, the abelian p-groups form an abelian subcategory of Ab (an abelian group is called a p-group if for every element a we have p*a = 0 for some k). (iii) For finding examples of additive categories satisfying (A4) but failing to be abelian, the following observation can be useful. Suppose f : X + Y is a morphism with ker f = 0 and coker f = 0, i.e. a monomorphism and an epimorphism. Then the coimage of f is the identity on X, the image of f is the identity on Y and hence the natural morphism coim f > im f is just f itself. So in this special case the axiom (A5) states that a morphism which is a monomorphism and an epimorphism must be invertible.Triangulated categories 9 (iv) Explicit examples of additive categories where axiom (AS) fails for the above reason are the category of topological abelian groups (with contin- uous group homomorphisms) or the category of Banach complex vector spaces (with continuous linear maps). In such categories the cokernel of amorphism f : X — Y is of the form Y/ imy where im, is the closure of the usual set-theoretic image of f. In particular, the natural morphism coim f — im f is the inclusion of the usual image of f into its closure, and this is in general not an isomorphism, Proposition 2.5. Let A be an abelian category. Then the category of complexes C(A) is also abelian. Proof. We have seen in Proposition 1.4 that C(.A) is an additive category, so it remains to verify the axioms (A4) and (A5). (A4) Let f : X — Y be a morphism in C(A), i.e. f = (frnez with fy : X, — Y, morphisms in A. We show the existence of a kernel and leave the details of the dual argument for the cokernel as an exercise. Since A is abelian, each morphism f,, : X, > Y, has a kernel K, := ker fy, in A, coming with a morphism k, : K, — X, Satisfying the above universal property. Note that for every n € Zwe have fy_1 0d* ok, =d? 0 fy okn =0. Then it follows by the universal property of kernels that there is a unique morphism dX : K, > Ky~1 such that ky, 0 d* = d* ok,. Note that ky-1 0 d* o dX, = dX oky o dk, = dX odX,, okna1 =0 since X is a complex. By uniqueness of the map in the universal property of K,-1 it follows that dK o dX, , = 0, ie. (Ky, dX) is a complex. Combining the universal properties of the kernels K,, it easily follows that the complex (K,,, ¢) indeed satisfies the universal property for the kernel of f in C(A). (A5) The crucial observation is that a morphism of complexes f = (fn) : X — ¥ is an isomorphism in C(A) if and only if each f, is an isomorphism in A. In fact, if each f, is an isomorphism, with inverse g,, then the family 8 = (gn) is automatically a morphism of complexes (and hence clearly an inverse to f in C(.A)): for all n € Z we have dX 0 Bnet = Bn fn dy O Bnet = Bn dN, O fast Bnet = Bn dN). The reverse implication is obvious. For axiom (AS) now consider the natural morphism coim f > im f. In the proof of (A4) above we have seen that kernels and cokernels in C(A), and10 Thorsten Holm and Peter Jorgensen hence also the morphism coim f — im f, are obtained degreewise. But since A is abelian by assumption, we know that for every n the natural morphism coim f, + im f, in A is indeed an isomorphism. Then, by the introductory remark, the morphism of complexes (coim f, — im f,)nez is an isomorphism in C(A). go An important observation is that the homotopy category K(A) is not abelian in general, even if A is abelian. Example 2.6. We provide an explicit example for the failure of axiom (A4) in a homotopy category. Consider the abelian category A = Ab of abelian groups. Let f : X > Y be the following morphism of complexes of abelian groups, with non-zero entries in degrees 1 and 0, +z We 7 9... In the category C(Ab) of complexes f is non-zero and has the zero complex as kernel (cf. the proof of Proposition 2.5). However, f is homotopic to zero (with the identity as homotopy map), ie. f = 0 in the homotopy category K(Ab). We claim that in the homotopy category f has no kernel. Recall the cate- gorical definition of the kernel of a morphism f : X — Y from Section 2. Suppose for a contradiction that our morphism f had a kernel in K(Ab). So there is a complex ... > Ki > Ko > K-, >... anda morphism k = ko : Ko > Z of abelian groups (in all other degrees the map k has to be zero since X is concentrated in degree 0). The image of k, being a subgroup of Z, has the form rZ for some fixed r € Z. Now choose K' = X and consider the morphisms 1: K' > X given by multiplication with / for any / € Z. Clearly, f of =Oin K(Ab) since f = 0 in K(Ab). According to the universal property of a kernel, there must exist (unique) morphisms u, : Z— Ko such that k ou; =/ up to homotopy. However, these maps are from K’ = X to X and this complex is concentrated in degree 0. Thus there are no non-zero homotopy maps and so k ou, =1as morphism of abelian groups. But the image of k 0 u; is contained in the image of k which is rZ for a fixed r, so ko u; =! can not hold for arbitrary / € Z, a contradiction. Hence axiom (A4) fails and therefore the homotopy category K(Ab) is not an abelian category.Triangulated categories 11 3. Definition of triangulated categories We have seen in the previous section that the homotopy category of complexes is not abelian in general. We shall see in Section 6 below that K(A) carries the structure of a triangulated category, a concept which we are going to define in this section. Roughly, one should think of the distinguished triangles occurring in this context as a replacement for short exact sequences (which do not exist in general since K(A) is not abelian). However, for an additive category to be abelian is purely an inherent property of the category. On the other hand a triangulated structure is an extra piece of data, consisting of a suspension functor and a set of distinguished triangles chosen suitably to satisfy certain axioms. In particular, an additive category can have many different triangulated structures; see [1] for more details and examples. A functor I between additive categories is called an additive functor if for every pair of objects X, Y the map Hom(X, Y) > Hom(Z(X), X(Y)) is a homomorphism of abelian groups. Let T be an additive category and let & : T > T be an additive functor which is an automorphism (i.e. it is invertible, thus there exists a functor D~! on T such that Z o Z~! and Z~! o ¥ are the identity functors). A triangle in T is a sequence of objects and morphisms in T of the form XSyY>Z4 ux. A morphism of triangles is a triple (f, g,h) of morphisms such that the following diagram is commutative in T x “ey +7 “yy ook le x Hey Be gr whe sy: If in this situation, the morphisms f, g and h are isomorphisms in 7, then the morphism of triangles is called an isomorphism of triangles. Definition 3.1. A triangulated category is an additive category T together with an additive automorphism ©, the translation or shift functor, and a collection of distinguished triangles satisfying the following axioms (TRO) Any triangle isomorphic to a distinguished triangle is again a distin- guished triangle. (TR1) For every object X in T, the triangle X 8 ¥ 305 EX is adistin- guished triangle.12 Thorsten Holm and Peter Jorgensen (TR2) For every morphism f : X -> Y in T there is a distinguished triangle of the form X 4 YoZ> xX. (TR3) If X + ¥ 3 Z4 EX is a distinguished triangle, then also Y > ZS ux 3 tYVisa distinguished triangle, and vice versa. (TR4) Given distinguished triangles X Y % Z-% EX and X'S y' > z 4 Ex! ’, then each commutative diagram x “ey Ye zy ok Py y Hey Mz ae sy can be completed to a morphism of triangles (but not necessarily uniquely). (TRS) (Octahedral axiom) Given distinguished triangles X 4y¥sZ7'3 EX, YS Z—> X'> BY and X 4 Z > Y' > EX, there exists a distinguished triangle Z' > Y' + X' — £2’ making the following diagram commutative xX “+y —+7' —-rx idx | |e | [cox vu xX —+Z —+Y' —~-EXx job |b y —+z —~-x’ —-xy Lol | Z' —+ Y' —+X' —+ 7’ Remark 3.2. The above version (TRS) of the octahedral axiom is taken from the book by Kashiwara and Schapira [7, Sec. 1.4]. There are various other versions appearing in the literature which are equivalent to (TRS), see for instance A. Neeman’s article [13] or his book [12]; a short treatment can also be found in A. Hubery’s notes [6] (which are based on the former references). We shall only mention two variations here. Mainly a reformulation of the axiom (TRS) is the following. Note that in (TRS) the given three distinguished triangles are placed in the first three rows, whereas in (TR5’) below they are placed in the first two rows and the second column.Triangulated categories 13 (TRS”) Given distinguished tiangles X > ¥ > Z’ > EX,¥5Z—> x4 EY and X 5 Z — Y’ ++ EX, then there exists a distinguished triangle Z’ > y’ + xX’ EZ’ making the following diagram commutative and satisfying (Zu)s = lv’. xX “ey —+7 —-rx “lok Lb xX “+7 —+y ++r5x lt y By | | xY— =z’ Itis not difficult to check that (TR5) and (TR5’) are indeed equivalent; we leave this verification as an exercise to the reader. The following version (TRS”) of the octahedral axiom can be found in Neeman’s book [12, Prop. 1.4.6]. It is less obvious that it is equivalent to (TRS); for details on this we refer the reader to [12], [13] and [6]. (TR5”) Given distinguished triangles X SYs7'525KX,Y3725 X’ + ZY and X 3 Z—> Y’ > EX, then there exists a distinguished tri- angle Z’ + Y’ —» X’— EZ’ making the following diagram commutative in which every row and every column is a distinguished triangle. X “+y —+7' —+2Ex ide | | ° | [ssa x +z —+y' —~xx f tot idx 0 —~+ x’ —+x' —0 f | | | ex. py—- Ez ex14 Thorsten Holm and Peter Jorgensen 4. Some formal properties of triangulated categories We shall draw some first consequences from the definition. Let T be a triangu- lated category with translation functor 2. Proposition 4.1 (Composition of morphisms). Let X 5 Y 5 Z-4 EX bea distinguished triangle. Then v o u = 0 and wo v = 0, i.e. any composition of two consecutive morphisms in a distinguished triangle vanishes. Proof. By the rotation property (TR3) it suffices to show that vo u = 0. Also by (TR3) we have a distinguished triangle Y 5 Z-% EX —3“ EY. By (TRI) and (TR4) the following diagram can be completed to a morphism of triangles. y te z 4 sy ~E4 yy | ja zo Z ad, Z o. Oo 2. =mZ In particular, —Z(v ou) = —Lvo Lu = 0 which implies v ou =0 since E is an automorphism. Qo Proposition 4.2 (Long exact sequences). Let X 5 Y > Z-4 EX bea dis- tinguished triangle. For any object T € T there is a long exact sequence of abelian groups zu. 4, ; . > Homy(T, 5'X) = Homy (7, E'¥) 2° Homy(T, E'Z) 2S omy (T, E't!X) >... Proof. For abbreviation we denote by f, :=Homz(T, f) the morphism induced by f under the functor Homz(T, —) on the additive category T. By the rotation property, it suffices to show that Homz(T, 2X) Ee Homz(T, ry) By Homz(T, £'Z) is an exact sequence of abelian groups. By Proposition 4.1 we have Z'v o Diu = 0 and hence also L'v, o Liu, = 0, i.e. the image of Z'u, is contained in the kernel of D'v,. Conversely, take f in the kernel of Z‘v,. Consider the following diagram whose rows are distinguished triangles by (TR1) and (TR3). rT oO, 0 oO. pity SS aid, pity [poy jo jy y tz xx F4 5yTriangulated categories 15 The left hand square is commutative by assumption on f. By (TR4) there exists amorphism h : D-‘+!T — © X completing the above diagram to a morphism of triangles. In particular, D+! f = Luo h andhence f = Liu o Li'hisin the image of Z'u, as desired. oO Proposition 4.3 (Triangulated 5-lemma). Suppose we are given a morphism of distinguished triangles as in the following diagram. x 4+ y +7 "oxy lk y Hey 2h gy whey If f and g are isomorphisms then also h is an isomorphism. Proof. We apply the functor Hom(Z’, —) := Homy7(Z’, --) to the distinguished triangles. By Proposition 4.2 this leads to the following commutative diagram whose rows are exact sequences of abelian groups. Hom(2', X) —»Hom(Z', Y) —-Hom(Z’,Z) —-Hom(Z', 2X) —* Hom(Z', ZY) |s |e. |r. [zs | 6. Hom(Z', X’) --Hom(Z’, ¥’) —--Hom(2', Z‘) --Hom(Z', £X') — Hom(Z’, EY’) By assumption, f and g are isomorphisms and hence also f,, g., Uf. and &g, are isomorphisms. So we can appeal to the usual 5-lemma in the category of abelian groups to deduce that A, is an isomorphism. In particular the identity idz has a preimage, i.e. there exists a morphism g € Homz(Z’, Z) such that hoq=idz. A similar argument using the functor Homz(—, Z’) produces a left inverse to h, thus A is an isomorphism. Oo Proposition 4.4 (Split triangles). Let X 5 Y > ZS XX bea distinguished triangle where w = 0 is the zero morphism. Then the triangle splits, i.e. u isa split monomorphism and v is a split epimorphism. Remark 4.5, The notion of split monomorphism is synonymous with that of a section, and a split epimorphism is also known as a retraction. Proof. We first show that u is a split monomorphism, i.e. there exists a mor- phism w’ such that u' ou = idy. We have the following commutative diagram of distinguished triangles.16 Thorsten Holm and Peter Jorgensen x Hey 2-7 O-xy lia lo lia x dey OL, O.sy By (TR3) and (TR4) it can be completed to a morphism of triangles, i.e. there exists u’: Y — X such that u’ ou = id. Similarly, one can show that v is a split epimorphism, i.e. there is a morphism vu’: Z— Y such that vo v’ = id. o 5. Abelian categories vs. triangulated categories As an application of the formal properties in the previous section we shall compare the notions of abelian categories and triangulated categories. Definition 5.1. An abelian category A is called semisimple if every short exact sequence in A splits. Example 5.2. (i) Let R bea semisimple ring. Then the module categories R-Mod and R-mod are semisimple. In particular, the category of vector spaces Vecx over a field K is semisimple. (ii) The category Ab of abelian groups is not semisimple. For instance, the short exact sequence 0 > Z/2Z —» Z/4Z + Z/2Z, + 0 does not split. The following result illustrates that the concepts of abelian and triangulated categories overlap only slightly. Theorem 5.3. Let T be a category which is triangulated and abelian. Then T is semisimple, Proof. Let 0 X 4 y 4. 20 be a short exact sequence in T. We have to show that it splits; to this end it suffices to show that f is a section, i.e. there exists a morphism f’ : Y + X such that f’o f = idy. By (TR2) and (TR3), f can be embedded into a distinguished triangle sly-Sxbryv. The composition of consecutive morphisms in a distinguished triangle is always zero by Proposition 4.1, in particular f ou = 0, But f is a monomorphism in T since itis the first map in a short exact sequence, hence u = 0. Thus we haveTriangulated categories 17 a distinguished triangle x4Syvyv%eux where Zu = 0. Now the triangle splits by Proposition 4.4. QO We shall see in the next section that the homotopy category K(A) of com- plexes over an additive category .A is a triangulated category. This, together with the preceding theorem, will then give a more structural explanation of the earlier observation that K(Ab) is not abelian in Example 2.6, where we have used an ad-hoc argument to show that morphisms do not necessarily have a kernel. 6. The homotopy category of complexes is triangulated Let A be an additive category, with corresponding category of complexes C(A) and homotopy category K(A). As discussed above, the homotopy category K(.A) is in general not abelian, even if A is abelian. We shall explain in this section how the homotopy category K(A) becomes a triangulated category. We first need an additive automorphism on K(A) which serves as transla- tion functor. This functor can already be defined on the level of the category C(A). Definition 6.1. In C(A) we construct a translation functor X =[1] by shifting any complex one degree to the left. More precisely, for an object X = (Xy.d*)nez in C(A) we set XU = (Xn A pez, with XUn = Xn and dXUl = —dX |. For a morphism of complexes f = (fu)nez in C(A) we set FU := (FU nex where f(Mn = fn-1- Remark 6.2. (i) The sign appearing in the differential of X[1] might look arbitrary; it will become clear later when discussing the triangulated structure of the homotopy category why this sign is needed. (ii) The functor £ = [1] defined above is an additive functor and moreover an automorphism of the category C(A). (iii) Note that the above definitions are compatible with homotopies so we have a well-defined induced functor £ = [1] on the homotopy category K(A).18 Thorsten Holm and Peter Jorgensen The next step for getting a triangulated structure on the homotopy category is to find a suitable set of distinguished triangles. To this end, the following construction of mapping cones is crucial. Definition 6.3. Let f be a morphism berween complexes X = (Xn,d*) and Y =(¥n, a"). The mapping cone M(f) is the complex in C(A) defined by -dy, 0 fot dy) Remark 6.4. (i) There are canonical morphisms in C(A) as follows M(x =Xn1® Yq and dU) ve ( a(f):¥ + M(f), a(f)n = (0, idy,) and Bf): M(f) > XU], BCA)n = (idx, 0). Note that B( f) is a morphism of complexes because the differential in X[1] carries a sign. From the above definitions we get a short exact sequence of chain complexes 0s ¥ 3 mp 28 xno. (ii) Let f : X — Y be a morphism of complexes. The short exact sequence o-Y a M(f) a X{1] — 0 splits (i.e. there is a morphism of com- plexes o : X[1] > M(f) such that B(f) oo = idy;y) if and only if f is homotopic to zero. In fact, a splitting map is given by o(x) := (x, —5(x)) where s is a homotopy map. Example 6.5. (i) For any complex X consider the zero map f : X — 0 to the zero complex. Then the mapping cone is M(f) = X[1]. On the other hand, the mapping cone of g : 0 — ¥ is just M(g) = ¥ itself. (ii) Let A and B be objects in A and view them as complexes X4 and Xz concentrated in degree 0. Any morphism f : A > B in A induces a morphism of complexes f : X4 — Xg. Its mapping cone is the complex OS ALB0-... where A is in degree 1 and B in degree 0.Triangulated categories 19 (iii) Let X = (Xn, 4%) be any complex in C(A). The mapping cone of the identity morphism idx has degree n term equal to X,_) @ X,, and differ- ential vara OD. y @Xn > Xn2 OX, Gan a) 1 Xa n n-2 © Xn-1- The identity morphism on the mapping cone M(idx) is homotopic to zero, 0 idx, 0 0 category K(A) the identity idya,) is equal to the zero map. As a conse- quence, in the homotopy category, the mapping cone M (idx ) is isomorphic to the zero complex. via the map 5 = (Sn)nez Where s, = ( ). Thus, in the homotopy It is easy to check that the morphisms a( f) and A( f) are also well-defined in the homotopy category K(A) (i.e. independent on the choice of representatives of the equivalence class of morphisms). This leads to the following definition. Definition 6.6. A sequence of objects and morphisms in the homotopy category K(A) of the form f af) BN = x Y —> M(f) —> X() is called a standard triangle. A distinguished triangle in K(A) is a triangle which is isomorphic (in K(A)!) to a standard triangle. With this class of distinguished triangles the homotopy category obtains a triangulated structure as we shall show next. Due to the technical nature of the axioms of a triangulated category, the proof that a certain additive category is indeed triangulated is usually rather long, can be partly tedious and can still be quite involved. In this introductory chapter we want to present such a proof at least once in detail. Theorem 6.7. Let A be an additive category. Then the homotopy category of complexes K(A) is a triangulated category. Proof. We have to show that with the above translation functor [1] and the set of distinguished triangles just defined, the axioms (TRO)-(TR5) are satisfied. The axioms (TRQ) and (TR2) hold by Definition 6.6. (TR1) From the mapping cone construction there is a standard triangle x © x — Miidx) — xD.20 Thorsten Holm and Peter Jorgensen By Example 6.5 above, M(idx) is isomorphic to the zero complex in the homotopy category. Hence we indeed have a distinguished triangle x 8s x0 xu. (TR3) Because the rotation property is compatible with isomorphisms of trian- gles, it suffices to prove (TR3) for a standard triangle a(f) BA) xv M(f)—> We shall show that the rotated triangle X{H. y 2 BA) aft —> M(f) —> X{1] > YO] is isomorphic in K(A) to the following standard triangle for a(f), y%Q ala f)) Bap) M(f) —> For constructing an isomorphism between the latter two triangles we take the identity maps for the first, second and fourth entries. Moreover, we define morphisms M(a(f))'—> Y(). $ = (on): XU] > M(@(f)) by setting $, = (—fn-1, idx,_,, 0) and conversely = (Wn): M(@(f)) > X[I] by setting ¥, = ©. idy,_,, 0). These yield morphisms of triangles since by definition B(a(f)) 0 ¢@ = —f{\}, and # o B( f) ~ a(a(f)) via the homotopy given by 0 —id (: 0 ) 1 M(Pyn = Xn-1 B Yn > M(@(f)nt1 = Yn @ Xn @ Png 0 0 Similarly, yy is a morphism of triangles since B(f) = y 0 a(ae( f)) by definition and —f[1] 0 w ~ B(a(f)) via the homotopy (0, 0, — id): M(@(f)n > ¥ (Un. Finally, and most importantly for proving (TR3), the above morphisms are isomorphisms in K(A) because we have y o ¢ = idx,1 (by definition) and $0 ~ idaacyy Via the homotopy map 00 id 0 0 0 |: M@(f)yn > M@(f)n+1 00 0 (recall that M(a(f))n = Yn) ® Xn—-1 @ Yn).Triangulated categories 21 (TR4) Again it suffices to prove the axiom for standard triangles. By assump- tion we have a diagram x +.y MO) yyy BO XU] |r |e [0 y “Ly au’) Mw’) Bu’) xT] where the left square commutes in K(A), i.e. there exist homotopy maps Sn Xn > Yi4, such that gan — Ufa = Sn + 5,-1a* for all n € Z. For completing the diagram to a morphism of triangles we define h = (hy)nez : M(u) > M(u’) by setting hy = (fe ° ) | M(W)y = Xq-1 @ Yq > Mn = XI OY. Sn-1 8n This is indeed a morphism of complexes because of the homotopy property of s given above. Moreover, the completed diagram commutes since by definition we have that h o a(u) = au’) og and B(u’') oh = f[1] o B(u); note that these are proper equalities, not only up to homotopy. (TRS) Again it suffices to prove the octahedral axiom for standard trian- gles. From the assumptions we already have the following part of the relevant diagram x ey ey 2 xm | 4 | x vu Zz a(vu) Mu) Buu) xt Tote y fez Sem 2 yy a(u) | o(vus) | | |«orn M(u) M(vu) Mv) M(u){1] We now define the missing morphisms as follows. Let f = (f,): M(u) > M(vu) be given in degree n by f, = ("5 °) and set g =(g,): In M(uu) > M(v) to be given by gn = (“5 i : ). Finally define A: Zn22 Thorsten Holm and Peter Jorgensen M(v) > M(u)[1] as the composition a(u)[1] o Bu), i.e. it is given by the matrix (; . 0): Then it is easy to check from the definitions that all Yn squares in the completed diagram commute (not only up to homotopy). For proving (TRS) it now remains to show that the bottom line Mu) & Mou) Me +> Mw{l) is a distinguished triangle in K(A). To this end we construct an isomorphism to the standard triangle Mu) > Mow “2 uy “2 Met. Note that only the third entries in the triangles are different. So it suffices to find morphisms o = (o,): M(v) > M(f) and t =(t,): M(f) > M(v) leading to commutative diagrams (in K(A)!), ie. we need that B(f)oo =h, hot =£(f),o 0g =a(f)andt oa(f) = g, up to homotopy. Moreover, we have to show that they are isomorphisms in the homotopy category. We set 0 0 idy,, 0 0 id 1 0 ome [ "Gg | and n= (5 “oF ia): 0 idz, First, let us check that o and t give commutative diagrams. Directly from the definitions we get that t oa(f) = g; in fact both are given in degree n by the map (“s' io ) : X,-1 ® Zn > Yn-1 ® Zy. Also by definition we see 2, that B(f) oo = h, both given by (iat 0) t ¥n-1 ® Zn > Xn-2 ®B Yn-1- Yo=v The remaining commutativities will now only hold up to homotopy. Note that a(f)—aog: M(vu) > M(f) is given in degree n by 0 0 —unz-) 0 cmt Ns Xn © Zn > Xn ® Vn @ Xn @ Zn. idxy,, 0 0 0 We claim that e(f) — o © g is homotopic to zero, ie. a(f) =o 0 g in K(A). In fact, a homotopy map s = (s,) where s, : M(vu), > M(f)n+1 is given by idy,, 0 0 0 0 0 1 Xn-1 @ Zp > Xn-1 ® Yn @ Xn @ Zngi- 0 0Triangulated categories 23 For verifying the details recall that the differential of the mapping cone M(f) is given by dX, 0 0 0 gin | 2 “Gy 0 0 " id, 0 -d¥, 0 0 Un-1 (Uu)p-1 d? Finally, consider B(f) — h o t : M(f) > M(u)[1] which in degree n is given by (ite 0 0 0 Oo mas )) Kaa @ Yan @ Kaas @ Zp > Kya @ Yous This can be seen to be homotopic to zero by using the homotopy map s = (s,) where (° 0 idx,, 0 Sn = 00 0 0) 2 Xn-2 B Yp-1 ® Xn-1 @ Zy > Xn-1 Bn. For the straightforward verification again use the differential of M(f) as given above. For completing the proof it now remains to show that o and Tr are iso- morphisms in the homotopy category. We have t oo = idsgy) by definition. Conversely, the composition o 0 7 is in degree n given by 0 0 0 0 0 idy,, unr 0 0 60 0 0 0 0 0 idz, If we then define homotopy maps s, : M(f)n > M(f)n+1 by setting 0 0 -idx,, 0 scx 00 0 0 "10 0 0 0 00 0 0 then we have oot —idycy) = OM 05 + 5n-10dn” which is easily checked using the differential of M(f) as given above. Thus o ot = idycy) in the homotopy category K(A) and we have proved the octahedral axiom for K(A). o Remark 6.8. We have seen that for every standard triangle x & y 2B uc 2S xy