A Mathematical Construction of an E6 Grand Unified Theory
AA Mathematical Constructionof an E Grand Unified Theory
A Thesis Presented for the Degree of
Master of Science
Vivian Anthony Britto under the supervision of
Prof. Dr. Mark Hamilton
Mathematisches InstitutLudwig-Maximilians-Universität MünchenSeptember 2017 a r X i v : . [ h e p - t h ] F e b a gloria di colui che tutto moveper l’universo penetra, e risplendein una parte piú e meno altrove. Dante Alighieri,
Paradiso , Canto I eclaration
I hereby declare that I am the sole author of this thesis. Where the work of others has beenconsulted, this is duly acknowledged in the references that appear at the end of this text. Ihave not used any other unnamed sources. All verbatim or referential use of the sourcesnamed in the references has been specifically indicated in the text. This work was notpreviously presented to another examination board and has not been published.21st September 2017, München V cknowledgements I am indebted to Prof. Hamilton for his guidance and patience during the writing of thisthesis, and for making me a better mathematician and researcher in the process. Gratitudeis also owed to Dr. Robert Helling for many enlightening discussions, and for providingthe inspiration for the final chapter of this paper.I was lucky to have the finest colleagues at LMU, some of whom deserve a special men-tion: Alex Tabler and Danu Thung for their companionship, and the endless conversationsabout mathematics, physics, LaTeX, and everything in between; Martin Dupont for hiscamaraderie and careful proofreading of the final draft, and Agnes Rugel, who gentlywalked me through my many existential crises. Finally, I must thank my parents, to whomI owe everything, and my brother, for being a constant source of joy in my life.
VII ntroduction
Nature is simple. This article of faith, often taken for granted, sometimes fought overbitterly, has ever been at the centre of physicists’ attempts to describe reality. A wonderfuldeclaration of this is found in the third book of the
Principia , in a passage in which Newtonsharpens the proverbial razor of that famous monk from Ockham: “We are to admit nomore causes of natural things than such as are both true and sufficient to explain theirappearances. To this purpose the philosophers say that Nature does nothing in vain, andmore is in vain when less will serve; for Nature is pleased with simplicity and affects notthe pomp of superfluous causes.”The Standard Model of particle physics is decidedly not simple. It is nevertheless anabsolute triumph of modern science, remarkable for its economy in asserting but the singleguiding principle of gauge symmetry within the framework of quantum field theory. Lesspleasing is the actual U ( ) × SU ( ) × SU ( ) gauge group itself, with the seemingly arbitrarycharges of the U ( ) hypercharge group. Other aspects also demand explanation: theHiggs mechanism and corresponding hierarchy problem, the observed non-zero massesof neutrinos, the 19 unrelated parameters within the Standard Model that need to befine-tuned, and the inability to account for a potential cold dark matter particle; theoreticalpuzzles include the mathematical validity of the path integral, the chirality of the leptonsand quarks, and the fact that these particles come in pairs. Perhaps most damning of all,the Standard Model cannot account for gravitation, because quantum field theories ofgravity generally break down before reaching the Planck scale.Grand unification attempts to answer some of these questions by positing that the sym-metry of the Standard Model is a broken one, a shadow of some other, more fundamentalsymmetry of nature that is only accessible at extremely high energies. Mathematically, thiscorresponds to embedding the symmetry group of the Standard Model 𝐺 SM into a larger,often simpler Lie Group, and picking a representation of the same such that it reduces to theStandard Model fermion representation when one restricts to 𝐺 SM . The first example camein 1974, when Georgi and Glashow proposed their SU ( ) theory; though it was definitivelydisproved some twenty years later, it remains the prototypical grand unified theory forits aesthetic simplicity; its most important feature was certainly its logical explanation ofthe fractional charges of the quarks. This is a virtue that can be extended to the Spin ( ) grand unified theory, another brainchild of Howard Georgi. The spinor representations inwhich it accommodates the Standard Model fermions are completely natural, separatingthe left- and right-handed particles in two different irreducible representations. We willstudy this theory in some detail in chapter 2, since it plays a vital role in the constructionof the E grand unified theory, the focus of this paper.An E grand unified theory first appeared in 1976, due to Gürsey, Ramond and Sikivie.Of the five exceptional groups, E is considered the most attractive for unification due to thefollowing reasons: (i) it contains both Spin ( ) × U ( ) and SU ( ) × SU ( ) × SU ( ) as maximalsubgroups, each of which admit embeddings of the Standard Model; (ii) uniquely amongthe exceptional groups, it admits complex representations; in particular, its 27 dimensionalfundamental representation accommodates one generation of left-handed fermions underthe usual charge assignments; (iii) all of its representations are anomaly-free. We will IX discuss each of these aspects in the coming pages.This thesis was originally conceived as an extension of Baez and Huerta’s analysis in The Algebra of Grand Unified Theories [8] to the case of the E theory. There, the authorsundertook a mathematical introduction to the representation theory of the StandardModel, and of grand unified theories—they treated in some detail the SU ( ) , Spin ( ) and Pati-Salam models, and then considered how these theories may be related to eachother. Their pedagogical approach forms the basis for most of chapter 1, and chapters 2.3and 2.5, where we, respectively, introduce the Standard Model, and prove that SU ( ) andSpin ( ) each extend it. Chapter 3.2, where we show that E is a grand unified theory, isthe culmination of this project.John Adams’ Lectures on Exceptional Lie Groups [4] was the other major influence onthis paper. His lucid presentation of the construction of these “curiosities of Nature”contained the three main ingredients we needed to prove that E extended the Spin ( ) theory: (i) an explicit realisation of the subgroup Spin ( ) × U ( )/ Z ⊂ E ; (ii) a route tothe 27-dimensional fundamental representations of E , and (iii) the characterisation ofthe restriction of these representations to Spin ( ) × U ( ) . Moreover, his introduction toClifford Algebras, and the Spin groups and their representations, enabled us to deepenBaez and Huerta’s discussion of the Spin ( ) grand unified theory; in particular, we wereable to make more precise the connection between the SU ( ) fermion representation Λ ∗ C ,and the spinor representations Δ ± of Spin ( ) Chapters 3.2 and 3.3 contain our modest contributions to the literature. In the former,we explicitly check that that Z kernel of the homomorphism Spin ( ) × U ( ) → E actstrivially on every fermion; this is absolutely essential (in the cascade of unified theoriesthat we consider) for E to extend the Spin ( ) theory, and hence the Standard Model. Webelieve the reason that this result does not appear anywhere in the (predominantly physics)literature on the subject is the same reason that the Z kernel of the homomorphism 𝐺 SM → SU ( ) is rarely mentioned: physicists are often content to deal with thesesymmetry groups at the level of Lie algebras, which are indifferent to finite quotientsof Lie groups. This affection for Lie algebras extends to their discussions of symmetrybreaking in grand unified theories, which are almost universally analysed using Dynkindiagrams and related techniques. While computationally preferable, we felt that followingthis method would break with the spirit of the rest of the paper, so we attempted tounderstand symmetry breaking, in particular the symmetry breaking of the exotic E fermions under Spin ( ) → SU ( ) , using a different approach: we explicitly embedded 𝔰𝔲 ( ) ↩ → 𝔰𝔬 ( ) (cid:27) 𝔰𝔭𝔦𝔫 ( ) , and then solved the related eigenvalue problem; this is the workof chapter 3.3. The result of this calculation is table 3.1, where one sees how the StandardModel fermions and their new exotic compatriots fit into the fundamental representationof E . This apparent bounty of new physics was the impetus for the final chapter, on thephenomenology of grand unified theories.To avoid getting lost in quantum field theory, we restricted ourselves to the followingquestion in chapter 4: are there any predictions of grand unified theories that comesolely out of representation theory? One of the most famous is certainly the Weinbergangle, and we treat this in section 4.1. We also consider in some detail, because it has arather nice mathematical interpretation, the issue of anomaly cancellation; this is not aphenomenological prediction of grand unified theories per se, but rather, a requirementon their fermion representations: in section 4.2, we present Okubo’s proof [70] that allrepresentations of E are anomaly-free. We devote the final section of this paper to abrief but general discussion on the signatures of grand unified theories, and their presentoutlook. ontents Spin ( ) Grand Unified Theory 13 ( ) Grand Unified Theory . . . . . . . . . . . . . . . . . . . . . . . . 222.4 Clifford Algebras . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 242.4.1 The Spin Groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 262.4.2 Clifford Modules and Representations . . . . . . . . . . . . . . . . . 282.5 The Spin ( ) Extension of SU ( ) . . . . . . . . . . . . . . . . . . . . . . . . . 30 E Grand Unified Theory 35 and E . . . . . . . . . . . . . . . . . . . . . . . . . . 353.2 The E Extension of Spin ( ) . . . . . . . . . . . . . . . . . . . . . . . . . . . 413.3 The New Fermions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43 Appendix The Construction of G XI hapter 1 The Standard Model
Any discussion of grand unification must begin with the Standard Model of particlephysics. This rather uninspiringly-named theory is in fact a theory of almost everything: itdescribes three of the four known fundamental forces in the universe (the electromagnetic,weak, and strong interactions), as well as classifying all known elementary particles. Itwas developed in stages throughout the latter half of the twentieth century, with thecurrent formulation being finalized in the mid-1970s upon experimental confirmation ofthe existence of quarks. The history of this development is a fascinating subject in its ownright, featuring brilliant scientists, and set against the backdrop of some of the darkestperiods of the last century. We refer the reader to [21] for the history, and to [51] for aremarkable collection of scientific essays from the pioneers of the field.Since those early days, experimental confirmation of Standard Model predictions haveonly added to its credence: highlights include the discover of the top quark in 1995, thetau neutrino in 2000, and the Higgs boson in 2012. Indeed, it can be said that the stunningexperimental success of the Standard Model is often a cause for frustration among moderntheoretical physicists, many of whom are holding out for evidence of new particles to lendsupport to the many projects of physics “beyond the Standard Model”; a highly-readableoverview of the major contenders can be found in [61].We have already encountered some of the shortcomings of the Standard Model: it doesnot fully explain baryon asymmetry, incorporate the full theory of gravitation as describedby general relativity, or account for the accelerating expansion of the universe as possiblydescribed by dark energy; the model does not contain any viable dark matter particlethat possesses all of the required properties deduced from observational cosmology; italso does not incorporate neutrino oscillations and their non-zero masses. Understandingthese difficulties is beyond the scope of this paper , but it is nevertheless clear that theyare a strong motivation to look for other, hopefully more complete theories. In any case,the Standard Model lies at the heart of all model-building, of which grand unified theoryis a part, so we absolutely must understand it before we move on. The basic theory of mathematical groups and their representations is really all that we willneed to understand the algebra of the Standard Model, and of grand unified theories; inthe first section below, we will briefly review the necessary concepts, and set the notation.We mention some references, noting that these represent but a sample of the literature: thebook by Hall [43] is a solid introduction to Lie groups, algebras and their representations;Fulton and Harris’ text [32] on representation theory goes even deeper; for an approach Reference [82] is a helpful starting point. 1 1 . T H E S TA N DA R D M O D E L more geared towards physicists, the book of Fuchs and Schweigert [31] is an excellentresource.In section 1.1.2, we will formulate and motivate the two fundamental principles ofthe representation theory of particle physics. Though these rules are surprisingly easy tostate and work with, their origins do require some preparation to appreciate, since theyare best encountered within the framework of mathematical gauge theory. Referencesfor this field abound, we mention but three: for the mathematically inclined, there is thevenerable text by Bleecker [17], and the lecture notes by Hamilton [45]; the physicist canturn to Nakahara [68] for his concise and clear presentation.
We follow [4] in these paragraphs. A
Lie group 𝐺 is a group which is also a smoothmanifold such that the maps 𝐺 × 𝐺 → 𝐺 , ( 𝑔, ℎ ) ↦→ 𝑔 ℎ and 𝐺 → 𝐺 , 𝑔 ↦→ 𝑔 − are smooth.A homomorphism 𝜃 : 𝐺 → 𝐻 of Lie groups is a homomorphism of groups which is also asmooth map. A subgroup 𝐻 ⊂ 𝐺 is said to be normal if and only if 𝑔𝐻 = 𝐻 𝑔 for all 𝑔 ∈ 𝐺 ;a Lie group is called simple if it possess no non-trivial connected normal subgroups.A representation 𝑉 of 𝐺 , where 𝑉 is a finite-dimensional vector space over a field K = R or C , can be thought of as a map 𝐺 × 𝑉 → 𝑉 such that for all 𝑔 ∈ 𝐺 , 𝑣 ∈ 𝑉 , we have • 𝑒𝑣 = 𝑣 and 𝑔 ( 𝑔 (cid:48) 𝑣 ) = ( 𝑔 𝑔 (cid:48) ) 𝑣 , and • 𝑔𝑣 is a continuous function of 𝑔 and 𝑣 , which is additionally K -linear in 𝑣 .By choosing a basis for 𝑉 , we get isomorphisms 𝑉 (cid:27) K 𝑛 for some integer 𝑛 andEnd ( 𝑉 ) (cid:27) 𝑀 𝑛 ( K ) = { 𝑛 × 𝑛 matrices with entries in K } . A linear subspace 𝑈 ⊂ 𝑉 is called 𝐺 -invariant if 𝑔𝑢 ∈ 𝑈 for all 𝑢 ∈ 𝑈 . A representation 𝑉 is said to be irreducible if its only 𝐺 -invariant subspaces are the trivial ones, 0 and 𝑉 ; else,it is said to be reducible . Irreducible representations will be fundamental in what follows;for brevity, we will call them irreps.The general linear group of 𝑉 ,GL ( 𝑉 ) = Aut ( 𝑉 ) = { 𝐴 ∈ End ( 𝑉 ) | ∃ 𝐴 − ∈ End ( 𝑉 )} , is a group which is an open subset of End ( 𝑉 ) , and hence a smooth manifold. The productand inverse map in GL ( 𝑉 ) are smooth, so GL ( 𝑉 ) is a Lie group. If the dimension of 𝑉 over K is 𝑛 , we will write GL ( 𝑛, K ) for GL ( 𝑉 ) .We can choose on 𝑉 a Hermitian form (cid:104) , (cid:105) such that for 𝑣, 𝑤 ∈ 𝑉 , 𝜆 ∈ K • (cid:104) 𝑣, 𝑤 𝜆 (cid:105) = (cid:104) 𝑣, 𝑤 (cid:105) 𝜆 , and • (cid:104) 𝑤, 𝑣 (cid:105) = (cid:104) 𝑣, 𝑤 (cid:105) ;this form then defines taking the conjugate transpose via 𝑥 † 𝑦 : = (cid:104) 𝑥, 𝑦 (cid:105) . The subgroups { 𝐴 ∈ GL ( 𝑛, K ) | 𝐴 † 𝐴 = Id } are denoted O ( 𝑛 ) , U ( 𝑛 ) , Sp ( 𝑛 ) respectively for K = R , C , H .For K = R or C , we also get the special linear subgroups { 𝐴 ∈ GL ( 𝑛, K ) | det 𝐴 = } denoted SL ( 𝑛, R ) and SL ( 𝑛, C ) . Finally, we have SO ( 𝑛 ) = SL ( 𝑛, R ) ∩ O ( 𝑛 ) and SU ( 𝑛 ) = SL ( 𝑛, C ) ∩ U ( 𝑛 ) . All these groups are collectively called classical groups , and are in fact Liegroups. The groups O ( 𝑛 ) , SO ( 𝑛 ) , U ( 𝑛 ) , SU ( 𝑛 ) , Sp ( 𝑛 ) are all compact, while GL ( 𝑛, K ) andSL ( 𝑛, K ) ( 𝑛 ≠
1) are not.Since Lie groups possess the structure of a manifold, it is sensible to talk about tangentvectors; if the Lie group also happens to be modelled on some vector space 𝑉 , as theclassical groups are, then the tangent spaces at each point, 𝑇 𝑝 𝐺 will in fact be isomorphicto 𝑉 . More will be said once we make the following . 1 . P R E L I M I NA R I E S 3 Definition 1.1.1 (Lie Algebra) . A Lie algebra is a vector space 𝑉 over a field K equippedwith an operation [ , ] : 𝑉 × 𝑉 → 𝑉 called a Lie bracket , which satisfies the followingaxioms. • Bilinearity: [ 𝑎𝑥 + 𝑏 𝑦, 𝑧 ] = 𝑎 [ 𝑥, 𝑧 ] + 𝑏 [ 𝑦, 𝑧 ] for all scalars 𝑎, 𝑏 ∈ K and vectors 𝑥, 𝑦, 𝑧 ∈ 𝑉 . • Alternativity: [ 𝑥, 𝑥 ] = 𝑥 ∈ 𝑉 . • The Jacobi identity: [ 𝑥, [ 𝑦, 𝑧 ]] + [ 𝑧, [ 𝑥, 𝑦 ]] + [ 𝑦, [ 𝑧, 𝑥 ]] = 𝑥, 𝑦, 𝑧 ∈ 𝑉 .We will follow the standard convention of denoting the Lie algebra of a group by the sameletters in lower-case Fraktur font, e.g. 𝑇 𝑒 𝐺 = : 𝔤 . Proposition 1.1.2.
For 𝐺 = O ( 𝑛 ) , U ( 𝑛 ) or Sp ( 𝑛 ) , a matrix 𝐴 ∈ 𝔤 if and only if 𝐴 † + 𝐴 = . Forthe groups 𝐺 = SL ( 𝑛, R ) , SL ( 𝑛, C ) , 𝐴 ∈ 𝔤 if and only if tr 𝐴 , the trace of 𝐴 vanishes. Theorem 1.1.3. (i) For any Lie group 𝐺 , the space 𝑇 𝑒 𝐺 of tangent vectors to the identity element is a Lie algebraover K .(ii) For any matrix group 𝐺 ⊂ GL ( 𝑉 ) , the Lie bracket is given by the commutator, [ 𝑋 , 𝑌 ] = 𝑋𝑌 − 𝑌𝑋 .(iii) If 𝜃 : 𝐺 → 𝐻 is a homomorphism of Lie groups, the induced map d 𝜃 : 𝔤 → 𝔥 is ahomomorphism of Lie algebras.(iv) For any representation 𝑉 of 𝐺 , 𝑉 is a representation of 𝔤 .(v) For a matrix group 𝐺 acting on 𝑉 as the endomorphism group, 𝔤 acts in the same way.(vi) If 𝐻 acts on 𝑉 and we are given a homomorphism 𝜃 : 𝐺 → 𝐻 of Lie groups so that 𝐺 actson 𝑉 , the resulting action of 𝔤 on 𝑉 is 𝑋 · 𝑣 = ( d 𝜃 ( 𝑋 )) · 𝑣 , where 𝑋 ∈ 𝔤 . The proof of these statements can be found in any of the references listed at thebeginning of this section. We note that in item (iv) above, a representation of a Lie algebrais defined in the obvious way: a map from 𝔤 × 𝑉 → 𝑉 that is linear for 𝑣 ∈ 𝑉 and respectsthe Lie bracket, [ 𝑋 , 𝑌 ] 𝑣 = 𝑋 ( 𝑌𝑣 ) − 𝑌 ( 𝑋𝑣 ) .The final definition in this section is of significant importance to us, and it goes asfollows. 𝐺 acts on itself by conjugation, 𝑐 𝑔 : 𝐺 → 𝐺 , ℎ ↦→ 𝑔 ℎ 𝑔 − , and this is clearly ahomomorphism. We hence obtain by item (iii) in the theorem above, for each 𝑔 ∈ 𝐺 , acorresponding Lie algebra automorphism d 𝑐 𝑔 : 𝔤 → 𝔤 . This is the adjoint representation of 𝐺 on its Lie algebra, Ad : 𝐺 → Aut 𝔤 . The differential of this representation gives the adjoint representation of the Lie algebra on itself; this map is ad : 𝔤 → GL ( 𝔤 ) , ad 𝑋 ( 𝑌 ) = [ 𝑋 , 𝑌 ] . In the mathematics of particle physics, all vector spaces must be complex because offoundational axioms in quantum mechanics. With this in mind, the two fundamentalprinciples of the representation theory of particle physics can be stated in a few words:given a gauge symmetry 𝐺 of a theory, the fermions (matter fields) of this theory are basisvectors of unitary irreps of 𝐺 , while the gauge bosons, which mediate forces, are basisvectors of its complexified adjoint representation (which is irreducible if 𝐺 is simple). In See the Dirac-von Neumann axioms [24, 86]. 1 . T H E S TA N DA R D M O D E L this section, we will try to motivate these claims, assuming some prior knowledge of gaugetheory, the arena in which classical field theory plays out.For a field theory on a spacetime, the
Lagrangian is the name given to a function thatdescribes the dynamics and all the interactions of the fields. There are a few core principlesthat one must follow when attempting to write down such a function, of which we willonly consider symmetry : the Lagrangian, and hence the laws of physics, should be invariantunder the transformations of some specified symmetry group 𝐺 (such as the Poincarégroup, which encodes the symmetry of Minkowski spacetime). It is clear at the outset thatthe fields (particles) out of which the Lagrangian is built must live in irreps of 𝐺 , to keepthe invariance under 𝐺 manifest. The unitarity requirement on the irreps then arises quitenaturally from the desire to compute observables (matrix elements). For example, a state | 𝜓 (cid:105) ↦→ 𝑃 | 𝜓 (cid:105) under a Poincaré transformation 𝑃 , so an observable would transform as (cid:104) 𝜓 | 𝜓 (cid:105) = 𝑀 𝑀 = (cid:104) 𝜓 | 𝑃 † 𝑃 | 𝜓 (cid:105) . 𝑃 Therefore, we need 𝑃 † 𝑃 = Id, i.e. 𝑃 must be unitary, if the matrix element is to be invariant. It is more involved to see why the gauge bosons live the adjoint representation; webegin with a plausibility argument from physics known as minimal coupling . Considera matter field 𝜓 ( 𝑥 ) : for local gauge transformations, 𝜓 ( 𝑥 ) ↦→ 𝑔 ( 𝑥 ) 𝜓 ( 𝑥 ) , and ordinaryderivatives transform as 𝜕 𝜇 𝜓 ( 𝑥 ) ↦→ 𝑔 ( 𝑥 ) 𝜕 𝜇 𝜓 ( 𝑥 ) + ( 𝜕 𝜇 𝑔 ( 𝑥 )) 𝜓 ( 𝑥 ) , that is, inhomogeneously. What we would like instead is a gauge-covariant derivative D 𝜇 ,which transforms as D 𝜇 𝜓 ( 𝑥 ) ↦→ 𝑔 ( 𝑥 ) D 𝜇 𝜓 ( 𝑥 ) . To achieve this, we defineD 𝜇 𝜓 ( 𝑥 ) = 𝜕 𝜇 𝜓 ( 𝑥 ) − 𝐴 𝜇 𝜓 ( 𝑥 ) , where 𝐴 𝜇 is a Lie algebra-valued 1-form written in a local basis. As the notation suggests,this is our gauge boson, and it is forced to have the transformation law 𝐴 𝜇 ↦→ 𝐴 (cid:48) 𝜇 ( 𝑥 ) = 𝑔 ( 𝑥 ) 𝐴 𝜇 𝑔 − ( 𝑥 ) + ( 𝜕 𝜇 𝑔 ( 𝑥 )) 𝑔 − ( 𝑥 ) . (1.1.1)The first term is the desired adjoint transformation of 𝐴 𝜇 . The second term vanishes if 𝑔 ( 𝑥 ) is taken to be constant locally; this is referred to as a rigid gauge transformation . Thegroup of rigid physical transformations form a group isomorphic to 𝐺 .By way of motivation, the above should suffice (and is often the last word in physicstextbooks). But let us go deeper. In mathematics, matter fields like 𝜓 ( 𝑥 ) are sections of thevector bundle 𝑃 × 𝜌 𝑉 → 𝑀 associated to a 𝐺 -principal bundle 𝑃 → 𝑀 by a representation 𝜌 : 𝐺 → GL ( 𝑉 ) . The derivative D 𝜇 above is usually denoted ∇ 𝜇 in mathematics, and is infact the covariant derivative on the vector bundle associated to a connection 1-form 𝐴 onthe principal bundle; in coordinate-free notation, for a vector field 𝑋 ∈ 𝔛 ( 𝑀 ) , it outputs asection ∇ 𝐴𝑋 𝜓 = d 𝜓 ( 𝑋 ) + d 𝜌 ( 𝐴 𝑠 ( 𝑋 )) 𝜓 , where 𝐴 𝑠 = 𝐴 ◦ D 𝑠 ∈ Ω ( 𝑈 , 𝔤 ) is the local gauge field for 𝑠 : 𝑀 ⊃ 𝑈 → 𝑃 a choice oflocal gauge, and d 𝜌 : 𝔤 → End ( 𝑉 ) is the induced Lie algebra representation. A cluethat something more needs to be said about the somewhat ad hoc imposition of thetransformation law (1.1.1) for 𝐴 can be found in the fact that in the parlance of gauge theory, We used the Poincaré group here since it is a natural example, but we will not be concerned with itin what follows. The reason for this is that the unitary irreps of this group are all infinite dimensional, asproved by Wigner in 1939 [91], and we wish to restrict to finite-dimensional representation theory. See [79,pp. 109-103] for further discussion. . 2 . T H E F U N DA M E N TA L F O RC E S 5 the corresponding result is purely a statement about what happens when we change gaugefrom 𝑠 𝑖 : 𝑈 𝑖 → 𝑃 to 𝑠 𝑗 : 𝑈 𝑗 → 𝑃 on the principal bundle , and has nothing whatsoever todo with the associated vector bundle, where all the physics takes place. Instead, what isneeded is the following: one can show that the difference between two connections is in facta section (field) on the base manifold 𝑀 with values in the vector bundle Ad ( 𝑃 ) : = 𝑃 × Ad 𝔤 ,associated to 𝑃 → 𝑀 by the adjoint representation of 𝐺 . (Heuristically, one can see thisfrom equation (1.1.1) by noting that the difference of two transformed gauge fields killsthe ( 𝜕 𝜇 𝑔 ( 𝑥 )) 𝑔 − ( 𝑥 ) term.) Naturally, sections of this bundle then transform in the adjointrepresentation, as desired. A rather beautiful physical interpretation of this result is foundin [44]: in quantum field theory, particles in general are described as excitations of a givenvacuum field; in the case of a gauge field, one has to declare the vacuum field to be a certainspecific connection 1-form 𝐴 on the principal bundle, with reference to which all othergauge fields would then by described ; by the result stated in the previous paragraph, thisdifference (excitation) 𝐴 − 𝐴 can then be identified with a 1-form on the spacetime 𝑀 ,with values in Ad ( 𝑃 ) that hence transforms in the adjoint representation of 𝐺 .Now that we have the transformation rules for all the particles in our theory, a naturalquestion arises: how can the gauge bosons be said to “mediate forces”? The mathematicalmechanism is in fact quite straightforward. When we say that a force is invariant under theaction of some group, this corresponds to the statement that any physical process causedby this force should be described by an “intertwining operator”, which is a linear operatorthat respects the action of the group under consideration. More precisely, suppose that 𝑉 and 𝑊 are finite-dimensional Hilbert spaces on which some group 𝐺 acts as unitaryoperators. Then a linear operator 𝐹 : 𝑉 → 𝑊 is an intertwining operator if 𝐹 ( 𝑔 𝜓 ) = 𝑔𝐹 ( 𝜓 ) forevery 𝜓 ∈ 𝑉 and 𝑔 ∈ 𝐺 . Now we saw in theorem 1.1.3 that a representation 𝜌 : 𝐺 × 𝑉 → 𝑉 of a group 𝐺 gives rise to a representation of its Lie algebra 𝔤 on 𝑉 ; we think of this as thelinear map d 𝜌 : 𝔤 ⊗ 𝑉 → 𝑉 . It is easy to check that this map is an intertwining operator,and it hence gives the gauge bosons agency to act on particles. We begin our brief exposition of the Standard Model proper with the representation theoryof quantum chromodynamics (QCD), since it is the most straightforward application of theprinciples that we encountered in the previous section. Many great minds contributed tothe development of this theory , but it was the trio of Fritzsch, Gell-Mann and Leutwylerwho formulated the concept of colour as the source of a “strong field” in a Yang-Millstheory in 1973 [29].This will be followed by a description of the weak force, which will then be expandedto include electromagnetism in the section on the electroweak interaction; this milestonein the history of unification was due to independent work by Glashow [38], Salam [78]and Weinberg [88], for which they were jointly awarded the Nobel prize in 1979. We willfollow the article of Baez and Huerta [8] in this section, and in the one following, on thefermion representation of the Standard Model. Let us begin with the nucleons of high-school chemistry, the protons and neutrons. It turnsout that they are not fundamental particles, but are instead made up of other particlescalled quarks , which come in a number of different flavours . It takes two flavours to make See [44, Theorem 5.25]. N.b. the form 𝐴 ≡ not a connection. Reference [21, Ch. 4] has an account of the history. 1 . T H E S TA N DA R D M O D E L protons and neutrons, the up quark 𝑢 , and the down quark 𝑑 : the proton can be written as 𝑝 = 𝑢𝑢𝑑 , and the neutron, 𝑛 = 𝑢𝑑𝑑 (the notation will be clarified momentarily). It followsfrom the charges of the proton (+1) and neutron (0) that 𝑢 has a charge 2 /
3, and 𝑑 , − / confinement is one of two defining characteristics of QCD, the other being asymptotic freedom . The latter is unfortunately outside the scope of this paper; we referthe reader to a review article by Gross, one of the discoverers of aymptotic freedom[39]. Now quark confinement, loosely speaking, is the statement that the force betweenquarks does not diminish as they are separated; thus, they are forever bound into hadrons such as the proton and the neutron. Let us try to understand this mathematically. Eachflavour of quark comes in three different states called colours : red ( 𝑟 ), green ( 𝑔 ), and blue ( 𝑏 ).This means that the Hilbert space for a single quark is C , with 𝑟 , 𝑔 , and 𝑏 the standardbasis vectors. The colour symmetry group SU ( ) acts on C in the obvious way, via itsfundamental representation. Since both up and down quarks come in three colour states,there are really six kinds of quarks in matter: three up quarks, spanning a copy of C ; 𝑢 𝑟 , 𝑢 𝑏 , 𝑢 𝑔 ∈ C , and similarly for down quarks. The group SU ( ) acts on each space. Allsix quarks taken together span the vector space C ⊕ C (cid:27) C ⊗ C , where C is spannedby 𝑢 and 𝑑 . Confinement amounts to the following decree: all observed states must bewhite, i.e. invariant under the action of SU ( ) . Hence, we can never see an individualquark, nor particles made from two quarks, because there are no vectors in C or C ⊗ C which transform trivially under SU ( ) . But we do see see particles made up of threequarks, such as nucleons, because there are unit vectors in C ⊗ C ⊗ C fixed by SU ( ) .Indeed, as a representation of SU ( ) , C ⊗ C ⊗ C contains precisely one copy of the trivialrepresentation: the antisymmetric rank-three tensors, Λ C . This one-dimensional vectorspace is spanned by the wedge product of all three basis vectors, 𝑟 ∧ 𝑏 ∧ 𝑔 ∈ Λ C , so upto normalisation, this must be colour state of a nucleon. We also now see that the colourterminology is well-chosen, since an equal mixture of red, green, and blue light is white.Hence, confinement is intimately related to colour. An explanation of the quark flavours ispostponed until the next section.We will have much more to say about the quarks, but as an introduction, what we haveabove suffices: the strong force is concerned with the quarks; the up and the down quarkstogether span the representation C ⊗ C of SU ( ) , where C is trivial under SU ( ) . In theprevious section, we took the trouble to understand how gauge bosons transform andact, and we now we reap the fruits of that labour: from the standpoint of representationtheory, all there is to say is that strong force is mediated by the gluons , usually denotedby 𝑔 , which live in C ⊗ 𝔰𝔲 ( ) = 𝔰𝔩 ( , C ) , the complexified adjoint representation of SU ( ) .They act on quarks via the standard action of 𝔰𝔩 ( , C ) on C . Our story of the weak force begins, interestingly enough, in early attempts to describe thestrong force, particularly in the work of Heisenberg in 1932 [49]. He hypothesised thatthe proton and nucleon were the two possible observed states of a nucleon; a nucleonwould hence live in the simplest Hilbert space possible for such a setup: C = C ⊕ C .Shortly thereafer, in 1936, Cassen and Condon [20] suggested that the C space of nucleonsis acted upon by SU ( ) , emphasising the analogy with the spin of an electron, which isalso described by vectors in C acted upon by SU ( ) . The property that distinguishes theproton from the neutron was hence dubbed isospin : the proton was declared to be isospinup, 𝐼 = + /
2, and the neutron isospin down, 𝐼 = − /
2. The charge and the isospin of thenucleons 𝑁 were seen to be related in the following simple way: 𝑄 ( 𝑁 ) = 𝐼 ( 𝑁 ) + . . 2 . T H E F U N DA M E N TA L F O RC E S 7 This turned out to be a special instance of what came to be called the
Gell-Mann–Nishijimaformula (abbreviated as the NNG formula): 𝑄 = 𝐼 + 𝑌 .𝑌 is a quantity called hypercharge which depends only on the “family” of the particle. Forthe moment, this simply means that 𝑌 is required to be constant on representations of theisospin symmetry group, SU ( ) . To now understand how all of this relates to the moderntheory of weak interactions, we have to introduce a new particle.Along with the electron 𝑒 − and the up and down quarks, the neutrinos 𝜈 form the firstgeneration of fundamental fermions. They carry no charge and no colour, and interactonly through the weak force, first proposed by Enrico Fermi in 1933 [26]. The weak forceis chiral , i.e. it cares about the handedness of particles: every particle thus far discussedcomes in left- and right-handed varieties, which we will denote by subscript- 𝐿 and - 𝑅 respectively. Remarkably, the weak force interacts only with the left-handed particles, andright-handed antiparticles . We have been silent about antiparticles until now, but they arequite simple to introduce: to each particle, there is a corresponding antiparticle, which isjust like the original particle, but with opposite charge and isospin; mathematically, thisjust means that we pass to the dual representation. Returning to the weak interaction,when the neutron decays for example, we always have 𝑛 𝐿 → 𝑝 𝐿 + 𝑒 − 𝐿 + 𝜈 𝑅 , and never 𝑛 𝑅 → 𝑝 𝑅 + 𝑒 − 𝑅 + 𝜈 𝐿 . This parity violation of the weak force, proposed by Yang and Lee in 1956 [58] is stillstartling; no other physics, classical or quantum, looks different when viewed in a mirror.One important corollary of this oddity is that the right-handed neutrino 𝜈 𝑅 has never beenobserved directly; we will discuss this particle in the context of grand unified theories insections 2.5 and 4.3.The isospin mentioned above is an extremely useful quantity since it is conservedduring quantum interactions; as such, we would like to extend it to weak interactions.First, for the proton and neutron to have the right isospins of ± , we must have the isospinof the up and down quarks defined to be ± respectively (making these particles the upand down states at which their names hint). A quick check then shows that isospin is notautomatically conserved in weak interactions; for example, in the above neutron decay, 𝑢 𝐿 𝑑 𝐿 𝑑 𝐿 → 𝑢 𝐿 𝑢 𝐿 𝑑 𝐿 + 𝑒 − 𝐿 + 𝜈 𝑅 , the right-hand side has 𝐼 = − / 𝐼 = /
2. What is neededis an extension of the concept of isospin to the leptons , i.e. the particles which do not feelthe strong force, 𝑒 − and 𝜈 ; simply setting 𝐼 ( 𝜈 𝐿 ) = and 𝐼 ( 𝑒 − 𝐿 ) = − does the trick. Thisextension of isospin is called weak isospin , and unlike the isospin of the nucleons, is anexact symmetry. We will simply refer to it as isospin from now on.We come to the description of the weak force. This is a theory with the isospinsymmetry group SU ( ) ; the particles in the same representation are paired up in doublets , (cid:18) 𝜈 𝐿 𝑒 − 𝐿 (cid:19) , (cid:18) 𝑢 𝐿 𝑑 𝐿 (cid:19) , with the particle with the higher 𝐼 on the top; this is just a shorthand way of writingthat these particles live in (and span) the same irrep C of SU ( ) . The fact that only the left-handed particles are combined into doublets reflects the fact that only they participatein weak interactions. Every right-handed fermion, on the other hand, is trivial underSU ( ) : they are called singlets , and span the trivial representation C .The particles in the doublets interact via the exchange of the so-called 𝑊 bosons, 𝑊 + = (cid:18) (cid:19) , 𝑊 = (cid:18) − (cid:19) , 𝑊 − = (cid:18) (cid:19) . As we would expect, these span the complexified adjoint representation of SU ( ) , 𝔰𝔩 ( , C ) ,and they act on each of the particles in the doublets via the action of 𝔰𝔩 ( , C ) .We close this section with the afore-promised explanation of quark flavour splitting.Recall that colour is related to confinement; in much the same way, flavour is relatedto isospin. Indeed, we can use quarks to explain the isospin symmetry of the nucleons:protons and neutrons are so similar, with nearly the same mass and strong interactions,because 𝑢 and 𝑑 quarks are so similar, with nearly the same mass, and truly identicalcolours. As mentioned above, the isospin of the proton and neutron arises from the isospinof the quarks, once we define 𝐼 ( 𝑢 ) = /
2, and 𝐼 ( 𝑑 ) = − /
2; we see that the proton obtainsthe right 𝐼 : 𝐼 ( 𝑝 ) = + − = , and a quick check shows the same for the neutron. This is a good start, but what we reallyneed to do is to confirm that 𝑝 and 𝑛 span a copy of the fundamental representation C of SU ( ) . It turns out that the states 𝑢 ⊗ 𝑢 ⊗ 𝑑 and 𝑢 ⊗ 𝑑 ⊗ 𝑑 do not span a copy of thefundamental representation of SU ( ) inside C ⊗ C ⊗ C ; what is needed, for the protonfor instance, is some linear combination of the 𝐼 = / 𝑢 ’s and one 𝑑 : 𝑢 ⊗ 𝑢 ⊗ 𝑑 , 𝑢 ⊗ 𝑑 ⊗ 𝑢 , 𝑑 ⊗ 𝑢 ⊗ 𝑢 ∈ C ⊗ C ⊗ C . The exact linear combination required to make 𝑝 and 𝑛 work also involves the spin ofthe quarks, which is outside the scope of our discussion. What we can do however,is see that this is at least possible, i.e. that C ⊗ C ⊗ C really does contain a copy ofthe fundamental representation C of SU ( ) . First note that any rank-2 tensor can bedecomposed into symmetric and antisymmetric parts, C ⊗ C (cid:27) Sym C ⊕ Λ C . NowSym C is the unique 3-dimensional irrep of SU ( ) , and Λ C , as the top exterior power ofits fundamental representation C , is the trivial 1-dimensional irrep; as a representation ofSU ( ) , we therefore have C ⊗ C ⊗ C (cid:27) C ⊗ ( Sym C ⊕ C ) (cid:27) ( C ⊗ Sym C ) ⊕ C . So indeed, C is a subrepresentation of C ⊗ C ⊗ C . As a final remark, we note that theNNG formula still works for quarks, provided we define the hypercharge for both quarksto be 𝑌 = / All the fermions have now been grouped into SU ( ) representations based on their isospin.Let us now consider the other piece of NNG formula, hypercharge. Just as we did forisospin, we can extend the notion of hypercharge to encompass the leptons, calling thisnew quantity weak hypercharge . It is a matter of simple arithmetic to see that we must have 𝑌 = − 𝐼 =
0, we must set 𝑌 = 𝑄 .How can we understand hypercharge? Let us frame the discussion by briefly discussingisospin again: it is an observable, and we know from quantum mechanics that it hence . 2 . T H E F U N DA M E N TA L F O RC E S 9 corresponds to a self-adjoint operator; indeed, from an eigenvalue expression like ˆ 𝐼 𝜈 𝐿 = 𝜈 𝐿 , it is easy to see that we must have ˆ 𝐼 = (cid:18) / − / (cid:19) . The story with hypercharge is similar: corresponding to hypercharge 𝑌 is an observable ˆ 𝑌 , which is also proportional to a gauge boson, although this gauge boson lives in thecomplexified adjoint representation of U ( ) .The details are as follows. Particles with hypercharge 𝑌 span irreps C 𝑌 of U ( ) ; by C 𝑌 we denote the one-dimensional vector space C with action of 𝛼 ∈ U ( ) given by 𝛼 · 𝑧 = 𝛼 𝑌 𝑧 . The factor of three is inserted because 𝑌 is not guaranteed to be an integer, but onlyan integer multiple of 1 /
3. For example, the left-handed leptons 𝜈 𝐿 and 𝑒 − 𝐿 both havehypercharge 𝑌 = −
1, so each span a copy of C − . Hence, 𝜈 𝐿 , 𝑒 − 𝐿 ∈ C − ⊗ C , where the C istrivial under U ( ) .Now, given a particle 𝜓 ∈ C 𝑌 , to find out how the gauge boson in C ⊗ 𝔲 ( ) (cid:27) C acts onit, we can differentiate the U ( ) action above. We obtain 𝑖 · 𝜓 = 𝑖𝑌 𝜓 = ⇒ ˆ 𝑌 = ∈ C . Following convention, we set the so-called 𝐵 boson equal to ˆ 𝑌 ; particles with hyperchargeinteract by exchanging this boson. Note that the 𝐵 boson is a lot like the familiar photon,and the hypercharge force which 𝐵 mediates is a lot like electromagnetism, except that itsstrength is proportional to hypercharge rather than charge.The unification of electromagnetism and the weak force is called the electroweakinteraction . This is a U ( ) × SU ( ) gauge theory, and we have now encountered it in fulldetail: the fermions live in representations of hypercharge U ( ) and weak isospin SU ( ) ,and we tensor these together to get representations of U ( ) × SU ( ) . These fermionsinteract by exchanging 𝐵 and 𝑊 bosons, which span C ⊕ 𝔰𝔩 ( , C ) , the complexified adjointrepresentation of U ( ) × SU ( ) .We close with a word on symmetry breaking . Despite electroweak unification, elec-tromagnetism and the weak force are very different at low energies, including mostinteractions in the everyday world. Electromagnetism is a force of infinite range that wecan describe by a U ( ) gauge theory with the photon as gauge boson, while the weak forceis of very short range and mediated by the 𝑊 and 𝑍 bosons: we “define” the photon andthe 𝑍 boson by the following relation: (cid:18) 𝛾 𝑍 (cid:19) = (cid:18) cos 𝜃 w sin 𝜃 w − sin 𝜃 w cos 𝜃 w (cid:19) (cid:18) 𝐵𝑊 (cid:19) . (1.2.1)We have introduced here the weak mixing angle , or Weinberg angle 𝜃 w ; it can be thought ofas the parameter that characterises how far the 𝐵 − 𝑊 boson plane has been rotated bysymmetry breaking, which is the mechanism that allows the full electroweak U ( ) × SU ( ) symmetry group to be hidden away at low energies, and replaced with the electromagneticsubgroup U ( ) . Moreover, the electromagnetic U ( ) is not the obvious factor U ( ) ×
1, butanother copy, wrapped around inside U ( ) × SU ( ) in a manner given by the NNG formula.Unfortunately, the dynamics of electroweak symmetry breaking is outside of our scope;we refer the reader to [79, Ch. 29.1] for the details. We will discuss symmetry breakingfrom a representation theoretic viewpoint in section 3.3, and return to the Weinberg anglein section 4.1. Since U ( ) is abelian, all of its irreps are one-dimensional.0 1 . T H E S TA N DA R D M O D E L We are now in a position to put the whole Standard Model together in a single picture. Ithas the gauge group 𝐺 SM = U ( ) × SU ( ) × SU ( ) , and the fundamental fermions described thus far combine in representations of this group.We summarise this information in the table below.Table 1.1: The Standard Model FermionsParticle Name Symbol U ( ) × SU ( ) × SU ( ) Rep.Left-handed leptons (cid:169)(cid:173)(cid:173)(cid:171) 𝜈 𝐿 𝑒 − 𝐿 (cid:170)(cid:174)(cid:174)(cid:172) C − ⊗ C ⊗ C Left-handed quarks (cid:169)(cid:173)(cid:173)(cid:171) 𝑢 𝑟𝐿 , 𝑢 𝑔𝐿 , 𝑢 𝑏𝐿 𝑑 𝑟𝐿 , 𝑑 𝑔𝐿 , 𝑑 𝑏𝐿 (cid:170)(cid:174)(cid:174)(cid:172) C / ⊗ C ⊗ C Right-handed neutrino 𝜈 𝑅 C ⊗ C ⊗ C Right-handed electron 𝑒 − 𝑅 C − ⊗ C ⊗ C Right-handed up quarks (cid:0) 𝑢 𝑟𝑅 , 𝑢 𝑏𝑅 , 𝑢 𝑔𝑅 (cid:1) C / ⊗ C ⊗ C Right-handed down quarks (cid:0) 𝑑 𝑟𝑅 , 𝑑 𝑏𝑅 , 𝑑 𝑔𝑅 (cid:1) C − / ⊗ C ⊗ C All the representations of 𝐺 SM in the right-hand column are irreducible, since they aremade by tensoring irreps of this group’s three factors. On the other hand, if we take thedirect sum of all these irreps, 𝐹 = ( C − ⊗ C ⊗ C ) ⊕ · · · ⊕ ( C − / ⊗ C ⊗ C ) , we get a reducible representation containing all the first-generation fermions in theStandard Model. We call 𝐹 the fermion representation . If we take the dual of 𝐹 , we get arepresentation describing all the antifermions in the first generation. Taking the directsum of these spaces, 𝐹 ⊕ 𝐹 , we get a representation of 𝐺 SM that we will call the StandardModel representation ; it contains all the first-generation elementary particles in the StandardModel. The fermions interact by exchanging gauge bosons that live in the complexifiedadjoint representation of 𝐺 SM .Table 1.2: The Standard Model Gauge BosonsForce Gauge Boson SymbolElectromagnetism Photon 𝛾 Weak Force 𝑊 and 𝑍 bosons 𝑊 + , 𝑊 − and 𝑍 Strong Force Gluons 𝑔 . 3 . T H E S TA N DA R D M O D E L R E P R E S E N TAT I O N 11 Generations
For the purposes of describing grand unified theories, the above description of the StandardModel is all we need. For completeness however, we tabulate below the second and third generations of fermions, evidence of which first arose in the 1930s; an elegant summary ofthe physics can be found in [46].Table 1.3: Quarks by Generation1st Generation 2nd Generation 3rd GenerationName Symbol Name Symbol Name SymbolUp 𝑢 Charm 𝑐 Top 𝑡 Down 𝑑 Strange 𝑠 Bottom 𝑏 Table 1.4: Leptons by Generation1st Generation 2nd Generation 3rd GenerationName Symbol Name Symbol Name SymbolElectronneutrino 𝜈 𝑒 Muonneutrino 𝜈 𝜇 Tauneutrino 𝜈 𝜏 Electron 𝑒 − Muon 𝜇 − Tau 𝜏 − Notice that we thus have a pattern in the Standard Model: there are as many flavoursof quarks as there are of leptons. The Pati-Salam model explains this pattern by unifyingquarks and leptons, but we will unfortunately not treat this theory here; the interestedreader is referred to [8, Ch. 3.3].The second and third generations of quarks and charged leptons differ from the firstby being more massive and able to decay into particles of the earlier generations. Thevarious neutrinos do not decay, and for a long time it was thought they were massless,but now it is known that some and perhaps all are massive. This allows them to changeback and forth from one type to another, a phenomenon called neutrino oscillation ; theStandard Model explain this phenomenon by recourse to the famous “Higgs mechanism” .For our purposes however, the generations are identical: as representations of 𝐺 SM , eachgeneration spans another copy of 𝐹 , with the corresponding generation of antifermionsspanning a copy of 𝐹 . All told, we thus have three copies of the SM representation, 𝐹 ⊕ 𝐹 .We will only need to discuss one generation, so we find it convenient to speak as if 𝐹 ⊕ 𝐹 contains particles of the first generation. This redundancy in the Standard Model, threesets of very similar particles, remains a mystery. See the article [15] for a brief review of the theoretical and experimental aspects. See [45] and [79, Ch. 28]. hapter 2
The
Spin ( ) Grand Unified Theory
Due to spontaneous symmetry breaking, not all of the symmetries of the Standard Modelare seen in everyday life—the symmetries encoded by 𝐺 SM are symmetries of the laws ofphysics, but not necessarily of the vacuum. Grand unified theories attempt to answer thequestion, what if this process continues? That is, could the symmetries of the StandardModel be just a subset of all the symmetries in nature? By way of motivation, considerthat from a representation theoretic standpoint alone, the Standard Model leaves a lot tobe desired: “The representations of 𝐺 SM seem ad hoc. Why these? What about all theseemingly arbitrary hypercharges? Why do both leptons and quarks come in left- andright- handed varieties, which transform so differently? Why do quarks come in chargeswhich are units 1 / ( ) grand unified theory. Thereafter, we will focus ourattention on Clifford algebras; it is a short step from there to the Spin groups; once wethen understand their representations, we will prove that Spin ( ) extends the StandardModel in section 2.5. The study of characters, and root and weight systems, is fundamental to representationtheory. Our modest goals in this section of simply defining these terms and stating themain results will doubtless do a severe injustice to this branch of mathematics; we point to[43, Ch. 8] for a lucid introduction and additional references. We will follow [4] here.Particle physics demands that we restrict to complex representations, so let us do soright at the outset, reaping the added benefit that over C , every irrep of a compact abeliangroup is 1-dimensional. Remark 2.1.1 (Structure Maps) . We note that this restriction involves no sacrifice of gener-ality: consider that a representation 𝑉 over the quanterions H is certainly a representationover C ; together with a conjugate linear structure 𝐺 -map 𝑗 : 𝑉 → 𝑉 such that 𝑗 = − 𝑖𝑗 = − 𝑗𝑖 , this representation does in fact return the original H -representation; on the otherhand, a representation 𝑉 over R gives 𝑉 ⊗ R C and this carries a conjugate linear structuremap 𝑗 : 𝑣 ⊗ 𝑧 ↦→ 𝑣 ⊗ 𝑧 such that 𝑗 =
1; we can regard 𝑉 as the +1 eigenspace of 𝑗 (or the -1eigenspace).
134 2 . T H E S P I N ( ) G R A N D U N I F I E D T H E O RY
Definition 2.1.2 (Characters) . Suppose 𝑉 is a representation of 𝐺 over C . Then its characteris given by 𝜒 𝑉 : 𝐺 C ,𝑔 tr C ( 𝑔 : 𝑉 → 𝑉 ) . It follows from the definition that characters are class functions , i.e. 𝜒 𝑉 ( 𝑔 ℎ 𝑔 − ) = 𝜒 𝑉 ( ℎ ) for all 𝑔, ℎ ∈ 𝐺 . Also, 𝜒 𝑉 ⊕ 𝑊 ( 𝑔 ) = 𝜒 𝑉 ( 𝑔 ) + 𝜒 𝑊 ( 𝑔 ) , 𝜒 𝑉 ⊗ 𝑊 ( 𝑔 ) = 𝜒 𝑉 ( 𝑔 ) · 𝜒 𝑊 ( 𝑔 ) . The following result clarifies the importance of characters.
Theorem 2.1.3. If 𝜒 𝑉 = 𝜒 𝑊 , then 𝑉 (cid:27) 𝑊 . A proof is found in [3, pp. 46–52]. Consider now the torus, 𝑇 = (cid:206) 𝑆 . Because 𝑇 is acompact, connected abelian group, the exponential map exp : 𝔱 → 𝑇 is a homomorphism,and we can thus regard 𝑇 as 𝑇 (cid:27) 𝔱 / Γ , where Γ : = ker exp is a discrete subgroup of 𝔱 , calledthe integer lattice of 𝑇 .Homomorphisms 𝜃 : 𝑇 → 𝑇 (cid:48) are easily described. We need only check for a linearmap 𝜑 : 𝔱 → 𝔱 (cid:48) such that 𝜑 ( Γ ) ⊂ Γ (cid:48) , and if so, then 𝜃 = ˜ 𝜑 : 𝔱 / Γ → 𝔱 (cid:48) / Γ (cid:48) . All continuoushomomorphisms arise in this way, and all 1-dimensional representations of 𝑇 arise fromlinear maps 𝜑 : 𝔱 → 𝔲 ( ) . Here we encounter the first connection to representation theory:given a representation 𝑉 of 𝑇 , there are linear maps 𝜑 : 𝔱 → 𝔲 ( ) such that 𝑉 decomposesas a direct sum of non-zero sub-representations 𝑉 𝜑 , where 𝔱 acts on 𝑉 𝜑 by 𝜏 ( 𝑥 ) = 𝜑 ( 𝜏 ) 𝑥 for 𝜏 ∈ 𝔱 , 𝑥 ∈ 𝑉 𝜑 . Definition 2.1.4 (Weights) . The linear maps 𝜑 on 𝔱 are called the weights of 𝑉 . Thedimension of 𝑉 𝜑 is the multiplicity of 𝜑 . The remarkable history of the more than one-hundred-and-fifty years of Lie theory isstudied in [84]; the Killing-Cartan classification of Lie groups is arguably the highpoint ofthis story, and certainly a significant achievement of modern mathematics. This sectionis the briefest of summaries of this classification scheme, and is important for us for tworeasons: (i) it motivates the existence of the exceptional Lie groups, and (ii) the roots andweights of the classical Lie groups that we will derive along the way will be instrumentalin constructing E . Definition 2.1.5 (Maximal Torus) . A maximal torus in a compact connected Lie group 𝐺 is a subgroup 𝑇 which is (i) a torus, and (ii) maximal, i.e. if 𝑇 ⊂ 𝑇 (cid:48) ⊂ 𝐺 for 𝑇 (cid:48) a torus, then 𝑇 (cid:48) = 𝑇 . Example 2.1.6.
The maximal torii of the classical Lie groups are as follows.(i) In U ( 𝑛 ) , consider the subgroup of matrices diag ( 𝑒 𝜋 𝑖𝑥 , . . . , 𝑒 𝜋 𝑖𝑥 𝑛 ) , for 𝑥 𝑗 ∈ R . Thisis a maximal torus in U ( 𝑛 ) : any matrix in U ( 𝑛 ) which commutes with all diagonalmatrices must be diagonal, and hence in 𝑇 . Thus, 𝑇 is maximal among all abeliansubgroups, connected or not.(ii) Since C ⊂ H , the matrices of (i) are in Sp ( 𝑛 ) , and they form a torus in Sp ( 𝑛 ) . Since C 𝑛 can be regarded as R 𝑛 , we get an embedding U ( 𝑛 ) → SO ( 𝑛 ) and we can again . 1 . C H A R AC T E R S A N D W E I G H T S O F R E P R E S E N TAT I O N S 15 take the corresponding matrices, namely (cid:169)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:171) cos 2 𝜋 𝑥 − sin 2 𝜋 𝑥 sin 2 𝜋 𝑥 cos 2 𝜋 𝑥 cos 2 𝜋 𝑥 − 𝜋 sin 𝑥 sin 2 𝜋 𝑥 cos 2 𝜋 𝑥 . . . cos 2 𝜋 𝑥 𝑛 − sin 2 𝜋 𝑥 𝑛 sin 2 𝜋 𝑥 𝑛 cos 2 𝜋 𝑥 𝑛 (cid:170)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:172) . These will form a torus in SO ( 𝑛 ) .(iii) We can embed R 𝑛 in R 𝑛 + and thus SO ( 𝑛 ) in SO ( 𝑛 + ) , where we map 𝐴 ↦→ (cid:0) 𝐴
00 1 (cid:1) .This is the corresponding torus in SO ( 𝑛 + ) .(iv) In SU ( 𝑛 ) we take the matrices of (i) subject to (cid:205) 𝑥 𝑖 = Theorem 2.1.7.
Let 𝑇 ⊂ 𝐺 be a maximal torus of a compact, connected Lie group 𝐺 . Then any 𝑔 ∈ 𝐺 is conjugate to some element of 𝑇 . That is, there exist elements 𝑡 ∈ 𝑇 , ℎ ∈ 𝐺 such that 𝑔 = ℎ𝑡 ℎ − . Corollary 2.1.8. If 𝑉 , 𝑊 are representations of a compact connected Lie group 𝐺 and 𝜒 𝑉 (cid:12)(cid:12) 𝑇 = 𝜒 𝑊 (cid:12)(cid:12) 𝑇 ,then 𝜒 𝑉 = 𝜒 𝑊 , so 𝑉 (cid:27) 𝑊 . Hence the weights (together with the multiplicities) of a representation 𝑉 of 𝐺 ,determine 𝑉 up to equivalence. We also have Corollary 2.1.9.
Given two maximal tori
𝑇 , 𝑇 (cid:48) in a compact connected Lie group 𝐺 , there existsan inner automorphism of 𝐺 taking 𝑇 to 𝑇 (cid:48) . It follows from this corollary that any property of 𝐺 defined by reference to a maximaltorus 𝑇 is independent of the choice of 𝑇 . The most important example of this is thefollowing Definition 2.1.10 (Rank) . The rank of a compact connected Lie group 𝐺 is the dimensionof the maximal torus of 𝐺 . We will usually write 𝑙 = rank 𝐺 .Suppose now that 𝑇 ⊂ 𝐺 is a torus (not necessarily maximal). Then 𝐺 acts on 𝔤 via the adjoint representation, so 𝑇 acts on 𝔤 by restriction and 𝔤 ⊗ C splits as a sumof 1-dimensional representations of 𝑇 , with 𝑇 acting trivially on 𝔱 ⊂ 𝔤 . Thus the trivial1-dimensional representation occurs at least 𝑑 = dim 𝑇 times. In fact, we have Proposition 2.1.11. If 𝐺 is compact, then 𝑇 is maximal if and only if the trivial 1-dimensionalrepresentation occurs exactly 𝑑 times. A proof of this result can be found in [3, p. 83]. Henceforth, we suppose that 𝑇 ismaximal and set 𝑑 = 𝑙 . Definition 2.1.12 (Roots) . The roots of a compact connected Lie group 𝐺 are the weightsof the adjoint representation, excluding 0 (which occurs 𝑙 times).The roots are thus R -linear functions on 𝔱 , that is, elements of 𝔱 ∗ . Since the adjointrepresentation of 𝐺 is real, the 1-dimensional summands of 𝔤 ⊗ C occur in conjugate pairsand the roots occur in pairs ± 𝜃 . ( ) G R A N D U N I F I E D T H E O RY
Example 2.1.13.
The roots of the classical Lie groups are as follows. • For the maximal torus of U ( 𝑛 ) described in example 2.1.6 above, the weights are 0, 𝑛 times, and ±( 𝑥 𝑖 − 𝑥 𝑗 ) , where 1 ≤ 𝑖 < 𝑗 ≤ 𝑛 . Proof.
First note that 𝔲 ( 𝑛 ) ⊗ C (cid:27) 𝔤𝔩 ( 𝑛, C ) (cid:27) End ( C 𝑛 ) since 𝐵 ↦→ − 𝐵 † is a conjugatelinear structure map and its +1 eigenspace is 𝔲 ( 𝑛 ) . Now take the basis { 𝑒 𝑗 } ofstandard column vectors for C 𝑛 ; for 𝑖 < 𝑗 , define linear maps 𝜃 𝑖𝑗 ∈ End ( C 𝑛 ) by 𝜃 𝑖𝑗 ( 𝑒 𝑗 ) = 𝑒 𝑖 , 𝜃 𝑖𝑗 ( 𝑒 𝑘 ) = 𝑘 ≠ 𝑗 . The matrix of 𝜃 𝑖𝑗 has a 1 in the 𝑖𝑗 -th place andzeroes elsewhere. The 𝜃 𝑖𝑗 are eigenvectors of the action of 𝑇 with eigenvaluesexp ( 𝜋 𝑖 ( 𝑥 𝑖 − 𝑥 𝑗 )) , so 𝑥 𝑖 − 𝑥 𝑗 are eigenvalues for the action of 𝔱 on 𝔲 ( 𝑛 ) . (Here, we aretaking 𝔱 to be the diagonal matrices diag ( 𝑖 𝑦 , . . . , 𝑖 𝑦 𝑛 ) , 𝑦 𝑗 ∈ R , and 𝑥 𝑖 ∈ 𝔱 ∗ is given by 𝑥 𝑖 ( diag ( 𝑖 𝑦 , . . . , 𝑖 𝑦 𝑛 )) = 𝑖 𝑦 𝑖 .) QED • The roots of the other matrix groups areSU ( 𝑛 ) : ±( 𝑥 𝑖 − 𝑥 𝑗 ) , 1 ≤ 𝑖 < 𝑗 ≤ 𝑛 ; 0, ( 𝑛 − ) times.SO ( 𝑛 ) : ± 𝑥 𝑖 ± 𝑥 𝑗 , 1 ≤ 𝑖 < 𝑗 ≤ 𝑛 ; 0, 𝑛 times.SO ( 𝑛 + ) : ± 𝑥 𝑖 ± 𝑥 𝑗 , 1 ≤ 𝑖 < 𝑗 ≤ 𝑛 ; 0, 𝑛 times; ± 𝑥 𝑖 , 1 ≤ 𝑖 ≤ 𝑛 .Sp ( 𝑛 ) : ± 𝑥 𝑖 ± 𝑥 𝑗 , 1 ≤ 𝑖 < 𝑗 ≤ 𝑛 ; 0, 𝑛 times; ± 𝑥 𝑖 , 1 ≤ 𝑖 ≤ 𝑛 . Definition 2.1.14 (Weyl Group) . The Weyl group 𝑊 of a compact, connected Lie group 𝐺 is the group of those automorphisms of a maximal torus which are given by innerautomorphisms of 𝐺 . Example 2.1.15.
In U ( ) , conjugation by (cid:0) −
11 0 (cid:1) is an element of 𝑊 , and (cid:18) −
11 0 (cid:19) (cid:18) 𝑒 𝜋 𝑖𝑥 𝑒 𝜋 𝑖𝑥 (cid:19) (cid:18) − (cid:19) = (cid:18) 𝑒 𝜋 𝑖𝑥 𝑒 𝜋 𝑖𝑥 (cid:19) ∈ 𝑇 .
The Weyl groups of the other classical matrix groups are as follows [3, pp.114–116]U ( 𝑛 ) and SU ( 𝑛 ) : 𝑊 = any permutation of 𝑥 , . . . , 𝑥 𝑛 .Sp ( 𝑛 ) and SO ( 𝑛 + ) : 𝑊 = the group generated by all permutations of 𝑥 , . . . , 𝑥 𝑛 and all sign changes of 𝑥 𝑖 .SO ( 𝑛 ) : 𝑊 = the group generated by all permutations of 𝑥 , . . . , 𝑥 𝑛 and an even number of sign changes of 𝑥 𝑖 .The Weyl group 𝑊 acts on 𝔱 and permutes roots. If we regard the roots as elementsof 𝔱 ∗ , they form a configuration with great symmetry and very distinctive properties [3,Ch. 5]. The Dynkin diagram encodes this configuration, as we proceed to describe.We may choose on 𝔱 ∗ a positive definite inner product invariant under 𝑊 , so thatwe can define the lengths and angles of roots. For each pair of roots ± 𝜃 , the kernel,ker 𝜃 = ker (− 𝜃 ) , is a hyperplane in 𝔱 called a root plane . Conversely, it can be shown thateach root plane comes from only one pair of roots, ± 𝜃 . The root planes form a figure in 𝔱 called the (infinitesimal) Stiefel diagram .The root planes divide 𝔱 into convex open sets called Weyl chambers , and the Weyl grouppermutes these chambers in a way which is simply transitive. We choose one and call itthe fundamental Weyl chamber (FWC); we denote it by 𝐶 . A root 𝜃 is positive (resp. negative )if 𝜃 > 𝜃 <
0) on 𝐶 . A positive root is simple if it defines a wall of 𝐶 ; in the Stiefeldiagram of SO ( ) shown in figure 2.1, the roots 𝑥 and 𝑥 − 𝑥 are simple. . 1 . C H A R AC T E R S A N D W E I G H T S O F R E P R E S E N TAT I O N S 17 𝑥 𝑥 𝑥 − 𝑥 𝐶 Figure 2.1: The Stiefel Diagram of SO ( ) The
Dynkin diagram has one node for each simple root (i.e. for each wall of thefundamental Weyl chamber). These nodes are joined by the following number of bonds: , ◦ or 120 ◦ , ◦ or 135 ◦ , ◦ or 150 ◦ . These are the only possibilities [3, pp.118–121]. By definition, the Dynkin diagram of thetorus is empty.
Example 2.1.16.
The Dynkin diagrams of the classical Lie groups are as follows.(i) For U ( 𝑛 ) , take 𝐶 to have 𝑥 < 𝑥 < · · · < 𝑥 𝑛 . We obtain − 𝑥 + 𝑥 − 𝑥 + 𝑥 − 𝑥 + 𝑥 − 𝑥 𝑛 − + 𝑥 𝑛 SU ( 𝑛 ) has the same Dynkin diagram as U ( 𝑛 ) , which is traditionally labelled 𝐴 𝑛 − .(We always take the usual inner product on R 𝑛 .)(ii) For SO ( 𝑛 ) , take 𝐶 to have − 𝑥 < 𝑥 < 𝑥 < 𝑥 < · · · < 𝑥 𝑛 . Then the diagram isdenoted 𝐷 𝑛 and is given by 𝑥 + 𝑥 − 𝑥 + 𝑥 − 𝑥 + 𝑥 − 𝑥 + 𝑥 − 𝑥 + 𝑥 − 𝑥 𝑛 − + 𝑥 𝑛 (iii) For SO ( 𝑛 + ) , take 𝐶 to have 0 < 𝑥 < 𝑥 < · · · < 𝑥 𝑛 . Then we have the Dynkindiagram 𝐵 𝑛 : 𝑥 short − 𝑥 + 𝑥 − 𝑥 + 𝑥 − 𝑥 𝑛 − + 𝑥 𝑛 (iv) For Sp ( 𝑛 ) , take 𝐶 to have 0 < 𝑥 < 𝑥 < · · · < 𝑥 𝑛 . Then we have the Dynkin diagram 𝐶 𝑛 : ( ) G R A N D U N I F I E D T H E O RY2 𝑥 long − 𝑥 + 𝑥 − 𝑥 + 𝑥 − 𝑥 𝑛 − + 𝑥 𝑛 Note that “short” (resp. “long”) means 𝑥 · 𝑥 = < 𝑥 · 𝑥 = > 𝐺 and 𝐺 (cid:48) are compact Lie groups, then 𝔤 is isomorphic to 𝔤 (cid:48) if and onlyif 𝔤 ⊗ C is isomorphic to 𝔤 (cid:48) ⊗ C ; thus the Dynkin diagram determines 𝔤 , and hence 𝐺 ,locally. In particular, corresponding to each Dynkin diagram, there is a unique compact,connected, simply connected Lie group, because to each of the diagrams in the Killing-Cartan classification , i.e. example 2.1.16, plus the exceptional Dynkin diagrams below, thereis a unique simple Lie algebra—in fact, every complex simple Lie algebra is isomorphic toone of the algebras in this classification scheme—and hence a unique connected, simplyconnected, compact, simple Lie group. As we saw above, the groups SU ( 𝑛 + ) , 𝑛 ≥
1, andSp ( 𝑛 ) , 𝑛 ≥ 𝐴 𝑛 and 𝐶 𝑛 respectively; the groups Spin ( 𝑛 + ) , 𝑛 ≥ ( 𝑛 ) , 𝑛 ≥ 𝐵 𝑛 and 𝐷 𝑛 . All these groups have rank 𝑛 and are pairwisenon-isomorphic. The non-classical or “exceptional” Dynkin diagrams are as follows; thenotation and conventions are explained in [4, Ch. 9].G F 𝑠 𝑠 𝑙 𝑙 E E E We are primarily interested in the Lie group corresponding to the diagram E , butto arrive at the same, we will need to construct the group E ; we do so in section 3.1.Additionally, we describe the construction of the smallest exceptional Lie group G in anappendix, by way of an illustrative example. By way of motivating grand unified theories, we have already raised several questions aboutthe unsatisfactory aesthetics of the Standard Model representation. We introduce nowconsiderations of a more technical nature, which will help us “classify” grand unificationgroups as it were, i.e. understand which groups are preferred from the plenitude ofavailable Lie groups that contain embeddings of the Standard Model. . 2 . P O S S I B L E G R A N D U N I F I C AT I O N G RO U P S 19
Coupling Constants in Gauge TheoriesDefinition 2.2.1 (Killing Form) . Let 𝐴 be a Lie algebra. The map 𝐴 × 𝐴 C , ( 𝑋 , 𝑌 ) 𝐾 tr ( ad ( 𝑋 ) ◦ ad ( 𝑌 )) is called its Killing form.The symmetry and bilinearity of this form are easy to check; it is also immediately clearthat the Killing form of an abelian Lie algebra is zero. More interesting is the following Proposition 2.2.2.
Let 𝜎 : 𝐴 → 𝐴 be a Lie algebra automorphism. Then ( 𝜎 𝑋 , 𝜎 𝑌 ) 𝐾 = ( 𝑋 , 𝑌 ) 𝐾 for all 𝑋 , 𝑌 ∈ 𝐴 . For 𝐴 = 𝔤 the Lie algebra of some Lie group, this holds in particular for theautomorphism Ad ( 𝑔 ) for an arbitrary 𝑔 ∈ 𝐺 . Proof.
Since 𝜎 is a Lie algebra automorphism, we havead ( 𝜎 𝑋 ) 𝑌 = [ 𝜎 𝑋 , 𝑌 ] = 𝜎 ([ 𝑋 , 𝜎 − 𝑌 ]) = ( 𝜎 ◦ ad ( 𝑋 ) ◦ 𝜎 − )( 𝑌 ) . We hence compute ( 𝜎 𝑋 , 𝜎 𝑌 ) 𝐾 = tr ( ad ( 𝜎 𝑋 ) ◦ ad ( 𝜎 𝑌 )) = tr ( 𝜎 ◦ ad ( 𝑋 ) ◦ ad ( 𝑌 ) ◦ 𝜎 − ) = tr ( ad ( 𝑋 ) ◦ ad ( 𝑌 )) = ( 𝑋 , 𝑌 ) 𝐾 QEDWe introduce some more terminology: if 𝔤 is the Lie algebra of a compact Lie group 𝐺 ,it is in turn called compact ; a subspace 𝔦 ⊂ 𝔤 is called an ideal if it is closed under the Liebracket, and satisfies [ 𝔤 , 𝔦 ] ⊆ 𝔦 ; a Lie algebra is called simple if its only ideals are 0 and itself.Now, one can show that for 𝔤 compact and simple, the negative of the Killing form is apositive-definite inner product; moreover, it turns out that up to a positive constant, it isthe unique such form. The proof of this is not very hard, and can be found in [31, Ch. 8.1],for example. From this, one can deduce the following result (see [44, Ch. 2.10] for a proof)which will in turn finally allow us to make the definition that we are after. Theorem 2.2.3.
Let 𝐺 be a compact, connected Lie group of the form 𝐺 = U ( ) × · · · × U ( ) × 𝐺 × · · · × 𝐺 𝑛 , (up to a finite quotient) where the 𝐺 𝑖 are simple. Let 𝑘 : 𝔤 × 𝑔 → C be an Ad -invariant positivedefinite scalar product on the Lie algebra 𝔤 . Then 𝑘 is the orthogonal sum of • a positive definite scalar product 𝑘 on the abelian algebra 𝔲 ( ) ⊕ · · · ⊕ 𝔲 ( ) , and • Ad 𝐺 𝑖 -invariant positive definite scalar products 𝑘 𝑖 ’s on the Lie algebras 𝔤 𝑖 .The scalar product 𝑘 is determined by a positive definite symmetric matrix, and the scalar products 𝑘 𝑖 are determined by positive constants relative to some fixed Ad -invariant positive definite scalarproducts on the corresponding Lie algebras (such as the negative Killing form). Definition 2.2.4 (Coupling Constants) . The constants that determine the positive definiteAd-invariant scalar products on the abelian ideal 𝔲 ( ) ⊕ · · · ⊕ 𝔲 ( ) and the 𝔤 𝑖 -summandsrelative to some standard Ad-invariant scalar products, are called coupling constants. ( ) G R A N D U N I F I E D T H E O RY
Some insight from physics is in order. Gauge couplings are simply numbers, determinedby experiment, that fix the interaction strength of the field that they correspond to. Theyare encountered most directly in pure Yang-Mills theories , which lie at the heart of bothelectroweak unification and QCD. We consider them briefly, returning to the frameworkof gauge theory; we follow [44, Ch. 7.2]. Let 𝐺 : 𝑃 → 𝑀 be a principal bundle withthe structure group 𝐺 compact and finite dimensional. Further, fix an Ad-invariantpositive-definite scalar product 𝑘 on 𝔤 as in the theorem above, and a 𝑘 -orthonormal basisfor 𝔤 . For 𝐴 a connection 1-form with curvature 2-form 𝐹 𝐴 ∈ Ω ( 𝑃, 𝔤 ) , in a local gauge 𝑠 : 𝑀 ⊃ 𝑈 → 𝑃 , the field strength is given by 𝐹 𝐴𝑠 : = 𝑠 ∗ 𝐹 𝐴 ∈ Ω ( 𝑀, 𝔤 ) . The Yang-Mills Lagrangian is then simply defined by ℒ YM = − 𝑘 ( 𝐹 𝐴𝑠 , 𝐹 𝐴𝑠 ) . In the case that 𝐺 is simple, for instance, there is a single coupling constant 𝑔 > ≈ renormalisation group running . Calculations show (see [66, Ch. 5.5]) that if the couplingconstants are normalised as in the previous paragraph, i.e. taken to be orthonormal withrespect to the Killing form on 𝔤 SM , the renormalisation group equation indicates that theyroughly converge at high energies. This is a plausibility argument for a grand unificationgroup with a single coupling constant, unifying the three forces of the Standard Modelat high energies; this can only occur if the unification group is simple, or a product ofidentical simple groups, where the coupling constant for each factor is set the same byforcing the theory to have some sort of discrete symmetry. This is the first demand that wewill make of any potential grand unification group. Chirality and Complex Representations
In mathematics, the term “complex representation” simply refers to a group representationon a complex vector space; the term as used in physics denotes something different, and itis related to chirality. As we have seen, the weak force, and hence the Standard Model, ischiral. This unexpected feature detracts significantly from the symmetry of the rest of thetheory, and one might expect that grand unified theories behave more naturally, or at leastsomehow explain this parity violation. But this is in fact not the case: Georgi [34] andBarbieri et al. [10] have argued that the fermions that would have to be introduced into anachiral grand unified theory to recover the chirality of the Standard Model on symmetrybreaking would be unacceptably heavy; this is an instance of the
Survival Hypothesis , whichwe will discuss in more detail in section 4.3. For the moment, we will content ourselveswith defining a complex representation, and seeing how it is concerned with chirality.
Definition 2.2.5 (Complex Representation) . Two representations 𝜋 : 𝐺 → GL ( 𝑉 ) and 𝜋 : 𝐺 → GL ( 𝑉 ) of a group 𝐺 are said to be equivalent if there is an intertwining operatorfrom 𝑉 to 𝑉 such that it is also a vector space isomorphism. If 𝜋 : 𝐺 → GL ( 𝑉 ) is arepresentation, the complex conjugate representation 𝜋 is defined over the complex conjugate . 2 . P O S S I B L E G R A N D U N I F I C AT I O N G RO U P S 21 vector space 𝑉 by 𝜋 ( 𝑔 ) = 𝜋 ( 𝑔 ) . A representation of a group is said to be complex if it isnot equivalent to its complex conjugate representation.The connection to handedness is pretty straightforward. We know that the way toget the antiparticle representation from the particle representation is simply to pass tothe dual; for example, 𝜈 𝑅 ∈ C ⊗ C ⊗ C (cid:27) C − ⊗ C ⊗ C (cid:51) 𝜈 𝐿 (we have used here thefact that C (cid:27) C under SU ( ) ). So if we take a direct sum of the representations of allthe left-handed fermions, call this 𝑓 𝐿 , it stands to reason that the direct sum of all theright-handed fermion representations is given by 𝑓 𝑅 = 𝑓 𝐿 . Therefore, if 𝑓 𝐿 (cid:27) 𝑓 𝐿 , i.e. if 𝑓 𝐿 isreal, the theory is manifestly achiral, since the right-handed particles transform as the left;such theories are called vectorlike ; on the other hand, if 𝑓 𝐿 is complex, the theory is chiral.We will hence demand that our grand unification groups admit complex representations,to preserve this feature of the standard model. Let us summarise our work in the following Definition 2.2.6 (Possible Unification Group) . We call a Lie group 𝐺 a possible unificationgroup if it satisfies the following properties. • 𝐺 is simple, or a product of several copies of the same simple group. • 𝐺 contains (perhaps up to a finite quotient) the Standard Model gauge group 𝐺 SM . • 𝐺 admits complex representations. Classification of Unification Groups
We will restrict our discussion to Lie groups with rank less than 7, since they are generallyconsidered the most interesting for unification ; they are listed as follows: • rank 1: SU ( ) ; • rank 2: SU ( ) , Spin ( ) , G ; • rank 3: SU ( ) , Spin ( ) , Sp ( ) ; • rank 4: SU ( ) , Spin ( ) , Spin ( ) , Sp ( ) , F • rank 5: SU ( ) , Spin ( ) , Spin ( ) , Sp ( ) ; • rank 6: SU ( ) , Spin ( ) , Spin ( ) , Sp ( ) , E .Of these, the only ones for which we have not explicitly computed the rank are theexceptional groups G , F and E . Since we will momentarily eliminate F as a possibleunification group, we will not bother with this computation ; the rank of E is computedin the proof of theorem 3.1.1, and of G in the appendix.Mehta and Srivastava have classified the complex representations of all the classicalLie groups: only the SU ( 𝑛 ) ’s, for 𝑛 >
2, the Spin ( 𝑛 + ) ’s, for 𝑛 ≥
1, and E admit complexrepresentations [64, 65]. Together with the fact that the gauge group of the Standard Model 𝐺 SM = U ( ) × SU ( ) × SU ( ) has rank equal to 1 + + =
4, and the further requirement onsimplicity from definition 2.2.6, we can immediately thin down the above list significantly. For a discussion on higher rank unification groups, see [56, Ch. 3.4] and [81, Ch. 3]. The interested reader may refer to [4, Ch. 8]. In section 4.2 we will consider the issue of anomaly cancellation, and its consequences for grand unifiedtheories. This rather subtle requirement from quantum field theory is hard to motivate from a representationtheoretic standpoint alone (though it does have a nice interpretation in the same), and was hence omitted inthis section. In any case, it has no bearing on our list of possible unification groups.2 2 . T H E S P I N ( ) G R A N D U N I F I E D T H E O RY
Proposition 2.2.7.
The only possible grand unification groups with rank less than are thefollowing: • rank : SU ( ) and SU ( ) ; • rank : SU ( ) and Spin ( ) ; • rank : SU ( ) , SU ( ) , SU ( ) and E . We provide references for these grand unified theories, where they exist. In the samepaper [36] in which they proposed the SU ( ) theory, Georgi and Glashow ruled out anSU ( ) theory for physical reasons, leaving SU ( ) the unique rank 4 unification group;we will turn to this theory in the next section. A theory with unification group SU ( ) was suggested in 2005 by Hartanto and Handoko [47], while the Spin ( ) grand unifiedtheory was put forward by Georgi in 1974 [33] and Fritzsch and Minkowski in 1975 [30].Finally, a theory with SU ( ) as gauge group, called trinification , was demonstrated by deRúluja et al. in 1984 [22], an SU ( ) grand unification theory was studied by Umemura andYamamoto in 1981 [83], and the subject of this thesis, the E grand unified theory, firstappeared in a 1976 paper by Gürsey et al. [40]. SU ( ) Grand Unified Theory
Georgi and Glashow’s SU ( ) extension of the Standard Model was the first grand unifiedtheory, and is still considered the prototypical example of the same. Unfortunately thistheory has since been ruled out by experiment: it predicts that protons will decay fasterthan the current lower bound on proton lifetime. Our focus here will be simply to showwhat exactly we mean when we say that SU ( ) is a grand unified theory; the questions wewill ask and methodology we will develop will be highly instructive for us when we laterconsider the Spin ( ) theory, and eventually the one of E . We closely follow [8] in thissection.For integers 𝑚, 𝑛 ≥
1, let us define S ( U ( 𝑚 ) × U ( 𝑛 )) = {( 𝐴, 𝐵 ) ∈ U ( 𝑚 ) × U ( 𝑛 ) | det 𝐴 · det 𝐵 = } . This Lie group is naturally a subgroup of SU ( 𝑚 + 𝑛 ) under theembedding S ( U ( 𝑚 ) × U ( 𝑛 )) SU ( 𝑚 + 𝑛 ) , ( 𝐴, 𝐵 ) (cid:18) 𝐴 𝐵 (cid:19) . The key to the whole SU ( ) theory is the following: the subgroup S ( U ( ) × U ( )) isisomorphic to 𝐺 SM , modulo a finite subgroup. More precisely, consider the map 𝜙 : U ( ) × SU ( ) × SU ( ) SU ( ) , ( 𝛼 , 𝐴, 𝐵 ) (cid:18) 𝛼 𝐴 𝛼 − 𝐵 (cid:19) ;this is clearly a homomorphism from 𝐺 SM to S ( U ( ) × U ( )) . Equally clear is the fact that itis not injective: its kernel is all elements of the form ( 𝛼 , 𝛼 − , 𝛼 ) . This kernel is Z , becausescalar matrices 𝛼 − and 𝛼 live in SU ( ) and SU ( ) simultaneously if and only if 𝛼 is a sixthroot of unity. So in short order, we have obtained 𝐺 SM / Z (cid:27) S ( U ( ) × U ( )) ↩ → SU ( ) . Detailed studies and reviews of this theory abound in the literature, see [66, 76] and references therein. . 3 . T H E S U ( ) G R A N D U N I F I E D T H E O RY 23
This sets up a test that the SU ( ) theory must pass for it to have any chance ofsuccess: not all representations of 𝐺 SM factor through 𝐺 SM / Z , but all those coming fromrepresentations of SU ( ) must do so. In particular, we have to check that Z acts triviallyon all the irreps inside 𝐹 , that is, it must act trivially on all fermions (and antifermions,but that amounts to the same thing). For this to be true, some non-trivial relationsbetween hypercharge, isospin and colour must hold. Consider for example the electron 𝑒 − 𝐿 ∈ C − ⊗ C ⊗ C ; for any 𝛼 ∈ Z we need ( 𝛼 , 𝛼 − , 𝛼 ) to act trivially on this particle. Wecompute, ( 𝛼 , 𝛼 − , 𝛼 ) · 𝑒 − 𝐿 = 𝛼 − 𝛼 − 𝑒 − 𝐿 = 𝛼 − 𝑒 − 𝐿 = 𝑒 − 𝐿 , since 𝛼 is a sixth root of unity. In principle, there are 15 other such cases to check, butthese can be reduced to just four hypercharge relations that must be satisfied: • for the left-handed quarks, 𝑌 = even integer + / • for the left-handed leptons, 𝑌 = odd integer, • for the right-handed quarks, 𝑌 = odd integer + /
3, and • for the right-handed leptons, 𝑌 = even integer.A glance at table 1.1 shows that all of these equalities hold, so our SU ( ) theory has passedits first test. We remark here that not only is Z contained in the kernel of the StandardModel representation, but it is in fact the entire kernel. Hence, one could say that 𝐺 SM / Z is the “true” gauge group of the Standard Model.Our next order of business is to find a representation of SU ( ) that extends the StandardModel representation, and there is a beautiful choice that works: the exterior algebra Λ ∗ C .We have to check that pulling back the representation from SU ( ) to 𝐺 SM using 𝜙 givesthe Standard Model representation 𝐹 ⊕ 𝐹 ; our strategy will be to use the fact that, beingrepresentations of compact Lie groups, both 𝐹 ⊕ 𝐹 and Λ ∗ C are completely reducible, andcan be written as the direct sum of irreps; we will then match them up one irrep at a time.We already know what the decomposition of 𝐹 ⊕ 𝐹 into irreps is, so let us look at Λ ∗ C . Anyelement 𝐴 ∈ SU ( ) acts as an automorphism of the exterior algebra: 𝐴 · ( 𝑣 ∧ 𝑤 ) = 𝐴𝑣 ∧ 𝐴𝑤 ,where 𝑣, 𝑤 ∈ Λ ∗ C . Since we know how 𝐴 acts on vectors in C , and these generate Λ ∗ C ,this rule is enough to tell us how 𝐴 acts on all of Λ ∗ C . This action respects grades in Λ ∗ C ,so each exterior power in Λ ∗ C (cid:27) Λ C ⊕ Λ C ⊕ Λ C ⊕ Λ C ⊕ Λ C ⊕ Λ C is a subrepresentation. More than that, they are all irreps of SU ( ) , though this is not soeasy to see; we refer the reader to [32, Ch. 15.2] for a proof. Λ C and Λ C are both trivial irreps of 𝐺 SM , and there are exactly two trivial irreps in 𝐹 ⊕ 𝐹 , namely (cid:104) 𝜈 𝑅 (cid:105) and (cid:104) 𝜈 𝐿 (cid:105) (we use the angle brackets to denote the Hilbert space spannedby a vector or vectors). Hence, these irreps must match up; we will select Λ C (cid:27) (cid:104) 𝜈 𝐿 (cid:105) and Λ C (cid:27) (cid:104) 𝜈 𝑅 (cid:105) for reasons that will be clear in a moment. Consider next the irrep Λ C (cid:27) C . The group 𝐺 SM acts on C via 𝜙 ; just by inspection, we see that this actionpreserves a splitting of C into C ⊕ C , with the C part transforming in the hyperchargerepresentation C , and the C piece transforming in C − / . From table 1.1 then, we see thatwe must have Λ C (cid:27) (cid:0) C ⊗ C ⊗ C (cid:1) ⊕ (cid:0) C − / ⊗ C ⊗ C (cid:1) (cid:27) (cid:28) 𝑒 + 𝑅 𝜈 𝑅 (cid:29) ⊕ (cid:104) 𝑑 𝑅 (cid:105) , ( ) G R A N D U N I F I E D T H E O RY where we once again used the self-duality of C under SU ( ) .The remainder of the irrep matching is similarly straightforward. The final result is asfollows: Λ C (cid:27) (cid:104) 𝜈 𝐿 (cid:105) , Λ C (cid:27) (cid:28) 𝑒 + 𝑅 𝜈 𝑅 (cid:29) ⊕ (cid:104) 𝑑 𝑅 (cid:105) , Λ C (cid:27) (cid:104) 𝑒 + 𝐿 (cid:105) ⊕ (cid:28) 𝑢 𝐿 𝑑 𝐿 (cid:29) ⊕ (cid:104) 𝑢 𝐿 (cid:105) , Λ C (cid:27) (cid:104) 𝑒 − 𝑅 (cid:105) ⊕ (cid:28) 𝑑 𝑅 𝑢 𝑅 (cid:29) ⊕ (cid:104) 𝑢 𝑅 (cid:105) , Λ C (cid:27) (cid:28) 𝜈 𝐿 𝑒 − 𝐿 (cid:29) ⊕ (cid:104) 𝑑 𝐿 (cid:105) , Λ C (cid:27) (cid:104) 𝜈 𝑅 (cid:105) . (2.3.1)Hence, Λ ∗ C (cid:27) 𝐹 ⊕ 𝐹 , as desired. Notice that our choice Λ C (cid:27) (cid:104) 𝜈 𝐿 (cid:105) has led to a ratherpleasing pattern: the left-handed particles transform in the even grades, while the righthanded particles transform in the odd ones. At the level of the SU ( ) theory, this is nicebut not essential; for the Spin ( ) theory, this is the only possibility; we will return to thispoint in section 2.5.We have now shown everything we needed to show: the mapping above defines a linearisomorphism 𝐹 ⊕ 𝐹 → Λ ∗ C between representations of 𝐺 SM , i.e. these representations arethe same when we identify S ( U ( ) × U ( )) with 𝐺 SM / Z using the isomorphism induced by 𝜙 . This can be neatly summarised in a commuting diagram, the main result of this section. Theorem 2.3.1. SU ( ) is a grand unified theory, i.e. the following square commutes: 𝐺 SM / Z SU ( ) 𝐹 ⊕ 𝐹 Λ ∗ C (cid:27) To approach the Spin ( ) grand unified theory, we need to understand Clifford algebras,which are the most natural environment in which to study the Spin groups. Moreover,many of the results that we will obtain will be required to construct E in due course.Clifford algebras are a generalisation of the complex numbers, quaternions andoctonions: indeed, they are sometimes constructed in the literature by adding the requirednumber of square-roots of − 𝑉 is a finite-dimensional vector space over K = R or C . Definition 2.4.1 (Tensor Algebra) . For a non-negative integer 𝑘 , we define the 𝑘 th tensorpower of 𝑉 to be the tensor product of 𝑉 with itself 𝑘 times: 𝑇 𝑘 𝑉 = 𝑉 ⊗ 𝑘 = 𝑉 ⊗ 𝑉 ⊗ · · · ⊗ 𝑉 (cid:124) (cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32) (cid:123)(cid:122) (cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32) (cid:125) 𝑘 times . By convention, 𝑇 𝑉 = K . Then the tensor algebra is given by 𝑇 ( 𝑉 ) = ∞ (cid:202) 𝑘 = 𝑇 𝑘 𝑉 .
We define multiplication as follows: if 𝑣 ⊗ · · · ⊗ 𝑣 𝑝 ∈ 𝑉 ⊗ 𝑝 and 𝑤 ⊗ · · · ⊗ 𝑤 𝑞 ∈ 𝑉 ⊗ 𝑞 ,then their product is 𝑣 ⊗ · · · ⊗ 𝑣 𝑝 ⊗ 𝑤 ⊗ · · · ⊗ 𝑤 𝑞 ∈ 𝑉 ⊗( 𝑝 + 𝑞 ) . For example, if 𝑉 has a basis { 𝑥.𝑦 } , then 𝑇 ( 𝑉 ) has a basis { , 𝑥, 𝑦, 𝑥 𝑦, 𝑦𝑥, 𝑥 , 𝑦 , . . . } (the tensor product symbol hasbeen omitted for brevity). In general, 𝑇 ( 𝑉 ) is a free associative algebra. . 4 . C L I F F O R D A L G E B R A S 25 Definition 2.4.2 (Clifford Algebra) . Let 𝑉 be endowed with a symmetric bilinear form (cid:104) , (cid:105) .Let 𝐽 denote the two-sided ideal in 𝑇 ( 𝑉 ) generated by the set { 𝑣 ⊗ 𝑣 − (cid:104) 𝑣, 𝑣 (cid:105) · | 𝑣 ∈ 𝑉 } ,and define Cl ( 𝑉 ) : = 𝑇 ( 𝑉 )/ 𝐽 ;this is the Clifford Algebra over ( 𝑉 , (cid:104)· , ·(cid:105)) .It is clear that this is equivalent to the characterisation one usually sees, namely, thatthe Clifford algebra is the associative algebra freely generated by 𝑉 with relations 𝑣𝑤 + 𝑤𝑣 = − (cid:104) 𝑣, 𝑤 (cid:105) . (2.4.1)A word on notation: though we should define a map 𝑉 → Cl ( 𝑉 ) denoting the composition 𝑉 ↩ → 𝑇 ( 𝑉 ) → 𝑇 ( 𝑉 )/ 𝐽 = Cl ( 𝑉 ) , this is usually omitted in practice, and we write 𝑣 for anelement of 𝑉 or its image in Cl ( 𝑉 ) .Now, 𝑇 ( 𝑉 ) is a Z + -graded algebra; let us set 𝑇 ( 𝑉 ) : = (cid:202) 𝑛 even 𝑉 ⊗ 𝑛 , 𝑇 ( 𝑉 ) : = (cid:202) 𝑛 odd 𝑉 ⊗ 𝑛 . Then 𝑣 ⊗ 𝑣 − (cid:104) 𝑣, 𝑣 (cid:105) · ∈ 𝑇 ( 𝑉 ) and 𝐽 = 𝐽 + 𝐽 , where 𝐽 𝑖 = 𝐽 ∩ 𝑇 𝑖 ( 𝑉 ) , andCl ( 𝑉 ) = Cl ( 𝑉 ) ⊕ Cl ( 𝑉 ) , where Cl 𝑖 ( 𝑉 ) = 𝑇 𝑖 ( 𝑉 )/ 𝐽 𝑖 . Hence, Cl ( 𝑉 ) is a Z -graded algebra. Proposition 2.4.3. If 𝑉 = 𝑉 (cid:48) ⊕ 𝑉 (cid:48)(cid:48) with (cid:104) 𝑣 (cid:48) , 𝑣 (cid:48)(cid:48) (cid:105) = for all 𝑣 (cid:48) ∈ 𝑉 (cid:48) , 𝑣 (cid:48)(cid:48) ∈ 𝑉 (cid:48)(cid:48) , then Cl ( 𝑉 ) (cid:27) Cl ( 𝑉 (cid:48) ) ⊗ Cl ( 𝑉 (cid:48)(cid:48) ) is its Clifford algebra. The proof of this standard result can be found in, for example, [4, pp. 14–15]. Itscorollaries bridge the gap between our abstract construction of Clifford algebras, and themotivation of generalising the complex numbers.
Corollary 2.4.4. If dim K 𝑉 = 𝑛 and { 𝑒 . . . 𝑒 𝑛 } is an orthogonal basis for 𝑉 with (cid:104) 𝑒 𝑖 , 𝑒 𝑗 (cid:105) = 𝜆 𝑗 𝛿 𝑖𝑗 ,then dim K Cl ( 𝑉 ) = 𝑛 , and (cid:110)(cid:206) 𝑒 𝑖 𝑗 𝑗 (cid:111) is a basis, where 𝑖 𝑗 is 0 or 1. Example 2.4.5.
Suppose 𝑉 is 1-dimensional with basis { 𝑒 } . Then Cl ( 𝑉 ) has basis { , 𝑒 } ,because 𝑇 ( 𝑉 ) has basis (cid:8) , 𝑒 , 𝑒 , . . . (cid:9) and 𝐽 has a basis (cid:8) 𝑒 − (cid:104) 𝑒 , 𝑒 (cid:105) · , 𝑒 − (cid:104) 𝑒 , 𝑒 (cid:105) · 𝑒 , . . . (cid:9) .Hence, Cl ( R ) = C , as desired. The next result shows that Cl ( C ) = H . Corollary 2.4.6.
Again, assume that { 𝑒 𝑖 } is a basis for 𝑉 and that (cid:104) 𝑒 𝑖 , 𝑒 𝑗 (cid:105) = − 𝛿 𝑖𝑗 . Then theproducts in Cl ( 𝑉 ) are determined by the following relations: • 𝑒 𝑖 = − , • 𝑒 𝑖 𝑒 𝑗 = − 𝑒 𝑗 𝑒 𝑖 , for 𝑖 ≠ 𝑗 . Structure Maps on Clifford Algebras
We define structure maps on Clifford algebras, analogously to remark 2.1.1. Consider 𝛼 : Cl ( 𝑉 ) → Cl ( 𝑉 ) , induced by − 𝑉 → 𝑉 ; we have that 𝛼 | Cl ( 𝑉 ) = +
1, and 𝛼 | Cl ( 𝑉 ) = − 𝛽 : 𝑇 ( 𝑉 ) → 𝑇 ( 𝑉 ) by 𝛽 ( 𝑣 ⊗ · · · ⊗ 𝑣 𝑛 ) = 𝑣 𝑛 ⊗ · · · ⊗ 𝑣 . This is an anti-automorphism, 𝛽 ( 𝑥 𝑦 ) = 𝛽 ( 𝑦 ) 𝛽 ( 𝑥 ) , and induces 𝛽 : Cl ( 𝑉 ) → Cl ( 𝑉 ) with 𝛽 | 𝑉 =
1. Finally, 𝛾 : = 𝛼𝛽 = 𝛽𝛼 : Cl ( 𝑉 ) → Cl ( 𝑉 ) is an anti-automorphism such that 𝛾 | 𝑉 = − ( ) G R A N D U N I F I E D T H E O RY
Example 2.4.7. • Cl ( R ) has generators { , 𝑖 } with 𝛼 ( ) = 𝛼 ( 𝑖 ) = − 𝑖 , 𝛽 ( ) = 𝛽 ( 𝑖 ) = 𝑖 , 𝛾 ( ) = 𝛾 ( 𝑖 ) = − 𝑖 . • The algebra H = Cl ( C ) has a basis { , 𝑖, 𝑗, 𝑖𝑗 = 𝑘 } , with the action of 𝛼 , 𝛽 , 𝛾 given by1 𝑖 𝑗 𝑘 𝛼 − 𝑖 − 𝑗 𝑘 𝛽 𝑖 𝑗 − 𝑘 𝛾 − 𝑖 − 𝑗 − 𝑘 Note that for Cl ( R ) = C or Cl ( C ) = H , 𝛾 is the usual conjugation map. We now posses the machinery to introduce the Spin groups. Let 𝑉 = K 𝑛 with (cid:104) , (cid:105) the standard inner product, and construct its Clifford algebra Cl ( 𝑉 ) with respect to −(cid:104) , (cid:105) . With { 𝑒 𝑖 } the standard basis for 𝑉 , we have the products 𝑒 𝑟 𝑒 𝑠 for 𝑟 < 𝑠 inCl ( 𝑉 ) —there are 𝑛 ( 𝑛 − )/ [ 𝑒 𝑟 𝑒 𝑠 , 𝑒 𝑡 𝑒 𝑢 ] = 𝑒 𝑟 𝑒 𝑠 𝑒 𝑡 𝑒 𝑢 − 𝑒 𝑡 𝑒 𝑢 𝑒 𝑟 𝑒 𝑠 . By corollary 2.4.6, we have that • if 𝑟, 𝑡, 𝑠, 𝑢 are different, the bracket vanishes, and also for 𝑟 = 𝑡, 𝑠 = 𝑢 ; • if 𝑟, 𝑡, 𝑢 are different, [ 𝑒 𝑟 𝑒 𝑢 , 𝑒 𝑡 𝑒 𝑢 ] = 𝑒 𝑟 𝑒 𝑡 .We wish to see this Lie algebra as the Lie algebra of a Lie group. Definition 2.4.8 (The Pin Groups) . Pin ( 𝑉 ) ⊂ Cl ( 𝑉 ) is the subset of elements 𝑥 such that • 𝑥 ( 𝛾 𝑥 ) = ( 𝛾 𝑥 ) 𝑥 =
1, and • the map 𝜋 𝑥 : 𝑉 → Cl ( 𝑉 ) defined by ( 𝜋 𝑥 ) 𝑣 = 𝑥𝑣 ( 𝛽 𝑥 ) maps 𝑉 ⊂ Cl ( 𝑉 ) into 𝑉 . Example 2.4.9.
Pin ( R ) = { 𝑧 ∈ Cl ( R ) | 𝑧𝑧 = , 𝑧 ∈ R } = {± , ± 𝑖 } . Proposition 2.4.10.
The
Pin groups satisfy the following properties. • Pin ( 𝑉 ) is a subgroup of the invertible elements of Cl ( 𝑉 ) and its Lie algebra is the one specifiedabove. • The map 𝜋 : Pin ( 𝑉 ) → O ( 𝑉 ) , where O ( 𝑉 ) is the group of K -linear maps 𝑉 → 𝑉 preserving −(cid:104) , (cid:105) , is a surjection with ker 𝜋 = {± } . • The closed subsets 𝜋 − ( det − ) and 𝜋 − ( det − − ) of Pin ( 𝑉 ) are in Cl ( 𝑉 ) and Cl ( 𝑉 ) respectively, and are connected for 𝑛 > . Proof.
It is straightforward to check that Pin ( 𝑉 ) is a closed subgroup of invertible elementsof Cl ( 𝑉 ) . We move on to showing that 𝜋 ( 𝑥 ) ∈ O ( 𝑉 ) : (cid:104)( 𝜋 𝑥 ) 𝑣, ( 𝜋 𝑥 ) 𝑣 (cid:105) = (( 𝜋 𝑥 ) 𝑣 ) = − 𝑥𝑣 ( 𝛽 𝑥 )( 𝛼 𝑥 ) 𝑣 ( 𝛾 𝑥 ) = − 𝑥𝑣𝑣 𝛾 ( 𝑥 ) = (cid:104) 𝑣, 𝑣 (cid:105) 𝑥 𝛾 𝑥 = (cid:104) 𝑣, 𝑣 (cid:105) , . 4 . C L I F F O R D A L G E B R A S 27 where in the second line we used the fact that −( 𝜋 𝑥 ) 𝑣 = −( 𝛼 𝑥 ) 𝑣 ( 𝛾 𝑥 ) .It is easy to see that 𝜋 is a homomorphism; let us try to identify its kernel. Suppose that 𝑥 ∈ ker 𝜋 . Then 𝑣 = 𝑥𝑣 𝛽 𝑥 for all 𝑥 ∈ 𝑉 , so 𝑣 𝛼 𝑥 = 𝑥𝑣 ( 𝛽 𝑥 )( 𝛼 𝑥 ) = 𝑥𝑣 ; by the lemma below, 𝑥 must be a scalar. But 𝑥 𝛾 𝑥 =
1, so 𝑥 = = ⇒ 𝑥 = ± = ⇒ ker 𝜋 ⊂ {± } ; since theinclusion certainly holds in the other direction, we conclude that ker 𝜋 is identically {± } . Lemma 2.4.11. If (cid:104) , (cid:105) on 𝑉 is non-singular and 𝑥 ∈ Cl ( 𝑉 ) is such that 𝑥𝑣 = 𝑣 ( 𝛼 𝑥 ) for all 𝑣 ∈ 𝑉 ,then 𝑥 is a scalar. Proof of Lemma.
Over K = R or C , we can diagonalise (cid:104) , (cid:105) and choose a basis { 𝑒 𝑖 } suchthat (cid:104) 𝑒 𝑟 , 𝑒 𝑠 (cid:105) = 𝛿 𝑟𝑠 𝜆 𝑟 , with 𝜆 𝑟 ≠
0. Then any 𝑥 ∈ Cl ( 𝑉 ) can be written as (cid:205) 𝐼 𝜆 𝐼 (cid:206) 𝑗 𝑒 𝑖 𝑗 𝑗 ,where 𝜆 𝐼 ∈ K . Now, 𝑒 𝑠 is invertible since 𝑒 𝑠 𝑒 𝑠 = 𝜆 𝑠 ≠
0, so that if 𝑥𝑒 𝑠 = 𝑒 𝑠 ( 𝛼 𝑥 ) , we have 𝑒 − 𝑠 𝑥𝑒 𝑠 = 𝛼 𝑥 . But 𝑒 − 𝑠 (cid:169)(cid:173)(cid:171)(cid:214) 𝑗 𝑒 𝑖 𝑗 𝑗 (cid:170)(cid:174)(cid:172) 𝑒 𝑠 = (cid:40) (− ) (cid:205) 𝑖 𝑗 (cid:206) 𝑒 𝑖 𝑗 𝑗 if 𝑖 𝑠 = , (− ) − + (cid:205) 𝑖 𝑗 (cid:206) 𝑒 𝑖 𝑗 𝑗 if 𝑖 𝑠 = , while 𝛼 ( (cid:206) 𝑗 𝑒 𝑖 𝑗 𝑗 ) = (− ) (cid:205) 𝑖 𝑗 (cid:206) 𝑒 𝑖 𝑗 𝑗 in all cases; thus 𝑥𝑒 𝑠 = 𝑒 𝑠 𝛼 𝑥 if and only if 𝜆 𝐼 = 𝑖 𝑠 =
1, i.e. 𝜆 𝐼 ≠ 𝐼 = ( , , . . . , ) , so 𝑥 is a scalar. QEDNext, we proceed to show that d 𝜋 from the Lie algebra of Pin ( 𝑉 ) ∩ Cl ( 𝑉 ) to 𝔬 ( 𝑉 ) is a surjection. The first step is of course to check that Pin ( 𝑉 ) ∩ Cl ( 𝑉 ) is a closedsubgroup, but this is straightforward. Then, for 𝑟 < 𝑠 , ( 𝑒 𝑟 𝑒 𝑠 )( 𝑒 𝑟 𝑒 𝑠 ) = − 𝑒 𝑟 𝑒 𝑟 𝑒 𝑠 𝑒 𝑠 = −
1, so 𝑥 = 𝑒 ( 𝑒 𝑟 𝑒 𝑠 ) 𝑡 = cos 𝑡 + ( 𝑒 𝑟 𝑒 𝑠 ) sin 𝑡 is defined for R or C . This means that 𝜋 ( 𝑥 ) = (cid:169)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:171) . . . 𝑡 − sin 2 𝑡 . . . 𝑡 cos 2 𝑡 . . . (cid:170)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:172) since ( cos 𝑡 + ( 𝑒 𝑟 𝑒 𝑠 ) sin 𝑡 ) 𝑒 𝑢 ( cos 𝑡 − ( 𝑒 𝑟 𝑒 𝑠 ) sin 𝑡 ) = (cid:40) 𝑒 𝑢 if 𝑢 ≠ 𝑟, 𝑠 , ( cos 2 𝑡 + ( 𝑒 𝑟 𝑒 𝑠 ) sin 2 𝑡 ) 𝑒 𝑢 if 𝑢 = 𝑟, 𝑠 . Thus 𝜋 ( 𝑥 ) maps 𝑉 to 𝑉 , so 𝑥 ∈ Pin ( 𝑉 ) ; in fact, 𝑥 ∈ Pin ( 𝑉 ) ∩ Cl ( 𝑉 ) . We see thatthe K -multiples of 𝑒 𝑟 𝑒 𝑠 lie in 𝔭𝔦𝔫 ( 𝑉 ) ∩ 𝔠𝔩 ( 𝑉 ) and map under d 𝜋 to the K -multiples of (cid:169)(cid:173)(cid:173)(cid:173)(cid:171) . . . − . . . (cid:170)(cid:174)(cid:174)(cid:174)(cid:172) = ( 𝜋 ( 𝑥 )) (cid:48) | 𝑡 = , which form a K -basis for 𝔬 ( 𝑉 ) , the skew-symmetric matrices.Hence, d 𝜋 is surjective.It follows from this that d 𝜋 : 𝔭𝔦𝔫 ( 𝑉 ) → 𝔬 ( 𝑉 ) is also a surjection; it must further be aninjection, since ker 𝜋 is finite. Thus, the Lie algebra of Pin ( 𝑉 ) ∩ Cl ( 𝑉 ) is really the Liealgebra 𝔭𝔦𝔫 ( 𝑉 ) , and d 𝜋 : 𝔭𝔦𝔫 ( 𝑉 ) → 𝔬 ( 𝑉 ) is an isomorphism. As a consequence of this, { 𝑒 𝑟 𝑒 𝑠 | 𝑟 < 𝑠 } is a K -basis for 𝔭𝔦𝔫 ( 𝑉 ) . ( ) G R A N D U N I F I E D T H E O RY
Let us make the transition from Lie algebra to Lie group. Using exp and log, we findthat 𝜋 maps a small neighbourgood of 1 ∈ Pin ( 𝑉 ) ∩ Cl ( 𝑉 ) onto the identity componentof O ( 𝑉 ) , i.e. onto SO ( 𝑉 ) . But {± } is contained in the identity component of Pin ( 𝑉 ) ∩ Cl ( 𝑉 ) if 𝑛 ≥
2, since cos 𝑡 + ( 𝑒 𝑒 ) sin 𝑡 for 𝑡 ∈ [ , 𝜋 ] is a path from 1 to − ( 𝑉 ) ∩ Cl ( 𝑉 ) . Thus, 𝜋 − ( SO ( 𝑉 )) = 𝜋 − ( det − ( )) is connected and contained in Cl ( 𝑉 ) . Toshow that 𝜋 − ( det − (− )) is connected and complete the proof, it suffices to produce anelement in 𝜋 − ( det − (− )) multiplication by which will send 𝜋 − ( det − ( )) to 𝜋 − ( det − (− )) .The element 𝑒 will do, for one checks that it lies in Pin ( 𝑉 ) and covers the reflectiondiag (− , , . . . , ) . QED Definition 2.4.12 (The Spin Groups) . We define Spin ( 𝑛 ) as the subgroup 𝜋 − ( det − ) = Pin ( 𝑉 ) ∩ Cl ( 𝑉 ) . It comes with a homomorphism 𝜋 : Spin ( 𝑉 ) → SO ( 𝑉 ) . Remark 2.4.13.
Over R , the maximal torus in Spin ( 𝑚 ) , 𝑚 = 𝑛 or 2 𝑛 +
1, is usually taken toconsist of the elements (cid:206) 𝑛𝑟 = (cid:0) cos 𝑥 𝑟 + ( 𝑒 𝑟 − 𝑒 𝑟 ) sin 𝑥 𝑟 (cid:1) in Cl ( 𝑉 ) , 𝑥 𝑟 ∈ R , which correspondsto (cid:169)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:173)(cid:171) cos 𝑥 − sin 𝑥 sin 𝑥 cos 𝑥 cos 𝑥 − sin 𝑥 sin 𝑥 cos 𝑥 . . . cos 𝑥 𝑛 − sin 𝑥 𝑛 sin 𝑥 𝑛 cos 𝑥 𝑛 (cid:170)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:174)(cid:172) . Here, row 𝑛 + 𝑚 = 𝑛 +
1, and is empty for 𝑚 = 𝑛 . Since Spin ( 𝑉 ) ⊂ Cl ( 𝑉 ) , any Cl ( 𝑉 ) -module is a representation of Spin ( 𝑉 ) , and someextremely important representations of Spin ( 𝑛 ) arise in this way. We study these now. Proposition 2.4.14.
The algebras Cl ( 𝑉 ) and Cl ( 𝑉 ) are semi-simple, so all their representationsare completely reducible. Proof.
Let { 𝑒 𝑖 } be the standard basis in 𝑉 = K 𝑚 , and consider 𝐸 = (cid:110) ± (cid:206) 𝑚𝑗 = 𝑒 𝑖 𝑗 𝑗 | 𝑖 𝑗 = (cid:111) ,a subgroup of order 2 𝑚 + of Cl ( 𝑉 ) , corresponding to the matrices diag (± , . . . , ± ) ofO ( 𝑚 ) . In Cl ( 𝑉 ) consider the subgroup 𝐸 of 2 𝑚 elements with (cid:205) 𝑗 𝑖 𝑗 even; we have 𝐸 ⊂ 𝐸 ⊂ Pin ( 𝑉 ) . Let 𝜈 denote − ∈ Cl ( 𝑉 ) when considered an element of 𝐸 , 𝐸 orPin ( 𝑉 ) ; a module over Cl ( 𝑉 ) gives a representation in which 𝜈 acts as −
1. Conversely,a representation of 𝐸 in which 𝜈 acts as − K ( 𝐸 )/(cid:104) 𝜈 + (cid:105) = Cl ( 𝑉 ) ,where K ( 𝐸 ) is the group ring over K of 𝐸 . The representation theory of Cl ( 𝑉 ) can thus beinferred from that of the finite group 𝐸 , and in particular all representations are completelyreducible. We argue similarly for Cl ( 𝑉 ) , replacing 𝐸 by 𝐸 . QED Proposition 2.4.15. If dim 𝑉 is odd, 𝑚 = 𝑛 + , say, then Cl ( 𝑉 ) has one irrep Δ of degree 𝑛 ,affording a representation Δ of Spin ( 𝑛 + ) with weights (± 𝑥 ± 𝑥 · · · ± 𝑥 𝑛 ) ; there are 𝑛 ofthese weights. If dim 𝑉 = 𝑚 = 𝑛 , then Cl ( 𝑉 ) has two irreps Δ + , Δ − of degree 𝑛 − , affordingrepresentations Δ + , Δ − of Spin ( 𝑛 ) having weights (± 𝑥 ± 𝑥 · · · ± 𝑥 𝑛 ) , with an even number of − signs for Δ + and an odd number of − signs for Δ − ; there are 𝑛 − such weights. If K = C , theseare complex-analytic representations of Spin C ( 𝑚 ) . . 4 . C L I F F O R D A L G E B R A S 29 Proof.
By Schur’s lemma, 𝜈 acts on any irrep as either 1 or -1. The ones in which it actsas 1 are representations of E /(cid:104) 𝜈 (cid:105) , which is an abelian group of order 2 𝑚 − , so there areexactly 2 𝑚 − 𝐸 in which 𝜈 acts as 1. Since the kernel of 𝐸 → 𝐸 /(cid:104) 𝜈 (cid:105) has exactly two elements, the conjugacy classes in 𝐸 are either one element(if the element is central) or two elements ± 𝑔 . For 𝐸 , the centre is {± } if 𝑚 = 𝑛 + (cid:8) ± , ± (cid:206) 𝑛 𝑒 𝑖 (cid:9) if 𝑚 = 𝑛 ; we can see this as follows. If we conjugate 𝑔 = (cid:206) 𝑚 𝑒 𝑖 𝑗 𝑗 with 𝑒 𝑟 𝑒 𝑠 where 𝑖 𝑟 = 𝑖 𝑠 =
0, we change its sign. So if 𝑔 is in the centre, 𝑔 = ± ± 𝑒 𝑒 . . . 𝑒 𝑚 . Thelatter is in the centre only for 𝑚 even.Recalling that the isomorphism classes of irreps (over C ) of a finite group are in a1 : 1 correspondence with the conjugacy clasees, we see that 𝐸 has one (resp. two) moreirreducible class(es) of representation(s) than 𝐸 /(cid:104) 𝜈 (cid:105) if 𝑚 = 𝑛 + 𝐹 ⊂ 𝐸 be the subgroup generated by 𝑒 𝑒 , . . . , 𝑒 𝑟 − 𝑒 𝑟 , . . . 𝑒 𝑛 − 𝑒 𝑛 . This is an abeliangroup of order 2 𝑛 + , so the index of 𝐹 in 𝐸 is 2 𝑛 if 𝑚 = 𝑛 +
1, and 2 𝑛 − if 𝑚 = 𝑛 .Choose now a complex 1-dimensional representation 𝑊 of 𝐹 in which 𝜈 acts as − 𝑒 𝑟 − 𝑒 𝑟 acts as 𝑖 𝜖 𝑟 , 𝜖 𝑟 = ± 𝑖 = √−
1. Then the induced representation Ind 𝐸 𝐹 𝑊 is arepresentation of 𝐸 with degree 2 𝑛 for 𝑚 = 𝑛 + 𝑛 − for 𝑚 − 𝑛 , with 𝜈 acting as −
1. It has a basis 𝑛 + (cid:206) 𝑖 = 𝑖 odd 𝑒 𝑗 𝑖 𝑖 with 𝑛 + (cid:205) 𝑗 𝑖 even, if 𝑚 = 𝑛 + 𝑛 − (cid:206) 𝑖 = 𝑖 odd 𝑒 𝑗 𝑖 𝑖 with 𝑛 − (cid:205) 𝑗 𝑖 even, if 𝑚 = 𝑛 .When 𝑚 = 𝑛 +
1, there are 2 𝑛 choices for 𝑊 (because we have 𝑛 choices for 𝜖 𝑟 ), 𝐸 / 𝐹 permutes them transitively, and by conjugating with 𝑒 𝑟 𝑒 𝑟 + , we can change the signof 𝜖 𝑟 without changing anything else. Each choice appears once in Ind 𝐸 𝐹 𝑊 . We thusget a representation Δ of 𝐸 with character 2 𝑛 at − − 𝑛 at − Δ is an irrep of 𝐸 .When 𝑚 = 𝑛 , there are 2 𝑛 choices for 𝑊 , and under 𝐸 / 𝐹 they fall into two orbits:those with (cid:206) 𝜖 𝑟 =
1, and those with (cid:206) 𝜖 𝑟 = −
1. We can only change the sign of an evennumber of 𝜖 𝑟 , since conjugating with 𝑒 𝑟 𝑒 𝑠 , 𝑟 < 𝑠 , changes the sign of both 𝜖 𝑟 and 𝜖 𝑠 .Hence we have one irrep of 𝐸 which as a representation of 𝐹 contains all the 𝑊 with 𝜖 : = (cid:206) 𝜖 𝑟 =
1, and another containing all the 𝑊 with 𝜖 = (cid:206) 𝜖 𝑟 = −
1. The character ofthese representations is 2 𝑛 − at 1, − 𝑛 − at − 𝑖 𝑛 𝜖 at (cid:206) 𝑛 𝑒 𝑖 , − 𝑖 𝑛 𝜖 at − (cid:206) 𝑛 𝑒 𝑖 , and zeroelsewhere. This is an irreducible character by the orthogonality relations, so these areinequivalent representations. QED Remark 2.4.16.
We calculate the weights of Δ for K = R as follows: the element 𝑛 (cid:214) 𝑟 = (cid:18) cos 2 𝜋 𝑥 𝑟 + ( 𝑒 𝑟 − 𝑒 𝑟 ) sin 2 𝜋 𝑥 𝑟 (cid:19) of the maximal torus acts on Δ with eigenvalues 𝑛 (cid:214) 𝑟 = (cid:18) cos 2 𝜋 𝑥 𝑟 + ( 𝑖 𝜖 𝑟 ) sin 2 𝜋 𝑥 𝑟 (cid:19) = exp (cid:32) 𝜋 𝑖 𝑛 (cid:213) 𝑟 = 𝜖 𝑟 𝑥 𝑟 (cid:33) and weights (cid:205) 𝑛𝑟 = 𝜖 𝑟 𝑥 𝑟 , where 𝜖 𝑟 = ± Definition 2.4.17 (Spinors) . For K = C , the representations Δ , Δ + and Δ − are called spinorrepresentations of the (complex) Clifford algebra. ( ) G R A N D U N I F I E D T H E O RY
Proposition 2.4.18.
The representation Δ of Spin ( 𝑛 + ) is self-dual. The representations Δ + , Δ − of Spin ( 𝑛 ) are self-dual if 𝑛 is even and dual to each other if 𝑛 is odd. Proof.
We have to prove the isomorphism in the various senses of 𝑀 ∗ with 𝑁 , wherewe write 𝑀 , 𝑁 for the spinor representations in the proposition statement. Considerthat by definition, 𝐸 acts on an ℎ in the dual representation 𝑀 ∗ = Hom C ( 𝑀, C ) as ( 𝑔 ℎ )( 𝑚 ) = ℎ ( 𝑔 − 𝑚 ) for 𝑚 ∈ 𝑀 ; generalising this, we see that Cl ( C 𝑚 ) acts on 𝑀 ∗ as ( 𝑎 ℎ )( 𝑚 ) = ℎ (( 𝛾 𝑎 ) 𝑚 ) . From the discussion in the final two paragraphs of the proof ofproposition 2.4.15, it is clear that we have an isomorphism of the representations 𝑁 and 𝑀 ∗ of the finite group 𝐸 , which is an isomorphism of the Cl ( C 𝑚 ) -modules. ButSpin ( 𝑚 ) ⊂ Cl ( C 𝑚 ) , so the isomorphism preserves the action of the elements of Spin ( 𝑚 ) ,provided this action is defined by ( 𝑔 ℎ ) 𝑚 = ℎ ( 𝑔 − 𝑚 ) , which is the usual action. QEDWe now in fact have almost everything we need to discuss the Spin ( ) grand unifiedtheory. But before we do so, we end this section by stating a technical result, requiredto construct E (and also G , in the appendix). In particular, we need to understandhow the representations of the Spin groups behave under certain inclusions. To this end,we first note that the inclusion K 𝑚 ↩ → K 𝑚 + induces an inclusion Cl ( K 𝑚 ) ↩ → Cl ( K 𝑚 + ) ,so we get Spin ( 𝑚 ) ↩ → Spin ( 𝑚 + ) covering the usual map SO ( 𝑚 ) ↩ → SO ( 𝑚 + ) , 𝐴 ↦→ (cid:0) 𝐴
00 1 (cid:1) . We also have, by proposition 2.4.3, that Cl ( K 𝑝 ) ⊗ Cl ( K 𝑞 ) = Cl ( K 𝑝 + 𝑞 ) , which givesSpin ( 𝑝 ) × Spin ( 𝑞 ) → Spin ( 𝑝 + 𝑞 ) . Proposition 2.4.19. (i) Under the inclusions
Spin ( 𝑛 ) ↩ → Spin ( 𝑛 + ) ↩ → Spin ( 𝑛 + ) , we have Δ + Δ + + Δ − Δ Δ − (ii) Under Spin ( 𝑟 ) × Spin ( 𝑠 ) → Spin ( 𝑟 + 𝑠 ) Δ + ⊗ Δ + + Δ − ⊗ Δ − ← (cid:0) Δ + Δ + ⊗ Δ − + Δ − ⊗ Δ + ← (cid:0) Δ − The proofs of these inclusions are found in [4, pp. 23–24]. We will henceforth denote, aswe have here, the direct sum of vector spaces (and representations) by a simple + insteadof an ⊕ . Spin ( ) Extension of SU ( ) Let us revisit the SU ( ) theory. Viewed from a different light, the core idea behind theembedding SU ( ) × SU ( ) ↩ → S ( U ( ) × U ( )) , which subsequently split each irrep of SU ( ) into an isospin and colour piece (each twisted with hypercharge), can be stated as follows:since the Standard Model representation is 32-dimensional, each particle or antiparticle inthe first generation of fermions can be named by a 5-bit code. Roughly speaking, these bitsare the answers to five binary queries. There are subtleties when we answer “yes” to both of the first two questions, or “yes” to more than oneof the last three, but we ignore this problem here; it has no bearing on our argument. . 5 . T H E S P I N ( ) E X T E N S I O N O F S U ( ) • Is the particle isospin up? • Is it isospin down? • Is it red? • Is it green? • Is it blue?This binary code interpretation of the SU ( ) theory requires the dimension of 𝐹 + 𝐹 to be32, and this raises some questions, as we shall see now.At the time of writing, there is no direct experimental evidence for the existence ofthe right-handed neutrino, even though they are extremely desirable theoretically, asthey could account for several phenomena that have no explanation within the StandardModel. The right-handed neutrino 𝜈 𝑅 has a direct bearing on our grand unified theories;in particular, it presents a mystery for the SU ( ) theory. SU ( ) does not require us touse the full 32-dimensional representation Λ ∗ C . It works just as well with the smallerrepresentation Λ C + Λ C + Λ C + Λ C , which is less-aesthetically pleasing, and moreover, clearly does not allow for the existenceof 𝜈 𝑅 . It would be nicer to have a theory that required us to use all of Λ ∗ C ; better still,if our theory were an extension of SU ( ) , our explanation for the arbitrary hyperchargesof the Standard Model particles would live on. The Spin ( ) grand unified theory is anattempt at such an extension; [30] and [33] are the original references for the same.In proposition 2.4.15, we constructed the spinor representations for Spin ( 𝑛 ) , Δ ± , eachof dimension 2 𝑛 − . It is perhaps not immediately apparent from the somewhat technicalproof of that result, but these irreps are intimately related to Λ ∗ C 𝑛 , and we will exploit thisfact to forge a path to the SU ( ) theory.Let 𝑉 be a complex vector space with dim 𝑉 = 𝑛 , equipped with the standard innerproduct (cid:104) , (cid:105) . Write 𝑉 = 𝑊 + 𝑊 (cid:48) , where the 𝑊 ’s are 𝑛 -dimensional isotropic spaces for (cid:104) , (cid:105) . In fact, under (cid:104) , (cid:105) , we can simply take 𝑊 to be spanned by the first 𝑛 standard basisvectors, and 𝑊 (cid:48) by the last 𝑛 . Proposition 2.5.1.
The decomposition 𝑉 = 𝑊 + 𝑊 (cid:48) determines an isomorphism of algebras, Cl ( 𝑉 ) (cid:27) End ( Λ ∗ 𝑊 ) . Proof.
We follow [32, p. 304]. Mapping Cl ( 𝑉 ) to the algebra 𝐸 = End ( Λ ∗ 𝑊 ) is the same asdefining a linear mapping from 𝑉 to 𝐸 , satisfying the relation (2.4.1). That is, we mustconstruct maps 𝑙 : 𝑊 → 𝐸 and 𝑙 (cid:48) : 𝑊 (cid:48) → 𝐸 such that 𝑙 ( 𝑤 ) = = 𝑙 (cid:48) ( 𝑤 (cid:48) ) and 𝑙 ( 𝑤 ) 𝑙 (cid:48) ( 𝑤 (cid:48) ) + 𝑙 (cid:48) ( 𝑤 (cid:48) ) 𝑙 ( 𝑤 ) = (cid:104) 𝑤, 𝑤 (cid:48) (cid:105) , (2.5.1)for any 𝑤 ∈ 𝑊 and 𝑤 (cid:48) ∈ 𝑊 (cid:48) . To do this, we will “deform” the usual wedge product on theexterior algebra (this is sometimes referred to as Clifford multiplication of forms ). For each 𝑤 ∈ 𝑊 , 𝑤 (cid:48) ∈ 𝑊 (cid:48) and 𝜉 ∈ Λ ∗ 𝑊 , define 𝑙 𝑤 ( 𝜉 ) = 𝑤 ∧ 𝜉 ,𝑙 (cid:48) 𝑤 (cid:48) ( 𝜉 ) = 𝜄 𝑤 (cid:48) 𝜉 , For thorough reviews of the current theoretical and phenomenological status of this elusive particle, see[25, 85] and references therein. Recall that a space is isotropic when the chosen symmetric bilinear form restricts to the zero form on it.2 2 . T H E S P I N ( ) G R A N D U N I F I E D T H E O RY where 𝜄 𝑤 (cid:48) : Λ 𝑘 𝑊 → Λ 𝑘 − 𝑊 is the usual contraction by 𝑤 (cid:48) ; on a basis vector it acts as 𝜄 𝑤 (cid:48) ( 𝑤 ∧ · · · 𝑤 𝑘 ) = 𝑘 (cid:213) 𝑗 = (− ) 𝑗 + (cid:104) 𝑤 (cid:48) , 𝑤 𝑗 (cid:105) 𝑤 ∧ · · · ∧ (cid:99) 𝑤 𝑗 ∧ · · · ∧ 𝑤 𝑘 . It is immediately clear that 𝑙 and 𝑙 (cid:48) vanish on their domains, and it is a straightforwardexercise to check that equation (2.5.1) holds. Finally, one confirms that the resulting mapfrom Cl ( 𝑉 ) → End ( Λ ∗ 𝑊 ) is an isomorphism by computing it on a basis set. QEDThe maps 𝑙 and 𝑙 (cid:48) are far more important than they perhaps appear. The first clue isthat if we extend them to all of 𝑉 = C 𝑛 , they are in fact adjoint with respect to the innerproduct induced on Λ ∗ C 𝑛 by (cid:104) , (cid:105) , i.e. for 𝑣 ∈ C 𝑛 , 𝑝, 𝑞 ∈ Λ ∗ C 𝑛 , (cid:104) 𝑝, 𝑙 𝑣 𝑞 (cid:105) = (cid:104) 𝑙 (cid:48) 𝑣 𝑝, 𝑞 (cid:105) . Adjointoperators are the bread and butter of quantum mechanics, so one might ask if these mapshave a physical interpretation; indeed, there is one readily available. In the parlance ofphysics, particles are vectors, so 𝑙 𝑣 = 𝑣 ∧ can be said to “create” a particle of type 𝑣 bywedging; analogously 𝑙 (cid:48) 𝑣 = 𝜄 𝑣 “destroys” a particle of type 𝑣 by contraction. In other words,these maps return, for each 𝑣 , the corresponding creation and annihilation operators . It iscustomary to denote the 𝑛 creation and annihilation operators corresponding to the 𝑛 basis vectors 𝑒 𝑗 of C 𝑛 by 𝑎 ∗ 𝑗 and 𝑎 𝑗 respectively, and we will do so below.Consider now the splitting Λ ∗ 𝑊 = Λ even 𝑊 + Λ odd 𝑊 into the sum of even and oddexterior powers; Cl ( 𝑊 ) clearly respects this splitting. Hence, we deduce that there is anisomorphism Cl ( 𝑉 ) (cid:27) End ( Λ even 𝑊 ) + End ( Λ odd 𝑊 ) . Restricting now to the case 𝑛 =
5, we conclude that since Spin ( ) ⊂ Cl ( C ) , the aboveClifford modules, i.e. the even- and odd-graded powers of the exterior algebra Λ ∗ C ,are representations of Spin ( ) . Moreover, by proposition 2.4.15, they are irreducible.Elements of these two irreps, Δ + and Δ − , are called left- and right-handed Weyl spinors respectively, while elements of their direct sum, Λ ∗ C , are called Dirac spinors .We are tantalisingly close now to the Spin ( ) grand unified theory; there remains butone question. Does the Dirac spinor representation of Spin ( ) extend the representationof SU ( ) on Λ ∗ C ? Or more generally, does the Dirac spinor representation of Spin ( 𝑛 ) ,which we will call 𝜌 (cid:48) , extend the representation of SU ( 𝑛 ) on Λ ∗ C 𝑛 ? Recall that this latterrepresentation 𝜌 : SU ( 𝑛 ) → Λ ∗ C 𝑛 acts as the fundamental representation on Λ C 𝑛 (cid:27) C 𝑛 and respects wedge products. The result that we need is answered in the affirmative by thefollowing theorem, which appears in a classic paper by Atiyah, Bott and Shapiro, whereinthey also founded the abstract theory of Clifford modules [7]. Theorem 2.5.2.
There exists a Lie group homomorphism 𝜓 that makes this triangle commute: SU ( 𝑛 ) Spin ( 𝑛 ) Λ ∗ C 𝑛 𝜓𝜌 𝜌 (cid:48) Proof.
We follow the proof as laid out in [8]. The connected component of the identityin O ( 𝑛 ) is SO ( 𝑛 ) ; since U ( 𝑛 ) is connected and U ( 𝑛 ) ↩ → O ( 𝑛 ) , it follows that there isan inclusion SU ( 𝑛 ) ↩ → U ( 𝑛 ) ↩ → SO ( 𝑛 ) . Passing to Lie algebras, we obtain an inclusion 𝔰𝔲 ( 𝑛 ) ↩ → 𝔰𝔬 ( 𝑛 ) . A homomorphism of Lie algebras gives a homomorphism of thecorresponding simply-connected Lie groups, so we now have a map 𝜓 : SU ( 𝑛 ) → Spin ( 𝑛 ) ;we must check that it makes the above triangle commute. . 5 . T H E S P I N ( ) E X T E N S I O N O F S U ( ) Since all the groups involved are connected, it suffices to check that this diagram 𝔰𝔲 ( 𝑛 ) 𝔰𝔬 ( 𝑛 ) Λ ∗ C 𝑛 d 𝜓 d 𝜌 d 𝜌 (cid:48) (2.5.2)commutes. Since the Dirac representation d 𝜌 (cid:48) is defined in terms of creation andannihilation operators, we should try to express d 𝜌 in this way. A good choice of basis for 𝔰𝔲 ( 𝑛 ) will be extremely helpful in this regard: we pick the so-called generalised Gell-Mannmatrices . Let 𝐸 𝑗𝑘 denote the matrix with 1 in the 𝑗 𝑘 th entry, and 0 everywhere else. Then 𝔰𝔲 ( 𝑛 ) has the basis 𝐸 𝑗𝑘 − 𝐸 𝑘𝑗 for 𝑗 < 𝑘 , 𝑖 ( 𝐸 𝑗𝑘 + 𝐸 𝑘𝑗 ) for 𝑗 > 𝑘 , and 𝑖 ( 𝐸 𝑗𝑗 − 𝐸 𝑗 + ,𝑗 + ) for 𝑗 = , . . . , 𝑛 − 𝐸 𝑗𝑘 ( 𝑒 𝑙 ) = 𝛿 𝑘𝑙 , which is in fact how 𝑎 ∗ 𝑗 𝑎 𝑘 acts on Λ C 𝑛 . So on thisspace at least, we have the simple relationsd 𝜌 ( 𝐸 𝑗𝑘 − 𝐸 𝑘𝑗 ) = 𝑎 ∗ 𝑗 𝑎 𝑘 − 𝑎 ∗ 𝑘 𝑎 𝑗 , d 𝜌 ( 𝑖 ( 𝐸 𝑗𝑘 + 𝑒 𝑘𝑗 )) = 𝑖 ( 𝑎 ∗ 𝑗 𝑎 𝑘 − 𝑎 ∗ 𝑘 𝑎 𝑗 ) , d 𝜌 ( 𝑖 ( 𝐸 𝑗𝑗 − 𝐸 𝑗 + ,𝑗 + )) = 𝑖 ( 𝑎 ∗ 𝑗 𝑎 𝑗 − 𝑎 ∗ 𝑗 + 𝑎 𝑗 + . (2.5.3)We claim that these hold on all of Λ ∗ C 𝑛 . To see this, first recall that 𝜌 preserves wedgeproducts: 𝜌 ( 𝑥 )( 𝑣 ∧ 𝑤 ) = 𝜌 ( 𝑥 ) 𝑣 ∧ 𝜌 ( 𝑥 ) 𝑤 ;differentiating this condition, we see that 𝔰𝔲 ( 𝑛 ) must act as a derivation:d 𝜌 ( 𝑋 )( 𝑣 ∧ 𝑤 ) = d 𝜌 ( 𝑋 ) 𝑣 ∧ 𝑤 + 𝑣 ∧ d 𝜌 ( 𝑋 ) 𝑤 . Since both the derivative and taking wedge products are linear, derivations on Λ ∗ C 𝑛 aredetermined by their action on Λ C 𝑛 ; hence, for equations (2.5.3) to hold on Λ ∗ C 𝑛 , it sufficesto check that all the operators on the right hand side of the equation are derivations. Theannihilation operator is given by contraction, which acts like so on a wedge product: 𝑎 𝑗 ( 𝑣 ∧ 𝑤 ) = 𝑎 𝑗 ( 𝑣 ) ∧ 𝑤 + (− ) 𝑝 𝑣 ∧ 𝑎 𝑗 𝑤 , where 𝑝 is the order of the tensor 𝑣 ; this is almost a derivative, but not quite. On the otherhand, the creation operators act in a completely different way: 𝑎 ∗ 𝑗 ( 𝑣 ∧ 𝑤 ) = 𝑎 ∗ 𝑗 𝑣 ∧ 𝑤 = (− ) 𝑝 𝑣 ∧ 𝑎 ∗ 𝑗 𝑤 , since 𝑎 ∗ 𝑗 acts by wedging with 𝑒 𝑗 , and moving this through 𝑣 introduces 𝑝 minus signs.This combines with the pervious relation to ensure that 𝑎 ∗ 𝑗 𝑎 𝑘 is a derivation for everycombination of 𝑗 and 𝑘 , as can be checked explicitly. Hence, d 𝜌 can be expressed directlyas a sum of creation and annihilation operators. Checking now that the diagram 2.5.2commutes is straightforward (though tedious). QEDThe homomorphism 𝜓 is precisely what allows us to extend the SU ( ) model toSpin ( ) , and makes this square commuteSU ( ) Spin ( ) Λ ∗ C Λ ∗ C 𝜓𝜌 𝜌 (cid:48) (cid:27) ( ) G R A N D U N I F I E D T H E O RY
From theorem 2.3.1 then, we have the final result of this chapter.
Theorem 2.5.3.
Spin ( ) is a grand unified theory, i.e. the following diagram commutes: 𝐺 SM / Z Spin ( ) 𝐹 ⊕ 𝐹 Λ ∗ C 𝜌 (cid:48) (cid:27) hapter 3 The E Grand Unified Theory
A grand unified theory based on the exceptional group E first appeared in a 1976 paper byGürsey, Ramond and Sikivie [40]. The authors were motivated by the fact that E has as amaximal subgroup SU ( ) × SU ( ) × SU ( ) : they took these components to be, respectively,the symmetry groups of the left- and right-handed quarks, and the colour group of thequarks, and considered two assignments of this subgroup into a 27 dimensional irrep ofE . We will not follow their treatment in this chapter, choosing instead to focus on thefollowing “cascade” of theories [11, 42, 48]:E → Spin ( ) → SU ( ) → 𝐺 SM . We will first construct E and E in section 3.1 below. In the process, we will see howthe group Spin ( ) × U ( )/ Z arises naturally as a maximal subgroup of E , which will leadus directly into the proof that E extends the Standard Model in section 3.2. Thereafter, wewill analyse the new fermions that appear in the E theory. E and E We closely follow [4] in this section. Our strategy will be the following: to describe anunknown group 𝐺 , it is useful to find a known subgroup of maximal rank 𝐻 ⊂ 𝐺 and togive an account of 𝐺 / 𝐻 . The main theorem of this section is the following, the proof ofwhich will be in stages.
Theorem 3.1.1.
There exist Lie groups 𝐺 with subgroups 𝐻 as specified in the following table. 𝐺 Rank Dim. Local type of 𝐻 Rank Dim. 𝔤 / 𝔥 as C Rep. Dim. E Spin ( ) × U ( )/ Z Δ + ⊗ 𝜉 + Δ − ⊗ 𝜉 − E Spin ( )/ Z Δ + 𝜉 is the fundamental representation of U ( ) on C . Remark 3.1.2.
We will see in due course that(i) in Spin ( ) × U ( )/ Z , the Z is generated by ( (cid:206) 𝑒 𝑗 , 𝑖 ) .(ii) in Spin ( )/ Z , the Z is generated by (cid:206) 𝑒 𝑖 , andThe first column we will fill is that of dim 𝔤 / 𝔥 , proceeding thereafter to find groups whichhave the required representations of these dimensions. We begin with the construction ofthe Lie algebra of E . See the construction of G in the appendix for a prototypical example.356 3 . T H E E G R A N D U N I F I E D T H E O RY
The Construction of a Lie Algebra of Type E For E , there is no representation of smaller degree than Ad, so let us use this fact. Take 𝐿 + Δ + , where we denote 𝔰𝔭𝔦𝔫 ( ) = 𝐿 , and consider this simultaneously over R and C ; itsdegree is 120 + 2 = 248, as required.For a while we can work with Spin ( 𝑛 ) ; let us try and define a suitable inner producton its Lie algebra 𝐿 . By proposition 2.4.10, 𝐿 ⊂ Cl ( 𝑉 ) has a basis { 𝑒 𝑟 𝑒 𝑠 | 𝑟 < 𝑠 } and Δ + is a representation of Spin ( 𝑛 ) , and hence of 𝐿 over R , i.e. for all 𝑎 ∈ 𝐿 , 𝑢 ∈ Δ + ,we have [ 𝑎, 𝑢 ] ∈ Δ + satisfying the Jacobi identity, where the multiplication is Cliffordmultiplication. Assume now that 2 𝑛 ≡ Δ + as a real representationof Spin ( 𝑛 ) . Choose ( , ) Δ + : Δ + ⊗ Δ + → R , a symmetric bilinear non-zero map, invariantunder Spin ( 𝑛 ) , i.e. ( 𝑔𝑢, 𝑔𝑣 ) = ( 𝑢, 𝑣 ) for 𝑔 ∈ Spin ( 𝑛 ) , 𝑢, 𝑣 ∈ Δ + ; the linearised formof this invariance is ([ 𝑎, 𝑢 ] , 𝑣 ) + ( 𝑢, [ 𝑎, 𝑣 ]) =
0. Since Spin ( 𝑛 ) is the double cover ofSO ( 𝑛 ) , we have 𝐿 = 𝔰𝔭𝔦𝔫 ( 𝑛 ) (cid:27) 𝔰𝔬 ( 𝑛 ) , which is the space of skew-symmetric matrices;on matrices the form ( 𝐴, 𝐵 ) = tr 𝐴𝐵 is symmetric, bilinear and real on real matrices. Theinvariance property for 𝑋 ∈ GL ( 𝑛 ) is tr 𝑋 𝐴𝑋 − 𝑋 𝐵𝑋 − = tr 𝐴𝐵 ; if we set 𝑋 = Id + 𝑡𝑌 and pass to the limit, we obtain the linearised version: tr ([ 𝑌, 𝐴 ] 𝐵 + 𝐴 [ 𝑌, 𝐵 ]) =
0; hence ([ 𝑌, 𝐴 ] , 𝐵 ) + ( 𝐴, [ 𝑌, 𝐵 ]) =
0. Under this identification, 𝑒 𝑟 𝑒 𝑠 corresponds to the matrix withall entries zero except in positions ( 𝑟, 𝑠 ) and ( 𝑠, 𝑟 ) where we have respectively − ( 𝑒 𝑟 𝑒 𝑠 , 𝑒 𝑟 𝑒 𝑠 ) = −
8, so to remove this undesirablefactor, we set ( 𝐴, 𝐵 ) 𝐿 : = −
18 tr 𝐴𝐵 so that ( 𝑒 𝑟 𝑒 𝑠 , 𝑒 𝑡 𝑒 𝑢 ) 𝐿 = 𝛿 𝑟𝑡 𝛿 𝑠𝑢 . We now transpose the action 𝐿 ⊗ Δ + → Δ + to get a map Δ + ⊗ Δ + → 𝐿 . Lemma 3.1.3.
For all 𝑢, 𝑣 ∈ Δ + , there is a unique [ 𝑢, 𝑣 ] ∈ 𝐿 such that ( 𝑎, [ 𝑢, 𝑣 ]) 𝐿 = ([ 𝑎, 𝑢 ] , 𝑣 ) Δ + for all 𝑎 ∈ 𝐿 and [ 𝑢, 𝑣 ] is bilinear in 𝑢, 𝑣 . Furthermore, if 𝑣, 𝑤 ∈ C ⊗ Δ + are such that 𝑒 𝑞 − 𝑒 𝑞 𝑣 = 𝑖𝑣 for all 𝑞 (cid:18) corresponding to a weight 𝑛 (cid:213) 𝑥 𝑖 (cid:19) , 𝑒 𝑞 − 𝑒 𝑞 𝑤 = − 𝑖𝑤 for all 𝑞 (cid:18) corresponding to a weight − 𝑛 (cid:213) 𝑥 𝑖 (cid:19) ,and ( 𝑣, 𝑤 ) = , then(i) [ 𝑣, 𝑤 ] = 𝑖 ( 𝑒 𝑒 + 𝑒 𝑒 + · · · 𝑒 𝑛 − 𝑒 𝑛 ) ;(ii) [ 𝑒 𝑞 𝑒 𝑟 𝑣, 𝑤 ] = (cid:0) 𝑒 𝑞 − + 𝑖𝑒 𝑞 (cid:1) ( 𝑒 𝑟 − + 𝑖𝑒 𝑟 ) , 𝑞 < 𝑟 ;(iii) [ 𝑒 𝑞 𝑒 𝑞 · · · 𝑒 𝑞 𝑚 𝑣, 𝑤 ] = if 𝑚 > and 𝑞 < 𝑞 < · · · 𝑞 𝑚 . Proof.
Clearly, ([ 𝑎, 𝑢 ] , 𝑣 ) Δ + is a linear function of 𝑎 . Since the inner product on 𝐿 isnon-singular, we must have ([ 𝑎, 𝑢 ] , 𝑣 ) Δ + = ( 𝑎, 𝑏 ) 𝐿 for some 𝑏 = [ 𝑢, 𝑣 ] ∈ 𝐿 . Since ([ 𝑎, 𝑢 ] , 𝑣 ) Δ + is bilinear in 𝑢 and 𝑣 , so is 𝑏 . We proceed to derive the explicit formulae. All the innerproducts in the following are over Δ + .(i) First we have ( 𝑒 𝑞 − 𝑒 𝑞 𝑣, 𝑤 ) = ( 𝑖𝑣, 𝑤 ) = 𝑖 and ( 𝑒 𝑟 𝑒 𝑠 𝑣, 𝑤 ) = 𝑒 𝑟 𝑒 𝑠 is not one of thebasis elements 𝑒 𝑞 − 𝑒 𝑞 . Thus [ 𝑣, 𝑤 ] is paired to 𝑖 if 𝑎 = 𝑒 𝑞 − 𝑒 𝑞 and to 0 for all theother basis elements.(ii) For simplicity of notation, consider [ 𝑒 𝑒 𝑣, 𝑤 ] . Then ( 𝑒 𝑟 𝑒 𝑠 𝑒 𝑒 𝑣, 𝑤 ) = ( 𝑟, 𝑠 ) = ( , ) , ( , ) , ( , ) or ( , ) , when we get respectively 1 , 𝑖, 𝑖, −
1. This yields [ 𝑒 𝑒 𝑣, 𝑤 ] = 𝑒 𝑒 + 𝑖𝑒 𝑒 + 𝑖𝑒 𝑒 − 𝑒 𝑒 = ( 𝑒 + 𝑖𝑒 )( 𝑒 + 𝑖𝑒 ) , as desired. (Notice that 𝑒 𝑒 𝑣 has weight (− 𝑥 − 𝑥 + 𝑥 + · · · + 𝑥 𝑛 ) while 𝑤 has weight (− 𝑥 − · · · − 𝑥 𝑛 ) sothat [ 𝑒 𝑒 𝑣, 𝑤 ] must have weight − 𝑥 − 𝑥 . In fact, 𝑒 + 𝑖𝑒 has weight − 𝑥 and 𝑒 + 𝑖𝑒 has weight − 𝑥 .) . 1 . T H E CO N S T RU C T I O N O F E A N D E (iii) It suffices to note that for all 𝑒 𝑟 𝑒 𝑠 , 𝑟 < 𝑠 , we have ( 𝑒 𝑟 𝑒 𝑠 𝑒 𝑞 𝑒 𝑞 · · · 𝑒 𝑞 𝑚 𝑣, 𝑤 ) =
0. QED
Remark 3.1.4.
Note that the map [ , ] : Δ + ⊗ Δ − → 𝐿 is invariant under Spin ( 𝑛 ) because everything in the construction is invariant under Spin ( 𝑛 ) . The linearised form ofinvariance, i.e. invariance under 𝐿 , is [ 𝑎, [ 𝑢, 𝑣 ]] = [[ 𝑎, 𝑢 ] , 𝑣 ] + [ 𝑢, [ 𝑎, 𝑣 ]] ;this is established as follows. It is sufficient to show that for all 𝑏 ∈ 𝐿 , (− 𝑏, [ 𝑎, [ 𝑢, 𝑣 ]]) 𝐿 + ( 𝑏, [[ 𝑎, 𝑢 ] , 𝑣 ]) 𝐿 + ( 𝑏, [ 𝑢, [ 𝑎, 𝑣 ]]) 𝐿 is zero. This expression, using the invariance of ( , ) 𝐿 under 𝐿 and the definition of [ 𝑢, 𝑣 ] ,is equivalent to ([ 𝑎, 𝑏 ] , [ 𝑢, 𝑣 ]) 𝐿 + ([ 𝑏, [ 𝑎, 𝑢 ]] , 𝑣 ) 𝐿 + ([ 𝑏, 𝑢 ] , [ 𝑎, 𝑣 ]) 𝐿 ;the invariance of ( , ) Δ + means that this in turn can be written as ([[ 𝑎, 𝑏 ] , 𝑢 ] , 𝑣 ) Δ + + ([ 𝑏, [ 𝑎, 𝑢 ]] , 𝑣 ) Δ + − ([ 𝑎, [ 𝑏, 𝑢 ]] , 𝑣 ) Δ + = , where we used the properties of the action of 𝐿 on Δ + .We now proceed to give 𝐿 + Δ + the inner product with 𝐿 and Δ + orthogonal, and ( 𝑎 + 𝑢, 𝑏 + 𝑣 ) = ( 𝑎, 𝑏 ) 𝐿 + ( 𝑢, 𝑣 ) Δ + for all 𝑎, 𝑏 ∈ 𝐿 and 𝑢, 𝑣 ∈ Δ + . The Lie bracket [ 𝑎, 𝑢 ] is as in 𝐿 ; [ 𝑎, 𝑢 ] as the action of 𝐿 on Δ + satisfies [ 𝑢, 𝑎 ] = −[ 𝑎, 𝑢 ] , and [ 𝑢, 𝑣 ] as in 3.1.3. Theorem 3.1.5. If 𝑛 = , 𝐿 + Δ + becomes a Lie algebra with an invariant inner product. Proof (Sketch).
The inner product is invariant under 𝐿 by definition, and under Δ + by thedefinition of [ 𝑢, 𝑣 ] . We need to prove anti-commutativity and the Jacobi identity.Clearly, [ 𝑎, 𝑏 ] = −[ 𝑏, 𝑎 ] since 𝐿 is a Lie algebra, and we define [ 𝑢, 𝑎 ] to be −[ 𝑎, 𝑢 ] . Tosee that [ 𝑢, 𝑣 ] = −[ 𝑣, 𝑢 ] , observe that for all 𝑎 ∈ 𝐿 , 𝑢, 𝑣 ∈ Δ + , we have ( 𝑎, [ 𝑢, 𝑣 ] + [ 𝑣, 𝑢 ]) 𝐿 = ([ 𝑎, 𝑢 ] , 𝑣 ) 𝐿 + ([ 𝑎, 𝑣 ] , 𝑢 ) 𝐿 = ([ 𝑎, 𝑢 ] , 𝑣 ) 𝐿 + ( 𝑢, [ 𝑎, 𝑣 ]) 𝐿 = , where the penultimate equality follows from the symmetry of ( , ) 𝐿 , and the last one fromthe invariance of ( , ) 𝐿 under 𝐿 . Since this is true for all 𝑎 , we have [ 𝑢, 𝑣 ] + [ 𝑣, 𝑢 ] = • Three variables in 𝐿 , none in Δ + . The identity will hold here since 𝐿 is a Lie algebra. • Two variables in 𝐿 , one in Δ + . We have [[ 𝑎, 𝑏 ] , 𝑢 ] = [ 𝑎, [ 𝑏, 𝑢 ]] − [ 𝑏, [ 𝑎, 𝑢 ]] =
0. Thus [ 𝑎, [ 𝑏, 𝑢 ]] + [ 𝑢, [ 𝑎, 𝑏 ]] + [ 𝑏, [ 𝑢, 𝑎 ]] = • One variable in 𝐿 , two in Δ + . Here, by the invariance of the bracket under 𝐿 , [ 𝑎, [ 𝑢, 𝑣 ]] = [[ 𝑎, 𝑢 ] , 𝑣 ] + [ 𝑢, [ 𝑎, 𝑣 ]] , which leads to the Jacobi identity. • All three variables in Δ + . This is where one needs the fact that 𝑛 =
8. Reference[4, pp. 40–42] argues this case in full detail, setting down a general procedure forchecking identities of this type by using symmetry. QED G R A N D U N I F I E D T H E O RY
The Construction of a Lie Group of Type E Our construction of the simple, connected, compact Lie group with Lie algebra of type E proceeds according to the following steps.(i) Take the Lie algebra 𝐿 + Δ + (over R or C ).(ii) Take the group of automorphisms of this Lie algebra; this is closed subgroup ofGL ( 𝐿 + Δ + ) preserving the Lie bracket.(iii) Take the identity component and call it E . (In fact, the result of step 2 is alreadyconnected.)All our constructions are invariant under Spin ( ) , over R or C , so we get a mapSpin ( ) → Aut ( 𝐿 + Δ + ) , and since Spin ( ) is connected, we get a homomorphism intoE . To find the kernel, note that 𝑒 𝑒 · · · 𝑒 ∈ Spin ( ) acts as 𝑖 = Δ + . It covers − Id ∈ SO ( ) , so it acts as − R but it acts as 1 on 𝐿 . Therefore it acts as 1 on 𝐿 + Δ + . This and the identity are the only elements which act as 1 on 𝐿 + Δ + , so we get anembedding Spin ( )/ Z → E . We now check that E has the required properties.Let 𝐴 be a finite dimensional algebra over R or C (for example, a Lie algebra) and letAut ( 𝐴 ) be the group of automorphism of 𝐴 , that is, linear bijections 𝛼 : 𝐴 → 𝐴 such that 𝛼 ( 𝑎𝑏 ) = 𝛼 ( 𝑎 ) 𝛼 ( 𝑏 ) . Then Aut ( 𝐴 ) is a closed subgroup of GL ( 𝐴 ) , hence a Lie group. Definition 3.1.6.
A linear map 𝛿 : 𝐴 → 𝐴 is a derivation if 𝛿 ( 𝑎𝑏 ) = ( 𝛿 𝑎 ) 𝑏 + 𝑎 𝛿 ( 𝑏 ) .The commutator [ 𝛿 , 𝛿 (cid:48) ] is a derivation if 𝛿 and 𝛿 (cid:48) are, so the derivations form a Liealgebra, 𝔡𝔢𝔯 ( 𝐴 ) . We then have Lemma 3.1.7.
The Lie algebra of
Aut ( 𝐴 ) , 𝔞𝔲𝔱 ( 𝐴 ) , is the algebra of derivations of 𝐴 . Proof.
First we show that 𝔞𝔲𝔱 ( 𝐴 ) ⊂ 𝔡𝔢𝔯 ( 𝐴 ) . To see this, take a short curve in 𝔞𝔲𝔱 ( 𝐴 ) , 𝛼 𝑡 = + 𝛾 𝑡 starting at the identity. Then 𝛼 𝑡 ( 𝑎𝑏 ) = 𝛼 𝑡 ( 𝑎 ) 𝛼 𝑡 ( 𝑏 ) , which gives us 𝛾 ( 𝑎𝑏 ) = 𝛾 ( 𝑎 ) 𝑏 + 𝑎 𝛾 ( 𝑏 ) , so 𝛾 ∈ 𝔞𝔲𝔱 ( 𝐴 ) is a derivation.For the reverse inclusion, consider a derivation 𝛿 : 𝐴 → 𝐴 ; we have (by induction) thestandard formula 𝛿 𝑛 ( 𝑎𝑏 ) = (cid:205) 𝑖 + 𝑗 = 𝑛 (cid:0) 𝑛𝑖 (cid:1) ( 𝛿 𝑖 𝑎 )( 𝛿 𝑗 𝑏 ) . Define now 𝛼 𝑡 : 𝐴 → 𝐴 by 𝛼 𝑡 = (cid:205) ∞ 𝑛 = 𝑡 𝑛 𝑛 ! 𝛿 𝑛 .Then 𝛼 𝑡 ( 𝑎𝑏 ) = (cid:213) 𝑖,𝑗 𝑡 𝑖 + 𝑗 𝑖 ! 𝑗 ! ( 𝛿 𝑖 𝑎 )( 𝛿 𝑗 𝑏 ) = ( 𝛼 𝑡 𝑎 )( 𝛼 𝑡 𝑏 ) , so 𝛼 𝑡 ∈ Aut ( 𝐴 ) , and the tangent vector to 𝛼 is 𝛿 ∈ 𝔞𝔲𝔱 ( 𝐴 ) . QEDThe most familiar example of a derivation is the map ad 𝑥 ( 𝑦 ) = [ 𝑥, 𝑦 ] for 𝑥 ∈ 𝐴 . It iseasy to see that we further have the formula [ ad 𝑥 , 𝛿 ] = ad 𝑥 𝛿 for 𝛿 ∈ 𝔡𝔢𝔯 ( 𝐴 ) . The next resultsays that ad 𝑥 is an isomorphism if the Killing form is non-degenerate. Lemma 3.1.8 (Zassenhaus) . If 𝐴 is a Lie algebra with non-degenerate Killing form, then itsalgebra of derivations is 𝐴 . Proof.
To see that ad 𝑥 is injective, suppose that ad 𝑥 = [ 𝑥, 𝑦 ] = 𝑦 ∈ 𝐴 . Then [ 𝑤, [ 𝑥, 𝑦 ]] = 𝑦 , i.e. [ 𝑤, [ 𝑥, ·]] is the zero function, so tr [ 𝑤, [ 𝑥, ·]] = ( 𝑤, 𝑥 ) 𝐾 = 𝑤 . This implies that 𝑥 =
0, as the Killing form is non-degenerate.To prove that ad 𝑥 is surjective , let us first identify ad 𝑥 with 𝑥 , and so extend the Killingform from elements 𝑥 to derivations. Now fix a derivation 𝛿 ; the linear map 𝑥 ↦→ tr ( ad 𝑥 ) 𝛿 is a then a linear mapping of 𝐴 into C , i.e. an element of the dual vector space 𝐴 ∗ ; since ( , ) 𝐾 We follow [52, p. 74]. . 1 . T H E CO N S T RU C T I O N O F E A N D E is non-degenerate, it follows that there exists an element 𝑑 ∈ 𝐴 such that ( 𝑑, 𝑥 ) 𝐾 = tr ( ad 𝑥 ) 𝛿 for all 𝑥 ∈ 𝐴 . Let us denote 𝜕 : = 𝛿 − ad 𝑑 . Then,tr ( ad 𝑥 ) 𝜕 = tr ( ad 𝑥 ) 𝛿 − tr ad 𝑥 ad 𝑑 = tr ( ad 𝑥 ) ◦ 𝛿 − ( 𝑑, 𝑥 ) 𝐾 = . Now consider, for 𝑥, 𝑦 ∈ 𝐴 , ( 𝑥 𝜕 , 𝑦 ) 𝐾 = tr ad 𝑥 𝜕 ad 𝑦 = tr [ ad 𝑥 , 𝜕 ] ad 𝑦 = tr (cid:0) ( ad 𝑥 ) 𝜕 ad 𝑦 − 𝜕 ad 𝑥 ad 𝑦 (cid:1) = tr ( 𝜕 ad 𝑦 ad 𝑥 − 𝜕 ad 𝑥 ad 𝑦 ) = tr 𝜕 [ ad 𝑦 , ad 𝑥 ] = tr 𝜕 ad 𝑦𝑥 = , by the above result. Since ( , ) 𝐾 is non-degenerate, this implies that 𝜕 =
0; hence 𝛿 = ad 𝑑 for some 𝑑 ∈ 𝐴 . QED Lemma 3.1.9.
The Killing form on 𝐿 + Δ + is non-singular: indeed, ( 𝑥, 𝑦 ) 𝐾 = − ( 𝑥, 𝑦 ) . Proof.
Both ( , ) 𝐾 and ( , ) are invariant under Spin ( ) and 𝐿 and Δ + are irreps of Spin ( ) which are not dual to one another. We can define a map 𝑓 : 𝐴 → 𝐴 , where 𝐴 = 𝐿 + Δ + ,by ( 𝑥, 𝑦 ) 𝐾 = ( 𝑓 𝑥, 𝑦 ) for all 𝑦 ∈ 𝐴 ; 𝐴 splits as a sum of eigenspaces of 𝑓 invariant underSpin ( ) . Thus ( 𝑎 + 𝑢, 𝑏 + 𝑣 ) 𝐾 = 𝜆 ( 𝑎, 𝑏 ) + 𝜇 ( 𝑢, 𝑣 ) . But ( , ) 𝐾 and ( , ) are invariant under 𝐴 and 𝐴 is an irrep of 𝐴 , for the only possible subspaces closed under 𝐿 are 𝐿 and Δ + andthey are not closed under Δ + . Thus 𝜆 = 𝜇 .To find 𝜆 , we calculate ( 𝑒 𝑒 , 𝑒 𝑒 ) 𝐾 = tr ( 𝑧 ↦→ [ 𝑒 𝑒 , [ 𝑒 𝑒 , 𝑧 ]]) = tr ( 𝛿 ) ,say. Nowfrom lemma 3.1.3, we have [ 𝑒 𝑒 , 𝑒 𝑒 ] = , [ 𝑒 𝑒 , 𝑒 𝑒 𝑟 ] = 𝑒 𝑒 𝑟 , [ 𝑒 𝑒 , 𝑒 𝑒 𝑟 ] = − 𝑒 𝑒 𝑟 and [ 𝑒 𝑒 , 𝑒 𝑟 𝑒 𝑠 ] = < 𝑟 < 𝑠 . So 𝛿 ( 𝑒 𝑒 ) = 𝛿 ( 𝑒 𝑒 𝑟 ) = − 𝑒 𝑒 𝑟 , 𝛿 ( 𝑒 𝑒 𝑟 ) = − 𝑒 𝑒 𝑟 , 𝛿 ( 𝑒 𝑟 𝑒 𝑠 ) = 𝐿 𝛿 = − Δ + the action is Clifford multiplication, so [ 𝑒 𝑒 , [ 𝑒 𝑒 , 𝑢 ]] = 𝑒 𝑒 𝑒 𝑒 = − 𝑢 , so tr Δ + ( 𝛿 ) = −
128 and we have tr 𝐴 ( 𝛿 ) = − 𝜆 = −
240 since ( 𝑒 𝑒 , 𝑒 𝑒 ) =
1. QED
Corollary 3.1.10.
With the above constructions, 𝔢 = 𝐿 + Δ + . Proof.
Immediate from the preceding three lemmas. QED
The Construction of E The map Spin ( ) × Spin ( ) → Spin ( ) givesSpin ( ) Spin ( ) × Spin ( ) Spin ( ) 𝐸 SU ( ) U ( ) SO ( ) Consider the centraliser of the image of this SU ( ) in E ; we will call its identity componentE . Note that since the identity component of a topological group is a closed (and normal)subgroup, the group obtained, E , is automatically compact. We proceed to check that ithas the subgroup of type Spin ( ) × U ( )/ Z , as claimed in theorem 3.1.1.Consider the diagramsSpin ( ) × Spin ( ) → Spin ( ) → E The centraliser of a subgroup 𝑆 ⊂ 𝐺 is the set of elements in 𝐺 which commute with 𝑆 .0 3 . T H E E G R A N D U N I F I E D T H E O RY and ( 𝑧, 𝑔 ) 𝑆 × SU ( ) Spin ( )( 𝑧 , 𝑔 ) 𝑆 × SU ( ) U ( ) SO ( ) where the maps are the obvious ones. We have Spin ( ) × 𝑆 × SU ( ) → Spin ( ) → 𝐸 where Spin ( ) × 𝑆 centralises the (image of) SU ( ) in E . It remains to identify the kernelof Spin ( ) × 𝑆 → 𝐸 .The kernel of Spin ( ) → E is Z generated by 𝑒 𝑒 · · · 𝑒 and the kernel of Spin ( ) × Spin ( ) → Spin ( ) is Z generated by (− , − ) . Thus ker ( Spin ( ) × Spin ( ) → E ) hasfour elements, generated by ( 𝑒 · · · 𝑒 , 𝑒 · · · 𝑒 ) and (− , − ) , so this kernel is Z . To seethat all four elements lie in 𝑆 , note that 𝑆 is the image of 𝑡 ↦→ ( cos 𝑡 + ( 𝑒 𝑒 ) sin 𝑡 )( cos 𝑡 +( 𝑒 𝑒 ) sin 𝑡 )( cos 𝑡 + ( 𝑒 𝑒 ) sin 𝑡 ) , and for 𝑡 = 𝜋 / , 𝜋 , 𝜋 /
2, this image goes through 𝑒 𝑒 · · · 𝑒 , −
1, and − 𝑒 𝑒 · · · 𝑒 respectively, corresponding to the points 𝑖, − − 𝑖 of 𝑆 in C . This completes the check of subgroups mentioned in theorem 3.1.1. Identification of 𝔢 Call the subgroups Spin ( ) × U ( )/ Z and SU ( ) 𝐻 and 𝐾 respectively. We know that 𝔢 = 𝔰𝔭𝔦𝔫 ( ) + Δ + as a representation of Spin ( ) , and we have a subgroup 𝐻 × 𝐾 mappinginto Spin ( ) . We wish to determine the centraliser E of 𝐾 in E , so we write 𝔢 as arepresentation of 𝐻 × 𝐾 , and take the part fixed under 𝐾 . This is 𝔢 , and we regard it as arepresentation of 𝐻 .When we restrict from Spin ( ) to Spin ( ) × Spin ( ) , 𝔰𝔭𝔦𝔫 ( ) restricts to 𝔰𝔭𝔦𝔫 ( ) + 𝔰𝔭𝔦𝔫 ( ) + Λ ⊗ Λ , where we have introduced the notation Λ 𝑛 : = Λ ( K 𝑛 ) . We can see thisas follows: since Spin ( ) is the double cover of SO ( 𝑛 ) , it shares its Lie algebra 𝔰𝔬 ( 𝑛 ) , theskew-symmetric 𝑛 × 𝑛 matrices; hence, 𝔰𝔬 ( ) can be decomposed into a block-diagonal 𝔰𝔬 ( ) + 𝔰𝔬 ( ) plus a leftover 10 × Λ ⊗ Λ . On the otherhand, Δ + restricts to Δ + ⊗ Δ + + Δ − ⊗ Δ − by proposition 2.4.19.Recall that we defined 𝜉 to be the fundamental representation of 𝑆 = U ( ) . Then, onrestricting Spin ( ) to 𝑆 × SU ( ) under our map 𝑆 × SU ( ) → Spin ( ) , we find that 𝔰𝔭𝔦𝔫 ( ) restricts to 𝔲 ( ) + 𝔰𝔲 ( ) + ( 𝜉 ⊗ Λ + 𝜉 − ⊗ Λ ) . By looking at weights, we see that Δ + , Δ − and Λ restrict respectively to 𝜉 ⊗ + 𝜉 − ⊗ Λ , 𝜉 − ⊗ + 𝜉 ⊗ Λ and 𝜉 ⊗ Λ + 𝜉 − ⊗ Λ .Putting this all together gives 𝔢 = 𝔰𝔭𝔦𝔫 ( ) + 𝔲 ( ) + 𝔰𝔲 ( )+ 𝜉 ⊗ Λ + 𝜉 − ⊗ Λ + Λ ⊗ Λ + Λ ⊗ 𝜉 − ⊗ Λ + Δ + ⊗ 𝜉 ⊗ + Δ + ⊗ 𝜉 − ⊗ Λ + Δ − ⊗ 𝜉 − ⊗ + Δ − ⊗ 𝜉 ⊗ Λ , where 𝔢 is the part on which SU ( ) acts trivially: 𝔢 = 𝔰𝔭𝔦𝔫 ( ) + 𝔲 ( ) + ( Δ + ⊗ 𝜉 + Δ − ⊗ 𝜉 − ) . This completes the proof of theorem 3.1.1.Finally, note that we have a map E × SU ( ) → E , so we may consider 𝔢 as arepresentation of E × SU ( ) . Regarded thus, 𝔢 = 𝔢 + 𝔰𝔲 ( ) + ( 𝜉 − + Λ ⊗ 𝜉 + Δ + ⊗ 𝜉 − ) ⊗ Λ + ( 𝜉 + Λ ⊗ 𝜉 − + Δ − ⊗ 𝜉 ) ⊗ Λ ; In proposition A.1, we show that Spin ( ) (cid:27) SU ( ) . Our strategy here is as in the proof of theorem A.5. The roots of Spin ( ) which are not roots of 𝑆 × SU ( ) are ±( 𝑥 𝑖 + 𝑥 𝑗 ) . . 2 . T H E E E X T E N S I O N O F S P I N ( ) this leads to a result of paramount importance for us. Corollary 3.1.11. E has two representations, whose restrictions to Spin ( ) × U ( ) are respec-tively 𝜉 − + Λ ⊗ 𝜉 + Δ + ⊗ 𝜉 − and 𝜉 + Λ ⊗ 𝜉 − + Δ − ⊗ 𝜉 . These are of degree 27 and complexconjugate. E Extension of
Spin ( ) We only need a few more things to be able to write down a theorem for E as a grandunified theory. Firstly, we have not shown that the above 27-dimensional representationsof E are irreducible. This is in fact the case, but the proof of this result is unfortunatelyquite involved, and we will omit it in this paper; the interested reader is referred to [4,Ch. 11].The second thing that we need to check is that these representations, call them 𝑁 and 𝑁 , are unitary. This seems problematic, since we have no direct description of them; theonly thing we know is their dimension, and how they reduce to Spin ( ) × U ( ) → E .Fortunately, there is a way to circumvent this difficulty. We have used several timesalready that an equivalent charecterisation of a unitary representation 𝑉 of a group 𝐺 isthe requirement that the action of 𝐺 on 𝑉 is an isometry—indeed, this is sometimes takento be the definition; with this in mind, we have the following handy result, often referredto as Weyl’s unitarian trick . It requires the notion of a
Haar measure , which we do not definehere; see [19, Ch. 1.5].
Proposition 3.2.1.
Any representation 𝑉 of a compact group 𝐺 possesses a 𝐺 -invariant innerproduct. Proof (Sketch).
Let 𝑏 : 𝑉 × 𝑉 → C be any inner product, and define 𝑐 ( 𝑢, 𝑣 ) : = ∫ 𝐺 𝑏 ( 𝑔𝑢, 𝑔𝑣 ) d 𝑔 , where the integral is normalised. 𝑐 : 𝑉 × 𝑉 → C is then linear in 𝑢 , conjugate linear in 𝑣 , 𝐺 -invariant since the integral is left-invariant, and positive definite since the integral of apositive continuous function is positive. QED 𝑁 + 𝑁 endowed with this natural E -invariant inner product is thus a direct sumof unitary irreps of the compact group E . To extend theorem 2.5.3 and prove that E is a grand unified theory however, we need to check something still further: we needa homomorphism Δ + + Δ − → 𝑁 + 𝑁 as unitary representations of Spin ( ) and E respectively. But since Spin ( ) ↩ → Spin ( ) × U ( ) → E , and we know how 𝑁 + 𝑁 restricts to Spin ( ) × U ( ) , it suffices to produce a homomorphism between Δ + + Δ − and these restricted representations. But first, let us run through our usual checklist:the restricted representations, as the direct sum of irreps, are clearly irreps. Are theyunitary? (i) 𝜉 was defined to be the fundamental representation of the unitary group U ( ) ,so there is nothing to check here. (ii) From proposition 2.5.1, the spinor representationscan be seen to be unitary: recall that these are defined via the creation and annihilationoperators, which are adjoint; we therefore have ( 𝑙 † 𝑣 𝑙 𝑣 )( 𝜓 ) = 𝜄 𝑣 ( 𝑣 ∧ 𝜓 ) = Id 𝜓 , so this is We use the word “restriction” here a little loosely. What we mean is that we obtain a representation onSpin ( ) × U ( ) as it is homomorphic to the subgroup Spin ( ) × U ( )/ Z of E . We will pick up this point inthe next section.2 3 . T H E E G R A N D U N I F I E D T H E O RY indeed a unitary representation. Lastly, (iii) Spin ( ) × U ( ) (cid:51) ( 𝑠, 𝑢 ) acts unitarily on thecomplex representation Λ ⊗ 𝜉 (cid:51) 𝑎 ⊗ 𝑝, 𝑏 ⊗ 𝑞 : (cid:104) 𝑠 · 𝑎 ⊗ 𝑢 · 𝑝, 𝑠 · 𝑏 ⊗ 𝑢 · 𝑞 (cid:105) Λ ⊗ 𝜉 = (cid:104) 𝑠 · 𝑎, 𝑠 · 𝑏 (cid:105) Λ (cid:104) 𝑢 · 𝑝, 𝑢 · 𝑞 (cid:105) 𝜉 = (cid:104) 𝑎, 𝑏 (cid:105) Λ (cid:104) 𝑝, 𝑞 (cid:105) 𝜉 = (cid:104) 𝑎 ⊗ 𝑝, 𝑏 ⊗ 𝑞 (cid:105) Λ ⊗ 𝜉 , where we simply used the definitions of the tensor product of representations and Hilbertspaces, and the fact that Spin ( ) and U ( ) are each isometries on these representations.So, (i), (ii) and (iii), together with the fact that the tensor product of unitary representationsis again unitary, means that we are done, and can write down the following commutingdiagram: Spin ( ) Spin ( ) × U ( ) E Δ + + Δ − ( Δ + ⊗ 𝜉 − ) + ( Δ − ⊗ 𝜉 ) + · · · 𝑁 + 𝑁 𝜌 (cid:48) We have but one final check. Recall that the homomorphism Spin ( ) × U ( ) → E hasthe kernel Z ; it is hence incumbent on us to verify, just as we did for the SU ( ) theory, thatthis kernel acts trivially on every fermion. Explicitly, the four elements of the kernel are { 𝑘 , 𝑘 , 𝑘 , 𝑘 } : = (cid:8) ( , ) , ( (cid:206) 𝑒 𝑗 , 𝑖 ) , (− , − ) , (− (cid:206) 𝑒 𝑗 , − 𝑖 ) (cid:9) ;let us begin with the easiest pieces of the restrictions of 𝑁 and 𝑁 . For 𝜉 − and 𝜉 , thereis nothing to check for Spin ( ) , and since the U ( ) components of the 𝑘 𝑖 ’s are preciselythe fourth roots of unity, they do in fact act trivially. Now what about the representations Δ + ⊗ 𝜉 − and Δ − ⊗ 𝜉 ? The elements 𝑘 and 𝑘 clearly act trivially. For 𝑘 and 𝑘 , recall theconstruction of the spinor representations in proposition 2.4.15: we saw there that (cid:206) 𝑒 𝑗 acts as ( (cid:206) 𝜖 𝑟 ) 𝑖 , where (cid:206) 𝜖 𝑟 = ± Δ ± respectively; coupling this with the fact that theU ( ) components act as 𝑖 ∓ , means that 𝑘 , for example, acts on Δ + as 𝑖 ⊗ 𝑖 − =
1. Theother three cases work out just as easily. The final representations we need to consider are Λ ⊗ 𝜉 ± , where Spin ( ) acts by conjugation. Once again, 𝑘 and 𝑘 pose no problem. Totackle 𝑘 and 𝑘 , we will need the following Claim 3.2.2. (cid:206) 𝑛 𝑒 𝑗 ∈ Spin ( 𝑛 ) acts on 𝑣 ∈ Λ 𝑛 as 𝑣 ↦→ − 𝑣 . Proof.
This is a direct computation. Since Clifford multiplication is linear, it suffices toshow this for 𝑣 = 𝑒 𝑘 , for some 1 ≤ 𝑘 ≤ 𝑛 . Consider then ( 𝑒 · · · 𝑒 𝑛 ) · 𝑒 𝑘 = ( 𝑒 · · · 𝑒 𝑛 ) 𝑒 𝑘 ( 𝑒 · · · 𝑒 𝑛 ) − = (− ) 𝑛 − 𝑘 𝑒 · · · 𝑒 𝑘 𝑒 𝑘 (cid:124)(cid:123)(cid:122)(cid:125) − 𝑒 𝑘 + · · · 𝑒 𝑛 (− 𝑒 𝑛 ) · · · (− 𝑒 𝑘 + ) (cid:124) (cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32) (cid:123)(cid:122) (cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32) (cid:125) · · · (− 𝑒 ) = (− ) 𝑛 − 𝑘 − 𝑒 · · · 𝑒 𝑘 − (− 𝑒 𝑘 )(− 𝑒 𝑘 − ) · · · (− 𝑒 ) = (− ) 𝑛 − 𝑘 − (− ) 𝑘 − (− 𝑒 𝑘 ) 𝑒 · · · 𝑒 𝑘 − (− 𝑒 𝑘 − ) · · · (− 𝑒 ) (cid:124) (cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32) (cid:123)(cid:122) (cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32)(cid:32) (cid:125) = − 𝑒 𝑘 . QEDThus, 𝑘 acts on Λ ⊗ 𝜉 as (− ) ⊗ 𝑖 =
1; the other cases are similar. We conclude thatthe kernel Z does indeed act trivially on all the fermions in the E theory. By theorem2.5.3 then, we can write down . 3 . T H E N E W F E R M I O N S 43 Theorem 3.2.3. E is a grand unified theory, i.e. the following diagram commutes: 𝐺 SM / Z E 𝐹 + 𝐹 𝑁 + 𝑁 We are now in uncharted territory: this latest extension of the Standard Model has, for thefirst time, yielded new particles. We started with a 32 dimensional representation 𝐹 ⊕ 𝐹 of all the standard model fermions, and found that they fit exactly into the irrep Λ ∗ C of SU ( ) ; this was in turn shown to be isomorphic to the spinor representation Δ + + Δ − of Spin ( ) . But now, we have added a significant number of dimensions: 𝑁 ⊕ 𝑁 is27 + =
54 dimensional, which means that we have 11 new fermions and antifermions.How can we understand them?Let us think again about the SU ( ) grand unified theory. There, we matched irreps,one by one, of SU ( ) and 𝐺 SM ; this perhaps obscured the fact that the particles of the SU ( ) theory as such, are not characterised by the 𝐺 SM charges. Said another way, if we lived in auniverse governed by an unbroken SU ( ) theory, there would be no need to think of theStandard Model charges, in the same way that the representation theory of the strong forceis remarkably simple because its symmetry group SU ( ) is unbroken at the vacuum. Butmore often than not, we find ourselves in the opposite situation, and so out of necessity,we characterise particles based on how they transform under the broken symmetry ofour vacuum, 𝐺 SM . In short, to understand these new fermions, we need to think aboutsymmetry breaking, and in particular, we need to understand what charges they carryunder 𝐺 SM .Happily, one can see a whole lot at the level of representation theory, without venturinginto the (complicated) dynamics of symmetry breaking; indeed, without saying so explicitly,we have laid most of the groundwork. Consider again the irrep matching of the SU ( ) theory, equation (2.3.1). Once we confirmed that that Z kernel of 𝜙 : 𝐺 SM → SU ( ) actedtrivially on 𝐹 , matching irreps was precisely the act of understanding how the SU ( ) symmetry broke down to a 𝐺 SM theory. In much the same way, theorem 2.5.2 was theattempt to see how Spin ( ) broke to SU ( ) . In both cases, we had no need for any newcharges; with E , the situation is different. The proof that E is a grand unified theoryrested on the inclusion Spin ( ) ↩ → Spin ( ) × U ( ) → E , so a new U ( ) charge seems tobe demanded by the mathematics; let us denote it with U ( ) (cid:48) to differentiate it from theU ( ) of electromagnetism. Then since we have no obvious reason to not do so, let us simplydeclare that each particle now carries the U ( ) (cid:48) charge 𝑄 (cid:48) dictated by the superscript ofthe 𝜉 representation to which it belongs. For example, the particles which live in therepresentation 1 ⊗ 𝜉 − of Spin ( ) × U ( ) (cid:48) will carry a charge 𝑄 (cid:48) of − ( ) (cid:48) symmetry wereto remain unbroken at the vacuum, this would imply the presence of a new force (similarto electromagnetism) mediated by a hitherto unobserved massless boson (akin to thephoton). No such force has been detected to date, so let us take this into account and posit G R A N D U N I F I E D T H E O RY the following cascade of theories: 𝐺 SM / Z SU ( ) × U ( ) (cid:48) Spin ( ) × U ( ) (cid:48) E 𝐹 + 𝐹 + · · · Λ ∗ C ⊗ 𝑄 (cid:48) + · · · Δ + ⊗ 𝜉 − + Δ − ⊗ 𝜉 + · · · 𝑁 + 𝑁 (cid:27) (cid:27) Following the discussion in the previous paragraphs, we have introduced here the notation Λ ∗ C ⊗ 𝑄 (cid:48) for the extended SU ( ) theory to indicate that the particle representations arenow tensored with an additional 𝜉 𝑄 (cid:48) : for the left-handed electron for example, we wouldwrite 𝑒 − 𝐿 ∈ Λ C ⊗ 𝜉 − , since in the Spin ( ) theory, 𝑒 − 𝐿 lives in Δ + , and this is now tensoredwith 𝜉 − . In fact, it should be clear that this analysis works for all the Standard Modelfermions: we know already which Weyl spinor representation they live in, so it is a simplematter to assign to them a 𝑄 (cid:48) = ∓
1, according to whether they are in Δ ± , respectively.The first legitimately new particles appear in 𝜉 ± , but these are easy to understand sincethey do not transform in any group other than U ( ) (cid:48) . Hence, at the level of SU ( ) × U ( ) (cid:48) ,we can simply state that they are the sole elements of the one-dimensional representations1 ⊗ 𝜉 ± ; under this assignment, they would be antiparticles of each other, and not interactwith any of the Standard Model particles. We will return to this interesting point in section4.3.The representations Λ ⊗ 𝜉 ± will take the most work to sort through. Clearly, the firststep is to understand how the Spin ( ) representation Λ breaks to SU ( ) . We make thefollowing Claim 3.3.1.
Under SU ( ) ↩ → Spin ( ) , the representation Λ restricts to Λ + Λ , where SU ( ) acts on the former as its fundamental representation, and on the latter as the complex conjugatethereof. The proof of this will be in stages. The first thing we will do is to ask whether it sufficesto consider the same question at the level of Lie algebras, since in that case we have theexplicit embedding (and corresponding eigenvalue problem), 𝜏 : 𝔰𝔲 ( 𝑛 ) 𝔰𝔬 ( 𝑛 ) (cid:27) 𝔰𝔭𝔦𝔫 ( 𝑛 ) 𝐴 + 𝑖𝐴 (cid:18) 𝐴 𝐴 − 𝐴 𝐴 (cid:19) , (3.3.1)where 𝐴 and 𝐴 are real 𝑛 × 𝑛 matrices such that 𝐴 𝑇 = − 𝐴 , 𝐴 𝑇 = 𝐴 , and tr 𝐴 =
0. Theresult that we will need comes from a classic query in the theory of representations: canevery representation of the Lie algebra of a Lie group be associated with a representation ofthe group itself, where we moreover require that the differential of the group representationreturns the one of the algebra? The answer turns out to be in the affirmative in the casewhere the Lie group is simply connected [87, p. 105], which works out nicely for us sinceboth SU ( 𝑛 ) and Spin ( 𝑛 ) are indeed simply connected: Spin ( 𝑛 ) is simply connected byvirtue of being the universal cover of SO ( 𝑛 ) ; for a proof for SU ( 𝑛 ) , see [92].Now as we saw above, Spin ( 𝑛 ) acts on Λ 𝑛 by conjugation; the differential of this actionis the commutator, 𝑋 · 𝑣 = [ 𝑋 , 𝑣 ] , for 𝑋 ∈ 𝔰𝔭𝔦𝔫 ( 𝑛 ) , 𝑣 ∈ Λ 𝑛 . Note that the multiplication Masiero’s paper [63] considers some of the phenomenological implications of such an extension to theSU ( ) theory. The article by King [54] is a general reference for extended SU ( ) theories. Some of theseextensions are still viable as grand unified theories [1]. . 3 . T H E N E W F E R M I O N S 45 on the right hand side of the equation is Clifford multiplication, where we canonicallyembed both Λ 𝑛 (cid:27) C 𝑛 and 𝔰𝔬 ( 𝑛 ) (cid:27) 𝔰𝔭𝔦𝔫 ( 𝑛 ) = span (cid:8) 𝑒 𝑟 𝑒 𝑠 ∈ Cl ( C 𝑛 ) | ≤ 𝑟 < 𝑠 ≤ 𝑛 (cid:9) into Cl ( C 𝑛 ) . How do we now relate this to our other embedding, (3.3.1)? Lemma 3.3.2.
For 𝑋 ∈ 𝔰𝔬 ( 𝑛 ) and 𝑣 ∈ C 𝑛 , 𝑋 · 𝑣 = [ 𝑋 , 𝑣 ] , where the on the left we have the standard action of 𝔰𝔬 ( 𝑛 ) on C 𝑛 , and on the right, Cliffordmultiplication. Proof.
As with all linear algebra results, it suffices to check this on a basis. As we haveseen, a natural one for 𝔰𝔬 ( 𝑛 ) , the space of skew-symmetric matrices, is { 𝜖 𝑟𝑠 = 𝐸 𝑠𝑟 − 𝐸 𝑟𝑠 | ≤ 𝑟 < 𝑠 ≤ 𝑛 } , where 𝐸 𝑟𝑠 is the 2 𝑛 × 𝑛 matrix with 1 in the 𝑟𝑠 entry, and 0 everywhere else. We have, for 𝑒 𝑗 a standard basis vector of C 𝑛 , 𝜖 𝑟𝑠 𝑒 𝑗 = 𝛿 𝑗𝑟 𝑒 𝑠 − 𝛿 𝑗𝑠 𝑒 𝑟 . From proposition 2.4.10, the isomorphism between 𝔰𝔭𝔦𝔫 ( 𝑛 ) and 𝔰𝔬 ( 𝑛 ) is given by 𝑒 𝑟 𝑒 𝑠 = 𝜖 𝑟𝑠 ; we thus compute, [ 𝜖 𝑟𝑠 , 𝑒 𝑗 ] = [ 𝑒 𝑟 𝑒 𝑠 , 𝑒 𝑗 ] = (cid:0) 𝑒 𝑟 (− 𝛿 𝑗𝑠 − 𝑒 𝑗 𝑒 𝑠 ) − (− 𝛿 𝑗𝑟 − 𝑒 𝑟 𝑒 𝑗 ) 𝑒 𝑠 (cid:1) = (cid:0) 𝛿 𝑗𝑟 𝑒 𝑠 − 𝛿 𝑗𝑠 𝑒 𝑟 (cid:1) = 𝜖 𝑟𝑠 𝑒 𝑗 . QEDTherefore, we now have an honest-to-goodness eigenvalue problem for the matrix (cid:16) 𝐴 𝐴 − 𝐴 𝐴 (cid:17) ∈ 𝔰𝔭𝔦𝔫 ( 𝑛 ) . A quick calculation shows that the two 𝑛 -dimensional eigenspaces ofthis matrix are spanned by ( 𝑢, ± 𝑖𝑢 ) , 𝑢 ∈ C 𝑛 : (cid:18) 𝐴 𝐴 − 𝐴 𝐴 (cid:19) (cid:18) 𝑢 ± 𝑖𝑢 (cid:19) = (cid:18) ( 𝐴 ± 𝑖𝐴 ) 𝑢 (± 𝑖𝐴 ∓ 𝐴 ) 𝑢 (cid:19) (cid:27) ( 𝐴 ± 𝑖𝐴 ) 𝑢 , whence we conclude that SU ( ) ↩ → Spin ( ) does indeed act as its fundamental repre-sentation (resp. complex conjugate fundamental representation) on Λ (resp. Λ ). Thiscompletes the proof of claim 3.3.1.We are almost done. The last step we must make is to understand how the Λ and Λ of SU ( ) break down to 𝐺 SM / Z so we can assign the Standard Model charges to theseparticles; but this is easy. Indeed, the homomorphism from before, 𝜙 : 𝐺 SM SU ( ) , ( 𝛼 , 𝑔, ℎ ) (cid:18) 𝛼 𝑔 𝛼 − ℎ (cid:19) G R A N D U N I F I E D T H E O RY contains all the information that we need. We can just read off the restricted representations(recall that because of how the hypercharge representation C 𝑌 was defined, we have todivide the exponent of 𝛼 by 3):SU ( ) ( U ( ) × SU ( )) + ( U ( ) × SU ( )) Λ ( C ⊗ C ) + ( C − / ⊗ C ) Let us consider a quick example to see how we might catalogue these particles: to theparticles in the C ⊗ C doublet of U ( ) × SU ( ) , we would assign as usual the isospins ± /
2, and they would each carry a hypercharge 𝑌 of 1. In addition, at the level of theSU ( ) theory and beyond, they would carry a new 𝑄 (cid:48) = ±
2, according whether the Λ came from the Λ ⊗ 𝜉 ± . Finally, we note that for the (antiparticle) representation Λ , onesimply passes to the complex conjugate of the representation on the bottom right of thecommuting diagram above.We summarise all of the information in this section in table 3.1. The hypercharge 𝑌 therein gives the corresponding Standard Model U ( ) representation C 𝑌 ; only thedoublets (and hence the particles with non-zero isospin) transform in SU ( ) ; the SU ( ) representations are written down explicitly. The electromagnetic charge 𝑄 can be obtainedfrom the 𝑌 and 𝐼 columns via the NNG formula. The 𝑄 (cid:48) column gives the correspondingU ( ) (cid:48) representation that should be tensored with the representations of 𝐺 SM , SU ( ) orSpin ( ) . Finally, the corresponding table for 𝑁 is easily obtained from this one by passingto the dual representations and taking the opposite charges throughout. Up to the sign of 𝑄 (cid:48) (he chooses the opposite convention), we have reproduced Table 21 in [81], forexample. . 3 . T H E N E W F E R M I O N S 47 Table 3.1: Particles in the Representation 𝑁 of E Symbol
𝑌 𝐼 SU ( ) Rep. 𝑄 (cid:48) SU ( ) Rep.
Spin ( ) Rep. 𝜈 𝐿 C − Λ Δ + 𝑒 + 𝐿 C − Λ Δ + (cid:169)(cid:173)(cid:171) 𝑢 𝐿 𝑑 𝐿 (cid:170)(cid:174)(cid:172) / ± / C − Λ Δ + 𝑢 𝐿 − / C − Λ Δ + (cid:169)(cid:173)(cid:171) 𝜈 𝐿 𝑒 − 𝐿 (cid:170)(cid:174)(cid:172) − ± / C − Λ Δ + 𝑑 𝐿 / C − Λ Δ + (cid:169)(cid:173)(cid:171) ?? (cid:170)(cid:174)(cid:172) ± / C Λ Λ ? − / C Λ Λ (cid:169)(cid:173)(cid:171) ?? (cid:170)(cid:174)(cid:172) − ± / C Λ Λ ? 2 / C Λ Λ ? 0 0 C − C C hapter 4
Aspects of Phenomenology
As mathematically interesting as grand unified theories are, they are ultimately statementsabout the real world. So in this section, we pose the following question: at the levelof group (representation) theory, what can we say about the phenomenology of thesetheories? Clearly, a healthy amount of physics is required to motivate and supplement anysuch discussion, but the aim is to stay as close to mathematics as possible; the relevantphysics is introduced where necessary.In the first section, we discuss a proper prediction of grand unified theories, theweak mixing angle, which has a simple closed formula in terms of the eigenvalues of theintertwining operators ˆ 𝐼 and 𝑄 . Following this, we will discuss anomalies, which arenot so much a phenomenological prediction as they are a basic physical requirement onunification groups. They have a rather nice interpretation in terms of a certain Casimiroperator on the Lie algebras of said groups, so this issue is completely reduced to amathematical property that we can understand fairly easily, given the machinery we havealready built. Finally, section 4.3 functions as something of a survey section, where weconsider other expected signatures of the E theory, and discuss its outlook. One of the unambiguous predictions of grand unified theories is the weak mixing angleor Weinberg angle, which we have already encountered in section 1.2.3. Recall thatequation (1.2.1) offered a rather geometric interpretation of this angle, as the parameterthat characterised the rotation of the 𝑊 − 𝐵 boson plane after symmetry breaking; it canalso be written in terms of the gauge couplings 𝑔 and 𝑔 , of the SU ( ) and U ( ) groups ofthe electroweak theory respectively, assin 𝜃 w = 𝑔 𝑔 + 𝑔 . (4.1.1)In 1974, Georgi, Quinn, and Weinberg derived a formula for the weak mixing angle 𝜃 w in grand unified theories [37]. The only assumption that they needed in the proof thereofwas that the U ( ) × SU ( ) group of the electroweak theory is embedded in the grandunification group 𝐺 in such a way that the NNG formula still holds. We have assumedthis throughout, so this theorem is applicable to all the grand unified theories we haveanalysed; let us hence state and prove their result. Theorem 4.1.1.
Let 𝑅 be some fermion representation of the unification group 𝐺 . Then the weak
490 4 . A S P E C T S O F P H E NO M E NO L O G Y mixing angle is given by sin 𝜃 w = (cid:213) fermions 𝐼 (cid:213) fermions 𝑄 . Proof.
We follow the lecture notes of Bjorken [16]. The 𝑊 = ˆ 𝐼 and 𝐵 bosons, appropriateto the broken SU ( ) and U ( ) theory, are gauge bosons of the full gauge group 𝐺 ; thecoupling of 𝑊 to any fermion is proportional to 𝐼 , and the coupling of the 𝐵 boson isproportional to the hypercharge 𝑌 . Because 𝑊 and 𝐵 are both gauge particles for thegroup 𝐺 , we must have, for any representation of 𝐺 , (cid:213) fermions 𝐼 = (cid:213) fermions 𝑌 , (4.1.2)since there is a symmetry operation of the group that can transform 𝑊 into 𝐵 , but thattransforms the representation 𝑅 into itself.Completing the proof is now a matter of simple algebra. From equation (1.2.1), wemust have that the electric charge is given by 𝑄 ∝ ( 𝑌 cos 𝜃 w + 𝐼 sin 𝜃 w ) ;in order to have the difference of 𝑄 between two members of the same isospin doublet be ±
1, we must set the constant of proportionality to ( sin 𝜃 w ) − , i.e. 𝑄 = 𝐼 + cot 𝜃 w 𝑌 .
We now square this equation, and sum over all fermions. The cross term (cid:205) 𝐼 𝑌 vanishes,because the only non-zero contributions to this sum come from isospin doublets, and foreach doublet this term is zero (since 𝑌 is constant on a doublet, while the 𝐼 ’s come withopposite signs). We hence obtain (cid:213) fermions 𝑄 = (cid:213) fermions 𝐼 + cot 𝜃 w (cid:213) fermions 𝑌 . Utilising equation (4.1.2) above, we obtain the formula stated in the theorem. QEDIn the same paper, Georgi et al. immediately applied this result to the SU ( ) theory,leading to the famous prediction sin ( ) 𝜃 w = /
8. It should be clear that since the Spin ( ) theory introduces no new fermions, the prediction for the Weinberg angle is same as forSU ( ) . In our E theory however, we do have new fermions, so we should see a differentvalue; indeed, on consulting table 3.1 and doing the necessary arithmetic, we obtain thefollowing: sin 𝜃 w = = . . As far as representation theory goes, this is all we can say. But it is too tempting to notcompare such a definite phenomenological prediction with the real world; unfortunately,the comparison is none too comforting: one standard estimate [67] for the weak mixingangle is sin 𝜃 w = . renormalisation theory , a catch-all term fortechniques used to deal with the infinities that plague quantum field theory. We haveneither the desire nor the pages to get into any details here , but we would like to at least We point the reader to [74, 79, 90] or any standard quantum field theory reference. . 1 . T H E W E A K M I X I NG A NG L E 51 state a result from Marciano [62] which succinctly accounts for renormalisation effects onthe value of sin 𝜃 w in grand unified theories. What follows is hence necessarily sketchy;the reader is encouraged to consult the original paper for an excellent discussion. The mainassumption that he needed in its derivation goes back to an earlier paper of Weinberg’s [89]:all gauge bosons in the grand unified theory must have large masses (on the order of somesuperheavy 𝑀 𝑆 , say) compared with the 𝑊 ± and the 𝑍 , (the order of which we denoteby 𝑀 𝑊 ) and also compared with the standard model fermions in the theory (also on theorder 𝑀 𝑊 ). The motivation for this was mostly phenomenological: effects mediated bythese gauge bosons had eluded detection thus far. Of course, there was a further technicalaspect to this assumption, but since it involes some quantum field theory, we relegate it toa footnote dedicated to the interested reader. In any case, Marciano’s result is written as sin 𝜃 w ( 𝑀 𝑊 ) = sin 𝜃 (cid:40) − 𝛼 ( 𝑀 𝑤 ) 𝜋 (cid:20)
223 cot 𝜃 − 𝑁 𝐻 (cid:18) 𝜃 − (cid:19) − 𝑁 𝑓 (cid:18) 𝜃 − (cid:19) (cid:21) log 𝑀 𝑆 𝑀 𝑊 (cid:41) . From left to right, here are the quantities we have not yet defined. The superscript 0 insin 𝜃 simply indicates that this is the theortical value of the weak angle given by thetheory, i.e. 9 /
20 in the case of E . 𝛼 ( 𝑀 𝑊 ) is the fine structure constant ; it depends on themass scale because it is defined through the electric charge as 𝑒 / 𝜋 , and 𝑒 has a massscale dependence. Marciano provides the estimate 𝛼 ( 𝑀 𝑊 ) ≈ / .
5; we will use the same.Next, the term 𝑁 𝐻 is the number of complex Higgs doublets in the theory; we set 𝑁 𝐻 = 𝑁 𝑓 is the number of fermion flavours: for the StandardModel (and SU ( ) and Spin ( ) ), this equals 6, as seen in table 1.3; similarly, from table 3.1,we see that we have 𝑁 𝑓 = × =
12 for E , since we add two new flavours (one each in 𝑁 and 𝑁 ) per generation. Plugging all this in, the above formula simplifies tosin 𝜃 w ( 𝑀 𝑊 ) = (cid:18) − .
015 log 𝑀 𝑆 𝑀 𝑊 (cid:19) . So for example, if we take the measured values sin 𝜃 w = . 𝑀 𝑊 = . theory is of the order 𝑀 𝑆 = . × GeV.We caution that the value obtained from the formula is is quite sensitive to changes inthe value of sin 𝜃 w because of the logarithm; it decreases by about 50% for each increase He leaves open the possibility that there might be exotic fermions in the theory with masses on the orderof 𝑀 𝑆 . As we will see in section 4.3, this is the case with E . See also [75]. Even today, the lower bounds on the masses of grand unified theory gauge bosons are at least two ordersof magnitude larger than the known masses of the 𝑊 and 𝑍 bosons [73]. The argument, as seen in [37], runs as follows. The gauge couplings—of the grand unified symmetrygroup 𝐺 , and the Standard Model subgroups U ( ) , SU ( ) and SU ( ) —are functions of the momentum scalewhich we denote by 𝜇 ; in particular, equation (4.1.1) only holds when 𝜇 is much larger than the superheavyboson masses, where the breaking of 𝐺 may be neglected. However, the observed values of the gaugecouplings refer to much smaller values of 𝜇 , of the order of the 𝑊 ± and 𝑍 masses, or even smaller. Theproblem is therefore to bridge the gap between superlarge values of 𝜇 , where 𝐺 imposes relations amongthe gauge couplings, and ordinary values of 𝜇 , where the gauge couplings are observed. In order to dealwith this, Georgi and collaborators employed a theorem from Appelquist et al. [6], which proved that allmatrix elements involving particles with masses much less than the superheavy scale could be calculated inan effective renormalisable theory. In this case, one could simply consider the original theory with all thesuperheavy particles omitted (but with coupling constants that could depend on the superheavy masses).All other effects of the superheavy particles are suppressed by factors of an ordinary mass divided by asuperheavy mass. We note that this formula is specifically for theories with sin 𝜃 ≠ / of 0 .
005 in sin 𝜃 w . One final remark is that we fixed the value of 𝑁 𝐻 = ; if this restriction isrelaxed, there is some wiggle room in the above formulae to increase the value of sin 𝜃 w by increasing the value of 𝑁 𝐻 , and this in turn obviously has a direct bearing on 𝑀 𝑆 ; tableII in [62] estimates the size of this effect. In section 1.1.2, we discussed Lagrangian symmetries, and saw their paramount importance;the transformation laws that we considered there were indeed the foundation for everythingthat came after. We return to this theme now, but with a different question as our startingpoint: which classical symmetries of the Lagrangian are elevated to quantum symmetries?The business of quantising a classical Lagrangian is a messy one. By way of illustration,consider the simplest case: given a Lagrangian ℒ of a (real) scalar field 𝜙 , one defines the generating functional as 𝑍 [ 𝐽 ] : = ∫ D 𝜙 exp (cid:20) 𝑖 ∫ d 𝑥 (cid:0) ℒ + 𝐽 𝜙 (cid:1) (cid:21) , where 𝐽 𝜙 suggestively denotes a source term, akin to electromagnetism. The measure ofintegration, D 𝜙 , represents an integration over all possible field configurations. We cannow define the effective action : Γ [ 𝜙 cl ] = 𝑊 [ 𝐽 ] − ∫ d 𝜏 d 𝑥𝐽 𝜙 cl , where 𝑊 [ 𝐽 ] is defined implicitly via 𝑍 [ 𝐽 ] = 𝑒 − 𝑊 [ 𝐽 ] , 𝜙 cl is the functional derivative 𝛿 𝑊 [ 𝐽 ]/ 𝛿 𝐽 ,and 𝜏 = 𝑖𝑡 is the so-called Wick-rotated time. Now, the effective action is given its name forobvious reasons: just as in classical mechanics, where the equations of motion are derivedfrom the principle of stationary action, the equations of motion for the vacuum expectationvalues of quantum fields can be derived from the requirement that the effective action bestationary. The details, and examples of such calculations can be found in [74, Ch. 9], orany other standard reference on quantum field theory; what we have here should sufficeto motivate an effective quantum action for a classical Lagrangian.The first anomaly that we will consider is the chiral anomaly ; to introduce the same,we will need the
Dirac equation . We have encountered in some detail already the Diracspinors in chapter 2, as elements of certain irreps of the Spin groups. This is a descriptionfree of dynamics, and therefore, far removed from Dirac’s original conception of theseparticles. The equation describes all spin-1 / See section 4.3 for references for the Higgs mechanism in E . See [55, § 28–30] for a cogent presentation of the same. It is common knowledge that foundational questions about the mathematical validity of this definition,and about the path integral formalism in general, remain. See [5] for a mathematical overview, and [50] for aphysicist’s take on the same. See [68, Chs. 1.1.2, 1.3.3] For example, if our Lagrangian includes a potential 𝑉 ( 𝜙 ) , at a low temperatures, the quantum field 𝜙 will not settle in a local minimum of 𝑉 ( 𝜙 ) as in the classical case, but rather in a local minimum of the effectivepotential. In the 1928 paper [23] presenting his equation for the first time, he begins by asking “why Nature shouldhave chosen this particular model for the electron, instead of being satisfied with the point charge;” hisremarkable solution to this quandary was a theory that, for the first time, fully accounted for special relativityin the context of quantum mechanics. For a captivating account of the history and a lucid derivation of theequation, see [90, Ch. 1.1]. . 2 . A NO M A LY C A NC E L L AT I O N 53 symmetry, i.e. the leptons. In symbols, for a field 𝜓 , which we take to be massless, and agauge boson 𝐴 𝜇 = (cid:205) 𝛼 𝐴 𝛼𝜇 𝑇 𝛼 , written in a basis of generators of the compact semi-simplesymmetry group 𝐺 , the theory is described by the Lagrangian ℒ = (cid:213) 𝜇 𝑖 𝜓𝛾 𝜇 ( 𝜕 𝜇 − 𝐴 𝜇 ) 𝜓 . (4.2.1)The 𝛾 𝜇 , 0 ≤ 𝜇 ≤
3, are the gamma matrices , which form a basis for the Clifford algebraCl , ( R ) ; the subscript 1,3 denotes that instead of relation (2.4.1), we use { 𝛾 𝜇 , 𝛾 𝜈 } = 𝜂 𝜇𝜈 ,where 𝜂 𝜇𝜈 is the Minkowski metric with the (physicists’) signature (+ − −−) . As insection 1.1.2, the Lagrangian is invariant under the local gauge transformation 𝜓 ↦→ 𝑔 𝜓 , 𝐴 𝜇 ↦→ 𝑔𝐴 𝜇 𝑔 − + 𝜕 𝜇 𝑔 𝑔 − , but there is now an additional global symmetry 𝜓 ↦→ 𝑒 𝑖 𝛼𝛾 𝜓 ,where 𝛼 ∈ R , and 𝛾 : = 𝑖 𝛾 𝛾 𝛾 𝛾 = (cid:16) Id − Id (cid:17) . This symmetry is chiral: in the standard(Dirac) basis, the left- and right-handed Weyl components of Dirac 4-spinor correspond tothe first two and second two components of the 4-vector respectively; the effect of the 𝑒 𝑖 𝛼𝛾 is then to rotate these Weyl spinors in opposite directions, by the angle 𝛼 .As with any other Lagrangian symmetry, the chiral symmetry corresponds to a current,which in this case can be shown to be 𝑗 𝜇 : = 𝜓𝛾 𝜇 𝛾 𝜓 . Recall that we need our theory of leptons to be chiral. The question of the hour istherefore the following: does this classically conserved quantity (i.e. (cid:205) 𝜇 𝜕 𝜇 𝑗 𝜇 =
0) stayconserved when we pass to the quantised Dirac Lagrangian? The answer turns out to be no . Unfortunately, deriving this result is a nuanced, technical calculation in quantum fieldtheory, far outside the scope of this paper; we list some references in a footnote. The finalresult of this computation is stated as (cid:213) 𝜇 𝜕 𝜇 𝑗 𝜇 = (cid:213) 𝜅 , 𝜆 , 𝜇 , 𝜈 𝜋 𝜖 𝜅𝜆𝜇𝜈 tr 𝐹 𝜅𝜆 𝐹 𝜇𝜈 = 𝜋 tr (cid:213) 𝜅 , 𝜆 , 𝜇 , 𝜈 𝜖 𝜅𝜆𝜇𝜈 𝜕 𝑘 (cid:18) 𝐴 𝜆 𝜕 𝜇 𝐴 𝜈 + 𝐴 𝜆 𝐴 𝜇 𝐴 𝜈 (cid:19) . This is not particularly illuminating as it stands. One can show however [14, § 4] that theright hand side can be rewritten such that it contains the term (recall that the 𝐴 𝜇 ’s arewritten in terms of the group generators 𝑇 𝛼 )tr ( 𝑇 𝑎 { 𝑇 𝑏 , 𝑇 𝑐 }) L − tr ( 𝑇 𝑎 { 𝑇 𝑏 , 𝑇 𝑐 }) R , (4.2.2)where the subscript L (resp. R) denotes the representation of the left-handed (resp. right-handed) fermions under consideration; our theory is said to be “anomaly-free” if thisquantity vanishes. We have already seen in section 1.1.2 that gauge bosons are Lie-algebra valued 1-forms; hence, they canbe expressed locally in a basis of (anti-Hermitian) generators { 𝑇 𝛼 } of the Lie algebra 𝔤 . For more details, c.f [31,Ch. 4.6]. We note here that adding a mass term 𝑚 𝜓𝜓 to the Lagrangian spoils this symmetry, which is why werestrict ourselves to the massless situation. One can show that only such massless fermions contribute toanomalies anyway; see [14, § 7.2]. The lectures of Bilal [14] are specifically on this topic; Schwartz’s book [79, Ch. 30] also has a detailedtreatment; Nakahara [68, Ch. 13] derives the same result from a geometric point of view, making the connectionto Atiyah-Singer-Index theory; the book of Nash [69] goes even deeper into the mathematics.4 4 . A S P E C T S O F P H E NO M E NO L O G Y
References [14, § 7.3] and [79, Ch. 30.4] show how the explicit check proceeds in thecase of the Standard Model gauge group; it is a surprisingly tame affair. In the case of theU ( ) − U ( ) − U ( ) anomaly for instance, equation (4.2.2) simply reduces to (cid:205) L 𝑌 − (cid:205) R 𝑌 ,and a glance at table 1.1 will confirm that this indeed vanishes. The checks for the othersubgroups of 𝐺 SM are just as straightforward. Closer to our purposes, the vanishing of equation (4.2.2) is also a requirement onthe groups used for grand unified theories; what can we say about these? Georgi andGlashow showed in [35] that if a group has only real (or pseudoreal) representations, it isautomatically anomaly-free—as far as simple Lie groups go, this immediately allowedall theories with gauge groups SO ( 𝑛 + ) (including SU ( ) (cid:27) SO ( )) , SO ( 𝑛 ) for 𝑛 ≥ ( 𝑛 ) for 𝑛 ≥
3, as well as all the exceptional Lie groups other than E (which we knowhas complex representations). In the same paper, they went on to prove that all SO ( 𝑛 ) ’s,excluding SO ( ) , are also anomaly-free, thus adding the case SO ( 𝑛 + ) for 𝑛 ≥ ( ) theory). This left only the SU ( 𝑛 ) ’s,for 𝑛 ≥
3, and E . The case of the SU ( ) theory is discussed in [66, Ch. 5.2] for example; itsuffices for us to note that the representation that we employed is indeed free of anomalies.As for E , the following fact was clear to Gürsey and collaborators [40] right at the adventof this theory: Theorem 4.2.1.
All representations of E are anomaly-free. The remainder of this section will be dedicated to sketching a proof of this result; wewill follow Okubo’s paper from 1977 [70]. At the crux of his proof is the following claim:the calculation of an 𝑛 -fermion closed-loop diagram is related to a study of the 𝑛 th orderCasimir invariant of the algebra 𝔤 of the symmtery group 𝐺 . We will slowly work towardsunderstanding what this means, and how the proof proceeds therefrom.Introducing Feynman diagrams in detail is outside our scope here , but we must say afew words: these diagrams are representations of certain mathematical expressions thatarise in perturbative (read: almost all) calculations in quantum field theory, usually of thescattering amplitudes of particles. The chiral anomaly we have been considering thus faris often called a triangular anomaly because to one-loop (first order), the Feynman diagramlooks like so.Here, the external legs represent any of the gauge bosons of the theory, while the fermionscirculating in the internal lines can be in any relevant representation of the gauge groups.So according to Okubo, since our anomaly arises in the 3-fermion closed-loop diagram,we need to concern ourselves with the 3rd order Casimir invariant of E . To introducethe same, we shift our analysis to the level of Lie algebras, recalling (section 3.3) that forsimply connected Lie groups, this involves no sacrifice of generality. Definition 4.2.2 (Structure Constants) . For a 𝑑 -dimensional Lie algebra, consider any setof 𝑑 basis vectors, or generators, { 𝑡 𝑎 } . Because of bilinearity, the Lie bracket is determined It is not necessary to carry out this check for all possible triples that can be made from U ( ) , SU ( ) andSU ( ) ; cf. [79, Ch. 30.4]. The reader is once again encouraged to consult [74, 79, 90] or any standard quantum field theory reference. . 2 . A NO M A LY C A NC E L L AT I O N 55 uniquely if it is known on a basis set; therefore one can define the Lie bracket, and hencethe Lie algebra 𝔤 abstractly through the expansions [ 𝑡 𝑎 , 𝑡 𝑏 ] = 𝑑 (cid:213) 𝑐 = 𝑓 𝑎𝑏𝑐 𝑡 𝑐 , where the 𝑓 𝑎𝑏𝑐 are called the structure constants of 𝔤 .Because of the antisymmetry of the Lie bracket, the structure constants satisfy 𝑓 𝑎𝑏𝑐 = − 𝑓 𝑏𝑎𝑐 ; from the Jacobi identity, we further have 𝑑 (cid:213) 𝑐 = (cid:16) 𝑓 𝑎𝑏𝑐 𝑓 𝑐𝑑𝑒 + 𝑓 𝑑𝑎𝑐 𝑓 𝑐𝑏𝑒 + 𝑓 𝑏𝑑𝑐 𝑓 𝑐𝑎𝑒 (cid:17) = . The next idea that we wish to consider requires the tensor algebra of the vector space(definition 2.4.1) over which 𝔤 is defined. To endow this very general product with thestructure that 𝔤 carries, we make an obvious identification: an element of the form (wesuppress the ⊗ symbol for brevity) 𝑥 𝑥 · · · 𝑥 𝑖 𝑥 𝑖 + 𝑥 𝑖 + · · · 𝑥 𝑛 − 𝑥 𝑥 · · · 𝑥 𝑖 + 𝑥 𝑖 𝑥 𝑖 + · · · 𝑥 𝑛 ∈ 𝑉 ⊗ 𝑛 is identified with 𝑥 𝑥 · · · [ 𝑥 𝑖 , 𝑥 𝑖 + ] 𝑥 𝑖 + · · · 𝑥 𝑛 ∈ 𝑉 ⊗( 𝑛 − ) . This quotient still has the structure of an associative algebra (with a unit element), and iscalled the universal enveloping algebra of 𝔤 . Definition 4.2.3 (Vector Operator) . A collection of 𝑑 elements { 𝑥 𝑎 } belonging to theuniversal enveloping algebra of 𝔤 is called a vector operator on 𝔤 if the following relationholds: [ 𝑡 𝑎 , 𝑥 𝑏 ] = 𝑑 (cid:213) 𝑐 = 𝑓 𝑎𝑏𝑐 𝑡 𝑐 . (4.2.3)Obviously, { 𝑡 𝑎 } is a vector operator, but it is not in general the only one. We can alsodefine vector operators { 𝑥 𝑎 } for a given 𝑛 -dimensional representation 𝜌 of 𝔤 , if { 𝑡 𝑎 } and { 𝑥 𝑎 } are 𝑛 × 𝑛 matrices satisfying the structure equation and the relation above.Let us restrict to the case that 𝜌 is an irrep of 𝔤 , and moreover demand that 𝔤 be simple.Then Okubo showed in [71] that there is a simple relationship between the number of alllinearly independent vector operators on the representation 𝜌 , and the highest weight Λ ofthat representation. This latter quantity is defined by Λ = 𝑚 Λ + 𝑚 Λ + · · · 𝑚 𝑙 Λ 𝑙 , where 𝑙 is the rank of 𝔤 , the Λ 𝑖 ’s are its roots (definition 2.1.12), and the 𝑚 𝑖 ’s are non-negative integers specified uniquely by the representation 𝜌 . Let us denote the numberof 𝑚 𝑖 ’s which are zero by 𝜈 ( 𝜌 ) ; then the number of linearly independent vector operators 𝜈 ( 𝜌 ) is given by 𝜈 ( 𝜌 ) = 𝑙 − 𝜈 ( 𝜌 ) . We have technically only defined the notions of rank and roots for Lie groups, but it should be clear thatthese concepts can be extended quite naturally to Lie algebras. For the details, see [43, Ch. 6]. It is not at all obvious that such a unique decomposition in terms of roots should exist for an arbitraryirrep of 𝔤 ; we refer the reader to [43, Ch. 7] for a proof of this, the so-called highest weight theorem .6 4 . A S P E C T S O F P H E NO M E NO L O G Y In other words, 𝜈 ( 𝜌 ) is equal to the number of 𝑚 𝑖 ’s which are positive. We can apply thistheorem immediately: for the standard Lie algebras, their roots (and weights) have beenstudied and tabulated [72]; consulting these, we see quickly that the algebras with Dynkindiagram type 𝐴 𝑛 , for 𝑛 ≥
2, have 𝜈 ( ad ) =
2, and 𝜈 ( ad ) = 𝔤 in some more detail. Set 𝑇 𝑎 = ad 𝑡 𝑎 , sothat the 𝑝𝑞 -th entry of this matrix is given by ( 𝑇 𝑎 ) 𝑝𝑞 = 𝑓 𝑝𝑎𝑞 ; the 𝑑 × 𝑑 matrices { 𝑇 𝑎 } clearlysatisfy the structure equation [ 𝑇 𝑎 , 𝑇 𝑏 ] = 𝑑 (cid:213) 𝑐 = 𝑓 𝑎𝑏𝑐 𝑇 𝑐 . Recall now the Killing form, definition 2.2.1; we will denote the same by 𝑔 𝑎𝑏 = tr 𝑇 𝑎 𝑇 𝑏 .Introduce now the vector operator { 𝑋 𝑎 } on the adjoint representation; by definition, itsatisfies [ 𝑇 𝑎 , 𝑋 𝑏 ] = (cid:205) 𝑑𝑐 = 𝑓 𝑎𝑏𝑐 𝑋 𝑐 . We further introduce the triple linear forms 𝑋 𝑝𝑎𝑞 = ( 𝑋 𝑎 ) 𝑝𝑞 ,𝑋 𝑎𝑏𝑐 = 𝑑 (cid:213) 𝑖 = 𝑔 𝑎𝑖 𝑋 𝑖𝑏𝑐 . In particular, for { 𝑋 𝑎 } = { 𝑇 𝑎 } , the second equation reduces to 𝑇 𝑎𝑏𝑐 = (cid:205) 𝑑𝑖 = 𝑔 𝑎𝑖 𝑓 𝑖𝑏𝑐 ; so 𝑇 𝑎𝑏𝑐 iscompletely antisymmetric in its indices. Now, Okubo showed that an equivalent way ofwriting the vector operator equation (4.2.3) is the following: 𝑑 (cid:213) 𝑎 = 𝑓 𝑎𝑏𝑐 𝑋 𝑎𝑑𝑒 + 𝑓 𝑎𝑏𝑑 𝑋 𝑐𝑎𝑒 + 𝑓 𝑎𝑏𝑒 𝑋 𝑐𝑑𝑎 = 𝑋 𝑎𝑏𝑐 satisfying these equations can be chosen to be eithercompletely symmetric or completely antisymmetric, and moreover, that the completelyantisymmetric 𝑋 𝑎𝑏𝑐 must be proportional to the structure constants 𝑓 𝑎𝑏𝑐 . Hence, by theresult in the previous paragraph, we conclude that for all simple Lie algebras other thanthe 𝐴 𝑛 ’s, the 𝑋 𝑎𝑏𝑐 ’s must be antisymmetric, because { 𝑋 𝑎 } must be proportional to { 𝑇 𝑎 } ,the one and only vector operator on the adjoint representation. For the 𝐴 𝑛 ’s, there is anadditional vector operator.In order to obtain the “upper-indices” version of the form 𝑋 𝑎𝑏𝑐 , we do the obviousthing, raising indices with the Killing form: 𝑋 𝑎𝑏𝑐 = (cid:205) 𝑝,𝑞,𝑟 𝑔 𝑎𝑝 𝑔 𝑏𝑞 𝑔 𝑐𝑟 𝑋 𝑝𝑞𝑟 ; the form werecover must clearly also be either symmetric or antisymmetric. If we then set 𝐼 = 𝑑 (cid:213) 𝑎,𝑏,𝑐 = 𝑋 𝑎𝑏𝑐 𝑇 𝑎 𝑇 𝑏 𝑇 𝑐 , this defines a Casimir operator (of the third order) of 𝔤 . These operators are sometimesreferred to as Casimir invariants , because they are invariant under the action of 𝔤 on itsadjoint representation—they are, in other words, intertwining operators, as defined insection 1.1.2. What is important for our purposes here is that the formula for 𝐼 above is infact the most general form of a third-order Casimir operator, and that the coefficients 𝑋 𝑎𝑏𝑐 must be symmetric. So from the result in the previous paragraph, we conclude that 𝐼 ≡ 𝐴 𝑛 ’s. For further discussion of these invariants, see [31, Ch. 17.8] and references therein. For completeness, we note that in the case of the antisymmetric 𝑋 𝑎𝑏𝑐 ’s, 𝐼 reduces to a scalar multiple ofthe Killing form, which is in fact the second-order Casimir invariant. . 3 . O T H E R S I G NAT U R E S 57 We are now ready to understand the claim that the calculation of an 𝑛 -fermion closed-loop diagram is related to a study of the 𝑛 th order Casimir invariant of 𝔤 . Let { 𝑇 𝑎 } be therepresentation matrices of the { 𝑡 𝑎 } for a generic irrep of 𝔤 ; set 𝑋 𝑎𝑏𝑐 = tr 𝑇 𝑎 { 𝑇 𝑏 , 𝑇 𝑐 } . Obviously, 𝑋 𝑎𝑏𝑐 is completely symmetric. Moreover, it satisfies equation (4.2.4) if wenote the trivial identity tr [ 𝑇 𝑝 , 𝑇 𝑎 { 𝑇 𝑏 , 𝑇 𝑐 }] =
0. Thus, 𝑋 𝑎𝑏𝑐 is a symmetric triple form,corresponding to some vector operator { 𝑋 𝑎 } ; it must hence vanish identically, and so mustequation (4.2.2). This completes the sketch of the proof that all representations of E areanomaly-free. Since grand unified theories have ever been within the purview of physicists, it is onlyfitting that we devote this final section of the paper to these tireless inquirers, for whom itis never good enough that a theory be beautiful, and rightly so; they demand that it bepredictive, and hence, falsifiable. So in the following paragraphs, we will attempt to paintin broad strokes the answers to some of the questions that have naturally arisen duringour analysis, but on which we have hitherto been silent; it is scarcely necessary to add thatwe are striving neither for comprehensiveness nor exhaustiveness here.Let us begin with the simplest question: what signatures might we expect from the E theory? We quote from a recent survey paper precisely about this topic that will set thetone for the rest of the discussion:Signatures of E include [an] extension of the Higgs sector; existence of neutral 𝑍 (cid:48) gauge bosons at masses above the electroweak scale. . . ; the productionof new vector-like quarks and leptons, and manifestations of the neutralfermion. . . through its mixing with other neutral leptons, giving rise to signa-tures of “sterile” neutrinos. Up to now, with the possible exception of weakevidence for sterile neutrinos there has been no indication of the extra degreesof freedom entailed by the 27-plet of E . [53]Setting aside for the moment that the outlook for the E theory is rather bleak, let us tryand understand the terms above that we have not yet encountered.We have refrained from speaking about the Higgs mechanism thus far, and unfortu-nately, our silence on the same will continue—references [9, 11, 18, 41, 42] are some earlypapers that examine this mechanism within the context of mass scales in the E model.One thread that runs through them all is worth examining, since it is directly concernedwith the most obvious thing one would think to look for to ratify the E theory, namely,new fermions. Here we introduce the Survival Hypothesis [10, 34]: stated succinctly, it saysthat low-mass fermions are those that cannot receive 𝐺 SM invariant masses. To understandwhat this means, recall from the previous section that mass terms spoil the invariance ofthe Lagrangian under chiral symmetry; the survival hypothesis thus postulates that whenthe grand unification symmetry group is broken down to the Standard Model gauge group,the fermions which do not acquire mass are those that cannot receive mass terms invariantunder 𝐺 SM ; in particular, this means that all the particles that do admit such a mass termwill receive a superheavy mass, since the symmetry breaking occurs at grand unificationscales. Put another way, most fermions in a grand unified theory should have masses on We have of course shown more: we have shown that all representations of every algebra other thanSU ( 𝑛 ) for 𝑛 ≥ In the previous section, for instance, the superheavy mass scale was found to be on the order of 10 GeV.8 4 . A S P E C T S O F P H E NO M E NO L O G Y unification scales; those that do not are associated with one of the two unbroken gaugesymmetry groups 𝐺 SM or U ( ) × SU ( ) , since both of these demand chiral symmetry. The upshot is the following: we do not see the new fermions of the E theory because theyare phenomenally heavy, many orders of magnitude outside the reach of even the mostpowerful detectors. For a thorough examination of mass scales in grand unified theoriesin general, and E in particular, see [75]. We note that this discussion is of course not validonly for E , but is formulated as a general principle in grand unified theories; this is the“fermion desert”. In Georgi’s words, “If [the above] picture is correct, physics between300 GeV [the Standard Model mass scale] and 10 GeV is boring. There is a grand plateauin momentum scale on which the world is well-described by an SU ( ) × SU ( ) × U ( ) gaugetheory. . . there will be no new interactions below 10 GeV.” [34]Let us turn our attention now to the aforementioned 𝑍 (cid:48) bosons. A detailed analysisof their phenomenology is outside the scope of this paper, but they arise quite naturallyin representation theory, and this we can certainly understand. If gauge bosons live inthe complexified adjoint representation of the symmetry group 𝐺 , it is clear that there issomething special about the maximal set of elements in 𝔤 that commute with each other,i.e. the Cartan subalgebra of 𝔤 . In general, these commuting elements make for goodquantum numbers (charges); the best way to see this is by choosing the Cartan-Weyl basis for 𝔤 , which we describe now. Let us consider the complexified adjoint representation of 𝔤 (while suppressing the use of the ad operator notation): if 𝔤 has rank 𝑙 and dimension 𝑑 ,let { 𝑥 𝑖 } , for 𝑖 = , . . . , 𝑙 , be the Cartan subalgebra; then one can show that the set { 𝑥 𝑖 } canbe completed to a basis { 𝑥 𝑖 , 𝑡 𝛼 } of 𝔤 such that [ 𝑥 𝑖 , 𝑡 𝛼 ] = 𝛼 ( 𝑥 𝑖 ) 𝑡 𝛼 for 𝑖 = , , . . . 𝑙 , where the eigenvalue 𝛼 ( 𝑥 𝑖 ) is non-vanishing for at least one value of 𝑖 . This is the Cartan-Weyl basis for 𝔤 , sometimes called the canonical or standard basis. The most pertinentexample of this construction is something that we have already encountered in the SU ( ) weak force. The relevant Lie algebra, 𝔰𝔩 ( , C ) , is of rank 1 and spanned by the 𝑊 bosons;they form a Cartan-Weyl basis since they satisfy [ 𝑊 , 𝑊 ± ] = ± 𝑊 ± , [ 𝑊 + , 𝑊 − ] = 𝑊 . Hence, we see that the basis for the 1-dimensional Cartan subalgebra is indeed given by aquantum number operator, the isospin matrix ˆ 𝐼 = 𝑊 .To extend the notion of charges to gauge bosons, recall first this aspect of ˆ 𝐼 : in astandard (doublet) basis for the space of weak-theory fermions C , the isospin of a particlewas simply given by the eigenvalue of the action of the ˆ 𝐼 operator on said fermion. Itshould be clear that the correct generalisation of the above concept is the following: sincethe gauge bosons transform in the adjoint representation of the symmetry group 𝐺 , theycan act on each other through the adjoint action of 𝔤 on itself ; moreover, in the Cartan-Weylbasis, the charges of the gauge bosons are given (up to normalisation) by their roots. In thecase of the SU ( ) theory for instance, this yields the correct isospin for 𝑊 ± , i.e. ±
1, since [ˆ 𝐼 , 𝑊 ± ] = ± 𝑊 ± . But notice that we now have a highly interesting statement about thesenumber-generating gauge bosons themselves: all of their quantum charges must vanishsince they belong to the Cartan subalgebra (and hence commute with each other). In otherwords, the number of neutral gauge bosons in a theory is given by the rank of its symmetry They could also be Higgs SU ( ) doublets, but we ignore this possibility here. There are of course many different choices of a Cartan subalgebra for a given semisimple Lie algebra, butall of these are related by automorphisms of 𝔤 , so we abuse terminology and speak here as though our choicewere unique. Clearly, this action is non-trivial only if 𝐺 is non-abelian. . 3 . O T H E R S I G NAT U R E S 59 group [59]. The bearing of this discussion on grand unified theories is straightforward:when we break from E → Spin ( ) → SU ( ) → 𝐺 SM , we break from a rank 6 group toa 5 to a 4, and then once again to a 4. Hence, while additional neutral gauge bosons areforbidden in the (standard) SU ( ) theory, both Spin ( ) and E each obtain one additionalneutral gauge boson when their symmetry is broken.This is where representation theory ends and quantum field theory begins. Theliterature on neutral 𝑍 (cid:48) bosons is vast and varied, and we will not undertake a review of thesame here; we point the reader instead to [57, 59] and references therein for summaries ofthe physics and the phenomenology respectively; both cover E in some detail. Reference[73] has the current exclusion limits on the masses of these 𝑍 (cid:48) bosons for the Spin ( ) andE theories: the lower bounds are all on the order of 10 GeV.Let us now consider sterile neutrinos. We have assumed throughout our analysisthat these exist, incorporating them into the SU ( ) theory, and then consequently into theSpin ( ) and E theories. Rosner [77] recently carried out a phenomenological analysisfor the sterile neutrinos in the E theory, of which there are three, from the three copiesof 𝑁 . In his framework, the traditional candidates for sterile neutrinos, the 𝜈 𝑅 ’s, obtainextremely heavy masses and become unimportant, while the Spin ( ) singlets (the finalentry in table 3.1) acquire light masses, and are promoted to sterile neutrino status. Theother exotic fermions in the theory remain heavy, and mix only weakly with the StandardModel fermions, as per the survival hypothesis. Finally, he notes that only two of thesethree Spin ( ) singlets are required to account for present data in neutrino oscillationexperiments, leaving one neutrino free to be a candidate dark matter particle. This wouldfit neatly into the picture of “dark electromagnetism” proposed by Ackerman et al. [2]:in this scenario, dark matter particles interact via a new gauge boson corresponding tosome U ( ) theory that is unbroken at the vacuum—the U ( ) (cid:48) gauge group that arises inthe breaking of the E theory is a natural candidate for the same. Schwichtenberg however,argues in [80] that the Spin ( ) singlet is not the correct choice for a candidate dark matterparticle in the E theory. Instead, he makes a case for the exotic neutrino in the Λ ⊗ 𝜉 representation of SU ( ) × U ( ) (cid:48) (this is the third entry from the bottom in table 3.1). Divinginto the details of this interesting debate is unfortunately beyond our scope here; we simplywanted to note that the exotic fermions in the E theory provide a playground to exploresuch ideas.We end with a brief discussion on perhaps the most famous prediction of grand unifiedtheories: proton decay. Simply put, since each inclusion of 𝐺 SM into the unification gaugegroups that we have been considering involves significant jumps in dimension—from 12 to24 to 45 and finally to the 78-dimensional E —we obtain at each step a huge number of newgauge bosons. These mediate new interactions between particles, one of which is the decayof the proton, which is stable in the standard model; the review article [56] by Langackertreats the subject of proton decay in depth. The original SU ( ) model predicted a maximumproton lifetime on the order of 10 years [28] which was subsequently disproved by theSuper-K(amiokande) experiment. Following a long period where the Spin ( ) theorywas thought to be as dead as the SU ( ) , Bertolini et al. [13] re-examined proton decay inSpin ( ) and discovered that it was still viable. For the E theory, [60] and [75] are two earlyreferences that go into great detail regarding proton decay via many possible symmetrybreaking chains of E ; one take-away point is that in almost every case, the proton decayrate for the SU ( ) theory is a lower-bound for the same in the E theory, with the possibilityof extending the proton lifetime by some orders of magnitude above the SU ( ) bounddepending on how certain parameters in the theory are chosen.This brings us to our final point: the parameter space of grand unified theories is [1] is the most recent publication from the collaboration, summarising data from the 20 years (!) that thisexperiment has been running.0 4 . A S P E C T S O F P H E NO M E NO L O G Y (usually) large enough for all manner of tinkering and fine-tuning to match data. In somesense then, they still have a shot at corresponding to reality. And yet, we steadily seemto be approaching a point where testing them is getting difficult to the point of beinginfeasible. As a recent article notes,But while [the aforementioned] Super-K could suddenly strike gold in thenext few years and confirm one of these models, it could also run for another 20years, nudging up the lower limit on the proton’s lifetime, without definitivelyruling out any of the models.Japan is considering building a $1 billion detector called Hyper-Kamiokande,which would be between 8 and 17 times bigger than Super-K and would besensitive to proton lifetimes of 10 years after two decades. It might startseeing a trickle of decays. Or it might not. “We could be unlucky,” [S. M.] Barr said. “We could build the biggest detector that anyone is ever going to buildand protons decay just a little bit too slow and then we’re out of luck.” [27]Indeed. So while the incompleteness and seeming arbitrariness of the Standard Modelremain strong motivators to seek a more complete, natural physics, it seems that the bestwe can do right now, at least as regards grand unified theories, is simply to wait. It wouldappear that nature is not so keen to give up her secrets just yet. One of the inventors of the flipped SU ( ) theory [12]. ppendix The Construction of G The reference for this appendix is [4]. For large values of 𝑛 , the Dynkin diagrams of type 𝐴 𝑛 , 𝐵 𝑛 , 𝐶 𝑛 , 𝐷 𝑛 are distinct. For small values of 𝑛 , we have the possibility of exceptionalisomorphisms between the classical groups as follows.(i) Spin ( ) (cid:27) SU ( ) . Both have dimension 15, rank 3 and the Dynkin diagram .(ii) Spin ( ) (cid:27) Sp ( ) . Both have dimension 10, rank 2 and the Dynkin diagram .(iii) Spin ( ) (cid:27) SU ( ) (cid:27) Sp ( ) = 𝑆 ⊂ H .(iv) Spin ( ) (cid:27) 𝑆 × 𝑆 .We prove the first two isomorphisms. Proposition A.1.
We have the following isomorphisms of Lie groups:
Spin ( ) (cid:27) SU ( ) and Spin ( ) (cid:27) 𝑆𝑝 ( ) . Proof.
We do both cases in parallel. Spin ( ) has the representation Δ of degree 4 over C and degree 2 over H . We can impose a Hermitian form, invariant under the compactgroup Spin ( ) , giving us a homomorphism from Spin ( ) → Sp ( ) which we also denotewith Δ . Similarly, Spin ( ) has the representation Δ + of degree 4 over C and we have Δ + : Spin ( ) → U ( ) . We first wish to show that Im Δ + ⊂ SU ( ) .Let 𝑡 ∈ 𝑇 ⊂ Spin ( ) . Then 𝑡 acts with eigenvalues defined by weights (cid:26) ( 𝑥 + 𝑥 + 𝑥 ) , ( 𝑥 − 𝑥 − 𝑥 ) , (− 𝑥 + 𝑥 − 𝑥 ) , (− 𝑥 − 𝑥 + 𝑥 ) (cid:27) ;these add to zero so the eigenvalues multiply to 1 and 𝑡 must act with determinant 1.Hence, any 𝑔𝑡 𝑔 − acts with det 1 and Δ + maps to SU ( ) .Now Δ is faithful: if 𝑔 ∈ Spin ( 𝑛 + ) and 𝑔 ↦→ C ( Δ , Δ ) (cid:27) Cl ( 𝑉 ) then 𝑔 is 1.Also Δ + is faithful if 𝑛 is odd, for if 𝑔 acts as 1 on Δ + , then 𝑔 ∈ Spin ( 𝑛 ) acts as 1 on thedual, Δ − , so 𝑔 ↦→ C ( Δ + , Δ + ) + Hom C ( Δ − , Δ − ) = Cl ( 𝑉 ) . Thus 𝑔 =
1. Hence thetwo maps Δ , Δ + are injective homomorphisms, and induce injective homomorphisms d Δ ,d Δ + ; they are thus isomorphisms for dimensional reasons. Hence, Δ and Δ + map smallneighbourhoods in Spin ( ) , Spin ( ) onto small neighbourhoods of Sp ( ) , SU ( ) respectively.But Sp ( ) and SU ( ) are connected, so Δ : Spin ( ) → Sp ( ) and Δ + : Spin ( ) → SU ( ) mustbe surjections. QED Corollary A.2.
The group
Spin ( ) acts transitively on the unit sphere 𝑆 ⊂ Δ . This is because Sp ( ) is defined to be the subgroup of GL ( , H ) under which the inner product on H isinvariant. 612 . T H E CO N S T RU C T I O N O F G Proof. Sp ( ) acts transitively on 𝑆 ⊂ H . QED Corollary A.3.
The group
Spin ( ) acts transitively on pairs ( 𝑥, 𝑧 ) , 𝑥 ∈ 𝑆 ⊂ R , 𝑧 ∈ 𝑆 ⊂ Δ + .Moreover the subgroup fixing 𝑧 ∈ 𝑆 ⊂ Δ + may be taken, by a suitable choice of 𝑧 , to be SU ( ) ⊂ Spin ( ) , where this inclusion arises by lifting the composite SU ( ) ↩ → U ( ) ↩ → SO ( ) to Spin ( ) . Proof.
Spin ( ) covers SO ( ) , which acts transitively on 𝑆 . Taking a suitable 𝑥 , e.g. 𝑥 = ( , . . . , , ) , the subgroup fixing 𝑥 is Spin ( ) . The restriction of Δ + to Spin ( ) is Δ (byproposition 2.4.19) and Spin ( ) acts transitively on 𝑧 ∈ 𝑆 ⊂ Δ by the last corollary. Theweights of Δ + are (cid:8) ( 𝑥 + 𝑥 + 𝑥 ) , ( 𝑥 − 𝑥 − 𝑥 ) , . . . (cid:9) and their restrictions to 𝑇 ⊂ SU ( ) are { , 𝑥 , 𝑥 , 𝑥 } ; hence the restriction of Δ + to SU ( ) is 1 + Λ . So this SU ( ) fixes pointsof 𝑆 ⊂ Δ + : in fact, a whole circle of them. For any such fixed point, the subgroupfixing it is no bigger, since in SU ( ) (cid:27) Spin ( ) the subgroup fixing a unit vector in C is anSU ( ) . QED Corollary A.4.
The group
Spin ( ) acts transitively on triples ( 𝑥, 𝑦, 𝑧 ) where 𝑥, 𝑦 are orthogonaland 𝑧 ∈ 𝑆 ⊂ Δ . Proof.
Spin ( ) covers SO ( ) , which is transitive on points 𝑦 ∈ 𝑆 . Choose 𝑦 = ( , . . . , , ) .Then the subgroup fixing 𝑦 is Spin ( ) . Now Δ restricts to Δ + + Δ − over C (by proposition2.4.19 again), but starting with Δ as a fixed vector space of dimension 8 over R , the restrictionto Spin ( ) is the representation of dimension 8 underlying Δ + (or Δ − ). Finally, note that bythe previous corollary, Spin ( ) is transitive on pairs ( 𝑥, 𝑧 ) , 𝑥 ∈ 𝑆 , 𝑧 ∈ 𝑆 ⊂ Δ + . QED Theorem A.5.
Consider the subgroup 𝐺 of Spin ( ) which fixes a point 𝑧 ∈ 𝑆 ⊂ Δ . Then 𝐺 isa compact, connected, simply connected Lie group of rank 2 and dimension 14, with the Dynkindiagram and commonly called G . Moreover, G is transitive on pairs ( 𝑥, 𝑦 ) of orthogonalvectors in 𝑆 ⊂ R . Proof.
The last sentence follows from the above corollary. Now clearly, 𝐺 is a closedsubgroup of Spin ( ) . Since Spin ( ) is transitive on 𝑆 , we have dim 𝐺 = dim Spin ( ) − dim 𝑆 = − =
14. Let 𝐻 ⊂ 𝐺 be the subgroup that fixes 𝑦 = ( , , . . . , ) . Then 𝐻 isthe same as the subgroup of Spin ( ) which fixes 𝑧 , and by a suitable choice of 𝑧 we cantake 𝐻 = SU ( ) ⊂ Spin ( ) (this was corollary A.3). Since 𝐻 is connected and 𝐺 / 𝐻 = 𝑆 isconnected, we find that 𝐺 is connected. Similarly, 𝐺 is simply connected.We determine the roots of 𝐺 . 𝐻 = SU ( ) acts on 𝔥 ⊂ 𝔤 by the adjoint action;we wish to know how 𝐻 acts on 𝔤 / 𝔥 , the tangent space to 𝑆 at 𝑦 . By construction, 𝑆 = Spin ( )/ Spin ( ) , so we need to look at the action of Spin ( ) on 𝔰𝔭𝔦𝔫 ( )/ 𝔰𝔭𝔦𝔫 ( ) . This isthe geometrically obvious action where the tangent space to 𝑆 at ( , , . . . , ) is R , thespace of the first 6 coordinates, and Spin ( ) acts on it as usual. The weights are hencegiven by {± 𝑥 , ± 𝑥 , ± 𝑥 } . Thus, 𝑇 ⊂ SU ( ) acts on 𝔤 with weights { , , ±( 𝑥 − 𝑥 ) , ±( 𝑥 − 𝑥 ) , ±( 𝑥 − 𝑥 ) , ± 𝑥 , ± 𝑥 , ± 𝑥 } . We conclude that 𝑇 is maximal (so 𝐺 has rank 2), thatthese are the roots of 𝐺 , and that the Dynkin diagram is . QED This follows from 2.1.13 and 2.1.6. The representation 1 + Λ = + C ⊂ C of SU ( ) has SU ( ) acting by diag ( , 𝐴 ) for 𝐴 ∈ SU ( ) . Hence,the action of SU ( ) on 𝑆 ⊂ C fixes a (complex) circle. A familiar analogy is SO ( ) rotating the 3-sphere aboutthe 𝑧 -axis, which fixes an 𝑆 , the north and south poles. Use the homotopy exact sequence of the fibration 𝐻 → 𝐺 → 𝐺 / 𝐻 . This can be seen directly: the maximal torus ˜ 𝑇 of Spin ( ) is given in remark 2.4.13. Since we are inan even-dimensional space, the eigenvalues of this rotation matrix occur in complex conjugate pairs, 𝑒 ± 𝑖𝑥 𝑗 .Hence, the eigenvalues of the action of ˜ 𝔱 on 𝑇 R (cid:27) R are ± 𝑥 𝑗 . See also remark 2.4.16. That is, with the (standard) weights of SU ( ) given in example 2.1.13 together with the weights justcomputed. The trivial representation occurs exactly twice in the list of the 14 (i.e. all) weights.3 G starts life with two obvious representations: a 7-dimensional representation 𝐺 ⊂ Spin ( ) acting on R with weights { , ± 𝑥 , ± 𝑥 , ± 𝑥 } , and the 14-dimensional adjointrepresentation, Ad with weights as in the above theorem. ibliography [1] K. Abe et al. Search for Proton Decay via 𝑝 → 𝑒 + 𝜋 and 𝑝 → 𝜇 + 𝜋 in 0 .
31 megaton · years Exposure of the Super-Kamiokande Water Cherenkov Detector. Physical ReviewD , :012004, Jan. 2017.[2] L. Ackerman, M. R. Buckley, S. M. Carroll, and M. Kamionkowski. Dark Matter andDark Radiation. Physical Review D , :023519, Jan. 2009.[3] J. F. Adams. Lectures on Lie Groups . University of Chicago Press, Chicago, 1983.[4] J. F. Adams.
Lectures on Exceptional Lie Groups . University of Chicago Press, Chicago,1996.[5] S. Albeverio and S. Mazzucchi. Path Integral: Mathematical Aspects.
Scholarpedia , (1):8832, 2011. Revision Physical Review D , :1747–1756, Sep. 1973.[7] M. F. Atiyah, R. Bott, and A. Shapiro. Clifford Modules. Topology , :3–38, July 1964.[8] J. Baez and J. Huerta. The Algebra of Grand Unified Theories. Bulletin of the AmericanMathematical Society , (3):483–552, Mar. 2010.[9] R. Barbieri, D. V. Nanopoulos, and A. Masiero. Hierarchical Fermion Masses in E . Physics Letters , :194–198, 1981.[10] R. Barbieri, D. V. Nanopoulos, G. Morchio, and F. Strocchi. Neutrino Masses in GrandUnified Theories. Physics Letters B , (1):91–97, 1980.[11] R. Barbieri and D.V. Nanopoulos. An Exceptional Model for Grand Unification. Physics Letters B , (3):369–375, 1980.[12] S. M. Barr. A New Symmetry Breaking Pattern for SO ( ) and Proton Decay. PhysicsLetters B , (3):219–222, 1982.[13] S. Bertolini, L. Di Luzio, and M. Malinský. Intermediate Mass Scales in the Nonsuper-symmetric SO ( ) Grand Unification: A Reappraisal.
Physical Review D , :015013,July 2009.[14] A. Bilal. Lectures on Anomalies. Available as arXiv:0802.0634 .[15] S. M. Bilenky. Neutrino Oscillations: Brief History and Present Status. Available as arXiv:1408.2864 .[16] J. D. Bjorken. Weak Interaction Theory and Neutral Currents. In Proceedings, SummerInstitute on Particle Physics, Aug. 2–13, 1976 , volume
C7608021 , pages 1–84, 1976.
B I B L I O G R A P H Y[17] D. Bleecker.
Gauge Theory and Variational Principles . Addison Wesley PublishingCompany, Reading, 1981.[18] M. J. Bowick and P. Ramond. Calculable Masses in Grand Unified Theories.
PhysicsLetters , :338–342, 1981.[19] T. Bröcker and T. tom Dieck. Representations of Compact Lie Groups . Springer-Verlag,New York, 1995.[20] B. Cassen and E. U. Condon. On Nuclear Forces.
Physical Review , (9):846–849, Nov.1936.[21] R. P. Crease. The Second Creation: Makers of the Revolution in Twentieth-Century Physics .Rutgers University Press, New Brunswick, 1996.[22] A. De Rújula, H. Georgi, and S. L. Glashow. Hadron Masses in a Gauge Theory.
Physical Review D , :147–162, July 1975.[23] P. A. M. Dirac. The Quantum Theory of the Electron. Proceedings of the Royal Society ofLondon A: Mathematical, Physical and Engineering Sciences , (778):610–624, 1928.[24] P. A. M. Dirac. The Principles of Quantum Mechanics . Oxford University Press, Oxford,1981.[25] M. Drewes. The Phenomenology of Right Handed Neutrinos.
International Journal ofModern Physics E , (08):1330019, Aug. 2013.[26] E. Fermi. Versuch einer Theorie der 𝛽 -Strahlen. I. Zeitschrift f ur Physik , (3-4):161–177,Mar. 1934.[27] R. Fisker. Grand Unification Dream Kept at Bay, QuantaMagazine . Retrieved from , Dec. 15,2016.[28] P.H. Frampton and S.L. Glashow. Staying Alive with SU ( ) . Physics Letters B , (4):340–342, 1983.[29] H. Fritzsch, M. Gell-Mann, and H. Leutwyler. Advantages of the Color Octet GluonPicture. Physics Letters B , (4):365–368, Nov. 1973.[30] H. Fritzsch and P. Minkowski. Unified Interactions of Leptons and Hadrons. Annalsof Physics , (1):193–266, 1975.[31] J. Fuchs and C. Schweigert. Symmetries, Lie Algebras and Representations . CambridgeUniversity Press, Cambridge, 2003.[32] W. Fulton and J. Harris.
Representation Theory; A First Course . Springer-Verlag, NewYork, 1999.[33] H. Georgi. The State of the Art—Gauge Theories.
AIP Conference Proceedings , :575–582, 1975.[34] H. Georgi. Towards a Grand Unified Theory of Flavor. Nuclear Physics B , (1):126–134, 1979.[35] H. Georgi and S. L. Glashow. Gauge Theories Without Anomalies. Physical Review D , :429–431, Jul. 1972. I B L I O G R A P H Y [36] H. Georgi and S. L. Glashow. Unity of All Elementary-Particle Forces. Physical ReviewLetters , (8):438–441, Feb. 1974.[37] H. Georgi, H. R. Quinn, and S. Weinberg. Hierarchy of Interactions in Unified GaugeTheories. Physical Review Letters , :451–454, Aug. 1974.[38] S. L. Glashow, J. Iliopoulos, and L. Maiani. Weak Interactions with Lepton-HadronSymmetry. Physical Review D , :1285–1292, Oct. 1970.[39] D. J. Gross. Twenty Five Years of Asymptotic Freedom. Nuclear Physics B - ProceedingsSupplements , (1-3):426–446, Mar. 1999.[40] F. Gürsey, P. Ramond, and P. Sikivie. A Universal Gauge Theory Model Based on E . Physics Letters , B60 :177–180, 1976.[41] F. Gürsey and M. Serdaro ˇglu. Basic Fermion Masses and Mixings in the E Model.
Lettere al Nuovo Cimento , :28, 1978.[42] F. Gürsey and M. Serdaro ˇglu. E Gauge Field Theory Model Revisited.
Nuovo Cimento , A65 :337, 1981.[43] B. Hall.
Lie Groups, Lie Algebras, and Representations: An Elementary Introduction .Springer-Verlag, New York, 2004.[44] M. J. D. Hamilton. Lecture Notes on Mathematical Gauge TheoryI. Retrieved from , May 2017.[45] M. J. D. Hamilton. The Higgs Boson for Mathematicians. Lecture Notes on GaugeTheory and Symmetry Breaking. Available as arXiv:1512.02632 .[46] H. Harari. Three Generations of Quarks and Leptons. In
V International Conference onMeson Spectroscopy Boston, Mass., Apr. 29–30, 1977 , page 0170, 1977.[47] A. Hartanto and L. T. Handoko. Grand Unified Theory Based on the SU ( ) Symmetry.
Physical Review D , :095013, May 2005.[48] J. A. Harvey. Patterns of Symmetry Breaking in the Exceptional Groups. NuclearPhysics B , :254–260, 1980.[49] W. Heisenberg. Über den Bau der Atomkerne. I. Zeitschrift für Physik , (1):1–11, 1932.[50] R. C. Helling. How I Learned to Stop Worrying and Love QFT. Available as arXiv:1201.2714 .[51] L. Hoddeson, L. Brown, M. Riordan, and M. Dresden, editors. The Rise of the StandardModel: A History of Particle Physics from 1964 to 1979 . Cambridge University Press,Cambridge, 2006.[52] Nathan Jacobson.
Lie Algebras . Dover Publications Inc., New York, 1979.[53] A. Joglekar and J. L. Rosner. Searching for Signatures of E . Available as arXiv:1607.06900 .[54] R. C. King. Fundamental Fermion Representations in Generalisations of the SU ( ) Grand Unified Gauge Theory.
Nuclear Physics B , 185(1):133 – 156, 1981.[55] L. D. Landau and E. M. Lifshitz.
The Classical Theory of Fields . Elsevier Ltd., Oxford,1995. B I B L I O G R A P H Y[56] P. Langacker. Grand Unified Theories and Proton Decay.
Phyics Reports , :185, 1981.[57] P. Langacker. The Physics of Heavy 𝑍 (cid:48) Gauge Bosons.
Reviews of Modern Physics , :1199–1228, Aug. 2009.[58] T. D. Lee and C. N. Yang. Question of Parity Conservation in Weak Interactions. Physical Review , (1):254–258, Oct. 1956.[59] A. Leike. The Phenomenology of Extra Neutral Gauge Bosons. Physics Reports , (3-4):143–250, 1999.[60] D. London and J. L. Rosner. Extra Gauge Bosons in E . Physical Review D , :1530–1546,Sep. 1986.[61] J. D. Lykken. Beyond the Standard Model. In CERN Yellow Report CERN-2010-002,101–109 , 2010.[62] W. J. Marciano. Weak Mixing Angle and Grand Unified Gauge Theories.
PhysicalReview D , :274–288, July 1979.[63] A. Masiero. On the Phenomenological Group in the Unified SO ( ) Model.
PhysicsLetters B , 93(3):295 – 298, 1980.[64] M. L. Mehta. Classification of Irreducible Unitary Representations of Compact SimpleLie Groups. I.
Journal of Mathematical Physics , (10):1824–1832, 1966.[65] M. L. Mehta and P. K. Srivastava. Classification of Irreducible Unitary Representationsof Compact Simple Lie Groups. II. Journal of Mathematical Physics , (10):1833–1835,1966.[66] R. N. Mohapatra. Unification and Supersymmetry . Springer-Verlag, New York, 2002.[67] P. J. Mohr, B. N. Taylor, and D. B. Newell. CODATA Recommended Values of theFundamental Physical Constants: 2010.
Reviews of Modern Physics , :1527–1605, Nov.2012.[68] M. Nakahara. Geometry, Topology and Physics . Taylor & Francis Ltd., Boca Raton, 2003.[69] C. Nash.
Differential Topology and Quantum Field Theory . Academic Press, London,1991.[70] S. Okubo. Gauge Groups without Triangular Anomaly.
Physical Review D , :3528–3534, Dec. 1977.[71] Susumu Okubo. Casimir Invariants and Vector Operators in Simple and Classical LieAlgebras. Journal of Mathematical Physics , (12):2382–2394, 1977.[72] J. Patera and R. T. Sharp. Branching Rules for Representations of Simple Lie Algebrasthrough Weyl Group Orbit Reduction. Journal of Physics A: Mathematical and General , (13):2329, 1989.[73] C. Patrignani et al. Review of Particle Physics. Chinese Physics , C40 (10):100001, 2016.[74] M. E. Peskin and D. V. Schroeder.
An Introduction to Quantum Field Theory . Addison-Wesley Publishing Company, Reading, 1995.[75] R. W. Robinett and J. L. Rosner. Mass Scales in Grand Unified Theories.
PhysicalReview D , :2396–2419, Nov. 1982. I B L I O G R A P H Y [76] C. Rosen. All About SU ( ) : Unification and the Standard Model of Particle Physics.Retrieved from
Quantum Field Theory and the Standard Model . Cambridge UniversityPress, Cambridge, 2014.[80] J. Schwichtenberg. Dark Matter in E Grand Unification. Available as arXiv:1704.04219 .[81] R. Slansky. Group Theory for Unified Model Building.
Physics Reports , (1):1–128,Dec. 1981.[82] S. V. Troitsky. Unsolved Problems in Particle Physics. Physics-Uspekhi , (1):72–95,Jan. 2012.[83] I. Umemura and K. Yamamoto. Symmetry Breaking Patters in an SU ( ) Grand UnifiedModel.
Progress of Theoretical Physics , :1430, 1981.[84] V. S. Varadarajan. Historical Review of Lie Theory. Retreived from , n.d.[85] S. Verma. Theoretical and Phenomenological Status of Neutrino Physics: A BriefReview. Advances in High Energy Physics , :1–15, 2015.[86] J. von Neumann. Mathematical Foundations of Quantum Mechanics . Princeton UniversityPress, Princeton, 1996.[87] D. P. Želobenko.
Compact Lie Groups and Their Representations (Translations of Mathe-matical Monographs Reprint) . American Mathematical Society, Providence, 1973.[88] S. Weinberg. A Model of Leptons.
Physical Review Letters , :1264–1266, Nov. 1967.[89] S. Weinberg. Mixing Angle in Renormalizable Theories of Weak and ElectromagneticInteractions. Physical Review D , :1962–1967, Apr. 1972.[90] S. Weinberg. The Quantum Theory of Fields Volume 1: Foundations . Cambridge UniversityPress, Cambridge, 2005.[91] E. Wigner. On Unitary Representations of the Inhomogeneous Lorentz Group.
TheAnnals of Mathematics , (1):149, Jan. 1939.[92] Y. Wong and Y. Au-Yeung. An Elementary and Simple Proof of the Connectedness ofthe Classical Groups. The American Mathematical Monthly ,74