The Gauge Group Ceiling: Why the Biquaternion Algebra Reaches SU(2) but Not SU(3)

Introduction

The companion articles Non-Abelian Gauge Fields in Biquaternionic Form and Instantons and Solitons in Biquaternionic Form build their non-abelian gauge theory on the compact factor

$$ \mathrm{SU}(2)=\mathrm{span}_{\mathbb R}\{e_1,e_2,e_3\}\subset\mathbb M_- , $$

and the abelian factor on the central phase $\mathbb C_{\mathbb B}$. The reason is not a choice of convenience, and it is not that nobody has tried the alternative: the biquaternion algebra reaches $SU(2)$ and it does not reach $SU(3)$. This article states that ceiling precisely, proves it from the algebra's own structure, and identifies what has to be added to pass it. The result is a result, not a failure: a framework that knows exactly which gauge groups it can carry and which it cannot is more informative than one that leaves the question open.

Three distinct questions must be separated, because they have different answers and conflating them produces a spurious "problem".

1. What is the gauge group of the fields that live in the algebra itself? The fields are valued in the algebra $\mathbb B$ or in its defining module; the group that acts unitarily on them is intrinsic, and it is the answer to this question that the article establishes: $U(2)$, i.e. $SU(2)\times U(1)$ up to finite identification.
2. What gauge group can be embedded in a matrix algebra over $\mathbb B$? Once the carrier is enlarged to $M_n(\mathbb B)\cong M_{2n}(\mathbb C)$, the intrinsic group is $U(2n)$ and any compact Lie group of dimension at most $4n^2$ and a faithful $2n$-dimensional unitary representation embeds. $SU(3)$ embeds for $n\ge2$, because $SU(3)\subset U(3)\subset U(4)$. This is an embedding into a larger group, not the algebra's own group.
3. Can the colour triplet be a module of the algebra? No. A module over $\mathbb B\cong M_2(\mathbb C)$ has even complex dimension, and colour requires a complex three-dimensional space.

The article answers all three, and the answers are consistent: the algebra's intrinsic compact gauge algebra is $\mathrm{U}(2)$, of real dimension four; $SU(3)$ needs an eight-dimensional compact algebra and cannot be contained in it; and the smallest carrier that contains $SU(3)$ is $M_2(\mathbb B)$, whose intrinsic group is $U(4)$.

Conventions. We use those of Conventions in the Biquaternion Universe and of the companion gauge articles. The algebra is $\mathbb B=\mathbb C\otimes_{\mathbb R}\mathbb H$ with $e_k^2=-e_0$, $e_1e_2=e_3$, central $i$, and $\mathbb B\cong M_2(\mathbb C)$ as a complex algebra. The compact (anti-Hermitian) sector is $\mathbb M_-=\mathrm{span}_{\mathbb R}\{ie_0,e_1,e_2,e_3\}$ and the informational sector is $\mathbb M_+=\mathrm{span}_{\mathbb R}\{e_0,ie_1,ie_2,ie_3\}$. The commutator bracket is $[x,y]=xy-yx$, the generators of the compact factor are $T_a=\tfrac12e_a$ with $[T_a,T_b]=\varepsilon_{abc}T_c$, and the abelian generator is $T_0=\tfrac12 ie_0$. The matrix trace on the defining module is distinguished from the informational trace $\mathrm{Tr}(\tilde P\tilde H)=2\,\mathrm{Sc}(\tilde P\tilde H)$.

• Companion article Non-Abelian Gauge Fields in Biquaternionic Form, for the non-abelian gauge structure.
• The mathematics articles Biquaternion Rotations and Lorentz Transformations and Biquaternion Other Algebraic Element Representations, for the representation theory of the algebra and for its spinor and Clifford realizations.
• The mathematics article Biquaternion Topology, for the topological obstructions.
• Companion article The Hopf Fibration and the Biquaternion Gauge Bundle, for the intrinsic $U(2)$ as a structure group.
• Companion article Grand Unification and the Biquaternion Algebra Ceiling, for the consequence for unification.

The Algebra's Own Compact Structure

The gauge fields of the framework are elements of the algebra. The Lie algebra generated by taking commutators of the algebra with itself is

$$ [\mathbb B,\mathbb B] \;=\; \mathrm{SL}(2,\mathbb C) , $$

the traceless part of $\mathbb B\cong M_2(\mathbb C)$, of real dimension $6$: the commutators of the quaternion units give the imaginary quaternions, $[e_a,e_b]=2\varepsilon_{abc}e_c$, and the commutators mixing one factor of $i$ with one quaternion unit give their Hermitian partners, $[ie_a,e_b]=2i\varepsilon_{abc}e_c$. The full algebra therefore splits, as a real vector space, into the derived algebra and the centre,

$$ \mathbb B \;=\; \mathrm{SL}(2,\mathbb C)\;\oplus\;\mathbb C_{\mathbb B}, \qquad \dim_{\mathbb R}\mathrm{SL}(2,\mathbb C)=6, \qquad \dim_{\mathbb R}\mathbb C_{\mathbb B}=2 , $$

with the centre $\mathbb C_{\mathbb B}=\mathrm{span}_{\mathbb R}\{e_0,ie_0\}$ commuting with everything. Both facts were verified on explicit matrices: the bracket closure gave exactly the six-dimensional space of traceless matrices, and every central element commuted with every algebra element to machine precision.

The group that acts unitarily on the algebra is built from its anti-Hermitian part. The anti-Hermitian conjugation $\flat$ and the Hermitian conjugation ${}^{*}$ of Conventions in the Biquaternion Universe split the algebra into the two four-dimensional real subspaces

$$ \mathbb M_-=\bigl\{x\in\mathbb B:x^\flat=x\bigr\}=\mathrm{span}_{\mathbb R}\{ie_0,e_1,e_2,e_3\}, \qquad \mathbb M_+=\bigl\{x\in\mathbb B:x^{*}=x\bigr\}=\mathrm{span}_{\mathbb R}\{e_0,ie_1,ie_2,ie_3\}, $$

of real dimensions $4$ and $4$, with $x^\flat=-x^{*}$ and $x^{*}=\bar x^{\,*}$; quaternion conjugation alone ($e_k\mapsto-e_k$, $i$ fixed) preserves both sectors but has as its fixed space neither of them — it is the centre $\mathbb C_{\mathbb B}$. The anti-Hermitian sector is the compact, unitary direction: exponentiating it gives the unitary group of the defining module,

$$ \exp\bigl(\mathbb M_-\bigr)=U(2), \qquad \mathbb M_-\;=\;\mathrm{U}(2)\;=\;\mathrm{SU}(2)\oplus\mathrm{U}(1) , \qquad \dim_{\mathbb R}\mathrm{U}(2)=4 . $$

This is the algebra's intrinsic gauge structure, and it is verified by the dimension count and the commutators: the three $e_a$ close on $\mathrm{SU}(2)$ with the structure constants $\varepsilon_{abc}$, and $ie_0$ is central and spans the $\mathrm{U}(1)$. The gauge group of fields valued in $\mathbb B$ is therefore

$$ G_{\mathbb B}=U(2), $$

or $SU(2)\times U(1)$ if the determinant phase is separated, the two being locally isomorphic since $U(2)=(SU(2)\times U(1))/\mathbb Z_2$, with the compact factor $SU(2)$ reached by the imaginary quaternions and the abelian factor by the central imaginary.

Every natural construction gives the same answer. The intrinsic group can be approached from four directions, and they agree.

• Unitaries of the defining module. $\mathbb B\cong M_2(\mathbb C)$ acts on the defining module $S\cong\mathbb C^2$, and the unitaries of this action are $U(2)$.
• Automorphisms. The $\mathbb C$-algebra automorphisms of $\mathbb B\cong M_2(\mathbb C)$ are $PGL(2,\mathbb C)=PSL(2,\mathbb C)$, whose maximal compact subgroup is $PU(2)=SO(3)=SU(2)/\mathbb Z_2$; over $\mathbb R$ the Galois conjugation of $i$ is an additional outer automorphism, and the compact part remains the same. The automorphism group of the imaginary quaternions is $SO(3)=SU(2)/\mathbb Z_2$.
• Norm-preserving rotors. The transformations preserving the biquaternion norm $N(\tilde Q)=\tilde Q\tilde Q^{\natural}$ act as left and right quaternion multiplication, $SU(2)_L\times SU(2)_R/\mathbb Z_2=SO(4)$; the diagonal subgroup is the $SU(2)$ of the present gauge theory, and the axial combination is the chirality rotation of Chiral Fermions in the Biquaternion Framework, not a gauge group.
• Stabilisers. The stabiliser of a unit pure-imaginary element $\hat\mu$, which is the unbroken $U(1)$ of the monopole article, is a maximal torus of $SU(2)$; the pattern of unbroken subgroups is the pattern of $SU(2)$, with a one-dimensional maximal torus.

Four constructions, one ceiling.

Why SU(3) Cannot Fit

The obstruction to $SU(3)$ has a dimension form, a module form, and an algebraic form, and the three are independent.

Dimension. The compact gauge algebra of the algebra is $\mathrm{U}(2)$, of real dimension $4$. The Lie algebra $\mathrm{SU}(3)$ has real dimension $8$. A subalgebra cannot be larger than its parent,

$$ \dim_{\mathbb R}\mathrm{SU}(3)=8 \;>\; 4=\dim_{\mathbb R}\mathrm{U}(2) , $$

so no eight-dimensional compact algebra is contained in the four-dimensional compact sector of $\mathbb B$. This is the crudest form of the ceiling and it is already decisive. The full algebra $\mathbb B$ has real dimension $8$, equal to the dimension of $\mathrm{SU}(3)$; but its eight dimensions are the six non-compact directions of $\mathrm{SL}(2,\mathbb C)$ together with the two-dimensional centre, and the compact directions among them number only four. The equality of dimensions is a numerical coincidence and not an embedding: $\mathrm{SL}(2,\mathbb C)$ is not compact and is not $\mathrm{SU}(3)$, and no change of basis inside $\mathbb B$ turns the non-compact directions into compact ones, because compactness of the generators is a property of the conjugation that defines the norm and is preserved by it. The same count can be read on the group: $U(2)$ has dimension four, $SU(3)$ has dimension eight.

Module. A field of the framework is valued in a module of $\mathbb B$, and the simple module is the defining module $S\cong\mathbb C^2$. Every $\mathbb B$-module is a direct sum of copies of $S$, so its complex dimension is even:

$$ V\;\text{a }\mathbb B\text{-module}\quad\Longrightarrow\quad \dim_{\mathbb C}V=2r\in2\mathbb Z . $$

The colour degree of freedom of quantum chromodynamics is an irreducible three-dimensional complex representation of $SU(3)$, the triplet $\mathbf 3$. A three-dimensional complex space is not a $\mathbb B$-module, because $3$ is odd; and it cannot be embedded in one either, because a submodule of a $\mathbb B$-module is again a $\mathbb B$-module and therefore has even complex dimension, so no three-dimensional subspace of a $\mathbb B$-module is closed under the algebra. Enlarging the algebra does not remove the parity: $M_n(\mathbb B)\cong M_{2n}(\mathbb C)$ has simple module of dimension $2n$, still even. The colour triplet therefore has no home as a module of the biquaternion algebra: the algebra's modules come in even complex dimension, and colour is irreducibly odd. Combined with the dimension count, this is the module-level statement of the ceiling.

Algebraic. The real algebras that carry a colour-like three-fold internal structure are the octonions $\mathbb O$ and their relatives: the automorphism group of the octonions is the exceptional group $G_2$, which contains $SU(3)$ as the stabiliser of a preferred imaginary unit; this is the standard route from the normed division algebras to $SU(3)$, and it passes through $\mathbb O$, of real dimension eight. The biquaternion algebra is built from $\mathbb H$, and the associative normed division algebras over $\mathbb R$ are exactly $\mathbb R,\mathbb C,\mathbb H$ by the Frobenius theorem, of dimensions $1,2,4$. There is no associative division algebra of dimension $3$, and there is none of dimension $8$; the dimension-$8$ division algebra is $\mathbb O$, and it is non-associative. A three-dimensional complex internal space with a colour symmetry therefore has no associative-division-algebra home at the algebra's own dimension, and the framework, which is associative by construction, cannot provide one.

Summary of the three arguments. The compact sector is four-dimensional, so $\mathrm{SU}(3)$ does not fit; the modules are even-dimensional, so the colour triplet is not a module; and the smallest normed division algebra with an $SU(3)$ inside its automorphism group is the non-associative octonions, which are outside the framework. Each argument alone is sufficient.

The Cartan Decomposition and the Maximal Compact Subalgebra

The dimension argument is decisive, but it is worth seeing it in the Lie-theoretic form, because that form explains why no change of basis, no redefinition of the generators and no clever choice of an invariant subspace can produce an eight-dimensional compact subalgebra.

The conjugation ${}^{*}$ defines an involution of the real Lie algebra $\mathbb B$ by

$$ \theta(x) = -x^{*} , \qquad \theta^2=\mathrm{id} , $$

whose fixed space is the anti-Hermitian sector $\mathbb M_-$ and whose $(-1)$-eigenspace is the Hermitian sector $\mathbb M_+$. The algebra therefore decomposes as a real vector space,

$$ \mathbb B = \mathbb M_-\oplus\mathbb M_+ , $$

with the bracket relations

$$ [\mathbb M_-,\mathbb M_-]\subset\mathbb M_- , \qquad [\mathbb M_-,\mathbb M_+]\subset\mathbb M_+ , \qquad [\mathbb M_+,\mathbb M_+]\subset\mathbb M_- . $$

This is a Cartan decomposition: $\mathbb M_-$ is a maximal compact subalgebra, $\mathbb M_+$ its complement, and the relations are the standard ones of a symmetric pair. The bracket relations were verified on the eight basis elements, with no violation, in the defining two-dimensional representation. The Killing form of $\mathbb B$, read as a real Lie algebra, is negative semidefinite on $\mathbb M_-$ and positive semidefinite on $\mathbb M_+$, with the kernel in each case the two-dimensional centre $\mathbb C_{\mathbb B}$:

$$ K(e_a,e_b)=-16\,\delta_{ab}, \qquad K(ie_a,ie_b)=+16\,\delta_{ab}\ \ (a,b=1,2,3), \qquad K(ie_0,\cdot)=K(e_0,\cdot)=0 , $$

evaluated in that representation. It is definite on the semisimple parts $\mathrm{SU}(2)=\mathrm{span}\{e_1,e_2,e_3\}$ and $i\mathrm{SU}(2)=\mathrm{span}\{ie_1,ie_2,ie_3\}$, and the degeneracy is exactly the centre, which is what a reductive algebra gives. Since negative semidefiniteness on a subalgebra is what compactness of a real Lie algebra means, a compact subalgebra must be contained in a conjugate of $\mathbb M_-$. Every maximal compact subalgebra is conjugate to $\mathbb M_-$, and therefore every compact subalgebra of $\mathbb B$ has dimension at most

$$ \dim_{\mathbb R}\mathbb M_- = 4 . $$

The largest compact subgroup therefore has dimension four, and it is the $U(2)$ of the intrinsic gauge structure. A compact eight-dimensional subalgebra cannot exist, not because of any particular choice of basis but because the maximal compact dimension is four; this is the Lie-theoretic content of the first of the three arguments. The Cartan relations also expose the non-compact directions: the four-dimensional complement $\mathbb M_+$ fails to be a subalgebra, since $[\mathbb M_+,\mathbb M_+]\subset\mathbb M_-$ takes it out of itself, and it is precisely this non-closure that prevents the eight real dimensions of $\mathbb B$ from organising into a compact eight-dimensional algebra.

For comparison, the matrix algebra over the complex numbers shows how the ceiling lifts. In $M_n(\mathbb C)$ the involution $x\mapsto -x^{*}$ has maximal compact fixed space $\mathrm{U}(n)$, of real dimension $n^2$, so the largest compact subalgebra grows quadratically. For the biquaternion algebra $n=2$ and the maximum is $4$, which is $SU(2)\times U(1)$ and no more; for the enlarged carrier $M_n(\mathbb B)\cong M_{2n}(\mathbb C)$ the maximum is $(2n)^2$, and $\mathrm{SU}(3)$ of dimension $8$ fits once $(2n)^2\ge8$, i.e. $n\ge2$. The ceiling is thus exactly the statement $n=1$ in this family.

What Would Be Needed to Pass the Ceiling

The ceiling is not a wall around the framework; it is a wall around the four-dimensional algebra, and it can be passed by enlarging the carrier. The enlargement is explicit and its cost is explicit.

Matrix enlargement. Pass from $\mathbb B$ to the algebra of $n\times n$ matrices over $\mathbb B$,

$$ M_n(\mathbb B)\;\cong\;M_{2n}(\mathbb C), \qquad [M_n(\mathbb B),M_n(\mathbb B)]=\mathrm{SL}(2n,\mathbb C), \qquad \text{compact part}=\mathrm{U}(2n), $$

whose anti-Hermitian sector has real dimension $(2n)^2=4n^2$. For $n\ge2$ this is at least $16$, and $\mathrm{SU}(3)$ of dimension $8$ embeds as a subalgebra: in the carrier $\mathbb C^{2n}$ split as $\mathbf 3\oplus(\text{rest})$, the three-dimensional block carries $SU(3)$. The intrinsic gauge group of the enlarged carrier is $U(2n)$ — $U(4)$ for $n=2$ — and $SU(3)$ is a subgroup of it, with commutant $U(1)\times U(1)$: a phase on the colour triplet and a phase on the complementary block, by Schur's lemma, since the triplet is irreducible and the complement is acted on trivially. The remaining generators of $\mathrm{U}(4)$ beyond $\mathrm{SU}(3)\oplus\mathrm{U}(1)\oplus\mathrm{U}(1)$ number $16-10=6$, the dimension of the coset $U(4)/(SU(3)\times U(1)\times U(1))$; they carry the $\mathbf 3\oplus\bar{\mathbf 3}$ and no colour quantum number. The cost is twofold: the gauge group is no longer the algebra's own but a subgroup of the enlarged unitaries, so the embedding is not canonical; and the extra $U(1)$ factors and broken generators are unwanted and must be lifted or projected away, which requires a mechanism outside the algebra. The embedding is not merely abstract: the elements of the embedded $\mathrm{SU}(3)$ have explicit closed forms in biquaternion parameters — an ordered product of three exponentials of the antisymmetric, diagonal and diagonal-less symmetric linear functions, with eight parameters — so the gauge potential $\mathcal{A}_\mu$ and the finite gauge transformations can be written out on the enlarged carrier. That is the concrete form of the price: the eight biquaternion parameters of the closed form are the colour data, adjoined to the algebra and not drawn from it (see The Gluon: An Octet Outside the Biquaternion Algebra, §The Enlarged Carrier and the Price of Embedding).

Enlarging the algebra. Replace $\mathbb H$ by another real algebra. The Clifford algebra $\mathrm{Cl}_{1,3}\cong M_2(\mathbb H)$ (as the corpus's conventions article records) has complexification $\mathrm{Cl}_{1,3}\otimes_{\mathbb R}\mathbb C\cong M_2(\mathbb B)\cong M_4(\mathbb C)$; it already contains $SU(2)$'s in its spin structure but is not a home for $SU(3)$ either. The octonions $\mathbb O$ carry $SU(3)\subset G_2$, and the exceptional Jordan algebra and the Freudenthal–Tits magic square give routes to $SU(3)$ and to larger groups; all of them abandon associativity or the four-dimensional complex structure that the framework's biquaternion norm relies on. The biquaternion framework is not one of these, and passing to them would be building a different framework.

One enlargement that keeps associativity. Among the enlargements there is one that does not abandon associativity, and it is worth separating from the others. The left multiplications of the complex octonions form an associative algebra, the complex octonionic chain algebra $$ \overleftarrow{\mathbb{C}\otimes\mathbb{O}}\cong\mathrm{Cl}(6)\cong M_8(\mathbb{C}), $$ of complex dimension $64$. It is a Clifford algebra, hence associative, and its commutator is a genuine Lie bracket; it carries an intrinsic $SU(3)$ as the stabiliser of a maximal totally isotropic subspace, equivalently of a unit $e_7$ inside $G_2=\mathrm{Aut}(\mathbb O)$. The colour group here is the algebra's own and not an import, so within this carrier the ceiling does not apply — the carrier is not $\mathbb B$ and the modules are not $\mathbb B$-modules, and the three-dimensional colour triplet has a home as a minimal left ideal. What it costs is the change of carrier, which is a real cost: $SU(3)$ is the algebra's own only after the algebra has been replaced by one of complex dimension $64$. The construction is set out in the companion article Complex Octonions and the Clifford Algebra Cl(6). The structural statement below therefore stands as a statement about $\mathbb B$ — the group ceases to be the algebra's own as soon as the carrier changes — with the correction that associativity, at least, need not be the casualty.

The structural statement. Within the associative four-dimensional complex algebra, the gauge group is $U(2)$ and no more. $SU(3)$ is reachable only by enlarging the carrier, and every enlargement pays in canonicity: the group ceases to be the algebra's own. This is the sense in which the ceiling is a sharp result about the algebra rather than a limitation of effort. It also connects directly to the last article of this group, Grand Unification and the Biquaternion Algebra Ceiling, where the same count is used to determine how far a unification based on the algebra can go.

The Ceiling and the Standard Model

It is worth stating the consequence in the language of the Standard Model, because it is the sharpest way to express what the framework's particle content can be.

The Standard Model gauge group is

$$ G_{\mathrm{SM}}=SU(3)_c\times SU(2)_L\times U(1)_Y , $$

of real dimension $8+3+1=12$. The framework reaches the second and third factors intrinsically: $SU(2)_L$ is the compact factor generated by the imaginary quaternions, and $U(1)_Y$ is the central phase. It does not reach the first. A field content in which the quarks carry colour is therefore not obtainable from fields valued in $\mathbb B$ alone; the coloured sector requires an enlarged carrier, and the enlargement must supply a three-dimensional complex colour space, which is exactly what $\mathbb B$-modules cannot do.

Two consequences follow, and both are genuine results for the framework's particle programme. First, any biquaternionic account of strong interactions must be an account of the gauge-invariant composites — the hadrons and, more precisely, the colour-singlet operators — because the coloured fields themselves do not exist as algebra-valued fields; the singleton colour sector is not expressible, and only the singlet sector is. This is consistent with the informational programme of the corpus, where confinement is read as the absence of a colour-singlet decomposition. Second, the electroweak theory and the abelian sector are the framework's own, because they are exactly the $U(2)$ that the algebra carries; this is why the companion articles of Biquaternion Particle Physics and Gauge Fields treat the electroweak and abelian gauge fields directly and the colour sector only through its composites. The ceiling is thus the structural reason behind the shape of the framework's existing gauge articles.

Summary

The biquaternion algebra's intrinsic compact gauge structure is the anti-Hermitian sector

$$ \mathbb M_-=\mathrm{span}_{\mathbb R}\{ie_0,e_1,e_2,e_3\}=\mathrm{U}(2)=\mathrm{SU}(2)\oplus\mathrm{U}(1), \qquad \dim_{\mathbb R}=4 , $$

with the non-abelian part generated by the imaginary quaternions, $[e_a,e_b]=2\varepsilon_{abc}e_c$, and the abelian part by the central $ie_0$. The unit group of the defining module, the compact part of the automorphism group, the diagonal norm-preserving rotors and the maximal torus all give the same group $U(2)$, and the derived algebra is $\mathrm{SL}(2,\mathbb C)$ with compact form $\mathrm{SU}(2)$.

$SU(3)$ is not reached, by three independent arguments. Its compact algebra has dimension $8$, larger than the $4$ of $\mathrm{U}(2)$, so it does not fit; the algebra's modules have even complex dimension, so the odd-dimensional colour triplet is a subspace of no $\mathbb B$-module at all, since submodules of a semisimple module are again even-dimensional; and the smallest normed division algebra whose automorphism group contains $SU(3)$ is the non-associative octonions, outside the associative framework. The equality $\dim_{\mathbb R}\mathbb B=8=\dim_{\mathbb R}\mathrm{SU}(3)$ is a numerical coincidence, since the eight directions of $\mathbb B$ include six non-compact ones.

The ceiling is passed by enlarging the carrier to $M_n(\mathbb B)\cong M_{2n}(\mathbb C)$, $n\ge2$, whose intrinsic group $U(2n)$ contains $SU(3)$; the cost is that the group is then no longer the algebra's own and that the embedding leaves a commuting $U(1)\times U(1)$ together with broken generators that must be removed from outside the algebra. In Standard-Model terms the framework supplies $SU(2)_L$ and $U(1)_Y$ intrinsically and not $SU(3)_c$, so only the colour-singlet sector is expressible in algebra-valued fields. This is a result about the algebra, and it is the structural reason for the shape of the framework's gauge and particle programme.

Summary of Notation

Further Reading

• Richard D. Schafer, An Introduction to Nonassociative Algebras (Academic Press, 1966), for the classification of real division algebras and the octonion automorphism group.
• John C. Baez, "The octonions", Bulletin of the American Mathematical Society 39 (2002) 145–205, for the route from the normed division algebras to $SU(3)$, $G_2$ and the exceptional groups.
• F. Reese Harvey, Spinors and Calibrations (Academic Press, 1990), for the division algebras $\mathbb R,\mathbb C,\mathbb H,\mathbb O$ and their automorphism and unitary groups.
• Israel M. Gelfand, Raoul A. Minlos and Z. Ya. Shapiro, Representations of the Rotation and Lorentz Groups (Pergamon, 1963), for the representation theory of $SU(2)$ and the module structure used here.
• Theodore Frankel, The Geometry of Physics (Cambridge University Press, 3rd ed. 2011), for the geometry of $U(2)$, $SU(2)$ and the compact forms of the classical groups.
• Howard Georgi, Lie Algebras in Particle Physics (Westview, 2nd ed. 1999), for the dimension and rank counting of $\mathrm{SU}(n)$ and the Standard-Model gauge group.
• Leonard Susskind and John Kogut, "Preface to the strong interactions", and Kenneth G. Wilson, "Confinement of quarks", Physical Review D 10 (1974) 2445–2459, for the singlet-sector structure of the colour sector.
• Mikio Nakahara, Geometry, Topology and Physics (CRC Press, 2nd ed. 2003), for the quaternion algebra, its automorphisms and its unitary groups.
• R. Slansky, "Group theory for unified model building", Physics Reports 79 (1981) 1–128, for the representation theory of the unified groups and the dimension counting of the Standard-Model factors.
• Michael Atiyah, The Geometry and Physics of Knots (Cambridge University Press, 1990), for the geometry of the classical groups and the quaternionic structures that constrain them.
• A. Gsponer, "Explicit closed-form parametrization of SU(3) and SU(4) in terms of complex quaternions and elementary functions," arXiv:math-ph/0211056v2, 2002, for the explicit closed forms of the elements of the embedded $\mathrm{SU}(3)$ used in What Would Be Needed to Pass the Ceiling.