Baum-Connes Conjecture

2 The reduced group \(C^*\)-algebra

Throughout this section, \(\Gamma \) is an arbitrary discrete group with identity \(e\). Countability is needed only for the separability statement. Scalars are complex; inner products are conjugate-linear in the first variable and linear in the second. All twenty-one mathematical nodes in this chapter are verified in Lean. This includes the concrete reduced group \(C^*\)-algebra and its canonical isometric \(*\)-isomorphism with the abstract reduced-norm completion of \(\mathbb C[\Gamma ]\), together with the generating property, countable-group separability, and the trivial-group case. These foundational results do not prove any case of the Baum–Connes conjecture.

2.1 The complex group algebra

The complex vector space \(\mathbb C[\Gamma ]\) consists of finitely supported functions \(a:\Gamma \to \mathbb C\), with pointwise vector operations. Write \(\operatorname {supp}(a)=\{ g:a(g)\ne 0\} \) and let \(\delta _g\) be \(1\) at \(g\) and \(0\) elsewhere. Define

\[ (a*b)(h)=\sum _g a(g)b(g^{-1}h),\qquad a^*(h)=\overline{a(h^{-1})},\qquad \| a\| _1=\sum _g|a(g)|. \]

These sums are finite. The proposed multiplicative identity is \(\delta _e\).

Convolution and involution preserve finite support, with

\[ \operatorname {supp}(a*b)\subseteq \operatorname {supp}(a)\operatorname {supp}(b),\qquad \operatorname {supp}(a^*)=\operatorname {supp}(a)^{-1}. \]

Moreover,

\[ a=\sum _g a(g)\delta _g,\qquad \delta _g*\delta _k=\delta _{gk},\qquad \delta _g^*=\delta _{g^{-1}}. \]
Proof ▶

If \((a*b)(h)\ne 0\), some summand is nonzero, so \(g\) belongs to \(\operatorname {supp}(a)\) and \(g^{-1}h\) belongs to \(\operatorname {supp}(b)\). Thus \(h\) lies in the indicated finite product set. The other assertions follow by evaluating at each group element.

The operations make \(\mathbb C[\Gamma ]\) a unital complex \(*\)-algebra: convolution is associative and bilinear, its identity is \(\delta _e\), and

\[ (a*b)^*=b^**a^*,\qquad (a^*)^*=a,\qquad (za+wb)^*=\overline z\, a^*+\overline w\, b^*. \]
Proof ▶

Bilinearity and conjugate-linearity follow from the defining formulas. Using the finite delta expansion, associativity reduces to \(\delta _{(gk)l}=\delta _{g(kl)}\), the identity laws reduce to \(eg=g=ge\), and reversal under involution reduces to \((gk)^{-1}=k^{-1}g^{-1}\). Inversion twice and complex conjugation twice give \((a^*)^*=a\).

2.2 The Hilbert space and left translations

Definition 4 Square-summable functions
✓

For a nonnegative family, define its sum as the supremum of its finite partial sums. Put

\[ \ell ^2(\Gamma )=\left\{ \xi :\Gamma \to \mathbb C: \sum _h|\xi (h)|^2{\lt}\infty \right\} ,\qquad \| \xi \| _2^2=\sum _h|\xi (h)|^2, \]

with pointwise vector operations and \(\langle \xi ,\eta \rangle =\sum _h\overline{\xi (h)}\eta (h)\). The notation \(\delta _g\) also denotes the corresponding vector in \(\ell ^2(\Gamma )\); the ambient space determines its meaning.

The inner product is well defined, and \(\ell ^2(\Gamma )\) is a complex Hilbert space. The delta vectors are orthonormal and have dense finite linear span. For every \(h\), \(|\xi (h)|\leq \| \xi \| _2\) and \(\langle \delta _h,\xi \rangle =\xi (h)\).

Proof ▶

Finite-sum Cauchy–Schwarz, followed by a supremum over finite sets, gives \(\sum _h|\overline{\xi (h)}\eta (h)|\leq \| \xi \| _2\| \eta \| _2\). Thus the inner-product sum is absolutely convergent. The inequality \(|z+w|^2\leq 2|z|^2+2|w|^2\) ensures closure under addition; the inner-product identities follow coordinatewise.

For completeness, a Cauchy sequence \((\xi _n)\) is Cauchy in each coordinate. Define \(\xi (h)=\lim _n\xi _n(h)\). If \(\| \xi _n-\xi _m\| _2\leq \varepsilon \) for \(n,m\geq N\), taking \(m\to \infty \) on any finite set \(F\) gives \(\sum _{h\in F}|\xi _n(h)-\xi (h)|^2\leq \varepsilon ^2\). Taking the supremum proves \(\xi -\xi _n\in \ell ^2(\Gamma )\) and \(\| \xi -\xi _n\| _2\leq \varepsilon \), hence \(\xi \in \ell ^2(\Gamma )\) and norm convergence.

The delta formulas are immediate from the inner product. For \(\varepsilon {\gt}0\), choose a finite \(F\) whose complement has \(\sum _{h\notin F}|\xi (h)|^2{\lt}\varepsilon ^2\). Then \(\sum _{h\in F}\xi (h)\delta _h\) approximates \(\xi \) within \(\varepsilon \). This argument does not require an enumeration of \(\Gamma \).

Definition 6 Left translation
✓

For \(g\in \Gamma \), prescribe

\[ (\lambda _g\xi )(h)=\xi (g^{-1}h). \]

The next lemma verifies that this defines an operator on \(\ell ^2(\Gamma )\).

Each \(\lambda _g\) is a complex-linear isometry, and

\[ \lambda _e=I,\qquad \lambda _g\lambda _k=\lambda _{gk},\qquad \lambda _g\delta _h=\delta _{gh},\qquad \| \lambda _g\| =1, \]

where the final norm is the operator norm.

Proof ▶

The bijection \(h\mapsto g^{-1}h\) preserves finite subsets and their supremum, so \(\sum _h|\xi (g^{-1}h)|^2=\sum _h|\xi (h)|^2\). This proves well-definedness and norm preservation; linearity is coordinatewise. Also \((\lambda _g\lambda _k\xi )(h)=\xi (k^{-1}g^{-1}h)=\xi ((gk)^{-1}h)\). The identity and delta formulas follow by evaluation. Norm preservation gives \(\| \lambda _g\| \leq 1\), and the unit vector \(\delta _e\) gives the reverse inequality.

Each \(\lambda _g\) is unitary, with \(\lambda _g^{-1}=\lambda _{g^{-1}}=\lambda _g^*\).

Proof ▶

The multiplication identities give the two-sided inverse \(\lambda _{g^{-1}}\). Reindexing the absolutely convergent inner product gives \(\langle \lambda _g\xi ,\eta \rangle =\langle \xi ,\lambda _{g^{-1}}\eta \rangle \), which identifies the adjoint.

Definition 9 Left regular representation
✓

The left regular representation is the group homomorphism \(\lambda :\Gamma \to U(\ell ^2(\Gamma ))\), \(g\mapsto \lambda _g\), where \(U(H)\) denotes the unitary operators on \(H\). It is distinct from the linear extension defined below.

2.3 The integrated representation

We use the standard functional-analysis theorem that the bounded complex-linear operators \(\mathcal B(H)\) on a complex Hilbert space form a unital \(C^*\)-algebra. Multiplication is composition, involution is the adjoint, the operator norm is complete, and \(\| T^*T\| =\| T\| ^2\). This is an explicit background prerequisite.

Definition 10 Integrated regular representation
✓

Define the complex-linear map

\[ \rho :\mathbb C[\Gamma ]\longrightarrow \mathcal B(\ell ^2(\Gamma )), \qquad \rho (a)=\sum _{g\in \operatorname {supp}(a)}a(g)\lambda _g. \]

This is a finite sum of bounded operators, and \(\rho (\delta _g)=\lambda _g\).

Lemma 11 Action formula and norm bound
✓

For \(a\in \mathbb C[\Gamma ]\), \(\xi \in \ell ^2(\Gamma )\) and \(h\in \Gamma \),

\[ (\rho (a)\xi )(h)=\sum _g a(g)\xi (g^{-1}h),\qquad \| \rho (a)\| \leq \| a\| _1. \]
Proof ▶

Apply the finite operator sum to \(\xi \) and evaluate at \(h\). The operator-norm triangle inequality gives \(\| \rho (a)\| \leq \sum _g|a(g)|\| \lambda _g\| =\| a\| _1\).

The map \(\rho \) is a unital complex \(*\)-algebra homomorphism:

\[ \rho (a*b)=\rho (a)\rho (b),\qquad \rho (a^*)=\rho (a)^*,\qquad \rho (\delta _e)=I. \]
Proof ▶

Distribute finite sums and group terms with \(gk=h\):

\[ \rho (a)\rho (b)=\sum _{g,k}a(g)b(k)\lambda _{gk} =\sum _h\left(\sum _g a(g)b(g^{-1}h)\right)\lambda _h =\rho (a*b). \]

Conjugate-linearity of adjoints and \(\lambda _g^*=\lambda _{g^{-1}}\) give

\[ \rho (a)^*=\sum _g\overline{a(g)}\lambda _{g^{-1}} =\sum _h\overline{a(h^{-1})}\lambda _h=\rho (a^*). \]

Linearity is built into the definition; \(\lambda _e=I\) gives unitality.

Regard \(a\in \mathbb C[\Gamma ]\) also as a finitely supported \(\ell ^2\)-vector. Then

\[ \rho (a)\delta _e=a,\qquad \| a\| _2\leq \| \rho (a)\| ,\qquad |a(h)|\leq \| \rho (a)\| \quad (h\in \Gamma ). \]

In particular, \(\rho \) is injective.

Proof ▶

In \((\rho (a)\delta _e)(h)=\sum _g a(g)\delta _e(g^{-1}h)\), only \(g=h\) contributes, giving \(a(h)\). Since \(\| \delta _e\| _2=1\), \(\| a\| _2=\| \rho (a)\delta _e\| _2\leq \| \rho (a)\| \). The coordinate bound gives the last inequality. If \(\rho (a)=0\), all coefficients of \(a\) vanish.

2.4 The reduced norm and operator-norm closure

Definition 14 Reduced norm
✓
#

For \(a\in \mathbb C[\Gamma ]\), put \(\| a\| _r=\| \rho (a)\| \).

The reduced norm is a norm, with

\[ \| a*b\| _r\leq \| a\| _r\| b\| _r,\qquad \| a^*\| _r=\| a\| _r,\qquad \| a^**a\| _r=\| a\| _r^2, \]
\[ \| \delta _e\| _r=1,\qquad \| a\| _2\leq \| a\| _r\leq \| a\| _1. \]

Completeness of \(\mathbb C[\Gamma ]\) in this norm is not asserted.

Proof ▶

Homogeneity and the triangle inequality come from the operator norm; faithfulness gives positive definiteness. The homomorphism identities transfer submultiplicativity, isometry of the involution, and the \(C^*\)-identity from bounded operators. The remaining bounds were proved above.

Definition 16 Reduced group C-star algebra
✓

Define

\[ C_r^*(\Gamma )=\overline{\rho (\mathbb C[\Gamma ])}^{\| \cdot \| } \subseteq \mathcal B(\ell ^2(\Gamma )), \]

with inherited operations and norm. The closure is in the operator norm, not the weak or strong operator topology.

The inherited operations make \(C_r^*(\Gamma )\) a unital complex \(C^*\)-algebra with identity \(I=\lambda _e\).

Proof ▶

The image of \(\rho \) is a unital \(*\)-subalgebra. Its closure is closed under linear combinations and adjoints by continuity. If \(A_n\to A\) and \(B_n\to B\) in the image, \((\| A_n\| )\) is bounded and

\[ \| A_nB_n-AB\| \leq \| A_n\| \| B_n-B\| +\| A_n-A\| \| B\| \longrightarrow 0. \]

This proves closure under multiplication; \(I\) already lies in the image. The closure is complete as a closed subspace of the Banach space of bounded operators. The \(C^*\)-identity is inherited from that algebra.

Proof ▶

Isometry follows from the reduced norm and density from the definition of the closure. Completeness gives a unique isometric linear extension to the completion. Its range is closed by completeness and dense because it contains \(\rho (\mathbb C[\Gamma ])\), hence is the whole target. Submultiplicativity and isometry of involution extend the algebra operations to the completion. Passing to limits shows that the extension preserves multiplication, involution, and the unit.

The algebra \(C_r^*(\Gamma )\) is the norm closure of \(\operatorname {span}_{\mathbb C}\{ \lambda _g:g\in \Gamma \} \). It is the smallest norm-closed \(*\)-subalgebra of \(\mathcal B(\ell ^2(\Gamma ))\) containing every \(\lambda _g\).

Proof ▶

The image of \(\rho \) is exactly the finite linear span of the generators. Any closed \(*\)-subalgebra containing them contains this span and its closure. Conversely, \(C_r^*(\Gamma )\) is such a subalgebra.

2.5 Countability and a first calibration

Proposition 20 Separability
✓
#

If \(\Gamma \) is countable, then \(C_r^*(\Gamma )\) is separable.

Proof ▶

The finite sums \(\sum _g q_g\lambda _g\) with \(q_g\in \mathbb Q+i\mathbb Q\) form a countable set. Approximate coefficients on any fixed finite support by such rationals. The bound \(\| \rho (a)-\rho (b)\| \leq \| a-b\| _1\) proves density in \(\rho (\mathbb C[\Gamma ])\), hence in its closure.

For the one-element group, \(z\mapsto zI\) is an isometric unital \(*\)-isomorphism \(\mathbb C\to C_r^*(\{ e\} )\).

Proof ▶

Evaluation at \(e\) identifies \(\ell ^2(\{ e\} )\) with \(\mathbb C\). Then \(\rho (z\delta _e)=zI\) and \(\| zI\| =|z|\). The image is complete and hence closed, so taking its closure adds nothing.

2.6 Interface for subsequent parts

Subsequent sections may use the unital complex \(C^*\)-algebra structure, the dense isometric group-algebra embedding, and separability when \(\Gamma \) is countable. This construction alone proves no assertion about the Baum–Connes assembly map.

For a standard reference and the relation with reduced crossed products, see ISem24, C-star-algebras and dynamics, Section 9.6. The proofs here construct the discrete-group algebra directly; crossed products are not a prerequisite for this section.