You are currently browsing the category archive for the ‘C*-Algebras & Operator Theory’ category.

In my pursuit of an Ergodic Theorem for Random Walks on (probably finite) Quantum Groups, I have been looking at analogues of Irreducible and Periodic. I have, more or less, got a handle on irreducibility, but I am better at periodicity than aperiodicity.

The question of how to generalise these notions from the (finite) classical to noncommutative world has already been considered in a paper (whose title is the title of this post) of Fagnola and Pellicer. I can use their definition of periodic, and show that the definition of irreducible that I use is equivalent. This post is based largely on that paper.


Consider a random walk on a finite group G driven by \nu\in M_p(G). The state of the random walk after k steps is given by \nu^{\star k}, defined inductively (on the algebra of functions level) by the associative

\nu\star \nu=(\nu\otimes\nu)\circ \Delta.

The convolution is also implemented by right multiplication by the stochastic operator:

\nu\star \nu=\nu P,

where P\in L(F(G)) has entries, with respect to a basis (\delta_{g_i})_{i\geq 1} P_{ij}=\nu(g_jg_{i^{-1}}). Furthermore, therefore

\nu^{\star k}=\varepsilon P^k,

and so the stochastic operator P describes the random walk just as well as the driving probabilty \nu.

The random walk driven by \nu is said to be irreducible if for all g_\ell\in G, there exists k\in\mathbb{N} such that (if g_1=e) [P^k]_{1\ell}>0.

The period of the random walk is defined by:

\displaystyle \gcd\left(d\in\mathbb{N}:[P^d]_{11}>0\right).

The random walk is said to be aperiodic if the period of the random walk is one.

These statements have counterparts on the set level.

If P is not irreducible, there exists a proper subset of G, say S\subsetneq G, such that the set of functions supported on S are P-invariant.  It turns out that S is a proper subgroup of G.

Moreover, when P is irreducible, the period is the greatest common divisor of all the natural numbers d such that there exists a partition S_0, S_1, \dots, S_{d-1} of G such that the subalgebras A_k of functions supported in S_k satisfy:

P(A_k)=A_{k-1} and P(A_{0})=A_{d-1} (slight typo in the paper here).

In fact, in this case it is necessarily the case that \nu is concentrated on a coset of a proper normal subgroup N\rhd G, say gN. Then S_k=g^kN.

Suppose that f is supported on g^kNWe want to show that for Pf\in A_{k-1}Recall that 

\nu^{\star k-1}P(f)=\nu^{\star k}(f).

This shows how the stochastic operator reduces the index P(A_k)=A_{k-1}.

A central component of Fagnola and Pellicer’s paper are results about how the decomposition of a stochastic operator:


specifically the maps L_\ell can speak to the irreducibility and periodicity of the random walk given by P. I am not convinced that I need these results (even though I show how they are applicable).

Stochastic Operators and Operator Algebras

Let F(X) be a \mathrm{C}^*-algebra (so that X is in general a  virtual object). A \mathrm{C}^*-subalgebra F(Y) is hereditary if whenever f\in F(X)^+ and h\in F(Y)^+, and f\leq h, then f\in F(Y)^+.

It can be shown that if F(Y) is a hereditary subalgebra of F(X) that there exists a projection \mathbf{1}_Y\in F(X) such that:


All hereditary subalgebras are of this form.

Read the rest of this entry »

This sandbox is going to take from a variety of sources, mostly Shuzhou Wang.


Let J\subset C(X) be a closed (two-sided) ideal in a non-commutative unital C^*-algebra C(X). Such an ideal is self-adjoint and so a non-commutative C^*-algebra J=C(S). The quotient map is given by \pi:C(X)\rightarrow C(X)/C(S), f\mapsto f+J, where f+J is the equivalence class of f under the equivalence relation:

f\sim_{J} g\Rightarrow g-f\in C(S).

Where we have the product


and the norm is given by:

\displaystyle\|f+J\|=\sup_{j\in C(S)}\|f+j\|,

the quotient C(X)/ C(S) is a C^*-algebra.

Consider now elements j_1,\,j_2\in C(S) and f_1,\, f_2\in C(X). Consider

j_1\otimes f_1+f_2\otimes j_2\in C(S)\otimes C(X)+C(X)\otimes C(S).

The tensor product \pi\otimes \pi:C(X)\otimes C(X)\rightarrow (C(X)/C(S))\otimes (C(X)/ C(S)). Now note that

(\pi\otimes\pi)(j_1\otimes f_1+f_2\otimes j_2)=(0+J)\otimes(f_1+J)+


by the nature of the Tensor Product (0\otimes a=0). Therefore C(X)\otimes C(S)+C(S)\otimes C(X)\subset \text{ker}(\pi\otimes\pi).


A WC*-ideal (W for Woronowicz) is a C*-ideal J=C(S) such that \Delta(J)\subset \text{ker}(\pi\otimes\pi), where \pi is the quotient map C(G)\rightarrow C(G)/C(S).

Let F(G) be the algebra of functions on a classical group G. Let H\subset G. Let J be the set of functions which vanish on H: this is a C*-ideal. The kernal of \pi:F(G)\rightarrow F(G)/J is J.

Let \delta_s\in J so that s\not\in H. Note that

\displaystyle\Delta(\delta_s)=\sum_{t\in G}\delta_{st^{-1}}\otimes\delta_t

and so

\displaystyle(\pi\otimes \pi)\Delta(\delta_s)=\sum_{t\in G}\pi(\delta_{st^{-1}})\otimes \pi(\delta_t).

Note that \pi(\delta_t)=0+J if t\not\in H. It is not possible that both st^{-1} and t are in H: if they were st^{-1}\cdot t\in H, but st^{-1}\cdot t=s, which is not in H by assumption. Therefore one of \pi(\delta_{st^{-1}}) or \pi(\delta_t) is equal to zero and so:


and so by linearity, if f vanishes on a subgroup H,

\Delta(f)\subset \text{ker}(\pi\otimes\pi).

In this way, WC*-ideals generalise functions which vanish on distinguished subgroups. In fact, without checking all the details, I imagine that first isomorphism theorem can show that F(G)/ J=F(H). Let \pi_H:F(G)\rightarrow F(H) be the ring homomorphism

\displaystyle\pi_H\left(\sum_{t\in G}a_t\delta_t\right)=\sum_{t\in H}a_t\delta_t.

Then \text{ker}\,\pi_H=J, \text{im}\,\pi_H=F(H), and so we have the isomorphism of rings, which presumably carries forward to the algebras of functions level…

Read the rest of this entry »

Just some notes on the pre-print. I am looking at this paper to better understand this pre-print. In particular I am hoping to learn more about the support of a probability on a quantum group. Flags and notes are added but mistakes are mine alone.


From this paper I will look at:

  • lattice operations on \mathcal{I}(G), for G a LCQG (analogues of intersection and generation)

1. Introduction

Idempotent states on quantum groups correspond with “subgroup-like” objects. In this work, on LCQG, the correspondence is with quasi-subgroups (the work of Franz & Skalski the correspondence was with pre-subgroups and group-like projections).

Let us show the kind of thing I am trying to understand better.

Let F(G) be the algebra of function on a finite quantum group. Let \nu,\,\mu\in M_p(G) be concentrated on a pre-subgroup S. We can associate to S a group like projection p_S.

Let, and this is another thing I am trying to understand better, this support, the support of \nu be ‘the smallest’ (?) projection p\in F(G) such that \nu(p)=1. Denote this projection by p_\nu. Define p_\mu similarly. That \mu,\,\nu are concentrated on S is to say that p_\nu\leq p_S and p_\mu\leq p_S.

Define a map T_\nu:F(G)\rightarrow F(G) by 

a\mapsto p_\nu a (or should this be ap_\nu or p_\nu a p_\nu?)

We can decompose, in the finite case, F(G)\cong \text{Im}(T_\nu)\oplus \ker(T_\nu)

Claim: If \nu is concentrated on S\nu(ap_S)=\nu(a)I don’t have a proof but it should fall out of something like p_\nu\leq p_S\Rightarrow \ker p_\nu\subseteq \ker p_S together with the decomposition of F(G) above. It may also require that \int_G is a trace, I don’t know. Something very similar in the preprint.

From here we can do the following. That p_S is a group-like projection means that:

\Delta (p_s)(\mathbf{1}_G\otimes p_S)=p_S\otimes p_S

\Rightarrow \sum p_{S(1)}\otimes (p_{S(2)}p_S)=p_S\otimes p_S

Hit both sides with \nu\times \mu to get:

\sum \nu(p_{S(1)})\mu(p_{S(2)}p_S)=\nu(p_S)\mu(p_S).

By the fact that \nu,\,\mu are supported on S, the right-hand side equals one, and by the as-yet-unproven claim, we have

\sum \nu(p_{S(1)})\mu(p_{S(2)})=1.

However this is the same as

(\nu\otimes\mu)\Delta(p_S)=1\Rightarrow (\nu\star \mu)(p_S)=1,

in other words p_{\nu\star \mu}\leq p_S, that is \nu\star \mu remains supported on S. As a corollary, a random walk driven by a probability concentrated on a pre-subgroup S\subset G remains concentrated on S.

Read the rest of this entry »

In May 2017 I wrote down some problems that I hoped to look at in my study of random walks on quantum groups. Following work of Amaury Freslon, a number of these questions have been answered. In exchange for solving these problems, Amaury has very kindly suggested some other problems that I can work on. The below hopes to categorise some of these problems and their status.


  • Show that the total variation distance is equal to the projection distance. Amaury has an a third proof. Amaury suggests that this should be true in more generality than the case of \nu being absolutely continuous (of the form \nu(x)=\int_G xa_{\nu} for all x\in C(G) and a unique a_{\nu}\in C(G)). If the Haar state is no longer tracial Amaury’s proof breaks down (and I imagine so do the two others in the link above). Amaury believes this is true in more generality and says perhaps the Jordan decomposition of states will be useful here.
  • Prove the Upper Bound Lemma for compact quantum groups of Kac type. Achieved by Amaury.
  • Attack random walks with conjugate invariant driving probabilitys: achieved by Amaury.
  • Look at quantum generalisations of ‘natural’ random walks and shuffles. Solved is probably too strong a word, but Amaury has started this study by looking at a generalisation of the random transposition shuffle. As I suggested in Seoul, Amaury says: “One important problem in my opinion is to say something about analogues of classical random walks on S_n (for instance the random transpositions or riffle shuffle)”. Amaury notes that “we are blocked by the counit problem. We must therefore seek bounds for other distances. As I suggest in my paper, we may look at the norm of the difference of the transition operators. The \mathcal{L}^2-estimate that I give is somehow the simplest thing one can do and should be thought of as a “spectral gap” estimate. Better norms would be the norms as operators on \mathcal{L}^\infty or even better, the completely bounded norm. However, I have not the least idea of how to estimate this.”

Results to be Improved

  • I have recently received an email from Isabelle Baraquin, a student of Uwe Franz, pointing out a small error in the thesis (a basis-error with the Kac-Paljutkin quantum groups).
  • Recent calculations suggest that the lower bound for the random walk on the dual of S_n is effective at k\sim (n-1)! while the upper bound shows the walk is random at time order n!.  This is still a very large gap but at least the lower bound shows that this walk does converge very slowly.
  • Get a much sharper lower bound for the random walk on the Sekine family of quantum groups studied in Section 5.7. Projection onto the ‘middle’ of the M_n(\mathbb{C}) factor may provide something of use. On mature reflection, recognising that the application of the upper bound lemma is dominated by one set of terms in particular, it should be possible to use cruder but more elegant estimates to get the same upper bound except with lighter calculations (and also a smaller \alpha — see Section 5.7).

More Questions on Random Walks

  • Irreducibility is harder than the classical case (where ‘not concentrated’ on a subgroup is enough). Can anything be said about aperiodicity in the quantum case? (U. Franz).
  • Prove an Ergodic Theorem (Theorem 1.3.2) for Finite Quantum Groups. Extend to Compact Quantum Groups. It is expected that the conditions may be more difficult than the classical case. However, it may be possible to use Diaconis-Van Daele theory to get some results in this direction. It should be possible to completely analyse some examples (such as the Kac-Paljutkin quantum group of order 8).This will involve a study of subgroups of quantum groups as well as normal quantum subgroups and cosets.
  • Look at a random walk on the Sekine quantum groups with an n-dependent driving probability and see if the cut-off phenomenon (Chapter 4) can be detected. This will need good lower bounds for k\ll t_n, some cut-off time.
  • Convolutions Factorisations of the Random Distribution: such a study may prove fruitful in trying to find the Ergodic Theorem. See Section 6.5.
  • Amaury mentions the problem of considering random walks associated to non-central states (in the compact case). “The difficulty is first to build non-central states (I do not have explicit examples at hand but Uwe Franz said he had some) and second to be able to compute their Fourier transform. Then, the computations will certainly be hard but may still be doable.”
  • A study of the Cesaro means: see Section 6.6.
  • Spectral Analysis: it should be possible to derive crude bounds using the spectrum of the stochastic operator. More in Section 6.2.

Future Work (for which I do not yet have the tools to attack)

  • Amaury/Franz Something perhaps more accessible is to investigate quantum homogeneous spaces. The free sphere is a noncommutative analogue of the usual sphere and a quantum homogeneous space for the free orthogonal quantum group. We can therefore define random walks on it and the whole machinery of Gelfand pairs might be available. In particular, Caspers gave a Plancherel theorem for Gelfand pairs of locally compact quantum groups which should apply here yielding an Upper Bound Lemma and then the problem boils down to something which should be close to my computations. There are probably works around this involving Adam Skalski and coauthors.
  • Amaury: If one can prove a more general total variation distance equal to half one norm result, then Amaury suggests one can consider random walks on compact quantum groups which are not of Kac type. The Upper Bound Lemma will then involve matrices Q measuring the modular theory of the Haar state and some (but not all) dimensions in the formulas must be replaced by quantum dimensions. The main problem here is to define explicit central states since there is no Haar-state preserving conditional expectation onto the central algebra. However, there are tools from monoidal equivalence to do this.


Distances between Probability Measures

Let G be a finite quantum group and M_p(G) be the set of states on the \mathrm{C}^\ast-algebra F(G).

The algebra F(G) has an invariant state \int_G\in\mathbb{C}G=F(G)^\ast, the dual space of F(G).

Define a (bijective) map \mathcal{F}:F(G)\rightarrow \mathbb{C}G, by

\displaystyle \mathcal{F}(a)b=\int_G ba,

for a,b\in F(G).

Then, where \|\cdot\|_1^{F(G)}=\int_G|\cdot| and \|\cdot\|_\infty^{F(G)}=\|\cdot\|_{\text{op}}, define the total variation distance between states \nu,\mu\in M_p(G) by

\displaystyle \|\nu-\mu\|=\frac12 \|\mathcal{F}^{-1}(\nu-\mu)\|_1^{F(G)}.

(Quantum Total Variation Distance (QTVD))

Standard non-commutative \mathcal{L}^p machinary shows that:

\displaystyle \|\nu-\mu\|=\sup_{\phi\in F(G):\|\phi\|_\infty^{F(G)}\leq 1}\frac12|\nu(\phi)-\mu(\phi)|.

(supremum presentation)

In the classical case, using the test function \phi=2\mathbf{1}_S-\mathbf{1}_G, where S=\{\nu\geq \mu\}, we have the probabilists’ preferred definition of total variation distance:

\displaystyle \|\nu-\mu\|_{\text{TV}}=\sup_{S\subset G}|\nu(\mathbf{1}_S)-\mu(\mathbf{1}_S)|=\sup_{S\subset G}|\nu(S)-\mu(S)|.

In the classical case the set of indicator functions on the subsets of the group exhaust the set of projections in F(G), and therefore the classical total variation distance is equal to:

\displaystyle \|\nu-\mu\|_P=\sup_{p\text{ a projection}}|\nu(p)-\mu(p)|.

(Projection Distance)

In all cases the quantum total variation distance and the supremum presentation are equal. In the classical case they are equal also to the projection distance. Therefore, in the classical case, we are free to define the total variation distance by the projection distance.

Quantum Projection Distance \neq Quantum Variation Distance?

Perhaps, however, on truly quantum finite groups the projection distance could differ from the QTVD. In particular, a pair of states on a M_n(\mathbb{C}) factor of F(G) might be different in QTVD vs in projection distance (this cannot occur in the classical case as all the factors are one dimensional).

Read the rest of this entry »

Just back from a great workshop at Seoul National University, I am just going to use this piece to outline in a relaxed manner my key goals for my work on random walks on quantum groups for the near future.

In the very short term I want to try and get a much sharper lower bound for my random walk on the Sekine family of quantum groups. I believe the projection onto the ‘middle’ of the M_n(\mathbb{C}) might provide something of use. On mature reflection, recognising that the application of the upper bound lemma is dominated by one set of terms in particular, it should be possible to use cruder but more elegant estimates to get the same upper bound except with lighter calculations (and also a smaller \alpha — see Section 5.7).

I also want to understand how sharp (or otherwise) the order n^n convergence for the random walk on the dual of S_n is — n^n sounds awfully high. Furthermore it should be possible to get a better lower bound that what I have.

It should also be possible to redefine the quantum total variation distance as a supremum over projections \sim subsets via G \supset S\leftrightarrow \mathbf{1}_S. If I can show that for a positive linear functional \rho that |\rho(a)|\leq \rho(|a|) then using these ideas I can. More on this soon hopefully. No, this approach won’t work. (I have since completed this objective with some help: see here).

The next thing I might like to do is look at a random walk on the Sekine quantum groups with an n-dependent driving probability and see if I can detect the cut-off phenomenon (Chapter 4). This will need good lower bounds for k\ll t_n, some cut-off time.

Going back to the start, the classical problem began around 1904 with the question of Markov:

Which card shuffles mix up a deck of cards and cause it to ‘go random’?

For example, the perfect riffle shuffle does not mix up the cards at all while a riffle shuffle done by an amateur will.

In the context of random walks on classical groups this question is answered by the Ergodic Theorem 1.3.2: when the driving probability is not concentrated on a subgroup (irreducibility) nor the coset of a normal subgroup (aperiodicity).

Necessary and sufficient conditions on the driving probability \nu\in M_p(\mathbb{G}) for the random walk on a quantum group to converge to random are required. It is expected that the conditions may be more difficult than the classical case. However, it may be possible to use Diaconis-Van Daele theory to get some results in this direction. It should be possible to completely analyse some examples (such as the Kac-Paljutkin quantum group of order 8).

This will involve a study of subgroups of quantum groups as well as normal quantum subgroups.

It should be straightforward to extend the Upper Bound Lemma (Lemma 5.3.8) to the case of compact Kac algebras. Once that is done I will want to look at quantum generalisations of ‘natural’ random walks and shuffles.

I intend also to put the PhD thesis on the Arxiv. After this I have a number of options as regard to publishing what I have or maybe waiting a little while until I solve the above problems — this will all depend on how my further study progresses.


Taken from C*-Algebras and Operator Theory by Gerald J. Murphy.

Although the principal aim of this section is to construct direct limits of C*-algebras, we begin with direct limits of groups.

If \{G_n\}_{n=1}^\infty is a sequence of groups, and if for each n we have a homomorphism \varphi_n:G_n\rightarrow G_{n+1}, then we call \{G_n\}_{n\geq1} a direct sequence of groups. Given such a sequence and positive integers n\leq m, we set \varphi_{nn}=I_{G_n} and we define \varphi_{nm}:G_n\rightarrow G_m inductively on m by setting


If n\leq m\leq k, we have \varphi_{nk}=\varphi_{mk}\varphi_{nm}.

If G' is a group and we have homomorphisms \theta^n:G^n\rightarrow G' such that the diagram

commutes for each n, that is \theta^n=\theta^{n+1}\varphi_n, then \theta^n=\theta^m\varphi_{nm} for all m\geq n.

Read the rest of this entry »

Taken from C*-Algebras and Operator Theory by Gerald Murphy.

This section is concerned with positive linear functionals and representations. Pure states are introduced and shown to be the extreme points of a certain convex set, and their existence is deduced from the Krein-Milman theorem. From this the existence of  irreducible representations is proved by establishing a correspondence between them and the pure states.

If (H,\varphi) is a representation of a C*-algebra A, we say x\in H is a cyclic vector for (H,\varphi) if x is cyclic for the C*-algebra \varphi(A) (This means that cyclic vector is a vector x\in H such that the closure of the linear span of \{\varphi(a)x\,:\,a\in A\} equals H). If (H,\varphi) admits a cyclic vector, then we say that it is a cyclic representation.

We now return to the GNS construction associated to a state to show that the representations involved are cyclic.

Theorem 5.1.1

Let A be a C*-algebra and \rho\in S(A). Then there is a unique vector x_\rho\in H_\rho\in H such that

\rho(a)=\langle a+N_\rho,x_\rho\rangle, for a\in A.

Moreover, x_\rho is a unit cyclic vector for (H_\rho,\varphi_\rho) and

\varphi_\rho(a)x_\rho=a+N_\rho, for a\in A.

Read the rest of this entry »

Taken from C*-algebras and Operator Theory by Gerald Murphy.

If H and K are vector spaces, we denote by H\otimes K their algebraic tensor product. This is linearly spanned by the elements x\otimes y (x\in Hy\in K).

One reason why tensor products are useful is that they turn bilinear maps (a bilinear map \varphi has \lambda\varphi(x,y)=\varphi(\lambda x,y)=\varphi(x,\lambda y)) into linear maps (\lambda\varphi(x,y)=\varphi(\lambda x,\lambda y)). More precisely, if \varphi:H\times K\rightarrow L is a bilinear map, where H,\,K and L are vector spaces, then there is a unique linear map \varphi_1:H\otimes K\rightarrow L such that \varphi_1(x\otimes y)=\varphi(x,y) for all x\in H and y\in K.

If \rho,\,\tau are linear functionals on the vector spaces H,\,K respectively, then there is a unique linear functional \rho\otimes\tau on H\otimes K such that

(\rho\otimes\tau)(x\otimes y)=\rho(x)\tau(y)

since the function

H\times K\rightarrow\mathbb{C}(x,y)\mapsto \rho(x)\tau(y),

is bilinear.

Suppose that the finite sum \sum_jx_j\otimes y_j=0, where x_j\in H and y_j\in K. If y_1,\dots,y_n are linearly independent, then x_1=\cdot=x_n=0. For, in this case, there exist linear functionals \rho_j:K\rightarrow \mathbb{C} such that \rho_j(y_i)=\delta_{ij}. If \rho:H\rightarrow\mathbb{C} is linear, we have

0=(\rho\otimes \rho_j)(\sum_i x_j\otimes y_j)=\sum_i\rho(x_i)\rho_j(y_i)=\rho(x_j).

Thus \rho(x_j)=0 for arbitrary \rho and this shows that all the x_j=0.

Similarly if the finite sum \sum_jx_j\otimes y_j=0 with the x_j linearly independent, implies that all the y_j are zero.

Read the rest of this entry »

Taken from C*-algebras and Operator Theory by Gerald Murphy.

We prepare the way for the density theorem with some useful results on strong convergence.

Theorem 4.3.1

If H is a Hilbert space, the involution T\mapsto T^* is strongly continuous when restricted to the set of normal operators of B(H).


Let x\in H and suppose that T,S are normal operators in B(H). Then

\|(S^*-T^*)(x)\|^2=\langle S^*x-T^*x,S^*x-T^*x\rangle

=\|Sx\|^2-\|Tx\|^2+\langle TT^*x,x\rangle-\langle ST^*x,x\rangle

+\langle TT^*x,x\rangle-\langle TS^*x,x\rangle

=\|Sx\|^2-\|Tx\|^2+\langle (T-S)T^*x,x\rangle+\langle x,(T-S)T^*x\rangle

\leq \|Sx\|^2-\|Tx\|^2+2\|(T-S)T^*x\|\|x\|.

If \{T_\lambda\}_{\lambda\in\Lambda} is a net of normal operators strongly convergent to a normal operator T, then the net \|T_\lambda x\|^2 is convergent to \|Tx\|^2 and the net \{(T-T_\lambda)T^*x\} is convergent to 0, so \{T_\lambda^*x-T^*x\} is convergent to 0. Therefore, \{T_\lambda^*\} is strongly convergent to T^* \bullet

Read the rest of this entry »

%d bloggers like this: