Aram Harrow asked: "Is there any place this is written up?"

Update  Partly in answer to Aram's question, the thermodynamical properties of varietal dynamical systems now are written-up in our Soldier Healing Seminar's notes for 2013; and various motivations for caring about these properties are surveyed in a post to Dick Lipton and Ken Regan's Gödel's Lost Letter weblog. For engineers, the practical question at-issue is whether thermodynamics "just works" on varietal state-spaces, not only at regular points, but at singular points.

Introduction  In considering the question "Are quantum states exponentially long vectors?" (arXiv:quant-ph/0507242) Scott Aaronson provocatively argued that:

The real reason to study quantum computing is not to learn other people’s secrets, but to unravel the ultimate Secret of Secrets: Is our universe a polynomial or an exponential place? …

For me, the main weakness in the arguments of quantum computing skeptics has always been their failure to suggest an answer to the following question: What criterion separates the quantum states we’re sure we can prepare, from the states that arise in Shor’s factoring algorithm?

In a (wonderfully interesting!) subsequent debate, the thesis that our universe is (in Aaronson's phrase) "an exponential place" has been ably defended by Aram Harrow, whereas Gil Kalai has marshalled cogent reasons — for example, in a UW/PIMS colloquium (Friday November 23; details here) — "Why quantum computers cannot work."

The questions asked  The questions asked concern the classification of secant varieties of Segre varieties of Veronese varieties (SSVs). Viewed as dynamical manifolds immersed in an embedding Hilbert state-space, SSVs find broad applications in quantum simulation problems.

The questions asked concretely concern a dynamical Hilbert space of 27 dimensions, which is viewed as a 26-dimensional projective space $\mathbb{P}^{26}$. Numerical studies indicate that the following five varieties are immersed submanifolds of dimension 25, that is, the following five varietal state-spaces are Hilbert-deficient by precisely one dimension:

$$\begin{alignedat}{2} \sigma_4\big(\,&\text{Seg}(\,\mathbb{P}^2\times\mathbb{P}^2\times\mathbb{P}^2\,)\big)\\ \sigma_5\big(\,&\text{Seg}(\,\mathbb{G}^2\times\mathbb{P}^2\times\mathbb{P}^2\,)\big)\\ \sigma_7\big(\,&\text{Seg}(\,\mathbb{G}^2\times\mathbb{G}^2\times\mathbb{G}^2\,)\big)\\ \sigma_9\big(\,&\text{Seg}(\,\mathbb{G}^8\times\mathbb{P}^2)\,\big)\\ \sigma_{13}\big(\,&\text{Seg}(\,\mathbb{G}^{26}\,)\big) \end{alignedat}$$

The Four Questions

Q1  In what mathematical respects (if any) do these five varietal spaces differ from one another, in regard to their topology, singularity structure, metric structure, and symplectic structure? Do they all pushforward/blowdown to the same varietal submanifold of Hilbert space? Or do they represent five essentially different "foamy" algebraic/metric/symplectic geometries?

Q2  By what physical measurement processes (if any) can dynamical trajectories on these manifolds be distinguished from one another, and from trajectories on the (flat, singularity-free) Hilbert space in which they are embedded?

Q3  In what quantum mechanical respects (if any) are trajectory unravelings on these manifolds sufficiently deficient in any and all forms of "quantum goodness" as to invalidate their claim to be "Sure-Shor separators" (in the sense of Aaronson)?

Q4  What classification theorems (if any) exist to help quantum systems engineers in exploiting these varietal spaces for purposes of large-scale quantum dynamical simulation?

Physical remarks  The state-spaces are stipulated to inherit by pullback the standard metric, symplectic, Hamiltonian, and Lindbladian structures of the immersing Hilbert space, such that (from a Harrow-esque perspective) any quantum experiment envisioned to be performed in the Hilbert space, has an algebraically natural pullback/blowup/resolution of singularities on the varieties named, such that simulated quantum trajectories unraveled on the varietal state-space are a reasonable approximation to the "true" quantum state-space.

Alternatively (from a Kalai-esque perspective), any quantum experiment envisioned to be performed in the varietal state-space, has an algebraically natural pushforward/blowdown onto a fictitious Hilbert space of exponentially many dimensions, that is a maximally symmetric (but computationally intractable) approximation to the "true" varietal state-space.

In a nutshell  These varietal spaces are candidate manifolds for Aaronson-style "Sure-Shor separators", and the difficulty of distinguishing geometric perturbations of dynamical trajectories on these "foamy" algebraic manifolds from noise perturbations helps us appreciate the depth and subtlety of Gil Kalai's reasons "why quantum computers cannot work."

Notational remarks  The question adopts and slightly generalizes the notation of Landsberg's Tensors: Geometry and Applications as follows: $\sigma_r$ is the usual rank-$r$ secant join, $\text{Seg}$ is the usual Segre embedding, and $\mathbb{P}^k$ is the usual projective space

$$\mathbb{P}^k(\,\boldsymbol{\xi}\,) = [\xi_0,\xi_1,\dots,\xi_k]$$

The projective space $\mathbb{G}^k$ is (what mathematicians call) a Veronese embedding

$$ \mathbb{G}^k(a,b) = [\,{% \textstyle{{k}\choose{0}}^{1/2}a^k b^0,% {{k}\choose{1}}^{1/2}a^{k-1} b^1,% \dots,% {{k}\choose{k-1}}^{1/2}a^{1} b^{k-1},% {{k}\choose{k}}^{1/2}a^{0} b^{k}% }\,]\hookrightarrow \mathbb{P}^k.$$

Physicists regard this Veronese embedding as the space of coherent states; the "$\mathbb{G}$" of the above notation is a mnemonic for Roy Glauber, who pioneered the application of coherent states to broad-ranging practical problems, first in quantum optics and nowadays in many branches of science and engineering; the connexion between the physics viewpoint and the algebraic geometry viewpoint is surveyed by Dorje Brody and Eva-Maria Graefe's recent Coherent states and rational surfaces; the complex numbers $a,b$ are associated to the quaternionic rotation coordinates $\boldsymbol{q} = \{q_0,q_1,q_2,q_3\}$ that are favored by engineers by

$$ \begin{aligned} a(\boldsymbol{q}) &= q_0 + i q_3\\ b(\boldsymbol{q}) &= i q_1 - q_2\ ; \end{aligned} $$

and finally, the binomial weights ${{k}\choose{0}}^{1/2}$ etc. that appear in the definition of $\mathbb{G}^k$ commonly are omitted in the mathematical literature. The binomial weights ensure that the Veronese embedding is a metric and symplectic isometry, as is desirable to preserve various aspects of "quantum goodness" for simulation purposes; Katsumi Nomizu (1976) seems to have been among the first to appreciate their rationale.

Motivation  This question is constructed as a tribute to Aaronson's fine "Sure-Shor Separator" question, and as a tribute to the outstanding Kalai/Harrow debate that Aaronson's question helped stimulate. For engineers, Gil Kalai's proposition "Why quantum computers CAN'T work" belongs to a broad class of strategically crucial propositions that include "How large-scale quantum simulation DOES work" and "How atomic-resolution microscopes CAN work", and include too unanswered fundamental physics questions like "Is Nature's quantum dynamical state-space EXPERIMENTALLY static and flat?"

Acknowledgements  Appreciation and thanks are extended to Scott Aaronson, Gil Kalai, and Aram Harrow, for their sustained and admirably collegial commitment to asking good questions and seeking good answers.


1 Answer 1


I doubt that these theories can recreate the causal properties of nonrelativistic QM, i.e. the feature of local evolution/measurement not causing observable effects far away. I think they are also incompatible with the traditional Born rule (i.e. probability = absolute value squared of amplitude), because these theories plus the Born rule would mean that certain correlations are impossible for classical probability distributions, which would quick lead to undesirable behavior.

As written your models are incomplete, because they describe only Hamiltonian evolution, but not (a) how to combine several such systems, or (b) how measurement works, including local measurement. I really can't see how to complete your models in this way without breaking lots of other things.

(Also, I think Gil believes in the standard model of Hilbert space, and disagrees only about whether the noise assumptions of the threshold theorem are realistic. So I think his skepticism is distinct from this line of reasoning.)

  • 1
    $\begingroup$ Hi Aram! Gil was just here in Seattle and it was fun talking with him. Your questions have natural answers, e.g., combining/separating systems is "rank-jumping", measurement is the pullback of Lindblad/Stratonovich processes, spectral theorems become Weinstein conjectures (etc). So it's emergently "known" (empirically per Feynman) that varietal dynamics can be a pretty good model of the Kalai postulates. I'll update the question later in the week! $\endgroup$ Nov 26, 2013 at 7:16
  • 1
    $\begingroup$ In regard to rank-jumping, Elizabet Allman et al Tensor rank, invariants, inequalities, and applications (2012), provide algebraic tools that help answer questions like: How do we couple/decouble qudits into/from a dynamical system? The key lesson (for me) is that (as Borges says in the story Averroes's Search) "What we seek is near to hand." We are seeking is Sure/Shor separators, and in varietal state-spaces that model the Kalai postulates, we find that these separators indeed are near-to-hand. $\endgroup$ Nov 26, 2013 at 7:50
  • 1
    $\begingroup$ Those papers look interesting! I wouldn't say that every analogue of Hilbert space needs to exist in these generalized theories, but it would be good to have an analogue of reduced density matrices - and specifically of dynamics (including measurement) that can be defined locally, without depending on far-away systems. $\endgroup$
    – aram
    Nov 27, 2013 at 4:09
  • $\begingroup$ Yes to all your remarks. On non-Hilbert state-spaces, it turns out that density matrices (which AFAICT don't pullback in any natural way) unravel as ensembles of trajectories that are integral curves of Lindblad/Stratonovich stochastic processes (that do pullback in a natural way). Then the same measurement-induced localization that is associated to qudit input/output serves broadly to localize thermodynamical transport processes. The resulting quantum-yet-nonHilbert dynamics turns out to be a pretty good model for Gil Kalai's postulates and it's also a practical simulation framework. $\endgroup$ Nov 27, 2013 at 16:31
  • 1
    $\begingroup$ Is there any place this is written up? Normal Hilbert spaces have the property that pure entangled states have mixed reduced states, and that mixed states can be viewed as reduced states of pure entangled states. Are the non-Hilbert versions of these worked out anywhere? $\endgroup$
    – aram
    Nov 28, 2013 at 16:58

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.