Catching up with the quantum-thermo crowd

You have four hours to tour Oxford University.

What will you visit? The Ashmolean Museum, home to da Vinci drawings, samurai armor, and Egyptian mummies? The Bodleian, one of Europe’s oldest libraries? Turf Tavern, where former president Bill Clinton reportedly “didn’t inhale” marijuana?

Felix Binder showed us a cemetery.

Of course he showed us a cemetery. We were at a thermodynamics conference.

The Fifth Quantum Thermodynamics Conference took place in the City of Dreaming Spires.Participants enthused about energy, information, engines, and the flow of time. About 160 scientists attended—roughly 60 more than attended the first conference, co-organizer Janet Anders estimated.

IMG_1177

Weak measurements and quasiprobability distributions were trending. The news delighted me, Quantum Frontiers regulars won’t be surprised to hear.

Measurements disturb quantum systems, as early-20th-century physicist Werner Heisenberg intuited. Measure a system’s position strongly, and you forfeit your ability to predict the outcomes of future momentum measurements. Weak measurements don’t disturb the system much. In exchange, weak measurements provide little information about the system. But you can recoup information by performing a weak measurement in each of many trials, then processing the outcomes.

Strong measurements lead to probability distributions: Imagine preparing a particle in some quantum state, then measuring its position strongly, in each of many trials. From the outcomes, you can infer a probability distribution \{ p(x) \}, wherein p(x) denotes the probability that the next trial will yield position x.

Weak measurements lead analogously to quasiprobability distributions. Quasiprobabilities resemble probabilities but can misbehave: Probabilities are real numbers no less than zero. Quasiprobabilities can dip below zero and can assume nonreal values.

Do not disturb 2

What relevance have weak measurements and quasiprobabilities to quantum thermodynamics? Thermodynamics involves work and heat. Work is energy harnessed to perform useful tasks, like propelling a train from London to Oxford. Heat is energy that jiggles systems randomly.

Quantum properties obscure the line between work and heat. (Here’s an illustration for experts: Consider an isolated quantum, such as a spin chain. Let H(t) denote the Hamiltonian that evolves with the time t \in [0, t_f]. Consider preparing the system in an energy eigenstate | E_i(0) \rangle. This state has zero diagonal entropy: Measuring the energy yields E_i(0) deterministically. Considering tuning H(t), as by changing a magnetic field. This change constitutes work, we learn in electrodynamics class. But if H(t) changes quickly, the state can acquire weight on multiple energy eigenstates. The diagonal entropy rises. The system’s energetics have gained an unreliability characteristic of heat absorption. But the system has remained isolated from any heat bath. Work mimics heat.)

Quantum thermodynamicists have defined work in terms of a two-point measurement scheme: Initialize the quantum system, such as by letting heat flow between the system and a giant, fixed-temperature heat reservoir until the system equilibrates. Measure the system’s energy strongly, and call the outcome E_i. Isolate the system from the reservoir. Tune the Hamiltonian, performing the quantum equivalent of propelling the London train up a hill. Measure the energy, and call the outcome E_f.

Any change \Delta E in a system’s energy comes from heat Q and/or from work W, by the First Law of Thermodynamics, \Delta E = Q + W.  Our system hasn’t exchanged energy with any heat reservoir between the measurements. So the energy change consists of work: E_f - E_i =: W.

Train

Imagine performing this protocol in each of many trials. Different trials will require different amounts W of work. Upon recording the amounts, you can infer a distribution \{ p(W) \}. p(W) denotes the probability that the next trial will require an amount W of work.

Measuring the system’s energy disturbs the system, squashing some of its quantum properties. (The measurement eliminates coherences, relative to the energy eigenbasis, from the state.) Quantum properties star in quantum thermodynamics. So the two-point measurement scheme doesn’t satisfy everyone.

Enter weak measurements. They can provide information about the system’s energy without disturbing the system much. Work probability distributions \{ p(W) \} give way to quasiprobability distributions \{ \tilde{p}(W) \}.

So propose Solinas and Gasparinetti, in these papers. Other quantum thermodynamicists apply weak measurements and quasiprobabilities differently.2 I proposed applying them to characterize chaos, and the scrambling of quantum information in many-body systems, at the conference.3 Feel free to add your favorite applications to the “comments” section.

Dinner

All the quantum ladies: The conference’s female participants gathered for dinner one conference night.

Wednesday afforded an afternoon for touring. Participants congregated at the college of conference co-organizer Felix Binder.3 His tour evoked, for me, the ghosts of thermo conferences past: One conference, at the University of Cambridge, had brought me to the grave of thermodynamicist Arthur Eddington. Another conference, about entropies in information theory, had convened near Canada’s Banff Cemetery. Felix’s tour began with St. Edmund Hall’s cemetery. Thermodynamics highlights equilibrium, a state in which large-scale properties—like temperature and pressure—remain constant. Some things never change.

IMG_1192

 

With thanks to Felix, Janet, and the other coordinators for organizing the conference.

1Oxford derives its nickname from an elegy by Matthew Arnold. Happy National Poetry Month!

2https://arxiv.org/abs/1508.00438https://arxiv.org/abs/1610.04285,
https://arxiv.org/abs/1607.02404,
https://journals.aps.org/prl/abstract/10.1103/PhysRevLett.118.070601,
https://journals.aps.org/prl/abstract/10.1103/PhysRevLett.120.040602

3Michele Campisi joined me in introducing out-of-time-ordered correlators (OTOCs) into the quantum-thermo conference: He, with coauthor John Goold, combined OTOCs with the two-point measurement scheme.

3Oxford University contains 38 colleges, the epicenters of undergraduates’ social, dining, and housing experiences. Graduate students and postdoctoral scholars affiliate with colleges, and senior fellows—faculty members—govern the colleges.

Rock-paper-scissors, granite-clock-idea

I have a soft spot for lamassu. Ten-foot-tall statues of these winged bull-men guarded the entrances to ancient Assyrian palaces. Show me lamassu, or apkallu—human-shaped winged deities—or other reliefs from the Neo-Assyrian capital of Nineveh, and you’ll have trouble showing me the door.

Assyrian art fills a gallery in London’s British Museum. Lamassu flank the gallery’s entrance. Carvings fill the interior: depictions of soldiers attacking, captives trudging, and kings hunting lions. The artwork’s vastness, its endurance, and the contact with a three-thousand-year-old civilization floor me. I tore myself away as the museum closed one Sunday night.

Lamassu

I visited the British Museum the night before visiting Jonathan Oppenheim’s research group at University College London (UCL). Jonathan combines quantum information theory with thermodynamics. He and others co-invented thermodynamic resource theories (TRTs), which Quantum Frontiers regulars will know of. TRTs are quantum-information-theoretic models for systems that exchange energy with their environments.

Energy is conjugate to time: Hamiltonians, mathematical objects that represent energy, represent also translations through time. We measure time with clocks. Little wonder that one can study quantum clocks using a model for energy exchanges.

Mischa Woods, Ralph Silva, and Jonathan used a resource theory to design an autonomous quantum clock. “Autonomous” means that the clock contains all the parts it needs to operate, needs no periodic winding-up, etc. When might we want an autonomous clock? When building quantum devices that operate independently of classical engineers. Or when performing a quantum computation: Computers must perform logical gates at specific times.

Clock

Wolfgang Pauli and others studied quantum clocks, the authors recall. How, Pauli asked, would an ideal clock look? Its Hamiltonian, \hat{H}_{\rm C}, would have eigenstates | E \rangle. The labels E denote possible amounts of energy.

The Hamiltonian would be conjugate to a “time operator” \hat{t}. Let | \theta \rangle denote an eigenstate of \hat{t}. This “time state” would equal an even superposition over the | E \rangle’s. The clock would occupy the state | \theta \rangle at time t_\theta.

Imagine measuring the clock, to learn the time, or controlling another system with the clock. The interaction would disturb the clock, changing the clock’s state. The disturbance wouldn’t mar the clock’s timekeeping, if the clock were ideal. What would enable an ideal clock to withstand the disturbances? The ability to have any amount of energy: E must stretch from - \infty to \infty. Such clocks can’t exist.

Approximations to them can. Mischa, Ralph, and Jonathan designed a finite-size clock, then characterized how accurately the clock mimics the ideal. (Experts: The clock corresponds to a Hilbert space of finite dimensionality d. The clock begins in a Gaussian state that peaks at one time state | \theta \rangle. The finite-width Gaussian offers more stability than a clock state.)

Disturbances degrade our ability to distinguish instants by measuring the clock. Imagine gazing at a kitchen clock through blurry lenses: You couldn’t distinguish 6:00 from 5:59 or 6:01. Disturbances also hinder the clock’s ability to implement processes, such as gates in a computation, at desired instants.

Mischa & co. quantified these degradations. The errors made by the clock, they found, decay inverse-exponentially with the clock’s size: Grow the clock a little, and the errors shrink a lot.

Reliefs

Time has degraded the lamassu, but only a little. You can distinguish feathers in their wings and strands in their beards. People portray such artifacts as having “withstood the flow of time,” or “evaded,” or “resisted.” Such portrayals have never appealed to me. I prefer to think of the lamassu as surviving not because they clash with time, but because they harmonize with it. The prospect of harmonizing with time—whatever that means—has enticed me throughout my life. The prospect partially underlies my research into time—perhaps childishly, foolishly—I recognize if I remove my blurry lenses before gazing in the mirror.

The creation of lasting works, like lamassu, has enticed me throughout my life. I’ve scrapbooked, archived, and recorded, and tended memories as though they were Great-Grandma’s cookbook. Ancient civilizations began alluring me at age six, partially due to artifacts’ longevity. No wonder I study the second law of thermodynamics.

Yet doing theoretical physics makes no sense from another perspective. The ancient Egyptians sculpted granite, when they could afford it. Gudea, king of the ancient city-state of Lagash, immortalized himself in diorite. I fashion ideas, which lack substance. Imagine playing, rather than rock-paper-scissors, granite-diorite-idea. The idea wouldn’t stand a chance.

Would it? Because an idea lacks substance, it can manifest in many forms. Plato’s cave allegory has manifested as a story, as classroom lectures, on handwritten pages, on word processors and websites, in cartloads of novels, in the film The Matrix, in one of the four most memorable advertisements I received from colleges as a high-school junior, and elsewhere. Plato’s allegory has survived since about the fourth century BCE. King Ashurbanipal’s lion-hunt reliefs have survived for only about 200 years longer.

The lion-hunt reliefs—and lamassu—exude a grandness, a majesty that’s attracted me as their longevity has. The nature of time and the perfect clock have as much grandness. Leaving the British Museum’s Assyrian gallery at 6 PM one Sunday, I couldn’t have asked for a more fitting location, 24 hours later, than in a theoretical-physics conversation.

 

With thanks to Jonathan, to Álvaro Martín-Alhambra, and to Mischa for their hospitality at UCL; to Ada Cohen for the “Art history of ancient Egypt and the ancient Near East” course for which I’d been hankering for years; to my brother, for transmitting the ancient-civilizations bug; and to my parents, who fed the infection with museum visits.

Click here for a follow-up to the quantum-clock paper.

Fractons, for real?

“Fractons” is my favorite new toy (short for quantum many-body toy models). It has amazing functions that my old toys do not have; it is so new that there are tons of questions waiting to be addressed; it is perfectly situated at the interface between quantum information and condensed matter and has attracted a lot of interest and efforts from both sides; and it gives me excuses and new incentives to learn more math. I have been having a lot of fun playing with it in the last couple years and in the process, I had the great opportunity to work with some amazing collaborators: Han Ma and Mike Hermele at Boulder, Ethan Lake at MIT, Wilbur Shirley at Caltech, Kevin Slagle at U Toronto and Zhenghan Wang at Station Q. Together we have written a few papers on this subject, but I always felt there are more interesting stories and more excitement in me than what can be properly contained in scientific papers. Hence this blog post.

How I first learned about Fractons

Qmem

Back in the early 2000s, a question that kept attracting and frustrating people in quantum information is how to build a quantum hard drive to store quantum information. This is of course a natural question to ask as quantum computing has been demonstrated to be possible, at least in theory, and experimental progress has shown great potential. It turned out, however, that the question is one of those deceptively enticing ones which are super easy to state, but extremely hard to answer. In a classical hard drive, information is stored using magnetism. Quantum information, instead of being just 0 and 1, is represented using superpositions of 0 and 1, and can be probed in non-commutative ways (that is, measuring along different directions can alter previous answers). To store quantum information, we need “magnetism” in all such non-commutative channels. But how can we do that?

At that time, some proposals had been made, but they either involve actively looking for and correcting errors throughout the time during which information is stored (which is something we never have to do with our classical hard drives) or going into four spatial dimensions. Reliable passive storage of quantum information seemed out of reach in the three-dimensional world we live in, even at the level of a proof of principle toy model!

Given all the previously failed attempts and without a clue about where else to look, this problem probably looked like a dead-end to many. But not to Jeongwan Haah, a fearless graduate student in Preskill’s group at IQIM at that time, who turned the problem from guesswork into a systematic computer search (over a constrained set of models). The result of the search surprised everyone. Jeongwan found a three-dimensional quantum spin model with physical properties that had never been seen before, making it a better quantum hard drive than any other model that we know of!

The model looks surprising not only to the quantum information community, but even more so to the condensed matter community. It is a strongly interacting quantum many-body model, a subject that has been under intense study in condensed matter physics. Yet it exhibits some very strange behaviors whose existence had not even been suspected. It is a condensed matter discovery made not from real materials in real experiments, but through computer search!

fractal

Excitations (stars) in Haah’s code live at the corner of a fractal.

In condensed matter systems, what we know can happen is that elementary excitations can come in the form of point particles – usually called quasi-particles – which can then move around and interact with other excitations. In Jeongwan’s model, now commonly referred to as Haah’s code, elementary excitations still come in the form of point particles, but they cannot freely move around. Instead, if they want to move, four of them have to coordinate with each other to move together, so that they stay at the vertices of a fractal shaped structure! The restricted motion of the quasi-particles leads to slower dynamics at low energy, making the model much better suited for the purpose of storing quantum information.

But how can something like this happen? This is the question that I want to yell out loud every time I read Jeongwan’s papers or listen to his talks. Leaving aside the motivation of building a quantum hard drive, this model presents a grand challenge to the theoretical framework we now have in condensed matter. All of our intuitions break down in predicting the behavior of this model; even some of the most basic assumptions and definitions do not apply.

Haahcode

The interactions in Haah’s code involve eight spins at a time (the eight Z’s and eight X’s in each cube).

I felt so uncomfortable and so excited at the same time because there was something out there that should be related to things I know, yet I totally did not understand how. And there was an even bigger problem. I was like a sick person going to a doctor but unable to pinpoint what was wrong. Something must have been wrong, but I didn’t know what that was and I didn’t know how to even begin to look for it. The model looked so weird. Interaction involved eight spins at a time; there was no obvious symmetry other than translation. Jeongwan, with his magic math power, worked out explicitly many of the amazing properties of the model, but that to me only added to the mystery. Where did all these strange properties coming from?

From the unfathomable to the seemingly approachable

I remained in this superposition of excited state and powerless state for several years, until Jeongwan moved to MIT and posted some papers with Sagar Vijay and Liang Fu in 2015 and 2016.

Xcube

Interaction terms in a nicer looking fracton model.

In these papers, they listed several other models, which, similar to Haah’s code, contain quasi-particle excitations whose motion is constrained. The constraints are weaker and these models do not make good quantum hard drives, but they still represent new condensed matter phenomena. What’s nice about these models is that the form of interaction is more symmetric, takes a simpler form, or is similar to some other models we are familiar with. The quasi-particles do not need a fractal-shaped structure to move around, instead they move along a line, in a plane, or at the corner of a rectangle. In fact, as early as 2005 – six years before Haah’s code, Claudio Chamon at Boston University already proposed a model of this kind. Together with the previous fractal examples, these models are what’s now being referred to as the fracton models. If the original Haah’s code looks like an ET from beyond the milky way, these models at least seem to live somewhere in the solar system. So there must be something that we can do to understand them better!

Obviously, I was not the only one who felt this way. A flurry of papers appeared on these “fracton” models. People came at these models armed with their favorite tools in condensed matter, looking for an entry point to crack them open. The two approaches that I found most attractive was the coupled layer construction and the higher rank gauge theory, and I worked on these ideas together with Han Ma, Ethan Lake and Michael Hermele. Each approach comes from a different perspective and establishes a connection between fractons and physical models that we are familiar with. In the coupled layer construction, the connection is to the 2D discrete gauge theories, while in the higher rank approach it is to the 3D gauge theory of electromagnetism.

I was excited about these results. They each point to simple physical mechanisms underlying the existence of fractons in some particular models. By relating these models to things I already know, I feel a bit relieved. But deep down, I know that this is far from the complete story. Our understanding barely goes beyond the particular models discussed in the paper. In condensed matter, we spend a lot of time studying toy models; but toy models are not the end goal. Toy models are only meaningful if they represent some generic feature in a whole class of models. It is not clear at all to what extent this is the case for fractons.

Step zero: define “order”, define “topological order”

I gave a talk about these results at KITP last fall under the title “Fracton Topological Order”. It was actually too big a title because all we did was to study specific realizations of individual models and analyze their properties. To claim topological order, one needs to show much more. The word “order” refers to the common properties of a continuous range of models within the same phase. For example, crystalline order refers to the regular lattice organization of atoms in the solid phase within a continuous range of temperature and pressure. When the word “topological” is added in front of “order”, it signifies that such properties are usually directly related to the topology of the system. A prototypical example is the fractional quantum Hall system, whose ground state degeneracy is directly determined by the topology of the manifold the system lives in. For fractons, we are far from an understanding at this level. We cannot answer basic questions like what range of models form a phase, what is the order (the common properties of this whole range of models) characterizing each phase, and in what sense is the order topological. So, the title was more about what I hope will happen than what has already happened.

But it did lead to discussions that could make things happen. After my talk, Zhenghan Wang, a mathematician at Microsoft Station Q, said to me, “I would agree these fracton models are topological if you can show me how to define them on different three manifolds”. Of course! How can I claim anything related to topology if all I know is one model on a cubic lattice with periodic boundary condition? It is like claiming a linear relation between two quantities with only one data point.

But how to get more data points? Well, from the paper by Haah, Vijay and Fu, we knew how to define the model on cubic lattices. With periodic boundary conditions, the underlying manifold is a three torus. Is it possible to have a cubic lattice, or something similar, in other three manifolds as well? Usually, this kind of request would be too much to ask. But it turns out that if you whisper your wish to the right mathematician, even the craziest ones can come true. With insightful suggestions from Michael Freedman (the Fields medalist leading Station Q) and Zhenghan, and through the amazing work of Kevin Slagle (U Toronto) and Wilbur Shirley (Caltech), we found that if we make use of a structure called Total Foliation, one of the fracton models can be generalized to different kinds of three manifolds and we can see how the properties of the model are related to certain topological features of the manifold!

Foliation

Foliation.

Foliation is the process of dividing a manifold into parallel planes. Total foliation is a set of three foliations which intersect each other in a transversal way. The xy, yz, and zx planes in a cubic lattice form a total foliation and similar constructions can be made for other three manifolds as well.

Things start to get technical from here, but the basic lesson we learned about some of the fracton models is that structural-wise, they pretty much look like an onion. Even though onions look like a three-dimensional object from the outside, they actually grow in a layered structure. Some of the properties of the fracton models are simply determined by the layers, and related

Onionto the topology of the layers. Once we peel off all the layers, we find that for some, there is nothing left while for others, there is a nontrivial core. This observation allows us to better address the previous questions: we defined a fracton phase (one type of it) as models smoothly related to each other after adding or removing layers; the topological nature of the order is manifested in how the properties of the model are determined by the topology of the layers.

staringThe onion structure is nice, because it allows us to reduce much of the story from 3D to 2D, where we understand things much better. It clarifies many of the weirdnesses of the fracton model we studied, and there is indication that it may apply to a much wider range of fracton models, so we have an exciting road ahead of us. On the other hand, it is also clear that the onion structure does not cover everything. In particular, it does not cover Haah’s code! Haah’s code cannot be built in a layered way and its properties are in a sense intrinsically three dimensional. So, after finishing this whole journey through the onion field, I will be back to staring at Haah’s code again and wondering what to do with it, like what I have been doing in the eight years since Jeongwan’s paper first came out. But maybe this time I will have some better ideas.

Machine learning the arXiv

Over the last year or so, the machine learning wave has really been sweeping through the field of condensed matter physics. Machine learning techniques have been applied to condensed matter physics before, but very sparsely and with little recognition. These days, I guess (partially) due to the general machine learning and AI hype, the amount of such studies skyrocketed (I admit to contributing to that..). I’ve been keeping track of this using the arXiv and Twitter (@Evert_v_N), but you should know about this website for getting an overview of the physics & machine learning papers: https://physicsml.github.io/pages/papers.html.

This effort of applying machine learning to physics is a serious attempt at trying to understand how such tools could be useful in a variety of ways. It isn’t very hard to get a neural network to learn ‘something’ from physics data, but it is really hard to find out what – and especially how – the network does that. That’s why toy cases such as the Ising model or the Kosterlitz-Thouless transition have been so important!

When you’re keeping track of machine learning and AI developments, you soon realize that there are examples out there of amazing feats. Being able to generate photo-realistic pictures given just a sentence. e.g. “a brown bird with golden speckles and red wings is sitting on a yellow flower with pointy petals”, is (I think..) pretty cool. I can’t help but wonder if we’ll get to a point where we can ask it to generate “the groundstate of the Heisenberg model on a Kagome lattice of 100×100”…

Another feat I want to mention, and the main motivation for this post, is that of being able to encode words as vectors. That doesn’t immediately seem like a big achievement, but it is once you want to have ‘similar’ words have ‘similar’ vectors. That is, you intuitively understand that Queen and King are very similar, but differ basically only in gender. Can we teach that to a computer (read: neural network) by just having it read some text? Turns out we can. The general encoding of words to vectors is aptly named ‘Word2Vec’, and some of the top algorithms that do that were introduced here (https://arxiv.org/abs/1301.3781) and here (https://arxiv.org/abs/1310.4546). The neat thing is that we can actually do arithmetics with these words encoded as vectors, so that the network learns (with no other input than text!):

  • King – Man + Woman = Queen
  • Paris – France + Italy = Rome

In that spirit, I wondered if we can achieve the same thing with physics jargon. Everyone knows, namely, that “electrons + two dimensions + magnetic field = Landau levels”. But is that clear from condensed matter titles?

Try it yourself

If you decide at this point that the rest of the blog is too long, at least have a look here: everthemore.pythonanywhere.com or skip to the last section. That website demonstrates the main point of this post. If that sparks your curiosity, read on!

This post is mainly for entertainment, and so a small disclaimer is in order: in all of the results below, I am sure things can be improved upon. Consider this a ‘proof of principle’. However, I would be thrilled to see what kind of trained models you can come up with yourself! So for that purpose, all of the code (plus some bonus content!) can be found on this github repository: https://github.com/everthemore/physics2vec.

Harvesting the arXiv

The perfect dataset for our endeavor can be found in the form of the arXiv. I’ve written a small script (see github repository) that harvests the titles of a given section from the arXiv. It also has options for getting the abstracts, but I’ll leave that for a separate investigation. Note that in principle we could also get the source-files of all of these papers, but doing that in bulk requires a payment; and getting them one by one will 1) take forever and 2) probably get us banned.

Collecting all this data of the physics:cond-mat subsection took right about 1.5 hours and resulted in 240737 titles and abstracts (I last ran this script on November 20th, 2017). I’ve filtered them by year and month, and you can see the result in Fig.1 below. Seems like we have some catching up to do in 2017 still (although as the inset shows, we have nothing to fear. November is almost over, but we still have the ‘getting things done before x-mas’ rush coming up!).

numpapers

Figure 1: The number of papers in the cond-mat arXiv section over the years. We’re behind, but the year isn’t over yet! (Data up to Nov 20th 2017)

Analyzing n-grams

After tidying up the titles (removing LaTeX, converting everything to lowercase, etc.), the next thing to do is to train a language model on finding n-grams. N-grams are basically fixed n-word expressions such as ‘cooper pair’ (bigram) or ‘metal insulator transition’ (trigram). This makes it easier to train a Word2Vec encoding, since these phrases are fixed and can be considered a single word. The python module we’ll use for Word2Vec is gensim (https://radimrehurek.com/gensim/), and it conveniently has phrase-detection built-in. The language model it builds reports back to us the n-grams it finds, and assigns them a score indicating how certain it is about them. Notice that this is not the same as how frequently it appears in the dataset. Hence an n-gram can appear fewer times than another, but have a higher certainty because it always appears in the same combination. For example, ‘de-haas-van-alphen’ appears less than, but is more certain than, ‘cooper-pair’, because ‘pair’ does not always come paired (pun intended) with ‘cooper’. I’ve analyzed up to 4-grams in the analysis below.

I can tell you’re curious by now to find out what some of the most certain n-grams in cond-mat are (again, these are not necessarily the most frequent), so here are some interesting findings:

  • The most certain n-grams are all surname combo’s, Affleck-Kennedy-Lieb-Tasaki being the number 1. Kugel-Khomskii is the most certain 2-name combo and Einstein-Podolksi-Rosen the most certain 3-name combo.
  • The first certain non-name n-gram is a ‘quartz tuning fork’, followed by a ‘superconducting coplanar waveguide resonator’. Who knew.
  • The bigram ‘phys. rev.’ and trigram ‘phys. rev. lett.’ are relatively high up in the confidence lists. These seem to come from the “Comment on […]”-titles on the arXiv.
  • I learned that there is such a thing as a Lefschetz thimble. I also learned that those things are called thimbles in English (we (in Holland) call them ‘finger-hats’!).

In terms of frequency however, which is probably more of interest to us, the most dominant n-grams are Two-dimensional, Quantum dot, Phase transition, Magnetic field, One dimensional and Bose-Einstein (in descending order). It seems 2D is still more popular than 1D, and all in all the top n-grams do a good job at ‘defining’ condensed matter physics. I’ll refer you to the github repository code if you want to see a full list! You’ll find there a piece of code that produces wordclouds from the dominant words and n-grams too, such as this one:

caltechwordcloud.png

For fun though, before we finally get to the Word2Vec encoding, I’ve also kept track of all of these as a function of year, so that we can now turn to finding out which bigrams have been gaining the most popularity. The table below shows the top 5 n-grams for the period 2010 – 2016 (not including 2017) and for the period 2015 – Nov 20th 2017.

2010-2016

2015 – November 20th 2017

Spin liquids  Topological phases & transitions
 Weyl semimetals  Spin chains
 Topological phases & transitions  Machine learning
 Surface states  Transition metal dichalcogenides
 Transition metal dichalcogenides  Thermal transport
 Many-body localization  Open quantum systems

Actually, the real number 5 in the left column was ‘Topological insulators’, but given number 3 I skipped it. Also, this top 5 includes a number 6 (!), which I just could not leave off given that everyone seems to have been working on MBL. If we really want to be early adopters though, taking only the last 1.8 years (2015 – now, Nov 20th 2017)  in the right column of the table shows some interesting newcomers. Surprisingly, many-body localization is not even in the top 20 anymore. Suffice it to say, if you have been working on anything topology-related, you have nothing to worry about. Machine learning is indeed gaining lots of attention, but we’ve yet to see if it doesn’t go the MBL-route (I certainly don’t hope so!). Quantum computing does not seem to be on the cond-mat radar, but I’m certain we would find that high up in the quant-ph arXiv section.

CondMat2Vec

Alright, finally time to use some actual neural networks for machine learning. As I started this post, what we’re about to do is try to train a network to encode/decode words into vectors, while simultaneously making sure that similar words (by meaning!) have similar vectors. Now that we have the n-grams, we want the Word2Vec algorithm to treat these as words by themselves (they are, after all, fixed combinations).

In the Word2Vec algorithm, we get to decide the length of the vectors that encode words ourselves. Larger vectors means more freedom in encoding words, but also makes it harder to learn similarity. In addition, we get to choose a window-size, indicating how many words the algorithm will look ahead to analyze relations between words. Both of these parameters are free for you to play with if you have a look at the source code repository. For the website everthemore.pythonanywhere.com, I’ve uploaded a size 100 with window-size 10 model, which I found to produce sensible results. Sensible here means “based on my expectations”, such as the previous example of “2D + electrons + magnetic field = Landau levels”. Let’s ask our network some questions.

First, as a simple check, let’s see what our encoding thinks some jargon is similar to:

  • Superconductor ~ Superconducting, Cuprate superconductor, Superconductivity, Layered superconductor, Unconventional superconductor, Superconducting gap, Cuprate, Weyl semimetal, …
  • Majorana ~ Majorana fermion, Majorana mode, Non-abelian, Zero-energy, braiding, topologically protected, …

It seems we could start to cluster words based on this. But the real test comes now, in the form of arithmetics. According to our network (I am listing the top two choices in some cases; the encoder outputs a list of similar vectors, ordered by similarity):

  • Majorana + Braiding = Non-Abelian
  • Electron + Hole = Exciton, Carrier
  • Spin + Magnetic field = Magnetization, Antiferromagnetic
  • Particle + Charge = Electron, Charged particle

And, sure enough:

  • 2D + electrons + magnetic field = Landau level, Magnetoresistance oscillation

The above is just a small sample of the things I’ve tried. See the link in the try it yourself section above if you want to have a go. Not all of the examples work nicely. For example, neither lattice + wave nor lattice + excitation nor lattice + force seem to result in anything related to the word ‘phonon’. I would guess that increasing the window size will help remedy this problem. Even better probably would be to include abstracts!

Outlook

I could play with this for hours, and I’m sure that by including the abstracts and tweaking the vector size (plus some more parameters I haven’t even mentioned) one could optimize this more. Once we have an optimized model, we could start to cluster the vectors to define research fields, visualizing the relations between n-grams (both suggestions thanks to Thomas Vidick and John Preskill!), and many other things. This post has become rather long already however, and I will leave further investigation to a possible future post. I’d be very happy to incorporate anything cool you find yourselves though, so please let me know!

Gently yoking yin to yang

The architecture at the University of California, Berkeley mystified me. California Hall evokes a Spanish mission. The main library consists of white stone pillared by ionic columns. A sea-green building scintillates in the sunlight like a scarab. The buildings straddle the map of styles.

Architecture.001

So do Berkeley’s quantum scientists, information-theory users, and statistical mechanics.

The chemists rove from abstract quantum information (QI) theory to experiments. Physicists experiment with superconducting qubits, trapped ions, and numerical simulations. Computer scientists invent algorithms for quantum computers to perform.

Few activities light me up more than bouncing from quantum group to info-theory group to stat-mech group, hunting commonalities. I was honored to bounce from group to group at Berkeley this September.

What a trampoline Berkeley has.

The groups fan out across campus and science, but I found compatibility. Including a collaboration that illuminated quantum incompatibility.

Quantum incompatibility originated in studies by Werner Heisenberg. He and colleagues cofounded quantum mechanics during the early 20th century. Measuring one property of a quantum system, Heisenberg intuited, can affect another property.

The most famous example involves position and momentum. Say that I hand you an electron. The electron occupies some quantum state represented by | \Psi \rangle. Suppose that you measure the electron’s position. The measurement outputs one of many possible values x (unless | \Psi \rangle has an unusual form, the form a Dirac delta function).

But we can’t say that the electron occupies any particular point x = x_0 in space. Measurement devices have limited precision. You can measure the position only to within some error \varepsilon: x = x_0 \pm \varepsilon.

Suppose that, immediately afterward, you measure the electron’s momentum. This measurement, too, outputs one of many possible values. What probability q(p) dp does the measurement have of outputting some value p? We can calculate q(p) dp, knowing the mathematical form of | \Psi \rangle and knowing the values of x_0 and \varepsilon.

q(p) is a probability density, which you can think of as a set of probabilities. The density can vary with p. Suppose that q(p) varies little: The probabilities spread evenly across the possible p values. You have no idea which value your momentum measurement will output. Suppose, instead, that q(p) peaks sharply at some value p = p_0. You can likely predict the momentum measurement’s outcome.

The certainty about the momentum measurement trades off with the precision \varepsilon of the position measurement. The smaller the \varepsilon (the more precisely you measured the position), the greater the momentum’s unpredictability. We call position and momentum complementary, or incompatible.

You can’t measure incompatible properties, with high precision, simultaneously. Imagine trying to do so. Upon measuring the momentum, you ascribe a tiny range of momentum values p to the electron. If you measured the momentum again, an instant later, you could likely predict that measurement’s outcome: The second measurement’s q(p) would peak sharply (encode high predictability). But, in the first instant, you measure also the position. Hence, by the discussion above, q(p) would spread out widely. But we just concluded that q(p) would peak sharply. This contradiction illustrates that you can’t measure position and momentum, precisely, at the same time.

But you can simultaneously measure incompatible properties weakly. A weak measurement has an enormous \varepsilon. A weak position measurement barely spreads out q(p). If you want more details, ask a Quantum Frontiers regular; I’ve been harping on weak measurements for months.

Blame Berkeley for my harping this month. Irfan Siddiqi’s and Birgitta Whaley’s groups collaborated on weak measurements of incompatible observables. They tracked how the measured quantum state | \Psi (t) \rangle evolved in time (represented by t).

Irfan’s group manipulates superconducting qubits.1 The qubits sit in the physics building, a white-stone specimen stamped with an egg-and-dart motif. Across the street sit chemists, including members of Birgitta’s group. The experimental physicists and theoretical chemists teamed up to study a quantum lack of teaming up.

Phys. & chem. bldgs

The experiment involved one superconducting qubit. The qubit has properties analogous to position and momentum: A ball, called the Bloch ball, represents the set of states that the qubit can occupy. Imagine an arrow pointing from the sphere’s center to any point in the ball. This Bloch vector represents the qubit’s state. Consider an arrow that points upward from the center to the surface. This arrow represents the qubit state | 0 \rangle. | 0 \rangle is the quantum analog of the possible value 0 of a bit, or unit of information. The analogous downward-pointing arrow represents the qubit state | 1 \rangle, analogous to 1.

Infinitely many axes intersect the sphere. Different axes represent different observables that Irfan’s group can measure. Nonparallel axes represent incompatible observables. For example, the x-axis represents an observable \sigma_x analogous to position. The y-axis represents an observable \sigma_y analogous to momentum.

Tug-of-war

Siddiqi lab, decorated with the trademark for the paper’s tug-of-war between incompatible observables. Photo credit: Leigh Martin, one of the paper’s leading authors.

Irfan’s group stuck their superconducting qubit in a cavity, or box. The cavity contained light that interacted with the qubit. The interactions transferred information from the qubit to the light: The light measured the qubit’s state. The experimentalists controlled the interactions, controlling the axes “along which” the light was measured. The experimentalists weakly measured along two axes simultaneously.

Suppose that the axes coincided—say, at the x-axis \hat{x}. The qubit would collapse to the state | \Psi \rangle = \frac{1}{ \sqrt{2} } ( | 0 \rangle + | 1 \rangle ), represented by the arrow that points along \hat{x} to the sphere’s surface, or to the state | \Psi \rangle = \frac{1}{ \sqrt{2} } ( | 0 \rangle - | 1 \rangle ), represented by the opposite arrow.

0 deg

(Projection of) the Bloch Ball after the measurement. The system can access the colored points. The lighter a point, the greater the late-time state’s weight on the point.

Let \hat{x}' denote an axis near \hat{x}—say, 18° away. Suppose that the group weakly measured along \hat{x} and \hat{x}'. The state would partially collapse. The system would access points in the region straddled by \hat{x} and \hat{x}', as well as points straddled by - \hat{x} and - \hat{x}'.

18 deg

Finally, suppose that the group weakly measured along \hat{x} and \hat{y}. These axes stand in for position and momentum. The state would, loosely speaking, swirl around the Bloch ball.

90 deg

The Berkeley experiment illuminates foundations of quantum theory. Incompatible observables, physics students learn, can’t be measured simultaneously. This experiment blasts our expectations, using weak measurements. But the experiment doesn’t just destroy. It rebuilds the blast zone, by showing how | \Psi (t) \rangle evolves.

“Position” and “momentum” can hang together. So can experimentalists and theorists, physicists and chemists. So, somehow, can the California mission and the ionic columns. Maybe I’ll understand the scarab building when we understand quantum theory.2

With thanks to Birgitta’s group, Irfan’s group, and the rest of Berkeley’s quantum/stat-mech/info-theory community for its hospitality. The Bloch-sphere figures come from http://www.nature.com/articles/nature19762.

1The qubit is the quantum analog of a bit. The bit is the basic unit of information. A bit can be in one of two possible states, which we can label as 0 and 1. Qubits can manifest in many physical systems, including superconducting circuits. Such circuits are tiny quantum circuits through which current can flow, without resistance, forever.

2Soda Hall dazzled but startled me.

Paradise

The word dominates chapter one of Richard Holmes’s book The Age of WonderHolmes writes biographies of Romantic-Era writers: Mary Wollstonecraft, Percy Shelley, and Samuel Taylor Coleridge populate his bibliography. They have cameos in Age. But their scientific counterparts star.

“Their natural-philosopher” counterparts, I should say. The word “scientist” emerged as the Romantic Era closed. Romanticism, a literary and artistic movement, flourished between the 1700s and the 1800s. Romantics championed self-expression, individuality, and emotion over convention and artificiality. Romantics wondered at, and drew inspiration from, the natural world. So, Holmes argues, did Romantic-Era natural philosophers. They explored, searched, and innovated with Wollstonecraft’s, Shelley’s, and Coleridge’s zest.

Age of Wonder

Holmes depicts Wilhelm and Caroline Herschel, a German brother and sister, discovering the planet Uranus. Humphry Davy, an amateur poet from Penzance, inventing a lamp that saved miners’ lives. Michael Faraday, a working-class Londoner, inspired by Davy’s chemistry lectures.

Joseph Banks in paradise.

So Holmes entitled chapter one.

Banks studied natural history as a young English gentleman during the 1760s. He then sailed around the world, a botanist on exploratory expeditions. The second expedition brought Banks aboard the HMS Endeavor. Captain James Cook steered the ship to Brazil, Tahiti, Australia, and New Zealand. Banks brought a few colleagues onboard. They studied the native flora, fauna, skies, and tribes.

Banks, with fellow botanist Daniel Solander, accumulated over 30,000 plant samples. Artist Sydney Parkinson drew the plants during the voyage. Parkinson’s drawings underlay 743 copper engravings that Banks commissioned upon returning to England. Banks planned to publish the engravings as the book Florilegium. He never succeeded. Two institutions executed Banks’s plan more than 200 years later.

Banks’s Florilegium crowns an exhibition at the University of California at Santa Barbara (UCSB). UCSB’s Special Research Collections will host “Botanical Illustrations and Scientific Discovery—Joseph Banks and the Exploration of the South Pacific, 1768–1771” until May 2018. The exhibition features maps of Banks’s journeys, biographical sketches of Banks and Cook, contemporary art inspired by the engravings, and the Florilegium.

online poster

The exhibition spotlights “plants that have subsequently become important ornamental plants on the UCSB campus, throughout Santa Barbara, and beyond.” One sees, roaming Santa Barbara, slivers of Banks’s paradise.

2 bouganvilleas

In Santa Barbara resides the Kavli Institute for Theoretical Physics (KITP). The KITP is hosting a program about the physics of quantum information (QI). QI scientists are congregating from across the world. Everyone visits for a few weeks or months, meeting some participants and missing others (those who have left or will arrive later). Participants attend and present tutorials, explore beyond their areas of expertise, and initiate research collaborations.

A conference capstoned the program, one week this October. Several speakers had founded subfields of physics: quantum error correction (how to fix errors that dog quantum computers), quantum computational complexity (how quickly quantum computers can solve hard problems), topological quantum computation, AdS/CFT (a parallel between certain gravitational systems and certain quantum systems), and more. Swaths of science exist because of these thinkers.

KITP

One evening that week, I visited the Joseph Banks exhibition.

Joseph Banks in paradise.

I’d thought that, by “paradise,” Holmes had meant “physical attractions”: lush flowers, vibrant colors, fresh fish, and warm sand. Another meaning occurred to me, after the conference talks, as I stood before a glass case in the library.

Joseph Banks, disembarking from the Endeavour, didn’t disembark onto just an island. He disembarked onto terra incognita. Never had he or his colleagues seen the blossoms, seed pods, or sprouts before him. Swaths of science awaited. What could the natural philosopher have craved more?

QI scientists of a certain age reminisce about the 1990s, the cowboy days of QI. When impactful theorems, protocols, and experiments abounded. When they dangled, like ripe fruit, just above your head. All you had to do was look up, reach out, and prove a pineapple.

Cowboy

Typical 1990s quantum-information scientist

That generation left mine few simple theorems to prove. But QI hasn’t suffered extinction. Its frontiers have advanced into other fields of science. Researchers are gaining insight into thermodynamics, quantum gravity, condensed matter, and chemistry from QI. The KITP conference highlighted connections with quantum gravity.

…in paradise.

What could a natural philosopher crave more?

Contemporary

Artwork commissioned by the UCSB library: “Sprawling Neobiotic Chimera (After Banks’ Florilegium),” by Rose Briccetti

Most KITP talks are recorded and released online. You can access talks from the conference here. My talk, about quantum chaos and thermalization, appears here. 

With gratitude to the KITP, and to the program organizers and the conference organizers, for the opportunity to participate. 

What Clocks have to do with Quantum Computation

Have you ever played the game “telephone”? You might remember it from your nursery days, blissfully oblivious to the fact that quantum mechanics governs your existence, and not yet wondering why Fox canceled Firefly. For everyone who forgot, here is the gist of the game: sit in a circle with your friends. Now you think of a story (prompt: a spherical weapon that can destroy planets). Once you have the story laid out in your head, tell it to your neighbor on your left. She takes the story and tells it to her friend on her left. It is important to master the art of whispering for this game: you don’t want to be overheard when the story is passed on. After one round, the friend on your right tells you what he heard from his friend on his right. Does the story match your masterpiece?

If your story is generic, it probably survived without alterations. Tolstoy’s War and Peace, on the other hand, might turn into a version of Game of Thrones. Passing along complex stories seems to be more difficult than passing on easy ones, and it also becomes more prone to errors the more friends join your circle—which makes intuitive sense.

So what does this have to do with physics or quantum computation?

Let’s add maths to this game, because why not. Take a difficult calculation that follows a certain procedure, such as long division of two integer numbers.

long-division

Now you perform one step of the division and pass the piece of paper on to your left. Your friend there is honest and trusts you: she doesn’t check what you did, but happily performs the next step in the division. Once she’s done, she passes the piece of paper on to her left, and so on. By the time the paper reaches you again, you hopefully have the result of the calculation, given you have enough friends to divide your favorite numbers, and given that everyone performed their steps accurately.

I’m not sure if Feynman thought about telephone when he, in 1986, proposed a method of embedding computation into eigenstates (e.g. the ground state) of a Hamiltonian, but the fact remains that the similarity is striking. Remember that writing down a Hamiltonian is a way of describing a quantum-mechanical system, for instance how the constituents of a multi-body system are coupled with each other. The ground state of such a Hamiltonian describes the lowest energy state that a system assumes when it is cooled down as far as possible. Before we dive into how the Hamiltonian looks, let’s try to understand how, in Feynman’s construction, a game of telephone can be represented as a quantum state of a physical system.

telephone-history-state

In this picture, | \psi_t \rangle represents a snapshot of the story or calculation at time t—in the division example, this would be the current divisor and remainder terms; so e.g. the snapshot | \psi_1 \rangle represents the initial dividend and divisor, and the person next to you is thinking of | \psi_2 \rangle, one step into the calculation. The label |t\rangle in front of the tensor sign \otimes is like a tag that you put on files on your computer, and uniquely associates the snapshot | \psi_t \rangle with the t-th time step. We say that the story snapshot is entangled with its label.

This is also an example of quantum superposition: all the |t\rangle\otimes|\psi_t\rangle are distinct states (the time labels, if not the story snapshots, are all unique), and by adding these states up we put them into superposition. So if we were to measure the time label, we would obtain one of the snapshots uniformly at random—it’s as if you had a cloth bag full of cards, and you blindly pick one. One side of the card will have the time label on it, while the other side contains the story snapshot. But don’t be fooled—you cannot access all story snapshots by successive measurements! Quantum states collapse; whatever measurement outcome you have dictates what the quantum state will look like after the measurement. In our example, this means that we burn the cloth bag after you pick your card; in this sense, the quantum state behaves differently than a simple juxtaposition of scraps of paper.

Nonetheless, this is the reason why we call such a quantum state a history state: it preserves the history of the computation, where every step that is performed is appropriately tagged. If we manage to compare all pairs of successively-labeled snapshots (without measuring them!), one can verify that the end result does, in fact, stem from a valid computation—and not just a random guess. In the division example, this would correspond to checking that each of your friends performs a correct division step.

So history states are clearly useful. But how do you design a Hamiltonian with a history state as the ground state? Is it even possible? The answer is yes, and it all boils down to verifying that two successive snapshots | \psi_t \rangle and | \psi_{t+1} \rangle are related to each other in the correct manner, e.g. that your friend on seat t+1 performs a valid division step from the snapshot prepared by the person on seat t. In fancy physics speak (aka Bra-Ket notation), we can for example write

local-check

The actual Hamiltonian will then be a sum of such terms, and one can verify that its ground state is indeed the one representing the history state we introduced above.

I’m glossing over a few details here: there is a minus sign in front of this term, and we have to add its Hermitian conjugate (flip the labels and snapshots around). But this is not essential for the argument, so let’s not go there for now. However, you’re totally right with one thing: it wouldn’t make sense to write down all snapshots themselves into the Hamiltonian! After all, if we had to calculate every snapshot transition like | \psi_2 \rangle \langle \psi_1 | in advance, there would be no use to this construction. So instead, we can write

local-check-2.png

Perfect. We now have a Hamiltonian which, in its ground state, can encode the history of a computation, and if we replace the transition operator \mathbf U_\text{DIVISION} with another desired transition operator (a unitary matrix), we can perform any computation we want (more precisely, any computation that can be written as a unitary matrix; this includes anything your laptop can do). However, this is only half of the story, since we need to have a way of reading out the final answer. So let’s step back for a moment, and go back to the telephone game.

Can you motivate your friends to cheat?

Your friends playing telephone make mistakes.

no-mistakes

Ok, let’s assume we give them a little incentive: offer $1 to the person on your right in case the result is an even number. Will he cheat? With so much at stake?

no-mistakes-bribe.png

In fact, maybe your friend is not only greedy but also dishonest: he wants to hide the fact that he miscalculates on purpose, and sometimes tells his friend on his right to make a mistake instead (maybe giving him a share of the money). So for a few of your friends close to the person at the end of the chain, there is a real incentive to cheat!

local-check-3.png

Can we motivate spins to cheat?

We already discussed how to write down a Hamiltonian that verifies valid computational steps. But can we do the same thing as bribing your friends to procure a certain outcome? Can we give an energy bonus to certain outcomes of the computation?

In fact, we can. Alexei Kitaev proposed adding a term to Feynman’s Hamiltonian which raises the energy of an unwanted outcome, relative to a desirable outcome. How? Again in fancy physics language,

local-check-4.png

What this term does is that it takes the history state and yields a negative energy contribution (signaled by the minus sign in front) if the last snapshot | \psi_T \rangle is an even number. If it isn’t, no bonus is felt; this would correspond to you keeping the dollar you promised to your friend. This simply means that in case the computation has a desirable outcome—i.e. an even number—the Hamiltonian allows a lower energy ground state than for any other output. Et voilà, we can distinguish between different outputs of the computation.

The true picture is, of course, a tad more complicated; generally, we give penalty terms to unwanted states instead of bonus terms to desirable ones. The reason for this is somewhat subtle, but can potentially be explained with an analogy: humans fear loss much more than they value gains of the same magnitude. Quantum systems behave in a completely opposite manner: the promise of a bonus at the end of the computation is such a great incentive that most of the weight of the history state will flock to the bonus term (for the physicists: the system now has a bound state, meaning that the wave function is localized around a specific site, and drops off exponentially quickly away from it). This makes it difficult to verify the computation far away from the bonus term.

So the Feynman-Kitaev Hamiltonian consists of three parts: one which checks each step of the computation, one which penalizes invalid outcomes—and obviously we also need to make sure the input of the computation is valid. Why? Well, are you saying you are more honest than your friends?

local-check-5.png

Physical Implications of History State Hamiltonians

If there is one thing I’ve learned throughout my PhD it is that we should always ask what use a theory is. So what can we learn from this construction? Almost 20 years ago, Alexei Kitaev used Feynman’s idea to prove that estimating the ground state energy of a physical system with local interactions is hard, even on a quantum computer (for the experts: QMA-hard under the assumption of a 1/\text{poly} promise gap splitting the embedded YES and NO instances). Why is estimating the ground state energy hard? The energy shift induced by the output penalty depends on the outcome of the computation that we embed (e.g. even or odd outcome). And as fun as long division is, there are much more difficult tasks we can write down as a history state Hamiltonian—in fact, it is this very freedom which makes estimating the ground state energy difficult: if we can embed any computation we want, estimating the induced energy shift should be at least as hard as actually performing the computation on a quantum computer. This has one curious implication: if we don’t expect that we can estimate the ground state energy efficiently, the physical system will take a long time to actually assume its ground state when cooled down, and potentially behave like a spin glass!

Feynman’s history state construction and the QMA-hardness proof of Kitaev were a big part of the research I did for my PhD. I formalized the case where the message is not passed on along a unique path from neighbor to neighbor, but can take an arbitrary path between beginning and end in a more complicated graph; in this way, computation can in some sense be parallelized.

Well, to be honest, the last statement is not entirely true: while there can be parallel tracks of computation from A to B, these tracks have to perform the same computation (albeit in potentially different steps); otherwise the system becomes much more complicated to analyze. The reason why this admittedly quite restricted form of branching might still be an advantage is somewhat subtle: if your computation has a lot of classical if-else cases, but you don’t have enough space on your piece of paper to store all the variables to check the conditions, it might be worth just taking a gamble: pass your message down one branch, in the hope that the condition is met. The only thing that you have to be careful about is that in case the condition isn’t met, you don’t produce invalid results. What use is that in physics? If you don’t have to store a lot of information locally, it means you can get away using a much lower local spin dimension for the system you describe.

Such small and physically realistic models have as of late been proposed as actual computational devices (called Hamiltonian quantum computers), where a prepared initial state is evolved under such a history state Hamiltonian for a specific time, in contrast to the static property of a history ground state we discussed above. Yet whether or not this is something one could actually build in a lab remains an open question.

Last year, Thomas Vidick invited me to visit Caltech, and I worked with IQIM postdoc Elizabeth Crosson to improve the analysis of the energy penalty that is assigned to any history state that cheats the constraints in the Feynman-Kitaev Hamiltonian. We identified some open problems and also proved limitations on the extent of the energetic penalty that these kinds of Hamiltonians can have. This summer I went back to Caltech to further develop these ideas and make progress towards a complete understanding of such “clock” Hamiltonians, which Elizabeth and I are putting together in a follow-up work that should appear soon.

It is striking how such simple idea can have so profound an implication across fields, and remain relevant, even 30 years after its first proposal.

feynman-clever.png

Feynman concludes his 1986 Foundations of Physics paper with the following words.

At any rate, it seems that the laws of physics present no barrier to reducing the size of computers until bits are the size of atoms, and quantum behavior holds dominant sway.

For my part, I hope that he was right and that history state constructions will play a part in this future.