Even with everything happening in the Middle East right now, even with (relatedly) everything happening in my own family (my wife and son sheltering in Tel Aviv as Iranian missiles rained down), even with all the rather ill-timed travel I’ve found myself doing as these events unfolded (Ecuador and the Galapagos and now STOC’2025 in Prague) … there’s been another thing, a huge one, weighing on my soul.
Ray Laflamme played a major role in launching the whole field of quantum computing and information, and also a major role in launching my own career. The world has lost him too soon. I’ve lost him too soon.
After growing up in Quebec—I still hear his French-Canadian accent, constantly on the verge of laughter, as I’m writing this—Ray went into physics and became a PhD student of Stephen Hawking. No, not a different Stephen Hawking. If you’ve read or watched anything by or about Hawking, including A Brief History of Time, you might remember the story where Hawking believed for a while that time would reverse itself as the universe contracted in a Big Crunch, with omelettes unscrambling themselves, old people turning into children, etc. etc., but then two graduate students persuaded him that that was totally wrong, and entropy would continue to increase like normal. Anyway, Ray was one of those students (Don Page was the other). I’d always meant to ask Ray to explain what argument changed Hawking’s mind, since the idea of entropy decreasing during contraction just seemed obviously wrong to me! Only today, while writing this post, did I find a 1993 paper by Hawking, Laflamme, and Lyons that explains the matter perfectly clearly, including three fallacious intuitions that Hawking had previously held. (Even though, as they comment, “the anatomy of error is not ruled by logic.”)
Anyway, in the mid-1990s, starting at Los Alamos National Lab and continuing at the University of Waterloo, Ray became a pioneer of the then-new field of quantum computing and information. In 1997, he was a coauthor of one of the seminal original papers that proved the possibility of fault-tolerant quantum computation with a constant error rate, what we now call the Threshold Theorem (Aharonov and Ben-Or had such a result independently). He made lots of other key early contributions to the theory of quantum error-correcting codes and fault-tolerance.
When it comes to Ray’s scientific achievements after his cosmology work with Hawking and after quantum fault-tolerance—well, there are many, but let me talk about two. Perhaps the biggest is the KLM (Knill-Laflamme-Milburn) Theorem. It would be fair to say that KLM started the entire field of optical or photonic quantum computation, as it’s existed in the 21st century. In one sentence, what KLM showed is that it’s possible to build a universal quantum computer using only
- identical single-photon states,
- a network of “linear-optical elements” (that is, beamsplitters and phaseshifters) that the photons travel through, and
- feedforward measurements—that is, measurements of an optical mode that tell you how many photons are there, in such a way that you can condition (using a classical computer) which optical elements to apply next on the outcome of the measurement.
All of a sudden, there was a viable path to building a quantum computer out of photons, where you wouldn’t need to get pairs of photons to interact with each other, which had previously been the central sticking point. The key insight was that feedforward measurements, combined with the statistical properties of identical bosons (what the photons are), are enough to simulate the effect of two-photon interactions.
Have you heard of PsiQuantum, the startup in Palo Alto with a $6 billion valuation and hundreds of employees that’s right now trying to build an optical quantum computer with a million qubits? Or Xanadu, its competitor in Toronto? These, in some sense, are companies that grew out of a theorem: specifically the KLM Theorem.
For me, though, the significance of KLM goes beyond the practical. In 2011, I used the KLM Theorem, together with the fact (known since the 1950s) that photonic amplitudes are the permanents of matrices, to give a new proof of Leslie Valiant’s celebrated 1979 theorem that calculating the permanent is a #P-complete problem. Thus, as I pointed out in a talk two years ago at Ray’s COVID-delayed 60th birthday conference, entitled Ray Laflamme, Complexity Theorist (?!), KLM had said something new about computational complexity, without any intention of doing so. More generally, KLM was crucial backdrop to my and Alex Arkhipov’s later work on BosonSampling, where we gave strong evidence that some classical computational hardness—albeit probably not universal quantum computation—remains in linear optics, even if one gets rid of KLM’s feedforward measurements.
(Incidentally, I gave my talk at Ray’s birthday conference by Zoom, as I had a conflicting engagement. I’m now sad about that: had I known that that would’ve been my last chance to see Ray, I would’ve cancelled any other plans.)
The second achievement of Ray’s that I wanted to mention was his 1998 creation, again with his frequent collaborator Manny Knill, of the One Clean Qubit or “DQC1” model of quantum computation. In this model, you get to apply an arbitrary sequence of 2-qubit unitary gates, followed by measurements at the end, just like in standard quantum computing—but the catch is that the initial state consists of just a single qubit in the state |0⟩, and all other qubits in the maximally mixed state. If all qubits started in the maximally mixed state, then nothing would ever happen, because the maximally mixed state is left invariant by all unitary transformations. So it would stand to reason that, if all but one of the qubits start out maximally mixed, then almost nothing happens. The big surprise is that this is wrong. Instead you get a model that, while probably not universal for quantum computation, can do a variety of things in polynomial time that we don’t know how to do classically, including estimating the traces of exponentially large unitary matrices and the Jones polynomials of trace closures of braids (indeed, both of these problems turn out to be DQC1-complete). The discovery of DQC1 was one of the first indications that there’s substructure within BQP. Since then, the DQC1 model has turned up again and again in seemingly unrelated investigations in quantum complexity theory—way more than you’d have any right to expect a priori.
Beyond his direct contributions to quantum information, Ray will be remembered as one of the great institution-builders of our field. He directed the Institute for Quantum Computing (IQC) at the University of Waterloo in Canada, from its founding in 2002 until he finally stepped down in 2017. This includes the years 2005-2007, when I was a postdoc at IQC—two of the most pivotal years of my life, when I first drove a car and went out on dates (neither of which I do any longer, for different reasons…), when I started this blog, when I worked on quantum money and learnability of quantum states and much more, and when I taught the course that turned into my book Quantum Computing Since Democritus. I fondly remember Ray, as my “boss,” showing me every possible kindness. He even personally attended the Quantum Computing Since Democritus lectures, which is why he appears as a character in the book.
As if that wasn’t enough, Ray also directed the quantum information program of the Canadian Institute for Advanced Research (CIFAR). If you ever wondered why Canada, as a nation, has punched so far above its weight in quantum computing and information for the past quarter-century—Ray Laflamme is part of the answer.
At the same time, if you imagine the stereotypical blankfaced university administrator, who thinks and talks only in generalities and platitudes (“how can we establish public-private partnerships to build a 21st-century quantum workforce?”) … well, Ray was whatever is the diametric opposite of that. Despite all his responsibilities, Ray never stopped being a mensch, a friend, an intellectually curious scientist, a truth-teller, and a jokester. Whenever he and I talked, probably at least a third of the conversation was raucous laughter.
I knew that Ray had spent many years battling cancer. I naïvely thought he was winning, or had won. But as so often with cancer, it looks like the victory was only temporary. I miss him already. He was a ray of light in the world—a ray that sparkles, illuminates, and as we now know, even has the latent power of universal quantum computation.