Israeli quantum startup Quantum Pulse Ventures has unwrapped a manufacturing line tweak that it says promises a ten times boost in quantum computer performance. The company unwrapped its QP2.0 platform at a conference in Edinburgh this week, and said its “composite pulse approach improves operational fidelity and robustness against fabrication variability.”

The platform centers on a redesigned universal directional coupler that it claims delivers “an order-of-magnitude improvement in operational fidelity.” The company likens its directional coupler to a “transistor for optical computing, allowing two qubits to interact in a photonic circuit.

Variations and inaccuracies in manufacturing of photonic circuits cause a cascade of problems that make errors more likely. This is one of the reasons that quantum computer designs must allow for a vast number of physical qubits and extensive error correction in order to deliver one logical qubit, meaning quantum computers, for now, are massive both in size and cost.

Quantum Pulse’s design uses composite wave guides, rather than traditional one segment uniform wave guides, to reduce physical errors and noise in the circuits of these gates. The design is also more resistant to fabrication errors, it claims.

The company said this can be used in a broad class of photonic integrated circuits.  And it can be adopted across a range of current silicon photonics, silicon nitride, thin-film lithium niobate, and related integrated photonics manufacturing processes. Here are the components of its stack:

Ofer Shapiro, co-founder and chief executive officer the Israeli company, said the coupler allowed light to jump from one wave guide to another, similar to a transistor in a traditional CMOS circuit.

The wave guides themselves are “very, very small, they are less than one micrometer in size, and the distance between them has to be accurate in a sub-nanometer level for this operation to be accurate.”

This in turn “becomes a huge manufacturing issue, and that's why many of those solutions are just at the cusp of functionality.”

Sticking with the transistor analogy, Shapiro said: “If someone told you I figured out a way to do microelectronics with one tenth of the energy, it would be huge. It's the same thing for photonics.”

The technology is also applicable to matrix multiply units and quantum routers, the company explained, and would deliver a four times speed boost over existing technologies. Here is how you use the tech to make a matrix multiply unit:

And here is how you use it to make an optical router:

Shapiro was at pains to point out that this doesn’t mean extensive rearchitecting of processor designs or fabrication processes.

“We have a recipe for a better transistor that is more accurate and produces better results, and that's not a different manufacturing process. You don't have to change anything in the fab, you just give it a different image to lithograph on to the chip.”

As for how the company will monetize the innovation, Shapiro said: “It's an IP play. If you want to use that IP in order to introduce it into your own chips, you have to pay royalties. It's a very, very simple model, and also very compelling.”

Given that some optical quantum computers are the size of a stadium, with costs to match, this means a lot. One fault tolerant qubit typically requires 10,000 physical qubit averaging to one fault tolerant qubit, so, Shapiro argued, “If they're ten times better, you only need a thousand and this thing means that everything shrinks in proportion.” Or, put another way, “for the same budget you get ten times more.”

This massively improves the ability to produce those elusive fault-tolerant qubits he claimed. “The first person that would have a commercial 100 fault tolerant qubits would win the race for first practical quantum computer. That's considered the threshold where actually starts to make sense.”

Implementing Quantum Pulse’s design is not a big deal, according to Shapiro. “It requires us to work with that partner to get some information about their manufacturing platform of choice, give them the design for this, let's say the coupler or the polarized controller, whatever the component, probably they need both in most cases. Give them the design for that, they drop it in, and it goes.” He said this should take a couple of manufacturing cycles, or six to nine months.

While Shapiro was adamant that the tech can speed the delivery of practicable quantum computers, the uncertainty, as always, is when that will be. “They're certainly telegraphing to the market that they are almost there, right?

“But the question is, what's almost? It's like we were on autonomous driving for a long time. AI was almost there. I think I remember reading about neural networks 30 years ago, maybe more than that, and, you know, it happened when it happened, and of course it changed the world.”

“I think in five to ten years you will have processors that run electrons and processors that run photons.”

Either way, he said, we are getting to the physical limits of “what you could squeeze from electrons.” This is playing out in both memory – which is seeing soaring prices – and in processing.

“People are looking at architecture of the processor rather than a better transistor because they're getting so small in the process that they are limiting, getting close to the physical barriers, right?”

“There's a limit of how much you can get in a single computer, and then you have to scale it, and the way to scale it is by quantum router.”

By contrast, when it comes to quantum computers, there are several options on how to implement the qubit, he said.

“Some of them are using semiconductors, some of them are using trapped ions, some of them are using optics. It may be that for different applications, different technologies win.”

But while there are a variety of technologies or particles to build a qubit, “when you talk about connecting them, there is no replacement for photons. It has to be based on light, right.”

“This like the race to the moon, almost - It just happens in the private sector in most part, and there's no points for second place.” He pauses for a moment, then adds “I think there may be some for second, but I don't know about the third place.”