Performing computations with neutral atoms comes down to realizing, controlling, manipulating and measuring neutral-atom qubits. Here, we dive into some details.
Atoms consist of a nucleus and their electron cloud. At QuEra, we use Rubidium atoms. When the positive and negative charges of an atom balance each other, the atom is said to be neutral, just like most atoms are found in nature. By shining a laser at the atom, we can increase the energy of the atom, and excited it. Each atom has a wide range of potential excited states, each at different energy levels. This offers a great platform for storing and processing quantum information. Any choice of two such levels can be named ‘0’ and ‘1’ to form a qubit.
Using lasers as optical tweezers, we trap individual atoms in their place. The lasers suppresses the atomic movement, effectively cooling atoms down to nearly absolute zero temperature. At these temperatures, the individual discrete energy levels of the atoms can be resolved and manipulated. Some of them leading to huge coherence times exceeding one second.
When atoms are excited to high-energy states, called Rydberg states, their electron clouds are puffed up to relatively enormous scale. Such atoms grow by more than a thousand times their usual size. In this Rydberg state, different atoms can interact over long distances, enabling quantum information to be transferred between them. This enables entanglement, one of the key ingredients for quantum information manipulation.
The mechanism by which Rydberg atoms interact is called the ‘van der Waals’ interaction, originating from the strong dipole moments of the puffed atoms. This interaction decays with the 6th power of the interatomic distance, meaning that atoms interact very strongly with each other only when they very close. In fact, that interaction can be so strong that it can lead to the so-called "Rydberg blockade" effect, under which no two neighboring atoms can be excited at a given time. Through this mechanism, conditional quantum logic and two-qubit gates can be implemented.
Remarkably, the size of Rydberg atoms can include several proximate qubits, enabling all of them to interact with each other. While most quantum computers can only implement native 1-qubit and 2-qubit gates, the Rydberg blockade mechanism enables the creation of native multi-qubit gates.
Such gates, like the Toffoli gate (see image on the left), play an important role in many quantum algorithms. Encoding these multi-qubit gates natively can substantially reduce the algorithm’s circuit depth, thus also strongly mitigating errors. A good example is the constant-depth implementation of Shor’s algorithm.
All of this happens on a miniature scale: one can pack thousands of laser-trapped atoms in a square millimeter. And with integrated photonic chips, precise control of multiple qubits is achieved by just a few lasers.
As lasers can move freely through space, neutral atoms can also be arranged in nearly-arbitrary configurations. This allows adjusting and adapting the qubit connectivity to the specific needs of each problem. It also leads to significantly shorter development cycles, as new applications can make use of new configurations without reassembly of the hardware.
With a proper choice of atomic levels for qubit encoding, atoms can even be coherently moved around during a calculation. This is a key advantage of neutral-atom technology. It enables an efficient memory bus service, allowing all-to-all connectivity of the qubits at large scale. This is particularly beneficial in error correction, as it offers new opportunities for choosing gates and error-correcting codes. One example is our first and only realization of Kitaev's paradigmatic toric code with periodic boundary conditions. Furthermore, qubit shuttling enables zoned architectures for memory and processing, further improving the ratio of control lines to number of qubits.
Together, these features make neutral-atom quantum computers the architecture of choice for utility and scalability.
Neutral atoms uniquely support different computation modes: digital and analog, each with their own advantages. Users of QuEra's computer will be able to choose the best mode for the particular problem they want to solve.
The digital gate-based mode decomposes a complex operation into a few elementary steps (gates) that operate on one or two qubits at a time. The gates move qubits between their two computational states. Gate sequences change the computational state of the entire system from one state into another. When measured, the final state becomes a bit string, capturing the outcome of the computation.
Gate-based operation modes provides an elegant recipe to enable universal functionality and programmability, with only a small set of elementary operations.
At the physical level, digital quantum operations are continuous state transitions expressed by a Hamiltonian: the function that describes the forces that make a physical system change over time. If one can control a Hamiltonian well-enough, the gate-based mode can be bypassed in favor of an analog computation mode. The analog mode is not as universal as the gate-based mode, but when it is possible, it leads us directly to answers without decomposing algorithms into elementary steps, avoiding many of the noise and coherence issues that are common in digital computation modes.
In the digital gate-based mode, errors can be introduced at each step, and accumulate over the course of a computation. In contrast, analog mode does not suffer from accumulation of faulty gates and is therefore much less susceptible to errors, making it very suitable for the current stage of quantum computing maturity.
Combined with simplified control requirements, the analog quantum mode enables efficient manipulation of many more qubits with significantly fewer control signals.