Читать книгу Principles of Superconducting Quantum Computers - Daniel D. Stancil - Страница 30

At this point, you may be asking: what kind of physical system exhibits the behavior that we can exploit for quantum computing? Any two-state quantum mechanical system can represent a qubit, and there are several possibilities, such as the spin of an electron, the polarization of a photon, or the energy level of an electron in a charged ion. Any of these systems can be used to build a quantum computer, but there are tradeoffs regarding how the qubits can be manufactured and controlled, and how they interact with one another.

In this book, we concentrate on one specific technology for creating qubits and quantum computing systems: superconducting circuits. Unlike many competing technologies, superconducting qubits are macroscopic in size and are based on well-known nanofabrication technologies. They represent the current technology of choice for several companies building quantum computer systems, including IBM, Google, and Rigetti.

A large part of this book, Chapters 2–8, is devoted to a detailed explanation of these devices and how to control them to carry out the fundamental operations of a quantum computer, described above. In this section, before diving into the details, we provide a high-level overview of a superconducting quantum computer.

As we will see in Chapter 2, we will need to couple the qubit to a signal whose frequency depends on the energy difference between the |0⟩ and |1⟩ states, i.e., the ground and excited states. In superconducting quantum computers, this energy difference corresponds to a microwave frequency near 5 GHz. Consequently, we must design a microwave system to control and measure superconducting qubit states.

The general features of the microwave system to control and readout superconducting qubits are shown in Figure 1.12. A key feature is that the qubits must be held at a very low temperature, near absolute zero. Consequently the qubits must be located in a cryogenic refrigerator. To understand why this is necessary, we want to make sure that if we put the qubit in the ground state, it stays in the ground state. In other words, we want to make sure that it is unable to absorb enough energy incidentally from its environment to make a transition. A circuit in equilibrium at temperature T can emit and absorb photons with the energy kT, where k is Boltzmann’s constant. The energy of a photon is ℏω, where ω is the frequency and ℏ is Planck’s constant divided by 2π. We want to make sure that kT<<ℏω. For ω/(2π)=5 GHz, this means that T<<0.24 K. In state-of-the-art dilution refrigerators, the temperature of the qubits can be held at 10–15 mK. In this range of temperatures, thermal excitation of 5 GHz qubits can be neglected.

Figure 1.12 System diagram for a superconducting quantum computer.

Referring again to Figure 1.12, the quantum processor (QP) containing the qubits is located at the bottom left of the refrigerator. In addition to being kept very close to absolute zero temperature, the quantum processor is also sensitive to stray magnetic fields, so it is further placed inside a magnetic shield within the coldest stage of the refrigerator.

The round component just above the quantum processor is a circulator. This is a non-reciprocal microwave component in which energy can only propagate between ports in the direction of the circulating arrow. By non-reciprocal, we mean that the behavior of the component is different if you interchange the input with the output. For example, the RF signal from the control electronics enters the circulator at the top port; the energy “circulates” around to the port to which the quantum processor is connected. Any reflected energy from the quantum processor, e.g., containing information about the state of a qubit, then re-enters the bottom port of the circulator. However, since the circulator is non-reciprocal, instead of returning to the input port on the top side of the circulator, it flows instead to the port on the right and is conveyed to the circulator located in the center. We will return to this center circulator in a moment, but let us first consider the chain of coaxial cables and attenuators conveying the control signal to the first circulator.

If we simply used a copper coaxial cable to carry the signal from the room-temperature electronics into the refrigerator, we would have at least two significant problems. First, copper is a good conductor of heat as well as electricity, so the copper cable would convey heat from the upper stages to the lower stages, making it very difficult to reach the temperatures required at the lower stages. To address this, coaxial cables made of an alloy of copper and nickel are used instead. This alloy has very low thermal conductivity to assist with thermally isolating the stages, while having an acceptable electrical conductivity.

The second problem is that a cable coming straight from the outside environment would convey significant noise from the environment into the refrigerator. To combat this, attenuators are placed in the lower stages. These attenuators reduce noise power from the upper stages, but introduce their own noise proportional to their equilibrium temperature. Consequently at the lowest stage, the thermal noise is minimized by the very low temperature of the attenuator. Of course these attenuators also reduce the amplitude of the control signal, so we must make sure that the signal level produced by the signal source is strong enough to produce a satisfactory control signal at the quantum processor.

Returning to the signal reflected from the quantum processor, upon entering the center circulator, the energy is transferred to the bottom port and delivered to a Josephson Junction Parametric Amplifier (JPA). This is a quantum-limited amplifier, meaning that the noise it introduces to the circuit is close to the fundamental minimum allowed by quantum theory. The JPA works in reflection mode, so the amplified reflected signal is returned to the circulator and transferred to the final circulator at the bottom right.

The circulator at the bottom right is operated as an isolator. Power applied to the input (left port) is delivered to the output (top port), but any power reflected from impedance mismatches further down line, e.g., at the input of the HEMT amplifier, will be delivered to the matched load attached to the bottom port. In this way, the very sensitive quantum processor and quantum limited amplifier are isolated from reflected power or noise from the upper stages.

The measured signal at this point is very weak, and we definitely would not want to use a stack of attenuators on the output line! Instead, a coaxial cable made of an alloy of Niobium and Titanium (NbTi) is used to convey the signal to the 3 K stage. NbTi is a Type II superconductor with a transition temperature of about 10 K, so it provides an extremely low loss path to the 3 K stage. At the 3 K stage, there is a more conventional low-noise amplifier (LNA), but made using a high electron mobility transistor (HEMT)—a type of transistor that is known for producing very low noise. The signal from the output of the HEMT LNA is further amplified at room temperature before delivery to the signal processing electronics.

Returning to the control-signal electronics, any band-limited signal centered on the frequency ω can be represented by sine and cosine components:

(1.70)

where I(t) and Q(t) are “in-phase” and “quadrature” functions whose time variations are slow compared with the period T=2π/ω. Digital samples of I and Q are converted to band-limited analog signals by digital-to-analog converters followed in general by low-pass filters to eliminate high-frequency components resulting from aliasing. The circular components in Figure 1.12 containing “×” are called mixers, and produce an output that is simply the product of the two input signals. For example, the bottom-most mixer produces Q(t)sin⁡ωt on its output. Similarly, the output of the next-to-bottom mixer is I(t)cos⁡ωt. The outputs of the two mixers are applied to a power combiner, shown as the square component with two terminals on the left, and one on the right. The output of the combiner is the sum of the two input signals, creating the desired general signal of the form given by Eq. (1.70). This circuit is referred to as an IQ modulator.

There are different ways of combining RF signals, but the particular one shown is known as a Wilkinson combiner, or Wilkinson divider. Note that this component is reciprocal, so it can be used either as a combiner or a divider. An attractive feature of the Wilkinson circuit when used as a divider is that if all of the ports are matched, it does not introduce any loss.

The output signal from the quantum processor is applied to a signal processing circuit at the top center of Figure 1.12 that is very similar to the IQ modulator used to generate the control signal. The signal is first split into equal parts using a Wilkinson divider, and the divided signals are multiplied by either cos⁡ωt or sin⁡ωt before being applied to low-pass filters. The effect of the low-pass filters is to integrate the applied signal, so since the cross term proportional to cos⁡ωtsin⁡ωt averages to zero, the filter outputs are proportional to either I(t) or Q(t). These signals are then digitized and analyzed by the classical computer.

With this high-level description as motivation, we are now ready to discuss in detail the principles that underlie the hardware and software of superconducting quantum computing.

 Chapters 2 and 3 explore the quantum physics that determine the behavior of qubits and gates. This gives us the tools to understand the fundamentals of quantum states and how they can be manipulated.

 Chapters 4–8 show how qubits are constructed from superconducting devices, how they are coupled to each other with microwave transmission lines, and how they are controlled and measured by external systems.

 Chapters 9 and 10 discuss how imperfections in the systems we build affect the quantum information we are trying to process. In the near term, we need mechanisms to characterize and compensate for errors, while the long-term hope is that we will have a sufficient number of high-quality qubits to correct errors dynamically and sustain long-running, fault-tolerant quantum computations.

 Finally, Chapters 11 and 12 describe the computations that can be accomplished using qubits and gates, and the potential for applications beyond the capabilities of classical computers.

Our hope is to lay a firm foundation for those new to the quantum computing field, whether students or practicing engineers, as a first step toward tackling the many research and engineering challenges that are needed to make large-scale quantum computers a reality.