C H A P T E R 4
e D-Wave Platform
Previous chapters have set out the theoretical framework that underlies the adiabatic model of
computation and the quantum annealing paradigm. e remainder of this book considers practical
issues: this chapter gives an overview of D-Wave¹ quantum annealing systems, and Chapter 5
surveys experimental research to study their properties.
Section 4.1 presents a walk-through of the user experience when solving an optimization
problem. Section 4.2 describes the D-Wave technology stack. Because of space limitations it
is only possible to give a bare sketch of the general properties of these components: for more
technical details see Bunyk et al. , Harris et al. , Harris et al. , and Johnson et al.
. Section 4.3 discusses challenges to success when using this approach to quantum computa-
tion. e chapter concludes with a brief survey of alternative quantum annealing systems built by
other research groups. e discussion throughout is aimed at a non-physicist and non-specialist
in quantum computing.
4.1 THE USER’S VIEW
A D-Wave platform comprises two main components:
1. A processor chip (“the hardware”) that solves Ising Model (IM) problem instances by phys-
ical realization of a quantum annealing algorithm. e chip is mounted in a dilution refrig-
erator and supercooled to a target operating temperature below 20mK, which is necessary
to achieve quantum eﬀects. e chip subsystem is housed within many layers of shielding
to protect against various types of environmental noise.
2. A conventional (Intel) front end server connected to the chip via control lines and an I/O
subsystem. e front end receives instructions from the user and is accessed using a cloud
computing model that supports job queuing and scheduling. e front end sends a problem
instance to the hardware and receives a set of solutions in reply.
We start with a walk-through of the user experience to solve a combinatorial optimization
problem, sketched in Figure 4.1. Note that quantum annealing can be used in many other domains
for which optimization is a subproblem. Chapter 5 surveys the wider ﬁeld of applications.
Suppose the user has an instance I in hand for a given problem P . If I is in native form, it
can be submitted to the front end to be loaded onto the quantum annealing chip for direct solu-
tion in hardware. e hardware can be programmed via a low-level Quantum Machine Instruction
¹D-Wave, D-Wave One, D-Wave Two, Vesuvius, and QSage are trademarks of D-Wave Systems Inc.
44 4. THE D-WAVE PLATFORM
Quantum Annealing HW
Quantum Machine Instruction
1 instance k solutions
Figure 4.1: An overview of the D-Wave Quantum Annealing System. Several paths from input in-
stance to solution can be identiﬁed.
interface, or through a Software Interface containing, for example, C, Python, and Matlab pro-
gramming tools. If the instance is not in native form, the user has two options: translate it into a
native instance (software tools to help with this task are part of the Formulation box in Figure 4.1);
or submit it as is to a hybrid classical-quantum solver called QSage. We consider each option in
Native instances. Recall the Ising Model (IM) from Section 3.2.1: Given a set of real weights
, ﬁnd an assignment to n spin variables S D s
: : : s
, with s
2 f˙1g, to minimize the
is problem is directly implemented in the quantum annealing hardware: that is, the weights
correspond to electronic signals called biases that are applied to the qubits and to the