Skip to content

Basic Concepts

In this section, we will introduce the basic concepts required to understand the Qruise software and what it does.

Digital twins

A Digital Twin is a highly detailed and extremely flexible, physics-based simulation of both the quantum device and the adjacent classical control equipment. It includes all the physical parameters, as well as considering imperfections and sources of distortion, to perfectly replicate the real-life system. This allows users to isolate and quantify the sources of error, providing actionable information on what must be done to improve the performance of the next generation device. “What-if” simulations can also be performed, so that the effects of varying specific parameters can be explored.

For a quantum processing unit, the Digital Twin includes, among other things, initialisation errors (e.g. non-zero temperature), readout confusion matrix, signal generation chain (including AWG discretisation (time, voltage), LOs and mixers (with errors), control line transfer functions, control line crosstalk, non-Markovian noise (amplitude, phase, etc.)) and other user-defined components.

Bring-up

In software and hardware development, bring-up is the process of testing, validating, and debugging a system to ensure it functions correctly. This typically involves configuring hardware and software components, and performing initial tests so that the system performance can be compared to the requirements.

For example, the bring-up of a quantum device requires:

  • Configuring the control electronics to allow communication between them and the quantum device
  • Defining control parameters, such as pulse amplitude, shape and duration
  • Calibrating the quantum device, i.e. fine-adjusting the control parameters to correct for any hardware imperfections not captured in the simulation
  • Characterising the device, i.e. performing specific experiments to obtain key parameters that define the system model
  • Recalibrating the quantum device to correct for drift ​​and maintain the performance and reliability of your system

QruiseOS significantly reduces the time and effort required to bring a quantum device online by automating and streamlining these processes.

Optimal control

The application of optimal control algorithms is essential for realising the true possible performance of a given device. In quantum control theory, these algorithms are instrumental in optimising the time-varying form of pulses or pulse sequences of classical electromagnetic fields that control the dynamics of a quantum system.

Our optimal control algorithms determine the best possible execution of basic operations on the hardware level - such as quantum gates or the initialisation or readout of information - to give the smallest possible error and shortest possible duration. In fact, much shorter operation durations can be determined using optimised modulation of the pulses/pulse sequences than with manually developed pulses, up to the physical limit - the so-called quantum speed limit.

These algorithms can also be used to compensate for spatial and temporal inhomogeneities and to determine pulses for hardware operations for which there is no simple mathematical solution. This latter point is particularly relevant for complex, multi-qubit cases, and when the effect of control electronics/photonics needs to be taken into account.

Open loop optimisation

Open loop optimisation also known as model-based optimal control relies on a system model to design optimal operations for the device. More concretely, it has the following constituent components:

  • a closed or open system model of the quantum device and its control stack
  • a parameterised ansatz used for the control pulse
  • a cost function to quantify the quality of the specific operation (gate, state preparation etc.)
  • an algorithm to optimise the control pulse for this cost function

The Qruise software provides a unified tool that includes all 4 of these components while also allowing users to mix and match these with their own custom-defined subroutines.

Closed loop optimisation and calibration

Model-based or open loop optimisation suffers from the obvious problem of the model not being fully representative of the real hardware. This means operations optimised using the model do not deliver their theoretically optimal performance. To overcome this deficit, we utilise a very wide selection of closed-loop optimisation algorithms to determine the best device parameters for optimal performance. These tools use a feedback loop to adjust the control parameters iteratively by optimising the cost function directly on the hardware. Our unique, ongoing recalibration tools can also continuously work in the background to fine-tune parameters for maintaining maximum fidelities and consistent performance against system drift.

Characterisation and model learning

Traditional characterisation of quantum devices involves fitting measurements (from carefully designed experiments) to simplified analytical models of the system. This technique reaches a bottleneck when trying to characterise large multi-qubit systems or estimate system parameters for which well-understood simplified models do not exist. We expand and augment traditional characterisation with our Model Learning tools (see QruiseML).

Our advanced digital twin technology comprises a fully differentiable physics-based simulation which can be fed experiment results for optimising the model parameters to best match the data. The differentiable simulator allows the use of back-propagation and gradient based optimisation algorithms. The model learning process then iteratively reduces the statistical distance between the output of the digital twin and the real quantum device, thus obtaining a close match between the two. This approach eliminates the need for fitting functions, making it possible to determine multiple parameter values simultaneously to fully-characterise the device. If there is insufficient data to determine all model parameters with a certain degree of certainty, specific experiments can be designed to optimally obtain the necessary information.