tqec architecture#

This document is mainly targeted at a developer wanting to contribute to the tqec library. It provides a high-level overview of the different pieces that compose the tqec library and how they interact with each other.

Warning

This is a rapidly evolving project — codebase might change. If you encounter any inconsistencies, please open an issue.

This is done by presenting each of the sub-modules and their main classes / methods in the order in which they are used internally to go from a topological computation to an instance of stim.Circuit that can be executed on compatible hardware.

The diagram below provides a high-level overview of the dependencies between tqec submodules. It is not 100% accurate, as for example the tqec.circuit submodule defines a few core data-structures that are used in tqec.plaquette, but this is the order in which the sub-modules will be covered here.

        ---
config:
  darkMode: 'true'
  theme: base

---
%%{
init: {
    'theme': 'base',
    'themeVariables': {
    'primaryColor': '#AFEEEE',
    'lineColor': '#6495ED'
    }
}
}%%

graph
    A[tqec.interop] --> B[tqec.computation]
    B --> E[tqec.compile]
    C[tqec.templates] --> E
    D[tqec.plaquette] --> E
    E --> F[tqec.circuit]
    F --> G[tqec.utils.noise_models]
    G --> H[tqec.simulation]

A user starts with a computation and an observable. The input computation can be provided as a BlockGraph structure while an observable is represented as a correlation surface. The end goal of tqec is to generate a stim.Circuit such that the input is transformed into a topologically quantum error corrected quantum computation protected by a surface code.

interop#

Allows a user to provide a .dae file or a pyzx graph as an input to tqec.

Standard color representations as described in Terminology are predefined in TQECColor.

Some conventional implementations of computations in .dae format are available in the Computation Gallery.

computation#

Defines data structures for the high-level BlockGraph representations of a fault-tolerant computation protected by the surface code.

The 3D structures discussed in detail in Terminology are defined in this module.

  • A CorrelationSurface represents a set of parity measurements between the input and output logical operators.

  • A Cube is a fundamental building block constituting of a block of quantum operations that occupy a specific spacetime volume. Quantum information encoded in the logical qubits can be preserved or manipulated by these blocks.

  • A Pipe is a block that connects Cube objects in a BlockGraph but does not occupy spacetime volume on its own. The exception here are temporal hadamard pipes that have a volume when compiled using the fixed bulk convention.

  • PipeKind helps determine the kind of a pipe in a BlockGraph based on the wall bases at the head of the pipe in addition to a Hadamard transition.

  • Port depicts the open ports in a BlockGraph.

  • A YHalfCube represents Y-basis initialization and measurements.

  • ZXCube defines cubes with only X or Z basis boundaries.

compile#

Responsible for translations between internal representations.

BlockGraph obtained from the functionality in interop is further translated into a TopologicalComputationGraph represented by Block instances.

        ---
config:
  darkMode: 'true'
  theme: base

---
%%{
init: {
    'theme': 'base',
    'themeVariables': {
    'primaryColor': '#AFEEEE',
    'lineColor': '#6495ED'
    }
}
}%%

graph
    A[BlockGraph] --> B[Topological <br> Computation Graph] --> C[ScheduledCircuit]--> D[stim.Circuit]

Multiple block builder protocols defined in tqec.compile.spec.library will take the high-level structure of a block to templates and plaquettes, that can in turn be used to generate a fully annotated stim.Circuit.

templates#

Generates an array of numbers representing a 2-dimensional, scalable, arrangement of plaquettes. This allows us to describe a circuit that can scale the desired code distance.

plaquette#

A plaquette is the representation of a local quantum circuit. A surface code patch implements one layer in time or one round of the surface code. Same as templates, this module allows us to define scalable quantum circuits.

circuit#

Implementation of ScheduledCircuit, a quantum circuit representation in tqec, where each and every gate of a regular quantum circuit is associated with the time of execution. This module maps from the numbered templates to some plaquettes that implement small local circuits to measure a stabilizer as a ScheduledCircuit instance.

NoiseModel#

Note

The code for this module was modified from the code for [1].

This module implements the following noise models for Stim simulations:

  1. Superconducting Inspired Circuit Error Model (SI1000): A modified version of the noise model introduced in [2] which represents the noise on Google’s superconducting quantum chip.

    In si1000():

    • Depolarizing noise on measured qubits from the noise modeil in [2] has been removed because tqec measurements are immediately followed by resets.

    • The measurement result is probabilistically flipped instead of the input qubit.

  2. Uniform Depolarizing Noise: Single qubit depolarizing noise is uniformly applied to both single qubit and two qubit Clifford gates.

    In uniform_depolarizing():

    • The result of dissipative gates is probabilistically bit or phase flipped.

    • Result of non-demolition measurements is flipped instead of the input qubit.

simulation#

Utilities related to quantum circuit simulations through sinter, a Python submodule in stim. Plotting functions are in this module too.

References#

[1]

Craig Gidney. Inplace access to the surface code y basis. Quantum, 8:1310, April 2024. URL: http://dx.doi.org/10.22331/q-2024-04-08-1310, doi:10.22331/q-2024-04-08-1310.

[2] (1,2)

Craig Gidney, Michael Newman, Austin Fowler, and Michael Broughton. A fault-tolerant honeycomb memory. Quantum, 5:605, December 2021. URL: http://dx.doi.org/10.22331/q-2021-12-20-605, doi:10.22331/q-2021-12-20-605.