You're reading the documentation for a development version. For the latest released version, please have a look at 8.1.

User guide#

This guide illustrates the main features of Qiskit on IQM. You are encouraged to run the demonstrated code snippets and check the output yourself.


At the moment IQM does not provide a quantum computing service open to the general public. Please contact our sales team to set up your access to an IQM quantum computer.

Hello, world!#

Here’s the quickest and easiest way to execute a small computation on an IQM quantum computer and check that things are set up correctly:

  1. Download the example file (Save Page As…)

  2. Install Qiskit on IQM as instructed below (feel free to skip the import statement)

  3. Install Cortex CLI and log in as instructed in the documentation

  4. Set the environment variable as instructed by Cortex CLI after logging in

  5. Run $ python --cortex_server_url – replace the example URL with the correct one

  6. If you’re connecting to a real quantum computer, the output should show almost half of the measurements resulting in ‘00’ and almost half in ‘11’ – if this is the case, things are set up correctly!


The recommended way is to install the distribution package qiskit-iqm directly from the Python Package Index (PyPI):

$ pip install qiskit-iqm

After installation Qiskit on IQM can be imported in your Python code as follows:

import qiskit_iqm

Running a quantum circuit on an IQM quantum computer#

In this section we demonstrate the practicalities of using Qiskit on IQM on an example of constructing and executing a simple quantum circuit on an IQM quantum computer.

Let’s consider the following quantum circuit which prepares and measures a GHZ state:

from qiskit import QuantumCircuit

qc = QuantumCircuit(3, 3)
qc.h(0), 1), 2)

        ┌───┐           ░ ┌─┐
   q_0: ┤ H ├──■────■───░─┤M├──────
        └───┘┌─┴─┐  │   ░ └╥┘┌─┐
   q_1: ─────┤ X ├──┼───░──╫─┤M├───
             └───┘┌─┴─┐ ░  ║ └╥┘┌─┐
   q_2: ──────────┤ X ├─░──╫──╫─┤M├
                  └───┘ ░  ║  ║ └╥┘
meas_0: ═══════════════════╩══╬══╬═
                              ║  ║
meas_1: ══════════════════════╩══╬═
meas_2: ═════════════════════════╩═

To execute this circuit against an IQM quantum computer you need to initialize an appropriate Qiskit backend instance that represents the IQM quantum computer under use, and simply use Qiskit’s execute function as normal:

from qiskit import execute
from qiskit_iqm import IQMProvider

provider = IQMProvider(iqm_server_url)
backend = provider.get_backend()

job = execute(qc, backend, shots=1000)


Note that the code snippet above assumes that you have set the variable iqm_server_url.

You can optionally set IQM backend specific settings as additional keyword arguments to the execute method (which passes the values down to For example, an IQM server uses the best available calibration set automatically; however, if you know an ID (UUID4) of a specific calibration set that you want to use, you can provide it as follows:

job = execute(circuit, backend, shots=1000, calibration_set_id="f7d9642e-b0ca-4f2d-af2a-30195bd7a76d")

Another example is disabling the server-side circuit duration check. If any circuit in a job would take too long to execute compared to the coherence time of the QPU, the server will disqualify the job and not execute any circuits. In some special cases, you may want to disable this as follows:

job = execute(circuit, backend, shots=1000, circuit_duration_check=False)

Disabling the circuit duration check may be limited to certain users or groups, depending on the server settings. In normal use, the circuit duration check should always remain enabled.

If the IQM server you are connecting to requires authentication, you will also have to use Cortex CLI to retrieve and automatically refresh access tokens, then set the IQM_TOKENS_FILE environment variable to use those tokens. See Cortex CLI’s documentation for details. Alternatively, authorize with the IQM_AUTH_SERVER, IQM_AUTH_USERNAME and IQM_AUTH_PASSWORD environment variables or pass them as arguments to the constructor of IQMProvider, however this approach is less secure and considered as deprecated.

The results of a job, that was executed with IQM quantum computer, contain the original request with the qubit mapping that was used in execution. You can check this mapping once execution has finished.

  SingleQubitMapping(logical_name='0', physical_name='QB1'),
  SingleQubitMapping(logical_name='1', physical_name='QB2'),
  SingleQubitMapping(logical_name='2', physical_name='QB3')

The backend instance we created above provides all the standard backend functionality that one expects from a backend in Qiskit. For this example, I am connected to an IQM backend that features a 5-qubit chip with star-like connectivity:

QB2 - QB3 - QB4

Let’s examine its basis gates and the coupling map through the backend instance

print(f'Native operations of the backend: {backend.operation_names}')
print(f'Coupling map of the backend: {backend.coupling_map}')
Native operations of the backend: ['r', 'cz', 'measure']
Coupling map of the backend: [[0, 2], [1, 2], [2, 3], [2, 4]]

At IQM we identify qubits by their names, e.g. ‘QB1’, ‘QB2’, etc. as demonstrated above. In Qiskit, qubits are identified by their indices in the quantum register, as you can see from the printed coupling map above. Most of the time you do not need to deal with IQM-style qubit names when using Qiskit, however when you need, the methods IQMBackend.qubit_name_to_index() and IQMBackend.index_to_qubit_name() can become handy.

Now we can study how the circuit gets transpiled:

from qiskit.compiler import transpile

qc_transpiled = transpile(qc, backend=backend, layout_method='sabre', optimization_level=3)

global phase: π/2
               ┌────────────┐┌────────┐                 ┌────────────┐┌────────┐ ░       ┌─┐
      q_2 -> 0 ┤ R(π/2,π/2) ├┤ R(π,0) ├─────────■───────┤ R(π/2,π/2) ├┤ R(π,0) ├─░───────┤M├
               └────────────┘└────────┘         │       └────────────┘└────────┘ ░       └╥┘
ancilla_0 -> 1 ─────────────────────────────────┼─────────────────────────────────────────╫─
               ┌────────────┐┌────────┐         │                                ░ ┌─┐    ║
      q_0 -> 2 ┤ R(π/2,π/2) ├┤ R(π,0) ├─■───────■────────────────────────────────░─┤M├────╫─
               └────────────┘└────────┘ │                                        ░ └╥┘    ║
ancilla_1 -> 3 ─────────────────────────┼───────────────────────────────────────────╫─────╫─
               ┌────────────┐┌────────┐ │ ┌────────────┐  ┌────────┐             ░  ║ ┌─┐ ║
      q_1 -> 4 ┤ R(π/2,π/2) ├┤ R(π,0) ├─■─┤ R(π/2,π/2) ├──┤ R(π,0) ├─────────────░──╫─┤M├─╫─
               └────────────┘└────────┘   └────────────┘  └────────┘             ░  ║ └╥┘ ║
          c_0: ═════════════════════════════════════════════════════════════════════╬══╬══╬═
                                                                                    ║  ║  ║
          c_1: ═════════════════════════════════════════════════════════════════════╬══╬══╬═
                                                                                    ║  ║  ║
          c_2: ═════════════════════════════════════════════════════════════════════╬══╬══╬═
                                                                                    ║  ║  ║
       meas_0: ═════════════════════════════════════════════════════════════════════╩══╬══╬═
                                                                                       ║  ║
       meas_1: ════════════════════════════════════════════════════════════════════════╩══╬═
       meas_2: ═══════════════════════════════════════════════════════════════════════════╩═

Simulating the execution of a transpiled circuit locally#

The execution of circuits can be simulated locally, with a noise model to mimic the real hardware as much as possible. To this end, Qiskit on IQM provides the class IQMFakeBackend that can be instantiated with properties of a certain QPU, or subclasses of it such as IQMFakeAdonis that represent certain quantum architectures with pre-populated properties and noise model.

from qiskit import execute, QuantumCircuit
from qiskit_iqm import IQMFakeAdonis

circuit = QuantumCircuit(2)
circuit.h(0), 1)

backend = IQMFakeAdonis()
job = execute(circuit, backend, shots=1000)

Above, we use an IQMFakeAdonis instance to run a noisy simulation of circuit on a simulated 5-qubit Adonis chip. If you want to customize the noise model instead of using the default one provided by IQMFakeAdonis, you can create a copy of the fake Adonis instance with updated error profile:

error_profile = backend.error_profile
error_profile.t1s['QB2'] = 30000.0  # Change T1 time of QB2 as example
custom_fake_backend = backend.copy_with_error_profile(error_profile)

Running a quantum circuit on a facade backend#

Circuits can be executed against a mock environment: an IQM server that has no real quantum computer hardware. Results from such executions are random bits. This may be useful when developing and testing software integrations.

Qiskit on IQM contains IQMFacadeBackend, which allows to combine the mock remote execution with a local noisy quantum circuit simulation. This way you can both validate your integration as well as get an idea of the expected circuit execution results.

To run a circuit this way, use the facade_adonis backend retrieved from the provider. Note that the provider must be initialized with the URL of a quantum computer with the equivalent architecture (i.e. names of qubits, their connectivity, and the native gateset should match the 5-qubit Adonis architecture).

from qiskit import execute, QuantumCircuit
from qiskit_iqm import IQMProvider

circuit = QuantumCircuit(2)
circuit.h(0), 1)

provider = IQMProvider("")
backend = provider.get_backend('facade_adonis')
job = execute(circuit, backend, shots=1000)


When a classical register is added to the circuit, Qiskit fills it with classical bits of value 0 by default. If the register is not used later, and the circuit is submitted to the IQM server, the results will not contain those 0-filled bits. To make sure the facade backend returns results in the same format as a real IQM server, checks for the presence of unused classical registers, and fails with an error if there are any.

More advanced examples#

In this section we demonstrate some less simple examples of using Qiskit on IQM and its interoperability with various tools available in Qiskit.

It is possible to run multiple circuits at once, as a batch. In many scenarios this is more time efficient than running the circuits one by one. For batch execution there are some restriction that we shall keep in mind, namely all circuits have to measure the same qubits, and all circuits will be executed for the same number of shots. For starters, let’s construct two circuits preparing and measuring different Bell states:

qc_1 = QuantumCircuit(2)
qc_1.h(0), 1)

qc_2 = QuantumCircuit(2)
qc_2.x(1), 1)

Now, we can execute them together in a batch:

job = execute([qc_1, qc_2], backend, initial_layout=[0, 2], shots=1000)

The batch execution functionality can be used to run a parameterized circuit for various concrete values of parameters:

import numpy as np
from qiskit.circuit import Parameter

qc = QuantumCircuit(2)
theta = Parameter('theta')
theta_range = np.linspace(0, 2*np.pi, 3)

qc.h(0), 1)
qc.rz(theta, [0, 1]), 1)

qc_transpiled = transpile(qc, backend=backend, layout_method='sabre', optimization_level=3)

circuits = [qc_transpiled.bind_parameters({theta: n}) for n in theta_range]
job = execute(circuits, backend, shots=1000, optimization_level=0)


Note that it is important to transpile the parameterized circuit before binding the values to ensure a consistent qubit measurements across circuits in the batch.

How to develop and contribute#

Qiskit on IQM is an open source Python project. You can contribute by creating GitHub issues to report bugs or request new features, or by opening a pull request to submit your own improvements to the codebase.

To start developing the project, clone the GitHub repository and install it in editable mode with all the extras:

$ git clone
$ cd qiskit-on-iqm
$ pip install -e ".[dev,docs,testing]"

To be able to build the docs graphviz has to be installed. Then to build and view the docs run:

$ tox -e docs
$ firefox build/sphinx/html/index.html

Run the tests:

$ tox

Tagging and releasing#

After implementing changes to Qiskit on IQM one usually wants to release a new version. This means that after the changes are merged to the main branch

  1. the repository should have an updated CHANGELOG.rst with information about the new changes,

  2. the latest commit should be tagged with the new version number,

  3. and a release should be created based on that tag.

The last two steps are automated, so one needs to worry only about properly updating the CHANGELOG. It should be done along with the pull request which is introducing the main changes. The new version must be added on top of all existing versions and the title must be “Version MAJOR.MINOR”, where MAJOR.MINOR represents the new version number. Please take a look at already existing versions and format the rest of your new CHANGELOG section similarly. Once the pull request is merged into main, a new tag and a release will be created automatically based on the latest version definition in the CHANGELOG.