| author | description | ms.author | ms.date | ms.service | ms.subservice | ms.topic | no-loc | title | uid | |||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
SoniaLopezBravo |
This document provides the technical details of the IonQ quantum computing provider |
sonialopez |
10/24/2024 |
azure-quantum |
computing |
concept-article |
|
IonQ quantum computing provider |
microsoft.quantum.providers.ionq |
Important
IonQ has retired their Harmony-2 quantum computer effective September 1, 2024.
[!INCLUDE Azure Quantum credits banner]
IonQ’s quantum computers perform calculations by manipulating the hyperfine energy states of Ytterbium ions with lasers. Atoms are nature's qubits — every qubit is identical within and between programs. Logical operations can also be performed on any arbitrary pair of qubits, enabling complex quantum programs unhindered by physical connectivity. Want to learn more? Read IonQ’s trapped ion quantum computer technology overview.
- Publisher: IonQ
- Provider ID:
ionq
The following targets are available from this provider:
| Target name | Target ID | Number of qubits | Description |
|---|---|---|---|
| Quantum simulator | ionq.simulator | 29 qubits | IonQ's cloud-based idealized simulator. Free of cost. |
| IonQ Aria 1 | ionq.qpu.aria-1 | 25 qubits | IonQ's Aria trapped-ion quantum computer. |
| IonQ Aria 2 | ionq.qpu.aria-2 | 25 qubits | IonQ's Aria trapped-ion quantum computer. |
| IonQ Forte | ionq.qpu.forte | 32 qubits | IonQ's Forte trapped-ion quantum computer. Available in Private Preview only. |
IonQ's targets correspond to a :::no-loc text="QIR Base"::: profile. For more information about this target profile and its limitations, see Understanding target profile types in Azure Quantum.
GPU-accelerated idealized simulator supporting up to 29 qubits, using the same set of gates IonQ provide on its quantum hardware—a great place to preflight jobs before running them on an actual quantum computer.
- Job type:
Simulation - Data Format:
ionq.circuit.v1 - Target ID:
ionq.simulator - Target Execution Profile: :::no-loc text="QIR Base":::
IonQ Aria is the flagship of IonQ's trapped-ion quantum computers, with a 25-qubit dynamically reconfigurable system. For more information, see IonQ Aria (ionq.com).
Important
Debiasing is enabled on Aria systems by default, and submitted jobs are subject to debiasing-based pricing. For more information about debiasing and how to disable/enable the service, see Error mitigation.
- Job type:
Quantum Program - Data Format:
ionq.circuit.v1 - Target ID:
ionq.qpu.aria-1,ionq.qpu.aria-2 - Target Execution Profile: :::no-loc text="QIR Base":::
| Parameter Name | Type | Required | Description |
|---|---|---|---|
shots |
int | No | Number of experimental shots. |
| Measure | Average time duration |
|---|---|
| T1 | 10-100 s |
| T2 | 1 s |
| Single-qubit gate | 135 µs |
| Two-qubit gate | 600 µs |
| Operation | Average fidelity |
|---|---|
| Single-qubit gate | 99.95% (SPAM corrected) |
| Two-qubit gate | 99.6% (not SPAM corrected) |
| SPAM* | 99.61% |
* State Preparation and Measurement (SPAM): This measurement determines how accurately a quantum computer can set a qubit into its initial state and then measure the result at the end.
IonQ Aria is available through Azure Quantum Credits plan and a separate billing plan. For more information, see Azure Quantum pricing.
IonQ Forte is IonQ's highest-performing, commercially available trapped-ion quantum computer. With a 32-qubit software-configurable system, IonQ Forte is available in Private Preview on Azure Quantum. For more information, see IonQ Forte (ionq.com).
Important
Debiasing is enabled on the Forte system by default, and submitted jobs are subject to debiasing-based pricing. For more information about debiasing and how to disable/enable the service, see Error mitigation.
- Job type:
Quantum Program - Data Format:
ionq.circuit.v1 - Target ID:
ionq.qpu.forte - Target Execution Profile: :::no-loc text="QIR Base":::
| Parameter Name | Type | Required | Description |
|---|---|---|---|
shots |
int | No | Number of experimental shots. |
In Q#, the output of a quantum measurement is a value of type Result, which can only take the values Zero and One. When you define a Q# operation, it can only be submitted to IonQ hardware if the return type is a collection of Results, that is, if the output of the operation is the result of a quantum measurement. The reason for this is because IonQ builds a histogram from the returned values, so it restricts the return type to Result to simplify creating this histogram.
IonQ's targets correspond to the :::no-loc text="QIR Base"::: profile. This profile can't run quantum operations that require the use of the results from qubit measurements to control the program flow.
When you submit a quantum program to the IonQ simulator, it returns the histogram created by the measurements. The IonQ simulator doesn't sample the probability distribution created by a quantum program but instead returns the distribution scaled to the number of shots. This is most apparent when you submit a single shot circuit. You will see multiple measurement results in the histogram for one shot. This behavior is inherent to IonQ simulator, while IonQ QPU actually runs the program and aggregates the results.
Additional capabilities supported by IonQ hardware are listed here.
| Capability | Description |
|---|---|
| Error mitigation | Use debiasing to minimize noise and maximize algorithmic performance on IonQ hardware |
| Native gates support | Define and execute circuits directly on IonQ hardware-native gates |
| Noise model simulation | Simulate the noise profile that circuits will encounter when you run them on different IonQ hardware. |
Users can take advantage of these additional capabilities via pass-through parameters in the Azure Quantum Q# and Qiskit providers.
IonQ provides the option to enable quantum error mitigation when submitting jobs to IonQ hardware. Error mitigation is a compiler-level process that runs and executes multiple symmetric variations of a circuit, and then aggregates the outcomes while mitigating the impact of hardware errors and qubit decoherence. Unlike quantum error correction techniques, error mitigation does not require large gate and qubit overhead.
Debiasing is the process of creating slight variations of a given circuit that should be identical on an ideal noiseless machine, using techniques such as different qubit assignments, gate decompositions, and pulse solutions, and then executing those variations.
Sharpening and Averaging are options for aggregating the results of the variations. Averaging is based equally on all the variation results, whereas Sharpening filters out the erroneous results, and can be more reliable for certain types of algorithms.
For more information, see Debiasing and Sharpening. For error mitigation pricing, see IonQ pricing.
Note
Debiasing is enabled by default on Aria and Forte systems.
On Azure Quantum, error mitigation can be enabled or disabled for jobs submitted with Q# or with Qiskit.
To enable error mitigation, add an optional parameter for the target machine:
option_params = {
"error-mitigation": {
"debias": True
}
}To disable error mitigation, set the parameter to False:
option_params = {
"error-mitigation": {
"debias": False
}
}Note
If you are also using IonQ's noise model simulation, those parameters can be included here, for example:
option_params = {
"error-mitigation": {
"debias": False
},
"noise": {
"model": "aria-1",
"seed": 100
}
}For more information, see Noise model simulation.
This example uses a simple random number generator.
First, import the required packages and initiate the base profile:
import qsharp
import azure.quantum
qsharp.init(target_profile=qsharp.TargetProfile.Base)Next, define the function.
%%qsharp
import Std.Measurement.*;
import Std.Arrays.*;
import Std.Convert.*;
operation GenerateRandomBit() : Result {
use target = Qubit();
// Apply an H-gate and measure.
H(target);
return M(target);
}
and compile the operation:
```python
MyProgram = qsharp.compile("GenerateRandomBit()")Connect to Azure Quantum, select the target machine, and configure the noise parameters for the emulator:
MyWorkspace = azure.quantum.Workspace(
resource_id = "",
location = ""
)
MyTarget = MyWorkspace.get_targets("ionq.qpu.aria-1")Specify the error-mitigation configuration
option_params = {
"error-mitigation": {
"debias": True
}
}Pass in the error mitigation configuration when submitting the job:
job = MyTarget.submit(MyProgram, "Experiment with error mitigation", shots = 10, input_params = option_params)
job.get_results()In Qiskit, you pass the optional parameters to the target machine configuration before submitting the job:
circuit.name = "Single qubit random - Debias: True"
backend.options.update_options(**option_params)
job = backend.run(circuit, shots=500)Note
If you do not pass in the error-mitigation parameter, the target machine will use its default setting, which is enabled for Aria and Forte systems.
By default IonQ allows you to specify a quantum circuit using an abstract set of quantum gates, called qis, which allows flexibility and portability when writing an algorithm without worrying about optimization for the hardware.
However, in some advanced usage cases, you might want to define a circuit directly on native gates in order to be closer to the hardware and bypass optimization. The native gate set is the set of quantum gates that are physically executed in the quantum processor, and they map the circuit to those as part of the execution.
For more information, see Getting Started With Native Gates (ionq.com).
In order to use the native gate set when submitting Qiskit jobs to Azure Quantum, you specify the gateset parameter when initializing the backend as in the example below:
# Here 'provider' is an instance of AzureQuantumProvider
backend = provider.get_backend("ionq.qpu.aria-1", gateset="native")| Parameter Name | Type | Required | Description |
|---|---|---|---|
gateset |
string | No | Specifies the set of gates that will be used to define a circuit. A value of qis corresponds to the abstract gates (default behavior) and native to the IonQ hardware native gates. |
For more information about Qiskit jobs, see Submit a circuit with Qiskit.
Even the best of today's quantum hardware has inherent noise, and knowing the noise characteristics of your target system can help you refine your algorithms and get a more realistic prediction of results when running the circuit on hardware. IonQ provides a noise model simulation that introduces noise into the circuit using a "noise fingerprint" specific to the target hardware. For more information, see Get Started with Hardware Noise Model Simulation.
| Parameter Name | Values | Description |
|---|---|---|
noise |
model, seed
|
Enables the noise model simulation |
model |
ideal, aria-1
|
Specifies the noise model for the target hardware.
|
seed |
Integer between 1 and |
Allows you to specify a seed value for pseudo-random noise and shot-sampling, creating reproducible noisy results. If the parameter is not specified, a random seed value is created. |
Noise model simulation is shot-aware; that is, it samples measurements from the output state based on the number of shots provided. In Azure Quantum, the shots parameter is submitted with the job, and is required for aria-1 noise models. If no shot value is specified, a default value of 1000 is used. If the ideal noise model is used, the shots parameter is ignored.
While the ideal noise model allows you to simulate up to 29 qubits with the IonQ quantum simulator, the hardware specific noise models are limited to the actual qubit capacity of the target hardware, which is 25 qubits for the aria-1 noise model.
On Azure Quantum, noise model simulation can be enabled or disabled for jobs submitted with Q# or with Qiskit.
To enable noise model simulation, add an optional parameter for the target machine, for example:
option_params = {
"noise": {
"model": "aria-1", # targets the Aria quantum computer
"seed" : 1000 # If seed isn't specified, a random value is used
}
}Note
If you are also using IonQ's error mitigation, those parameters can be included here, for example:
option_params = {
"error-mitigation": {
"debias": False
},
"noise": {
"model": "aria-1",
"seed": 1000
}
}For more information, see Error mitigation.
You can use the same example program shown previously in error mitigation and add or replace the noise model configuration in option_params;
option_params = {
"error-mitigation": {
"debias": True
},
"noise": {
"model": "aria",
"seed": 1000
}
}Then pass the optional parameters when you submit the job:
job = MyTarget.submit(MyProgram, "Experiment with noise model simulation", shots = 10, input_params = option_params)
job.get_results()In Qiskit, you pass the optional parameters to the target machine configuration before submitting the job:
circuit.name = "Single qubit random - Debias: True"
backend.options.update_options(**option_params)
job = backend.run(circuit, shots=500)To see IonQ billing plan, visit Azure Quantum pricing.
IonQ quotas are tracked based on the QPU usage unit, which is qubit-gate-shot (QGS). The resource usage is credited against your account.
Every quantum program consists of
where:
-
$N$ is the number of one- or two-qubit gates submitted -
$C$ is the number of execution shots requested
Quotas are based on plan selection and can be increased with a support ticket. To see your current limits and quotas, go to the Credits and quotas blade and select the Quotas tab of your workspace on the Azure portal. For more information, see Azure Quantum quotas.
Note
If you are using an Azure Quantum Credits plan, and not a billing plan, the quotas information maps to your allocated credits. In that case, the quota lists the total number of credits you have received.
For information about IonQ QPU job processing delays, see IonQ status page.
To see recommended best practices for the IonQ QPU, see IonQ Best Practices (ionq.com).