FAQs

Job Execution

Questions about about job execution.

What determines job queue time?

Average queue time is dependent on the number of jobs in the queue for a given emulator or QPU and the number of HQCs that have been run by your organization. Quantinuum implements a fair queue to keep access to the hardware fair across all organizations using the systems. As the total number of HQCs run by an organization increases, the queue time may also increase as jobs from other organizations are selected by the fair queueing system. Within an organization, Group and User priority levels can be used to give a specific group’s or user’s jobs higher priority. Lower priority jobs may see an increase in queue time.

What is the average job execution time?

Execution time is determined by the size of circuit submitted, the operations used, and the number of shots.

What is the average time required for different operations?

The times required for different operations are described in this paper. While not exact, the times reported are still close for how long it takes to do quantum operations and measurements. In the System Model H1 and H2 systems, a fair amount of time is spent transporting ions around the trap to arrange them for the next round of gating. We do not currently expose the exact time it takes for all transport operations for a given circuit. We provide an average depth-1 circuit time value that can be found in the Product Data Sheets.

Can I see information about the number of jobs in the queue and their wait times?

No, users cannot see the number of jobs in the queue or their wait times. Quantinuum H-Series jobs operate with a Fair Queue where the jobs that run are picked based on the HQC accumulation for an organization. This is an active, dynamic calculation that is changing as organizations add jobs to the queue, therefore it is difficult to provide wait times.

Can I know which error rates were used during execution?

We have all specification information located in our Product Data Sheets on the Quantinuum website:

• System Model H1 Product Datasheet

• System Model H2 Product Datasheet

Our devices run consistently such that we guarantee their performance will always be at or better than the specifications reported in the Product Data sheets. We consistent run better than the minimum specification listed. You can also find links to our benchmarking data and specification data on Github on these pages.

Machine Questions

Questions about specifications or operations of the H-Series quantum computers.

Is it beneficial to parallelize quantum circuits on the machine if my quantum circuits use fewer qubits than available on the device?

It depends on the circuit. We recommend that rather than thinking about idling qubits, think about idling zones. System Model H1 can run up to 5 gates in parallel while System Model H2 can run 4 gates in parallel. If your circuit is already using all gate zones, there is no speed up by running copies of a circuit in parallel as this will incur extra memory error at each step. If your circuits run less than the number of gate zones in parallel, then you might see benefit, but it is not guaranteed.

Where can I find T1 or T2 times?

We do not provide T1 and T2 times because they are not very representative of machine performance. H-Series devices physically move qubits around the trap to achieve arbitrary connectivity, so the timing of a circuit can be quite complex such that simple ratios such as gate time or T2 are not descriptive. For this reason we specify a memory error per depth-1 circuit time in the Product Data Sheets. The depth-1 circuit time is the time it takes to arbitrarily pair all qubits, perform 1-qubit gates on all qubits, and do a round of 2-qubit gates on all pairs. For example, a 20-qubit trap can have up to 10 2-qubit gates and 20 1-qubit gates in a single depth-1 circuit time. The Product Data Sheets can be found at the following pages:

• System Model H1

• System Model H2

H-Series T1 times are greater than one minute. With such a long T1 time, it is time consuming to measure with high precision and knowing the time is \(>> 1\) minute is sufficient. Part of the trapped ion quantum computing community may say this time is infinite because in the dark the atom will never decay from one hyperfine state to the other, but in practice there is always light around that can unintentionally drive transitions between the 2-qubit states. Our latest T2 measurement was around 4 seconds. However, as noted above, this is not the most useful number when trying to estimate performance.

For more information about circuit timing, see the following papers:

• A Race Track Trapped-Ion Quantum Processor

• Demonstration of the trapped-ion quantum-CCD computer architecture

What would be the difference between using the General SU(4) Entangler gate vs. three Arbitrary Angle ZZ gates?

The General SU(4) Entangling gate, OpType.TK2 in pytket, has faster execution time than three Arbitrary Angle \(ZZ\) gates, OpType.ZZPhase in pytket. With three Arbitrary Angle \(ZZ\) gates, the compiler can cause out-of-gate zone transport between the Arbitrary Angle ZZ gates, but with the General SU(4) Entangling gate it won’t allow that type of transport. So the General SU(4) Entangling gate might be faster and have lower memory errors. The biggest difference appears on System Model H2 devices where the transport might have larger shifts that take longer.

Can I see the error rate per qubit?

There aren’t any per-qubit error rates since qubits in our ion trap architecture are identical. We’re using ions that are identical (“nature’s qubit”) rather than manufactured qubits. This is an advantage of our devices. Users don’t need to worry about qubit-specific error rates such as they do in other architectures and can simply focus on expressing the quantum circuits they want to run as-is, without needing extra care for qubit-specific errors.

Can I see information about frequency and timing of calibrations?

H-Series quantum computers are calibrated amidst operations, which provides several benefits to users and is unique compared to other architectures. Because of the limited and small number of gate zones that need to be calibrated, H-Series devices can perform calibrations during the middle of operations and occur between system-level chunk executions of customer jobs. This ensures users’ jobs are always run at high performance levels. Users do not need to worry about when these occur.

The local emulator is slow. What can I do?

Note

Only valid for pytket-quantinuum users. Nexus users have access to Nexus-hosted emulators.

The local H-Series emulator can be installed using the command pip install pytket-quantinuum[pecos]. projectq is a dependency of the local H-Series emulators. For speed issues using the H-Series emulators, the end-user should consult this page.

User Portal Questions

Questions about User Management and User Interface.

Nexus User Portal Questions

For nexus FAQs on the user portal, please see here.

What is the data retention policy for the Nexus User Portal?

Job data is stored indefinitely.

Legacy User Portal Questions

Is it possible to resend the user portal invitation email?

Yes. Work with your organization administrator to resend the invitation email. For administrators, find the user under User in the user portal and click Resend Invitation under the actions button.

How do I enable multi-factor authentication (MFA)?

For accounts that were not created using a Microsoft Single Sign On (SSO), account MFA can be enabled on the Account tab under SECURITY. Accounts created using Microsoft SSO are not eligible to use MFA through Quantinuum’s portal and instead would set up MFA through Microsoft.

How do I update my account email and password?

The email and password for your account can be updated in the Account tab in the user view. Underneath the ACCOUNT section are buttons to update the password and email separately. Both actions will sign a user out and require them to login using the new credentials.

What is the data retention policy for the User Portal?

Job data is stored for 40 days. Because of this, Quantinuum strongly recommends that users download job results as soon as they are available. This is referenced in the section. Additionally, if an organization would like to track usage metrics, organization administrators need to download monthly usage reports in the Reports tab.

Is data recoverable after the data retention period has passed?

It is unlikely that job results will be recoverable after the data retention period has passed, but if this happens please reach out to QCsupport@quantinuum.com and we will aim to assist you.