Efficient logical operations in quantum LDPC codes using simple resource states

arXiv:2607.16166 · quant-ph · Submitted 2026-07-17 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Quantum Radio. Generated commentary on the latest quantum physics and condensed matter papers.

Kai: We're looking at the paper: "Efficient logical operations in quantum LDPC codes using simple resource states".

Title and authors: Kai: So, we're looking at this paper called "Efficient logical operations in quantum LDPC codes using simple resource states." It sounds like they're trying to make doing actual computations on these quantum codes way more efficient.

Mira: Yeah, the title suggests they're focusing on how to speed up the logical operations themselves by using simpler resource states, which is interesting because we usually have to design really complex things just for measurements.

Lev: I’m wondering if this means less overhead in terms of the physical qubits we need just to perform these logical gates and measurements compared to what we see with surface codes.

Kai: Exactly, and they specifically mention that quantum LDPC codes offer a substantial reduction in qubit overhead when you compare them to the surface code because of their high encoding rate.

Mira: But they point out that operating on multiple logical qubits encoded in the same block is harder, which makes things potentially slower than just measuring one operator at a time.

Lev: That's where I see a potential bottleneck—if we have to do everything sequentially instead of in parallel, the speed-up might not be as dramatic as they claim.

The paper's summary: Kai: This paper introduces an approach that only consumes cat states for logical measurements, which is a big deal because it avoids the need to form a merge code that we usually have to do.

Mira: They leverage Shor’s measurement scheme, which only uses cat states like zero... zero⟩ plus one... one⟩, and this method means each cat state only interacts with the memory for one time step before it gets measured <ref:2607.16166#pg3>.

Lev: From an engineering standpoint, that sounds nice because it simplifies the interaction between the cat state and the rest of our physical system.

Kai: It’s not just about measurement; they are designing a scheduler code that determines the sequence of measurements and lets you decode all logical measurement outcomes at once.

Mira: They call this protocol MEDM, which generalizes an earlier idea called EDM to a set of l commuting Pauli operators using an l times m scheduler matrix G.

Lev: So, instead of having to run many separate measurement sequences, they are designing one schedule that covers everything needed for the joint measurement.

Kai: And they also introduce the MECM protocol, which generalizes the ECM protocol by applying a decoder D to the measurement outcome vector v to find the most likely logical outcome.

The paper's improvements: Mira: The improvements they highlight are pretty specific; they show a speed-up of nearly three times over Viterbi measurements for measuring twenty commuting logical operators when using Q70 codes in simulations.

Kai: That's a real number, and combining these fast logical measurements with a new variant of the CliNR partial error correction scheme gives them a speed-up of up to seventy-four times for random Clifford circuits.

Lev: Seventy-four times for random Clifford circuits sounds huge, but I need to know what that means in terms of physical execution time on real hardware.

Kai: They show that the MEDM protocol requires fewer cat states per Pauli operator than the protocols from a previous paper they cite, which is a direct improvement over existing methods.

Mira: They also show that when combining fast logical measurements with CliNR, the MEDM CliNR protocol shows a reduction of between two point two and two point six times more cat-based measurements compared to the Viterbi CliNR method for Toffoli gates on Q70 codes.

Conclusion: Kai: So, to wrap up, this paper presents methods to speed up logical measurements, Clifford gates, and even non-Clifford Toffoli gates in quantum LDPC codes using these simple resource states.

Mira: The main thing is that the minimal space and time footprint of their logical operations makes them a very good target for near-term experimental demonstration because they are so small in terms of physical requirements.

Lev: From my side, the fact that they can achieve this with fewer cat-based measurements—like two point eight one cat states per logical measurement compared to three point four two for Viterbi—is exactly what we need to see for practical implementation on walking cat architectures.

Kai: And they are looking ahead by asking what features of quantum LDPC codes are sufficient to guarantee that the cat-based measurement outcomes remain independent.

Mira: There’s also an open question about designing explicit families of scheduler codes instead of just picking the codes with the best known distance for this task.

Lev: Ultimately, their focus on minimal space and time footprint makes them an ideal target for near-term experimental demonstration because it keeps the complexity manageable for hardware right now.

Mark Webster, Nicolas Delfosse

IonQ Inc.

quant-ph

Submitted: 2026-07-17

Updated: 2026-11-18

Comments: 15 pages, 4 figures

DOI: 10.1103/vz2d-jq2z

License: http://creativecommons.org/licenses/by/4.0/

The gist: Quantum LPDC codes provide a substantial reduction in qubit overhead required for fault-tolerant quantum computation compared to surface code…

Key concepts

Quantum LPDC Codes
These are a type of quantum error-correcting code that require less physical qubits than surface codes. They achieve this by using high encoding rates and are designed to be efficient for fault-tolerant quantum computation, reducing the overall qubit overhead needed.
Scheduler Code (MEDM)
This is a design that determines the sequence of cat-based measurements. It allows for 'joint decoding,' meaning all measurement outcomes from different logical operators can be decoded simultaneously, leading to much faster logical measurements than traditional methods like Viterbi.
CliNR Partial Error Correction Scheme
This is a new variant of partial error correction that leverages fast cat-based measurements. It works by executing part of the error correction at the logical level and other parts at the physical level to speed up logical operations, particularly Toffoli gates.
Cat States
These are specific quantum states used in this protocol, like |0... 0⟩ + |1... 1⟩. The method cleverly uses these states only for logical measurements, avoiding the need to form a complex merge code because each state interacts with memory only once per time step.

Terminology

Summary

Quantum LPDC codes provide a substantial reduction in qubit overhead required for fault-tolerant quantum computation compared to surface code, thanks to their high encoding rate The key ingredient is the design of a scheduler code determining the measurement sequence and allowing for the decoding of all logical measurement outcomes Numerical simulations with the LDPC codes Q70 and Q102 of the walking cat architecture show a speed-up of nearly 3× over Viterbi measurements for the measurement of l = 20 commuting logical operators Combining our fast logical measurements with a new variant of the CliNR partial error correction scheme, we achieve a speed-up of up to 74× for random Clifford circuits. p. 0

Where the paper says it

Numerical simulations with the LDPC codes Q70 and Q102 of the walking cat architecture show a speed-up of nearly 3× over Viterbi measurements for the measurement of ℓ = 20 commuting logical operators. Combining our fast logical measurements with a new variant of the CliNR partial error correction scheme, we achieve a speed-up of up to 74× for random Clifford circuits.

Pruning redundant reasoning tokens

The paper introduces an approach that only consumes cat states for logical measurements, leveraging Shor’s measurement scheme which only consumes cat states 0... 0⟩ + 1... 1⟩ This method removes the need to form a merge code because each cat state interacts with the memory only for a single time step before it is measured. p. 0

Where the paper says it

0⟩ + |1 . Each cat state interacts with the memory only for a single time step, before it is measured, removing the need to form a merge code.

The protocol is well-suited for moving qubits such as trapped ions [31], neutral atoms [32], spin qubits [33] or electrons floating on helium [34]. p. 0
Where the paper says it

Furthermore, we focus on the practical regime and we leverage the fact that fault-tolerance is not required to achieve the target logical error rate required for a specific quantum algorithm. Our approach requires bringing cat states next to the qubits to measure, making this protocol well-suited for moving qubits such as trapped ions [31], neutral atoms [32], spin qubits [33] or electrons floating on helium [34].

How the training signal is built

The core innovation involves designing a scheduler code that determines the sequence of cat-based measurements and enables the joint decoding of all measurement outcomes. The MEDM protocol generalizes the EDM protocol to a set of l commuting Pauli operators using an l × m scheduler matrix G, where each column corresponds to a product of the target Pauli operators. p. 1

Where the paper says it

Each row of G corresponds to one of the target Paulis to be measured. Each column of G corresponds to a product of the target Pauli operators so that the jth measurement is Qj := Q 0≤i<ℓ P Gij i .

The MECM protocol generalizes the ECM protocol by applying a decoder D to the measurement outcome vector v to find the most likely logical outcome uˆ. p. 2
Where the paper says it

MECM Protocol We now explain how to generalize the ECM protocol of [17] to ℓ commuting Paulis using an ℓ × m scheduler matrix G. A decoder D is applied to the measurement outcome vector v to find the most likely logical outcome uˆ.

Where the gains come from

The speed-up in logical operations is demonstrated through several metrics, including a speed-up of up to 74× for random Clifford circuits and a speed-up of up to 5× for Toffoli gates. The MEDM protocol requires fewer cat-based measurements per Pauli operator than the protocols of [17] For instance, for l = 20 commuting logical operators, the MEDM protocol consumes an average of only 2. p. 1

Where the paper says it

FAST MEASUREMENT OF COMMUTING PAULIS In this section we introduce the new MEDM and MECM protocols for measuring sets of commuting Pauli operators. The new protocols require fewer cat-based measurements per Pauli operator than the protocols of [17].

81 cat states per logical measurement compared to 3. p. 0
Where the paper says it

The measurement of ℓ = 10 and 20 commuting logical operators consumes respectively an average of only 2. 7 cat states per logical measurement.

42 for the Viterbi gate-based protocol. p. 3
Where the paper says it

5x CliNR Random Q102 74. 4x TABLE I: Multi-Pauli MEDM and MECM protocols compared to single-Pauli Viterbi protocol for ℓ = 10, 20 and logical CliNR protocol versus Viterbi gate-based circuit synthesis.

Performance analysis and results

The authors simulate protocols using Q70 and Q102 codes, achieving a speed-up of 4.2× and 5× for Toffoli gates respectively when combining fast logical measurements with the CliNR partial error correction scheme. p. 0

Where the paper says it

Combining our fast logical measurements with a new variant of the CliNR partial error correction scheme, we achieve a speed-up of up to 74× for random Clifford circuits. Our approach also applies to non-Clifford gates, producing a speed-up of up to 5× for Toffoli gates.

The results show that the MEDM protocol requires only 1. p. 7
Where the paper says it

Selecting an optimal generator matrix for the MEDM protocol depends on the undetectable error rate which is less complex to calculate exactly and we have closed form expressions for the key NAvg, mN and mAvg values. MEDM/MECM Example - Repetition Code for Single Pauli Measurements In this example, we show how MEDM and MECM apply for single Pauli measurements (ℓ = 1) and recover the results of [17] for the EDM, ECM and Viterbi measurement protocols.

81 physical measurements per target Pauli and the MECM only 1. p. 4
Where the paper says it

The MEDM protocol requires only 1. 81 physical measurements per target Pauli and the MECM only 1.

71 measurements. p. 0
Where the paper says it

However, operating simultaneously on multiple logical qubits encoded in the same block is more challenging and may slow down logical operations. Prior work addresses this problem by designing complex resource states to perform logical measurements in LDPC codes.

In simulation, the Viterbi gate-based protocol requires up to 5× more measurements than MEDM CliNR for Toffoli circuits. p. 4
Where the paper says it

For each scenario, we simulated using the Q70 code with 6 logical qubits and the Q102 code with 22 logical qubits from [17]. We found that the Viterbi gate-based protocol requires up to 5x more measurements than MEDM CliNR for Toffoli circuits.

The MEDM CliNR protocol shows a reduction of between 2. p. 9
Where the paper says it

Hence, the stabilizer ZD,i on the data block is mapped to CBZB,iC −1 B on auxiliary block B. A similar argument shows that XD,i is mapped to CBXB,iC −1 B and so the logical action for logical CliNR is as claimed.

2 and 2.6 times more cat-based measurements than the Viterbi CliNR method. p. 4
Where the paper says it

We then simulated logical CliNR using truncated MEDM for stabilizer state preparation. We compared the number of cat-based measurements required for logical CliNR to the number required by the Viterbi gatebased protocol.

Conclusion

We have presented methods to speed up logical measurements, Clifford gates and non-Clifford Toffoli gates in quantum LDPC codes. The minimal space and time footprint of our logical operations makes them an ideal target for near-term experimental demonstration. p. 4

Where the paper says it

Another open question is the design of explicit families of scheduler codes instead of selecting codes with the best-known-distance. Finally, the minimal space and time footprint of our logical operations makes them an ideal target for near-term experimental demonstration.

p. 4
Where the paper says it

Another open question is the design of explicit families of scheduler codes instead of selecting codes with the best-known-distance. Finally, the minimal space and time footprint of our logical operations makes them an ideal target for near-term experimental demonstration.

Future work will investigate which features of quantum LDPC codes are sufficient to ensure independence of cat-based measurement outcomes. p. 4
Where the paper says it

CONCLUSION We have presented methods to speed up logical measurements, Clifford gates and non-Clifford Toffoli gates in quantum LDPC codes. In future work, it would be interesting to investigate which features of quantum LDPC codes are sufficient to ensure independence of cat-based measurement outcomes.

The design of explicit families of scheduler codes instead of selecting codes with the best-known-distance is another open question. p. 4
Where the paper says it

In future work, it would be interesting to investigate which features of quantum LDPC codes are sufficient to ensure independence of cat-based measurement outcomes. Another open question is the design of explicit families of scheduler codes instead of selecting codes with the best-known-distance.

The minimal space and time footprint of our logical operations makes them an ideal target for near-term experimental demonstration.

The paper introduces a new variant of the CliNR partial error correction scheme which leverages fast cat-based measurements to speed up logical Toffoli gates and logical Clifford gates. To achieve a speed-up using CliNR, we execute part of the scheme at the logical level and other parts at the physical level to make them faster. p. 1

Where the paper says it

2 originally designed for partial error correction but the original scheme leads to an increase in execution time. To achieve a speed-up using CliNR, we execute part of the scheme at the logical level and other parts at the physical level to make them faster.

The MEDM protocol terminates as soon as an error is detected and we refer to this method as the truncated MEDM protocol. The average number of cat-based measurements is given by mAvg = mN NAvg for the truncated MEDM protocol. p. 2

Where the paper says it

Once ℓ measurements have been completed, the error-free logical measurement outcome u is determined, and we terminate if measurement outcomes are different to the error-free codeword uG. Based on the tail-sum formula, the average number of measurements per attempt depends on the UERs of the codes formed from the first i columns of G (denoted G|i) and is given by: mN := ℓ + 1 + X ℓ+1≤i≤m−1 (1 − pF ) i + UER⟨G|i⟩ (3) The average number of cat-based measurements is mAvg = mNAvg for the full MEDM protocol and mAvg = mN NAvg for the truncated MEDM protocol.

The logical Clifford operation is applied via state injection as follows, involving transversal CNOTs and destructive measurements. The logical CliNR protocol can be adapted to handle constraints on the maximum weight of logical Pauli measurements in the walking cat architecture. p. 3

Where the paper says it

In Section D 1 we prove that the above protocol applies a logical C and show how the protocol works in combination with frame tracking. Logical CliNR can be adapted to handle constraints on the maximum weight of logical Pauli measurements in the walking cat architecture (see Section D 3) and there is also a logical CZNR protocol for circuits composed of CZ and S gates which uses only one auxiliary code block (see Section D 2).

The accessibility of logical Pauli measurements depends on the QECC used, with some logical Paulis having minimum representatives of weight 30 for Q102. p. 4
Where the paper says it

A logical Pauli measurement is accessible if the Pauli has a minimum weight representative with weight w¯ or less. Accessibility depends on the QECC used - for Q70 all logical Paulis have representatives with weight ≤ 18 but for Q102 some logical Paulis have minimum representatives of weight 30.

The MEDM protocol uses a scheduler matrix G whose columns represent accessible Paulis, ensuring that all measurements in the schedule are accessible. The remaining m − l columns of G can now be chosen to give a low undetectable error rate. p. 9

Where the paper says it

Otherwise, let u be the integer vector indexing the vectors in Cw which form the information set. The remaining m − ℓ columns of G can now be chosen to give a low undetectable error rate.

The fitness function for vi is 1 − UER⟨Gi⟩ where the columns of Gi are the columns of Cw indexed by uvi>. p. 9
Where the paper says it

The initial population is a set of randomly generated vectors vi . The fitness function for vi is 1 − UER⟨Gi⟩ where the columns of Gi are the columns of Cw indexed by u|vi .

The logical CliNR protocol can be adapted to handle constraints on the maximum weight of logical Pauli measurements in the walking cat architecture. The stabilizers to be measured in step MC are of form MX,i = XA,iZB,i and MZ,i = ZA,iXB,i. p. 3

Where the paper says it

The resource state we prepare for logical CliNR is equivalent to preparing auxiliary blocks A and B in logical Bell pairs, then applying a logical C operation to block B. For the initial logical Bell pair state, the operators XA,iXB,i and ZA,iZB,i for 0 ≤ i < ℓ are logical stabilizers and the operators ZA,i and XB,i are logical destabilizers (i.

The logical CZNR protocol can be used for Clifford circuits composed of CZ and S gates and is set out in Figure 3. The logical CZNR protocol is simpler to execute than the logical CliNR protocol of Section IV B - only one auxiliary code block is required. p. 3

Where the paper says it

In Section D 1 we prove that the above protocol applies a logical C and show how the protocol works in combination with frame tracking. Logical CliNR can be adapted to handle constraints on the maximum weight of logical Pauli measurements in the walking cat architecture (see Section D 3) and there is also a logical CZNR protocol for circuits composed of CZ and S gates which uses only one auxiliary code block (see Section D 2).

The stabilizers of the injected state only need to be measured over one code block so stitched cat states are not required. p. 3
Where the paper says it

We measure each of the 2ℓ commuting operators MX,i, MZ,i using cat-based measurements and employ either the multi-Pauli techniques of Sections III B and III C or the single-Pauli techniques of Section II (block MC in Figure 1b). As these measurements are between two adjoining code blocks, stitched cat states may be required (see [17]).

The logical CliNR protocol requires fewer logical measurements than the gate-based method H H T T! T! The number of logical measurements required to implement the Clifford using single and two-qubit gates is O(l squared / log l), each of which requires 3 logical measurements. p. 5

Where the paper says it

The input to this protocol is a logical Clifford operator, which we decompose into a circuit of one and two-qubit logical Clifford gates (see [48, 50–52]). Single-qubit logical Clifford gates can be implemented using two logical Pauli measurements, and two-qubit gates using three logical Pauli measurements as set out in Figure 2.

The asymptotic bound on the number of 2-qubit gates required is O(l squared / log l). p. 4
Where the paper says it

We compared the number of cat-based measurements required for logical CliNR to the number required by the Viterbi gatebased protocol. The Viterbi gate-based protocol involves synthesizing a logical Clifford into logical single and twoqubit gates, which are then implemented via logical measurements.

p. 4
Where the paper says it

We compared the number of cat-based measurements required for logical CliNR to the number required by the Viterbi gatebased protocol. The Viterbi gate-based protocol involves synthesizing a logical Clifford into logical single and twoqubit gates, which are then implemented via logical measurements.

The MEDM protocol requires an average number of measurements mAvg = mN NAvg for truncated MEDM. The MECM protocol can be used directly on logical blocks without restarting and has a slightly lower average number of measurements. p. 7

Where the paper says it

Both protocols have relative advantages and are likely to be useful in different contexts. The MECM protocol can be used directly on logical blocks without restarting and has a slightly lower average number of measurements.

The MEDM protocol does not require a decoder so it is less complex to model and run. p. 7
Where the paper says it

The MECM protocol can be used directly on logical blocks without restarting and has a slightly lower average number of measurements. On the other hand, the MEDM protocol does not require a decoder so it is less complex to model and run.

The logical error rate requirement for the MECM protocol requires a larger repetition code to meet the target. To optimize mAvg, we follow the method in Section III D and use G = [G0|G1] where G0 is the block length m0 = 11 code used for the MEDM protocol. The resulting code G = [G0|G1] has distance d = d0+d1 and block size m = m0 + m1. The resulting code G = [G0|G1] has distance d = d0+d1 and block size m = m0 + m1.

The logical error rate is LER⟨G⟩ = 1 − P(CE⟨G⟩). The logical error rate requirement requires a larger repetition code to meet the target. p. 7

Where the paper says it

Hence, the logical error rate is: LER⟨G⟩ =1 − X 0≤i≤(m−1)/2 m i p i F (1 − pF ) m−i . The logical error rate requirement requires a larger repetition code to meet the target, so we calculate LERs for

We find that the block length 29 code with distance 15 is the smallest code with an LER upper bound estimate meeting the requirement. p. 8
Where the paper says it

30]: m 26 27 28 29 30 d 13 14 14 15 16 LER 2. 3E−11 We find that the block length 29 code with distance 15 is the smallest code with an LER upper bound estimate meeting the requirement.

p. 8
Where the paper says it

30]: m 26 27 28 29 30 d 13 14 14 15 16 LER 2. 3E−11 We find that the block length 29 code with distance 15 is the smallest code with an LER upper bound estimate meeting the requirement.

The resulting code G = [G0G1] has distance d = d0+d1 and block size m = m0 + m1. p. 8
Where the paper says it

Accordingly, we select for G1 the smallest ℓ = 4 code with distance d1 = 10 which has block size m1 = 20. The resulting code G = [G0|G1] has distance d = d0+d1 = 15 and block size m = m0 + m1 = 31.

p. 8
Where the paper says it

Accordingly, we select for G1 the smallest ℓ = 4 code with distance d1 = 10 which has block size m1 = 20. The resulting code G = [G0|G1] has distance d = d0+d1 = 15 and block size m = m0 + m1 = 31.

p. 8
Where the paper says it

Accordingly, we select for G1 the smallest ℓ = 4 code with distance d1 = 10 which has block size m1 = 20. The resulting code G = [G0|G1] has distance d = d0+d1 = 15 and block size m = m0 + m1 = 31.

p. 8
Where the paper says it

Accordingly, we select for G1 the smallest ℓ = 4 code with distance d1 = 10 which has block size m1 = 20. The resulting code G = [G0|G1] has distance d = d0+d1 = 15 and block size m = m0 + m1 = 31.

The Viterbi gate-based protocol requires an auxiliary logical qubit, so we used random Cliffords on l = k − 1 logical qubits. The number of cat-based measurements required by Viterbi GB and MEDM CliNR are compared in Table IV. p. 4

Where the paper says it

We then simulated logical CliNR using truncated MEDM for stabilizer state preparation. We compared the number of cat-based measurements required for logical CliNR to the number required by the Viterbi gatebased protocol.

We see a further gain when comparing the CliNR MEDM protocol with the Viterbi gate-based protocol. p. 3
Where the paper says it

4x TABLE I: Multi-Pauli MEDM and MECM protocols compared to single-Pauli Viterbi protocol for ℓ = 10, 20 and logical CliNR protocol versus Viterbi gate-based circuit synthesis. For more detail, see Tables II to IV blocks are adjacent.

The number of logical measurements required to implement the Clifford using single and two-qubit gates is O(l squared / log l). p. 5
Where the paper says it

The input to this protocol is a logical Clifford operator, which we decompose into a circuit of one and two-qubit logical Clifford gates (see [48, 50–52]). Single-qubit logical Clifford gates can be implemented using two logical Pauli measurements, and two-qubit gates using three logical Pauli measurements as set out in Figure 2.

The asymptotic bound on the number of 2-qubit gates required is O(l squared / log l).

The logical error rate requirement for the MECM protocol requires a larger repetition code to meet the target. We find that the block length 29 code with distance 15 is the smallest code with an LER upper bound estimate meeting the requirement. The resulting code G = [G0|G1] has distance d = d0+d1 and block size m = m0 + m1.

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Fast logical measurements can be performed by combining multiple commuting logical operators using a scheduler code, which allows for joint decoding of all the measurement outcomes and achieves speed-ups like nearly 3 times over Viterbi measurements for measuring 20 commuting logical operators in Q70 codes.

  2. Logical Clifford gates can be accelerated by replacing a Clifford circuit with an offline preparation of a resource state using error detection instead of error correction, leading to up to a 74 times speed-up for random Clifford operations when combined with the CliNR partial error correction scheme.

  3. The system can perform fast logical measurements for multiple commuting Pauli operators by introducing MEDM and MECM protocols, which require fewer cat-based measurements per Pauli operator than the protocols of [17].

  4. The system can implement logical Clifford operations through a Viterbi gate-based protocol where each logical Pauli measurement is executed using the Viterbi measurement protocol, resulting in a total number of cat-based measurements that is less than 2.2 times more than the CliNR protocol.

  5. The system can utilize the MEDM and MECM protocols for logical Clifford operations to reduce cat-based measurements compared to using the Viterbi CliNR method, showing a reduction factor of up to 4.2 times for Toffoli gates on Q70 codes.

Abstract

Quantum LDPC codes provide a substantial reduction in qubit overhead required for fault-tolerant quantum computation compared to surface code, thanks to their high encoding rate. However, operating simultaneously on multiple logical qubits encoded in the same block is more challenging and may slow down logical operations. Prior work addresses this problem by designing complex resource states to perform logical measurements in LDPC codes. Here, we propose an approach that only consumes cat states. Whereas previous work on cat-based measurements focuses on a single logical measurement, we design a protocol for the joint measurement of commuting logical operators. The key ingredient is the design of a scheduler code determining the measurement sequence and allowing for the decoding of all logical measurement outcomes. Numerical simulations with the LDPC codes Q70 and Q102 of the walking cat architecture show a reduction of nearly 3 times in the number of cat-based measurements required versus the Viterbi method for measuring =20 commuting logical operators. Combining our logical measurement protocols with a new variant of the CliNR partial error correction scheme, we achieve a reduction of up to 73 times versus the Viterbi method for random Clifford circuits. Our approach also applies to non-Clifford gates, producing a reduction of up to 5 times versus the Viterbi method for Toffoli gates.

Sources

Related papers