Next-Generation Topology of D-Wave Quantum Processors

“• Simulated annealing: a standard heuristic algorithm that exploits thermal excitations to search a classical energy landscape and relax upon a minimizing solution [13].”

“One measure of the complexity of a graph is its treewidth [3].”

“For example, the minimum energy of an Ising model defined on a graph of treewidth t with n vertices can be found in time O ( n 2 t ) using dynamic programming [4].”

“the minima as system size increases. We consider instead a Hamze de-Freitas Selby (HFS) greedy descent method [17, 15], using Nrandom trees of treewidth 14, determined by the minimum-degree heuristic [18]. We perform an exact minimization over variables in a randomly selected tree, and iterate until energy no longer d…”

“This inspired several quantum and classical devices, including the D-Wave quantum annealer (25–27), the coherent Ising machine (28, 29), the recently-introduced Fujitsu-led application-specific CMOS-based digital annealer (30), and even more recently, the Toshiba simulated bifurcation algorithm…”

“1 device has roughly 5000 hardware qubits in a topology called Pegasus depicted in Figure 2 (bottom row) [51, 52].”

“Grain 128 and Grain 128a ciphers’ coefficients in their QUBO matrix Q are within the range [ − 40 , 64 ] for quadratic monomials and within the range [1,64] for linear monomials.”

“The Pegasus architecture [78], denoted by P M , was used for various system sizes.”

“For instance, the D-Wave Advantage [45] can handle more than 5000 qubits as a sparse graph and 177 logical spins as a clique.”

“The latter is based on the so-called Pegasus architecture [22], which consists of a lattice of physical qubits and connections between pairs of them.”

“2 Pegasus topology of the D-Wave Advantage where curved blue lines, long red lines and short red lines are internal, external and odd couplers, respectively [30].”

“To make the embedding search faster, we raster-scan across 2 × 2 blocks of unit cells in the QPU’s Pegasus graph [16], then greedily construct a large set of non-intersecting embeddings.”

“The next generation of D-Wave devices (D-Wave Advantage) has an improved connectivity with the Pegasus graph of degree 15 [37], but was not available for testing as of writing time.”

“This quantum computer is equipped with 5,554 working qubits and over 35,000 couplers, organized in a Pegasus topology [20].”

“The latest quantum annealer from D-Wave features fewer than 6000 flux qubits arranged, in a Pegasus topology (Boothby et al. 2020).”

“Pegasus has a fixed described structure in which each qubit is connected with other fifteen qubits [56].”

“As of the time of the crafting of this article, the D-Wave Advantage 4.1 quantum processing unit (QPU) [6] hardware show-cases 5,640 functional qubits intricately connected through a sparsely connected Pegasus topology and a technique called minor embedding [27].”

“Prior chip-based approaches towards this goal focussed on mapping problems onto a fixed connectivity topology such as Chimera and Pegasus [13] and King’s graphs [15, 6].”

“…can reach the moment of O (10 − 12 ) seconds after big-bang, the realm of “quantum universe”, Nature will reveal its secrets when we face it with quantum technologies. squeeze performance in this limited situation, Advantage has a connectivity structure among qubits, called as Pe-gasus graph [32].”

“More generally, many NPcomplete and NP-hard problems can be formulated as finding energy minima in Lenz-Ising models [38], with practical applications leveraging quantum devices [39].”

“The quantum annealing market has existed for over a decade, with current quantum an-nealers consisting of superconducting devices containing a few-thousand qubits [3].”

“Unlike D-Wave that requires the problem to be embedded in a Chimera graph ([7] or (more recently) PEGASUS graph [8], DA has the advantage that it works with a fully connected graph; in [9], the scaling advantage of all-to-all connectivity was discussed.”

“t [26]. Most experimental investigations of AQC have thus far been conducted on devices from D-Wave Systems [8], a Canadian company which recently announced a 5000-qubit programmable quantum annealer [27]. Although quantum speedup on their devices has not been conclusively demonstrated, they have successfully exhibite…”

“NoticethattheRBMdataarebinaryvectorssuchthateachelementcan takethevalueof0or1,whereasthequbitscanhavevalues − 1or1.”

“ing to a Pegasus graph [55], as shown in the bottom panel of Fig.”

“Finally we note that commercial annealers do not yet offer the ability to control the transverse fields of the qubits independently[44].”

“Future work may consider how to adapt these findings to new generations of hardware, such as the Pegasus architecture announced by D-Wave that features 15 connections per qubit and 5000 qubits [24].”

“However, for artificial spin systems with couplers, such as the ones used for quantum annealing (see for example [9]), our results with a finite number of coordination spheres can be viewed as exact.”

“Furthermore, we present a comparative analysis focusing on embedding efficiency using the topologies of both the D-Wave 2000Q Quantum Processing Unit (QPU) of 2000 qubits and the recently released D-Wave Advantage platform based on a 5000-qubit QPU [14].”

“[34].”

“The hybrid algorithms we tried and the one heuristic we tried ran much faster but always found sub-optimal solutions.”

“The D-Wave Advantage system uses a Pegasus (Boothby et al. 2020) topology graph.”

“As a prime example, commercial quantum annealers from D-Wave Systems provide opportunities to map optimization problems into a 5000-qubit system [15].”

“[56] Kelly Boothby, Paul Bunyk, Jack Raymond, and Aidan Roy, Next-Generation Topology of D-Wave Quantum Processors, Tech.”

“Figure 2 shows the disjoint clique minor embeddings on the Pegasus connectivity graph of sizes ranging from two 100-vertex clique embeddings to twelve 50-vertex clique embeddings.”

“While Pegasus architecture QPUs [5] will increase both the number of qubits and qubit connectivity, embedded problems may still be challenging, as they often require numerous variables due to one-hot encoding of integer variables.”

“These QPUs are described by coupler connectivity graphs known as Pegasus [68, 69] and Zephyr [70].”

“…32/64 bits SA GPU Coherent Ising machine [43] 100,000 Complete 2 bits − Laser, FPGA D-Wave 2000Q [13] 2,048 Chimera − QA SQUID D-Wave Advantage [44], [45] 5,627 Pegasus − QA SQUID SQA: Simulated Quantum Annealing, MA: Momentum Annealing, PT: Parallel Tempering, bSB: ballistic SB, dSB: discrete…”

“This paradigm has been realized in practical hardware via systems like the D-Wave Advantage, which uses thousands of superconducting qubits [4, 5].”

“Since the ferromagnetic Ising model is invariant under such transformations (upon appropriate decoding), this procedure preserves physical observables while averaging over residual hardware biases 39 .”

“Current quantum annealers natively encode and solve Ising problems, so a QUBO problem can be solved by translating it into an Ising one —according to (14)—and handling the partial connectivity of the backend with minor embedding techniques [18].”

“In Section IV we first discuss the issues related to directly solve the integer linear problems via unrestricted algorithm on the quantum hardware.”

“The new Pegasus connection graph of next-generation D-Wave chip allow for 15 connections per qubits and therefore decreases the need for duplication of logical variables (Boothby et al., 2020).”

“Once the QUBO model was created using the DWAVE API [22], it was converted to the form of a binary quadratic model and then run on the Quantum Annealing Sampler.”

“The quantum annealing hardware embeds a maximum of 180 spins on a complete graph [33].”

“Importantly, these three Isings are natively embeddable onto the D-Wave Pegasus Quantum Processing Unit (QPU) topology [55, 89, 90].”

“For larger problems with n = 32 , we can use D-Wave’s Pegasus architecture [5], which has more intercon-nections (qubit couplings) between its K 4 , 4 unit cells than Chimera.”

“According to [54], which uses the formulas of [7], the maximal fully connected hardware (complete graph topology) that can be simulated on the D-Wave Advantage is for 182 logical qubits (with maximal chain lengths of 17) and 64 logical qubits on the D-wave 2000X (also with maximal chain length of 17).”

“Chain overhead consumes qubits and, if chain strengths are not tuned correctly, can degrade solution quality; although the higher degree of Zephyr reduces the average chain length, the effective instance size still falls to only a few thousand binary variables for most transport encodings [71,72].”

“The Pegasus topology supports 16 couplers per qubit [48], meaning dense, highly connected problems require careful embedding to avoid qubit fragmentation.”

“Thesetopologiesdefinehowlogicalproblem graphscanbeembeddedintothephysicalhardware.”

“…research has typically employed classical graph metrics such as degree centrality, betweenness centrality, closeness centrality, eigenvector centrality, and clustering coefficients to evaluate quantum computing topologies, including D-Wave’s Chimera, Pegasus, and Zephyr architectures [49,50,67–69].”

“[42], showing how the fully connected graph is mapped onto the limited qubit connectivity of the quantum hardware.”

“This computer boasts 5,616 qubits and over 35,000 couplers arranged in a Pegasus topology [9].”

“This parameter λ , known as chain strength in D-Wave quantum annealers, must be chosen with a sufficiently high value to ensure that the quantum annealing process identifies solutions where a = a ′ as viable.”

“[51, 52], native embeddings of 127 qubit heavy-hex lattices onto P 16 Pegasus graphs [55–57] were created, so we can use these tiled embeddings to embed the Ising model in question onto the current D-Wave Pegasus graph devices.”

“The reason is that [15] embedding methods in Chimera topology may be translated to Pegasus topology without modification, because Chimera”

“where zi, zj are called logical variables, and Qij includes a bias term hi and the interaction strength of the logical variables gij [16].”

“However, for artificial spin systems with couplers, such as the ones used for quantum annealing (see for example [14]-[20]), the obtained expressions are exact.”

“The current generation of the D-Wave quantum annealer (the Chimera architecture) includes more than 2,000 qubits and about 6,000 couplers, while the next generation (the Pegasus topology of the Advantage) will include more than 5,000 qubits and about 40,000 couplers 11 .”

“The drawbacks include hardware limitations such as sparse, local qubit connectivity, which limits scalability (e.g., D-Wave’s Pegasus graph architecture [20]), limited coherence times [21], which constrain the duration of quantum e ff ects, and high qubit error rates [22].”

“This process involves a randomly generated lattice dimerized at each point, and then subjected to a rewiring of the edges to fit the Pegasus architecture [17,71].”

“Most of the computational effort is classical and can therefore be offloaded to high-performance computing (HPC) infrastructure, while the optimization stage is handled by specialized solvers.”

“Fixed sparse architectures like the Chimera [14], Pegasus [15], and King’s Graphs [16] allow for efficient sparse problems to be mapped but exacerbate inefficiencies when more dense connectivity is required [17].”

“In contrast to QA, gate-based quantum computing, inspired by classical logic gates, operates by driving the quantum system within Hilbert space through a finite sequence of unitary gates, enabling universal computation.”

“As expected, the overhead produced by the embedder constantly increases with the number of arguments n .”

“Minor embedding is an NP-Hard problem but polynomial-time heuristic algorithms are available (Choi 2008; Cai et al. 2014; Boothby et al. 2020) 3 .”

“The space of embedding orders which are permutations of the set 𝑆 with the prefix 𝑂 𝐴 . modification, since Chimera is a subgraph of Pegasus [3].”

“We denote v i ( t + 1) as the updated membrane potential for neuron i at time t + 1.”

“The D-Wave Advantage can handle QUBO models with a 5760-node Pegasus graph [15].”

“This seemingly simple question calls for a reconsideration of our training algorithms, due to an important distinction between classical optimisation and the state-of-the-art quantum annealers (QA), such as the D-Wave Advantage [10, 18] 1 .”

“In particular, the physical qubits across currently existing D-Wave generations use connections determined by Chimera, Pegasus, or Zephyr graphs [29, 30].”

“The D-Wave devices used in this study have coupler hardware graphs known as Pegasus [33, 34] and Zephyr [35].”

“This marks a substantial improvement over the earlier Pegasus-based Advantage system [37].”

“Figure 4: Problems defined on Chimera, Pegasus, and Zephyr graphs [11, 25, 12, 26], employed in the past and current D-Wave quantum annealers, can be mapped to the Potts Hamiltonian on king’s graph upon grouping 8, 24, and 16 spins, respectively.”