Reconstructing Quantum Dot Charge Stability Diagrams with Diffusion Models

Abstract

Efficiently characterizing quantum dot (QD) devices is a critical bottleneck when scaling quantum processors based on confined spins. Measuring high-resolution charge stability diagrams (or CSDs, data maps which crucially define the occupation of QDs) is time-consuming, particularly in emerging architectures where CSDs must be acquired with remote sensors that cannot probe the charge of the relevant dots directly. In this work, we present a generative approach to accelerate acquisition by reconstructing full CSDs from sparse measurements, using a conditional diffusion model. We evaluate our approach using two experimentally motivated masking strategies: uniform grid-based sampling, and line-cut sweeps. Our lightweight architecture, trained on approximately 9,000 examples, successfully reconstructs CSDs, maintaining key physically important features such as charge transition lines, from as little as 4\% of the total measured data. We compare the approach to interpolation methods, which fail when the task involves reconstructing large unmeasured regions. Our results demonstrate that generative models can significantly reduce the characterization overhead for quantum devices, and provides a robust path towards an experimental implementation.

Figures (9)

• Figure 1: Charge stability diagram of a (gate-defined) double quantum dot.(a) Left: schematic of a double quantum dot. Right: (simulated) charge stability diagram as a function of two gate voltages. Colored regions correspond to fixed charge occupations $(N_1, N_2)$, while sharp transition lines indicate changes in the charge configuration. These transition lines encode the essential information required for device tuning (but are costly to acquire experimentally). (b) Measurement workflow. (c) Diffusion model reconstruction process.
• Figure 2: Training loss over 30 epochs for grid masks (top) and line-cut masks (bottom) at different sparsity levels using 140 diffusion steps. The solid line shows the average training loss per epoch, while the shaded region indicates the training loss variation across training steps (mini-batches), and the square markers show validation loss for selected epochs. Reconstructions at the right side correspond to epochs 1, 10, 20, and 30 for both masking strategies.
• Figure 3: Reconstruction results for the grid mask strategy, presented for 140 diffusion steps and a selected example from the test set. Columns show different reduce factors. From top to bottom, the rows show: masked measurement, original full measurement, diffusion reconstruction, biharmonic interpolation reconstruction, ridges detection for diffusion model compared to original and ridges detection for biharmonic interpolation compared to original. Predicted ridges matching original are shown in white, ridges that are detected by the prediction but not the original are shown in blue, ridges that are present in the original example but missing in the prediction are shown in red.
• Figure 4: Reconstruction results for the line-cut mask strategy, presented for 140 diffusion steps and a selected example from the test set. Columns show different line-cut factors. From top to bottom, the rows show: masked measurement, original full measurement, diffusion reconstruction, biharmonic interpolation reconstruction, ridges detection for diffusion model compared to original (in white predicted ridges match original, in blue ridges are detected by the prediction but not the original, and in red ridges are present in the original example but missing in the prediction), and ridges detection for biharmonic interpolation compared to original.
• Figure 5: Reconstruction results for grid mask case, for reduce factor (3,5,7,9), using diffusion models with number of diffusion steps (20,60,100,140).