Novel techniques of imaging interferometry analysis to study gas and plasma density for laser-plasma experiments
F. Filippi, M. Cipriani, S. Mastrostefano, M. Scisciò, F. Consoli · 2025
Read Novel techniques of imaging interferometry analysis to study gas and plasma density for laser-plasma experiments by F. Filippi, M. Cipriani, S. Mastrostefano, M. Scisciò, F. Consoli (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2507-13907.txt.
Introduction
Filippi and co-authors present new ways to read interferograms, the fringe patterns that reveal gas and plasma density in laser plasma experiments. Such measures do not disturb the plasma and work in a single shot, which suits laser acceleration and fusion work. Demand is rising as labs move to high repetition rates.
Setup design is tricky, and hand reading of the patterns invites error, the team notes. Their 2025 work builds machine learning routines to read the patterns fast, with no operator input. They show the network designs and how training and test data are made. Early results use synthetic data, and the study is still in progress.
Catalog record
| CREATOR | F. Filippi, M. Cipriani, S. Mastrostefano, M. Scisciò, F. Consoli |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 14 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2507-13907
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Novel techniques of imaging interferometry analysis to study gas and plasma density for laser-plasma experiments F. Filippi, M. Cipriani, S. Mastrostefano, M. Scisciò, F. Consoli Abstract Laser-plasma based experiments are always more demanding about the plasma features which need to be generated during the interaction. This is valid for laser-plasma acceleration as well as for inertial confinement fusion experiments. Most of these experiments are moving toward high repetition rate operation regimes, making even more demanding the requests on the plasma sources and the diagnostics to be implemented. Interferometry is one of the most used methods to characterize these sources, since it allows for non-perturbative, single-shot measurements either of the neutral gas or the plasma density. The design of the interferometric setup is non-trivial and needs to be shaped on the actual conditions of the experiment. Similarly, the analysis of the raw data is a complex task, prone to many sources of error and dependent on the manual inputs. In this work, we will present the techniques we are developing for the analysis of the interferograms to measure both the gas and plasma density. We will show the methods, the progress and the problems we encountered in the development of novel routines of analysis based on machine learning. The architectures and the methods to obtain data used for training and testing them will be introduced. The study is ongoing and preliminary results with synthetic data will be presented. The goal is to set up a fast and operator independent diagnostic for the feedback of plasma sources toward high repetition rate experiments. 1 Motivation Through the interference among two coherent laser beams, namely the reference and the probe beams, it is possible to measure the phase shift experienced by the probe caused by the variation of the refractive index due to the presence of plasma free-electrons or a non-ionized gas. The phase shift can be expressed by the following formula \Delta\phi=\frac{2\pi}{\lambda}\int_{L}1-N_{\lambda}(z)dz (1) where L is the length of the path that the laser has crossed through the plasma, z is the propagation axis along the probe, \lambda is the laser wavelength and N_{\lambda} is the refractive index of the sample, meanwhile the refractive index encountered by the reference beam is assumed to be 1. N_{\lambda} depends on the type of sample analyzed. In typical laser-plasma experiments this may be a non-ionized gas, in this case N_{\lambda} is function of the pressure, temperature and polarizability of the gas molecule, or a plasma, in which N_{\lambda} depends on the plasma free-electron density and on the probe laser wavelength. In imaging interferometry, the probe and the reference beams, usually originated by splitting the same laser beam, are recombined on a specific plane to create an interference pattern, called interferogram . The phase shift is made evident by the fringe deflection, which highlights the area where the different refractive index causes a phase shift in the probe. This method permits to have a quantitative measurement of the plasma electron/gas density and its spatial distribution. Still, the analysis of the raw data is a non-trivial task, prone to many sources of error and dependent on manual inputs whose consequences we tried to mitigate in other works [ 1 ] . In future laser-plasma experiments the repetition rate of the laser shots will be dramatically increased, creating a severe constraint on the imaging diagnostics and their analysis. High repetition rate requires tools able to furnish analyzed data in real-time with high affordability. At the present days, most of the used routines, based on the Fast Fourier Transform and Abel or other inversion algorithms [ 2 , 3 ] , require an action from the operator and even minor changes (e.g. the region of interest of the image) require the operator to adjust the analysis parameters, resulting in time loss and potentially causing errors. This is particularly important for the characterization of laser-produced plasma density, and it is crucial for the tuning of the laser and target properties to reach the optimal laser-plasma interaction in inertial confinement fusion, high energy density research and laser-driven plasma wakefields accelerators. In this article, we will show the progress in developing an automated tool for a fast and operator-independent analysis of the interferograms [ 4 ] . This could enable a real time feedback to help the operation of laser-plasma at high repetition rate. In this framework, we are developing a machine learning (ML) routine, which has been trained to fast analyse the interferograms and retrieve the corresponding phase shift. We will introduce the architecture we decided to use, the cGAN structure, and the methods we used to train the ML network. This architecture has been already used for the translation of aerial images into maps, or the colorization of pictures [ 5 ] . Similar approaches have been developed for scientific data analysis (e.g. [ 6 , 7 , 8 ] ). The program we have developed so far is able to retrieve the phase shift of a synthetic interferogram. We will discuss the further steps to improve it for the analysis of raw interferograms, acquired in real experimental conditions. This is the first step toward the implementation of an automated tool to retrieve information from an interferogram in a fast and robust way for high repetition rate experiments. 2 Methods: Network and Dataset Machine learning provides fast networks which map the inputs to the outputs through a process called training . In the conditional Generative Adversarial Network (cGAN) [ 5 ] two models are trained simultaneously by an adversarial process. A network called generator learns to create an image starting from an input image. A different network, called discriminator , learns to recognize real images, associated to the input image, and distinguish them from ”fakes”, actually produced by the generator (as shown in Fig. 1 ). During the training , the generator progressively becomes better in producing images similar to the expected outputs. Meanwhile, discriminator becomes better in distinguish the original outputs from the ones generated by the generator. Training is guided by loss functions that balance the ability of both the networks. Details on this function can be found in [ 5 ] . The process repeats those steps iteratively until it reaches an equilibrium when the discriminator can no longer distinguish real images from fakes. The generator we used is based on a ”U-Net” architecture [ 8 ] . The generator has to map a high resolution input image to a high resolution output image. The input is passed through a series of layers that progressively downsample until a bottleneck layer, from which the process is reversed and the output image can be progressively reconstructed through a series of similar layers. This process may cause a loss of ”low-level” information which should be better to transfer from the input to output. Then, a series of shortcuts let the information to skip some connections and to pass through corresponding layers that have the same sized feature maps. The discriminator uses a convolutional ”PatchGAN” classifier [ 5 ] . The discriminator takes the input image and an image which may be produced by the generator or be the real image expected to be associated to the input and predicts whether the