SOLIS Architecture: Integrated Scientific Systems

• SOLIS Architecture is a multi-domain framework that integrates advanced instrumentation and algorithms to support solar observations, astrochemical imaging, and autonomous lab screening.
• It employs modular designs, high-throughput data handling, and parallel processing to ensure precise calibration, robust gap correction, and reliable NLFFF extrapolation.
• Its applications span flare and CME modeling, chemical abundance diagnostics, and scalable machine learning operations, driving actionable insights across diverse scientific fields.

The acronym SOLIS refers to several distinct architectures and systems in contemporary research—ranging from solar observational platforms to machine learning and autonomous laboratory automation. Each implementation of SOLIS represents advanced system design in its domain, featuring integrated pipelines, high-throughput data handling, and rigorous algorithmic frameworks.

1. Overview and Definitions

SOLIS, depending on the context, denotes:

• Synoptic Optical Long-term Investigations of the Sun: A multi-instrument solar observing facility delivering full-disk vector magnetograms, Sun-as-a-star spectra, and high-level magnetic field products for solar physics and space weather applications.
• SOLIS in Machine Learning Operations (MLOps): A modular, cross-platform deployment pipeline for ML model serving, inference, and business logic execution within cloud or edge environments (Ciobanu et al., 2021).
• SOLIS in Laboratory Automation: An autonomous, vision-based solubility screening system using cascaded deep neural networks on robotic platforms for chemical materials discovery (Pizzuto et al., 2022).
• SOLIS in Astrochemistry: The Seeds Of Life In Space survey, designed for spatially resolved line imaging for astrochemical analysis using interferometry (Codella et al., 2021).

While these share the SOLIS label, each is independently architected to meet specific scientific or engineering demands and is characterized by precise system integration, algorithmic rigor, and adaptability.

2. SOLIS in Solar Physics: Instrumentation, Data Products, and Processing

Instrumentation and Architecture

The Synoptic Optical Long-term Investigations of the Sun (SOLIS) facility consists of three primary instruments mounted on a common pointing platform (Pevtsov et al., 2014):

• Vector Spectromagnetograph (VSM): 50-cm aperture, generates 2048×2048 full-disk spatially resolved magnetograms using curved spectrograph slits, multi-state spectropolarimetry (Stokes I, Q, U, V), and high spectral resolution (0.05 Å).
• Full-Disk Patrol (FDP): 14-cm tunable imager for contextual imaging.
• Integrated Sunlight Spectrometer (ISS): 8-mm fiber-fed double-pass spectrograph providing R ≈ 300,000 for disk-integrated Sun-as-a-star observations.

Data Acquisition and Processing

VSM scans the disk via declination stepping, forming spectral image cubes for multiple polarization states. Image processing includes handling of dual-CCD camera acquisitions, alignment, gap identification and removal using calibrated intensity thresholds and local polynomial rescaling (Marble et al., 2013), and robust Stokes inversion via Milne–Eddington atmospheric models (VFISV code) (Harker, 2017). Key architectural features include:

• Multi-threaded, MPI-parallelized computational flows for inversion.
• Automation in calibration, wavelength referencing (telluric O₂ lines), and error characterization.
• Standardized Level-1 FITS products with metadata for downstream analyses.

Integrated Data Products

SOLIS outputs include:

• Full-disk vector magnetograms.
• Synoptic vector (Carrington) maps capturing full vector field structure per solar rotation (Gosain et al., 2013).
• Polarimetric and spectral intensity cubes for the Fe I, Ca II, and He I lines.
• Sun-as-a-star integrated spectra and time series for linking spatial and global stellar features (Pevtsov et al., 2014).

3. Nonlinear Force-Free Field Modelling and Spherical Extrapolation

Physical and Mathematical Context

Routine photospheric magnetic measurements contradict the force-free assumption typical of the coronal field. SOLIS architecture supports advanced nonlinear force-free field (NLFFF) extrapolation in spherical geometry, a necessity for modeling large-scale or full-disk regions with proper solar surface curvature.

• Preprocessing Functional:

$$L_p = \mu_1 L_1 + \mu_2 L_2 + \mu_3 L_3 + \mu_4 L_4 \ L_3 = \sum_p (H - H_\text{obs}) \cdot W(\theta,\phi) \cdot (H - H_\text{obs})$$

where $H$ is the preprocessed field, $H_\text{obs}$ the measured field, and $W$ weights components by precision (Tadesse et al., 2010).

• Optimization Functional:

$L_\omega = L_f + L_d + \nu L_\text{photo}$

with

$$L_f = \int_V \omega_f B^{-2} |(\nabla\times B)\times B|^2 r^2\sin\theta\, dr\, d\theta\, d\phi \ L_d = \int_V \omega_d |\nabla\cdot B|^2 r^2\sin\theta\, dr\, d\theta\, d\phi \ L_\text{photo} = \int_S (B - H_\text{obs})\cdot W(\theta,\phi)\cdot (B - H_\text{obs}) r^2\sin\theta\, d\theta\, d\phi$$

Lagrange multiplier $\nu$ regularizes boundary relaxation, allowing the model field $B$ to approach force-free and solenoidal conditions while remaining consistent with data uncertainties; data gaps are handled by zeroing corresponding weights.

Validation and Comparison

High vector correlation ($\sim0.93$ for $B_r$) and improved force-free/solenoidal compliance are demonstrated. The estimated free magnetic energy for modeled active regions (e.g., 5×10³² erg, with NLFFF energy exceeding the potential field model by ~15%) confirms both accuracy and physical suitability (Tadesse et al., 2010, Tadesse et al., 2012). Parallel treatments with SDO/HMI and SOLIS/VSM data, after harmonizing pixelation and preprocessing, reveal consistency in coronal reconstructions and free‑energy budgets, despite systematic differences in field measurements (Thalmann et al., 2012, Tadesse et al., 2012).

Significance

The spherical optimization scheme permits accurate full-disk and large-region coronal field extrapolation—critical for flare and CME modeling. The incorporation of data uncertainty and robust gap management renders the solution well-suited for variable-quality vector synoptic products, strengthening its utility for operational and science applications.

4. Data Integration, Calibration, and Imaging Techniques

Disk-Integrated vs. Disk-Resolved Spectroscopy

SOLIS’s dual instrument design allows direct cross-comparison and modeling between spatially resolved (VSM) and Sun-as-a-star (ISS) spectra. Disk-integrated profiles are reconstructed by summing corrected disk-resolved spectra, factoring in limb darkening ($f(\phi, \mathrm{CMD})$), Doppler shifts (due to rotation), and scattered light (~10% baseline, adjusted per feature). Principal Components Analysis (PCA) reveals that two profile classes—Quiet Sun and plage—suffice for variance explanation (Pevtsov et al., 2014).

• Linear modeling relates core intensity $I_c$ to fractional area coverage by each class:

$I_c(t) = a + \sum_{j=0}^{J-1} c_j A_j(t)$

Camera Gap Correction

SOLIS/VSM dual-CCD design creates a central vertical gap with variable position and width. Gap calibration relies on intensity thresholding and local median measurements; adjacent columns are corrected by dividing by a quadratic fit to intensity ratios from unaffected reference regions, ensuring geometric and photometric continuity (Marble et al., 2013).

Astrochemical Imaging

SOLIS (Seeds Of Life In Space) in astrochemistry employs the IRAM-NOEMA array for high-resolution molecular line imaging. Systematic mapping (e.g., for S-bearing species) applies dual wavelength setups and LVG (large velocity gradient) non-LTE radiative transfer analysis, enabling spatial disentanglement of core, envelope, and jet components and measurement of physical conditions and abundance ratios (Codella et al., 2021).

5. SOLIS for Machine Learning Operations (MLOps) and Autonomous Laboratory Screening

Unified MLOps Pipeline Architecture

SOLIS as an MLOps deployment pipeline implements a sequential, parallelizable flow (Ciobanu et al., 2021):

• Configuration and Data Acquisition: External JSON configurations and protocol-agnostic communications (MQTT, HTTP, AMQP) for online/offline and multi-modal data streams.
• Inference Engine: Framework-agnostic serving (Tensorflow, PyTorch, etc.), parallel DAG execution, and multi-process isolation for robust GPU memory usage:

$$T_I = \max(T_{DAG_1}, T_{DAG_2}, \ldots, T_{DAG_N}) + \varepsilon$$

• Business Logic Plugins: Low-code/no-code Python scripting for post-processing, aggregation, and automated dispatch to APIs or IoT stacks.

Autonomous Vision-Based Laboratory Screening

SOLIS as an autonomous solubility screening platform leverages a three-stage architecture (Pizzuto et al., 2022):

1. Image Acquisition: Eye-in-hand RGB camera on a 7-DOF robotic manipulator (Franka Emika Panda) for flexible sample imaging.
2. Segmentation: Mask R–CNN (ResNet backbone) for vial region of interest extraction.
3. Classification: Deep CNN (ResNet, VGG, Inception, Densenet architectures, transfer learning, feature extraction/fine-tuning) for dissolved/undissolved binary prediction.

Empirical results using real-lab datasets show 99.13% test accuracy in solubility determination, confirming the effectiveness of the cascaded segmentation–classification pipeline, even under variable illumination and scene clutter.

6. Applications and Scientific Impact

SOLIS architectures advance research capabilities in multiple domains:

• In solar physics, they produce physically consistent global magnetic field models, enable synoptic mapping of current helicity (including hemispheric and field-strength–dependent variations (Gosain et al., 2013)), and support multi-instrument calibration and cross-validation.
• In astrochemistry, spatially resolved line imaging with high spectral resolution enables detailed studies of chemical evolution in star-forming regions, abundance diagnostics, and insight into planetary system formation environments (Codella et al., 2021).
• In computational sciences and laboratory automation, modular pipelines for scalable deployment, GPU resource optimization, and autonomous real-world vision systems contribute to the development of reliable, autonomous experimentation platforms with rigorous performance guarantees (Ciobanu et al., 2021, Pizzuto et al., 2022).

The design choices—rigorous uncertainty quantification, flexible data/model handling, computational parallelization, and robust calibration—render SOLIS frameworks exemplary for integrated, high-throughput scientific discovery pipelines across physics, chemistry, and engineering.