Overview
Dielectric and optoelectronic materials must satisfy coupled—and often competing—requirements across polarization, band structure, absorption, transport, stability, process compatibility, and safety. CD ComputaBio combines curated materials data, physics-informed machine learning, and targeted first-principles calculations to move from a broad chemical space to a short list of testable candidates.
Projects can begin with a composition family, crystal structures, measured spectra, device-level targets, or a partially characterized internal dataset. We define the operating condition first: frequency range, temperature, electric field, wavelength window, film or bulk form, substrate, and fabrication constraints. Predictions are reported with method provenance, uncertainty, and an applicability-domain assessment rather than treated as experimentally verified performance.
Core Services
Dielectric Tensor Prediction
Calculate or learn electronic and ionic contributions to the static dielectric tensor, refractive index, anisotropy, and frequency-dependent response. DFPT calculations are prioritized for high-value insulating candidates and interpreted in the context of crystal symmetry and phonon stability.
Optical Response Modeling
Evaluate complex dielectric functions, absorption coefficients, reflectivity, refractive indices, and polarization dependence across a defined spectral window. Higher-fidelity methods can be introduced where semilocal DFT band-gap bias would alter candidate ranking.
Band Alignment and Excitation Screening
Screen direct versus indirect gaps, orbital character, effective masses, vacuum-referenced band edges, and interface alignment. Hybrid DFT, spin–orbit coupling, GW, or excitonic treatments are scoped when chemistry and the required decision justify their cost.
AI-Guided Candidate Discovery
Build composition- and structure-aware models for permittivity, band gap, absorption onset, refractive index, or device-relevant composite objectives. Nested validation, calibration, uncertainty, and domain checks are included for small or heterogeneous datasets.
Ferroelectric and Polar Materials
Assess symmetry breaking, spontaneous polarization, competing phases, soft modes, and approximate switching pathways. Results distinguish intrinsic bulk tendencies from thin-film, domain, electrode, and processing effects that require experimental confirmation.
Multi-Objective Materials Ranking
Balance dielectric or optical response with thermodynamic stability, synthesizability, abundance, toxicity flags, defect sensitivity, temperature range, and process constraints. Pareto analysis makes trade-offs visible instead of collapsing them into an opaque score.
Integrated Project Workflow
| Stage | Key Activities | Decision Output |
|---|---|---|
| 1. Scope and target definition | Define material class, operating conditions, device context, exclusion rules, acceptable computational cost, and experimental decision thresholds. | Project specification and success criteria |
| 2. Data assembly and curation | Normalize structures, compositions, units, polymorph identity, temperature/frequency metadata, calculation settings, and experimental provenance. | Traceable, analysis-ready dataset |
| 3. Baseline screening | Apply stability, gap, chemistry, symmetry, toxicity, and manufacturability filters; establish simple and interpretable baselines. | Feasible candidate space |
| 4. AI prediction and uncertainty | Train structure/composition models, use grouped or nested validation, calibrate uncertainty, and flag out-of-domain predictions. | Prioritized candidates with confidence tiers |
| 5. Physics-based refinement | Run targeted relaxations, DFPT, band-structure, optical, phonon, interface, or higher-level electronic-structure calculations. | Validated property and stability evidence |
| 6. Multi-objective ranking | Integrate performance, robustness, cost, sustainability, and synthesis constraints; perform sensitivity analysis. | Shortlist and experimental validation plan |
Decision-Ready Deliverables
Every package is configured around the project question and may include:
- Curated structures, compositions, metadata, and provenance records
- Predicted dielectric, optical, electronic, and stability properties
- Tensor components, spectra, band structures, and orbital analyses
- Model cards, validation results, uncertainty, and applicability domain
- Ranked candidates, Pareto fronts, and sensitivity analyses
- Reproducible calculation settings and machine-readable result tables
- Technical report with assumptions, limitations, and interpretation
- Prioritized synthesis and characterization recommendations
Applications
High-κ and Low-Loss Dielectrics
Candidate selection for gate stacks, capacitors, embedded passives, RF components, and insulating layers under explicit gap, loss, breakdown, and processing constraints.
Photovoltaic Absorbers
Screen absorbers and transport layers for appropriate gaps, absorption strength, carrier transport, band alignment, stability, and restricted-element criteria.
LED and Detector Materials
Prioritize direct-gap semiconductors, emitters, scintillators, and photodetector materials by spectral response, effective masses, dielectric screening, and defect tolerance.
Integrated Photonics
Evaluate refractive index, birefringence, transparency window, anisotropy, and electro-optic potential for waveguides, modulators, filters, and nonlinear optics.
Ferroelectric Electronics
Explore polar oxides and related compounds for memories, sensors, actuators, and negative-capacitance concepts while accounting for phase competition and switching barriers.
Thin Films and Interfaces
Assess strain, orientation, substrate compatibility, band offsets, interfacial polarization, and likely defect effects before film growth or stack fabrication.
Scientific Evidence
High-throughput studies demonstrate that dielectric tensors, refractive indices, band gaps, and frequency-dependent dielectric functions can be calculated at database scale and used to prioritize materials for targeted refinement.1,2 For perovskite-related systems, machine learning is most useful when thermodynamic stability, structural descriptors, and higher-fidelity electronic calculations are integrated instead of optimized independently.3
1 Petousis, I.; Mrdjenovich, D.; Ballouz, E.; et al. High-throughput screening of inorganic compounds for the discovery of novel dielectric and optical materials. Scientific Data 2017, 4, 160134. https://doi.org/10.1038/sdata.2016.134. (Open Access, CC BY 4.0)
2 Choudhary, K.; Zhang, Q.; Reid, A. C. E.; et al. Computational screening of high-performance optoelectronic materials using OptB88vdW and TB-mBJ formalisms. Scientific Data 2018, 5, 180082. https://doi.org/10.1038/sdata.2018.82. (Open Access, CC BY 4.0)
3 Tao, Q.; Xu, P.; Li, M.; Lu, W. Machine learning for perovskite materials design and discovery. npj Computational Materials 2021, 7, 23. https://doi.org/10.1038/s41524-021-00495-8. (Open Access, CC BY 4.0)
The cited works are used for scientific context; no endorsement is implied.
Why Work with CD ComputaBio?
Our modular workflow connects fast screening with the minimum sufficient physics needed for the next decision. Data provenance, model limitations, uncertainty, and validation handoffs remain visible from initial scope to final shortlist. To discuss a material family, target spectrum, device stack, or internal dataset, please contact us or submit the Online Inquiry form below.
