ISAE-SUPAERO · Université de Toulouse

Materials & the Environment

An open, interactive book on choosing what things are made of when mass, cost and the planet all sit on the axes. From Ashby's charts to learned latent spaces, eco-indicators, 1 516 simulated jet-engine brackets and a small language model taught by the book itself.

Start reading Chapter 1 Table of contents GitHub
Drag the guideline: E⅓/ρ, the light-panel indexBest now: —

In memoriam

Pierre Morlier

Died 29 October 2025

This book is dedicated to my dad, Pierre Morlier, professor, engineer and researcher in wood science and structural mechanics. A graduate of the École des Mines de Paris, then researcher at the École Polytechnique, he worked at the Laboratoire de Mécanique des Solides, then moved to the University of Bordeaux. Founder of the Laboratoire de Rhéologie du Bois de Bordeaux, his studies across civil engineering and wood sciences (timber creep, tropical wood density and hardness, and historic timber structures) became references across Europe. A pioneer in multidisciplinary research, he helped organise the XYLOFUTUR competitiveness cluster, led the European COST E8 action on forests and forestry products, and mentored generations of students with patience, humility and integrity.

Why wood matters here. Wood is the original eco-material, and the thread that runs through this book. It grows from sunlight, and about half of its dry mass is carbon drawn from the atmosphere, so a building or a product made of wood stores carbon for as long as it lasts. It needs little energy to process compared with metals, polymers or glass, it can be reused, recycled or burnt for energy at the end of its life, and along the grain its stiffness per unit weight rivals that of steel. On the Ashby charts of Chapter 1, softwood, hardwood and plywood come out on top for light, stiff panels; Chapter 5 follows the same logic to the plant fibres of flax, hemp and jute. Making wood and its composites reliable for engineers, through their creep, variability and durability, was Pierre's life's work.

One problem, two objectives

The environmental impact of a product is largely determined in its earliest stages, using graphs and rules of thumb confined to proprietary software and static textbooks. Materials science education needs open and usable textbooks that present data, code, and environmental impact on a single page. Don't you agree? Behind every chapter sits the same optimisation problem. A part is described by its shape \mathcal{D}, its material m and its manufacturing process p. Classical lightweight design, problem \mathcal{P}_1, minimises its mass under a stiffness constraint. Eco-design, problem \mathcal{P}_2, keeps the same design space and the same constraint but changes the objective to the CO2 footprint of producing and processing the part. A set of new constraints could be added and represent the sizing (stress, buckling etc...) and manufacturing loop.

\begin{array}{llcll} \mathcal{P}_1:\ \displaystyle\min_{m,\,p,\,\mathcal{D}} & \operatorname{mass}(\mathcal{D}, m, p) & \quad\longrightarrow\quad & \mathcal{P}_2:\ \displaystyle\min_{m,\,p,\,\mathcal{D}} & \mathrm{CO}_2^{\mathrm{PP}}(\mathcal{D}, m, p)\\[6pt] \quad\text{s.t.} & \delta \le \delta_{\max} & & \quad\text{s.t.} & \delta \le \delta_{\max}\\[2pt] & \{m, p\} \in \Phi_{\text{material}} \times \Phi_{\text{process}} & & & \{m, p\} \in \Phi_{\text{material}} \times \Phi_{\text{process}} \end{array}
\begin{array}{ll} \mathcal{P}_1:\ \displaystyle\min_{m,\,p,\,\mathcal{D}} & \operatorname{mass}(\mathcal{D}, m, p)\\[2pt] \text{s.t.} & \delta \le \delta_{\max}\\[2pt] & \{m, p\} \in \Phi_{\text{mat}} \times \Phi_{\text{proc}}\\[4pt] & \Big\downarrow\\[4pt] \mathcal{P}_2:\ \displaystyle\min_{m,\,p,\,\mathcal{D}} & \mathrm{CO}_2^{\mathrm{PP}}(\mathcal{D}, m, p)\\[2pt] \text{s.t.} & \delta \le \delta_{\max}\\[2pt] & \{m, p\} \in \Phi_{\text{mat}} \times \Phi_{\text{proc}} \end{array}
Mixed and discreteThe material and the process are chosen from finite catalogues \Phi_{\text{material}} and \Phi_{\text{process}}, while the shape \mathcal{D} is continuous: Ashby charts (Chapters 1–3) screen the catalogues, topology optimisation (Chapters 4–5) shapes the part.
Same constraint, new objectiveThe displacement limit \delta \le \delta_{\max} is unchanged. Only what is minimised changes, and with it the answer: the lightest design is rarely the one with the lowest footprint.
Whole life nextReplacing production CO2 by a life-cycle total, as in the eco audit of Chapter 6, adds the use phase, where for anything that moves mass comes back into the objective.

The book

Six chapters, each a Distill-style article with live figures you can drag, filter and re-run. They read in order: each one lifts a limit of the one before.

Chartone index, two properties
Value functionmany objectives, one score
Eco-axisenvironment priced in
Latent spaceany index, solved by search
Natural fibresrenewable reinforcement, charted
Eco auditwhole life, five phases

  1. Ashby's Maps

    Charts, value functions and the eco-cost axis

    • 1.1 A live Ashby chart with a draggable guideline
    • 1.2 Exchange constants and the trade-off surface
    • 1.3 An eco-Ashby chart
    Joseph Morlieroriginal article

  2. Ashby's Maps, Renewed

    Selecting materials in a learned latent space

    • 2.1 The VAE and its training
    • 2.2 Gradient walks through latent space, in your browser
    • Tables: dataset and 16 extended cases
    J. Morlier, A. Cobo-Urios, Á. Silva-Vilela-Caridadeoriginal article

  3. Mapping the Ecolizer

    Eco-indicators, eco-indices and a masked VAE

    • 3.1 Ten orders of magnitude of production impact
    • 3.2 Eco-index per unit of stiffness
    • 3.3 Production route against electricity mix
    • 3.5 What imputation reveals
    mid2SUPAEROoriginal article

  4. Brackets, Metals and Carbon

    The TOSM benchmark of 1 516 bracket–alloy pairs

    • 4.1 Production against use-phase CO₂
    • 4.3 Utilisation and the safety factor
    • 4.4 An exact Pareto explorer, and a constraint sign bug
    R. Maeso Orti, P. Ballesteros Martín, J. Morlieroriginal article

  5. Natural Fibre Composites

    Ashby charts for plant fibres and their composites

    • 5.1 Twelve plant fibres against E-glass
    • 5.3 35 composites by stiffness, weight, cost and energy
    • 5.5 Fatigue strength at a million cycles
    after D. U. Shah (2014)original articlenotebook

  6. A Simple Eco Audit

    Energy and CO₂ over five life phases

    • 6.3 An eco audit calculator with seven products
    • 6.5 PET versus a reusable bottle
    • 6.6 Steel or aluminium bumper: the break-even
    after Ashby et al.notebook

Backup chapters

Earlier material kept for reference, outside the main sequence.

  1. Ask the Virtual Prof

    A chatbot that answers from Chapters 1–6

    • A.1 Ask a question, get cited passages or an explanation
    • A.2 Retrieve, then explain: how it works
    • A.3 Inside the knowledge base
    J. Morliermodels on Hugging Face

The lab

Materials & the Environment gathers teaching and research on sustainable structural design at ISAE-SUPAERO, Toulouse. Students and researchers build open notebooks and datasets; this site turns them into readable, explorable chapters.

Material selection

Ashby charts, performance indices and multi-objective value functions, extended with eco-costs and eco-indicators.

Machine learning for design

Variational autoencoders that turn material databases into continuous, searchable spaces, with honest tests of what they cannot predict.

Ecodesign & topology optimisation

Shape and material chosen together, with lifetime carbon, energy and water attached to every design.

Code and data

RepositoryWhat it holdsUsed in
MaterialsandtheEnvironmentThis website and book (GitHub Pages)All
mid2SUPAERO/TOSMFinalResults.xlsx, Materials.xlsx, OptiStruct and Pareto notebooksCh. 4
mid2SUPAERO/HybML-EvoMatDesEcoEco-material database and hybrid VAE + evolutionary design codeCh. 2, 4
mid2SUPAERO/ashby-vaeAshby-VAE case-study notebooksCh. 2
MaterialsandtheEnvironment on Hugging FaceVirtual Prof models and knowledge base; earlier fine-tuned SmolLM2Backup A
Shah (2014), supplementary tablesPlant fibre, PFRP and fatigue data, as data/nfc.jsCh. 5
ecolizer_vae.ipynbDigitised Ecolizer 2.0 tables and the masked VAECh. 3

People