Nature of Materials, CAE

THE NATURE OF MATERIALS

Part One

Understanding Material Character for Better Simulation

For any finite element tool

Developed in collaboration with Claude (Anthropic)

START HERE.

This is where we begin.

Not with keywords. Not with unit systems. Not with which flags to set in which solver. We begin with the materials themselves — what they actually are, how they behave, what kind of character each one has — and why understanding that character is the foundation of every honest simulation you will ever build.

This audiobook is Part One of two. It applies to every finite element tool ever written — Abaqus, Ansys, Nastran, LS-Dyna, Simcenter, Radioss, any of them. The mathematics differs between solvers. The keywords differ. The material card interface differs. The nature of the material does not differ. Steel is steel in every solver. Polycarbonate is polycarbonate. Glass is glass. And understanding what these materials actually are — not just what numbers describe them — is what determines whether your simulation is a faithful representation of physical reality or an expensive fiction.

Part Two — the technical guide for Abaqus and other solvers — will follow. You can go there next, or you can return to it whenever the need arises. But come here first. Because when you understand the nature of what you are trying to model, the technical requirements in Part Two stop being a checklist and start being an expression of that nature. The density requirement is not bureaucracy. It is capturing presence. The rate-dependent plasticity flag is not a warning box. It is the tool telling you the material's patience changes under different conditions and you have not told it how. Everything in Part Two means more once you have spent time here.

WE UNDERSTAND THINGS BY POKING THEM.

Here is the most fundamental thing I can tell you about simulation, and about engineering, and honestly about understanding anything at all.

We understand things by poking them.

You apply a force — an impact, a vibration, a temperature change, a sustained load — and the system responds. But how it responds depends entirely on what it is. A block of steel responds differently from a rubber pad. A thin glass panel responds differently from a thick aluminum chassis. And the same material responds differently depending on how it was made, what it has been through, and how fast you poke it. The response comes from the nature of the thing. And that nature — that character — is what we are trying to capture when we define a material in a simulation model.

Most FEA training teaches you what to type. Enter Young's modulus here. Enter density here. Add a plasticity curve. Tick the rate-dependent box. But it does not always explain what those numbers mean. What character trait each one expresses. What happens to the simulation's honesty when one of them is wrong — not in a units sense, but in a physical sense. When the number is correct for the test specimen but wrong for the part as it actually exists in the product.

That is what this discussion is about.

MATERIAL SELF-KNOWLEDGE. THE CHARACTER BEHIND THE NUMBERS.

Before a part can interact with anything else in a simulation — before it can push against a floor, load a PCB, transmit force through a bond — it must know itself. And in finite element analysis, self-knowledge means material properties.

Think about the human parallel. A person's response to any situation depends fundamentally on their own nature. Are they resilient or brittle? Do they bend under pressure or snap? Can they absorb energy or do they transfer it instantly to whoever is nearby? Do they behave the same way under gentle pressure as they do under extreme stress? Do they change as they age, as they absorb what the environment gives them, as they carry residual stress from what they have been through?

Every one of these human qualities has a direct material property counterpart. And understanding the correspondence — deeply, not just metaphorically — changes how you approach every material card you ever write.

Young's modulus is confidence.

It determines how much a material resists deformation under load. A high modulus material like steel stands firm. It takes enormous force to change its shape. A low modulus material like rubber yields readily. Neither is better. They simply have different roles in the assembly. Steel is the confident, immovable backbone. Rubber is the adaptable, compliant interface that lets the backbone tolerate misalignment without fracturing something rigid. In an assembly, you need both. The question is whether the modulus value in your model reflects the actual confidence of the material in the actual part. A polymer housing molded with a gate position that created glass fiber alignment perpendicular to the primary load direction can have an effective modulus in that direction 30 to 60 percent or more lower than the isotropic datasheet value. Its confidence in the direction that matters most has been reduced by the very process that created the part.

Poisson's ratio is empathy.

When you compress a material in one direction, it expands in the perpendicular directions. Poisson's ratio quantifies how much a material's response in one axis is affected by what is happening in another. A material with a Poisson's ratio near 0.5 — like rubber — is highly empathetic. Squeeze it here, it bulges there. Every action creates an equal and visible reaction elsewhere. A material like cork, with a Poisson's ratio near zero, is stoic. Compress it and the deformation stays local. It does not transfer its burden outward. This is why cork works for wine bottle stoppers — a highly empathetic material would transmit the seating force to the glass neck and risk cracking it. In FEA, near-incompressible materials approaching a Poisson's ratio of 0.49 or above require special element formulations, because that high empathy creates volumetric coupling that standard displacement elements cannot handle correctly. Note that only exactly 0.5 produces a mathematically singular bulk modulus — values like 0.495 or 0.499 are physically valid and acceptable when used with the appropriate hybrid element formulation. Rubber in standard elements locks up regardless. Its empathy cannot be expressed through conventional displacement interpolation.

Yield strength is patience.

It is the threshold beyond which a material's behavior permanently changes. Below yield, the material is elastic. It springs back. It forgives. Load it, release it, it returns to where it was. Above yield, the material plastically deforms. It remembers the insult. The damage is done. In human terms, everyone has a point beyond which the relationship changes permanently. Knowing where that threshold lies — for both the material and the person — is critical to predicting outcomes. And just as some people become more patient under certain conditions and less under others — calmer in familiar environments, quicker to break under sustained pressure or cold — yield strength changes with temperature, with strain rate, and with the history of what the material has already been through. A material that has been partially yielded carries residual stress and a shifted yield surface. It has already spent some of its patience. It will yield again at a lower additional load than it did the first time if the loading direction changes.

Density is presence.

It determines how much mass occupies a given volume. And in dynamic analysis, mass is everything. A dense part has inertia. It resists acceleration. It dominates vibratory response. In an assembly under impact loading, the densest parts are the heaviest voices in the room — they set the pace of wave propagation, they determine how much of the impacting energy is absorbed locally versus transmitted to neighboring components. A PCB mounted on a magnesium chassis responds differently from the same PCB on a polycarbonate chassis, even with identical geometry and contact conditions, because the chassis presence — its density, its inertia, its mass — is different. When density is wrong in the model, the dynamics are wrong. The voice of that part in the assembly is speaking at the wrong pitch.

THE STRESS-STRAIN CURVE IS THE MATERIAL'S BIOGRAPHY.

The stress-strain curve is not just a graph. It is the biography of the material under load. It tells you where it lives comfortably, where it begins to struggle, when it crosses a point of no return, and how it carries itself under the accumulated weight of what it has been through.

But here is something that is almost never said clearly in FEA training. The stress-strain curve you measure in the laboratory and the stress-strain curve that belongs in your FEA model are not the same curve. They look similar at small strains. They diverge meaningfully at large strains. And understanding why they diverge — and what that divergence means about the material's true nature — is fundamental to honest simulation.

A standard tensile test records engineering stress and engineering strain. Engineering stress is force divided by the original cross-sectional area. Engineering strain is the change in length divided by the original gauge length. These are easy to measure. They are also a polite fiction at large strains. As the specimen stretches, the cross-section shrinks — the material has Poisson contraction, it empathetically redistributes its volume. The actual stress on the material — force divided by the current area — is higher than the engineering stress once meaningful contraction has occurred. The actual strain the material has experienced — accumulated as the natural log of incremental stretch ratios — is also different from the engineering strain. True stress and true strain are what is physically happening to the material. Engineering stress and engineering strain are what the test machine conveniently measures. Your FEA solver works in true stress and true logarithmic strain. Feed it engineering values and you are giving it a biography with the facts changed.

Now the monotonic condition — and I want you to understand this deeply, not just as a rule to follow.

A material has character. Its stress-strain curve expresses that character. And most FEA plasticity models — in every solver — are built on a specific assumption about that character: that the material gets stronger as it deforms plastically. Each increment of plastic strain is accompanied by an increment of true stress. The tangent modulus — the slope of the true stress-true strain curve at any point — is positive. Always. This is the monotonic condition. It is not just a mathematical requirement of the solver. It is a statement about the physical stability of the deforming element. In the framework developed by Drucker, a stable material element releases more energy to its surroundings than it absorbs — it hardens as it deforms. A material that softens — that becomes weaker as it deforms — concentrates deformation rather than spreading it. It forms a band. It localizes. It fails in a way the standard hardening framework was never designed to capture.

For most structural metals, the true stress-strain curve is genuinely monotonically increasing right through necking and up to fracture. The engineering curve shows a descending branch after the ultimate tensile strength — but that is because you are dividing by the original area rather than the shrinking current area. The true stress continues to rise. The material is still hardening. The geometry is softening — the neck is forming — but the material at the neck is not. This distinction between geometric softening and material softening is one of the most important in all of structural mechanics. If you enter the engineering stress-strain curve directly into your plasticity model without converting to true values, the descending post-ultimate branch violates the monotonic condition, and the solver either flags it or silently misbehaves at strains beyond that point. The material's biography has been mistranslated. The solver is reading fiction as fact.

Polymers are more complex and more physically interesting. Many semicrystalline thermoplastics — nylon, polypropylene, polyethylene — show a pronounced upper yield point. At that point, the crystalline structure begins to break down. The polymer chains start to disentangle from their ordered arrangement. And the material genuinely, physically softens — the true stress drops as plastic strain increases through that transition zone. This is not a measurement artefact. It is the material's character changing. It is patience running out in a particular way — not a sudden brittle snap, but a deliberate yielding of structure followed by a new equilibrium: cold drawing. In cold drawing, uncoiled chains align in the loading direction, and a long plateau of large plastic strain at nearly constant stress carries the material to fracture. The descending branch from upper yield to lower yield plateau violates the monotonic condition in any standard isotropic hardening model. The standard approach is to regularize: replace the descending branch with a horizontal step directly from the upper yield point to the lower yield plateau. Understand what you are doing when you regularize. You are not modelling the strain-softening instability that drives shear band formation in the real material. For most engineering drop test applications where you need total energy absorption and peak force, the regularized curve is adequate. For analyses where localization and shear banding are the actual failure mechanism, it is not — and you need constitutive models with damage mechanics and regularization length-scale parameters.

One more implication worth stating clearly: extrapolation. Your plasticity data ends at the last test point — the strain at fracture in the tensile test. In an explicit dynamic simulation, local strains at impact zones and stress concentrations can exceed that test strain. The solver extrapolates — extending the last tangent slope forward into territory the test never covered. Check the maximum plastic strain in your results against the maximum strain in your input data. If elements are operating beyond your data range, the extrapolated response is an assumption. Own it. Document it. Do not let it hide.

THE NATURE OF EACH MATERIAL FAMILY.

Let us walk through each material family and ask not what the numbers are, but what the material's nature is. What kind of character are we trying to capture? And what does that character demand from the model?

Metals — the confident, history-carrying materials.

Metals are confident — high Young's modulus, high yield strength. They resist deformation strongly and yield only after sustained effort. But metals carry their history. A metal that has been cold-formed remembers the plastic strain it absorbed during forming as a shifted yield surface and a hardened microstructure. A metal that has been welded carries a heat-affected zone where its confidence has been locally reduced by thermal annealing. A die-cast metal carries porosity from its birth in the mould — microscopic voids that reduce ductility and fatigue patience without changing the nominal yield stress. The character of a metal is inseparable from its manufacturing biography. The alloy name tells you the chemistry. The temper designation tells you the current state of the character given that chemistry. The manufacturing process tells you what local deviations from that character exist in the real part. Metals are also rate-sensitive in a way that reflects their atomic-scale deformation mechanisms. At higher loading rates, dislocations have less time to move and multiply. The material resists more. Its patience under rapid loading is higher — it becomes temporarily more confident under pressure. This rate sensitivity is not a peculiarity to be corrected for. It is the material's natural response to being poked quickly. Capture it and the model reflects the real character. Ignore it and the model describes a metal that is the same in a crash test as in a slow-speed press — which it is not.

Polymers — the rate-sensitive, moisture-absorbing, history-dependent materials.

Polymers are deeply sensitive to conditions in a way that no metal is. Their confidence changes by a factor of two or more with moisture content. Their patience changes with temperature and rate in ways that can shift the failure mode entirely — from ductile cold drawing to brittle fracture — as a function of how fast you load them or how cold the environment is. Their history — weld lines from injection moulding, fibre orientation from the flow pattern, residual stress from differential cooling — can reduce their strength at specific locations by 30 to 70 percent without any change in the bulk material numbers. A polymer is the most context-dependent structural material in common use. Its character in the finished part is a product of the material's chemistry, the mould designer's decisions, the process engineer's settings, and the environment the product will live in. The datasheet captures the chemistry in a standard specimen under standard conditions. The other factors are invisible in the datasheet. They must be known by the analyst. The upper yield point and cold drawing of semicrystalline polymers is one of the most physically interesting phenomena in structural materials — a genuine phase transition under mechanical stress, the crystalline order breaking down and chains realigning in a lower-energy configuration. Standard FEA plasticity assumes the material's mechanism does not change. Here it does. Regularizing the data is the engineering workaround. Understanding why you are regularizing is the difference between applying a rule and understanding the physics.

Glass — the patient accumulator that fails without warning, always at its weakest point.

On the macroscopic scale that matters for engineering simulation and drop testing, glass does not yield. It remains linear elastic right up to fracture. There is no plastic deformation, no cold drawing, no forgiveness once the critical stress is reached. This is glass's defining character: enormous confidence paired with zero macroscopic ductility.

But I want to be careful here, because I do not promote black boxes and a simplification should be understood as a simplification. At the atomic scale, all materials yield to some degree. Even glass experiences localized bond breaking and rearrangement at crack tips. Glass technically flows over very long timescales — it is a supercooled liquid, and old cathedral windows are sometimes thicker at the bottom because of centuries of slow viscous movement. The difference between glass and a ductile metal is not that glass is incapable of any irreversible atomic rearrangement. The difference is that in glass these microscopic processes do not produce meaningful plastic flow before catastrophic failure at engineering timescales and temperatures. The material does not spread the damage — it concentrates it. One flaw under sufficient tension becomes a running crack before any other region has time to redistribute load plastically. For engineering simulation, we model glass as linear elastic to fracture — and that is the correct approximation. But know that it is an approximation, and know why it is valid.

This brings us to glass's deepest truth: it fails at its weakest link. The fracture strength of a glass panel is not determined by the average strength of the material. It is determined by the severity of the worst surface flaw present under tension at the moment of loading. A pristine chemically strengthened panel may withstand 700 to 800 megapascals. The same panel with a single five-micron scratch from a key in your pocket may fail at 150 to 200 megapascals. Not because the material changed chemically. Because the flaw changed the local stress intensity — the crack tip stress relative to the material's fracture toughness. This is the Griffith criterion, and it means that glass strength is inherently statistical. The same panel, the same load, different surface histories — different outcomes. This enormous variability is why glass strength follows Weibull statistics rather than a single deterministic value. The surface history — invisible scratches, handling marks, manufacturing imperfections — writes the final chapter of the glass's patience. Chemically strengthened glass adds a powerful defense: the ion-exchange process creates deep surface compression of 700 to 900 megapascals. The glass spends patience in advance so that external tension must first overcome that built-in compression before the surface goes into net tension. Modelling glass as linear elastic with a fixed fracture strength is honest only if you remember that the real strength lives in the flaws and the surface history, not in the bulk material alone.

Elastomers — the empathetic, compliant adaptors.

Elastomers have low confidence, enormous empathy approaching the incompressible limit, and extraordinary patience — they stretch to several times their original length and return to the original state. Their patience is energetic, not structural: stored strain energy, not plastic deformation. The stress-strain biography of an elastomer is a curve — nonlinear from the start, stiffening at large strains as the polymer chains approach full extension. There is no yield point in the traditional sense because the deformation is always recoverable until rupture. The hyperelastic model is not an approximation of elastomer behaviour — it is the mathematically correct description of it. Strain energy density as the fundamental quantity, stress derived from it. Any other approach is using the wrong language to describe the material's character.

Foam — the compressible, absorbing, rate-sensitive cushion.

Foam's character is defined by its architecture as much as its constituent material. The three-regime stress-strain biography — gentle elastic cell bending, the long energy-absorbing plateau of cell buckling and collapse, the stiffening densification as collapsed cells contact each other — is the story of that architecture progressing through three distinct mechanical states. A FEA model that uses a linear elastic material for foam is saying the architecture never changes under load. It is wrong by design. The plateau regime is where foam does its work in a drop test — absorbing energy at nearly constant force, protecting what is behind it. Miss the plateau, miss the engineering purpose of the foam.

RATE AND TEMPERATURE. WHEN CHARACTER CHANGES.

We said yield strength is patience. But patience is not fixed. A person who is normally patient can lose patience under sustained pressure, in extreme cold, or when events happen faster than they can process. Materials behave the same way.

Rate changes character. Metals poked quickly are more confident and more patient — higher modulus and higher yield. Polymers poked quickly become more confident but less patient in ductile terms — they may switch from cold drawing to brittle fracture. The same polycarbonate that absorbs enormous energy at slow rates can fracture without meaningful yield at very high impact rates. Its character has not changed permanently — put it back at slow rates and it is ductile again. But in the moment of the drop test, its character is different from what the datasheet describes at quasi-static rates. The model must reflect the character the material actually has at the rate of the event. Rate and temperature effects can also change the monotonic character of the stress-strain response qualitatively, not just quantitatively. A polymer that is monotonically hardening at low rates may show a pronounced upper yield drop at high rates. A metal that is monotonically hardening at room temperature may show thermal softening and localization at very high rates where adiabatic heating is significant. These are not just quantitative shifts in parameter values. They are qualitative changes in the nature of the material's response — changes that require different constitutive frameworks, not just different numbers.

Temperature for polymers defines the boundary between two entirely different materials in terms of behaviour. Above the glass transition temperature, the chains are mobile — soft, creep-dominated, easily deformed. Below it, the chains are frozen — rigid, brittle at stress concentrations, sensitive to notches and rate. The same polycarbonate housing that is tough and impact-absorbing at 23 degrees Celsius can fracture in a brittle manner at minus 40 degrees — not because the material changed chemically, but because the temperature crossed the threshold where its molecular mobility changed. Moisture for hygroscopic polymers acts as a plasticizer — it lowers the effective glass transition temperature, increases chain mobility, increases ductility, and reduces both modulus and yield strength. In every case the question is the same: what is the character of this material at the conditions of the event you are simulating?

THE THREE QUESTIONS.

Before accepting any material card, ask three questions.

Where did this property come from? We are asking: does this number describe the character of the material in this part, or the character of a carefully prepared laboratory specimen under carefully controlled conditions? The property came from somewhere — a test, a datasheet, a literature value, an assumption. Each source has a scope and a set of conditions. The question is whether that scope matches the part.

What process created the part? We are asking: what has this material been through since it was in its standard datasheet condition? A die-casting process gave it porosity. An injection molding process gave it weld lines and fibre orientation and residual stress. A heat treatment changed its temper and its patience. A welding process locally annealed it. An ion exchange process pre-loaded its surface in compression. The manufacturing biography of the part is written into its material character — whether or not the FEA model acknowledges it.

What environment will it see? We are asking: at what rate will this material be loaded, at what temperature, in what moisture state, after how much service life? The material's character at the moment of the event is determined by all of these. A material model that uses the room-temperature, quasi-static, dry-as-molded datasheet value for a part that will be dropped at minus 40 degrees after two years of humidity exposure is not describing the same material. It is describing a different material that happens to have the same name.

The Holistic Analyst asks all three questions — not because a checklist demands it, but because the simulation is a representation of reality, and reality includes the full character of the material. Every property you enter is a claim about reality. The question is whether that claim is honest.

CLOSING. THE HONEST SIMULATION AND THE TESTING MINDSET.

A simulation with carefully considered, clearly documented material assumptions is honest engineering. It tells the reviewer: here is what I know, here is what I assumed, here is where I have uncertainty, and here is the basis for the confidence I am claiming. A simulation with default, unexamined material values is not honest engineering. It may produce numbers. It may even produce numbers that pass a safety factor. But those numbers are a claim about a reality that was never examined.

When you poke a system in simulation, the response you get back is determined by the character you gave it. If you gave it the wrong character — wrong density, wrong yield strength, wrong rate sensitivity, wrong surface condition, wrong manufacturing history — the response is not the response of your part. It is the response of a fictional part that happens to have the right shape.

Let me share two approaches I use in my own work, because they illustrate what the holistic analyst mindset looks like in practice.

The first is a homemade stress relaxation device. I bend a plastic bar and hold it at a fixed strain while recording the change in relaxation force over time. When I want to evaluate the effect of a chemical on the material under realistic service conditions, I apply the chemical to the outer face — which is in tension — so it gets absorbed exactly where the stress is highest. This picks up the effect of chemical exposure much faster than standard ASTM methods that test the material unstressed. I can control the strain level, which means I can calibrate viscoelastic Prony series parameters from the relaxation curves and use them in simulation. I have used this approach successfully to model stress relaxation in plastic triggers where long-term force retention was the critical design variable. The key insight is that you are not using generic datasheet viscoelastic properties. You are using properties measured on material in a condition that reflects what the part actually experiences.

The second is a homemade impact bending tester. I load a dogbone specimen on its edge — one-eighth inch thick face taking the impact — with a soft notch of one-eighth inch radius on the opposite face from the impact point. The impactor drops from a controlled height in free fall, so I can directly control the impact velocity. I measure the impact force over time. And I film the yield and fracture sequence with a high-speed camera at approximately 6400 frames per second. Then I run the same experiment in FEA. I simulate the impactor dropping from the same height, the same specimen geometry, the same notch, and I compare the simulated force-time history to the measured one — and the simulated deformation and fracture sequence to the high-speed camera footage. When those two match — the force history and the visible fracture timing — I have reverse-engineered the material's rate-dependent plastic behavior and damage properties from an experiment that directly replicates the loading mode of the real product. Not a generic uniaxial tension test. A bending impact test designed for the way the part actually breaks.

When you can match the force-time history and the visible fracture behavior simultaneously, you have moved well beyond datasheet modelling. You have demonstrated that the material card in your model describes a material that behaves the way your real material behaves, at the rates that matter, in the failure mode that matters. That is not a black box. That is an honest simulation.

Understanding material nature — deeply, not just numerically — is what makes this level of validation possible. Young's modulus is confidence. Poisson's ratio is empathy. Yield strength is patience. Density is presence. Behind every number on a material card is a physical reality — a character — that your model is trying to faithfully represent. The better you understand that character, the more faithful the representation. The more faithful the representation, the more honest the simulation. And the more honest the simulation, the more trustworthy the engineering decision it supports.

That is the Holistic Analyst. That is what combating engineering mind blindness actually means.

When you are ready for the technical guide — the keywords, the unit systems, the solver checks, the rate-dependent flags, the Abaqus INP Analyzer workflow — that is Part Two. It will mean something different now.