Epoxy resins are one of those materials nobody talks about and everybody depends on. They cure when you want them to, they stick to almost anything, and they keep their shape under load. The catch is what happens when things get hot. A normal epoxy will happily do its job in a phone, but ask it to sit next to a silicon carbide power chip running above 250°C for thousands of hours and it falls apart. What follows is how chemists have been rebuilding epoxy molecule by molecule so it can actually survive that, and why the fix turned out to be a lot less obvious than just “add more heat-resistant stuff.”
Why heat became the wall nobody could get around
In a laptop, the resin in a package mostly has to survive one solder reflow and not deform. Power devices in cars are a different sport entirely. A SiC switch built for an EV is supposed to dissipate real power past 250°C and stay chemically intact through endless thermal cycling. That is far beyond what the usual encapsulation materials can take.
Silicone has been the safe pick for this job, and not without reason. It seals out moisture and relaxes stress so well that the package rarely delaminates, and the Si-O bond is genuinely tough, around 100 kcal/mol against the roughly 80 you get from the C-C bonds that make up most of an epoxy. But silicone has a dirty secret. The stray alkoxy groups left on chain ends can peel off under long exposure to high heat, and once you’re above 150°C the material can crack internally or let the interface peel, which is the exact insulation failure you cannot forgive in a traction inverter. That is the opening epoxy has been waiting for, because a stiffer epoxy handles the stress from mismatched expansion (silicon die, aluminum wire, copper sink) far better, and people have data suggesting it holds up longer.
Two meanings of “heat resistant” that everyone conflates
Before getting into design, it is worth splitting two properties that share a name but not a mechanism.
Physical heat resistance is the glass transition temperature, Tg. Below it the cured resin is a rigid glass. Above it the thing turns rubbery and loses stiffness, dimensional stability, and grip. Tg is your operating ceiling.
Chemical heat resistance is about decomposition. The resin might never soften and still have its bonds quietly break down over a thousand hours, and that is what kills the device in the field. Decomposition temperatures sit much higher than Tg, but they answer to a different set of rules. This is where epoxy design gets mean: the levers that lift one tend to drag the other down.
What actually moves the glass transition temperature
The structure-property work on epoxy has settled on a short list of features that push Tg up. More epoxy groups per molecule give you a denser network after cure, and denser networks resist the tiny Brownian jiggling that softens a polymer. Rigid aromatic cores hold shape better than floppy aliphatic chains. Symmetry quiets the local motion that lets chains slide. Bulky groups physically block movement, and polar groups hydrogen-bond to the hydroxyls formed during curing, pinning the whole network.
Phenolic novolac epoxy is the textbook case: high functionality, lots of groups, rigid backbone. Pushing Tg further means adding repeat units during the phenol-formaldehyde prepolymer step, which raises crosslink density. In practice that stalls near 220°C for a cresol-novolac cure, and you pay for it, which is the next problem.
The trap nobody warned you about
Here is the part that bites. The same moves that raise Tg also multiply the weakest links in the molecule. Glycidyl ether groups and the methylene bridges in a novolac are the first things to decompose. So when you pile on functional groups or repeat units to lift Tg, you are also piling on the very bonds that fail under heat. The two kinds of heat resistance pull against each other.
The smarter framing is to dilute the weak linkages instead of just adding more of them. Swap a methylene bridge for a dicyclopentadiene unit and the 5% weight-loss temperature climbs. Stick a methyl onto the phenol ring and it does the opposite, which is a genuine surprise and tells you the methyl itself is an easy break point. Counts matter less than where you put the atoms.
Why maxing out Tg wrecks everything else
Chase Tg the old way and the side effects stack up fast. More functional groups means higher molecular weight and melt viscosity, and past a point the resin will not dissolve at the 150-200°C cure temperature, so the cure goes bad. Heavy steric hindrance drops conversion, and uncured epoxy groups re-crosslink during a 250°C reflow, swelling the part and spiking residual stress. More crosslinking makes more hydroxyls, and hydroxyls drink water, which hurts electricals and dimensional stability. Higher Tg means higher internal stress, which fights the adhesion you were relying on. And more flammable epoxy groups mean weaker flame performance. Every one of those is a property the device also needs, so the old playbook just traded them off. The good work now is finding structures that let you keep all of them.
Thermal conductivity is its own headache
Getting heat out of a power device matters as much as surviving it. A thermal interface material sits between the hot die and the copper sink to close the air gaps from surface roughness. It has to conduct heat and flexibly track the warping surfaces, which is two contradictory jobs. Hard ceramic fillers like aluminum nitride conduct but will not deform. Soft polymers deform but scatter phonons and barely conduct. One material that does both has been nearly impossible, which is why silicone grease ran the show despite bleeding volatiles, pumping out, and going stiff over time.
For epoxy to take that role it needs flexibility and enough conductivity that a little filler does the work. The lever is different from the Tg story. Heat moves through an electrical insulator by phonons, not electrons, so the enemy is phonon scattering. Amorphous cured epoxy scatters phonons badly because of its disorder. The fix is to build in structures that self-order at the nanoscale, like nematic skeletons such as biphenyl, so phonons actually get through.
Five designs that hold up in the real world
Naphthalene bifunctional epoxy. Made from 1,6-dihydroxynaphthalene, it uses a rigid naphthalene core instead of bisphenol. Even at low crosslink density it gives a high Tg, and its flat, aligned structure keeps thermal expansion low and toughness high. It is one of the few epoxies that is both heat resistant and genuinely adhesive and tough.
Naphthalene tetrafunctional epoxy. Same 2,7-dihydroxynaphthalene starting point with formaldehyde, but the trick was picking that isomer and a tight reaction window to get a dimer that is rigid, symmetric, and low molecular weight at once. You end up with a rare epoxy whose Tg clears 300°C, which puts it right where SiC devices want to be.
Dicyclopentadiene-structured epoxy. The bulky dicyclopentadiene residue between phenol cores keeps functional-group concentration low, so hydroxyl content and water uptake stay strikingly small. Low crosslink density notwithstanding, the rigid backbone lifts Tg well past what rubber-elasticity theory predicts. No methylene bridges and a diluted aryl-ether concentration also push the decomposition temperature up. It carries physical heat resistance, chemical heat resistance, and low moisture absorption together.
Naphthalene-ether oligomer epoxy, or E-NEO. Selective self-condensation of 2,7-dihydroxynaphthalene gives a trimer used as a 2-3 functional epoxy. Its claim to fame is pairing a high Tg with real flame retardancy, the thing ordinary epoxies cannot do at the same time. The idea is that naphthalene-ether bonds break and recombine during burning to build a graphitic charred foam that puts the fire out. It keeps the low thermal expansion from the flat naphthalene units and still dissolves in solvent, which matters when you actually have to process it.
Mesogenic-structure epoxy. This one blends a liquid-crystal skeleton with a long chain. It shows a layered phase between about 60 and 90°C, is crystalline and easy to handle at room temperature, and has a very low melt viscosity near 150°C, so it takes high filler loading. By skipping the vulnerable ester, imine, and alkyl-ether bonds it reaches decomposition above 300°C while giving roughly 1.5 times the thermal conductivity of a standard epoxy and staying flexible enough to follow surface warpage.
Where this goes next
All five examples share one idea: stop treating Tg, chemical stability, moisture resistance, flame retardancy, and thermal conductivity as a zero-sum game. Pick skeletons that order themselves, resist a specific bond scission, or supply their own rigidity, and you hit numbers novolac chemistry could not touch. Epoxy already beats the higher-temperature contenders on processability and adhesion, so closing the heat gap is what lets it into SiC power modules, EV inverters, and energy hardware that has to run hot for a decade. The molecular toolbox is not close to empty, and I would bet the 300°C barrier goes from a curiosity to a baseline sooner than people expect.

