How do self-healing polymers work, and what can they fix?

Why it matters

A self-healing polymer is built so that damage can start a repair, either with no outside help or with a small trigger such as heat. An early example came in 2001, when a University of Illinois team described an epoxy composite that heals cracks autonomically, without manual intervention. If this works in real parts, a component could seal small cracks before they grow.

How they work

There are three broad routes.

Capsules. The material contains tiny capsules of liquid healing agent, along with a catalyst spread through the polymer. When a crack passes through, it breaks the capsules. The released liquid meets the catalyst and sets, bonding the crack faces. The 2001 epoxy recovered up to 75% of its fracture toughness in laboratory tests (White et al., 2001). The weakness is that a capsule can only be used once, so a second crack in the same place has nothing left to release (Toohey et al., 2007).

Vascular networks. Here, healing agent flows through a network of channels, rather than sitting in fixed capsules. Toohey and colleagues built a three-dimensional microvascular network under an epoxy coating, and the coating healed repeatedly.

Intrinsic chemistry. In this route the polymer's own bonds do the work. Some bonds can break and re-form. In a thermally re-mendable cross-linked polymer, heating above 120°C makes about 30% of the linkages between monomers disconnect, and they reconnect on cooling, so a fractured part can be re-mended several times. A hydrogen-bonded rubber can be cut, then healed at room temperature by pressing the faces together, and the cycle can be repeated. Vitrimers are a related idea: their networks rearrange through exchange reactions, so they stay insoluble but can be reshaped and welded by local heating.

What has been demonstrated

These results come from laboratory experiments, or from full-scale demonstrators, reported in peer-reviewed papers. Each holds only for the conditions tested.

• Epoxy composites. The capsule system recovered up to 75% of its fracture toughness in laboratory fracture tests (White et al., 2001).
• Repeated healing. The microvascular epoxy coating repeatedly healed crack damage (Toohey et al., 2007). The hydrogen-bonded rubber showed repeated break-and-heal cycles at room temperature (Cordier et al., 2008).
• Conductive skin-like composite. A supramolecular polymer filled with nickel microparticles healed electrically and mechanically at ambient conditions. The authors propose it for electronic skin (Tee et al., 2012).
• Fatigue in a hydrogel. A hydrogel sample without a pre-cut crack recovered its stress-stretch curve after thousands of loading cycles. A sample with a pre-cut crack still extended with each cycle. Above the fatigue threshold, noncovalent interactions slowed crack growth under cyclic loads (Bai et al., 2018).
• Ceramics. A small amount of MnO activator in a three-dimensional network within a structural ceramic sped up self-crack healing by more than 6,000 times in laboratory specimens, and lowered the required temperature (Osada et al., 2017). A Ti2AlC ceramic went through repeated cracking and repair at 1,500 K. The healed regions were strong enough that new cracks formed elsewhere on reloading (Sloof et al., 2016).
• Metals. Researchers directly observed nanoscale fatigue cracks in pure metals, and some cracks healed through cold welding of the crack flanks (Barr et al., 2023). This is a result in metals in controlled experiments, not in polymers.
• Concrete. A paper evaluating full-scale demonstrators of bacteria-based concrete from Delft University of Technology reports that, in earlier laboratory work, a healing agent of bacterial spores and lactate healed cracks up to 0.8 mm wide, depending on dosage. In two repair mortar and two concrete demonstrator projects, the authors observed no negative side effects on construction performance. They found it difficult to show better crack healing, because the demonstrator structures hardly cracked (Mors and Jonkers, 2019).

What is only reported

This section covers a claim made in an operator's announcement and a result that so far exists only in computer simulation. Neither is confirmed by physical measurement in the sources checked.

• A 2024 tunnel pilot. In a news item dated 16 July 2024, ProRail reported a 25-metre bacteria-based concrete wall in the Vijf Eikentunnel at Rijen, in the Netherlands, as a pilot to test water-tightness in practice (ProRail, 2024). According to ProRail, moisture and oxygen entering a crack activate the bacteria, which fill it with limestone. The announcement reports no results, so it does not show that the repair works or lasts. We have not checked for later updates, so the pilot's current status is unknown.
• A simulated polymer crack. A 2025 peer-reviewed Nature Communications paper reports that, in numerical simulations of a polymer glass, controlled oscillatory deformation restored the mechanical properties around a cylindrical crack. This is a computer model. It does not show repair of a manufactured polymer part.

What is speculative

The following uses have not been shown to work in practice in any of the sources checked. Except where a source is given, they are our own examples of where the idea might apply:

• composite parts in aircraft or wind turbines that seal small internal cracks between inspections. The ceramics authors have suggested a similar approach for lightweight ceramic turbine blades in aircraft engines, but that is a suggestion, not a test (Osada et al., 2017);
• phone or wearable casings that close minor scratches at room temperature;
• protective coatings for pipelines or bridges that reduce corrosion after minor damage;
• vitrimer composites that could be reshaped or rewelded at the end of their life. The welding property is shown in the vitrimer work (Montarnal et al., 2011), but none of the sources checked tests end-of-life recycling.

Limits and open problems

Healing is often partial. The first capsule system recovered up to 75% of its toughness, not all of it.

Repeat healing is not guaranteed. Capsules work once per location. Vascular and intrinsic systems can repeat.

Many systems need a trigger. The re-mendable polymer in the intrinsic section needs heating above 120°C. The ceramic healing needed 1,500 K. Room-temperature systems, such as the hydrogen-bonded rubber, healed by pressing pieces together.

Repeated loading is a harder test. The hydrogel result shows that a sample can recover its curve under uncut cyclic loading, yet a pre-cut crack still grows. Healing a surface is not the same as stopping a crack under fatigue.

Sealing is not the same as restoring strength. The concrete paper reports that cracks can be sealed in the laboratory, but the full-scale demonstrators could not show better healing in use. The ProRail pilot, as announced in July 2024, was a water-tightness test with no published results.

Speed and cost are not well measured. The sources checked give few speed figures and no cost data, so this article does not estimate either.

What to watch next

• Field data from pilots. Measured crack closure, leak rates and maintenance records from the Rijen wall would show whether the bacterial concrete approach works beyond the lab.
• Fatigue results in polymers. Tests that track a pre-cut crack under repeated loading, rather than only uncut samples, will show whether healing slows real damage growth.
• Physical tests of simulated healing. The 2025 simulation needs a laboratory test on a real polymer before its result can be treated as a demonstration.
• Replication in structural metals. The 2023 cold-welding observations in pure metals would need to be reproduced under service conditions before design rules change.