What a bow is made of: what the materials of limbs, riser, string and arrows change

When you choose a bow, or change a part of the one you have, the question of material comes up early. Wooden or carbon limbs, a lighter riser, a Dacron string or a modern fibre, aluminium or carbon shafts: behind each choice lies the same curiosity, whether the difference can be felt in the hand and whether it shows on the target.

The tests exist, but nearly all of them take place on a bench: a specimen bent until it breaks, a limb fitted with an accelerometer, an arrow set vibrating by an actuator, a bow in a rack shot with a mechanical release. They are direct measurements on equipment, and the registry counts them as evidence, but they almost always involve one piece or a handful, and in none of them did a recurve archer shoot with different materials and count the score. How these tests are done, and why so few pieces are ever tested, is explained on the page about material testing.

Shape and material

The oldest question is also the one with the clearest numbers. A mathematical model of the bow, checked against replicas, compared six types of bow, from the ancient straight-end bow to the modern recurve, at equal draw weight, draw length and limb mass, with arrows and strings of the same mass: all of them launch the arrow at almost the same speed, with coefficients between 1.92 and 2.09. The same paper gives the energy each material can store per kilogram, and here the gaps are wide: 2,500 newton metres per kilogram for sinew, 1,500 for horn, 1,100 for yew, 700 for maple and 4,300 for glass fibre. It follows that a bow made from the materials that store more reaches the same draw weight with shorter, lighter limbs, and according to the calculation that is where the advantage of composite bows lies, in the materials rather than the shape.

Scroll the figure→

The test on bows actually shot dates from 1988. Six bows, five of them replicas of historical originals, were shot several times by the same archer at a ballistics station, with two screens measuring the arrow's speed: a horn-and-sinew composite of 27 kilograms, or 60 lb, launched the same 50 gram field arrow as fast as a 36 kilogram yew longbow, 80 lb, and sent it more than 90 metres further. In the same series an Apache bow of wood backed with sinew, at 17.2 kilograms, launched a 28 gram arrow at 43 metres per second, while a Sioux bow of plain hickory, more powerful, launched a 30 gram arrow at about 30.

That is why the claim that a horn-and-sinew composite bow shoots better than a wooden bow of the same draw weight is reported. There is a single direct measurement, with one bow of each type, and the bows compared differ in several ways at once. In 2020 a comparison on a mechanical rig, with two bows of similar shape at 34 lb and the same draw length, found the elm bow up to 14% slower than a bow with a wooden core between two strips of glass fibre; that one, though, is a modern laminate rather than a historical composite, and the registry keeps it as context.

The wood of self bows

For a bow cut from a single piece of wood, the self bow, the choice of wood is nearly everything. Which woods work is a matter for the bowyers' literature: Osage orange and the two yews at the top, then maple, hickory, ash, oak and elm. A 2024 analysis set out to find what they have in common, pooling public data on 431 samples of 326 species. Density, bending strength and stiffness go together, whereas the ratio of bending strength to stiffness varies on its own: a high ratio means a wood that is strong in tension relative to its stiffness. The three woods considered the best in the world cluster where that ratio is highest, next to the secondary woods, while poplar and lime end up in the low-density, low-strength zone.

The claim that woods suited to self bows have a high ratio of bending strength to stiffness is reported, with a limit that lies upstream: which woods are suitable is decided by the bowyers' experience, not by a test, and no bow has been built and shot to check it. The species, moreover, tells you where a wood falls but not how good a single stave is, because two samples of the same species fall in the same zone, with correlations between 0.85 and 0.90, while within a species the properties vary by up to 30%.

The same study looks at Australian timbers, 127 samples of 106 species, and finds them generally dense and stiff but weak in tension relative to their stiffness, far from the woods known to be used, with few exceptions. The builders who write on an Australian forum describe the same difficulty: no native species tried has held up as an English longbow or a narrow flatbow, and many need a backing. The claim that Australian timbers are poorly suited to self bows is therefore reported, even though the two sources do not agree in the details: the species the forum rates as good fall almost entirely outside the zone of suitable woods, and the author leaves the explanation open.

Bamboo and woods in the Japanese bow

The kyudo bow is not made from a single piece: it has bamboo on the outer and inner faces and, between them, a core of bamboo and wood strips. A group working with a Kyoto workshop tested the pieces the master bowyer had already assigned to the two faces. In bending over a 400 millimetre span the inner bamboo came out stiffer, a difference that did not appear over a short 80 millimetre span, and the cross-section explains why: hard fibre took up 18% of the inner piece against 9% of the outer one, and the elastic modulus rose in proportion to the fibre.

The claim that in the Japanese bow the bamboo chosen for the inner face is stiffer than that for the outer is reported: one bowyer and one batch of bamboo, and the measurement does not say whether this is a rule of the craft or one master's choice. The master's words about the core woods were also set beside a bending test on specimens, and on two materials out of three they agree: hazenoki, which he calls brittle, breaks with minimal strain, while higo, which he describes as hard and brittle, behaves like maple, which he calls tough. The claim that what the bowyer says about the core materials only partly holds up to mechanical testing is also reported, with one curve per material and no stated repeats.

Carbon in the limbs

Modern limbs use maple and glass or carbon fibres embedded in synthetic resin, and the carbon comes in sheets that are already impregnated, the prepregs. A Korean group designing competition limbs took three unidirectional prepregs from three manufacturers, made with the same fibre and the same stated resin content, 38%, laid them up the same way and cured them in an autoclave under the same conditions. In tension the three were alike. In bending and compression they were not: one had more fibre, less measured resin and the best values, and under the electron microscope the machined surface of another showed numerous pinholes. Bending is precisely the load that matters in a limb, which is why the claim that carbon prepregs that match on paper but come from different suppliers do not give the same limb material is reported. It is a test on specimens, though, and nobody built a limb from the other two materials.

The same group also measured how much the limb vibrates after the shot, changing the damping layer laminated inside it. With 1.5 millimetres of glass fibre in place of bakelite, on a 68.42 lb carbon limb and four shots per condition, the vibration fell from 500 to 280 m/s² vertically, from 2,000 to 1,500 fore and aft and from 1,200 to 800 sideways. The claim that the material of the limb's inner layers changes how much the bow vibrates after the shot is reported, with two limits: the strongest vibration, fore and aft, falls by only a quarter, and the test concerns a single limb per condition, with no statistics, no archers and no measurement of the arrow. The same test, drawn in proportion, appears in the article on stabilisers.

The riser

A riser is often chosen by feel, heavy and steady or light. The only test of a lightened riser comes from a German group in 2004: a 25 inch riser designed with an evolutionary algorithm, and then built, weighed 871 grams against the 1,114 of the Hoyt Radian then used by the German national team, 22% less. In the structural analysis, at 200 newtons of draw weight, it deflected by at most 1.94 millimetres against 1.85, almost as little as the Radian; the margin existed because the stresses in the aluminium alloy stayed at 160 newtons per square millimetre, against a yield point of about 400.

The weight was measured on the actual piece, while the deflection comes from the computer model, and nobody loaded the riser that was built. That is why the claim that a competition riser can be made a fifth lighter with almost no loss of stiffness is transferred, meaning it rests on a calculation, and a second model from 2016 does not try to lighten anything. Of how that riser shoots we have only the opinion of three national team archers, two of them after nine arrows, and the authors themselves write that no study has ever shown that a stiffer riser makes for better shooting.

The string

For the string the mechanism has been known since 1937: to calculate the arrow's speed, about a third of the string's mass is added to the arrow's, so a lighter string leaves a larger share of the bow's energy to the arrow. Weight, however, is not the only thing that changes with the material. A 2007 chapter reports two series of laboratory tests. In the first, on a recurve in a rack, shot with a mechanical release and always with the same 18.1 gram arrow, four strings of different material and almost equal weight were tried; in the second, on a different recurve, two, four and then six strands were taken out of a sixteen-strand Fast Flight string.

With Dacron the arrow left 2.36 metres per second slower than with Fast Flight, and the author links this to Dacron's stretch: to lengthen by 1% one of its strands needs about 22 newtons, against about 70 for Kevlar. Taking strands out left the speed between 51.43 and 51.88 metres per second with no trend, because the lighter string is also less stiff and the two effects cancel out. On a shooting machine, in 2009, going from twenty strands to twelve sent the arrow off faster in every set-up tested, but by half a percentage point at most. In kyudo, finally, the synthetic strings passed about 69% of the energy to the arrow against 67.7% for hemp, less than the variation from shot to shot, and among the synthetics there was no gap.

On material, then, the two measurements, in the rack and in kyudo, point the same way, while on the number of strands they disagree, and the claim that the string's material and number of strands change the arrow's speed is contested. Nobody has measured it with archers shooting an Olympic recurve.

The same kyudo study, with ten university archers and fifty shots per string, also looked at vibration, with strain gauges near the grip. The string material did not change the frequency at which the bow vibrated, but how quickly the vibration died away: the logarithmic decrement in bending ran from 0.61 with Kevlar to 0.70 with hemp, and in torsion from 0.32 with Technora to 0.51 with hemp, so hemp settled the vibration sooner than the synthetics. The claim that string material changes the bow's vibration after release is reported: a single study, on a Japanese bow, and there are no measurements on a modern recurve or compound with stabilisers and dampers.

The shafts

For arrows the question of material is mostly a question of weight. A 2019 test bench, published as a preprint, sets the shaft vibrating with an actuator and derives its stiffness, damping and an equivalent mass. Taking the pairs with the closest spine, aluminium 2219 came to 22.7 grams against 13.5 for carbon 300, and aluminium 1916 to 14.8 against 9.8 for carbon 600: from one and a half to 1.7 times as much. The reason lies in the material properties used in a 2012 model, in which the carbon and resin layer is about three times as stiff as aluminium, 222 gigapascals against 72, at a little over half the density.

Scroll the figure→

The claim that at a similar spine, an aluminium shaft weighs more than a carbon one is reported. The limit is that the bench mass is the one estimated by the dynamic model, not the grain weight of a finished shaft, and that the shafts are few and were tested on a single bench; the grain weight of a shaft is on the manufacturer's chart, and the article on choosing competition arrows explains what changes on the field.

The same bench also estimates damping: carbon has the lowest values, from 0.32 to 0.51 depending on spine, aluminium the highest, from 0.53 to 0.67, and cedar sits at around 0.5. The claim that carbon arrows damp vibration less than aluminium and wooden ones is reported, with blurred edges: the most flexible carbon arrow overlaps with cedar, and the rig's resonances, between 31 and 41 Hz, are far from the 85 of an arrow's free bending, so the values probably describe the test more than the arrow in flight. Another laboratory, driving a carbon arrow at resonance, derived a damping time of about a second and a half, the time it takes the shaft, vibrating freely, to lose almost two thirds of its oscillation: longer than the flight time at competition distances. This is one of the bases of the claim, also reported, that carbon arrows damp their own vibration little and reach the target still oscillating; the arrow reaching the target still vibrating remains a deduction, though, because damping by the air has not been measured.

What reaches the hand

That leaves the question I started from. The measurements show that something nearly always changes, from the vibration of the limb to the mass of a shaft at a similar spine. Whether the archer notices has been asked only twice. In kyudo the free comments gathered after each shot departed from what was expected just past the statistical threshold, with a chi-square of 12.69 against 12.59, driven mainly by remarks about vibration and impact with hemp; for the lightened riser there are the impressions of three archers. Nobody has had a recurve archer shoot with limbs, risers or strings of different materials while measuring the score or the group, so between the material and the points there is always a step that no test has covered.

On the field I take three things from this, and none of them is a promise of points. The first concerns traditional bows, which are where material weighs most, because it decides how much draw weight you get from a limb of a given length and mass. Anyone building a self bow does well to start from the woods the bowyers' literature recommends, knowing that within one species one stave can be very different from another, and that a dense, stiff wood is not necessarily a good bow wood.

The second concerns the string. Between Dacron and the modern fibres the difference in speed measured on the bench is there, a little over two metres per second, while among the modern fibres it is small or absent, and taking strands out changes the speed by half a percentage point at most. If you change the string material, or the number of strands, it makes more sense to treat it as a change of set-up and check nocking point, brace height and sight marks again than to expect a faster arrow that shows on the target.

The third concerns vibration, which is what the hand feels most. The tests show that the material of the limb and, in the Japanese bow, that of the string change how much and for how long the bow vibrates after the shot, when the arrow has already gone, but none of them links that vibration to the point of impact. Choosing limbs or a string that make the bow feel quieter in the hand therefore remains a matter of feel, a reasonable one when it helps you shoot a long session calmly, and no test has so far shown that a shorter vibration tightens the group.

@misc{campagna2026_materialidellarcolegnocarboniobambu,
  author = {Campagna, Matteo},
  title = {What a bow is made of: what the materials of limbs, riser, string and arrows change},
  year = {2026},
  howpublished = {matteocampagna.com},
  url = {https://matteocampagna.com/en/articoli/what-a-bow-is-made-of-wood-carbon-bamboo},
  note = {accessed 2026-10-16}
}