A flexure bearing is a mechanical pivot that allows motion through the elastic deflection of a slender flexible element rather than through sliding or rolling contact. The defining component is the flexure blade — a thin elastic strip that bends to permit one direction of motion while resisting all others. It exists to eliminate friction, backlash, stiction, and wear in precision applications. The result is sub-nanometre repeatability, infinite life within fatigue limits, and motion you can trust in vacuum or cryogenic environments where lubricated bearings fail.
Flexure Bearing Interactive Calculator
Vary flexure blade thickness, length ratio, travel angle, and allowable stress to see blade length, peak bending stress, and fatigue margin.
Equation Used
The calculator estimates maximum bending stress in a thin flexure blade using sigma_max = E*t*theta/(2*L). Because L = t*(L/t), the stress is controlled mainly by rotation angle, material modulus, and the blade length-to-thickness ratio.
- AISI 301 stainless modulus is fixed at E = 193000 MPa.
- Blade bends with approximately constant curvature.
- Travel angle is the one-sided peak rotation from neutral.
- Stress concentration at the root fillet is not included.
How the Flexure Bearing Actually Works
A flexure bearing works by storing strain energy in a thin elastic element — usually a blade, notch, or leaf — and releasing that energy as the part returns to its rest position. There are no sliding surfaces, no rolling elements, no lubricant. You bend the flexure and it bends back. That single fact eliminates every friction-related problem you would normally fight in a precision pivot: no stiction breakaway force, no Coulomb friction step at zero crossing, no wear particles, no backlash. A monolithic flexure cut from a single block of titanium or beryllium-copper has zero assembly tolerance because there is no assembly. The pivot axis is defined by the blade geometry to within a few microns.
The geometry has to be right or the mechanism fails in two ways. If the blade is too thick, the bending stiffness goes up with the cube of thickness and the flexure resists motion you wanted. If too thin, axial and lateral stiffness collapses and the flexure buckles under load. For a typical cross-spring pivot using 0.25 mm spring-steel blades, you want length-to-thickness ratios of 30:1 to 80:1 and you size the blade so the maximum bending stress at full deflection stays below 40% of the material's fatigue limit — for AISI 301 stainless that means peak stress under roughly 280 MPa for indefinite life. Push past that and you get fatigue cracks at the blade root, usually visible as a hairline within a few hundred thousand cycles.
Parasitic motion is the other thing that bites you. A simple cantilever blade rotates about a centre that drifts as it deflects — the centre of rotation moves along an arc, not a fixed point. Cross-spring and double-parallelogram designs cancel this drift to first order. If you measure unexpected lateral translation at the tool point, your flexure geometry is single-stage and you need a compound or symmetric layout to get a true fixed pivot.
Key Components
- Flexure Blade: The thin elastic element that bends to allow motion. Typical thickness 0.1 to 0.5 mm in spring steel or beryllium-copper, with length-to-thickness ratio between 30:1 and 80:1. Blade root fillets must be smooth — any tool mark above Ra 0.4 µm becomes a fatigue crack initiation site.
- Mounting Flanges: The rigid sections at each end of the blade that bolt or weld to the moving and fixed bodies. These must be at least 5 times stiffer than the blade itself, otherwise compliance migrates into the flange and your pivot axis drifts under load.
- Cross-Spring Geometry: Two blades intersecting at right angles defines the rotation axis at their crossing point. The crossing is virtual — the blades typically don't touch — and the geometry cancels first-order centre drift, holding the pivot axis to within ±10 µm over ±5° of rotation.
- Stress-Relief Fillets: Radii at the blade-to-flange transition. Minimum recommended radius is 3× blade thickness. Sharp internal corners concentrate stress by factors of 2-3 and slash fatigue life by an order of magnitude.
- Travel Stops: Hard mechanical limits that prevent over-deflection. Critical because a flexure has no inherent stop — you can yield or snap the blade in a single overload event. Stops are sized to engage at 80% of the calculated yield deflection.
Where the Flexure Bearing Is Used
Flexure bearings show up wherever friction, wear, lubricant, or backlash would ruin the job. That means precision optics, scanning probe microscopy, vacuum and cryogenic systems, MEMS, and any pivot that has to survive a billion cycles without service. They are not used where large continuous rotations are needed — flexures only work over limited angular ranges, typically ±5° to ±15°. When the application fits, nothing else comes close on repeatability or stiction-free behaviour.
- Space optics: JWST primary mirror segment actuators use blade flexures to position each hexagonal segment in 6 degrees of freedom with nanometre-class resolution at cryogenic temperatures.
- Atomic force microscopy: Bruker Dimension Icon AFM stages use double-parallelogram flexures for X-Y scanning over ±90 µm with sub-Ångström noise floor.
- Semiconductor lithography: ASML wafer-stage fine-positioning loops use flexure-guided actuators to hold position to ±0.5 nm during exposure.
- MEMS sensors: Silicon notch flexures in Analog Devices ADXL accelerometers form the proof-mass suspension — etched in a single lithography step with no moving contact.
- Watchmaking: The Zenith Defy Lab oscillator replaces the traditional sprung balance with a monolithic silicon flexure running at 15 Hz instead of 4 Hz, eliminating balance-staff pivots entirely.
- Cryogenic coolers: Stirling-cycle pulse-tube coolers on satellite IR sensors use spiral flexure springs to suspend the piston, giving 10+ year contactless service in vacuum.
The Formula Behind the Flexure Bearing
The single most useful number for a flexure designer is the angular stiffness of a cantilever blade — how much torque it takes to deflect by one radian. At the low end of the typical operating range, very thin blades give you compliant pivots that need almost no actuator force but won't survive any side load. At the high end, thick blades support real payloads but eat your actuator's torque budget. The sweet spot is whatever stiffness keeps your peak bending stress under the material fatigue limit while still letting your actuator drive the thing. This formula tells you which side of that line you sit on.
Variables
| Symbol | Meaning | Unit (SI) | Unit (Imperial) |
|---|---|---|---|
| Kθ | Angular stiffness of the flexure blade about its bending axis | N·m/rad | lbf·in/rad |
| E | Young's modulus of the blade material | Pa | psi |
| b | Blade width (perpendicular to bending direction) | m | in |
| t | Blade thickness (in bending direction) | m | in |
| L | Active blade length between rigid flanges | m | in |
Worked Example: Flexure Bearing in a Swiss precision instrument lab
A precision instrument lab in Neuchâtel is designing a cross-spring flexure pivot for a tilting mirror mount in a femtosecond laser delivery head. The pivot has to allow ±3° of tilt with sub-microradian repeatability and survive 10 million actuation cycles. The team picks AISI 301 full-hard stainless blades, 20 mm wide, with an active length of 25 mm. They need to evaluate angular stiffness and peak stress across the realistic blade-thickness range of 0.15 mm to 0.35 mm to find the design sweet spot.
Given
- E = 193 GPa
- b = 20 mm
- L = 25 mm
- tnom = 0.25 mm
- θmax = 3 °
- σfatigue = 280 MPa
Solution
Step 1 — at the nominal blade thickness of 0.25 mm, compute the angular stiffness of one blade. Convert to SI: E = 193×109 Pa, b = 0.020 m, t = 0.00025 m, L = 0.025 m.
For a cross-spring pivot using two blades, total stiffness doubles to ≈ 0.40 N·m/rad. To drive ±3° (±0.0524 rad), the actuator needs about 0.021 N·m of torque — well within the range of a small voice coil.
Step 2 — check peak bending stress at full deflection. For a cantilever blade of length L deflected through angle θ, peak surface stress is σ = 3 × E × t × θ / (2 × L).
That's 54% of the 280 MPa fatigue limit for AISI 301 — comfortable margin for 10 million cycles.
Step 3 — evaluate the low end of the design range, t = 0.15 mm:
At 0.15 mm the blade is ultra-compliant and stress is only 33% of fatigue limit — sounds great, but lateral and axial stiffness drop with the same cube law, so the mirror sags visibly under its own weight and any vibration above 30 Hz couples straight into the pointing loop.
Step 4 — high end, t = 0.35 mm:
Stress is now 76% of the fatigue limit — borderline for 10 million cycles, and any stress concentration at the blade root will push it over. Stiffness has nearly tripled, eating most of the actuator torque budget.
Result
The nominal 0. 25 mm blade gives an angular stiffness of 0.20 N·m/rad per blade and a peak stress of 152 MPa at ±3° — the design sweet spot. In practice this feels like a pivot that returns to centre with no perceptible hysteresis and lets the voice coil drive crisp transitions without saturating. The 0.15 mm version is three times more compliant but loses lateral stiffness so badly that ambient vibration shows up as pointing jitter; the 0.35 mm version doubles your actuator power demand and parks you uncomfortably close to the fatigue cliff. If your built pivot shows higher stiffness than predicted, check three things: (1) blade root fillet radius — a sharp corner stiffens the effective root and shifts L, (2) flange clamp screws over-torqued so the blade is being pinched outside the intended L, or (3) blade thickness measured with a micrometer rather than nominal — sheet stock typically runs +0.02 mm and stiffness scales with t3.
Choosing the Flexure Bearing: Pros and Cons
Flexure bearings beat every other pivot on repeatability and stiction, but they lose hard on travel range and load capacity. The decision usually comes down to whether your application needs nanometre repeatability over a small angle, or many degrees of motion at coarser precision.
| Property | Flexure Bearing | Ball Bearing Pivot | Bushing Pivot |
|---|---|---|---|
| Angular travel range | ±5° to ±15° typical | Unlimited (continuous rotation) | Unlimited |
| Repeatability | Sub-microradian, sub-nanometre at tool point | 1-10 µrad limited by ball roundness and preload | 50-500 µrad limited by clearance |
| Friction / stiction | Zero — purely elastic | Low Coulomb friction, measurable stiction breakaway | Higher Coulomb friction, large stiction step |
| Load capacity (lateral) | Low — limited by buckling, typically <10× pivot weight | High — radial loads to thousands of N | High — well-distributed bearing surface |
| Lifespan | Effectively infinite below fatigue limit, finite above | 10⁶ to 10⁹ cycles depending on load | Wear-limited, lubricant-dependent |
| Vacuum / cryo compatibility | Excellent — no lubricant needed | Poor without specialised dry lubricant | Poor — lubricant outgasses or freezes |
| Cost (precision-grade pivot) | Moderate to high — wire EDM or precision grinding | Low to moderate — commodity part | Low |
| Backlash | Zero | Preload-dependent, typically 1-5 µm | Clearance-dependent, 10-100 µm |
Frequently Asked Questions About Flexure Bearing
First-order cancellation in a cross-spring pivot assumes the two blades meet at exactly 90° and have identical stiffness. In real builds, drift comes from blade-thickness mismatch — even 5% difference in t between the two blades unbalances the geometry and shifts the rotation centre by tens of microns over ±5°.
Measure both blades with a micrometer. If they differ by more than 0.01 mm, swap one. Also check the crossing angle with a calibrated square — wire EDM fixturing slop commonly leaves you at 89° or 91° instead of 90°, and the centre then walks linearly with deflection.
Notch flexures concentrate bending into a short, thin neck and give you a very well-defined pivot axis with low parasitic motion — but stress concentration at the neck limits angular range to roughly ±2° before fatigue becomes a problem. Leaf flexures spread bending over a long blade, tolerate ±10° or more, but the rotation centre drifts unless you use a compound geometry.
Rule of thumb: if your travel is under 2° and you need the tightest possible axis definition, use a notch. If travel exceeds 3° or you need symmetric loading, use leaf blades in a cross-spring or double-parallelogram layout.
You're seeing the natural frequency of the moving mass on its flexure suspension. fn = (1/2π) × √(K/m), and most flexure stages land somewhere between 30 and 200 Hz depending on payload. If 80 Hz is unacceptable, you have three levers: increase blade thickness (stiffness scales with t3), shorten blade length, or reduce moving mass.
Doubling t lifts the resonance by a factor of 2.83. Halving the moving mass lifts it by √2. Adding active damping with a voice coil and velocity feedback is often easier than redesigning the flexure if you only need to kill the Q peak.
No. Each cycle still accumulates fatigue damage, and a flexure rated for 10⁷ cycles at ±3° will fail at the same cycle count regardless of how slowly you traverse. Continuous unidirectional rotation is fundamentally outside the flexure design envelope — there is no way to unwind the strain.
If you need slow continuous motion plus high precision, the standard answer is a coarse-fine architecture: a bearing-supported coarse stage carries a flexure-guided fine stage that handles the last micron of motion within its angular limit.
The formula Kθ = E·b·t3/(12·L) assumes the rigid flanges are infinitely stiff and the blade transitions abruptly from flange to bending region. In a real part with fillets and bolted flanges, the effective bending length L is shorter than the nominal — typically by 1.5 to 2 times the fillet radius on each end.
If your measured stiffness is 20-30% high, recompute L using Leffective = Lnominal − 1.7·r at each end. If it's 50%+ high, your flange clamping is biting into the blade and the effective L is much shorter than you think — back the clamp screws off and re-measure.
Beryllium-copper C17200 in the precipitation-hardened temper is the standard choice for cryogenic flexures. It keeps a high yield strength (≈1100 MPa) and decent fatigue limit at 4 K, while remaining non-magnetic — critical near SQUID sensors. Titanium 6Al-4V also works and is cheaper, but its fatigue limit drops more steeply with cycle count at cryo.
Avoid martensitic stainless steels — they go brittle below about 80 K and crack at the blade root without warning. Phosphor bronze is fine for low-stress applications but its yield strength is too low for tight geometries.
References & Further Reading
- Wikipedia contributors. Flexure bearing. Wikipedia
Building or designing a mechanism like this?
Explore the precision-engineered motion control hardware used by mechanical engineers, makers, and product designers.
