Stiction is the static friction force that resists the start of motion between two surfaces in contact, and it's larger than the kinetic friction force that resists motion once sliding begins. The gap between the two creates breakaway force — the input must overcome stiction before the surfaces slip. That step change is what causes stick-slip oscillation, valve hunting, and servo dead-band. In a typical control valve, stiction of 5-10% of valve span produces visible limit cycles in the loop output that no amount of PID tuning will fix.
Stiction Interactive Calculator
Vary normal load, friction coefficients, and applied force to see breakaway force, sliding force, and the stiction force drop.
Equation Used
The calculator uses the article's stiction relationship: static friction sets the breakaway threshold, Fs = mu_s*N, and kinetic friction sets the lower sliding force, Fk = mu_k*N. The difference is the force drop that creates stick-slip behavior.
- Coulomb dry friction model.
- Normal load is constant during breakaway.
- Default coefficients are dry steel on steel from the article.
- Speed, lubrication regime, temperature, and wear effects are not modeled.
The Stiction in Action
Push a stationary object and nothing happens until your applied force crosses a threshold — then it suddenly jumps into motion and the force needed to keep it moving drops. That threshold is stiction, and the drop is the difference between the coefficient of static friction (μs) and the coefficient of kinetic friction (μk). For a steel-on-steel dry contact, μs sits around 0.74 and μk around 0.57. The asperities — microscopic peaks on each surface — interlock and form cold-welded micro-junctions while at rest. Breaking those junctions takes more force than shearing the boundary lubricant film once sliding starts.
The Stribeck curve explains why stiction matters most at low speed. At zero velocity you sit in the boundary lubrication regime where metal contacts metal through a thin oil film. As speed climbs, the lubricant gets entrained between the surfaces and friction drops to a minimum in the mixed regime, then climbs again in the hydrodynamic regime as viscous drag takes over. The stiction-to-kinetic transition is the cliff at the left edge of that curve. If your control loop tries to hold a position near zero velocity, you sit on the cliff edge — any disturbance pushes you over it, the actuator slips, overshoots, sticks again, and you get a limit cycle.
Tolerances and surface condition drive the magnitude. A control-valve stem with 5 µm of corrosion product on the packing contact will show 3-4× the stiction of a freshly polished stem. Boundary lubrication failure — additive depletion in a gear oil, water washout in a hydraulic ram — converts a clean Stribeck curve into a stiction-dominated mess. The classic failure mode is a control loop that tunes fine on the bench then hunts at ±2% in the field six months later because the packing has dried out.
Key Components
- Contacting Surfaces: Two solid surfaces with asperities of 0.1-10 µm height, depending on Ra finish. Real contact area is a tiny fraction of apparent area — typically 0.01-1% — concentrated at the asperity tips. Surface finish below Ra 0.4 µm reduces stiction by roughly half compared to Ra 1.6 µm because fewer asperities interlock at rest.
- Normal Load: The clamping force pressing the surfaces together. Stiction force scales linearly with normal load through Fs = μs × N. Doubling the bolt preload on a clamped joint doubles the breakaway force needed to make it slip — which is why preload sets static friction in friction-grip bolted connections.
- Boundary Lubricant Film: A 1-100 nm layer of oil, grease, or anti-friction additive that separates the asperities at zero velocity. Properly formulated EP (extreme pressure) additives like ZDDP shear before the metal does, dropping μs from 0.7 dry to 0.10-0.15 lubricated. Additive depletion or water contamination kills this film and stiction climbs back toward the dry value.
- Dwell Time: Time the surfaces sit stationary under load. Stiction grows logarithmically with dwell — a joint sitting overnight breaks free at 1.2-1.5× the value measured after a 1-second pause. This is why valves that cycle continuously stay smooth and valves that sit for weeks slam open when finally commanded.
- Tangential Compliance: Elastic deformation of the surfaces and structure before slip. Under a sub-breakaway tangential force the contact deflects by 1-50 µm without sliding — this is presliding displacement. Servo control loops see this as a soft spring, and the LuGre friction model exists specifically to capture this regime for simulation.
Where the Stiction Is Used
Stiction shows up wherever a controlled element has to start, stop, or reverse direction at low speed. The places it causes the worst problems are process control valves, precision linear stages, and any servo loop tracking near-zero velocity. The mechanism that fails is always the same — the control system commands a small move, the actuator can't break free, integral wind-up builds, the actuator finally slips and overshoots, and the cycle repeats. Naming it correctly is the first step to fixing it.
- Process Control: Globe valves on Emerson Fisher control valve trim where graphite packing stiction of 3-8% of valve travel causes loop oscillation on flow and pressure controllers in refinery service.
- Precision Motion: Aerotech ABL air-bearing stages spec stiction below 0.1 N specifically because semiconductor lithography stages need sub-nanometre positioning that any breakaway force would destroy.
- Hydraulics: Parker D1FP proportional valves use dither — a 200-300 Hz small-amplitude signal — superimposed on the command to keep the spool in continuous micro-motion and avoid stiction at the null position.
- Robotics: Harmonic Drive strain-wave gearboxes in collaborative robots like the Universal Robots UR5 show stiction at the wave generator that limits force-control bandwidth to roughly 30 Hz before stick-slip dominates.
- Automotive: Disc brake calipers on a Brembo M4 caliper use rolled-back piston seals that retract the pad 0.1-0.3 mm to break stiction-induced drag after release — without it the pads ride against the rotor.
- HVAC and Building Services: Belimo damper actuators in VAV boxes oversize the breakaway torque by 2-3× the steady-state torque because the damper shaft sits stationary for hours and accumulates dwell-time stiction.
The Formula Behind the Stiction
The classic Coulomb friction model splits friction into a static (stiction) value and a lower kinetic value. What the model gives you is the breakaway force needed to start motion as a function of normal load and the static coefficient. At the low end of typical operating conditions — well-lubricated, polished surfaces, short dwell — μs drops near μk and stiction is barely noticeable. At the nominal condition for a typical industrial contact, μs sits 30-50% above μk and you get a clean breakaway step. At the high end — dry contact, corroded surfaces, long dwell — μs can climb to 2-3× μk and the system flat-out refuses to move below a hard threshold. The sweet spot for stable control is keeping the contact in the mixed-lubrication regime where μs/μk stays below 1.3.
Variables
| Symbol | Meaning | Unit (SI) | Unit (Imperial) |
|---|---|---|---|
| Fbreakaway | Tangential force required to initiate sliding | N | lbf |
| μs | Coefficient of static friction (dimensionless) | — | — |
| N | Normal force pressing the surfaces together | N | lbf |
| μk | Coefficient of kinetic friction (for comparison; once moving) | — | — |
Worked Example: Stiction in a tape-storage cleanroom robotic library
An archival tape library integrator in Boulder is sizing the linear stage that moves an LTO-9 cartridge picker across a 4 m rail in a Spectra Logic TFinity-class library. The picker carriage rides on two THK SHS25 linear guides, total preload-plus-payload normal load of 180 N across the four blocks, and the customer reports the carriage occasionally fails to start moving from rest after sitting idle over a weekend. They need to know the breakaway force at three lubrication states to size the AC servo correctly.
Given
- N = 180 N
- μs (well-lubricated, fresh grease) = 0.05 —
- μs (nominal, after 6 months service) = 0.10 —
- μs (degraded, dry rail after weekend dwell) = 0.20 —
- μk (running) = 0.04 —
Solution
Step 1 — compute breakaway force at the well-lubricated low-end of the operating range, fresh THK AFB-LF grease:
At 9 N breakaway, the servo barely notices. The carriage glides off the rest position and the position loop tracks cleanly with no observable dead-band. This is what the system feels like the day it ships.
Step 2 — at nominal operating condition after 6 months of duty cycle, grease partially depleted at the wiper seals:
18 N is still well within the 50 N continuous thrust of a typical 200 W servo on a 5 mm-pitch ballscrew, but you'll start to see a 20-50 ms hesitation on small commanded moves under 0.5 mm. Not a fault — just a hint the lube schedule is overdue.
Step 3 — at the high end of the operating range, after a 60-hour idle dwell with grease aged out and a thin oxide film forming on the rail:
36 N breakaway is 4× the kinetic running force of μk × N = 0.04 × 180 = 7.2 N. The servo command-current spikes hard, the carriage breaks free, overshoots by 100-300 µm, and the position loop hunts for half a second before settling. This is exactly the symptom the customer reported. The fix is either a regreasing interval before that point or a small dither command (±5 µm at 100 Hz) to keep the blocks in continuous micro-motion through the weekend.
Result
Nominal breakaway force is 18. 0 N. That's the threshold the servo must clear before the carriage moves — small enough to ignore in steady-state operation but large enough to chew up the bottom 1-2% of the position loop's small-signal response. Across the operating range you see 9 N when fresh, 18 N at nominal, and 36 N when degraded — a 4× swing driven entirely by lubrication state, not load. If the measured breakaway in the field reads 50 N or higher instead of the predicted 36 N, check three things in order: (1) wiper-seal damage letting cleanroom dust into the ball tracks, which spikes μs well past 0.20; (2) rail misalignment over the 4 m span causing block side-loading that increases effective N beyond the nominal 180 N; (3) servo brake not fully releasing, which adds a constant offset force the friction model doesn't capture.
When to Use a Stiction and When Not To
Stiction is a property to manage, not a mechanism to select — but the engineering decision in front of a designer is usually "how do I reduce stiction in this contact" with three viable options: stick with sliding contact and lubricate well, switch to rolling-element bearings, or eliminate contact entirely with air or magnetic bearings. Each has clear trade-offs on cost, achievable stiction, load capacity, and lifetime.
| Property | Lubricated Sliding Contact (stiction managed) | Rolling-Element Bearing | Air Bearing |
|---|---|---|---|
| Typical breakaway force (180 N normal load) | 9-36 N depending on lube state | 0.5-2 N | <0.05 N |
| Cost per linear metre (industrial grade) | $50-150 | $300-800 (THK SHS25 class) | $3,000-8,000 (Aerotech ABL) |
| Load capacity | Very high — limited by surface fatigue | High — 10-50 kN per block | Low to medium — 100-2,000 N typical |
| Maintenance interval before stiction doubles | 3-12 months (regrease) | 2-5 years (relube) | Filter change at 6-12 months |
| Suitability for sub-micron positioning | Poor — stick-slip dominates | Marginal — preload helps but not enough | Excellent — sub-nm achievable |
| Failure mode when neglected | Breakaway climbs 3-4×, loop hunting | Brinelling, then catastrophic seizure | Crash on air supply loss |
| Control-loop friendly? | Requires dither or feed-forward compensation | Mostly — small residual stiction remains | Yes — friction effectively zero |
Frequently Asked Questions About Stiction
Because no PID tuning will compensate for stiction larger than about 2% of valve travel — you're chasing a non-linear discontinuity with a linear controller. The integral term winds up while the valve sits stuck, the output finally crosses the breakaway threshold, the valve jumps past the setpoint, integral unwinds, valve sticks again, repeat. This is a textbook stiction-induced limit cycle and the fix is mechanical, not algorithmic.
Quick diagnostic: plot valve command vs valve position. If you see a parallelogram instead of a straight line, that's stiction. The width of the parallelogram is the stiction band. Above 5% width you need to repack the valve, replace the stem, or fit a smart positioner with a stiction-compensation algorithm like the one in the Fisher FIELDVUE DVC6200.
Dwell-time effect. Stiction grows logarithmically with how long the surfaces have been stationary under load. The asperity micro-junctions creep and grow over hours to days — by Monday morning you can see 20-50% higher breakaway than after a 10-second pause on Friday. This is well-documented in tribology literature and it's why dither signals exist.
If a process needs predictable startup torque after long idle, either run a brief exercise cycle before the production cycle or apply a continuous low-amplitude dither during idle. A 1% command modulation at 50-100 Hz is usually enough to keep the contact in presliding micro-motion and reset the dwell clock.
Counter-intuitively, no. Slideway oils like Mobil Vactra No. 2 contain specific tackifier and anti-stick-slip additives precisely because thin oils squeeze out of the contact at zero velocity and let metal-to-metal asperity contact happen — which spikes stiction. The high-viscosity slideway formulation maintains a boundary film at rest and during the critical breakaway moment.
The general rule: at low sliding speeds you want a lubricant with strong boundary additives, not low viscosity. Save the thin oils for high-speed hydrodynamic applications where film thickness, not boundary chemistry, governs friction.
Cost and required positioning resolution drive the decision. If your application tolerates 10-50 µm positioning error, dither on a sliding contact at $50-150 per metre is cheap and reliable. If you need sub-5 µm, the residual stick-slip even with dither will cost you — switch to a recirculating-ball linear guide like THK SHS or HIWIN HG which knocks breakaway down by 10-20×.
Sub-micron territory means rolling-element bearings won't get you there either — preload friction in the balls themselves becomes the floor. That's where air bearings or flexure stages take over. The honest break-even point: above 10 µm resolution use sliding plus dither; 1-10 µm use rolling-element; below 1 µm use non-contact.
Seal stiction. Pneumatic cylinder seals — Buna-N or polyurethane lip seals — adhere to the cylinder bore over hours of dwell. The breakaway pressure on the first stroke can be 2-3× the running pressure. Once the seal is moving, it rides on a thin grease film and stiction drops back to normal.
Standard fixes: switch to a low-friction seal like a Hallite 605 or a PTFE-faced seal, increase the bore finish to Ra 0.2-0.4 µm, and make sure the cylinder grease is rated for long-term static contact (lithium-complex EP greases are good; standard chassis grease is not). On critical axes, a brief exercise pulse at startup eliminates the symptom entirely.
A constant-μ Coulomb model can't capture stick-slip because it has no memory and no presliding regime. You need a dynamic friction model — the LuGre model is the industry standard. It treats the contact as a bristle with internal state variables that capture presliding displacement, the Stribeck velocity-dependence, and the static-to-kinetic transition. Most servo simulation toolchains including Simscape Multibody have a LuGre block ready to drop in.
If LuGre is overkill, the Karnopp model is a simpler alternative that handles the zero-velocity dead-band correctly without modelling presliding. It's enough to reproduce limit cycles and tune compensators, and it runs about 5× faster than LuGre in fixed-step simulation.
Run a step response at two amplitudes — a large step (20% of stroke) and a small step (0.5% of stroke). A pure tuning problem scales — both responses show similar overshoot percentage and settling time. Stiction does not scale — the small step shows pronounced dead-band, delayed response, and a ratcheting overshoot pattern while the large step looks clean.
If the small-step response shows the spool sitting still for 50-200 ms before suddenly jumping to position, that's spool stiction in the proportional valve, almost always caused by silt lock from contaminated fluid. ISO 4406 cleanliness worse than 18/16/13 will produce this within weeks of operation.
References & Further Reading
- Wikipedia contributors. Stiction. Wikipedia
Building or designing a mechanism like this?
Explore the precision-engineered motion control hardware used by mechanical engineers, makers, and product designers.
