Flatness of a Mechanical Seal Face

Seal face flatness specifications call for two helium light bands — 23 microinches of total deviation. During operation, thermal and pressure distortion routinely push that number past ten light bands. That five-fold gap between the bench reading and operating reality is where most “unexplained” seal leakage originates.

Flatness is not a single pass/fail checkpoint. It is a dynamic property that changes the moment the seal starts running, and the pattern of deviation — concave, convex, or localized — tells you more about the root cause than the HLB count alone. The API 682 standard specifies flatness tolerances for a reason: the sealing gap between faces is measured in millionths of an inch, and small deviations produce disproportionately large leakage increases.

What Seal Face Flatness Means in Quantitative Terms

One helium light band (HLB) equals 0.0000116 inches, or 0.3 microns — half the wavelength of helium light at 587.6 nm. A typical lapped mechanical seal face is flat to within 23 millionths of an inch, which translates to two light bands. At this scale, you are measuring deviations smaller than a bacterium.

Flatness describes the macro-geometry of the face: whether the surface is truly planar or has warped into a concave, convex, or saddle shape. It is measured in HLB using an optical flat and monochromatic light. This is fundamentally different from surface finish.

Flatness vs Surface Finish

Surface finish (Ra) measures micro-roughness — the tiny peaks and valleys across the face at the microscopic scale. Flatness (HLB) measures the overall shape deviation from a true plane. A face can have excellent surface finish (Ra 0.1 microns) and still be badly out of flat, or it can be perfectly flat with a rough surface. Both parameters matter, but they measure different things and use different instruments.

Typical Ra values vary by material: tungsten carbide at 0.01 microns, silicon carbide at 0.04 microns, ceramic at 0.07 microns, and hard carbon at 0.1 microns. These numbers coexist with but are independent of the flatness reading. A face with Ra of 0.01 microns and 5 HLB of flatness deviation will leak — the smooth surface finish does not compensate for a warped geometry.

How Flatness Is Measured with an Optical Flat

Place a reference-grade optical flat on the seal face and illuminate it with monochromatic light — sodium or helium. Light waves reflect off both the flat’s bottom surface and the seal face. Where the gap between surfaces changes by half a wavelength, an interference fringe appears. These fringes form visible bands that map the face topography like contour lines on a terrain map.

Each band represents a height change of one half-wavelength: 0.0000116 inches for helium light. Lay an imaginary line tangent to one band and parallel to the contact line (where the flat touches the face); the number of bands it crosses is your flatness reading in HLB.

Reading the Band Pattern

The band count tells you the magnitude of deviation. The band shape tells you the type.

Straight, parallel, equally spaced bands mean the face is flat. This is what you want to see. Any curvature in the bands indicates a deviation from true flatness.

Bands curving away from the contact line indicate a convex surface — the face is higher at the center than the edges. Bands curving toward the contact line indicate a concave surface — the face is dished. A face showing heat check marks alongside curved bands has likely suffered thermal damage that affects both surface integrity and flatness.

Localized band distortion — where bands compress or fan out in one area — points to a localized defect: a chip, a hard spot in the material, or damage from improper handling during installation.

I’ve seen technicians count two light bands and call it a pass without ever looking at the band shape. A two-band reading with severely curved bands means a convex or concave face that happens to have low total deviation — but the non-uniform contact pattern will still cause accelerated wear on the high side of the face.

Material-Specific Flatness Tolerances

Not all seal face materials share the same flatness requirement. Applying a single tolerance to every material — a common oversimplification — risks accepting out-of-spec hard faces or needlessly rejecting normal carbon faces.

The tighter tolerance for hard face materials exists because SiC, TC, and ceramic faces do not flex. Carbon, being softer, conforms slightly against its mating face under spring and hydraulic load. A hard face at 3 HLB stays at 3 HLB in service — the material has no give.

Carbon Face Relaxation

What the spec sheet doesn’t tell you is that a new carbon face straight from the manufacturer will often read 2 to 3 light bands on the bench, even though it was lapped to less than one light band at the factory. This is normal carbon relaxation, not a defect. Carbon releases internal residual stress after lapping, causing the face to “spring back” slightly. These faces return to flat once clamped against a hard mating surface under spring load.

I’ve seen maintenance teams reject perfectly good carbon faces over a 3 HLB bench reading. The context matters: a 3 HLB reading on a new carbon face is normal relaxation, while 3 HLB on a used silicon carbide face is a reject. Carbon and hard face materials require different evaluation criteria, and the flatness tolerance tables reflect this material-dependent reality.

Why Flatness Degrades During Operation

A seal face that passes bench inspection at 2 HLB does not stay at 2 HLB once the pump starts. Three mechanisms work against flatness during operation, and each leaves a distinct signature on the face.

Thermal Distortion

Friction between rotating seal faces generates heat at the contact interface. The thermal conductivity mismatch between typical face pairs concentrates that heat unevenly: carbon sits at 8.6 Btu/hr.ft.F while tungsten carbide conducts at 58 Btu/hr.ft.F. The carbon face absorbs and retains heat while the hard face conducts it away rapidly. This asymmetric heating warps the carbon face into a concave shape, creating an ID-preferential contact pattern.

When you pull a face and see heavy wear concentrated at the inside diameter, thermal distortion is your primary suspect. Inadequate flush flow, high face pressure, or operating above design temperature all accelerate this mechanism.

Pressure-Induced Coning

Hydraulic pressure acting on the seal face creates a bending moment that distorts the face geometry. High-pressure applications — above 40 bar — push the face into a cone shape, concentrating contact at the outside diameter. This is why high-pressure seals require flatness within 1 HLB: the operating conditions will add deviation, and you need the tightest possible starting point.

An OD contact pattern on a used face is the diagnostic fingerprint of pressure-induced coning. Where thermal distortion pushes contact inward, pressure coning pushes it outward. The wear pattern is not cosmetic — it is a diagnostic tool.

Installation and Assembly Damage

Even manufacturing processes can destroy carefully achieved flatness. Research on aircraft engine carbon seals found that 30 to 50 percent of seals failed to meet the required flatness tolerance of three light bands after shrink-fit assembly. The residual stresses from the assembly process itself warped the faces.

In pump applications, the same principle applies on a smaller scale. Over-tightening gland bolts, uneven bolt torque, piping strain transmitted through the pump casing, and contamination between the face and its seat all introduce flatness deviations before the seal ever sees process fluid.

Why a Face That Passes Inspection Still Leaks in Service

A face reads 2 HLB on the bench — well within spec. The seal gets installed, the pump starts, and leakage appears within hours. The face “passed inspection.” What went wrong?

The bench reading measures flatness at room temperature, with no pressure, no rotation, and no thermal load. Operating conditions change every one of those variables simultaneously. The Fluid Sealing Association documented that operational distortions routinely produce flatness deviations of ten light bands or more — five times the specification that the face passed at the bench.

The consequences follow a cubic relationship. Leakage is proportional to the cube of the face separation. Double the face separation and leakage increases by a factor of eight. Triple it and leakage increases twenty-seven-fold. A face operating at 10 HLB instead of the specified 2 HLB is not leaking five times more — the relationship is far more severe than that because the gap increase is not uniform, and the highest-deviation zones dominate the total leakage path.

This cubic sensitivity is why I never treat flatness as a one-time pass/fail check. The bench reading establishes the starting point. What determines seal life is how quickly thermal distortion, pressure coning, and operational stresses push that starting point beyond where the fluid film can maintain a seal.

Before replacing the seal, check the contact pattern on the failed faces. An ID wear band points to thermal distortion — review flush plan adequacy and operating temperature. An OD wear band points to pressure coning — evaluate whether the seal design is rated for the actual system pressure. A localized worn spot points to contamination, a chip, or assembly damage. The deviation pattern tells you the root cause. Fix that, and the replacement seal lasts.

What to Do with a Flatness Reading

When a used face comes back out of spec, the contact pattern drives the next step. Uniform wear across the full face width often means the face can be restored by lapping. ID-concentrated or OD-concentrated wear means the root cause — thermal or pressure distortion — must be addressed before lapping will help. Relapping a thermally distorted face without fixing the flush plan just gives you the same failure on a shorter timeline.

The API 682 standard specifies these tolerances because they are the maximum starting deviation the seal design can absorb before the cubic leakage relationship overwhelms the fluid film. Most flatness failures I investigate turn out to be system problems — inadequate flush, excessive pressure, or piping strain — not bad seal faces. Fix the system, and the flatness holds.