Carbon vs Silicon Carbide Mechanical Seal: 3 Duties Where SiC Is the Wrong Upgrade

API 682’s default face pair is premium blister-resistant carbon against silicon carbide — not SiC against SiC. The industry standard governing 25,000-hour continuous service contradicts the marketing line that SiC/SiC is automatically the upgrade.

I’ve pulled more SiC/SiC pairs with thermal-shock edge chipping than carbon/SiC pairs with wear, and every one cost roughly 1.6 to 1.7 times as much as the carbon/SiC pair that would have outlasted it. The carbon vs silicon carbide decision is not a hardness contest. It’s a three-axis duty match, and getting it wrong usually means over-specifying SiC where carbon/SiC would have won.

Carbon Face Seal vs SiC Face: Head-to-Head Properties

The Property Table

The spec-sheet comparison favors silicon carbide on every axis engineers instinctively look at, and that is exactly the trap. A carbon face seal and a silicon carbide face solve different problems under different duties.

The Rows That Actually Decide

The rows that matter are not hardness and conductivity. They are mounted temperature, free silicon content, and dry-run tolerance.

The carbon graphite seal face brings one thing SiC cannot: graphite basal planes that shear into a lubricating film when the fluid stops cooperating. That film is why API 682 kept carbon as the default on one side. Treating RBSiC and SSiC as interchangeable silicon carbide seals hides a specification error the field reveals as a destroyed carbon mating face.

What Disadvantages Does Silicon Carbide Have?

Silicon carbide has three disadvantages the spec sheet never prices in: dry-run failure, free-silicon leaching in caustic and HF, and a mounted-temperature derating that erases most of the advertised “800 °F capability.” Every one has sent me back to carbon/SiC on a pair I originally specified as SiC/SiC.

Dry-Run and Vapor-Flash Failure

SiC/SiC pairs do not survive loss of lubrication. I’ve seen this failure pattern dozens of times: a flush line vapor-locks, the pump runs seconds dry, and both faces glaze, crack, and seize before anyone reaches the panel.

Carbon/SiC in the same duty usually survives because graphite’s basal planes shear into a solid-film lubricant the moment liquid film breaks down. That’s not marketing — it is the reason API 682 keeps carbon on one side. Carbon’s primary value in a seal is self-lubrication, not strength.

The same logic applies within 20-30 °F of the fluid’s vapor point, where local flashing across the face gap is effectively intermittent dry-run. Any application needing dry-running materials rules out SiC/SiC — the cost multiplier is wasted.

RBSiC Free-Silicon Leaching in Caustic and HF

RBSiC should not be used at pH above about 10. The 8-15% free silicon metal in reaction-bonded SiC is leached out by NaOH, KOH, and HF, leaving a porous skeleton of SiC grains. McNally Institute’s description is vivid and accurate: the leached surface acts like a grinding wheel against the softer carbon mating face.

SSiC does not have this problem — no free metal phase, chemically resistant across nearly the full pH range. If your chemistry is caustic or HF, specifying “silicon carbide” without saying SSiC is a specification error. This is a core criterion in corrosive media material selection.

SSiC Mounted-Temperature Derating

The spec sheet says SSiC handles 800 °F. The reality in a 316SS cartridge body says 250 °F. The steel holder has roughly 3-4 times the thermal expansion coefficient of SSiC, so every thermal transient pulls hoop stress on the face OD corner.

Edge chipping initiates there. I’ve pulled cartridges with neat circumferential edge-chip patterns at the OD the operator blamed on solids — it was thermal shock from a 100 °F startup transient on a face the datasheet said could tolerate 800 °F.

Tungsten carbide is the pragmatic middle: 750 °F solid, still 500 °F in 316SS. For pumps with vibration, cavitation, or thermal cycling, TC is less fragile than SSiC and keeps twice the mounted-temperature margin — the broader carbon vs ceramic vs carbide landscape matters when the duty is unsettled.

When Carbon/SiC Is the Correct Pair (Not the Cheap Fallback)

Carbon/SiC is the correct specification — not the fallback — any time the duty includes dry-run exposure, vapor-point proximity within 30 °F, or thermal-transient startup and shutdown cycles. These describe the majority of clean-fluid utility and process duties in real plants.

Thermal Transients

The axis most engineers miss is thermal transient. A pump that sees a 100 °F cold-to-hot ramp every Monday morning is better served by one forgiving material (carbon) plus one hard material (SiC) than by two hard materials with mismatched CTEs against their holders.

Carbon absorbs the ΔT by micro-conforming; SiC distributes the load through its high conductivity. SiC against SiC has neither escape mechanism.

Vapor-Point Proximity and Clean Duty

For light hydrocarbons, boiler feedwater, or any duty within ~20 °F of flash, carbon/SiC wins because graphite’s solid-film lubrication keeps the faces from scoring the moment liquid film breaks. SiC/SiC accumulates flash damage on every excursion until edge chipping begins, then fails.

For low-solids clean-fluid duties — most cooling water, utility, general process — there is no mechanical argument for SiC/SiC. The longest-lasting pair is the one API 682 specifies as default: premium blister-resistant carbon against RBSiC or SSiC, code N or R.

When SiC/SiC Is the Right Upgrade

SiC/SiC earns the 1.6-1.7× cost multiplier in exactly three scenarios: high-solids abrasive slurry above ~5% by weight, hard-crystalline service where crystals will abrade carbon faster than the film can rebuild, and duties kept continuously wet with no dry-run exposure.

Abrasive Slurry (the Defensible Case)

Abrasive slurry is the cleanest case for SiC/SiC. Supplier wear-resistance figures above 40% in abrasive media are directional, not precise, but the physics is real — SiC’s hardness survives mineral particles that would shred a carbon face in weeks. For hard-mineral abrasive slurry service, seal face material selection should start with SiC/SiC, not end there.

A counter-intuitive caveat from McNally: properly lapped faces (flatness under one helium light band, roughly 0.3 micron) already filter solids through the face gap. For marginal slurry — under 3% by weight, soft particles, lubricating carrier — carbon/SiC with tight flatness often outperforms SiC/SiC because the carbon film adds a margin SiC/SiC cannot match. The threshold crosses over around 5%.

Crystallizing Service

For high-concentration brines near saturation or specific polymer streams, SiC/SiC avoids the failure mode where crystals embed in a carbon face and act as third bodies against the hard mate. Narrower case than slurry, but real.

The Cost and API 682 Reality

SiC costs roughly 4 to 6 times as much as carbon per face, so a SiC/SiC pair runs only about 1.6 to 1.7 times the cost of a carbon/SiC pair — industry-estimated figures. That multiplier has to be justified by a duty condition, not absorbed as a default upgrade.

What the Standard Actually Specifies

API 682 4th Edition makes the default explicit: one ring shall be premium grade, blister-resistant carbon graphite, except for services requiring two hard materials. The 25,000-hour service target in the standard was set with carbon-against-SiC as the baseline, not SiC/SiC. The exception proves the default.

The 4th Edition task force could have specified SiC/SiC universally and didn’t. Carbon stayed on one side because field data showed it delivered longer service life across the refinery and chemical duties the standard covers.

The Procurement Logic

The procurement argument for SiC/SiC standardization — single-SKU, simpler inventory — has real logic. It just isn’t field-margin logic.

If stock-room pressure is real and your duty population is clean fluids with no dry-run exposure, standardizing on carbon/SiC gives the same SKU simplicity and costs roughly 40% less than the SiC/SiC alternative. For a standard pump seal with Cowseal’s standard mechanical seal face options, the default spec — carbon against RBSiC or SSiC — is also the cheapest spec that meets API 682’s service life target.

The Three-Axis Verdict

Before you specify SiC/SiC, check three things: dry-run exposure, thermal-transient severity, and vapor-point proximity. If any one is present, stay with carbon/SiC — the 1.6-1.7× cost multiplier buys you a new failure mode, not longer life.

Then check the fourth axis, chemistry: if pH exceeds about 10, or HF is in service, RBSiC is out entirely, and the pair is carbon against SSiC.

Everything else — hardness, thermal conductivity, surface finish — is secondary to those four questions. API 682 kept carbon as the default for a reason, and the reason shows up on every failure pull where someone upgraded what didn’t need upgrading.