A centrifugal pump mechanical seal is usually the first part to fail and the last one to be specified. Most pump quotations carry it as a single line — “mechanical seal” — and the face grade, the elastomer and the flush arrangement stay open until the seal is already leaking onto the baseplate. The Texas A&M Turbomachinery Laboratory records why that matters: industrial pumps are most often pulled from service when a seal leak is detected, so pump reliability depends on seal reliability.
What follows is not a catalogue of seal types. It works from two facts — that a centrifugal pump seal fails in only two ways, and that its faces come from a small set of material grades with published limits — and turns them into the four questions to settle before the order is placed.
Key takeaways
- A mechanical seal is two static joints and one running interface, not a single part. A leak from any of the three looks identical from outside.
- The DOE names two failure mechanisms: degradation of the face material, and loss of spring or bellows tension. Every other cause is a route into one of those two.
- Of the eight causes the USACE table lists for a leaking seal, only two are about the seal. The other six are alignment, shaft, balance, bearings, speed and particles.
- Whether the faces need to be self-lubricating decides soft-against-hard or hard-against-hard, and that question comes before material grade.
- Reaction-bonded and sintered silicon carbide are different materials: 8 to 12 percent free silicon in the first lowers friction by roughly 10 percent and reacts with strong acids and bases.
- The flush plan is the heat exchanger a plastic pump does not have. Polypropylene conducts heat far less readily than metal, so the seal chamber depends on the flush.
- A quotation that says only “mechanical seal” has not specified a seal. Face grade, elastomer, spring position and flush plan are the four lines to demand.
What a Centrifugal Pump Mechanical Seal Has to Seal
The DOE’s own diagram of the stuffing box names the parts that make up a centrifugal pump mechanical seal: the dynamic seal face, the stationary seal face, the seal rings, the spring, the O-rings, and the seal housing or gland. Each of those is a seal in its own right, doing a different job.
Two Static Joints and One Running Interface
Two of the joints do not move against each other. The stationary face is held to the pump housing, usually with an O-ring or a gasket; the rotating assembly is held to the shaft, again with an O-ring. Both are static seals in the sense that the surfaces they join do not slide. The third joint is the only one that runs: the two lapped faces turn against each other, separated by a film of the pumped liquid.
That arrangement explains a confusing field symptom. A leak at the gland joint, a leak past the shaft O-ring and a leak across the faces all present the same way — fluid on the baseplate — which is why replacing the seal does not always stop the leak.
What the Fluid Film Is Doing
The film between the faces carries heat away and keeps the faces from touching. It comes either from the process liquid or from an external source. When it is missing, the faces run in contact and generate heat faster than the material can shed it — the DOE’s “overheating due to inadequate lubrication” in practice.
Packing is the older alternative. The DOE’s comparison is worth keeping in view: packing is cheaper and needs less skill, but it has to leak to stay cool, and allowable leakage is usually 2 to 60 drops per minute. A mechanical seal is chosen when that drip is not acceptable.
Two Failure Mechanisms, Not a Catalogue of Causes
The DOE reduces mechanical seal failure to two mechanisms, and the reduction is the most useful sentence in the literature: “degradation of the face material and loss of spring or bellows tension, which allows the faces to separate more easily.” Everything else that gets called a cause is a route into one of those two.
Mechanism One: Face Material Degradation
The DOE attributes it to debris: “Degradation of the seal face is usually caused by debris that wedges into a seal face and causes damage.” A particle that lodges between two lapped faces holds them apart, and the gap it opens becomes a leak path that widens as the faces wear against the trapped particle.
This is why a centrifugal pump seal in a dirty service fails for reasons that have nothing to do with the material being cheap. The face may be the correct grade and still be destroyed by what is between it and its partner.
Mechanism Two: Loss of Spring or Bellows Tension
Springs and bellows lose their compressive force to fatigue, fouling and corrosion. The DOE’s answer for corrosive duty is a design change rather than a material upgrade: “In systems with highly corrosive fluids, mechanical seals with external springs are recommended.” An external spring sits outside the pumped fluid, so the fluid cannot corrode it and fouling cannot pack around it.
The Eight Causes of a Leaking Seal, Sorted
USACE EM 1110-2-3105 reproduces the Pump Handbook troubleshooting table. For the symptom “mechanical seal leaks excessively” it lists eight causes. Sorted by what they actually are, the list reads differently from how it is usually quoted.
| Cause of a leaking mechanical seal | Is it the seal? |
|---|---|
| Misalignment | No — shaft and coupling |
| Bent shaft | No — shaft |
| Rotor out of balance | No — rotor |
| Shaft running off-centre from worn bearings | No — bearings |
| Pump running at or near critical speed | No — rotor dynamics |
| Hard solid particles in the running clearances | Partly — the seal is the victim |
| Incorrect type of mechanical seal | Yes |
| Mechanical seal improperly installed | Yes |
Six of the eight are not seal problems at all. That is the practical value of the table: when a seal has failed twice in a year, the second replacement is usually the wrong response, because the cause was already on the list and was not the seal. A worn bearing moves the shaft, the shaft moves the faces, and the faces open.
The Lubrication Regime Decides the Face Pair
The first question in face selection is not which material is best. It is whether the faces will ever touch.
Phase First, Viscosity Second
The TAMU tutorial puts the phase question first: whether the sealed product enters and leaves the sealing gap as a liquid, as a gas, or as a liquid that vaporises before it exits. In the two cases where the product leaves as a gas, “the lubrication mode is going to be boundary lubricated or mixed lubricated”, so the faces will touch and the material pair has to tolerate contact.
Most seals sit in the liquid case, and there viscosity decides. The tutorial’s guidance, considering viscosity alone, is short enough to memorise.
| Viscosity (cSt) | Face combination |
|---|---|
| 0.3 | Carbon against a hard face |
| 3 | Any combination |
| 30 | Hard against hard |
The table carries its own limit — it considers viscosity only, not phase, abrasives, temperature or face topography — and it has one counter-intuitive edge. Carbon is not recommended in high-viscosity service, because the viscous shear between the faces can cause a localised binder failure known as blistering. A thicker fluid is not gentler on a carbon face. It is harder on it.
Soft Against Hard, or Hard Against Hard
Soft against hard is for boundary and mixed lubrication, where the faces touch and one of them has to give. Hard against hard is for full fluid film, where the faces should not touch at all and hardness buys abrasion resistance instead of forgiveness. Choosing hard against hard in a boundary-lubricated service removes the self-lubrication and puts nothing in its place.
One velocity note is worth keeping: below about 350 RPM, with lower pressures, self-lubricating faces may run fully boundary lubricated for a long period. That describes a condition, not a limit to design to.
What the Hard Face Materials Are Actually Good At
Hard faces are bought for three properties, and they do not move together.
Hardness Buys Abrasion Resistance, Not Life
The TAMU hardness comparison puts sintered silicon carbide at 21.6 GPa and reaction-bonded silicon carbide at 16.7, above nickel-bound tungsten carbide at 11.6 and aluminium oxide at 14.0. Carbon sits near 6. The tutorial’s conclusion is direct: both SiC grades “are the hardest solid composition seal face materials available”, and for the harshest slurries SiC is chosen for both faces.
Hardness protects against particles. It says nothing about heat or chemistry.
Fracture Toughness and Heat Checking
Tungsten carbide’s fracture toughness is 3 to 5 times that of silicon carbide and aluminium oxide, which is why it survives light impact that would crack a ceramic. Its weakness is thermal: the expansion mismatch between the metal binder and the carbide is over 2 to 1, which makes tungsten carbide prone to surface thermal cracking known as heat checking.
Resistance to a sudden temperature change separates the ceramics sharply.
| Face material | Max instantaneous temperature change (°C) |
|---|---|
| Tungsten carbide | 285 |
| Silicon carbide | 205 |
| Aluminium oxide | 63 |
Aluminium oxide’s 63 °C is why it belongs in clean, steady, low-duty service and not in anything that starts cold and runs hot.
The Chemical Ratings Nobody Prints
Every seal supplier publishes a list of face materials. Very few publish what each one is not for. The two tables below come from Huebner’s Material Selection for Mechanical Seals, reproduced in the TAMU tutorial, and they rate eight materials from A to D against the conditions that decide seal life.
Reaction-Bonded and Sintered Silicon Carbide Are Not the Same Material
Both are sold as “silicon carbide”. The difference is free silicon. Reaction bonding leaves 8 to 12 percent of the finished material as free silicon, which lowers the coefficient of friction against common opposing faces by roughly 10 percent — and which “may react with strong acids and bases”. Direct sintering leaves no free silicon at all.
That is why the two grades rate differently in caustic. In the compatibility table, reaction-bonded silicon carbide is rated D against bases; the sintered grade is rated A. Buying “SiC” without the grade is buying a coin flip.
| Medium | Resin carbon | Antimony carbon | Dry-running carbon | Acid-grade carbon | Reaction-bonded SiC | Sintered SiC | Tungsten carbide | Aluminium oxide |
|---|---|---|---|---|---|---|---|---|
| Acids | B | D | B | A | C | A | B | A |
| Bases | B | D | B | A | D | A | B | A |
| Overall chemical resistance | B | C | B | A | B | A | B | A |
A is excellent, B good, C fair, D poor. The ratings describe a face running against common opposing faces, not a bare coupon in a beaker.
The Carbon Grade Matters as Much as the Hard Face
The soft face is usually carbon, and carbon is not one material either. Antimony-impregnated carbon is rated D against both acids and bases; acid-grade carbon is rated A against both. The cheapest carbon in the catalogue is the one to leave out of chemical duty, and the grade name on the data sheet is what tells you which one you have.
| Property | Resin carbon | Antimony carbon | Dry-running carbon | Acid-grade carbon | Reaction-bonded SiC | Sintered SiC | Tungsten carbide | Aluminium oxide |
|---|---|---|---|---|---|---|---|---|
| General duty service | A | C | D | D | A | A | B | C |
| Dry running | B | D | A | C | C | C | D | B |
| Abrasive service | D | D | D | D | A | A | A | C |
| Blister resistance | C | A | D | C | A | A | C | C |
| Impact resistance | B | A | B | B | C | C | A | C |
| Thermal shock resistance | A | A | A | A | A | A | A | D |
Read across the abrasive row and the pattern is unambiguous: all four carbon grades are D. A carbon face in a slurry is a maintenance schedule, not a seal. Read down the tungsten carbide column and the trade appears: A for abrasion and impact, D for dry running, C for blister resistance.
The Standard Default Became a Constraint
API 682’s first edition in 1994 named carbon against reaction-bonded silicon carbide as the default pairing. The TAMU tutorial notes the unintended effect: a default inside a widely adopted standard effectively disallowed most other combinations. The fourth edition in 2014 added language allowing reaction-bonded and sintered SiC, plus performance-enhancing coatings.
The Flush Plan Is the Heat Exchanger You Do Not Have
Mechanism two — lost spring tension — is a materials problem. Mechanism one, face degradation, is a cleanliness and temperature problem, and the DOE’s remedy for it is plumbing: “mechanical seals are often serviced by special flushing lines that have filters to catch debris.”
What Each Plan Family Does
The numbering comes from API 682. The choice among these groups is made by what the fluid is doing to the seal chamber.
| Plan family | What it does | Use when |
|---|---|---|
| Recirculation from discharge (Plan 11) | Feeds process liquid from the discharge side into the seal chamber through an orifice | Clean liquid; the chamber needs cooling and venting |
| Recirculation through a cooler (Plan 21) | Plan 11 with a heat exchanger in the line | The liquid is hot enough that the chamber would otherwise approach its vapour pressure |
| Recirculation with a separator (Plan 31) | Takes liquid from discharge, drops the solids out, sends clean liquid to the seal | Moderate solids that would otherwise reach the faces |
| Clean external flush (Plan 32) | Injects a compatible clean fluid from outside the pump | Abrasive or dirty process, or where the process must not reach the faces |
| External quench (Plan 62) | Washes the atmospheric side of the faces with a clean fluid | Crystallising or salting liquids that would build up as the leak dries |
| Dual pressurised barrier (Plan 53 or 54) | Holds a barrier fluid between two seal pairs at a pressure above the process | Toxic, hazardous or near-zero-leak duty |
Why a Polypropylene Pump Changes the Calculation
A metal pump sheds seal-chamber heat through the casing. Polypropylene conducts heat far less readily than metal, so the heat generated between the faces stays near the faces instead of spreading into the housing. On a PP centrifugal pump the flush is not an accessory. It is the only route out of the seal chamber for that heat.
The same property explains why a dry run is so destructive in a plastic pump. With no liquid there is no film and no coolant, and the heat that the faces generate has nowhere to go. A dry-run interlock costs less than a seal and a casing cover.
It also caps the assembly. Our polypropylene centrifugal pump series is rated to 60 °C liquid temperature, and the seal’s elastomer and face limits can bind before that ceiling is reached — which is why the seal has to be chosen against the fluid, not against the pump. The material comparison in polypropylene pump vs stainless steel makes the same point from the other direction: the seal is usually the binding constraint before the casing is.
What the Quotation Must State
The centrifugal pump mechanical seal on a PP centrifugal pump is application-matched, and that phrase only means something if the application is written down. Four lines decide whether a quotation describes a seal or a placeholder.
One: the face pair by grade. Not “SiC”, but reaction-bonded or sintered. Not “carbon”, but which grade. The ratings above show that the grade, not the family, is what survives the medium.
Two: the elastomer by family, confirmed in writing. The elastomer is the second seal, with its own temperature and chemical limits — not the resin’s limits and not the pump’s. Ask the supplier to confirm compatibility for the actual medium at the actual temperature.
Three: the spring position. External for corrosive duty, for the reason the DOE gives. A spring inside the fluid is a spring in the fluid’s corrosion schedule.
Four: the flush plan by number, with its flow and pressure. A plan name without a flow rate is a label, not a specification.
A quotation that lists “mechanical seal” and nothing else has left all four open. The seal that arrives will be the one the factory stocks, not the one the duty needs — and the difference shows up as the second failure in a year.
Frequently Asked Questions
How long should a centrifugal pump mechanical seal last?
There is no single figure, and the DOE says so directly: because there are many types of mechanical seal for many applications, “it is difficult to state how long a seal should last”. The more useful question is which of the two mechanisms ended the last one: debris points to a filter or a plan change, lost spring tension to a design change.
Can a mechanical seal run dry?
Not without damage. The faces are separated by a film of liquid; with no liquid there is no film, the faces run in contact, and the heat they generate has nowhere to go. The DOE lists overheating from inadequate lubrication among the three common causes of seal problems. On a polypropylene pump the heat is worse, because the casing will not carry it away.
Is a carbon face always the right choice?
No. Carbon is the usual soft face because it conforms to its partner and self-lubricates, but the ratings show where it stops. All four carbon grades are rated D for abrasive service, and antimony-impregnated carbon is rated D against both acids and bases. The grade matters as much as the material.
Why does a new mechanical seal still leak?
Two of the eight causes the USACE table lists are about the seal itself — incorrect type and improper installation — and either will make a new seal leak. The other six, from misalignment to a worn bearing, will make a correctly installed seal leak as well. A new seal that leaks immediately is usually telling you the cause was never the old seal.
Does a mechanical seal always need a flush plan?
No, but the decision has to be made rather than defaulted. A clean liquid at moderate temperature may need only recirculation from the discharge. Abrasives, crystallising liquids and high temperatures each point to a different arrangement, and on a polypropylene pump the flush is carrying heat the casing cannot remove.
Can a polypropylene pump use a mechanical seal?
Yes, and it usually does. The constraint is not the seal but the temperature at the seal chamber: the pump’s liquid limit applies to the seal’s elastomer and faces as well, so the seal has to be chosen against the fluid and its temperature rather than against the pump alone. For the wider selection logic behind that, see polypropylene pump selection and centrifugal pump parts.
Sources
- U.S. Department of Energy, Industrial Technologies Program, and the Hydraulic Institute, Improving Pumping System Performance: A Sourcebook for Industry, 2nd ed. The two primary failure mechanisms, the three common causes of seal problems, the debris mechanism, the filtered flushing lines, the external-spring recommendation for corrosive fluids, the micron alignment tolerance, the packing leakage figures, and the statement that seal life cannot be stated as a single figure. energy.gov
- U.S. Army Corps of Engineers, Mechanical and Electrical Design of Pump Stations, EM 1110-2-3105, 30 April 2020. Table 17.2, reproduced from the Pump Handbook (McGraw Hill), for the causes listed against “mechanical seal leaks excessively”. publications.usace.army.mil
- Dietzel, W. and Vasko, J., The Evolution and Application of Mechanical Seal Face Materials. Turbomachinery Laboratory, Texas A&M University. The phase and viscosity selection logic, the blistering limit on carbon, the 350 RPM condition, the free-silicon content of reaction-bonded SiC, the fracture toughness and heat checking of tungsten carbide, the hardness, modulus and thermal shock comparisons, and the API 682 edition history. oaktrust.library.tamu.edu
- Huebner, M., Material Selection for Mechanical Seals. Proceedings of the Twenty-Second International Pump Users Symposium, Turbomachinery Laboratory, Texas A&M University. The chemical compatibility ratings and the strengths and weaknesses ratings for the eight face materials, as reproduced in the TAMU tutorial.
- American Petroleum Institute, Pumps — Shaft Sealing Systems for Centrifugal and Rotary Pumps, API Standard 682, 4th ed., 2014. The flush plan numbering system referenced in this article. api.org
