Products

News

Learn more industry news

Which SF6 Purity Do You Need? 99.5%, 99.9%, 99.95% or 99.995% by Application

For new gas going into switchgear, 99.9% is the working grade and 99.95% is what you buy when the project writes a moisture or dew-point limit into acceptance. 99.5% is the reconditioned grade for reuse loops and top-ups on gas that was already in service, not a cheap substitute for new gas. 99.995% belongs to semiconductor, analytical and research work, where the impurity floor is part of the measurement rather than part of the insulation. If you are buying 99.995% to insulate a 145 kV bay, you are paying for something the equipment cannot use.


Purity is a specification, not a safety ladder

The instinct is to treat the four grades as rungs: 99.5 is basic, 99.995 is best, so buy as high as the budget allows. That logic works for gold. It does not work here, for two reasons.

First, the grades are defined by different standards serving different purposes. IEC 60376:2018 covers new technical-grade SF₆ for electrical equipment. IEC 60480:2019 covers gas recovered from equipment and prepared for reuse, and 99.5% is that standard’s grade. Buying 99.5% new gas and putting it into a new GIS under warranty is not a saving. It is a warranty argument waiting to happen, because the OEM specified a new-gas grade and you delivered a reuse grade.

Second, the number is a residual. 99.9% means 1,000 ppm of “everything else”, and the identity of that 1,000 ppm matters more than its size. A lot with 900 ppm of nitrogen and 60 ppm of water behaves differently from a lot with 500 ppm of nitrogen and 400 ppm of water, even though both certify at 99.9%. This is why the certificate of analysis matters more than the headline figure, and why a supplier who can only quote a percentage is telling you something about their lab.

So the right question is never “what is the highest purity I can afford”. It is “what does this equipment, process or standard require, and what is the cheapest grade that meets it with margin”.


The four grades and the work each one actually does

99.5% — reuse grade, not a budget new-gas grade

99.5% is the reconditioned grade. Gas that has been recovered from equipment, dried, filtered and tested per IEC 60480 can legitimately go back into service at this specification. It is cheaper and it has a lower carbon footprint than new gas, and for a utility running a recovery loop it is the sensible default for refills.

Where it also shows up legitimately: purge and leak-test practice, training rigs, university demonstrations, and non-critical top-ups where the compartment was filled with reuse-grade gas in the first place. Where it should not show up: first fill, warranty work, or any compartment whose OEM manual names a new-gas grade.

One caution on reconditioned gas. The grade is only as good as the treatment behind it. A recovery stream that got mixed across equipment types, or that sat wet in a receiver, can be dried to pass a purity check while carrying decomposition products that a percentage figure will never show. Ask for the decomposition-product results (SO₂F₂, SOF₂, SO₂, HF) on the lot, not just the assay.

99.9% — the standard industrial grade

This is what most SF₆ in the world is sold as, and for good reason. It meets IEC 60376′s requirements for new gas with room to spare, and it covers medium-voltage switchgear, ring main units, load-break switches, and the great majority of 145 kV GIS first fills.

If you are commissioning standard MV equipment and the manual says “new SF₆ per IEC 60376″, 99.9% is the answer. It is also the grade to use for magnesium and aluminium casting cover gas, where SF₆ is used at fractions of a percent in a carrier and the purity requirement is driven by moisture and by what else is in the mixture, not by the last two nines.

99.95% — when the project writes a number into acceptance

The step from 99.9% to 99.95% halves the impurity budget, from 1,000 ppm to 500 ppm, and in practice it shows up as tighter moisture and lower air content. That matters in four situations:

  • HV GIS and GIL projects where the tender specifies a dew point or a moisture limit tighter than the IEC default, and the acceptance test will measure it.
  • Dielectric and partial-discharge test benches, where air content changes the breakdown behaviour you are trying to measure and a 1,000 ppm budget is enough to skew repeatability.
  • OEM acceptance and type testing, where the test report has to state the gas specification and a reviewer will read it.
  • Long-hold compartments, where gas will sit for years between services and the moisture margin is the difference between passing and failing the next test.

Outside those cases, 99.95% is insurance. Insurance is not irrational. Just book it as insurance in your cost model rather than as a technical requirement.

99.995% — process gas, not insulation gas

At 50 ppm total impurity, this grade is bought by people whose problem is measurement or process control.

Semiconductor fabs use SF₆ as an etch gas and for chamber cleaning, and there the impurity profile affects the process, not just the insulation. Analytical labs use it as a high-purity background or as a blend component, where a 50 ppm floor keeps the baseline quiet. Research groups working on gas mixtures, dielectric behaviour or decomposition chemistry buy it because they need to know that what they observe came from the experiment and not from the cylinder. Tracer-gas studies use it where a clean background matters.

What these buyers have in common is that they do not mainly care about the percentage. They care about metals, moisture, particulates and lot-to-lot consistency, and they will often ask for limits that go beyond the SF₆ standard entirely. If that is your world, the conversation is about the impurity panel, not the grade name.

A note on medical use. SF₆ has a legitimate medical application as an intraocular tamponade gas, and that supply chain is pharmaceutical-grade with its own qualification requirements. It is not something to source from an industrial cylinder catalogue, and no purity percentage from an industrial CoA makes it suitable.


The mapping, in one table

Application Grade Typical package What actually drives the choice
MV switchgear, ring main units, load-break switches 99.9% 8L or 40L New-gas grade per IEC 60376; moisture inside 200 ppmv
145 kV GIS first fill 99.9%, 99.95% if the tender specifies 40L OEM manual plus project acceptance criteria
HV GIS / GIL with written dew point limit 99.95% 40L Acceptance test measures moisture; margin is the point
Circuit breaker OEM production and type tests 99.95% 40L Test report has to state the gas spec
Reuse loop, refills after recovery 99.5% 40L IEC 60480 grade, clean lot data
Purge, leak-test training, demo rigs 99.5% 8L Cost per useable kilogram, not purity
Magnesium / aluminium casting cover gas 99.9% 8L or 40L Moisture and mixture accuracy, not last-nines
Semiconductor etch and chamber clean 99.995% (or a fab-specific panel) 40L Process impurity limits, metals, consistency
Analytical background, calibration blending 99.995% 8L Baseline cleanliness
Research on mixtures or decomposition 99.995% 8L Known starting composition

The two failure modes, and what each one costs

Over-specifying costs money on every kilogram, and it costs it quietly. The per-kg premium between 99.9% and 99.995% is not small, and on a 50 kg cylinder it compounds across a production year. It also narrows your supplier pool, which shows up later as longer lead times and less willingness to hold safety stock for you. If nobody in your organisation can point to the clause that requires the higher grade, you are probably over-specified.

Under-specifying costs equipment. Water is the mechanism. Excess moisture drives acid formation, acid attacks surfaces inside the compartment, and the resulting decomposition products are what turn a routine gas check into a repair. The bill arrives years after the purchase order, which is exactly why the purchase order is where the decision should be made carefully.

The asymmetry is worth stating plainly: over-specifying wastes a measurable amount of money, under-specifying risks an unmeasurable amount. When the equipment is a 20-year asset and the gas is a small fraction of its cost, the conservative choice is usually correct. When the gas is a consumable in a process running daily, the calculation flips.


What to write in the purchase order

Write the grade and the impurity limits, not a percentage.

SF₆, new gas per IEC 60376:2018. SF₆ content ≥ 99.9 vol%. Water ≤ 200 ppmv (target ≤ 100 ppmv). Air ≤ 500 ppmv. CF₄ ≤ 500 ppmv. Mineral oil ≤ 10 ppmw. Total acidity ≤ 7 ppmv. Per-lot certificate of analysis naming the lot number, with results for each parameter. Net fill weight stated per cylinder on the packing list. CGA 590 valve (or DIN, as applicable), adapter responsibility assigned to seller. UN 1079, class 2.2, dangerous goods declaration and SDS section 14 to agree.

Two lines in that block do most of the work: the per-lot CoA and the net fill weight. The first tells you whether the supplier tests, the second tells you whether they fill. Everything else is standard practice that a competent supplier already does without being asked.


Questions we get on this one

Can I mix grades in the same equipment?

Top up with the grade the equipment was filled with, or higher. Going from 99.9% to 99.95% on a top-up is not a problem. Going the other way, from a new-gas fill to a reuse-grade top-up, can put you outside the OEM’s specification without anything visibly going wrong.

Does 99.95% give better dielectric strength than 99.9%?

Not measurably, at the pressures used in switchgear. The difference is in the impurity profile, mainly moisture and air, and its effect shows up over time and in test repeatability, not in a step change in breakdown voltage.

Is 99.995% worth it for a lab doing IEC 62271 tests?

Only if the test protocol requires it or if you have evidence that cylinder impurities are affecting repeatability. Most type-test labs run 99.9% or 99.95% and control moisture at the gas cart instead.

Why is my 99.9% certificate showing 99.87%?

Because purity is measured, not guaranteed to the second decimal, and the specification is a floor on each parameter rather than an exact assay. What matters is whether each impurity is inside its limit. Ask for the parameter table, not just the assay figure.

Do you supply reconditioned gas?

We fill and export new technical-grade SF₆ from 99.5% to 99.995%. For reuse loops the specification work is the same, and the lot data is what we would want to see if we were the buyer. If you are setting up a recovery programme, the questions to ask any supplier are the ones in this article, in this order.

Which grade for a first fill on equipment that will sit in storage for two years?

99.95%, with the moisture result on the lot certificate stated, and the compartment left under dry nitrogen or under SF₆ at a holding pressure rather than open. Storage is where moisture gets in.


Sources

  • IEC 60376:2018, Specification of technical grade sulphur hexafluoride (SF₆) and complementary gases for use in electrical equipment — webstore.iec.ch/en/publication/33028
  • IEC 60480:2019, Guidelines for the checking and treatment of sulphur hexafluoride (SF₆) taken from electrical equipment and specification for its re-use — webstore.iec.ch/en/publication/61024
  • Regulation (EU) 2024/573 on fluorinated greenhouse gases (quota and phase-out framework for SF₆ equipment) — eur-lex.europa.eu/eli/reg/2024/573/oj
  • ASTM D2472, Standard Specification for Sulfur Hexafluoride

About the author

Newradar Gas fills SF₆ from 99.5% to 99.995% in 8L and 40L cylinders and ships it as UN 1079 dangerous goods with per-lot analysis. We have been doing this since 2007 from Wuhan, Hubei. If you are arguing internally about which grade to specify, send us the equipment manual line and the tender clause and we will tell you what the gas actually has to do. nrdgas.com/products


Post time: Nov-20-2026