Gas Switching and Isobaric Counterdiffusion

The switch you make on every trimix dive

You are hanging at 18 m after a dive to 60 m on Tx 18/45. The plan says EAN50 from here. You unclip the regulator, confirm the label, confirm your depth, and switch.

Nothing about the water changed. The pressure around you is the same as it was a breath ago. But the gas going into your lungs just changed completely: helium went from 45 percent to zero, and nitrogen went from 37 percent to 50.

Almost every trimix diver makes this exact switch, and almost every one of them has been told, at some point, that it can hurt them. The injury has a name, inner-ear decompression sickness, and a mechanism with a longer name, isobaric counterdiffusion, or ICD. There are at least six rules in circulation for avoiding it, and they disagree with each other about the very switch you just made.

So here it is from the bottom up, one layer at a time, ending with what I do at 18 m and why.

If you have not read how decompression models work, start there. Everything below assumes you know what a tissue compartment and a half-time are.

Layer 1: gas goes into you in proportion to its partial pressure

Partial pressure is fraction times ambient pressure. At 60 m the ambient pressure is 7 bar, and 18/45 splits like this.

Every tissue slowly fills with helium and nitrogen until its dissolved gas tension matches those inspired numbers. Fast tissues match in minutes, slow ones in hours. When you ascend, ambient pressure falls faster than tissues can unload, the tissue tension ends up above ambient, and that gap is supersaturation. Too much of it and bubbles form. All of decompression is about keeping that gap small enough.

Layer 2: a switch changes the gas without changing the depth

Now the switch itself. At 18 m the ambient pressure is 2.8 bar. Before and after look like this.

Helium in your tissues now sees zero helium in your blood, so it leaves. Nitrogen in your blood is now higher than in your tissues, so it enters. Two gases crossing in opposite directions at the same pressure. Iso-baric, same pressure. Counter-diffusion, opposite directions.

For most tissues this is fine or even helpful. Helium leaves quickly, nitrogen arrives slowly, and the tissue's total gas tension dips for a while. That dip is why people once believed a nitrogen switch accelerated decompression.

Layer 3: why crossing gases can make a tissue fuller

Two properties of the gases can flip that dip into a rise.

Helium diffuses about 2.65 times faster than nitrogen through tissue. And nitrogen dissolves about 4.5 times more readily in fat and in blood than helium does.

Whether a tissue's total goes up or down after a switch depends on its geometry and on which of those two effects wins there. Each rule picks one of the two to worry about, which is why the rules exist and why they disagree.

Layer 4: the inner ear is where it goes wrong

The inner ear has a small blood-supplied part, the membranous labyrinth, and two fluid-filled chambers, the perilymph and endolymph, with no blood supply of their own. The fluids exchange gas only by diffusion through the blood-supplied part. In the model Doolette and Mitchell built in 2003, that blood-supplied part behaves like a compartment with a half-time of about 8.8 minutes [1].

After an hour on helium the fluids hold a reservoir of it. On the switch, two things happen together. Helium in the fluids diffuses back into the blood-supplied part faster than nitrogen diffuses out, because helium is the faster gas. And blood arriving from the lungs now carries a lot of nitrogen and takes away little helium, because nitrogen is the more soluble gas. Both routes add gas to the same small tissue. Its total tension rises for a while, even though depth never changed.

The symptom is vertigo, often with nausea and vomiting, sometimes tinnitus or one-sided hearing loss, arriving a few minutes after the switch at a constant depth. The timing is the tell. Ear barotrauma happens while the pressure is changing. Ordinary DCS shows up after surfacing. Inner-ear ICD arrives at the stop, shortly after the regulator swap.

Layer 5: the finding that changed the picture

One finding decides which of the rules below matter, so it comes before them.

When Doolette and Mitchell ran their model on a deep dive, they found the inner ear was already heavily supersaturated during the early ascent, before any switch. On a 110 m, 25 minute profile the membranous labyrinth peaked at 1.7 bar above ambient with no gas change anywhere in the plan [2]. The switch, when they added it, produced only a transient slowing of washout on top of that.

By 2015 their position was explicit: the counterdiffusion effects of typical technical switches "would rarely, if ever, be sufficient to produce inner-ear bubble formation in their own right", and inner-ear DCS during deep decompression "may be explained primarily by inadequate decompression" [2]. If a switch happens while the ear is still near its peak, it can add to that peak. Otherwise it usually does nothing.

The rest of the evidence lines up behind that.

• The one human trial, at the US Navy Experimental Diving Unit, switched divers from heliox to air at 30 m after an hour at 46 to 67 m and found low inner-ear DCS risk [3]. That is a nitrogen step of about 3 bar, far larger than any technical switch.
• GUE reports no signal of elevated DCI across thousands of Tx 18/45 to EAN50 switches in its community [4].
• The nine inner-ear cases collected from GUE-trained trimix divers mostly presented after surfacing, most had a PFO, and the one diver who was hit underwater was judged to have omitted decompression earlier in the ascent, not to have been hurt by the switch [5].
• Rebreather Forum 4 in 2024 put it in one word: the contribution of gas switches to inner-ear DCS during decompression is "uncertain" [6].

Layer 6: the rules, one at a time

Each rule below is scored on the same switch, Tx 18/45 to EAN50 at 18 m, so you can watch them disagree.

Rule A: the rule of fifths

Steve Burton, a chartered engineer, published this in 2004 [7]. Nitrogen holds about 4.5 times more dissolved gas per bar than helium, so if you drop five points of helium and add one point of nitrogen the total dissolved load stays about level. So the nitrogen percentage may rise by no more than one fifth of the helium percentage you drop.

Our switch drops helium 45 points, so nitrogen may rise 9. It rises 13, so it fails.

The rule reads only the cylinder labels, so it gives the same verdict at 6 m and at 60 m. Its natural fix is to leave helium in the deco gas, and that is exactly what GUE's 35/25 and 50/25 do. Its evidence is Burton's own count of "hundreds of deep trimix dives". No agency standard uses it. Subsurface offers it as an optional warning.

On the physics it is half right. The solubility effect is real and is one of the two routes in the inner-ear model. But bubbles form on tension, not on the quantity of gas, and the one fifth is Burton's own round number rather than a fit to any data. The two largest bodies of real-world evidence, the NEDU trial and the GUE community, both break it wholesale without a signal.

Rule B: the 0.5 bar rule

BSAC skipped the solubility argument and asked how big the nitrogen step is in bar, which is what the tissue feels [8]. Keep it to about half a bar at the switch depth.

At 18 m our switch passes comfortably. The same cylinders at 30 m are borderline. Scaling with depth is more honest physics than rule A manages. It is also the number MultiDeco and V-Planner use for their default alarm [9]. Mike Rowley, who ran BSAC's technical group, calls the figure advisory: a 0.7 bar step "isn't going to bring the Sword of Damocles down on you" [4]. No paper produced the 0.5.

Rule C: depth limits

NAUI Tec, following Bruce Wienke, sets hard cut-offs: no helium-to-nitrogen switch deeper than 30 m, or in the stricter form deeper than 21 m, and in the "zero order" version no switch at all, decompress on bottom gas and go to oxygen at 6 m [4]. Our switch at 18 m passes. NAUI itself says the rule "has certainly not been formally tested". It agrees with Doolette's "the shallower the better", which is the only reason to respect it.

Rule D: keep helium in the deco gas

A design choice rather than a check. GUE and UTD standard gases put 35/25 in at 36 m and 50/25 or EAN50 at 21 m, then oxygen at 6 m. Every switch then drops both inert gases at once, and there is nothing to compute. TDI teaches the same idea as "hyperoxic trimix". The cost is helium and a third cylinder. The Guenzani cases show it does not make inner-ear DCS impossible [5].

Rule E: the timing rule

This is the only rule that came from a model of the organ that gets hurt. Doolette's summary in 2013 has four parts [3].

1. Decompress the inner ear before you switch. The algorithm "should adequately decompress the inner ear, which has a half-time of about 8.8 minutes". In practice: do not trim the first stops on back gas, do not run a high GF High on trimix, and give the ear a few minutes at the switch depth before you change regulators.
2. Switch shallow. The nitrogen step is smaller shallow, and by the time you are shallow the ear has had more minutes to unload. "No undue risk found with switches at 30 msw."
3. Highest safe oxygen after the switch. Every bar of oxygen in the inspired mix is a bar of inert gas that is not there, so the gradient out of the ear is steeper. And oxygen breathing is thought to reduce venous bubbles crossing the lungs into the arteries, the other route to the same injury.
4. A profile that does not throw a lot of venous bubbles, since a right-to-left shunt is the other big factor.

Our switch passes, provided the deco before it was adequate. No numbers are given for "adequate" or "a few minutes". Layer 9 fills that gap.

Rule F: tissue-based computer warnings

The OSTC computers and, as a second check, Subsurface watch the leading tissue and warn when helium and nitrogen are moving in opposite directions while the tissue is supersaturated and rising [9]. It sounds rigorous. Robert Helling, a Subsurface developer, found the criterion fires during an ordinary 60 m ascent with no gas switch at all [10]. Treat it as something a student may see on their screen, not as a rule.

Layer 7: the scorecard

They disagree because they measure different things: A reads labels, B reads the step in bar, and E reads the state of the ear when the step arrives. The evidence says E is the one that matters, B is the arithmetic check that scales properly, and A is the story of why helium-bearing deco gases exist.

Layer 8: phased switching, the technique divers like

Before any of these rules were written, Sheck Exley solved the problem by feel. On his 1989 dive to 268 m at Nacimiento del Río Mante he had to switch at 73 m from a mix of 50 percent helium and 40 percent nitrogen to air, a nitrogen step of about 3.2 bar. In his own words: "To make the switch as gradual as possible, I took only a single breath of air from the tank we had earlier left at 240 feet, then switched back to the helium mix for two breaths. Then came two air breaths, then back to the deep mix for one inhalation" [11]. One, two, two, one, then air, over about half a minute.

Burton credits this as the original ICD-avoidance technique. I like it and I teach it, with two caveats. No agency has adopted it and no paper has tested it. And the inner-ear compartments have half-times of minutes, so a thirty-second ramp is short relative to the tissue and its effect on the modelled peak is small. Its real merit is probably the narcosis and oxygen transition, and the pause it forces at the switch depth.

Three modern practices do the same job with less task loading. An intermediate gas, so no single step is large, which is rule D. The rebreather loop, where helium falls continuously as oxygen is added on the ascent; Simon Mitchell's 2016 review saw "little compelling reason, and perhaps some small risk" in doing diluent switches on a rebreather at all [12]. And a few minutes on back gas at the switch stop, so the ear is past its peak when the step arrives, which is rule E with a number on it, below.

Layer 9: the checklist, with every number tagged

Doolette's rule is the right one and it gives no numbers. Here they are, each tagged by where it comes from. Trial means human data. Model means the 2003 inner-ear model. Convention means an agency or planner default with no trial. Derived means arithmetic from the model that no paper states.

The oxygen ceilings are mine: 1.4 bar for any mix up to 80 percent oxygen, 1.6 bar for pure oxygen only. I inferred them from the NOAA exposure data; NOAA publishes time limits, not a rule about which mix gets which ceiling. They put EAN50 at 18 m and 35/25 at 30 m.

Layer 10: at the stop

If vertigo arrives a few minutes after a switch at a constant depth, it is inner-ear DCS until proven otherwise. The diver cannot hold depth or read a gauge. The team holds them. Do not ascend to get them out. The remaining obligation is real and surfacing does nothing for the vertigo. Keep the highest safe oxygen going, complete the decompression with hands-on support, and get to a chamber. Doolette and Mitchell call this the technical diver's "Catch 22", and it is the strongest argument for switching shallow and late: if it happens at 18 m the remaining deco is short and warm.

What is settled, and what is not

Settled: the inner ear is the target tissue; it is supersaturated by the ascent before any switch; a helium-to-nitrogen switch at 30 m or shallower carries low measured risk; a right-to-left shunt is a major factor; switching does not speed up decompression on a bounce dive.

Contested: how much a switch contributes at all, which Rebreather Forum 4 leaves as "uncertain"; whether any of the numeric rules predict anything; whether the oxygen window explains any of the benefit of a rich deco gas, or whether the inert-gas gradient explains all of it.

Not consensus, but convergent: the same two researchers, Doolette and Mitchell, have said the same thing in five venues over twelve years, and no peer-reviewed work contradicts them. The field rules are agency conventions with experience behind them and no trial.

The honest summary

The profile is what hurts the ear, and the switch only trims the risk. Decompress the ear first, switch shallow and late, breathe the richest gas the depth allows, and check that the nitrogen step is under half a bar. Keep the rule of fifths as the reason helium-bearing deco gases exist. Keep Exley's one-two-two-one because it is good history and because I happen to like it.

References

1. Doolette DJ, Mitchell SJ. Biophysical basis for inner ear decompression sickness. J Appl Physiol. 2003;94(6):2145–2150.
2. Mitchell SJ, Doolette DJ. Pathophysiology of inner ear decompression sickness: current concepts. Diving Hyperb Med. 2015;45(2):105–110.
3. Doolette DJ, Mitchell SJ. Recreational technical diving part 2: decompression from deep technical dives. Diving Hyperb Med. 2013;43(2):96–104. Citing Doolette DJ, Gerth WA. Safe inner ear gas tensions for switch from helium to air breathing during decompression. NEDU TR 12-04, 2013.
4. Kas S. How two tech agencies address isobaric counterdiffusion. InDEPTH, May 2020.
5. Guenzani S, Mereu D, Messersmith M, Olivari D, Arena M, Spanò A. Inner-ear decompression sickness in nine trimix recreational divers. Diving Hyperb Med. 2016;46(2):111–116.
6. Mitchell SJ, Pollock NW, eds. Rebreather Forum 4 Proceedings. 2024. Decompression theory and practice, inner-ear DCS.
7. Burton S. Isobaric counter diffusion. scubaengineer.com, December 2004.
8. British Sub-Aqua Club. BSAC recommended diving gas mixtures for open circuit diving. bsac.com.
9. HHS Software, MultiDeco FAQ; Subsurface source, core/gas.cpp and core/deco.cpp; Heinrichs Weikamp forum, IBCD calculations, 2021.
10. Helling R. Isobaric counter diffusion criteria and More confusion from isobaric counter diffusion. The Theoretical Diver, 2018.
11. Exley S. Caverns Measureless to Man. Cave Books, 1994. p. 259. Quoted in Burton S, Sheck Exley Mexico dive, scubaengineer.com.
12. Mitchell SJ. Decompression science: critical gas exchange. In: Rebreathers and Scientific Diving Proceedings, 2016. pp. 163–174.
13. Doolette DJ. Gradient factors in a post-deep stops world. InDEPTH, 2019.

Gas switching under a ceiling is the skill every technical course I teach is built around. Enquire about training →