XERF · RF Technique · A Physician's Perspective

XERF: Deep mode all the way?

First, what the modes actually are

The Shallow, Middle and Deep labels suggest a simple ladder: same treatment, three depths. The published pulse structures are more interesting than that.

In the 2025 study by Ko and colleagues, Shallow is a 6.78 MHz train of twelve 100 ms sub-pulses over roughly 1,930 ms. Middle is 6.78 MHz again, but six 200 ms sub-pulses over roughly 1,950 ms. Deep is five 150 ms sub-pulses at 6.78 MHz, followed by seven 120 ms sub-pulses at 2 MHz, over about 2,400 ms.

Shallow and Middle use the same frequency and the same total on-time, differing only in how that time is chopped up. Deep is the only setting where 2 MHz appears at all. That is the first reason I stopped treating the three as points on a scale.

Where each frequency puts its heat

Penetration depth in RF is inversely proportional to the square root of the frequency, so a lower frequency reaches further. That much is textbook. What the modelling added was where the energy concentrates once it gets there.

At 6.78 MHz, heating ran preferentially along the fibrous septa of the subcutaneous fat, and septa running perpendicular to the surface heated more than those lying horizontally. At 2 MHz, the reaction was broader and deeper, with bulk heating through the adipose tissue itself. The authors put this down to the difference in dielectric properties and loss tangent between fat and fibrous tissue.

One detail from that modelling reframed the mode names for me. Shallow and Middle both put their highest tissue temperature at the fibrous septa. Deep put its highest temperature in the subcutaneous fat, with heat extending into the deepest layers of the model. None of the three is superficial in the way the word implies. What changes is which structure takes the peak.

Deep is the only setting that covers both patterns at once, and that is what made me want to stay in it rather than move between three partial ones.

What I actually do

The forehead and cheek comparison above is the easy half of it. Jawline and submental take more attention, because thickness varies enough between patients that treating them as a fixed region does not really work. I adapt to what is in front of me rather than to where I am on the face.

What I like about working this way is that the delivery architecture stays constant. Velocity is a continuous variable, so it gives finer control than three discrete buttons, and I am not re-learning how the machine behaves every time I change region.

Speed changes dose, not depth

One distinction does most of the work in that paragraph, and it is worth stating plainly.

Sliding speed does not change how deep the energy goes. It changes how much accumulates where it lands.

If I move slowly over the cheek, I am not treating deeper. I am delivering a larger cumulative dose to the same distribution. Frequency and the electrical properties of the tissue set where the energy goes; velocity, pass count and overlap set how much builds up there. Conflating the two is how people end up believing they can steer depth with their hands.

This is also not a licence to use identical settings everywhere. Tissue thickness, underlying structures, contact quality, treatment boundaries and cumulative exposure all still apply, and none of them are solved by picking one mode.

The published protocol, and where I diverge

The paper closest to what I am doing came out in August 2026. Erlich, Dahan and Wolf retrospectively reviewed 16 women, median age 54.5 with a range of 25 to 73, Fitzpatrick II to V, laxity ranging from mild to severe. Each had one XERF session and nothing else done to the face for six months before or three months after, which is unusually clean for aesthetic work.

The entire treatment was Deep mode, 6.78 MHz followed by 2 MHz, including the upper face.

They did not drop out of Deep mode for thin tissue either. Forehead, temple and infrabrow were all treated in Deep dual-frequency, and what changed was tip size rather than mode. On that point we agree, and it was useful to see it done in a published series.

Where we part company is the hand. Their gliding was deliberately non-overlapping — one full-face gliding pass, then one full-face stamping pass, alternating, with the gliding run in a single direction along superolateral vectors and no overlap between adjacent passes. Read that as gliding for broad distribution and stamping for concentrated delivery, alternating between the two. I slide throughout and vary speed instead.

The pain numbers differ sharply. Theirs reported a mean VAS of 4.31 out of 10, distributed as 38% mild, 50% moderate and 12% severe, with two patients scoring 7 and 8. Only four of the sixteen used topical anaesthetic. Mine usually report 1 to 2, also without numbing.

I do not read that as Deep mode being comfortable. I read it as where the energy lands.

A stationary pulse deposits a whole sequence into essentially one footprint. Continuous sliding spreads successive exposures across tissue, so peak accumulation at any single point can be a good deal lower even while a broad thermal field builds over repeated passes. Their non-overlapping passes would reduce that spreading effect further.

There are too many uncoupled variables between their protocol and mine to call that more than a hypothesis. Energy level, speed, overlap, cooling, anatomy and endpoint all differ. It is the part of this I would most like to see someone measure properly.

What their 3D data showed

Running 2 MHz across the whole face raises an obvious objection. If you are putting more energy deeper everywhere, including over thin tissue, are you simply heating fat that you would rather leave alone?

The best answer I have seen to that comes from the same study, which measured with 3D stereophotogrammetry rather than relying on photographs. Lateral facial compartments gained volume. Medial compartments lost it. Immediately after treatment the combined lateral gain was about 1.37 mL against a medial loss of about 0.49 mL, and across the follow-up there was no clinical or imaging evidence of fat atrophy.

A net loss would show up as both compartments falling. What they recorded was the envelope moving, which is what you would expect from tissue contracting and repositioning rather than fat being destroyed.

It also matches what I see in the chair. The faces do not look smaller. Medial and lower-face heaviness comes down while the cheek structure holds, which gives a cleaner jawline without the hollowing you would associate with indiscriminate volume loss. Standardised 2D photographs cannot establish redistribution, so I would call that a consistent visual impression rather than a measurement.

The same study recorded progressive brow and upper-eyelid elevation in regions treated in Deep mode, and computerised analysis showed improvement in wrinkle severity, evenness, pore size and oiliness. Pigmentation did not change significantly at any timepoint, and patients often assume it will.

Most of it was there at one month

FACE-Q facial satisfaction improved by 15.3 points at one month and stayed significantly improved at three. Aging appraisal improved 16.3 points. Patients judged themselves 2.81 years younger at one month and 3.28 years at three.

The part that changes how I talk to patients: there was no statistically significant difference between the one-month and three-month FACE-Q results. The measurable effect was largely established by one month and then held.

RF is usually described as something that pays off at around three months, on the basis of collagen remodelling timelines. In this cohort the subjective gain arrived earlier and then plateaued. It is one small retrospective series, so I would not rebuild a whole consent conversation around it, but it is a useful corrective to telling people the good part is still coming.

The wider clinical picture

A prospective multicentre study by Weiss and colleagues, published in Cureus in 2026, gives the larger dataset: 39 subjects across four sites, two sessions four weeks apart, depth setting chosen by zone, sliding on broad areas and stamping on contoured ones. Mean total energy 56.4 kJ, mean 568.8 shots, mean pain 4.2 out of 10 without anaesthesia. Clinicians rated 84.6% at GAIS 3 or higher at 30 days and 92.3% at 90 days; 79.5% and 84.6% of patients rated satisfaction 4 or higher at the same points. Adverse events were mild and transient.

That protocol varied depth by region and mixed sliding with stamping, so it is the conventional approach rather than mine. Its pain figure sits close to the Erlich number, which is at least consistent with technique being the variable that separates both of them from what I see.

On the fat question

The mechanistic work points the same way as the 3D measurements. In the porcine study, specimens treated with the dual-frequency Deep pulse showed preserved adipocyte morphology at 7, 15 and 30 days with no structural disruption, and TUNEL assay found no apoptosis in the subcutaneous fat or fibrous components. Collagen and elastic fibre remodelling increased across the dermis and septa over the same period, most pronounced in the Deep-treated specimens at 30 days.

Sixteen patients over three months, plus an animal model at one set of energy settings, is not enough to establish that an all-Deep protocol cannot cause fat loss. It is particularly not enough when my own technique produces a different cumulative exposure from the one that was studied. I have written separately about when RF does start affecting fat, which is a question of thermal dose rather than frequency alone.

What is still open

The published evidence supports several things: Deep dual-frequency across the whole face including thin upper-face tissue, a single session producing measurable change, Deep combined with gliding, energy individualised to tolerance, and no observed fat atrophy in these cohorts.

What it does not support is my specific version of it. Nobody has compared a Deep-only, sliding-dominant protocol against conventional multi-depth treatment on efficacy, comfort or safety. Their forehead was stamped with smaller tips; their mid-face was half stamped. That is exactly where I diverge, and that divergence is untested.

After years with monopolar RF I have become more interested in how energy is delivered than in how much of it there is. Keeping one delivery architecture and varying the hand is, for me, a more controllable way to work than changing the architecture every time the anatomy changes. Whether it produces better outcomes than the conventional approach is not something I can answer yet.

Sources

Accurate as of September 2026. Published evidence on this device is still accumulating; this page is dated rather than left to age quietly.