Surface Wave UT for Rail: Physics, Uses and Limits

Surface wave UT is an ultrasonic method that sends Rayleigh waves along a steel surface at about 3,000 m/s, roughly 0.92 times the shear velocity, so the energy stays within about one wavelength of the surface: 1.5 mm at 2 MHz and 0.6 mm at 5 MHz. It is sensitive to surface-breaking cracks, but it is not part of standard revenue-service rail testing.

That last point matters because the method is often oversold. Rail test cars search for internal defects with 0°, roughly 37–45° and 70° beams. Gauge-corner rolling contact fatigue (RCF) is found and managed mainly with eddy current and ACFM, visual and magnetic particle checks, and rail grinding. Surface-wave inspection of the rail head is still a research and laboratory technique, with phased array and EMAT systems under development beside it.

This guide gives the physics with worked numbers, shows what actually handles gauge-corner cracking in practice, corrects the "near-surface blind zone" claim, and sets out where surface waves earn their place.

What is surface wave UT, and how do Rayleigh waves behave in rail steel?

A Rayleigh wave is a sound wave that travels along the surface of a solid, with elliptical particle motion that dies away quickly with depth. The US Federal Highway Administration's NDT glossary defines it this way:

"The particle motion is elliptical in a plane perpendicular to the surface, decreasing rapidly with depth below the surface. The effective depth of penetration is considered to be about one wavelength." — FHWA NDE Web Manual, "Rayleigh waves"

Evident's flaw detection tutorial says the same: surface waves travel "to a depth of one wavelength", have a velocity "similar to shear waves", and "can be used to detect surface-breaking cracks" (Evident, Wave Propagation).

In practice a surface wave is usually made the same way as a shear wave: an angle-beam wedge is set past the second critical angle, so neither longitudinal nor shear waves enter the steel and the energy is trapped at the surface. Couplant is still needed between the wedge and the rail. EMATs and lasers can generate surface waves without couplant, which is why most of the rail research uses them.

Velocities in steel

The Rayleigh velocity follows from the longitudinal and shear velocities and Poisson's ratio. Evident lists 5,890 m/s longitudinal for 1020 steel and about 3,250 m/s shear.

Wave mode	Velocity in carbon steel	How it is obtained
Longitudinal (compression)	5,890 m/s (0.232 in/µs)	Evident velocity table, 1020 steel
Shear (transverse)	3,250 m/s (0.128 in/µs); some guides use 3,230	Evident
Rayleigh (surface)	about 3,000 m/s (0.118 in/µs)	Calculated: 0.924 × 3,250 = 3,004 m/s

How the Rayleigh figure was checked: from the two Evident velocities, Poisson's ratio is 0.281. Solving the exact Rayleigh equation gives v_R/v_S = 0.924, so v_R = 3,004 m/s. The common Viktorov approximation, v_R ≈ v_S × (0.862 + 1.14ν)/(1 + ν), gives 3,000 m/s. With a 3,230 m/s shear velocity, the result is 2,988 m/s. The often-quoted "0.92 × shear, about 2,980 m/s" is therefore close but slightly low. Real rail steel can differ by 1% or more with composition, hardness and microstructure, so calibrate on the actual material.

Key takeaway: in steel, a Rayleigh wave runs at about 3,000 m/s, about 8% slower than a shear wave, and it carries its energy in the top wavelength of the surface.

How deep does a Rayleigh wave reach at 2 and 5 MHz?

Wavelength is velocity divided by frequency, λ = v/f, and the Rayleigh wave's useful depth is about one wavelength. So frequency sets depth: higher frequency gives better sensitivity to tiny cracks but a shallower look.

Worked example: Rayleigh wavelength and depth in rail steel

Inputs: v_R = 3,000 m/s (3.00 mm/µs), as calculated above.

Frequency	Rayleigh wavelength λ = v/f	Effective depth (≈ 1 λ)	For comparison: shear λ at 3,250 m/s
200 kHz (0.2 MHz)	15.0 mm (0.59 in.)	about 15 mm	16.3 mm
0.5 MHz	6.0 mm (0.24 in.)	about 6 mm	6.5 mm
1 MHz	3.0 mm (0.12 in.)	about 3 mm	3.25 mm
2 MHz	1.50 mm (0.059 in.)	about 1.5 mm	1.63 mm
5 MHz	0.60 mm (0.024 in.)	about 0.6 mm	0.65 mm

The arithmetic at the two common frequencies:

2 MHz: λ = 3.00 mm/µs ÷ 2 per µs = 1.50 mm. The wave samples roughly the top 1.5 mm of the rail head.
5 MHz: λ = 3.00 ÷ 5 = 0.60 mm. The wave samples roughly the top 0.6 mm.

Every depth in the table is far shallower than a rail head, so a Rayleigh wave never reaches the interior where transverse and detail fractures grow to critical size.

Worked example: matching frequency to RCF crack depth

TTCI sectioned five rails removed from service for poor UT results. It counted 880 RCF cracks and spalls and found that 90% reached no more than 0.040 in. (1.02 mm) below the surface (TTCI TD-18-027, 2018). To sample a crack to its tip, the wavelength should be at least the crack depth:

Maximum frequency for a 1.02 mm crack: f = v/λ = 3.00 ÷ 1.02 = 2.9 MHz. A 2 MHz wave (λ 1.50 mm) covers that depth. A 5 MHz wave (λ 0.60 mm) sees the crack mouth but not its tip.
RCF cracks run at a shallow angle. TTCI's Norfolk Southern sections showed crack angles of 18–28° to the surface (TTCI TD-14-007, 2014). A crack 5 mm long at 25° reaches 5 × sin 25° = 2.11 mm deep. Reaching that tip needs λ ≥ 2.11 mm, or f ≤ 3.00 ÷ 2.11 = 1.4 MHz.
A heavily cracked curve: the same digest reports Draisine eddy current depths of 0.14 in. (3.56 mm) before grinding on a 6° curve. Sampling 3.56 mm needs f ≤ 3.00 ÷ 3.56 = 0.84 MHz, which costs sensitivity to the smallest cracks.

This is why surface-wave research on rail often uses low frequencies. Hesse and Cawley modeled surface wave modes up to 350 kHz to get "a high penetration depth" (Hesse and Cawley, JASA 2006).

Key takeaway: choose the surface-wave frequency from the crack depth you need to see; 2 MHz suits the 1 mm RCF band, and deeper RCF needs 1 MHz or less.

Is surface wave UT part of standard rail flaw testing?

No. Revenue-service rail flaw testing in North America uses bulk-wave UT, not surface waves. Test cars and hi-rail systems use fluid-filled wheel probes or sleds with a typical beam set of 0° (horizontal defects and loss of back-wall), roughly 37–45° (bolt-hole cracks and web defects) and 70° shear (transverse and detail fractures in the head). Some systems add side-looking probes. Exact angles vary by equipment.

The federal rule is method-neutral. 49 CFR 213.237 requires a continuous search for internal rail defects at set intervals on Class 3–5 track, and the equipment must detect defects between joint bars. Part 213 never names a UT technique, frequency or angle. A surface wave can't search the rail interior, so it can't provide the valid search that 213.237 requires. At most, it could supplement a valid search. For the full rule, see 49 CFR 213.237 rail flaw detection requirements.

Where RCF does affect the federal test is coupling. A valid search must not be "compromised due to environmental contamination, rail conditions, or equipment malfunction" (49 CFR 213.238(b)(4)). FRA's 2011 RCF review describes the result: surface fatigue "prevents the successful transmission of ultrasonic waves to the base of the rail and back", which causes a no-test and usually a hand test (FRA DOT/FRA/ORD-11/24).

Key takeaway: surface wave UT isn't a 213.237 inspection method; the rail flaw test is still a 0°/37–45°/70° bulk-wave search for internal defects.

What actually finds and manages gauge-corner RCF?

Gauge-corner cracking is found mainly with electromagnetic methods (eddy current and ACFM) and visual or magnetic particle checks, and it is controlled by grinding. FRA's comprehensive RCF review sums up the practice:

"current best practice is to combine nondestructive evaluation with preventative rail grinding to optimize the tradeoff between maintenance cost and structural reliability" — Magel, Rolling Contact Fatigue: A Comprehensive Review, FRA DOT/FRA/ORD-11/24, November 2011

The same report puts the safety stake at about 100 FRA-reportable derailments a year in North America attributed to RCF. For the defect itself, see rolling contact fatigue in rail and head checks in rail.

Method	What it does for gauge-corner RCF	Status and evidence
Visual, dye penetrant, magnetic particle	Shows surface cracks; MPI and DPI make fine head checks visible	TTCI: labor-intensive, weather-dependent, and "not capable of measuring any key characteristics below the surface" (TD-14-008)
Eddy current (walking-stick, trolley or train-mounted)	Detects surface-breaking cracks; estimates depth from crack length and an assumed angle	TTCI/NS evaluation of a commercial trolley: depth error 5–33% against sections, with an assumed 25° angle (TD-14-007). Needs a steady 1–2 mm standoff (FRA ORD-11/24)
ACFM	Electromagnetic crack detection and sizing; FRA ORD-11/24 notes that with ACFM "the deeper the crack, the larger the response"	Commercial walking-stick systems; high-speed trials published (Papaelias et al., NDT&E Int. 2009)
Rail grinding	Removes the cracked layer before cracks grow; restores UT coupling	The main control. TTCI judged three to five production grinder passes would likely have restored UT confidence on its five test rails (TD-18-027)
Conventional 70° and 37° UT	Finds the deeper defects that RCF leads to, mainly detail fractures	Can detect larger (>4 mm) surface-originated defects, but surface cracks can shadow deeper defects (FRA ORD-11/24, citing Papaelias et al.)
Phased array UT (research)	Steers and focuses beams through and under surface damage	A TTCI head-contact prototype found transverse defects under shells, but can't inspect the base flanges (TD-16-050)
EMAT and surface-wave UT (research)	Non-contact Rayleigh waves across the railhead to detect and estimate RCF depth	Laboratory and proof-of-concept results only (see the next sections)

Two points in the table deserve emphasis. First, no method sizes RCF well. TTCI wrote in 2014 that crack angle and depth are "still a primary limitation" for eddy current and ultrasonic systems. Second, masking cuts both ways. Shells and head checks hide detail fractures from the 70° probe, which is the main reason grinding sits beside inspection. NTSB found that a lower gauge-corner fatigue crack at Lynchburg (2014) was "not easily scanned by ultrasonic equipment" (NTSB RAB-16/01). For how shells and spalls interfere with testing, see rail shelling and spalling.

Key takeaway: electromagnetic methods and grinding manage gauge-corner RCF today; surface-wave and phased array UT are research tools that may join them.

A single-element contact probe does have a dead zone just under the surface, but that isn't the main limit on rail head inspection, because rail testing doesn't depend on a single-element 0° probe for the head. The real near-surface problems are noise, RCF scattering and coupling.

Three design features cover the upper head:

Twin-crystal (dual) probes separate the transmitter from the receiver, so the receiver isn't blinded by the initial pulse. Evident lists "Improves near surface resolution" as the first advantage of duals, and rates two of its 7.5 MHz duals for "optimal near-surface resolution in the range of 0.030 – 2 in. (0.76 – 50.8 mm) in steel" (Evident dual element transducers).
Angle probes reach the upper head along an inclined path, so a reflector a few millimeters deep is met at a sound path well beyond the probe's ringdown. A 70° beam travels about 2.9 times the depth to reach a point (1 ÷ cos 70° = 2.92).
Wheel and sled geometry puts a fluid or delay path between the element and the rail. The initial pulse is spent in the wheel, in the same way that a delay-line probe is used, in Evident's words, "to separate the excitation pulse recovery from back wall echoes" (Evident, thickness gauge transducers).

The genuine limit is different. FRA's review says ultrasonic systems are often set to ignore "the first few millimeters of the rail surface" because the near-surface signal is too noisy and complex, not because the beam can't reach it (ORD-11/24). Shallow RCF also scatters sound and can stop coupling altogether. Surface waves don't fix either problem, since they are scattered by the same dense cracking. For the bulk-wave physics, see shear wave ultrasonic testing and the science behind ultrasonic testing.

Key takeaway: duals, angle beams and wheel or sled delay paths cover the head; the near-surface problem on rail is noise and RCF, not an uncovered dead zone.

Where is surface wave UT really used?

Surface wave UT is used where the defect is surface-breaking and the surface is reachable: weld toes and other fatigue-crack sites in structural steel, the rail foot in research systems, and laboratory or prototype railhead systems.

Application	What the surface wave does	Evidence and maturity
Weld toes and fatigue-crack sites in structural steel	Screens a run of surface for surface-breaking fatigue cracks, which often start at weld toes	Recognized technique: Evident says surface waves "can be used to detect surface-breaking cracks". A WVU dissertation on steel members found the Rayleigh wave "the most effective in detecting fatigue cracks in uncoated, painted, and rusted members" (Franklin, 1998, lab beams)
Rail foot (base)	Travels along the base to reach areas head-mounted probes can't	Research: EMAT pseudo-Rayleigh waves at 300–400 kHz suited a rail-base specimen with artificial defects (Wang et al., IEEE FENDT 2014); laser-generated surface waves found transverse defects at the outer edge of the base in the EU U-RAIL project, 2004–2007 (CORDIS 507622)
Railhead RCF, EMAT	Non-contact Rayleigh waves across the head to detect and characterize surface-breaking defects	Laboratory studies: Edwards, Dixon and Jian, NDT&E Int. 2006; Fan, Dixon, Edwards and Jian, NDT&E Int. 2007
Railhead, low-frequency guided surface modes	Modes below 350 kHz that fill the upper half of the head	Modeling and experiment; the authors warn that "accurate defect sizing will be difficult" (Hesse and Cawley, JASA 2006)
RCF depth for grinding decisions	Benchtop surface-wave system on a steel block with machined features	"Proof of concept" stage (Li, Oldknow and Brunskill, JRC 2024)

Rail welds are a related case. Field welds need hand UT of the base because test cars miss most of it, and that work is usually done with angle-beam shear probes placed on the foot. Surface waves on the foot remain a research option, not a standard weld procedure.

Key takeaway: surface wave UT is an established manual technique on structural steel, but on rail it is still a research method, mainly for the foot and for RCF depth.

What are the limits of surface wave UT on rail?

Surface wave UT sees only the top wavelength, struggles to size what it finds, and is very sensitive to the surface it travels on. Before using or specifying it, plan around these limits:

It can't find internal defects. At 2 MHz it samples about 1.5 mm. Transverse fissures start inside the head, and vertical and horizontal split heads lie deeper. None of these is in range.
Sizing is weak. Rail heads support several surface modes with similar velocities, so the received amplitude changes with position (Hesse and Cawley). Crack angle is unknown, which is also eddy current's main sizing problem.
Dense RCF limits range. Each crack mouth reflects and scatters energy. In a curve with continuous head checks, the wave may not get past the first few cracks to reach a deeper or larger one behind them.
The surface carries the wave. Lubricant, couplant pooling, rust, grinding facets and worn gauge-corner profiles all change amplitude. A worn gauge face changes the wave path, so read gauge face wear in rail before assuming a new-rail geometry.
Geometry echoes. Rail ends, bolt holes, weld flashing and changes in head contour all reflect surface waves, and each has to be separated from a crack.
Coupling and speed. Contact surface-wave probes need couplant and steady wedge contact. Non-contact EMAT and laser systems are the research answer, but none has replaced a production test or eddy current trolley in North America.
No acceptance basis. Part 213 and the 213.113 remedial action table size defects as a percentage of head area or by length. No federal or industry criterion rates a surface-wave indication, so findings feed maintenance decisions such as grinding, not defect disposition.

Key takeaway: treat surface wave UT as a shallow, qualitative screen, never as a replacement for the internal rail search.

How does surface wave UT compare with other methods?

Each method covers a different depth band, so a rail program combines them. The table below corrects the original article's ratings, which gave surface waves "excellent" surface sensitivity and phased array a "good" rating without evidence.

Method	Depth reached	Rail use today	Couplant	Main limitation
Surface wave UT (Rayleigh)	About one wavelength (1.5 mm at 2 MHz)	Research and lab on rail; manual on structural steel	Yes for contact probes; no for EMAT or laser	Shallow only; hard to size
Conventional UT (0°, 37–45°, 70°)	Whole head, web and part of the base	Standard internal rail testing (213.237)	Yes (wheel fluid or water)	Masking and no-tests on RCF; misses most of the base
Phased array UT	Whole head, steerable	Verification and research (TTCI prototype)	Yes	Not yet a production rail test
Eddy current	Surface-breaking cracks; depth estimated	Walking-stick, trolley and train-mounted RCF measurement	No	Assumed crack angle; 1–2 mm standoff
ACFM	Surface-breaking cracks; sized	Walking-stick RCF inspection	No	Slower; calibration for rail crack shapes
Magnetic particle	Surface and near-surface, ferromagnetic only	Spot checks, crack confirmation	No	Labor and weather; no depth

For more on the electromagnetic side, see eddy current testing vs ultrasonic testing. For steered beams, see phased array ultrasonic testing.

Key takeaway: surface wave UT fills a narrow surface band that eddy current and ACFM already cover in rail service, while bulk-wave UT covers the interior that no surface method can reach.

How EMA Industries applies this

EMA Industries performs ultrasonic rail flaw testing for transit agencies, short lines and industrial track owners, and rail weld inspection. We're based in Southern California and mobilize nationwide. On rail with heavy RCF, the report should mark every stretch where a valid search couldn't be made, so the owner can plan grinding or a hand test before the 213.237 deadline. To scope rail testing or weld inspection, see our ultrasonic examination services or contact us.

Frequently asked questions

How fast does a Rayleigh wave travel in steel?

About 3,000 m/s (0.118 in/µs), roughly 0.92 times the shear velocity. Calculated from Evident's 5,890 m/s longitudinal and 3,250 m/s shear values, the exact Rayleigh equation gives 3,004 m/s; with a 3,230 m/s shear value it gives 2,988 m/s. Calibrate on the actual rail steel.

How deep does surface wave UT inspect?

About one wavelength. In steel that is about 1.5 mm at 2 MHz and 0.6 mm at 5 MHz. Lower frequencies reach deeper (6 mm at 0.5 MHz) at the cost of sensitivity to the smallest cracks.

Can surface wave UT replace rail flaw testing under 49 CFR 213.237?

No. Section 213.237 requires a continuous search for internal defects, and a surface wave can't reach the interior of the rail head, web or base. Test cars use 0°, roughly 37–45° and 70° bulk-wave probes for that search.

What finds gauge-corner cracking in practice?

Eddy current and ACFM systems, supported by visual, dye penetrant and magnetic particle checks, with rail grinding to remove the cracked layer. Conventional 70° UT finds the deeper detail fractures that RCF can lead to. Phased array and EMAT surface-wave systems are still in research.

Why does RCF cause problems for ultrasonic rail testing?

Dense shallow cracks scatter the beam and can stop sound reaching the base and back, which produces a no-test. Shells and head checks can also hide detail fractures beneath them. Grinding the damaged layer usually restores testability.

Surface Wave UT for Rail: What It Finds and Where It Fits