Vertical Split Head in Rail: Definition, Causes and Ultrasonic Detection

A vertical split head is a rail defect defined in 49 CFR 213.113(d)(16): a vertical split running lengthwise through or near the middle of the rail head. Side-looking ultrasonic channels find it, and the 0° channel often shows a lost base echo. A split over 4 inches long falls under remedial action B, at 30 mph at most.

This guide is for track supervisors, transit engineers and rail flaw test operators who need to recognize, verify and size the defect.

What is a vertical split head under 49 CFR 213.113?

A vertical split head (VSH) is a lengthwise split in a vertical plane through or near the middle of the rail head, as defined in 49 CFR 213.113(d)(16). The FRA definition, as published on Cornell's Legal Information Institute copy of the eCFR, reads in full:

"Vertical split head means a vertical split through or near the middle of the head, and extending into or through it. A crack or rust streak may show under the head close to the web or pieces may be split off the side of the head." — 49 CFR 213.113(d)(16)

Each part of the definition tells the inspector something:

• "A vertical split" fixes the plane. The crack face stands upright and runs along the rail, parallel to the direction of travel.
• "Through or near the middle of the head" fixes the position. The split usually sits close to the head's centerline, which is also where a centered 0° probe looks.
• "Extending into or through it" means the split can stay inside the head or run through its full depth, toward the running surface above and the web below.
• "A crack or rust streak may show under the head close to the web" is the main visual sign. When the split reaches the underside of the head, a line or a rust stain can appear along the fillet on either side.
• "Pieces may be split off the side of the head" describes the late stage.

The definition is one of 16 in paragraph (d). Under 213.113(a), a person designated under 49 CFR 213.7 decides whether track with a defective rail may stay in use, and paragraph (c) holds the remedial action table. The current text is on eCFR, 49 CFR 213.113; Cornell's copy carries the amendment line "79 FR 4256, Jan. 24, 2014; 79 FR 4633, Jan. 29, 2014, as amended at 85 FR 63388, Oct. 7, 2020."

Why the vertical split head has extra regulatory weight

The vertical split head is one of only four defects that count as a service failure under the internal rail inspection rule. Section 213.237(j)(3) defines a service failure as "a broken rail occurrence, the cause of which is determined to be a compound fissure, transverse fissure, detail fracture, or vertical split head" (49 CFR 213.237). Those failures feed the service failure rate that each owner must hold below 0.1, 0.09 or 0.08 per mile per year, depending on class, passenger service and hazardous materials traffic. A horizontal split head, by contrast, is not on that list. So a broken rail from a VSH can push a test segment over its rate and force the owner to test more often, while a break from an HSH alone does not.

How the industry and FRA code the defect

FRA research report ORD-11/24, Rolling Contact Fatigue: A Comprehensive Review (November 2011), lists the standard railroad defect classifications, including VSH for vertical split head (FRA ORD-11/24, Table 2). In accident reporting, FRA cause code T221 covers a broken rail from a vertical split head. FRA's train accident data on data.transportation.gov list 55 main-track derailments coded T221 from 2016 to 2025 (pulled October 24, 2026), the third-largest broken-rail code after T207 detail fracture (110) and T220 transverse or compound fissure (57). The same pull shows 10 for T212 horizontal split head (FRA train accident data). These are counts of derailments with VSH as the primary cause, not counts of defects found in testing, and they change as late reports are filed.

Key takeaway: the vertical split head is defined by its plane and position, and it is the only split head that counts toward the 213.237 service failure rate.

How does a vertical split head form?

A vertical split head forms when a lengthwise internal discontinuity in the rail head, most often a vertically oriented inclusion stringer or seam from manufacture, grows along the rail under repeated wheel loads. It is a steel-quality defect first and a fatigue defect second.

FRA's ORD-11/24 review is the clearest public statement of the link. In its section on steel cleanliness it says that inclusion stringers "are strongly implicated in the formation of vertical split heads," along with deep-seated shells and the transverse defects that grow from them (ORD-11/24, section 4.1.2, printed p. 64). It cites a 1978 ASM failure analysis of railroad rail by Savage and Steele for that point. The same report credits cleaner steel with the decline of the defect: vacuum degassing, together with "a requirement to test every batch of rail both ultrasonically and (samples) destructively has eliminated many of the vertical slit [sic] heads, split web, and other problems that have arisen in the past." A later passage adds that cleaner steels, made by cutting hydrogen, sulfur and alumino-silicate inclusions, "have significantly reduced the number of catastrophic failures such as shattered rims, tache ovals, vertical split heads, and others."

A rail inspection patent describes the mechanism in nearly the same terms: "A railhead stringer that is vertically oriented can grow in the vertical plane along the axis of the rail," and the inventors call the result "potentially an extremely serious type of defect as it can result in the loss of the running surface of the rail" (US 6,600,999 B2, granted 2003). A patent is its inventors' technical description, not an independent study, but it agrees with the FRA review.

Seams are often named alongside inclusions as a mill origin. The public FRA and TTCI documents we found name inclusion stringers for the vertical split head, and the seam wording appears for the horizontal split head in US 6,600,999 ("usually originates from a longitudinal seam or inclusion"); none gives a separate VSH mechanism for seams (search, October 25, 2026). This guide therefore treats a seam as a possible lengthwise weak plane of the same kind, not as a separately documented cause.

What drives growth after the split starts:

Why length matters so much for this defect: ORD-11/24 summarizes a 2003 review by Cannon and others as saying that "defects that either run the length of the rail or appear in clusters are the most dangerous," and it places the vertical split head and head-web separation in the first group. The same review notes that losing a length of rail head "from a few millimeters to a couple of meters" is enough to cause a derailment. ORD-11/24 also reports, from a 2010 Zarembski paper analyzing FRA data from the 1980s, that one broken-rail derailment occurred for every 770 defects overall, and one for every 526 for "contact stress-related defects (such as transverse defects and vertical split heads)" (ORD-11/24, section 1.2, printed p. 8). That is a 1980s figure for a mixed group, not a current VSH rate.

No public FRA or TTCI growth rate for vertical split heads was found (search, October 25, 2026); the rule handles growth through reinspection notes F and G.

Key takeaway: a vertical split head starts from a lengthwise weak plane built into the steel at the mill, and cleaner rail steel is the main reason it is less common than it once was.

How is a vertical split head different from a horizontal split head and a split web?

A vertical split head stands upright in the head, a horizontal split head lies flat in the head, and a split web is a lengthwise crack in the web below the head. All three run along the rail and share one remedial action row in 213.113, but they sit in different planes and places, so ultrasonic testing finds them in different ways.

Vertical split head vs horizontal split head

Section 213.113(d)(11) defines the horizontal split head as "a horizontal progressive defect originating inside of the rail head, usually 1/4 inch or more below the running surface and progressing horizontally in all directions." The two defects are at right angles to each other. That one difference decides the probe. A 0° beam travels straight down from the running surface. It meets a horizontal crack face-on and gets a strong echo back, but it runs parallel to a vertical split and sees it nearly edge-on. So the HSH is a 0° target, and the VSH is mainly a side-looking target. Our article on the horizontal split head covers the HSH signals, sizing and worked example in detail.

Vertical split head vs split web and piped rail

Section 213.113(d)(14) defines a split web as "a lengthwise crack along the side of the web and extending into or through it," and (d)(13) defines piped rail as "a vertical split in a rail, usually in the web, due to failure of the shrinkage cavity in the ingot to unite in rolling." A split web and piped rail sit in the web; a vertical split head sits in the head. A VSH that grows downward through the head can reach the web, which is why one rail inspection patent says the 0° transducers can show "some vertical split heads that occur in the web region" (US 9,010,186 B2, granted 2015).

Side-by-side comparison

Key takeaway: the plane decides the probe: horizontal is a 0° echo target, vertical is a side-looker target, and the remedial row is the same for both.

How does ultrasonic testing find a vertical split head?

Ultrasonic testing finds a vertical split head mainly with side-looking channels aimed across the head, backed by the 0° channel, which loses the base echo when the split is wide, leaning or deep enough to block the beam. The 70° channels can also respond. No single channel is reliable alone, so the call comes from the channels together.

Why the 0° beam is weak on a vertical split head

A crack reflects best when the beam meets its face at a right angle. A vertical crack face is parallel to a 0° beam, so the beam slides down beside it and returns little direct echo. Tight splits right under the beam can let most of the sound pass to the base. That is why rail test equipment descriptions treat the 0° response to a vertical split head as indirect.

The 0° channel still matters. A 1998 test-system patent describes the 0° beam as aimed at horizontal defects and adds: "It also detects many vertical split heads. This beam is directed normal to the rail running surface, penetrating through the web to the rail base. It will also detect any defect which masks the bottom of rail echo" (US 5,777,891, granted July 7, 1998). The same patent describes a separate data channel that watches the base echo, and an alarm, a "Lack-Of-Expected Response," when that echo disappears.

The 0° signs of a vertical split head

A vertical split head shows on the 0° channel in three ways, depending on how wide it is and where it runs:

1. Loss of base echo when the split is wide. An open split, or one that leans, puts a steel-to-air boundary across part of the beam. Less sound reaches the base, so the base echo drops or disappears over the length of the split.
2. A "dark" 0° trace. On a B-scan or strip display, a stretch where the base echo is missing and no early echo replaces it shows as a blank, or dark, band in the place where the steady base-echo line should be. Over a VSH this can be the main 0° sign: the base line goes dark, but nothing appears in the head gate, because the split gives no flat face to reflect from.
3. Scattered or doubled echoes where the split reaches the web. A split that has turned or run into the web can give irregular echoes in the web gate, which is what the 2015 patent's "vertical split heads that occur in the web region" describes.

Lost coupling looks the same: a dry or lubricated patch, debris, or a worn head that steers the beam off the web can all kill the base echo. If the side-looking channels show an indication over the same stretch, there is a reflector. If no channel shows anything and the base echo is gone, the rail over that stretch has not been tested, and 213.241(c)(6) requires the record to show the location of any track not tested under 213.237(g).

Side-looking channels

Side-looking channels aim the beam across the head, in the transverse plane, rather than along the rail. A beam crossing the head meets a vertical split nearly face-on, which is the geometry the 0° beam lacks. The same 1998 patent describes optional 40° shear-wave channels "aimed at the field and gage sides of the rail (thus, called side looking)," and says: "These channels are used mainly for testing for vertical split head defects."

A 1995 side-looker patent explains how the call is made. Its beam is aimed at a lower side corner of the head, and "the presence of premature reflections or the or absence of reflections can be used to detect flaws such as vertical fractures in the rail head" (US 5,419,196). On sound rail, the side-looker returns a steady corner echo at a known time. A vertical split between the entry point and the corner gives an earlier echo, blocks the corner echo, or both. The same patent notes that an older 40° side-looker "can be effective for detecting vertically split rail heads" but doesn't allow for the curvature of the rail head and head wear, which it was designed to address with a 45° to 75° beam.

Other designs use the same idea. A 2003 patent describes "cross-rail" transducers "mounted laterally away from the centerline of the rail and angled back toward the center" to help detect vertical split heads (US 6,600,999 B2), and a 2011 patent describes earlier proposals to send sound through the head on a "Z path" to reach vertical longitudinal flaws (US 8,020,446 B2). These are inventors' descriptions; no public source sets one North American rail channel list, so treat the channel set as typical practice.

The 70° channels

The 70° channels look along the rail at transverse defects. The 1998 patent says the 70° beam "is also useful in detecting vertical splits in the head area." A long vertical split can block or redirect the 70° beam as it crosses the head, so a run of 70° indications that line up with a side-looker indication supports the VSH call.

Confirming and sizing by hand

A test vehicle or cart flags a suspected location. Under 213.113(b), the owner must then verify the indication within four hours, or immediately if the suspected defect would need remedial action A, A2 or B. For a VSH that may be over 4 inches long, that means immediately. Owners testing under 49 CFR 213.240 follow that section's verification windows instead.

The hand test usually has four parts:

• Look under the head on both sides for a crack or rust streak near the web, and along the side of the head for any piece starting to split away.
• Scan across the head with an angle probe from the running surface toward each side, checking for an early echo or a lost corner echo.
• Check the 0° trace along the suspect length for a lost base echo with clean coupling, which confirms the dark stretch is the rail and not the probe.
• Measure length by moving along the rail and marking where the cross-head indication starts and stops, using the edge method the owner's procedure specifies.

Key takeaway: the side-looker finds the vertical split, and the 0° dark trace shows how far it blocks the section, but only once coupling has been ruled out.

How well did test vehicles find a vertical split head in benchmark trials?

In TTCI's 1998 benchmark of six North American test systems, all six runs found the single vertical split head on the test track, which TTCI reported as 100 percent reliability, but it was one flaw, 120 inches long.

TTCI tested six hi-rail rail flaw detection systems between February and May 1998 on the Rail Defect Test Facility at FRA's Transportation Technology Center near Pueblo, Colorado. The 49 flaws had been found in revenue service and verified by TTCI. They were two bolt-hole cracks, four defective welds, 40 transverse defects, two horizontal split heads and one vertical split head. TTCI's digest states: "The reliability of finding the vertical split head was 100 percent" (TTCI TD-99-006, Garcia and Reiff, February 1999). Its flaw list gives the vertical split head's size as 120 inches (304.8 cm).

Three cautions apply. First, n = 1: one defect, seen six times. Even if the six runs were independent, six hits in six tries gives a 95% interval of about 61% to 100%, and they aren't independent, because every system looked at the same defect. Second, a 120 inch split is about thirty times longer than the 4 inch boundary for action B, so it is the easiest case. Third, these are 1998 systems; the result isn't a detection rate for today's equipment. TTCI also wrote that "broken rails, transverse defects (TDs), and vertical split heads comprise the third highest cause of train derailments."

No public benchmark of detection for short vertical split heads (under 4 inches) was found (search, October 25, 2026).

Key takeaway: quote the 1998 VSH result as "one long VSH found on all six runs," never as a 100% detection rate.

Worked example: side-looker timing and length on 136RE rail

A side-looker echo time gives the position of a vertical split across the head, and the length along the rail sets the remedial action. This example uses nominal 136RE dimensions, a shear velocity of 0.1280 in./µs (about 3,250 m/s) for the side-looker and a longitudinal velocity of 0.2320 in./µs (about 5,890 m/s) for the 0° channel. Times are steel-only round trips; wheel and wedge delays are excluded, and real instruments add them. The probe geometry is illustrative, not one vendor's design.

Inputs

Step 1: expected echoes on sound rail

For a beam at 60° from vertical, the horizontal distance crossed is the sound path × sin 60° (0.8660), and the depth reached is the sound path × cos 60° (0.5000).

On sound rail the side-looker shows a steady far-side echo near 40 µs. The exact time depends on head shape and wear, so it is checked on known-good rail at the start of each run.

Step 2: position of the indication

Sound path = 0.1280 × 18.0 ÷ 2 = 1.152 in.

Horizontal distance = 1.152 × 0.8660 = 0.998 in. from the entry point, so the reflector sits 0.998 − 0.75 = 0.248 in. to the field side of the centerline, at a depth of 1.152 × 0.5000 = 0.576 in.

That is "near the middle of the head," as 213.113(d)(16) describes. It isn't an HSH, because a horizontal crack at that depth would also give the 0° channel an early echo at 2 × 0.576 ÷ 0.2320 = 4.97 µs, and the 0° head gate is clean. The far-side echo at 40 µs is reduced or gone over the same stretch, because the split shadows it, which is the "absence of reflections" sign in the 1995 patent.

Step 3: length and the remedial action

The operator follows the indication along the rail by hand and marks where it begins and ends, using the owner's edge method. The side-looker indication runs 14.0 inches. Over the middle 3.5 inches of that length, where the split leans and opens, the 0° base echo also goes dark, with good coupling confirmed on both sides of the stretch.

If the length were taken from the 0° dark stretch alone, it would read 3.5 inches, which falls in the more than 2, up to 4 inch band: I and G, 30 mph and reinspection within 30 days. The side-looker length of 14.0 inches is over 4 inches: B, a speed set by a designated person with at least one year of supervisory experience, never over 30 mph or the class maximum, and verification immediately under 213.113(b).

The 0° loss of base echo marks only the part of the split that blocks the beam, so it undersizes the defect. Length should come from the channel that sees the split itself, and near a 1, 2 or 4 inch boundary the conservative choice is the more restrictive action until a second measurement settles it.

Step 4: what the record needs

Under 213.241(c), the inspection record must show the date, the track inspected, the location and type of each defect (VSH, not "split head"), its size if not removed before the next train, and the initial remedial action and its date. In this example the entry would read as a 14.0 in. vertical split head about 0.25 in. to the field side of center, with remedial action B. If the rail later breaks at that spot, the break counts as a service failure under 213.237(j)(3).

Key takeaway: size a vertical split head from the cross-head channel, not the 0° dark stretch, and check every position against the head width.

What remedial action does 213.113 require for a vertical split head?

For a vertical split head that is not replaced or repaired, the 213.113(c) table sets the action by length: more than 1 to 2 inches H and F, more than 2 to 4 inches I and G, more than 4 inches B, and a break out in the rail head A. The vertical split head shares this row with horizontal split head, split web, piped rail, head and web separation and longitudinal defective weld. The table is published as an image on eCFR and Cornell, so the values below come from the Federal Register text of the 2014 rule (79 FR 4256–4257). The 2020 amendment changed only the verification sentence in 213.113(b).

The lettered notes work the same way for every defect in the row. Note A is visual supervision of each operation by a designated person, not a stop order. Notes F and G restart the reinspection clock at each check, and a missed reinspection drops speed to Class 2 until the rail is measured again. Notes H and I cap speed at 50 and 30 mph or the 213.9 class maximum, whichever is lower. The horizontal split head guide quotes the notes in full, and our guide to FRA defect codes and the 213.113 table explains the whole table.

Three points specific to the vertical split head

Long splits go straight to B. The defect grows along a stringer, and the one in the 1998 benchmark was 120 inches long, so many VSH calls land in B, with immediate verification.

A break out is the end point to watch for. The rule text says pieces "may be split off the side of the head," and the 2003 patent warns of "loss of the running surface." Once a side of the head has broken away, the defect is "break out in rail head" and the action is A.

A break counts against the owner's rate. Because the vertical split head is a 213.237 service failure defect, a VSH left in track under B that then breaks adds to the segment's service failure rate. An HSH that breaks does not, unless it has turned into a compound fissure.

The rail flaw test operator, qualified under 49 CFR 213.238, finds and sizes the defect; the person designated under 213.7 applies the action. On Class 6 and higher track, the equivalent table is in 49 CFR 213.337.

Key takeaway: the vertical split head uses the HSH length row, but its length is easy to undersize and its break counts as a service failure.

What are the limits of ultrasonic testing for this defect?

Ultrasonic testing can miss a tight vertical split head, undersize a long one and confuse a dark 0° trace with lost coupling. These are the limits a Level III would expect an owner and an operator to plan around.

Edge-on to the 0° channel. A normal 0° trace is not evidence that no vertical split head is present.

Side-looker coverage and head wear. The side-looker relies on a known echo from the far side of the head. Head curvature, side wear and a worn gauge face move that echo and can hide the change a split makes. The 1995 patent was written because the earlier 40° design didn't handle curvature and wear. If a system's side-looking channels are weak or switched off, early vertical split heads can pass.

The dark trace is ambiguous. Treating every dark stretch as coupling, or every one as a defect, are both errors; the operator resolves each one and records what can't be retested.

Undersizing. Sizing from the 0° dark stretch can move a VSH from B down to I and G, as the worked example shows.

Short splits. The table starts at more than 1 inch, and a short tight split gives only a short run of side-looker indications at test speed. There is no public detection data for vertical split heads under 4 inches.

Surface conditions. Shells, head checks, engine burns and corrugation scatter the beam and weaken coupling. ORD-11/24 notes that surface fatigue "prevents the successful transmission of ultrasonic waves to the base of the rail and back," which produces a no-test and a hand test.

What the method can't tell you. Ultrasonic testing doesn't show whether a split has opened through the running surface or how much load the two halves of the head can still carry. That takes a look at the rail, and a break out decides the action.

For how these limits compare with other NDT methods on rail, see detection of rail defects using NDT methods, and for an inspector's view of the full defect list, see detecting rail defects: a guide for inspectors. The test intervals that decide how long a split can grow between searches are in our guide to 49 CFR 213.237.

Key takeaway: plan for the side-looker's sensitivity to head wear and for the dark-trace judgment; those two cause most missed or undersized VSH calls.

How EMA Industries applies this

EMA Industries performs ultrasonic rail flaw testing for transit agencies, short lines and industrial track owners from its base in Southern California, and mobilizes nationwide. For a vertical split head, the useful deliverable is a sized and classified defect: its position across the head from the side-looking channel, its length by a stated method, the 213.113 action band it falls in, and any dark 0° stretch that couldn't be explained and must be recorded as not tested. If you need suspected head defects verified and sized, or a test program planned around them, contact EMA Industries.

FAQ

Is a vertical split head the same as a horizontal split head?

It isn't. A vertical split head is an upright split along the middle of the head (213.113(d)(16)); a horizontal split head lies flat inside the head, usually at least 1/4 inch down (213.113(d)(11)). They share a remedial action row, but the vertical split head is a 213.237 service failure defect and is found mainly with side-looking channels, while the HSH is a 0° target.

What does a vertical split head look like on the rail?

The rule names two signs: a crack or rust streak under the head close to the web, and, later, pieces split off the side of the head. A split still inside the head shows nothing from the running surface, which is why internal testing under 213.237 is needed.

Why doesn't the 0° probe always find a vertical split head?

The 0° beam runs parallel to a vertical crack face, so it sees the split nearly edge-on and gets little echo back. It responds only when the split is wide, leans, or reaches the web enough to block the base echo. The side-looking channels, which cross the head, are the main tool.

What does a "dark" 0° signal mean?

It means the base echo has disappeared and nothing has replaced it, so the base-echo line on the display goes blank. Over a vertical split head that can be the defect blocking the beam, but lost coupling looks the same. The operator checks the side-looking channels and the coupling before calling it a defect or recording the stretch as not tested.

Does a broken vertical split head count against the service failure rate?

It does. Section 213.237(j)(3) lists vertical split head with compound fissure, transverse fissure and detail fracture as the causes of a service failure. A break from a VSH counts toward the segment's rate of 0.1, 0.09 or 0.08 per mile per year, depending on the track.

Sources

• 49 CFR 213.113, Defective rails (Cornell LII copy, read October 25, 2026)
• 49 CFR 213.113 on eCFR
• 49 CFR 213.237 on eCFR (service failure definition, 213.237(j)(3))
• 49 CFR 213.238, 213.240 and 213.241 on eCFR
• FRA final rule, Track Safety Standards; Improving Rail Integrity, 79 FR 4234, January 24, 2014 (remedial action table and notes, pp. 4256–4257)
• FRA ORD-11/24, Rolling Contact Fatigue: A Comprehensive Review, November 2011
• TTCI Technology Digest TD-99-006, Benchmarking Results of North American Rail-Flaw Detection Technologies, February 1999
• US Patent 5,777,891 (1998), real-time ultrasonic rail testing system
• US Patent 5,419,196 (1995), ultrasonic side-looker for rail head flaw detection
• US Patent 6,600,999 B2 (2003), hi-rail rail inspection system
• US Patent 8,020,446 B2 (2011), detecting flaws in a railhead
• US Patent 9,010,186 B2 (2015), detecting internal rail defects
• FRA train accident data, data.transportation.gov dataset byy5-w977 (pulled October 24, 2026)