What Glass Actually Does to a Bullet — Windshields, Windows, and Bulletproof Glazing

Ask someone whether a car windshield will stop a bullet and you'll get a confident answer either way — usually learned from a movie. The real answer is that "glass" isn't one material in terminal ballistics. It's three completely different problems, and the difference between them decides whether a bullet keeps flying, changes course, or stops dead.

Glass is three problems, not one

A residential window pane, a laminated windshield, and a sheet of bulletproof glazing all look like glass. Ballistically, they have almost nothing in common. The thin pane is a speed bump that shaves velocity and sprays fragments. The laminated windshield is a trajectory modifier that bends the bullet's path and tears bullets apart. The bulletproof laminate is an energy sink engineered around a measurable stopping threshold. Treat them as one material and every prediction you make — for home defense, for a vehicle engagement, for a shooting reconstruction — will be wrong in a different way.

Here's how each regime actually behaves, with the measured numbers behind it.

Regime 1: Thin glass — binary penetration, real consequences

A residential window is typically 3–4 mm of annealed glass. A car's side window is 3–5 mm of tempered glass — the same soda-lime material, but pre-stressed so it shatters into small dice instead of long shards. Against any centerfire bullet, both answer the penetration question the same way: yes. There is no realistic pistol or rifle load that a single thin pane stops. The interesting part is what the pane does to the bullet on its way through.

First, it taxes velocity. Waghmare et al. measured a 9mm bullet dropping from 373 m/s to 269 m/s after defeating a glass pane — roughly a 28% velocity loss. Because kinetic energy scales with velocity squared, that ~28% velocity cut is close to a 48% energy cut. The bullet still penetrates like a bullet on the far side, but it arrives with half the energy budget it left the muzzle with. Whether that matters depends entirely on what's behind the glass and how much energy the job needs.

Second, thin glass is a fragment generator. Annealed glass breaks into shards; tempered glass "dices" — the entire pane converts into thousands of small cubes at once. Those fragments carry enough energy at close range to cause secondary injuries, and in forensic work the total collapse of a tempered side window is a headache of its own: the pane that recorded the bullet hole no longer exists.

Third — and this is the part almost nobody expects — thin tempered glass can throw a bullet badly off course. In the classic study, Thornton and Cashman (Journal of Forensic Sciences, 1985) fired .38 Special bullets through tempered automobile glass while replicating a real case and recorded deflections as large as 26.4 degrees. That is not a rounding error. At 10 meters past the glass, a 26° deflection moves the impact point by more than 4.5 meters. A slow, round-nose handgun bullet meeting a pre-stressed pane can leave it pointed somewhere genuinely different from where it was aimed.

Regime 2: The laminated windshield — a trajectory modifier

A windshield is a different construction: two plies of annealed glass bonded to a polyvinyl butyral (PVB) interlayer, typically about 5–6 mm total. The PVB exists to keep the pane in one piece during a crash. Ballistically, it changes the interaction in three ways.

It holds together instead of vanishing

Where a tempered side window converts entirely into dice, a laminated windshield stays in the frame with a hole in it. The bullet punches a crater, radial and concentric fractures spread from the impact, and the PVB holds the fractured plies together. For forensics this is gold: the crater's cone fracture is wider on the exit side (telling you the direction of fire), and where two shots' fracture patterns intersect, the later crack terminates against the earlier one — letting examiners sequence multiple shots on the same pane.

It deflects the bullet — modestly, but systematically

This is the effect people overrate in magnitude and underrate in consequence. Controlled tests on laminated windshields — Wilgus, White and Berry (Journal of Forensic Identification, 2013) — found that bullet deflection through windshield glass stayed within roughly ±5 degrees for all but two of their test shots. Data compiled across North American shooting-reconstruction courses puts the average for handgun rounds fired from outside the vehicle at about 3 degrees downward. Shots fired from inside the vehicle out are trained around the opposite expectation: deflection upward, away from the glass surface.

Three degrees sounds trivial next to Thornton's 26.4° on tempered glass. It isn't. The windshield sits at a 30–45° rake, the bullet crosses it obliquely, and the target is usually several meters beyond. At 15 meters past the glass, a 4-degree deflection displaces the impact point by over a meter — the width of a car seat. Aim at the driver through the windshield and the bullet can arrive at the passenger's headrest. For a shooting reconstruction the logic runs in reverse: a trajectory rod pushed straight through a windshield defect does not point back at the shooter. Any reconstruction through laminated glass has to model the deflection, the glass angle, and the bullet type, or it locates a shooter who was never there.

It attacks the bullet's structure

The oblique, layered impact does something a flat steel plate doesn't: it shears at the bullet. Jackets separate from cores; hollow-point cavities collapse or pack with glass; the bullet often exits yawing, deformed, and lighter than it entered, with the jacket sometimes taking a separate path entirely. This is where regime 2 hands off to the most practically important section of this article — what glass does to defensive ammunition. More on that below.

Regime 3: Bulletproof glazing — engineering around a threshold

The third regime stops pretending glass is an obstacle and turns it into armor. Bulletproof glazing (more honestly: bullet-resistant glazing) is a thick multi-layer laminate — glass plies bonded with polymer interlayers, almost always finished with a polycarbonate layer on the protected side. The glass plies do the hard work: they blunt and erode the bullet's nose and spread the load. The polycarbonate does the ductile work: it stretches, catches the slowed bullet and the glass spall, and keeps fragments off the person behind it.

V50: the number that defines the regime

Unlike a thin pane, armor glass has a measurable threshold — the V50 ballistic limit, the impact velocity at which a given projectile perforates the target 50% of the time. Below it, the glazing wins most engagements; above it, the bullet does. Everything in transparent armor design is about pushing V50 above the threat's actual impact velocity.

Laboratory data shows how far ordinary laminated glass is from that goal. In the NTNU test program on laminated float-glass panes, Osnes and colleagues measured a ballistic limit of about 232 m/s for 9mm ball — while a 9mm leaves a service pistol at roughly 360 m/s. For 7.62 mm armor-piercing bullets, Osnes et al. (International Journal of Impact Engineering, 2021) determined a ballistic limit around 395 m/s against laminated glass plates, with test shots spanning 375–700 m/s — while rifle AP arrives at over 800 m/s. A windshield-class laminate sits far below every realistic threat velocity. Certified glazing closes that gap the only way physics allows: more plies, more thickness, more mass. Handgun-rated transparent armor runs roughly 20–40 mm thick; rifle-rated panels run 40–80+ mm and weigh accordingly. There is no thin, light, rifle-proof window. The threshold is bought in millimeters and kilograms.

Ratings: what the levels actually mean

In the US, transparent armor is commonly rated under NIJ Standard 0108.01 (with UL 752 and Europe's EN 1063 as the industry equivalents). The levels map to threats, not marketing:

The jump from IIIA to III is the expensive one. Handgun bullets arrive below ~450 m/s with soft construction; rifle ball arrives near 850 m/s, and AP adds a hardened core that erodes glass instead of being eroded by it. That's why a Level IIIA teller window and a Level IV embassy window are barely the same product category.

The FBI protocol: what glass does to hollow points

Here's the section that matters most if you carry a firearm or load one for home defense — and it's the reason the FBI tests ammunition through glass at all.

A jacketed hollow point works by hydraulics: soft material flows into the nose cavity and forces the bullet to mushroom. That mechanism assumes the cavity arrives at the target open and empty. Glass breaks both assumptions at once. The fractured plies shear at the exposed hollow-point nose and pack the cavity with glass and interlayer material. A plugged cavity can't take in tissue, so it can't expand. The bullet that exits the windshield is, functionally, a flat-nosed FMJ — it penetrates deep and doesn't mushroom. Laminated glass at an angle is even harsher: it's the barrier most likely to strip the jacket off the core entirely, sending two projectiles downrange instead of one.

If this sounds familiar, it's the same failure mode we documented for wallboard in our article on JHP after drywall — the cavity clogs, expansion dies, penetration grows. Glass just does it more violently, and adds jacket separation on top.

This is exactly why the FBI's ammunition test protocol — built after the agency's 1986 Miami shootout forced a hard look at handgun terminal performance — doesn't test ammunition in bare gelatin alone. The protocol fires every candidate load through a series of intermediate barriers, including laminated automobile glass set at a steep angle, into calibrated ordnance gelatin behind it, and requires adequate penetration (the familiar 12–18 inch window) after the barrier. A load that expands beautifully in bare gel and falls apart after glass fails the protocol.

The ammunition industry's answer is what marketing calls "barrier blind" ammunition: bonded designs where the jacket is metallurgically fused to the core (Speer Gold Dot, Federal Tactical Bonded, Winchester Ranger Bonded) or monolithic all-copper hollow points with no jacket to lose. Bonding doesn't stop the cavity from plugging — nothing does — but it keeps the bullet in one piece through the glass, so the mass that was supposed to arrive, arrives. When a duty load is advertised as "FBI protocol tested", the glass stage is the one it's bragging about surviving.

What this means in practice

Home defense

Your window will not stop anything. But it isn't ballistically transparent either: it takes a quarter of the bullet's velocity, half its energy, converts a JHP toward FMJ-like behavior, and adds a fragment shower. For over-penetration planning the direction matters too — a defensive round that misses and exits through a window keeps most of its lethality on the far side. The pane changes the bullet's behavior; it does not change your responsibility for where the bullet lands.

Forensics and shooting reconstruction

Glass type is not a detail — it's the first branch in the decision tree. Tempered glass can deflect a slow bullet by tens of degrees and then destroy itself as evidence. Laminated glass deflects modestly but systematically, preserves fracture evidence, sequences multiple shots, and reliably alters the bullet (mass loss, jacket separation) between the muzzle and the wound. A reconstruction that strings a straight line through a windshield defect without modeling deflection and glass angle is producing a confident wrong answer.

Vehicles

A car is two glass regimes bolted to one chassis. The windshield is laminated: it bends trajectories, strips jackets, plugs hollow points — and stays in place. The side windows are tempered: they offer one pane's worth of resistance and then cease to exist, with real deflection risk for slower rounds while they're being defeated. Same car, same bullet, two different ballistic problems depending on which pane it meets. Anyone who trains around vehicles learns this distinction early, because the aim correction that's right through the windshield is wrong through the door glass.

Run the numbers: glass in the BallisticEngine model

BallisticEngine models tempered glass as one of its 18 calibrated materials, and multi-layer targets let you stack plies — glass over polycarbonate over glass — the way real glazing is actually built. The engine also models the clogged-cavity effect: a JHP that defeats a hard barrier is computed as a non-expanding bullet downstream, which is exactly the FBI-protocol behavior described above.

Three scenarios, straight from the engine:

A note on modeling honesty: the engine's glass material is calibrated for tempered glass, and a multi-ply stack approximates a laminate's energy budget — the PVB interlayer's contribution to stopping the bullet is small next to the glass plies, though it's everything for holding the pane together. What the penetration engine deliberately does not predict is deflection angle; that depends on obliquity, bullet construction and velocity in ways the forensic literature above measures empirically. Use the engine for the energy question, the literature for the geometry question.

The bottom line

"Will glass stop a bullet?" is the wrong question, because glass isn't one thing. A thin pane says yes, pass — but slower, dirtier, and maybe not where you aimed. A windshield says pass, minus a few degrees and possibly your jacket. Bulletproof glazing is the only glass that gets to say no — and only up to a velocity you can look up, for a threat level printed on the spec sheet. Once you sort every glass question into one of those three regimes, the movie-logic answers fall away and the physics gets predictable. That's the whole point of modeling it.