Why Glass Breaks During Tempering: The Role of Heating, Quenching and Edge Quality

What actually causes tempering failures, and how to diagnose the real cause instead of guessing at furnace settings

Glass tempering is designed to increase strength. But the same thermal process that builds strength can also trigger fracture if the temperature gradient, cooling conditions, or existing flaws in the glass are not properly controlled. Most tempering-breakage investigations can be organized around three interacting areas: the thermal process, defects within the glass, and flaws associated with edge processing.

The short version

If your glass keeps breaking during tempering, the cause is usually one of three interacting areas: the thermal process itself, an internal defect (like an inclusion), or an edge flaw from cutting, grinding or polishing. Furnace temperature alone rarely explains it. The rest of this article breaks down how each of these actually works, and how to tell which one is happening on your line.

How Tempering Actually Builds Strength

Tempering places the glass surface in compression while the interior holds tensile stress. That compression is what makes the glass resistant to surface cracks growing. But glass is brittle. It can’t relieve concentrated stress the way metal does by bending or deforming. So the real question during tempering isn’t “did we cool it fast enough.” It’s “did the thermal cycle create a controlled stress state without pushing an existing flaw past its breaking point.”

Two panes run through the exact same recipe can behave completely differently if their thickness, edge condition, internal defects, or actual local cooling vary even slightly.

Heating: Setting Up the Thermal Condition

Heating isn’t just about getting the glass hot. It sets the starting condition that determines how the quench will behave. Fan et al. modelled an initial glass temperature of approximately 650°C, while Iglesias et al. found that 625°C gave better agreement with experimental results for their analysed case (Fan et al., 2016; Iglesias et al., 2022). These figures describe particular study conditions, not a universal furnace setpoint.

What matters more than the exact number is uniformity. If different areas of the pane enter the quench at different temperatures, they’ll cool differently too, and that shows up later as uneven residual stress.

Quenching: Where the Stress Is Actually Generated

After heating, the glass hits the quench zone, typically arrays of air jets cooling the surface fast. The surface loses heat much faster than the core. That mismatch is what creates the strengthening stress profile:

Hot glass → rapid surface cooling → temperature difference builds between surface and core → thermal contraction is constrained → transient stress develops → cooling continues through the thickness → final result: compressive surface, tensile core.

That’s the goal state. But Monnoyer and Lochegnies, studying a real industrial air-jet tempering system, found that the cooling field is not uniform across the pane; their numerical heat-transfer model was validated against surface-temperature and photoelastic stress measurements (Monnoyer & Lochegnies, 2008). That variation means that the cooling system can contribute to differences in the thermal and residual-stress history across the pane.

Why Glass Thickness Changes Everything

Thickness has a major effect on how fast the temperature gradient develops and how long it takes. Fan et al., using nonlinear finite-element analysis on glass from 1.8mm to 6mm, found the maximum temperature difference dropped sharply as thickness decreased (Fan et al., 2016):

• 6mm glass: approximately 76°C
• 4mm glass: approximately 65°C
• 3mm glass: approximately 31°C
• 2mm glass: approximately 20°C
• 1.8mm glass: approximately 10°C

These values are study-specific results from Fan et al.’s model and shouldn’t be read as universal targets (Fan et al., 2016). What matters is the pattern: the time available to establish the temperature gradient decreased sharply as thickness decreased. For the conditions studied by Fan et al., increasing heat transfer increased the temperature gradient in thin glass; the authors therefore identified improved cooling capability as one possible approach to the thin-glass tempering problem (Fan et al., 2016). That also means a recipe tuned for a thicker pane will not necessarily transfer cleanly to a thinner one.

The Quench Is a Heat-Transfer Problem, Not an Air-Pressure Number

It’s tempting to think of the quench as one setting: air pressure, fan speed, air volume. In reality the glass experiences a spatial field of heat transfer that varies jet by jet. That chain looks like this:

Nozzle and jet configuration → local airflow → local heat-transfer coefficient → local cooling rate → local temperature history → local residual stress.

Monnoyer and Lochegnies found substantial local variation in heat-transfer coefficient around the impinging jets in the industrial system they investigated (Monnoyer & Lochegnies, 2008). This means the cooling system itself becomes part of the mechanism generating breakage risk, not just a background setting.

Non-Uniform Cooling Creates Non-Uniform Stress

Iglesias et al., using a fluid-structure-interaction methodology, modelled the unsteady, non-uniform heat-transfer field produced by impinging jets and coupled it with the development of residual stress; their model agreed with prior experimental work to within an average error below 10% (Iglesias et al., 2022). Areas of the pane directly facing a jet cool differently than areas between jets, and that difference in cooling rate translates directly into a difference in residual stress across the same pane. So the glass isn’t experiencing one uniform quench. It’s experiencing a distributed process, and how even that process is determines how reliable your temper actually is.

Why an Internal Defect Can Turn Normal Stress Into a Break

Fracture in glass is governed by flaws, not just stress levels. A small, benign flaw and a deeper, sharper one behave completely differently under identical stress. Chang and Chou, investigating tempered glass fracture, found that stones or impurities located in the internal tensile region (not near the compressive surface) could act as fracture origins, with a strong relationship between defect location relative to the residual-stress profile and the probability of fracture (Chang & Chou, 1982).

Technical note for glass scientists and QC engineers: the governing relationship is the mode-I stress-intensity approximation, K_I = Y·σ·√(πa), where Y is a geometry factor, σ is the relevant tensile stress, and a is the characteristic flaw dimension. When the stress intensity at a flaw exceeds the material’s fracture resistance, crack propagation begins. This is why nominal average stress across a pane cannot predict failure on its own. Flaw location relative to the tensile/compressive zone matters as much as the flaw itself.

Edge Quality Is a Mechanical Property, Not a Cosmetic One

Cutting, grinding and polishing all mechanically interact with a brittle material and can introduce or worsen cracks, even when the edge looks fine visually. Bukieda et al., using microscopic examination and four-point bending tests on the edge surface, chamfer, and edge/chamfer transition, found that grinding and polishing can introduce flaws and that edge-processing quality has a measurable relationship with edge strength; changing polishing cup-wheel design was enough to improve chamfer surface quality and increase measured edge strength (Bukieda et al., 2020).

The practical implication: an edge can pass a visual inspection and still carry fracture-critical flaws.

Flaw Depth and Geometry Matter More Than People Assume

Vandebroek et al., using four-point bending on polished and cut edges, found meaningfully different flaw characteristics between the two finishing methods (Vandebroek et al., 2012):

• Mean initial critical flaw depth: approximately 41μm for polished edges vs. 117μm for cut edges
• Geometry factor: approximately 0.76 for polished edges vs. 0.86 for cut edges

These are study-specific results from Vandebroek et al.’s dataset, not universal constants (Vandebroek et al., 2012). But the underlying point holds broadly: how you finish an edge changes the flaw population on it, and the flaw population directly determines fracture strength. A tempering furnace cannot fix a critical edge flaw that was already introduced during cutting or edge processing before the glass ever reached the furnace.

How This Actually Plays Out on a Line

Edge and thermal factors don’t act in isolation. A common real failure chain looks like this:

Poor edge quality → deeper or sharper edge flaw → higher local stress concentration → lower fracture strength, combined with non-uniform or rapid cooling → increased local tensile stress → higher crack-driving force → fracture.

This is why chasing furnace settings alone after a breakage event often doesn’t fix anything. The furnace may be running exactly as designed, and the actual cause is sitting in the edge-prep station or in the raw glass itself.

Stored Energy and Why Tempered Glass Fragments the Way It Does

Tempering doesn’t just add strength, it stores elastic energy in the glass. Pourmoghaddam et al. developed a theoretical framework combining linear-elastic fracture mechanics with stochastic geometry to relate fragmentation morphology to stored elastic strain energy in tempered glass (Pourmoghaddam et al., 2019). Stewart and Prakash, using three-dimensional peridynamic simulations, developed an analytical relationship between residual-stress characteristics, stored energy and fragment density, and identified a modelled threshold associated with the onset of fragmentation susceptibility (Stewart & Prakash, 2023). Together, this body of work indicates that the amount and distribution of stored elastic energy influence the subsequent fracture and fragmentation behaviour of tempered glass (Pourmoghaddam et al., 2019; Stewart & Prakash, 2023).

Technical note: for the parabolic residual-stress approximation used by Pourmoghaddam et al., stored elastic strain energy per unit area is modelled as U = [(1−ν)/5E] × t × σ_s², where ν is Poisson’s ratio, E is Young’s modulus, t is glass thickness, and σ_s is surface compressive stress magnitude (Pourmoghaddam et al., 2019). This describes one specific stress-profile model, not every tempered-glass profile.

Measuring Whether Your Temper Is Actually Correct

ASTM C1279-23 is the current ASTM standard for non-destructive photoelastic measurement of edge and surface stress in annealed, heat-strengthened and fully tempered flat glass (ASTM C1279-23). ASTM C1048-25 is the current ASTM specification for heat-strengthened and fully tempered flat glass and addresses fabrication operations performed before strengthening or tempering (ASTM C1048-25).

One measurement caveat worth knowing: Efferz, Schuler and Siebert found that chamfered edges physically block direct retardation measurement all the way to the true edge, requiring extrapolation rather than a direct reading; their criticism concerned the ASTM C1279-09 evaluation approach for architectural glass specifically, not every provision of the current C1279-23 edition (Efferz et al., 2024). If you’re measuring edge stress on architectural glass with a chamfer, that limitation is worth understanding before trusting the number at face value.

How Can It Be Prevented and Improved?

No single universal tempering setting prevents every breakage mechanism. The evidence supports controlling the thermal process, cooling uniformity, glass thickness, edge quality, and verification methods as one integrated system, not as separate fixes.

• Match the thermal process to glass thickness. Fan et al. showed that thickness strongly changes the temperature gradient and quenching period, and thin glass needs particular attention because its thermal response is much faster (Fan et al., 2016). Develop and validate parameters for the specific thickness you’re running rather than assuming a recipe from another thickness will transfer.
• Control the actual cooling field, not just the setting. Industrial cooling is spatially uneven. Monnoyer and Lochegnies demonstrated real variation in local heat transfer, and Iglesias et al. showed how that variation shows up in residual stress (Monnoyer & Lochegnies, 2008; Iglesias et al., 2022). Maintain the cooling system so the intended jet configuration and airflow distribution are preserved, and investigate spatial cooling non-uniformity whenever you see abnormal stress patterns or breakage.
• Treat edge preparation as strength control, not cosmetic finishing. Vandebroek et al. showed flaw depth and geometry drive edge strength, and Bukieda et al. showed grinding and polishing parameters directly affect measured strength (Vandebroek et al., 2012; Bukieda et al., 2020). Establish reproducible edge-processing conditions and monitor edge quality rather than relying on visual inspection alone.
• Inspect the chamfer and transition regions specifically. Bukieda et al. identified the chamfer and the edge/chamfer transition as important fracture-origin areas (Bukieda et al., 2020). When diagnosing edge-related breakage, check these zones directly rather than just the nominal edge face.
• Investigate the fracture origin before assuming a furnace fault. Chang and Chou demonstrated that internal inclusions can initiate fracture within the tensile region (Chang & Chou, 1982). When breakage is unexplained, determine whether the origin is an internal defect, an edge flaw, or a process-related stress condition before changing furnace settings.
• Verify residual stress with a measurement method appropriate to the product. ASTM C1279-23 provides the non-destructive photoelastic framework for edge and surface stress (ASTM C1279-23), but Efferz et al. showed that architectural geometry, particularly chamfers, complicates edge-stress evaluation (Efferz et al., 2024). Interpret edge measurements with the limitations of the method in mind.
• Don’t treat one nominal stress reading as the full picture. The cooling-field studies show the underlying heat-transfer field can be heterogeneous, and the modelling studies show this produces spatially varying residual stress (Monnoyer & Lochegnies, 2008; Iglesias et al., 2022). When investigating repeated breakage, evaluate the spatial uniformity of the process rather than relying on one nominal stress value.

Diagnostic Checklist: Finding the Real Cause of a Breakage Event

Use this order when investigating an unexplained tempering failure, rather than starting with furnace temperature:

• Locate the fracture origin first. Edge origin and center/internal origin point to completely different root causes.
• If the origin is at the edge: check recent changes to the grinding or polishing wheel, inspect the chamfer and transition zone specifically (not just the edge face), and check whether flaw depth has increased versus your normal baseline.
• If the origin is internal and away from the edge: suspect an inclusion or stone in the raw glass rather than a process setting.
• If breakage is inconsistent across a batch rather than isolated to one flaw type: suspect non-uniform cooling. Check for blocked or degraded jets, uneven nozzle spacing, or airflow drift in the quench section before touching temperature settings.
• If you recently changed glass thickness without revalidating the recipe: treat this as the first suspect. A recipe proven on one thickness does not transfer automatically to another.
• Don’t rely on a single nominal stress reading to close the investigation. A pane can show an acceptable average stress and still have a dangerous local defect or a chamfer that isn’t being measured correctly.

The Bottom Line

Tempering reliability isn’t controlled by one variable. It’s controlled by the full chain: glass quality, heating uniformity, cooling uniformity, residual-stress distribution, edge quality, and how you verify the result. When something breaks unexpectedly, the fracture origin tells you more than the furnace log does.

This article is an original synthesis of the cited research. No source-paper paragraphs, figures, or table layouts have been reproduced. Numerical values reported from individual studies are identified as study-specific results.

---

References

ASTM International. (2023). ASTM C1279-23 — Standard Test Method for Non-Destructive Photoelastic Measurement of Edge and Surface Stresses in Annealed, Heat-Strengthened, and Fully Tempered Flat Glass.

ASTM International. (2025). ASTM C1048-25 — Standard Specification for Heat-Strengthened and Fully Tempered Flat Glass.

Bukieda, P., Lohr, K., Meiberg, J., & Weller, B. (2020). Study on the optical quality and strength of glass edges after the grinding and polishing process. Glass Structures & Engineering, 5, 411–428. DOI: 10.1007/s40940-020-00121-x (Open Access, CC BY 4.0)

Chang, J., & Chou, J. (1982). Fracture strength of tempered glass. Journal of Non-Crystalline Solids, 52, 395–404. DOI: 10.1016/0022-3093(82)90314-3

Efferz, L., Schuler, C., & Siebert, G. (2024). Photoelastic measurements and evaluation methods for edge stress in architectural tempered glass. Glass Structures & Engineering, 9, 439–454. DOI: 10.1007/s40940-024-00270-3 (Open Access, CC BY 4.0)

Fan, B.-W., Zhu, K.-Q., Shi, Q., Sun, T., Yuan, N.-Y., & Ding, J.-N. (2016). Effect of glass thickness on temperature gradient and stress distribution during glass tempering. Journal of Non-Crystalline Solids, 437, 72–79. DOI: 10.1016/j.jnoncrysol.2016.01.008

Iglesias, A., Martinez-Agirre, M., Torca, I., Llavori, I., & Esnaola, J. A. (2022). Numerical methodology based on fluid-structure interaction to predict the residual stress distribution in glass tempering considering non-uniform cooling. Computers & Structures, 264, 106757. DOI: 10.1016/j.compstruc.2022.106757

Monnoyer, F., & Lochegnies, D. (2008). Heat transfer and flow characteristics of the cooling system of an industrial glass tempering unit. Applied Thermal Engineering, 28, 2167–2177. DOI: 10.1016/j.applthermaleng.2007.12.014

Pourmoghaddam, N., Kraus, M. A., Schneider, J., & Siebert, G. (2019). Relationship between strain energy and fracture pattern morphology of thermally tempered glass for the prediction of the 2D macro-scale fragmentation of glass. Glass Structures & Engineering, 4, 257–275. DOI: 10.1007/s40940-018-00091-1 (Open Access, CC BY 4.0)

Stewart, R. J., & Prakash, N. (2023). Modeling dynamic fragmentation of tempered glass. Engineering Fracture Mechanics, 289, 109422. DOI: 10.1016/j.engfracmech.2023.109422

Vandebroek, M., Belis, J., Louter, C., & Van Tendeloo, G. (2012). Experimental validation of edge strength model for glass with polished and cut edge finishing. Engineering Fracture Mechanics, 96, 480–489. DOI: 10.1016/j.engfracmech.2012.08.019