Glass: Types, Properties and Manufacture

Glass is a non-crystalline (amorphous) inorganic solid produced by cooling a molten silicate mixture through its glass-transition temperature (Tg) rapidly enough to prevent crystalline ordering. The defining structural characteristic is the short-range order of the SiO4 tetrahedra that form the backbone of every silicate glass, combined with the absence of long-range periodic order that characterises crystalline minerals. This amorphous structure is not an accident of rapid cooling; it is the thermodynamically metastable state that silicon dioxide and its oxide modifiers occupy when quenched from the melt.

The pure SiO2 network is a three-dimensional random assembly of corner-sharing tetrahedra. Each silicon atom bonds to four oxygen atoms, and each bridging oxygen connects two silicon atoms, creating a continuous random network. The term "network former" applies to oxides that participate in this tetrahedral backbone (SiO2, B2O3, Al2O3 in appropriate concentrations, GeO2). "Network modifiers" are metal oxides (Na2O, K2O, CaO, MgO, PbO, BaO) that introduce sodium, potassium, calcium, or other metal cations into the interstices of the network. Each modifier ion associates with a non-bridging oxygen, effectively breaking a Si-O-Si linkage and reducing the network connectivity. This reduction in connectivity is what makes glass workable: pure SiO2 melts above 1700 degrees Celsius; adding 15% Na2O drops the working temperature to around 900 degrees Celsius, which is the range accessible to commercial furnaces.

The Tg of a glass composition marks the temperature below which the melt is kinetically frozen. Above Tg, the glass behaves as a very viscous liquid on laboratory timescales; below Tg, viscosity is so high that structural relaxation is effectively impossible. For soda-lime glass (the dominant window and container class), Tg is approximately 530-560 degrees Celsius. For borosilicate glass, the higher B2O3 content and lower modifier loading push Tg to approximately 820 degrees Celsius. These Tg differences matter forensically because they determine whether a glass sample deforms during thermal events (fires, vehicle fires) in ways that help date and sequence the damage.

The density and refractive index of a glass composition are both determined by the atomic packing density and the polarisabilities of the constituent ions. High-density ions (lead, barium) produce high-RI, high-density glass. Light network formers (boron) produce low-density, low-RI glass. This direct link between composition and physical properties is what gives elemental analysis, RI measurement, and density measurement their discriminating power in glass comparison casework.

Soda-lime float glass accounts for roughly 90% of all flat glass produced globally. Its near-ubiquity in windows, mirrors, picture frames, tabletops, and automotive side and rear glass makes it the most frequently encountered glass class in forensic laboratories. Understanding its manufacture is not merely academic: the float process introduces compositional gradients and surface characteristics that affect physical-property measurements.

The Pilkington float process, commercialised from 1959 by Alastair Pilkington at Pilkington Brothers Ltd in the UK, produces flat glass by pouring a continuous ribbon of molten glass onto a bath of molten tin at approximately 1000 degrees Celsius. The glass spreads under gravity to a naturally equilibrium thickness of about 6-7 mm, floating on the denser tin surface. The glass ribbon is then drawn forward through a controlled-temperature gradient, cooling from the molten state through the annealing range (where residual internal stresses are relieved) and finally to room temperature. The result is a ribbon of glass with two optically flat surfaces: the tin-side surface (bottom), which picks up a very thin layer of diffused tin, and the air-side surface (top), which remains compositionally pure.

The tin-surface layer has forensic significance. UV fluorescence testing (using a UV lamp at 254 nm) distinguishes the tin side (which fluoresces bluish-white due to tin oxide diffusion into the surface) from the air side (which does not fluoresce). This allows the analyst to determine which face of a pane fragment faced downward during manufacture, which may help orient a broken window fragment within a larger fracture pattern.

The bulk composition of standard architectural soda-lime float glass is approximately: SiO2 70-74%, Na2O 12-14%, CaO 8-12%, MgO 1-4%, Al2O3 0.5-2%, K2O 0.01-0.5%, and trace quantities of Fe2O3 (responsible for the characteristic green tint in thick sections), SO3, TiO2, and other oxides introduced by raw-material impurities. The exact ratios vary between manufacturers, between batches from the same furnace, and between furnace campaigns (the period between furnace relining). LA-ICP-MS can detect batch-level and furnace-campaign-level variation in elements such as Ba, Sr, Zr, Mn, Ti, and the rare-earth elements that are present at parts-per-million levels but vary reproducibly between manufacturing lots.

The refractive index (RI) of standard soda-lime float glass at the sodium D-line (589 nm) lies in the range 1.510-1.520. The density is 2.48-2.52 g/cm3. These are narrow ranges relative to the between-class differences discussed below, but within the soda-lime class the GRIM-3 instrument can resolve differences of 0.0001 RI units, making within-class discrimination possible when fragment populations are large enough to support statistical comparison.

Borosilicate glass is a silicate glass in which B2O3 replaces a significant fraction (12-15%) of the Na2O present in soda-lime glass. The substitution profoundly changes two properties. First, the coefficient of thermal expansion drops to approximately 3.3 x 10-6 per degree Celsius for standard borosilicate (versus 8-9 x 10-6 for soda-lime glass), giving the material its defining characteristic of thermal shock resistance. Second, the density drops to 2.23-2.26 g/cm3 and the RI drops to approximately 1.47-1.48 at the sodium D-line: both values are clearly distinguishable from soda-lime glass.

The Schott company introduced the trade name Duran (and later Pyrex was licensed to Corning) for borosilicate laboratory glassware in the late nineteenth century. Today, borosilicate appears in laboratory flasks, pharmaceutical packaging, optical components, kitchen cookware (Pyrex baking dishes), high-borosilicate glass for sodium-borosilicate-sealed beam headlamp envelopes, and specialist applications in fibre optics and telescope mirror blanks.

Forensic submissions of borosilicate glass arise in several contexts: break-ins involving laboratory glassware theft (drug-manufacturing casework), road-traffic debris from broken headlamp covers, fires in chemistry laboratories, and pharmaceutical counterfeiting involving the substitution of lower-quality glass vials for Type I borosilicate pharmaceutical containers. The RI and density values are so distinct from soda-lime glass that the identification can be made in a single GRIM-3 run; the confirmation comes from the elemental boron content, which LA-ICP-MS detects immediately.

The A-type alkali-barium-silicate glass used in traditional cathode-ray tube envelopes contains high concentrations of BaO and some SrO, with essentially no boron. This composition falls outside both soda-lime and borosilicate categories and was encountered in forensic submissions involving older television or computer monitor damage until CRT displays were largely displaced by flat panels in the 2000s.

Tempered (toughened) glass is produced by taking an annealed glass panel and subjecting it to a controlled thermal treatment: heating to approximately 620-650 degrees Celsius (above the soda-lime Tg of 530-560 °C but below the softening point at around 720 °C, so the glass is mobile enough for stress redistribution without losing dimensional stability) and then rapidly quenching the surfaces with cold air jets while the core remains hot. When the surfaces cool and contract before the core, they establish residual compressive stress at the surfaces and residual tensile stress in the core interior. The result is glass that is approximately four to five times stronger in bending than annealed glass of the same thickness, because any surface crack must first overcome the compressive skin before it can propagate.

The characteristic forensic signature of tempered glass is its failure mode. When the tensile core stress is released by a crack that penetrates through the compressed surface layer, the stored elastic energy shatters the entire panel into small, roughly cuboid fragments (called "dicing" or "cuboid fragmentation") with no large sharp shards. The size of these dice is inversely related to the level of temper: higher surface compression, finer dice. Standard automotive tempered side-door glass produces particles approximately 3-10 mm in their longest dimension, without the long blades seen in annealed glass fracture.

Tempered glass appears in automotive side and rear windows (post-1970s mandated safety glazing in the US under FMVSS 205, in the EU under ECE Regulation No. 43, and in India under IS 2553 Part 2 and the Central Motor Vehicles Rules Schedule X), shower screens, toughened glass doors and partitions, and some furniture glass. The characteristic dicing pattern distinguishes it from laminated windshield glass (which holds together) and from annealed glass (which fractures into large irregular shards). A crime-scene analyst observing cuboid glass fragments at a vehicle impact site can immediately narrow the likely glass source to a side or rear window.

The composition of tempered glass is identical to the base annealed glass from which it is made; tempering is a stress treatment, not a compositional change. RI and density values are therefore the same as for the base soda-lime glass, and LA-ICP-MS composition will match the same float-glass manufacturing lot. The distinguishing feature is mechanical: the fragment geometry and the absence of large shards.

The RCMP forensic laboratories in Canada, the FBI Laboratory in the US, and the UK Home Office forensic laboratories have all established reference collections of tempered and laminated glass fragment populations to support automotive hit-and-run casework.

Laminated glass consists of two or more glass plies bonded together by an interlayer of polyvinyl butyral (PVB), thermoplastic polyurethane (TPU), or ionoplast (SentryGlas, a DuPont product). Standard automotive windshields are two plies of 2.0-2.5 mm soda-lime glass around a 0.38 mm or 0.76 mm PVB interlayer. The assembly is produced by sandwiching the PVB film between the glass plies and autoclave-processing at approximately 140 degrees Celsius and 12 bar pressure to produce the optically clear laminate.

The PVB interlayer serves two forensic-casework-relevant functions. First, it prevents the glass from separating into flying shards on impact: the craze pattern (a network of radial and concentric cracks limited by the interlayer adhesion) is retained in the laminate. This makes forensic reconstruction of the fracture pattern more straightforward than with tempered or annealed glass, because the crack geometry is preserved in situ. The fracture pattern analysis that allows direction-of-force and sequence-of-impact reconstruction (covered in depth in the companion fragmentation topic) is most directly applicable to laminated windshields.

Second, the PVB interlayer retains glass fragment populations against the interlayer surface even after the pane is broken. This matters for evidence collection: fine glass fragments, body cells from the occupant, blood, hair, and fibre trace evidence may all be deposited on and in the crazed interlayer. A shattered windshield from a hit-and-run vehicle should be collected and packaged intact, with the interlayer preserved, to recover secondary trace evidence from the surface.

Compositionally, windshield glass is soda-lime float glass, so its RI, density, and elemental composition fall within the soda-lime range. LA-ICP-MS can often distinguish between the inner and outer plies of a windshield because small compositional differences arise from separate float-glass manufacturing batches. The PVB interlayer itself is recoverable and chemically characterisable by FTIR, which is relevant when a disputed interlayer needs to be sourced to a specific manufacturer or batch.

Bullet-resistant glazing adds further layers. Standard bullet-resistant glass for bank counter use (typically rated to BS EN 1063 SR-B2 or UL 752 Level 3 in the US) may have four to six plies of glass totalling 19-38 mm, with PVB interlayers between each ply and a polycarbonate backing layer to prevent spalling on the interior side. The polycarbonate component is distinguishable from glass by its density (1.20 g/cm3 versus 2.5 g/cm3 for glass) and by FTIR, which gives the characteristic carbonate ester spectrum.

Lead crystal glass contains PbO in concentrations from 24% (traditional English full lead crystal, as defined under EU Regulation EC 69/2001 for the crystal glassware designation) to 32% or above in some optical-grade compositions. The substitution of lead oxide for calcium oxide in a soda-lead-silicate base glass raises both the density (2.90-3.10 g/cm3 for full lead crystal versus 2.50 for soda-lime) and the refractive index (1.56-1.58 at the sodium D-line versus 1.515 for soda-lime). The high dispersion (Abbe number around 29-30 versus 66 for soda-lime) gives lead crystal its characteristic brilliance and rainbow play. Forensic submissions involving lead crystal arise in domestic violence and assault cases where decorative items are the weapon, and the high RI value immediately distinguishes a fragment from window or bottle glass.

Optical glass is a precisely formulated category with tightly controlled optical constants. Crown glass (low RI, low dispersion, Abbe number 50-65) and flint glass (high RI, high dispersion) are the classical categories; modern catalogues from Schott, CDGM, and Ohara list hundreds of optical glass types with specified RI values to four decimal places and specified dispersion values. Forensic submissions involving optical glass arise in camera and lens theft, scope and binocular damage in hunting and shooting casework, and spectrometer component analysis in laboratory equipment theft.

Coloured glasses are soda-lime or borosilicate base compositions with added colorant oxides: CoO for blue, Cu2O for ruby-red, Fe2O3 / Cr2O3 for green, Mn3O4 for purple. These colorant concentrations are detectable by LA-ICP-MS and can help narrow the source of a coloured glass fragment (blue-green cobalt glass from a specific pharmaceutical bottle, for instance) beyond what RI and density alone can provide.

The elemental composition of glass beyond the major oxides (SiO2, Na2O, CaO, Al2O3, MgO) contains a set of trace and minor elements whose concentrations are determined by raw material sourcing (sand quarry, soda ash origin, limestone provenance), furnace condition (refractories contribute trace metals), and intentional additions (colorants, decolorisers, fining agents). These elements carry manufacturing-lot fingerprints that persist in the glass indefinitely.

The elements with the highest discriminating power, as established by research from the FBI Laboratory (Almirall and Trejos, multiple publications 2006-2020), the ENFSI Glass Working Group, the University of Central Florida Chemistry Department, and the RCMP Chemistry Section, include: Ba, Sr, Ti, Zr, Mn, Fe, Mg, Al, Ca, Na, K at the major/minor level, and rare-earth elements (La, Ce, Nd, Sm), heavy metals (Pb, Bi, Sb), and transition metals (Cr, Co, Ni, Cu, Zn) at the trace level.

The ASTM E2927 standard for LA-ICP-MS glass analysis specifies a minimum element suite that provides the statistical power to discriminate between glass from different manufacturing sources. The multivariate statistical framework (Hotelling's T2 test for within-batch homogeneity assessment, Mahalanobis distance for between-source discrimination) applied to this element suite can resolve glass fragments from different float-glass batches produced at the same plant, even when their RI and density values are indistinguishable.

A critical forensic constraint is within-pane and between-pane variation. Glass from a single float-glass ribbon is not perfectly homogeneous; RI and elemental composition vary slightly across the width and along the length of the ribbon. Understanding this within-source variation (by analysing multiple locations on a reference pane) is essential before claiming that two fragments are inconsistent with a common origin. The ENFSI ENG3-2013 guideline requires the analyst to document within-pane variation when submitting a glass comparison opinion.

In India, the CFSL New Delhi and Hyderabad laboratories, along with several State FSLs, have adopted RI measurement by GRIM as standard; LA-ICP-MS capability is concentrated at the CFSL Chandigarh and New Delhi laboratories. In the US, the FBI and the ATF forensic laboratories, along with several state crime laboratories certified under the ASCLD/LAB and OSAC frameworks, routinely conduct LA-ICP-MS glass comparison. In the UK, the Forensic Science Service (now incorporated into private providers including Eurofins Forensic Services and Key Forensic Services after the FSS closure in 2012) and in Northern Ireland the FSNI, maintain glass comparison capability under the UKAS accreditation framework. ENFSI member laboratories across Germany (BKA), Netherlands (NFI), Spain (INTCF), and Sweden (NFC) apply the ENG3 guideline.

Glass evidence collection and documentation follow established protocols that differ somewhat between jurisdictions but converge on the same core requirements. The ENFSI ENG3-2013 guideline, the SWGMAT glass guidelines (the US equivalent, developed under the FBI and ATF), and the FSNI and Eurofins UK protocols all require documented sample number, scene location, collection method, packaging material, and chain-of-custody transfer record for each glass fragment population.

Physical collection of fine glass fragments at scenes requires forceps or tape-lift collection, not vacuuming, because vacuum collection mixes fragment populations from different objects. Packaging uses rigid-walled containers (not plastic bags, which allow fragment migration and create static-electricity contamination issues). Reference glass from the putative source (a known pane, a recovered broken container) is collected separately from scene fragments, and both populations are packaged, labelled, and forwarded under independent chain-of-custody documentation.

The distinction between the questioned sample (Q, from the suspect or crime scene) and the known/reference sample (K, from the putative source object or location) is the fundamental design of any glass comparison. In hit-and-run cases, Q comes from the suspect vehicle and K from the victim's vehicle glass or from the scene debris field. In break-and-enter cases, Q comes from glass found on the suspect's clothing or tool, and K comes from the broken window. This Q/K design mirrors the comparison framework used in all trace-evidence disciplines and determines whether the statistical result supports association or exclusion.