Electronics Guide

NFPA 921 and the Governing Framework

The National Fire Protection Association established a Technical Committee on Fire Investigations in 1985, and the committee produced the first edition of NFPA 921 in 1992. Before that, fire investigation was largely an apprenticeship trade in which rules of thumb passed from experienced investigators to novices without experimental validation, and many of those rules turned out to be wrong. Crazed glass was widely taught as an indicator of rapid heating by an accelerant; controlled testing showed it results from rapid cooling, typically when hose streams strike hot glass. The guide's principal work since has been to replace such lore with statements experiment supports.

NFPA 921 is revised on a roughly three-year cycle, and investigators should work from the current edition, since guidance on specific indicators has changed materially between revisions. The 2021 edition rewrote the fire effects and fire patterns chapter around observable or measurable changes to materials, organized into discoloration, deformation, deposition, and mass loss, each with an explicit statement of its limitations. The same edition reclassified arc mapping as a form of fire pattern rather than an independent source of origin information, emphasized the probabilistic nature of arcing, and reinforced the warning that arc sites are not necessarily located at the area of origin. A 2024 edition followed, and the association has issued tentative interim amendments against it, a reminder that citing an edition year is part of citing the guide at all.

NFPA 921 is designated a guide rather than a standard, meaning its provisions are recommendations rather than requirements. That designation is misread in both directions. It does not mean an investigator may disregard the guide at will, since courts in the United States have repeatedly treated it as a peer-reviewed benchmark for reliable methodology and expect a departure to be explained. Nor does it settle every technical question; it documents the state of knowledge, including the many places where knowledge is thin. The companion qualification document is NFPA 1033, Standard for Professional Qualifications for Fire Investigator, which states minimum job performance requirements for investigators in both the public and private sectors. Where NFPA 921 describes the method, NFPA 1033 describes the knowledge an investigator must hold before applying it.

The Scientific Method Applied to a Fire Scene

NFPA 921 frames origin-and-cause work as the scientific method applied to a specific physical event: recognize the need, define the problem, collect data, analyze it, develop hypotheses, test each one, and select a final hypothesis only if it survives. Two properties matter more than the steps. Hypotheses are formed after data collection rather than before, which is why premature theories corrupt the process. And testing is deductive: the investigator asks what else must be true if the hypothesis holds, then looks for that consequence in the physical record. Cognitive testing compares the hypothesis against the collected data and established fire science; experimental testing produces new data. The two carry different evidentiary weight, and an investigator who has done only the former should say so.

Expectation Bias and the Negative Corpus Problem

NFPA 921 devotes explicit attention to cognitive bias, which is unusual among engineering documents and reflects hard experience. Expectation bias occurs when an investigator reaches a conclusion before collecting sufficient data and then interprets later observations to fit it; confirmation bias reads ambiguous observations as supporting the favored hypothesis. In electrical work the signature is characteristic: an investigator who arrives believing a particular appliance failed will find arc damage on that appliance, because in a serious fire the appliance cord will very likely show arc damage regardless of what started the fire. The defenses are to form multiple hypotheses, document observations before interpreting them, and expose the reasoning to someone who has not worked the scene.

The most consequential methodological correction in the guide concerns the practice known as negative corpus, in which an investigator names an ignition source by eliminating the alternatives without producing affirmative physical evidence for the source asserted. The classic form runs: the area of origin contained a receptacle, a lamp, and a candle; the candle and lamp are excluded; therefore the receptacle caused the fire. NFPA 921 rejects this as inconsistent with the scientific method, on the ground that it generates a hypothesis nothing can test. The correction was carried through in the 2011 edition and reinforced since.

The distinction that matters is between elimination as a tool and elimination as a substitute for evidence. Eliminating candidate sources is legitimate where the evidence supports elimination; treating the residue of that elimination as proof is not. If the physical evidence for the remaining source is absent, the correct classification is undetermined, a professional finding rather than a failure. The point applies with force to electronics: a scorched switching supply in the area of origin is equally consistent with having caused the fire and with having been destroyed by it, and separating the two requires evidence internal to the supply.

Origin Before Cause

The most common structural error in fire investigation is reversing the order of origin and cause. NFPA 921 places origin first for a straightforward reason: the cause is the ignition source, the first fuel ignited, and the circumstance that brought them together, and all three lie in the area of origin. An investigator who identifies a suspected source first and then reasons backward to place the origin around it has abandoned the constraint that makes the analysis testable, because any candidate source can be surrounded by an origin area drawn to suit.

Origin determination draws on four categories of information, treated as complementary rather than ranked: witness observations, fire patterns, arc mapping evidence read as a form of fire pattern, and fire dynamics analysis. Where these agree, the origin area can be stated with confidence; where they conflict, the conflict is itself data and must be reported rather than resolved by preference. Origin is properly expressed as an area, no smaller than the evidence supports. In a fire extinguished early, narrowing it to a specific appliance is sometimes achievable, but in one that reached full-room involvement a stated point of origin is a claim the evidence rarely supports.

Fire Patterns, Heat and Flame Vectors, and Their Limits

Fire patterns are the visible or measurable effects of the fire on materials, and pattern analysis is the traditional core of origin determination. The physical basis is real: a fire grows from a small area, plumes rise, and the surfaces nearest and longest exposed to the plume sustain the greatest damage. Vector analysis formalizes this by drawing arrows on a scene diagram for the direction of heat and flame movement inferred from each pattern, then looking for convergence.

Two quantitative surveys support the vectors. Char depth surveys measure the depth of charring in wood at a grid of points, using a blunt probe with consistent pressure, and compare readings relatively across the grid. Calcination surveys do the equivalent for gypsum wallboard, where the depth to which the gypsum has lost its chemically bound water is assessed by probe resistance or by sectioning. Both compare exposure severity from point to point within one scene, and neither yields burn duration, because the charring rate of wood depends strongly on heat flux, species, density, moisture content, and geometry, and rates from standard furnace exposures do not transfer to compartment fires. A report that converts depth of char into elapsed time has overstated the measurement.

What Full-Room Involvement Destroys

Pattern analysis degrades sharply once a compartment reaches flashover and burns beyond it. In the post-flashover regime the whole compartment is at flame temperature, ventilation rather than fuel geometry controls where burning is most intense, and new patterns overwrite the ones produced during growth. Ventilation-generated patterns can appear near doorways and windows nowhere near the origin, and can be more pronounced than the origin pattern. Extended burning also produces clean burn and full-height damage over broad areas, erasing the boundaries that make vector analysis possible.

The magnitude of the problem was demonstrated in training exercises reported by Steven Carman in 2008. Fifty-three experienced fire investigators from the public and private sectors, who had not observed the burns, examined post-flashover compartments and were asked to identify the quadrant in which each fire had started. About five and seven-tenths percent selected the correct quadrant in each cell. Critics note real limitations: participants could not conduct a full investigation or move contents, judged visually from the doorway, and the study reported no demographics and applied no statistical rigor. Even discounted, the result stands as a caution, and Carman attributed it to incomplete professional understanding of post-flashover fire behavior. The consequence for electrical work is that in a severely burned structure, patterns may support only a room-level origin.

Arc Mapping and What It Can Support

Arc mapping, also called arc surveying or arcing fault pattern analysis, is the systematic evaluation of a structure's electrical circuits combined with the identification and plotting of arc sites, undertaken to help identify the area of origin and analyze the fire's spread. The method entered NFPA 921 in the 2001 edition as a fourth source of origin information, was expanded through the 2004, 2008, and 2011 editions, received its first formal definition in 2014, and was reclassified as a fire pattern in 2021.

The Underlying Hypotheses

The method rests on hypotheses worth stating explicitly, because in practice they are usually left implicit. Marcel Delplace and Emiel Vos, whose 1983 paper in Fire Technology is the usual starting point for the modern technique, proposed two: that the pattern of short circuits parallels the pattern of combustion damage, and that a short circuit normally occurs wherever the fire first damages a cable. An earlier Japanese formulation by Miyake, published in 1975, added a directional claim: where two sever points appear on one cable, the point farther from the supply was damaged first, since after the near point severs the far segment carries no current. A 1955 text by Straeter and Crawford already warned investigators to determine whether the short caused the fire or the fire caused the short.

The directional claim is the strongest part of the method and rests on unimpeachable physics. Current cannot flow past a severed conductor, and cannot flow at all once the protective device opens. If a circuit shows a sever or weld arc at one location and a further arc site downstream, the downstream arc must have occurred first, and the local direction of fire progression follows. That inference is deductive rather than statistical, and it is why arc mapping earned a place in the guide at all.

What Full-Scale Testing Found

The experimental record is less supportive of the broader spatial claim. Larry West and David Reiter reported three full-scale room fire tests in 2005 in which twelve to fifteen branch circuits of 14/2 nonmetallic-sheathed cable ran on the ceiling under direct fire exposure. In all three, arcing occurred near the area of origin but also remote from it, in some cases at the opposite side of the room, and the authors observed that arcing was strongly influenced by entrained air velocity, that is, by ventilation rather than proximity to origin. Arc bead size varied from very large to barely perceptible even though every circuit had the same protective device and an available short-circuit current of roughly three hundred ten to three hundred twenty amperes, indicating that bead size is probabilistic rather than diagnostic. Christopher Wood and Andrew Kimball, running four tests in a furnished bedroom, found arc artifacts typically one to four millimeters across and concluded that no arc mapping pattern clearly indicated a point of origin.

The largest study is the 2009 doctoral dissertation of Nicholas Carey at the University of Strathclyde, comprising thirty-nine compartment burns. Carey concluded that there is a high probability of arcing damage occurring to conductors close to the area of origin. His data support that, but it does not establish the discriminating claim that matters, namely that arcing is sparse away from the origin. He also reported that twenty-nine percent of arc sites involved severing and seven percent welding, figures that bound how often the directional inference is available at all. Most of these studies ran cables exposed on ceilings; wiring behind gypsum wallboard fails later and more uniformly, so testing with unprotected cable is biased toward positive findings. Vytenis Babrauskas, reviewing the literature in 2017, argued that the probability of a branch circuit exhibiting the conditions permitting a valid directional conclusion is under one percent, and urged NFPA to subsume arc mapping under fire patterns; the 2021 edition did so.

The position that survives is narrow but genuinely useful. Plotting arc sites on an accurate circuit diagram is uncontroversial as documentation, and conclusions about local direction of fire travel along a circuit are supportable where a sever or weld arc has a further arc downstream of it. Conclusions that a concentration of arc beads marks the area of origin are not, because arc concentration tracks fuel load, ventilation, and burn duration together, and only the last relates to origin. Absence of arcing is likewise weak evidence, since a circuit may have been de-energized by an upstream sever or a tripped breaker before the fire reached it.

Arcing Through Char

A related phenomenon bears on both origin and cause. When polymeric insulation is heated it chars, and the char is electrically semiconductive. Polyvinyl chloride, which insulates a large share of low-voltage building wiring, chars on the order of two hundred to three hundred degrees Celsius and resists the resulting leakage poorly. Once a carbonized path bridges two conductors, sustained arcing across it is easy to establish at one hundred twenty volts. This is arcing through char, the mechanism by which most fire-caused arcing occurs. Because the path has appreciable resistance, the fault current is much lower than the available bolted short-circuit current. Laboratory work by Frank Ablenas and Steven Bodzay for the Canadian Electrical Association found that lower fault currents tend to produce larger arc beads, inverting the intuition that a large bead indicates a violent fault. The low, intermittent current also means the overcurrent device may not operate: Yasuaki Hagimoto and colleagues measured occasional peaks up to about two hundred fifty amperes but typical peaks no greater than fifty.

Arcing through char is also a cause mechanism in its own right. Susumu Nagata and Shozo Yokoi found that virgin PVC heated to only one hundred sixty degrees Celsius would ignite when one hundred volts was impressed across one millimeter of insulation thickness, and that insulation previously preheated to two hundred to three hundred degrees Celsius could be ignited at temperatures as low as forty degrees Celsius during the voltage test. An episode of overheating that does not itself start a fire can therefore leave insulation permanently predisposed to arc tracking, with the fire following much later at ordinary temperatures. No common polymer is entirely immune, and arc tracking has been a persistent problem in automotive wiring at voltages far below those the standard tracking classifications contemplate.

Cause Versus Consequence in Electrical Artifacts

The recurring problem in electrical fire investigation is that the artifact produced by an arc that started a fire and the artifact produced by an arc caused by a fire are, in general, the same artifact. The literature calls these cause beads and victim beads, and separating them has attracted more research than any other metallurgical question in the field.

The first and easier distinction is between melting caused by an arc and melting caused by fire exposure. Arc melting is localized and abrupt, producing a rounded bead or a notch with a sharply defined boundary between melted and unmelted metal, often with expelled spatter, and leaving the adjacent conductor undamaged. Fire melting is gradual and diffuse, producing tapering, thinning, and drooping over a length of conductor. Copper melts at one thousand eighty-five degrees Celsius and aluminum at about six hundred sixty, and post-flashover fires can approach or exceed the copper figure, so fire melting of building conductors is evidence of nothing electrical. The distinction is not as robust as its frequent use implies, since an arc bead later subjected to prolonged fire exposure may lose the features that identify it, and NFPA 921 is explicit that visual examination of a melted conductor cannot by itself establish whether the arcing caused the fire or the fire caused the arcing.

Why the Cause-Versus-Victim Question Has Resisted Solution

Numerous techniques have been proposed to separate the two populations: internal void and gas porosity distribution, dendrite size and orientation in the resolidified metal, grain structure at the bead boundary, the ratio of cuprous to cupric oxide, surface morphology, carbon content, and elemental analysis for atmospheric constituents, on the theory that a cause bead forms in relatively clean air while a victim bead forms in a smoke-laden, oxygen-depleted atmosphere. Babrauskas reviewed this work in a 2004 paper in the Journal of Fire Protection Engineering and reported that none of the proposals had yielded a reliable technique; later reviews have not overturned that assessment. Local atmosphere, current, and cooling rate at an arc site overlap between the two populations, so the measured distributions overlap as well. A metallurgical examination can establish that a bead is an arc bead rather than fire melting, but it cannot, on present knowledge, establish that a given bead caused the fire.

Alloying and Other Confounders

Several additional effects mimic electrical damage. Eutectic alloying occurs when a conductor contacts a dissimilar metal well below either metal's melting point, and copper in contact with aluminum, zinc, or tin can form low-melting alloys that produce localized melting and pitting resembling arc damage. Molten metal dripping from an overhead source can freeze on a conductor, producing a bead-like feature with no electrical origin. Mechanical damage during overhaul severs and deforms conductors after any electrical event, which is why early scene examination and documentation are worth more than any laboratory technique applied afterward.

Fire-caused arcing is also easier than intuition suggests, because flames reduce the dielectric strength of air. Air breaks down at roughly three megavolts per meter under ordinary conditions, but measurements by Mesina indicate the value falls to about one tenth of that within flames, although the work examined conditions at sixteen hundred volts and above. Ionized gas ejected from one arcing fault can also initiate breakdown at a second, separate location. Arcing at a given point is therefore weak evidence that anything electrical happened there first.

Ignition Mechanisms in Wiring and Wiring Devices

If arc artifacts do not identify a cause, what does? The answer is mechanism-specific evidence: physical signatures characteristic of a particular failure process, corroborated by the installation configuration and by whether the mechanism is competent to ignite the fuel present. There are only a few ways electrical energy ignites a structure, and they leave different traces.

Parallel and Series Arcing

A parallel arc occurs between conductors, or between a conductor and ground, with the load not in series with the fault. Fault current is limited by source impedance, so the overcurrent device can respond, and in the bolted case it does. It is in fact very difficult to ignite building materials with a bolted short in a properly protected branch circuit. Jean-Pierre Béland hammered cables, armored cables, and conduit until the breaker operated and could never ignite wood, although loose wood fiberboard fibers did ignite in some cases. Kinoshita and colleagues did ignite cotton gauze with bolted shorts, but only with a thermal-only twenty-ampere breaker; with a thermal-magnetic breaker of the same rating, no ignitions occurred.

The ignitable variant is the arcing short with ejected molten metal. When conductors make momentary contact, magnetic forces push them apart and the liquid bridge breaks, throwing molten copper. Franklin reported readily starting fires in blankets and paper by cutting an energized cord with diagonal cutters, and produced as many as thirty successive short circuits on one cord before a twenty-ampere breaker tripped. Underwriters Laboratories has used a guillotine test with cheesecloth as the target to simulate the sawing accident. The forensic signature is a cut or crushed conductor with arc damage at the point of mechanical insult, combined with low-thermal-inertia fuel nearby.

A series arc occurs in line with the load, typically at a broken conductor or a failing connection. Because the load remains in series, the fault reduces circuit current rather than increasing it, and no overcurrent device will respond. Series arcing is therefore invisible to conventional protection and is one of the mechanisms arc-fault circuit interrupters were introduced to address, with the National Electrical Code progressively extending AFCI requirements for dwelling-unit branch circuits under Article 210.12. Whether an AFCI or a ground-fault circuit interrupter was installed, and whether it operated, is itself evidence worth recovering.

Overload, Thermal Confinement, and Stray Currents

Gross overload is the mechanism the public most readily assumes and the one that least often applies. Experimental studies indicate that currents on the order of three to seven times the rated load are required to ignite cable insulation. Since North American branch circuits are protected at fifteen or twenty amperes and the smallest cords in general use are 18 AWG rated at ten amperes, that overload is normally interrupted long before ignition. Gross overload is accordingly a rare cause of fire in protected branch-circuit wiring, and an investigator asserting it should explain why the protective device did not operate.

Excess thermal insulation is a far more plausible variant of the same physics, because a conductor carrying rated or even sub-rated current can reach ignition temperatures if its heat cannot escape. Laboratory demonstrations have produced ignition from a cord coiled a few times and covered with cloth. The historical case is knob-and-tube wiring, in which conductors strung on porcelain standoffs depend on free air circulation for their ampacity; burying such wiring in later-added thermal insulation has caused fires. The signature is generalized, long-length insulation degradation rather than a localized fault, together with a configuration that explains the confinement. Stray currents form a related category: Kinoshita and colleagues documented ignition requiring only about five amperes where a three-conductor PVC-insulated cable contacted a galvanized iron roof. Such cases are identified by tracing the unintended current path, not by examining a single artifact.

Poor Connections and Glowing Connections

A loose or oxidized connection is the mechanism with the most distinctive physics and the most substantial supporting literature. The process is a positive feedback loop: contact resistance produces localized heating, heating accelerates oxidation and metal creep, the connection loosens further, and resistance rises again. Past a threshold it enters a self-sustaining glowing state. The mechanism in copper is oxide growth. Hotta identified by X-ray analysis that the high resistance in a degrading copper-to-copper connection comes from progressive formation of cuprous oxide at the junction, and measured roughly fifteen watts dissipated at one ampere. Kawase described the sequence: after repeated make and break, a cuprous oxide breeding process begins, layers grow on both contacts, and a single bright filament of molten material emerges and meanders along the oxide layer. Jan Sletbak and colleagues measured the filament at twelve hundred to thirteen hundred degrees Celsius, consistent with the melting point of cuprous oxide near twelve hundred thirty.

The currents required are strikingly small. Work dating to 1961 found that four to six amperes were needed to initiate glowing at a freshly made connection, with about fifty watts dissipated in a copper-brass connection at twenty amperes and about thirty-five watts in copper-iron, the power depending on the materials rather than the contact size. William Meese and Richard Beausoleil, studying screw terminals of duplex receptacles, observed visible glow at currents as low as three-tenths of an ampere in a one hundred twenty volt circuit, saw a connection glow continuously for one hundred twenty-nine hours, measured twenty to forty watts dissipated at twenty amperes against a fraction of a watt for a sound connection, and found steel screws considerably more likely than brass to produce glowing. Because the drop across a glowing receptacle connection may be only one or two volts, the condition can persist indefinitely unnoticed.

Whether a glowing connection ignites a structure depends on what is adjacent to it, and here the experimental base is thinner than the mechanism literature. Jesse Aronstein reported igniting low-thermal-inertia furnishings placed directly against a receptacle face from a connection dissipating twenty-eight watts, thermoset cover plates at thirty watts, and wood studs at thirty-five to fifty watts, with combustion beginning as smoldering. Ontario Hydro tests on duplex outlets wired with aluminum conductors and cycled under a fifteen-ampere load ignited wood paneling and cellulose insulation after four cycles with a back-wired connection, while a similar assembly with gypsum wallboard and fiberglass produced no ignition after forty-two cycles. The assembly around the connection is as much a part of the cause as the connection itself.

Two further signatures belong here. Béland found that inadequately tightened twist-on connectors joining copper conductors commonly failed by metal loss several inches away from the connector rather than at it, a corrosion effect in which overheating liberates hydrogen chloride from the PVC insulation, which then attacks the copper. And repairs made by twisting conductors together and taping them were measured by Hijikata and Ogawara at one hundred thirty to three hundred degrees Celsius at twenty amperes. Both are recoverable when the connection survives, and both are strong affirmative evidence in the area of origin.

Ignition Competence and the First Fuel

Naming an ignition source is only half of a cause determination. NFPA 921 requires that the source be competent, meaning it possessed sufficient energy, at a sufficient temperature, for sufficient duration, in the required geometric relationship, to ignite the specific first fuel. Competence analysis is the most common place where an otherwise plausible electrical hypothesis fails. Temperature determines whether ignition is thermodynamically possible, energy and power determine whether enough heat can be delivered, and duration matters because ignition of solids is a heating process rather than an instant. A spark may reach several thousand degrees and ignite nothing, while a thirty-watt glowing connection far below flame temperature will ignite wood given hours of contact.

The fuel's thermal inertia, the product of thermal conductivity, density, and specific heat, governs how readily it responds. Low-thermal-inertia materials such as bedding, drapery, loose paper, and cellulose insulation ignite from modest sources, which is why the Aronstein and Ontario Hydro results depend so heavily on what was placed against the device. Lumber, gypsum board, and thermoset plastics require far more sustained input and typically ignite first in the smoldering mode, which can persist for hours before turning to flaming and bears on reconciling a hypothesis with witness timelines. Geometry completes the analysis: a glowing connection inside a metal enclosure has no path to any fuel, and molten droplets fall, so targets below an arcing fault matter more than those beside it.

Lithium Battery Incidents as a Separate Discipline

Fires involving lithium-ion cells have become a large and growing share of electronics-related thermal incidents, and they demand methods that differ from conventional wiring investigation. A cell in thermal runaway is simultaneously the ignition source, the first fuel, and a pressurized ejection mechanism, and it can propagate to its neighbors and consume the evidence of its own initiation.

Thermal Runaway Signatures

The internal sequence is well characterized. Between roughly ninety and one hundred twenty degrees Celsius the solid electrolyte interphase on the negative electrode begins to break down, generating heat and gas. Separator materials melt near one hundred thirty degrees Celsius. Above about two hundred degrees Celsius the exothermic reactions become rapid, and near two hundred twenty degrees graphite phase breakdown and cathode decomposition proceed in earnest. In work published in 2015, Donal Finegan and colleagues used high-speed synchrotron X-ray tomography to observe commercial 18650 cells with nickel manganese cobalt oxide cathodes during runaway, reporting gas-induced delamination between electrode layers, electrode layer collapse, and copper globules visible as highly attenuating spots, indicating internal temperatures exceeding the melting point of copper at one thousand eighty-five degrees Celsius while the outer layers remained comparatively intact.

That observation carries direct forensic weight, because melted copper current collectors inside a cell indicate internal temperatures characteristic of runaway rather than external fire exposure: the casing would fail long before external fire raised the interior that high. The same study showed how much construction matters, since the cell with an internal support mandrel retained its wound architecture and vented progressively, while the cell without one distorted severely and completed initiation and ejection in under a tenth of a second. The evidentiary yield from a recovered cell therefore depends on its format and construction as well as on the severity of the event.

Identifying the Initiating Cell

In a multi-cell pack the governing question is which cell failed first, since propagation obliterates the distinction between initiator and victim just as fire obliterates the distinction between cause and victim beads. The initiating cell typically shows the most complete internal consumption, with the greatest loss of jellyroll structure and the most extensive current collector melting. Propagation is directional, so the spatial gradient of damage across the pack can be read as a vector much as fire patterns are read in a compartment, and vent gas deposition on adjacent cells and on the enclosure indicates the direction of the initial ejection. External heating produces a damage gradient tied to the exterior of the pack, while an internal initiation produces one centered on a cell.

Computed tomography of the recovered pack before any disassembly is the standard first step, because it preserves spatial relationships that disassembly destroys. A scan can locate deformed, vented, and collapsed cells, reveal foreign particles and separator damage, show whether interconnects failed by melting or mechanically, and identify the most consumed cell, all while every component remains in position. Where the pack has been consumed, computed tomography of individual cells still resolves internal structure that sectioning would destroy. Disassembly should follow a documented plan and be treated as the destructive examination it is.

Internal Defect, External Abuse, or System Fault

The three competing hypotheses carry very different consequences, and distinguishing them is the analytical core of the work. An internal cell defect implicates the cell manufacturer, through metallic particle contamination introduced during electrode manufacture, separator defects, electrode misalignment or burrs, or lithium plating from charging at low temperature or high rate; contamination is detectable by scanning electron microscopy with energy-dispersive X-ray spectroscopy at the initiation site if that site survives. External abuse implicates the user or the environment through crush, penetration, external heating, or immersion, and crush damage leaves directionally consistent deformation that computed tomography will show. Charging-system faults implicate the charger, the protection circuitry, or the battery management system, since overcharge drives lithium plating and eventually dendritic growth while over-discharge dissolves copper current collectors. Separating these requires the cell, the charger, and the control electronics together, which is why recovering the whole system matters at the scene.

The battery management system is often the most informative single artifact, because it holds a record of conditions the physical evidence can no longer show. Its non-volatile memory may retain per-cell voltages, pack current, temperature readings, charge throughput, cycle count, and fault logs. A cell developing an internal soft short usually shows a drifting voltage relative to its siblings before failure, and that drift, if logged, is direct evidence of an internal defect and of when it began, while records of overcharge, over-temperature, or charging outside the permitted window point the analysis elsewhere. Recovery has the usual order-of-volatility problem, since re-energizing a damaged pack can propagate damage or start a second thermal event, and removing a memory device to read it directly is a destructive examination in the sense of ASTM E860. Timestamps deserve skepticism, since a device clock may be unset, drifted, or reset by the power loss under investigation. Where product-level test data exists under UL 9540A, the test method for evaluating thermal runaway fire propagation in battery energy storage systems, it provides a baseline for comparison.

Evidence Preservation, Spoliation, and Notice

The moment a fire may generate a claim against a product manufacturer, an installer, or a utility, the physical evidence acquires a legal status alongside its technical one. Spoliation, the loss, alteration, or destruction of evidence relevant to a proceeding, can support sanctions ranging from an adverse inference instruction to exclusion of expert testimony or dismissal of the claim. The duty to preserve attaches when litigation is reasonably anticipated, often well before suit is filed, and binds the party in possession of the evidence including a first-retained expert.

Two ASTM practices structure the response. ASTM E1188 addresses collection and preservation of information and physical items by a technical investigator and directs that evidence be collected expeditiously and identified so that it remains traceable to the incident. ASTM E860 governs the examination and testing of items involved in litigation, and its central requirement is procedural: document the nature, state, and condition of an item before examining it, and where a planned examination would alter that condition enough to limit later examination by others, notify the other known interested parties so that they may attend or object. In fire cases this drives the joint examination protocol, in which all parties attend, the sequence is agreed in advance, and each destructive step is documented as it occurs.

Practical preservation for electrical evidence begins at the scene. Circuits should be traced and documented before conductors are cut, and each cut end labeled and mapped so that recovered segments can be reassembled into their original topology. The service equipment, panelboard, and every protective device should be recovered with the position of each device recorded as found. Receptacles, switches, and luminaires from the area of origin should be recovered intact and still connected to their conductors, and appliances recovered whole with cords and plugs attached, since the plug and the receptacle it occupied are frequently the most informative pair of artifacts in the scene. Everything should be packaged to control corrosion, since fire debris is acidic and hygroscopic. Photographs of each item in place, a scaled diagram carrying the circuit topology, notes on the as-found condition of every protective device, and an unbroken chain of custody are the minimum record.

Laboratory Examination in Sequence

Laboratory work on recovered artifacts follows an order of increasing destructiveness, and that order is a requirement of the notice obligations rather than a preference. Nothing irreversible should be done while a non-destructive technique can still answer the question. Visual and stereomicroscopic examination comes first, with the item otherwise untouched. Radiography follows, revealing internal construction, conductor continuity, foreign objects, and component position without opening anything. Computed tomography extends radiography into three dimensions and suits anything whose internal spatial arrangement matters, including battery packs, sealed relays and contactors, potted assemblies, and appliances that opening would disturb. It also produces a permanent digital record other parties can examine independently.

Scanning electron microscopy with energy-dispersive X-ray spectroscopy is the workhorse of the next phase, resolving surface morphology far beyond optical magnifications and identifying elemental composition at points and across maps. Its applications include identifying the metals present at a fault site, detecting eutectic alloying by finding the alloying element, characterizing corrosion products, and identifying foreign particles at a battery initiation site. Elemental analysis also underlies several of the proposed cause-versus-victim discriminators, and its limits there should be stated whenever it is used for that purpose.

Metallography is the most destructive step and the most informative about what happened inside the metal. Sectioning a conductor through an arc site, then mounting, polishing, and etching the section, reveals the resolidified zone, its dendritic structure and orientation, porosity, the boundary between melted and unmelted metal, and the heat-affected zone in the parent conductor. A properly prepared section distinguishes arc melting from fire melting far more reliably than surface examination. What it does not do, on present knowledge, is establish that the arc caused the fire. Sectioning consumes the artifact at that location, so it should be preceded by full documentation and notice.

Electrical testing of recovered devices deserves separate mention because it is so often mishandled. Protective devices can be tested for calibration and trip characteristics, and the results bear on whether a claimed overcurrent event would have been interrupted, but a circuit breaker's as-found condition is itself evidence and testing alters it. The handle position, the internal condition of the contacts, and any evidence of arcing within the device should be documented, ideally radiographically, before the device is operated at all.

Reconstruction and Exemplar Testing

Physical reconstruction and exemplar testing convert a hypothesis from an assertion into a proposition that can fail. Exemplar testing uses unfailed units of the same make, model, and preferably the same production period as the subject article, either to establish the normal condition against which the subject's anomalies are judged or to reproduce the hypothesized failure. Provenance, date codes, and revision level should be documented as carefully for the exemplar as for the evidence.

Reconstruction testing at the assembly or scene level attempts to reproduce the ignition sequence under controlled and instrumented conditions: a receptacle wired with the same fault geometry and adjacent materials, a cell subjected to the hypothesized abuse, an appliance operated with the suspected defect induced. A successful reproduction demonstrates that the mechanism is competent to produce the observed result under conditions resembling those at the scene. It does not demonstrate that this is what occurred, since other mechanisms may produce the same result, but a failure to reproduce under favorable conditions is strong evidence against the hypothesis. Computational fire and thermal models play the same discriminating role, provided the uncertainty in their assumed fuel loads and ventilation histories is stated. All such work should be planned before execution and notified to the other parties where it will consume evidence.

Reporting, the Standard of Proof, and Reliability Limits

The report is where the discipline of the investigation becomes visible or fails to. A defensible report separates observation from inference, states each hypothesis considered and the basis on which it was eliminated or retained, identifies which testing was cognitive and which experimental, and states its limitations explicitly rather than leaving them to be discovered. Where the area of origin could be resolved only to a room, the report should say so, and where an artifact is consistent with two mechanisms, it should name both.

The standard of proof for a stated cause is higher than plausibility, and considerably higher than the absence of an alternative. In United States federal practice, Federal Rule of Evidence 702 requires the proponent of expert testimony to establish that it rests on sufficient facts or data, that it is the product of reliable principles and methods, and that those methods were reliably applied to the facts. An amendment effective December 1, 2023 made explicit that these requirements must be shown by a preponderance of the evidence. The Daubert factors guiding the reliability inquiry, including whether a technique has been tested and its known or potential error rate, apply awkwardly to several traditional fire indicators precisely because their error rates are either unmeasured or unfavorable.

An honest account of the field's reliability limits therefore belongs in the practice and not only in the literature. Bead size does not indicate fault severity. Concentration of arc beads does not locate the origin. Visual examination of a bead does not distinguish cause from consequence, and neither, on present evidence, does any metallurgical technique yet proposed. Char depth does not yield burn duration. Post-flashover pattern analysis showed poor accuracy in the one widely cited exercise that measured it. Each method simply supports a narrower claim than practitioners have historically made for it. The obligation that follows is to state conclusions at the confidence the evidence supports and no higher, and to classify a cause as undetermined when affirmative evidence is lacking, which is harder to honor than to state because the parties who retain investigators generally want a named cause. The handling of a report in a proceeding is addressed in Legal and Litigation Support.

Conclusion

Electrical fire investigation is governed less by a catalog of indicators than by an order of operations. Establish the area of origin from the physical evidence before considering any ignition source. Treat every electrical artifact as ambiguous between cause and consequence until specific evidence resolves it. Require that a proposed source be competent for the first fuel actually present, in the geometry present, for the duration available. Preserve evidence in a form another engineer can re-examine, and give notice before any step that forecloses re-examination. Test hypotheses in ways that could refute them.

The electrical evidence rewards the investigator who knows which mechanisms leave which traces. Glowing connections, arc tracking across carbonized insulation, ejected molten metal from an arcing short, thermally confined conductors, and lithium cell thermal runaway all have characteristic signatures and well-documented energy requirements. Gross overload and bolted shorts in protected branch circuits, by contrast, rarely start fires, and asserting them requires explaining the protective device. The field's own research has retired more indicators than it has established, so an investigator whose method rests on the current state of the literature, and who says plainly what the evidence does not show, will produce work that holds up.

Related Topics