Percolation threshold in filled conductive plastics

Filled conductive plastics switch from insulator to conductor at a percolation threshold, the loading where filler particles first form an unbroken network. That threshold is set mostly by filler aspect ratio and by how the matrix crystallizes, not by how conductive the filler itself is. Carbon black needs 15 to 25 wt%. Carbon nanotubes need 0.05 to 1 wt%. Almost every formulation decision that follows is a consequence of that gap.

Where the conductivity actually comes from

Plastic is an insulator. Three routes exist for making a part conduct, and only one of them ships in volume.

The filled route mixes carbon black, carbon fiber, nanotubes, graphene, metal powder or metal fiber into an insulating matrix such as PP, PA, PC, PPS or PEEK. Conduction happens through a network of touching or nearly touching particles, and the matrix stays an insulator. More than 95 percent of commercial conductive plastic products are made this way. This article deals only with this route.

Intrinsic or structural polymers, meaning polyaniline, polypyrrole, polythiophene and PEDOT:PSS, conduct through conjugated backbones and need no filler. Poor processability, limited stability and high cost keep them out of most volume applications. They appear here only as a reference point.

Ionic conduction, in which a PEO or PVDF-HFP matrix dissolves a lithium salt and charge moves as ions, is a different physical picture entirely. It belongs to the solid electrolyte family and is out of scope.

Conductivity comes from the filler. Whether a part conducts depends on whether the network is connected, and how much filler went in matters less than the connectivity. A compound at 20 wt% carbon black with a broken network can still measure as an insulator. A compound at 3 wt% nanotubes with a good network can measure as a conductor.

Units that decide whether a number is usable

Most failures in this field come from units rather than from data. The same number gets attached to the wrong one.

Two disciplines follow from that table.

Weight percent and volume percent are not interchangeable. Carbon black sits near 1.8 to 1.9 g/cm3 and nanotubes near 2.1 g/cm3, while a polyolefin matrix is close to 0.9 g/cm3. The same recipe written in wt% and in vol% differs by nearly a factor of two. A literature statement such as a nanotube threshold of 0.1 percent is unusable until the basis is stated.

Volume resistivity and surface resistivity are also not interchangeable. They differ by a specimen thickness factor. Acceptance testing is often written in surface resistance bands, for example 10^6 to 10^9 ohm/sq for static dissipative parts, while material data sheets describe intrinsic behavior in volume resistivity. The two lines cannot be compared directly.

How electrons cross an insulating gap

The macro picture is a connected or disconnected network. The micro picture is how an electron gets from one particle to the next, and that second picture is routinely underestimated, because adjacent particles are usually not in contact at all. A polymer layer a few nanometers thick sits between them.

Tunneling carries more of the current than most formulators assume. The Simmons model makes tunneling resistance grow exponentially with barrier width, so widening a gap from 2 nm to 4 nm can raise that resistance by more than an order of magnitude. This explains an effect that looks backwards at first: a matrix that wets the filler well can end up less conductive. Complete wetting coats every particle in a continuous insulating shell and pushes apart particles that would otherwise touch.

Surface treatment has the same double edge. A coupling agent improves dispersion and reduces agglomeration, which helps build a uniform network. It also thickens the insulating layer on the filler surface, which raises both contact resistance and tunneling resistance. In real systems the optimum sits at an intermediate dosage rather than at the maximum, and the curve rises before it falls.

Resistivity bands and what each one buys

Copper sits at about 1.7 x 10^-6 ohm-cm and an unfilled plastic above 10^14 ohm-cm, a spread of roughly 20 orders of magnitude. The engineering value of the filled route is that it places a molded part anywhere along that axis on demand. Antistatic, shielding and high conductivity are three different application lines with very different filler requirements, and carrying an antistatic recipe over to a shielding part is a common way to fail.

Band edges shift slightly between standards. The numbers above are working engineering conventions, not legal limits, and surface resistivity cannot be converted into volume resistivity.

Resistivity is also not a constant, which design work often ignores. Raise the voltage and gaps that were tunneling start emitting, so measured resistivity falls. Heat a semicrystalline part toward its melting range and the matrix expands sharply, breaking contacts and sending resistivity up by orders of magnitude. Apply alternating current and the composite shows resistive and capacitive behavior at once, so dielectric loss dominates the high frequency reading. A single number measured under one set of conditions cannot be extrapolated to another.

Percolation threshold: what actually sets the number

Percolation theory is where the quantitative side of this field starts. On a typical resistivity against loading curve, the region left of the percolation threshold is almost flat, because particles are isolated islands and adding more of them only creates more islands. Past the percolation threshold, a three-dimensional connected path appears for the first time and resistivity drops by several orders of magnitude within a narrow loading window.

The scaling law is written as sigma = sigma zero times (phi minus phi c) to the power t, valid above the threshold. The critical exponent t is about 1.3 in two dimensions and about 1.9 in three. Reported experimental values run 1.6 to 2.2, and elastomeric matrices can stretch the range to 1.3 to 3.5. The exponent depends on dimensionality rather than on chemistry, which is what lets it transfer across filler systems.

The 16 vol% anchor for spheres. For an ideal random dispersion of spheres, theory puts the critical volume fraction near 16 vol%, with lattice models giving 16 to 18 vol%. That number is the empirical anchor for carbon black systems. Battery electrodes using carbon black as a conductive additive are usually quoted near 3 percent, and the difference comes from the chain-like high structure of carbon black, which is not a sphere at all.

Aspect ratio is the only large lever. A long fiber can occupy several adjacent lattice sites at once, while a sphere occupies one. Excluded volume theory gives a threshold proportional to 1 divided by (2 x AR) for random orientation, and approximately 1 divided by AR for aligned fillers. Every order of magnitude in aspect ratio removes roughly one order of magnitude from the theoretical percolation threshold.

Aligned fillers reach percolation at lower loading than random ones because contact probability rises, but alignment brings anisotropy along with it. That trade shows up again in processing.

Published estimates for the same classes scatter with the orientation assumption. At aspect ratio 100 the quoted percolation threshold is 3.6 to 4 vol%, and at aspect ratio 1000 it is 0.3 to 0.8 vol%. Both ranges are reported here rather than averaged, because the orientation and dispersion assumptions behind them differ.

Why measured thresholds always sit above theory. Plot measured data against aspect ratio and a systematic bias appears. Measured percolation thresholds are consistently higher than the theoretical lower bound, and the gap widens as aspect ratio increases. The fitted trend follows AR to the power minus 0.65, while excluded volume theory predicts AR to the power minus 1.

The exponent degrades because theory assumes independent, evenly dispersed particles and reality delivers agglomerates. Nanotubes tangle into bundles under van der Waals attraction, and a tangled bundle behaves electrically much more like a sphere than like a fiber. Its effective aspect ratio collapses. This is why dispersion, not loading, is the central process problem on the nanotube route.

There is a second, dynamic loss. The percolation threshold is not a thermodynamic constant. During melt compounding, shear breaks weak agglomerates apart and reduces the density of conductive paths. Raising shear rate from 50 to 200 per second can push the percolation threshold of an identical recipe up by 0.15 to 0.3 vol%. The effect is called dynamic percolation hysteresis, and it is the widest gap between bench data and production data. Samples made in solution routinely show a much lower threshold than the same recipe run through a twin screw extruder.

The nine filler families and their percolation threshold ranges

Selecting a filler comes down to trades among aspect ratio, specific surface, cost, density penalty and corrosion resistance. Sorting the options into two families and nine classes collapses most of the apparent complexity.

The first question is not which filler conducts best. It is which resistivity band the part must hit, how much mechanical performance the part can lose, and what loading and abrasion the available equipment will tolerate. Answer those three and the family is usually decided.

Carbon black: the mature and heavy route

Carbon black is cheap, reliably supplied and shear resistant, and it has been the default answer for as long as conductive plastics have existed. The price is loading. At 15 to 25 wt% a tough plastic turns brittle.

Almost all of the route’s value sits in the structure of the carbon black itself. Rubber grades such as N330 and N550 are built for reinforcement, with low surface area and low structure, and they make poor conductive fillers. Conductive carbon black splits into two generations. Conventional grades such as acetylene black and conductive channel black run from tens to 250 m2/g of surface area and need 15 to 25 wt% to work. High structure grades such as Ketjenblack, Printex XE2 and the Ensaco series build primary particles into chain-like dendrites, reach 800 to 1400 m2/g, and network by branching, so a small number of branches covers a large volume.

That table yields one conclusion that goes straight into a formulation. A high structure grade reaches the same conductivity at one sixth to one third of the loading a conventional grade needs, and EC-600JD needs roughly half of what EC-300J needs. The reason is contact geometry. A branched three-dimensional network gives area and line contact between particle and matrix, while a network of spheres gives point contacts that break easily and do not survive processing.

Three numbers are worth carrying into a formulation review. In PP with EC-300J the useful window is 11 to 25 parts per 100 parts of resin, with 13 to 23 preferred and 17 to 22 optimal; below 11 parts conductivity barely appears, and above 25 parts foam cell size shrinks and flame retardancy drops. In PA46 the percolation threshold is about 20 wt% before annealing and about 10 wt% after 126 degrees Celsius annealing raises crystallinity from 39 to 67 percent, which compresses the amorphous fraction and doubles the local filler concentration. In PVDF the conductivity transition runs across 30 to 40 wt%, reaching 2.36 S/cm at 40 wt%, equivalent to 0.42 ohm-cm. That last number is far above polyolefin systems and shows that the matrix carries as much of the threshold as the filler does.

The route has three real weaknesses. Loading of 15 to 25 wt% corresponds to roughly 8 to 14 vol%, and elongation at break typically falls 60 to 90 percent, moving the material from ductile to brittle. DBP values of 310 to 510 mL/100 g mean very high oil absorption, so melt viscosity climbs and thin wall or long flow path parts become hard to mold. And the resistivity ceiling lands at 10^0 to 10^1 ohm-cm, which is nowhere near what a bipolar plate or a high current electrode needs.

What keeps carbon black in place despite all that is shear resistance. A branched three-dimensional network survives high shear extrusion and repeated recycling with its conductivity largely intact, while nanotube networks get broken and carbon fibers get shortened. Where a part must survive repeated processing or very high shear, carbon black is often the only dependable option. It remains the workhorse for antistatic parts, static dissipative parts and cable semiconductive shields.

Carbon fiber and nanotubes: trading aspect ratio for loading

Both routes work the same logic, using aspect ratio to buy loading down. Finer and longer fibers need less material to connect a network, so less mechanical damage follows. The differences are whether aspect ratio survives processing and whether the cost can be carried.

Carbon fiber delivers two functions at once. Chopped PAN based carbon fiber starts at 5 to 8 micrometers in diameter with a cut length of 3 to 12 mm, so the theoretical aspect ratio runs into the hundreds or higher. Reality intervenes. Twin screw compounding followed by injection molding shortens the fibers to 100 to 300 micrometers, leaving an effective aspect ratio of only 20 to 100. That is why carbon fiber compounds still need 10 to 20 wt% in practice, rather than the very low loadings theory suggests.

The value sits elsewhere. Carbon fiber conducts and reinforces at the same time. Adding 10 to 15 wt% to PPS, PA or PC crosses the electrical threshold while raising modulus and heat resistance, which is an economical single step for automotive and electronics housings. Molded graphite composite bipolar plates commonly use graphite with 10 to 15 wt% carbon fiber for the same reason.

Nanotubes reach the lowest threshold and pose the hardest dispersion problem. Nanotubes are the most efficient filler on this route. Single walled tubes run 0.8 to 2 nm in diameter, multiwalled tubes 5 to 30 nm, and lengths of 1 to 50 micrometers put aspect ratio in the hundreds to thousands. Reported percolation thresholds range from 0.05 to 1 wt%, and polyolefin systems usually land at 1 to 3 wt%.

Masterbatch concentration is the second largest unit trap in the field. A great deal of published material states a nanotube addition of 15 wt% when it means the concentration inside a masterbatch. Once diluted for the part, the real content is 3 to 8 wt%. Reading masterbatch concentration as a part recipe throws off every cost and property estimate downstream, so the first question about any nanotube number is whether it describes a masterbatch, a slurry or a finished part.

Dispersion is the hard problem on this route. Van der Waals attraction bundles the tubes, and a bundle behaves like a sphere, which is exactly the mechanism behind the measured thresholds that sit above theory. Available tools include ultrasonic dispersion for solution routes, bead milling and three roll milling, high shear mixing sections inside the extruder, in situ polymerization, and pre-dispersion into a masterbatch upstream. Solvent assisted wet mixing is a documented example: dissolving an epoxy in acetone before adding graphite and nanotubes distributes the tubes along the graphite interfaces, reduces microvoids, and produces a sample that clears both conductivity and flexural targets at once. Differences in dispersion quality ultimately show up as a factor of several in conductivity at identical loading.

The supply side. Nanotube conductive paste is one of the most concentrated parts of the Chinese conductive materials supply chain. One listed supplier reported 2025 revenue of 1.30 billion yuan and net profit of 228 million yuan, with multiwalled paste shipments of 71,000 tonnes and a domestic share of about 51.6 percent, according to a market research firm. Single walled tubes moved from roughly 100 tonnes in 2024 to more than 4000 tonnes in 2025, at a per tonne profit near 10,000 yuan against 2,500 to 3,000 yuan for multiwalled paste. That company has a 450 tonne single walled powder project in three phases and a 10,000 tonne paste line under construction abroad.

Two demand drivers explain the interest. Solid state batteries using high nickel cathodes and silicon carbon anodes are expected to need 3 to 5 times the conductive additive of current designs. Flexible sensors and electronic skin in robotics have also begun sampling nanotube materials.

Graphene and hybrid fillers: whether the synergy is real

Two-dimensional platelets should cover a large area at very low loading, and on paper they do. On an injection molded part the gain usually shrinks. The useful discipline here is separating what a laboratory film can do from what a molded part can reproduce.

Claims of a large conductivity gain at 0.1 to 0.5 wt% graphene deserve a process check. Most of them come from small specimens made by solution blending and casting, where slow solvent evaporation gives platelets time to self-assemble into a network. Move the same recipe to melt compounding and injection molding, and shear scatters the network, so the percolation threshold typically rises several fold. A solution route threshold cannot be used to predict a molded part.

Hybrids make mechanical sense. Points, meaning carbon black, handle short range contact. Lines, meaning nanotubes and fibers, carry current over long distances. Planes, meaning graphene and graphite, provide large contact area. The lithium battery cathode field has used carbon black, nanotube slurry and graphene as a three component package for years, and the logic came from there.

The strongest evidence in plastics comes from composite bipolar plates. A PVDF, graphite and multiwalled nanotube system was designed so that graphite and nanotubes sit selectively at PVDF interfaces, forming a brick and mortar structure rather than an even dispersion. At 35 wt% graphite plus 5 wt% nanotubes the compound reached 161.57 S/cm in plane with an area specific resistance of 7.5 milliohm cm2, while keeping flexural strength, hydrophobicity and corrosion resistance. By comparison, PP/epoxy and PE/epoxy systems that rely on graphite alone reach only 68 to 73 S/cm even at 80 wt% graphite.

Synergy comes from morphology covering a transport weakness, which is a different thing from adding species. Carbon black is weak at long range transport, tubes and fibers are weak at short range contact between particles, and graphene is weak at dispersion. Placing a filler where its geometry helps, and especially confining fillers to interfaces instead of spreading them evenly, matters more than the number of fillers used. Simply blending two fillers without a distribution design usually delivers close to nothing.

Metal and metal coated fillers

Metal fillers are often treated as the high performance, high cost option. That is half right. Their real value is in cases where carbon systems cannot reach the target, either because the resistance must be very low or because the mechanism needs magnetic loss.

The important reading of that table is the gap between column two and column three. Bulk metals are all highly conductive, but once made into a compound the achievable resistivity rises three to four orders of magnitude, because contact resistance between particles dominates the system. Changing spheres into flakes and powders into fibers is really a way of reducing the number of contact points and increasing contact area.

Coated fillers apply one idea consistently: a cheap core carries the shape and the volume, and a thin precious metal layer carries the current. Silver coated copper uses the copper core for conduction and low cost, with a 1 to 3 micrometer silver layer preventing oxidation and bringing contact resistance close to pure silver; it is the mainstream choice for conductive adhesives and gaskets. Nickel coated carbon fiber takes a fiber whose weakness is high contact resistance to the matrix and limited conductivity at high frequency, converts the surface into a metal conductor, keeps the aspect ratio, and reaches lower resistivity than an uncoated fiber system at 10 to 25 wt% while still reinforcing. Silver coated glass beads exploit a density of about 2.5 g/cm3 against 10.5 for silver and 8.9 for copper, which is irreplaceable where a shielding gasket has to lose weight, at the cost of stopping at 10^-3 to 10^-2 ohm-cm.

One widely repeated claim needs correcting. Metal fillers are not automatically the right choice for high frequency shielding. Shielding effectiveness is the sum of reflection loss and absorption loss. Reflection loss scales with conductivity, where metals do beat carbon. Absorption loss depends on the product of frequency and conductivity, and a thin metal layer sees its skin depth shrink as frequency rises, so reflection efficiency falls away. Measured shielding effectiveness of thin metal filled materials often drops with increasing frequency. In the Ka band from 26.5 to 40 GHz and the W band from 75 to 110 GHz, structured nanotube networks and segregated MXene structures have reported 74 to 84 dB, better than most thin metal approaches.

The accurate statement is that metals earn their place through high reflection loss at lower frequencies and through magnetic loss mechanisms, not through being better as frequency climbs. A supplier claiming a metal filler built for high frequency should be asked for a measured shielding effectiveness against frequency curve.

The route carries four costs. Density is the first, with silver at 10.5 and copper and nickel at 8.9 g/cm3, which fights every lightweighting target. Oxidation and migration come second, since oxide skins on copper and aluminum raise contact resistance and silver migrates under moisture and electric fields. Abrasion comes third, because metal fibers and hard particles wear screws and molds. Cost is fourth, and silver price moves pass straight into part cost, which is the whole reason silver coated copper exists.

Matrix resin: the hidden variable

Change the resin grade and keep the filler recipe, and a conductive compound can turn into an insulator. The matrix is a main variable of the same rank as filler morphology, and it works through three channels.

Channel one: crystallization squeezes fillers into a network. As a semicrystalline matrix cools, growing spherulites push insoluble filler particles into grain boundaries and amorphous regions. The filler stays evenly dispersed at the macro scale but gets compressed into local regions at the micro scale, forming what is called a segregated network. Local volume fraction ends up far above the nominal value, and this is the fundamental reason semicrystalline matrices show lower percolation thresholds than amorphous ones.

The processing consequence is direct. Cooling rate and spherulite size can be controlled, so the percolation threshold is designable rather than a fixed material constant. Fast quenching gives fine spherulites, slow cooling gives coarse ones, and the two produce different conductivity from the same recipe.

Channel two: thermal expansion mismatch creates PTC behavior. A polymer matrix expands more than an order of magnitude more than its filler. As temperature rises the matrix swells and pushes particles apart, conductive paths break, and resistivity climbs. Near the melting point of a semicrystalline matrix, crystal melting brings a sudden volume jump and resistivity can rise by orders of magnitude across a narrow temperature window. This is the positive temperature coefficient effect, and it is the basis of self-regulating heat tracing cable and polymeric resettable fuses.

Two rules govern PTC design. Switch temperature follows melting point, not crystallinity: PE melts at 100 to 130 degrees Celsius and switches near 125, PP at 160 to 170 and near 155, PVDF at 165 to 190 and near 168, PTFE at 327 to 342 and near 320. To get a given self-limiting temperature, choose a matrix with the matching melting point. EVA is adjustable through its vinyl acetate content and is often blended with PE for low temperature heat tracing.

PTC intensity, defined as the log of peak resistivity over room temperature resistivity, follows thermal expansion rather than crystallinity alone and can reach seven orders of magnitude. Adding POE to a system drops crystallinity from 56.1 to 44.3 percent while raising PTC intensity, because POE expands more than HDPE and tears the network apart more completely. That corrects the common simplification that higher crystallinity always means stronger PTC.

An NTC region usually follows the PTC peak. Once crystals have fully melted, filler reagglomerates or internal stress relaxes, resistivity falls back, and the device loses its switching behavior. Irradiation crosslinking is the standard fix: the crosslinked network holds fillers in place, largely suppresses NTC, and improves PTC reproducibility.

Channel three: melt viscosity and surface tension set dispersion quality. Higher melt viscosity transfers more shear stress to the filler, which opens agglomerates more effectively but also breaks aspect ratio faster. Surface tension decides whether the polymer wets and encapsulates the filler or separates from it, which sets contact resistance and barrier width directly. Neither has a universal optimum, so the pair has to be tested.

Thermosets behave differently. Epoxy and phenolic do not crystallize and get no exclusion effect, but they accept far more filler than thermoplastics, with graphite loading reaching 60 to 80 wt%, which is decisive for bipolar plates. Expanded graphite, flake graphite and phenolic resin compression molded at 75:25 reaches 250 S/cm in plane, among the highest of any composite plate route. Thermoplastics bring injection molding and automation, thermosets bring filler capacity, and that is a clean trade.

How processing rewrites the network

The same batch of compound can differ by an order of magnitude in resistivity between a compression molded part and an injection molded one. Within a single injection molded part, the flow direction and the cross flow direction can differ tenfold. Conductivity is a joint product of recipe and process, and a resistivity figure without a process attached is not a material property.

The first gap between bench and plant sits at the top of that table. Solution and latex routes reach much lower thresholds than melt compounding for two reasons: no shear damage, so orientation and aspect ratio survive, and slow solvent evaporation, which gives fillers time to self-assemble. Using a solution threshold to estimate the loading for a twin screw recipe will almost always come out low. Rising shear rate in melt compounding adds dynamic hysteresis on top of that, pushing the threshold up again.

Shear and orientation: why molded parts underperform. Melt flows fast through a mold cavity and aligns fibrous and tubular fillers along the flow direction. Dense conductive paths form along the flow and sparse ones form across it, so the part becomes strongly anisotropic.

Relative order of magnitude, with a compression molded part set at 1: injection molded along flow 2.5, injection molded across flow 12, and injection molded then annealed 0.5.

Measured on micro injection molded PS with carbon black, the anisotropy ratio of parallel to perpendicular conductivity is 12.6 at 3 wt%, 5.6 at 5 wt%, and 3.2 at 10 wt%. Anisotropy is worst at low loading, because a network near the percolation edge amplifies any orientational disturbance. Thin wall parts are the highest risk combination, since they are usually forced to low filler content while seeing the strongest shear.

Annealing brings the network back. Holding a part below the matrix melting point gives chain segments mobility, so fillers scattered by shear reconnect and resistivity falls. In PA46 with carbon black, annealing at 126 degrees Celsius raised crystallinity from 39 to 67 percent and dropped the percolation threshold from about 20 wt% to about 10 wt%, a halving. For PTC devices, slow cooling and quenching behave completely differently: slow cooled samples shift the PTC peak to higher temperature, narrow the peak and raise its intensity.

Foaming concentrates filler in a different way. As cells grow, they push filler out of the growing volume and into the cell walls, raising local concentration and conductivity at the same nominal loading. That is the process window behind lightweight conductive foams and flexible piezoresistive sensors. Mechanical performance and surface quality pay for it, and cell structure is difficult to control.

Recycling exposes a structural problem in this route. Fiber and tube aspect ratio is cut down step by step and cannot be recovered, so repeated extrusion and regrind steadily lower conductivity. Mitigations include masterbatch dosing to shorten residence time in the dispersion section, screw designs that lower peak shear, and side feeding so the filler avoids the harshest shear zone. The industry consensus is blunt: shear that cuts fiber length destroys the network instantly, and that often decides success more than loading does.

The master table and the four step selection

Bipolar plates push the graphite route to its limit. Hydrogen fuel cell bipolar plates are one of the most demanding applications in the field. The automotive target sets in plane conductivity above 100 S/cm together with flexural strength above 40 MPa. Both must be met at once, and they conflict: more filler means better conduction and worse mechanics.

High graphite loading of 60 to 80 wt% is the standard answer, and only thermosets such as epoxy and phenolic can hold that much. Two improvement paths are visible. The first is bridging with a high aspect ratio filler: adding 5 wt% nanotubes to a PVDF and graphite system builds a segregated network and reaches 161.57 S/cm with 7.5 milliohm cm2 area specific resistance, while solvent assisted wet mixing plus 7 wt% nanotubes reaches 151.6 S/cm and 56.7 MPa in an epoxy system. The second is chopped carbon fiber at 10 to 15 wt%, which is the common practice in domestic molded graphite plates. A pure carbon black system reaches only 2.36 S/cm at 40 wt% loading, two orders of magnitude short and effectively excluded from this application.

Four steps that decide the filler. First, fix the target resistivity band. Antistatic parts need 10^6 to 10^9 ohm/sq surface resistance and carbon black is sufficient. Bipolar plates need above 100 S/cm and require high graphite loading plus a bridging filler. Selecting across bands is the most common waste in the field.

Second, check how much mechanical performance the part can lose. The line between a structural part and a functional part is clear enough. Elongation losses of 60 to 90 percent are the normal price of conductivity, so any load bearing location should avoid the high loading routes.

Third, check process constraints. Thin wall parts need low viscosity, which rules out high DBP carbon black. Repeated processing or regrind rules out high aspect ratio fillers. Injection molding rules out thermosets and loadings much above 60 wt%.

Fourth, check cost and supply. Nanotubes are efficient but expensive per kilogram, carbon black is cheap but used in large amounts, and the difference in part cost is often smaller than intuition suggests.

What conductivity costs

Gains land in orders of magnitude. Losses land in percentages. That asymmetry sets where these materials can be used.

Resistivity moves across about ten orders of magnitude while every mechanical and processing loss sits in the percent range. The trade is therefore excellent on the electrical axis and a real structural weakening on the mechanical axis, which is why conductive plastics are mostly non load bearing functional parts: crates, carrier tape, shielding housings, bipolar plates and heaters.

The loss is hard to remove because conduction and mechanics want opposite filler distributions. Conduction needs fillers close to each other or touching, so a network can form. Mechanics needs each filler fully encapsulated and firmly bonded, so stress can transfer. One wants fillers to approach, the other wants the matrix between them, and the two demands collide in the same volume.

Coupling agents, surface grafting and compatibilizers therefore only go so far. They improve dispersion and recover part of the toughness, but they cannot remove the conflict. It is also why low loading, high aspect ratio routes such as nanotubes and graphene outperform high loading routes such as carbon black on mechanical terms: low loading simply needs fewer interfacial compromises.

Shielding numbers need decoding before comparison. Shielding effectiveness is defined as 20 times the base ten logarithm of incident power over transmitted power. Decoded: 20 dB blocks 90 percent of the energy, 40 dB blocks 99 percent, 60 dB blocks 99.9 percent and 80 dB blocks 99.99 percent. The industry phrase that 20 dB counts as effective shielding is widely repeated, and reading it as almost fully blocked is wrong. At 20 dB a tenth of the energy still passes, and the step to 40 dB is a factor of ten, not a factor of two. Any comparison of dB values has to be converted back into a blocked fraction first.

Measuring it: three ranges and three standards

Resistivity spans 20 orders of magnitude and no single instrument covers that range. Choosing a method outside its intended range produces numbers that cannot be compared with anything.

The four point method separates current injection from voltage measurement, which cancels the contact resistance between probe and specimen. That matters for highly conductive material, where contact resistance can rival the specimen resistance itself. Two probe setups are simpler and suit films and coatings but carry large errors on good conductors. A high resistance meter with a three electrode system covers the insulating and dissipative range up to 10^6 to 10^16 ohm.

Shielding measurement has its own conventions. ASTM D4935 uses a coaxial transmission line for flat sheet, usually across 30 MHz to 1.5 GHz, and the measured value can be separated into reflection loss, absorption loss and a multiple reflection correction. Waveguide methods with a vector network analyzer cover X band at 8.2 to 12.4 GHz, Ka band at 26.5 to 40 GHz and W band at 75 to 110 GHz, and they are the standard route for high frequency evaluation. Thickness matters throughout: shielding effectiveness scales strongly with specimen thickness, so cross study comparison has to be normalized to dB per mm or it means nothing.

Static control has a separate standards family. ANSI/ESD STM11.11 covers surface resistance measurement, IEC 61340-5-1 covers protection of electrostatic sensitive devices, and IEC 61340-2-3 covers resistance and resistivity test methods. These differ from insulating material standards in electrode configuration, electrification time and conditioning, so applying the conditions of an insulation standard to an antistatic acceptance test is a common compliance error.

Six measurement traps account for most bad numbers. Mixing wt% and vol% is one, and mixing ohm-cm with ohm-m is another, since they differ by 100. Volume resistivity gets confused with surface resistivity, temperature and humidity go unrecorded even though a PA matrix absorbs moisture, electrification time and test voltage stay unstated on a non ohmic material, and specimen thickness and molding method are left out when injection molded and compression molded parts are not comparable.

Five results that break the usual rules of thumb

Better wetting can mean lower conductivity. A matrix that thoroughly wets the filler coats every particle in a complete insulating shell, pushing apart particles that might otherwise touch. Good interfacial compatibility is not automatically good for conduction.

A coupling agent has an optimum rather than a direction. Surface treatment improves dispersion and thickens the insulating layer at the same time, so the dosage curve rises then falls. More is not better, and the peak has to be found experimentally.

Metal fillers are not the answer at higher frequency. Reflection loss favors metals, but skin depth shrinks with frequency and measured shielding effectiveness of thin metal layers often declines as frequency rises. Structured carbon and MXene networks have reported 74 to 84 dB in the Ka and W bands, above most thin metal approaches.

Low aspect ratio wins on strain sensitivity. At 10 percent strain, a zero dimensional carbon black in PDMS composite shows a gauge factor of 15.75, while a one dimensional nanotube in PDMS composite shows 4.36. Sensitivity depends on how completely the network breaks under tension, not on conduction efficiency. Carbon black point contacts snap in large groups, while nanotube networks carry redundant connections and resist breaking. Selection criteria invert with the application.

Annealing can beat compression molding. Slow cooled compression molding is usually assumed to give the best network. An injection molded part annealed below its melting point measured 0.5 against 1 for the compression molded reference, which means lower resistivity, because annealing let fillers reconnect after shear had scattered them.

Edge of the field: MXene, liquid metal and supply

MXene, the Ti3C2Tx family, is the most active conductive filler of the past few years. Its intrinsic conductivity depends on etch and delamination quality and is reported anywhere from 1 to 24,000 S/cm, so the intrinsic figure is not the useful one. Composite performance is.

A BINOL functionalized MXene in TPU reaches a percolation threshold of 0.3 vol% and shows strong shielding in the X, Ka and W bands, with the surface functionalization solving the incompatibility between hydrophilic MXene and hydrophobic polymers. A segregated polydisulfide at MXene structure built by electrostatic self-assembly confines MXene to polymer particle interfaces and reaches 0.063 vol% threshold, 1782.9 S/m at 3.8 vol%, and 83.9 dB in the X band with absorption dominating. A cellulose nanofiber and MXene aerogel in epoxy, built by directional freezing, reaches 0.20 vol% threshold, 1672 S/m and 74 dB at 1.38 vol%, which is 37 dB/mm normalized to thickness.

Liquid metals and silver nanowires pull in two other directions. Gallium based alloys such as EGaIn and Galinstan filled into elastomers give self-healing behavior and conductivity that survives extreme strain, which suits wearables and soft robotics. Silver nanowire transparent films still beat most alternatives on the trade between transmittance and sheet resistance, and remain the main competitor to indium tin oxide in flexible devices. Both share the same unsolved problems of long term stability and process consistency, and neither has displaced incumbents at scale.

Limitations

This overview is a working map, not a design manual. Loading windows and resistivity ranges are typical values that shift with matrix grade, filler grade, dispersion quality and process, so any number here needs confirmation against the actual system being molded.

Three summary tables in the underlying source material had been exported as unreadable graphics, covering carbon family form parameters, the consolidated quantitative table and application thresholds. Those tables were rebuilt here from the figure callouts and body text only, and no intermediate values were estimated to fill gaps.

Several numeric sets in the literature genuinely disagree, and the disagreement is reproduced rather than smoothed. Thresholds for the same aspect ratio class scatter between 3.6 and 4 vol% at aspect ratio 100 and between 0.3 and 0.8 vol% at aspect ratio 1000, depending on orientation and dispersion assumptions. Market figures for nanotube paste come from a single supplier and a single research firm and should be checked against other sources.

Standard designations are current at the versions cited, but they are revised on their own schedule and a compliance review should confirm the edition in force for the target market. Nothing here is a selection recommendation for a specific part.

Bottom line

Conductivity in a filled plastic is a network property, and the percolation threshold is where that network appears. Aspect ratio is the one lever that moves the threshold by orders of magnitude, and it is the first thing to check when a recipe underperforms. Measured thresholds sit above theory because fillers agglomerate, so dispersion quality decides outcomes as much as loading does.

The matrix does more than host the filler. Crystallization exclusion lowers the threshold, thermal expansion mismatch creates the PTC effect, and melt rheology sets dispersion. Processing is part of the recipe, with anisotropy up to a factor of twelve in molded parts and annealing able to recover much of what shear destroyed. Gains arrive in orders of magnitude and mechanical losses in percentages, which is why these materials end up as functional parts rather than structural ones.

FAQ

What is a percolation threshold in a conductive plastic?

A percolation threshold is the filler loading at which particles first form a continuous conductive path through the matrix. Below it, fillers sit as isolated islands and adding more barely changes resistivity. Above it, resistivity falls by several orders of magnitude within a narrow loading window. The value is normally reported in wt% or vol%, and the two bases are not interchangeable.

Why does carbon black need 15 to 25 wt% while carbon nanotubes need under 1 wt%?

Aspect ratio explains almost all of the gap. A nanotube thousands of times longer than it is wide can bridge several lattice sites at once, while a carbon black particle occupies roughly one. Excluded volume theory predicts a threshold falling in proportion to aspect ratio. High structure carbon black narrows the gap somewhat but never closes it.

What filler loading gives antistatic performance?

Antistatic and static dissipative parts normally need surface resistance between 10^6 and 10^9 ohm/sq. Conventional carbon black reaches that band at 15 to 25 wt%, or 8 to 15 wt% for a high structure grade. Permanent antistatic parts often use carbon black because it survives repeated processing, and carbon nanotubes reach the same band at 2 to 8 wt%.

How much filler does an EMI shielding plastic need?

Shielding in the 20 to 40 dB range, which blocks 90 to 99 percent of incident energy, is reached with carbon black at the upper end of its loading window or with 10 to 20 wt% carbon fiber. Above 60 dB requires structured networks or segregated morphologies rather than simple loading. Normalize any comparison to thickness in dB per mm.

Does a coupling agent improve conductivity?

Not reliably. A coupling agent improves dispersion and reduces agglomeration, which helps build a uniform network, but it also thickens the insulating shell around each particle. That raises contact and tunneling resistance. Real systems usually show an intermediate optimum where dispersion gains and barrier losses balance, so dosage has to be optimized rather than maximized.

Why is my injection molded part less conductive than the test coupon?

Shear during molding aligns fillers along the flow direction, so paths are dense along flow and sparse across it. Measured anisotropy in a PS and carbon black system runs 12.6 at 3 wt% and 3.2 at 10 wt%. Thin wall parts combine low loading with the strongest shear, and annealing below the melting point partly restores the network.

What causes the PTC effect in conductive plastics?

Thermal expansion of the matrix pulls filler particles apart as temperature rises, breaking conductive paths. Near the melting point of a semicrystalline matrix, crystal melting adds a sudden volume jump and resistivity rises by several orders of magnitude. Switch temperature follows melting point rather than crystallinity, while PTC intensity follows thermal expansion.

Which matrix resin gives the lowest percolation threshold?

Semicrystalline matrices generally beat amorphous ones because growing spherulites push filler into amorphous regions and grain boundaries, creating a segregated network with high local concentration. PVDF and PA46 are common low threshold hosts, while ABS and PC sit higher. Grade and cooling rate within one polymer can shift the threshold substantially.

Can conductive plastics be recycled without losing conductivity?

Not without loss. Chopped fiber and nanotube aspect ratio is cut down at every extrusion pass and cannot be recovered, so conductivity falls steadily with regrind content. Countermeasures include masterbatch dosing, gentler screw designs and side feeding. Carbon black networks survive shear far better, which is one reason they remain dominant in recyclable parts.

How is volume resistivity measured on a conductive plastic?

The range decides the method. High resistance meters with a guard ring cover 10^6 to 10^16 ohm-cm under ASTM D257, two or four electrode direct current covers 10^0 to 10^7 ohm-cm under ASTM D4496, and four point probes cover below 10^0 ohm-cm under ISO 3915. Using a method outside its range produces incomparable numbers, and electrode setup differs between the three as well.

What does 20 dB of shielding effectiveness actually block?

20 dB blocks 90 percent of incident energy, leaving 10 percent to pass through. The scale is logarithmic, so 40 dB blocks 99 percent, 60 dB blocks 99.9 percent and 80 dB blocks 99.99 percent. The step from 20 to 40 dB is a factor of ten in transmitted energy, not a factor of two, and most commercial carbon filled parts land between 20 and 40 dB.

Why do measured percolation thresholds exceed the theoretical values?

Theory assumes independent, evenly dispersed particles. Real fillers agglomerate, and a nanotube bundle conducts more like a sphere than a fiber, so effective aspect ratio collapses. Fitted data follow aspect ratio to the power minus 0.65, against a theoretical minus one. Melt compounding adds shear driven hysteresis that pushes the threshold up further.