Which insulator for circuit boards and electronics?

Insulators and substrates at a glance. No data means no value is stored under that name (alumina loss tangent, FR-4 electric strength and relative thermal index are quoted in the text). FR-4 expansion of 13 to 230 spans in-plane and through-thickness values. PEEK, PEI, PET and silicon nitride are left out as mostly blank.

Which circuit-board insulator has the lowest dielectric constant?

Polytetrafluoroethylene (PTFE) has the lowest dielectric constant in our data, 2.1, from a supplier sheet for machining stock. Next come Kapton HN polyimide film at 3.4 (1 kHz, 25 micrometre film) and FR-4 at 4.04 to 4.6 (1 to 2 GHz, two laminate grades), then Borosilicate glass at 4.6 (1 MHz). The ceramics sit far higher: Aluminium nitride (AlN) reads 8.5 to 9.05 and Alumina (Al2O3) reads 9.7, both at 1 MHz. Frequencies and methods differ, so take the order, not design values. Statements outside our profiles are marked as general engineering knowledge.

A lower dielectric constant lets a signal travel faster and a trace be wider for the same impedance and substrate thickness (general engineering knowledge). FR-4 shows how far grades differ: Isola 370HR reads 4.17 at 1 GHz and 4.04 at 2 GHz by different methods, so that drop is not a trend, and Shengyi S1000-2M reads 4.6 at 1 GHz.

Silicon nitride (Si3N4), Fused silica (SiO2 glass), Polyetheretherketone (PEEK), Polyetherimide (PEI) and PET have no dielectric constant in our data, and PC has 3.1 only at 100 Hz. The dies sit higher: Silicon reads 11.7 to 11.9 and Gallium arsenide (GaAs) 12.9 to 13.1, above every substrate here, while Gallium nitride (GaN) reads 8.9 static and 5.35 at high frequency, below alumina.

Which circuit-board insulator loses the least signal energy?

Our data points to the ceramics, but it cannot rank all the candidates at one frequency. Alumina (Al2O3) has a loss tangent of 0.0001 and Aluminium nitride (AlN) 0.0003, both at 1 MHz. Borosilicate glass has 37 x10^-4 as printed, also at 1 MHz. FR-4 has 0.0161 at 1 GHz and 0.021 at 2 GHz on the Isola 370HR page, and 0.018 at 1 GHz on the Shengyi sheet. Among these four, alumina is lowest, then aluminium nitride, borosilicate and FR-4. The FR-4 values are at GHz and the rest at 1 MHz, so this orders the numbers and is not a like-for-like ranking.

The FR-4 profile lists high-frequency loss as its weak point and calls the Isola laminate a standard-loss grade. It also names Panasonic MEGTRON 6 as a low-loss, high-speed laminate, but its values are not recorded as properties in our data. Real high-speed boards use mid-loss or RF laminates, and glass-weave skew and copper roughness matter too (general engineering knowledge). Ask the laminate maker for Df at your frequency.

Polytetrafluoroethylene (PTFE) carries no dissipation factor in our data, and Kapton HN polyimide film has none stored as a property (its profile note gives 0.0018 at 1 kHz for 25 micrometre film, not comparable to the GHz values), so we cannot say whether PTFE, with the lowest dielectric constant we hold, also has the lowest loss. PTFE is chosen in practice for low, stable loss, and RF PTFE laminates are filled, so they differ from our figure, which comes from a machining-stock sheet (general engineering knowledge). The other candidates have none either.

Which insulator has the highest dielectric strength?

Kapton HN polyimide film shows the highest value in our data, 303 kV/mm on 25 micrometre film at 60 Hz, but it is a thin-film figure and the thick parts in our data read far lower. FR-4 prints an electric strength of 54 kV/mm for the laminate and prepreg, with no thickness stated. Polytetrafluoroethylene (PTFE) reads 30 to 45 kV/mm at 0.5 mm. Aluminium nitride (AlN) reads 16 to 20 kV/mm in AC tests, with 6.35 mm labelled for the CoorsTek grades. Borosilicate glass reads 16 kV/mm at 50 Hz.

Do not rank these by size. Dielectric strength falls as the sample gets thicker (general engineering knowledge), and the profiles print the test thickness for some values and not others. The Alumina (Al2O3) profile leaves it out because its sources differ in AC against DC and do not state thickness. PET prints a minimum of 7.7 kV for 50 micron film, a different unit, which we leave unconverted.

A breakdown number is not a working voltage. One practical limit in FR-4 under humid conditions is conductive anodic filament: the profile records that copper-salt filaments grow along the epoxy-glass interface under DC voltage, that vias closer together fail sooner, and that the NPL test ran at 85 °C and 85 % RH for up to 1000 hours. Specify voltage, spacing and humidity, not just the strength.

Which circuit-board insulator moves heat best?

Aluminium nitride (AlN) moves heat best, at 140 to 180 W/(m·K) in the Accuratus table. At the top of that range it is 6 times the figure for Alumina (Al2O3) (30 W/(m·K)) and 450 times that of FR-4 (0.4 W/(m·K)); the multiples are smaller at the low end, 140.

Silicon nitride (Si3N4) sits near alumina at 24 to 29 W/(m·K). Fused silica (SiO2 glass) reads 1.3 to 1.4 W/(m·K) and Borosilicate glass 1.2 W/(m·K). The polymers conduct least: Kapton HN polyimide film 0.12 W/(m·K), PET 0.15 W/(m·K), Polyetherimide (PEI) 0.22 W/(m·K) and Polyetheretherketone (PEEK) 0.29 W/(m·K). Polytetrafluoroethylene (PTFE) and PC have none in our data.

For scale, Silicon reads 150 W/(m·K), Gallium nitride (GaN) 130 W/(m·K) and Gallium arsenide (GaAs) 46 to 55 W/(m·K). Aluminium nitride is the only insulator in our data in the range of the silicon and gallium nitride dies, and its profile lists electronic package substrates, power transistor bases and heat sinks. It says thermal values differ by grade, and the CoorsTek grade conductivities are not recorded, so the range is one supplier's typical grade.

FR-4 conducts so little that boards rely on copper planes and vias to carry heat; metal-core or insulated-metal boards are another route (general engineering knowledge, not in our data).

Which substrate expands the least, and which matches a chip?

Fused silica (SiO2 glass) expands least at 0.55 ppm/K, followed by Silicon nitride (Si3N4) at 1.4 to 3.2 ppm/K, Borosilicate glass at 3.25 ppm/K and Aluminium nitride (AlN) at 4.5 in the Accuratus table (no temperature stated) or 5.5 to 5.6 for the CoorsTek grades (a 25 to 1000 °C mean). Alumina (Al2O3) reads 8.2 ppm/K. Then come the polymers: PET 17 ppm/K, Kapton HN polyimide film 20 ppm/K, Polyetherimide (PEI) 52 ppm/K and Polytetrafluoroethylene (PTFE) 120 to 130 ppm/K.

FR-4 is the awkward case. The Isola page prints an in-plane expansion of 13 and 14 ppm/°C, but through the thickness it is 45 ppm/°C below the glass transition and 230 ppm/°C above it, with a 2.8 % total expansion from 50 to 260 °C. The profile lists this as a failure route: plated-through-hole and copper stress when a board is heated past its Tg.

Assembly takes a board past that point: the profile cites reflow peaks near 250 °C. So ask for the delamination times it prints for Isola 370HR, 60 min at 260 °C (T260) and 30 min at 288 °C (T288), and the decomposition temperature, 340 °C for Isola 370HR and 355 °C for Shengyi S1000-2M.

To match a die, only the compound semiconductors can be compared. Gallium nitride (GaN) reads 5.59 ppm/K (no temperature stated) and Gallium arsenide (GaAs) 5.73 (no temperature stated) to 6.86 (300 K). The CoorsTek aluminium nitride figure is similar to gallium nitride's, though one is a 25 to 1000 °C mean and the other has no temperature, and Alumina (Al2O3) is further off. The expansion of Silicon is not in the database, and we do not quote one from memory, so we cannot say which substrate matches a silicon die. Ask for die and substrate expansion over the same temperature range.

Which insulator should I pick, and how hot can each run?

Pick by the first requirement you cannot trade away: heat to aluminium nitride, flexing or temperature to polyimide, lowest dielectric constant to PTFE, otherwise FR-4. On temperature, Alumina (Al2O3) goes highest, with a no-load limit of 1,750 °C. Silicon nitride (Si3N4) is stated at 1,200 to 1,400 °C, depending on whether the atmosphere is air or inert. Fused silica (SiO2 glass) has a mechanical limit of 1,080 °C. Borosilicate glass is stated at 450 °C for 10 hours or more and 500 °C for less. Kapton HN polyimide film has a maker-stated successful use of 400 °C as bare film, Polytetrafluoroethylene (PTFE) 260 °C and PET 150 °C.

The polymers and laminates use other measures, so do not rank them by these numbers. FR-4 has a glass transition of 180 °C by DSC and 185 °C by DMA, a decomposition temperature of 340 °C (Isola) or 355 °C (Shengyi) and a relative thermal index of 130 °C (Isola), all for high-Tg grades. Polyetherimide (PEI) has a continuous service temperature of 170 °C. Polyetheretherketone (PEEK) has a glass transition of 150 °C and melts at 343 °C. PC deflects under load at 123 °C.

Ceramics stay insulating when hot, but not without limit: the alumina profile prints a volume resistivity above 10^14 ohm-cm at 25 °C that falls to 2 x 10^6 ohm-cm at 1000 °C. Two moisture points matter. Aluminium nitride shows slow hydrolysis in water with no limit printed, so ask about handling. Polyimide takes up moisture, which can shift its dielectric constant (general engineering knowledge); the Kapton profile holds no moisture figure. Moisture absorption is recorded for FR-4 (0.15 %), PEI (0.25 to 1.25 %) and polycarbonate (0.58 % at equilibrium), but no Dk or Df against humidity.

For orbit, the data is uneven. NASA outgassing screening, as the PTFE profile records it, accepts a total mass loss of at most 1.0 % and condensable material of at most 0.10 % (an ASTM E595 type test). Kapton HN polyimide film shows 0.96 % and 0.01 % for the 200HN grade, but the entry gives no test conditions, so treat it as an indication, and its atomic-oxygen erosion is about 20 times that of Polytetrafluoroethylene (PTFE) in the same MISSE 2 table (the PTFE profile gives the figure against Kapton H), so exposed polyimide needs cover. PTFE tubing shows 0 % for both, but its profile records radiation damage in a 1968 NASA report and a predicted embrittlement in a NASA Glenn ground analysis. Printed Polyetheretherketone (PEEK) shows 0.19 % and 0.01 %. The FR-4 profile holds no space data, and the Alumina (Al2O3) profile no vacuum result.

Terms used

Dielectric constant (Dk)

how strongly an insulator stores electric field compared with vacuum. Also called relative permittivity. It sets signal speed and trace width for a given impedance.

Dissipation factor (Df)

also called loss tangent. The share of signal energy an insulator turns into heat each cycle. Lower is better for high-frequency signals.

Dielectric strength

the electric field an insulator withstands before it breaks down, in kV/mm. It falls as the sample gets thicker, so only compare values taken at similar thickness.

CTE

coefficient of thermal expansion, in ppm/K. How much a material grows for each degree of heating. A mismatch between board and chip strains the solder joints.

Tg (glass transition)

the temperature where a polymer or laminate turns from stiff to soft. Above it, expansion through the board thickness rises sharply.

CAF

conductive anodic filament. A copper-salt thread that grows along the glass-epoxy interface under humidity and voltage and can short two vias.

Silicone and silicon

silicone is a polymer used in sealants, coatings and potting. Silicon is an element and a semiconductor. Silicon nitride and fused silica contain silicon but are not silicones.

What to check before you order

1. Ask for dielectric constant and dissipation factor at your own frequency and temperature, with the test method. Values in our data come from different frequencies and methods and cannot be mixed.
2. For FR-4, ask for the Tg of your grade by DSC, through-thickness expansion above and below it, T260 and T288 delamination times, the decomposition temperature, and CAF test results for the board as your fabricator builds it.
3. Ask for moisture absorption and for Dk and Df at your operating humidity. Our profiles hold none of the latter.
4. For ceramics, name the grade and purity, and ask for thermal conductivity, expansion and dielectric strength with the test thickness. For aluminium nitride, also ask about moisture handling.
5. For polyimide, state the gauge and ask for the temperature rating of the finished flex laminate, not the bare film. In orbit, ask whether the outer layer will see atomic oxygen.
6. For spacecraft hardware, ask for outgassing results to the standard your programme uses, with test conditions. The screening limits quoted above are NASA's, not necessarily yours.
7. Ask whether adhesives, coatings, potting and release agents on the board contain silicone, even if the substrate does not.

What do we not know yet?

• Dielectric constant is missing for Silicon nitride (Si3N4), Fused silica (SiO2 glass), Polyetheretherketone (PEEK), Polyetherimide (PEI) and PET, and PC has it only at 100 Hz. We cannot say where these sit on the permittivity axis.
• Dissipation factor exists for only FR-4 (1 to 2 GHz), Aluminium nitride (AlN), Borosilicate glass (1 MHz) and a differently named loss tangent for Alumina (Al2O3) (1 MHz). Polytetrafluoroethylene (PTFE) has none and Kapton HN polyimide film has none stored as a property (note: 0.0018 at 1 kHz), so the database cannot rank loss at one frequency.
• Dielectric strength lacks Alumina (Al2O3) and a matched thickness for any pair of materials. The FR-4 and PET values are stored in other forms, and no value is tied to a working voltage or to humidity.
• Silicon has no thermal expansion value, so we cannot say which substrate matches a silicon die. Only the two compound semiconductors have one.
• The database holds no metallisation, copper adhesion, metal-core board or ceramic-substrate process data (thick film, direct bonded copper), no price, no polyimide moisture or Dk drift, and no space-flight data for FR-4, Alumina (Al2O3) or Aluminium nitride (AlN). A silicone potting compound exists as a separate profile but is not a board substrate and was not compared.

Questions people ask