Performance Grade (PG) Bitumen: Superpave Grading Explained

Performance grading came out of the Strategic Highway Research Program (SHRP), a United States federal research programme that ran from 1987 to 1993. Its binder research produced the Superpave system — Superior Performing Asphalt Pavements — and the binder specification within it was adopted by AASHTO as M320. The programme was commissioned because the existing grading systems were not predicting pavement failure. Roads built with binders that passed every specification line were still rutting in summer and cracking in winter, and the specification could not explain why.

What penetration and viscosity grading could not do

Penetration grading has three structural limitations, and they are limitations of principle rather than of laboratory practice:

• It measures at 25 °C. No pavement failure mode occurs at 25 °C. Rutting happens when the surface is near 60 °C or hotter; thermal cracking happens well below 0 °C. Penetration at 25 °C is used as a proxy for both and is a reliable proxy for neither.
• It is an empirical test. A needle of a specified geometry under a specified load for a specified time returns a number in tenths of a millimetre. That number has no engineering units and cannot be fed into a pavement response model.
• It grades fresh binder. The material tested is the material as supplied. The material that carries traffic has passed through a mixing plant at 160 °C and then oxidised in the road for years.

Viscosity grading, under AASHTO M226 and IS 73, fixed part of the first two problems. It measures absolute viscosity at 60 °C, which is close to a hot pavement service temperature, and it reports in poise, a real unit. The AR grades of AASHTO M226 go further and classify the binder on aged residue rather than on the material as supplied, which addresses part of the third problem; the Indian VG grades of IS 73 classify on the original binder and use RTFOT only as a secondary check, through a viscosity ratio limit. But no viscosity grading system says anything whatever about low-temperature behaviour, and all of them characterise the binder under a single loading condition.

The four changes SHRP made

1. Test temperature equals service temperature. The grade names the pavement temperature at which the criterion is applied, so the specification moves with the climate instead of being the same everywhere.
2. Test the ageing state that matters for each failure mode. Construction-period behaviour is checked on original binder, early-life rutting on RTFOT residue, and long-term fatigue and thermal cracking on PAV residue.
3. Report in fundamental units. Complex shear modulus G* in pascals, phase angle δ in degrees, creep stiffness in megapascals. These are properties an engineer can calculate with.
4. Separate the failure modes. Rutting, fatigue cracking and thermal cracking each get their own parameter, at their own temperature, on their own ageing state. A binder cannot pass by averaging.

What a PG grade does not tell you

It is a binder specification, not a mix design and not a chemical description. It does not describe crude source, wax content or asphaltene fraction. It does not predict how the binder will bond to a particular aggregate, and it says nothing about moisture damage. Two binders carrying the same PG grade can have quite different chemistries. What the grade does tell you, and no other system tells you, is the climatic envelope inside which the binder has been demonstrated to work.

One consequence of that follows immediately and is worth stating in plain terms, because it is the single most common error in PG enquiries: a penetration grade cannot be converted to a PG grade on paper. There is no conversion table, no correlation factor and no equivalence chart that turns 60/70 into PG 64-22 or 40/50 into PG 70-10. Penetration is one empirical measurement at 25 °C on unaged material; a PG grade is eight measurements at five temperatures across three ageing states. The information needed to produce the second simply is not present in the first. Two cargoes that both test 60/70 on penetration can grade differently on the DSR and the BBR, because their crude sources and their ageing behaviour differ. The only way to know a binder's PG grade is to grade it, and the only acceptable evidence is the measured DSR and BBR data on the material being offered. The same is true in the other direction — a PG grade does not tell you the penetration.

A grade written PG H-L makes two statements. The binder meets the high-temperature criteria at H °C, the average seven-day maximum pavement design temperature. It meets the low-temperature criteria at L °C, the minimum pavement design temperature. Standard grades step in 6 °C increments on both sides: high temperatures of 46, 52, 58, 64, 70, 76 and 82 °C, low temperatures of −10, −16, −22, −28, −34, −40 and −46 °C.

Everything downstream follows arithmetically from those two numbers. The dynamic shear rheometer high-temperature tests are run at H. The bending beam rheometer is run at L + 10 °C. The intermediate-temperature fatigue test sits between them, at a temperature calculated from both.

The intermediate temperature: (H + L) ÷ 2 + 4

The fatigue criterion is applied in mid-range service conditions — warm enough that the pavement is not brittle, cool enough that the binder is stiff and repeated flexing under traffic accumulates damage. AASHTO M320 fixes that temperature arithmetically rather than leaving it to judgement:

Intermediate test temperature = (H + L) ÷ 2 + 4 °C

For PG 64-22 that is (64 + (−22)) ÷ 2 + 4 = 42 ÷ 2 + 4 = 21 + 4 = 25 °C. For PG 58-22 it is (58 + (−22)) ÷ 2 + 4 = 36 ÷ 2 + 4 = 18 + 4 = 22 °C. For PG 70-10 it is (70 + (−10)) ÷ 2 + 4 = 60 ÷ 2 + 4 = 30 + 4 = 34 °C. The full set for every grade covered on this site is tabulated further down, because a DSR report with no test temperature printed on it cannot be checked against the grade being claimed.

Why every BBR temperature is 10 °C warmer than the grade

Open any PG certificate and the bending beam line will show a test temperature ten degrees above the low number in the grade name: a −10 grade tested at 0 °C, a −16 grade at −6 °C, a −22 grade at −12 °C, a −28 grade at −18 °C, a −34 grade at −24 °C. The offset is the same 10 °C at every rung of the ladder. It is not a tolerance, a rounding allowance or a concession, and the report is not defective. Buyers meeting it for the first time usually read it as the laboratory testing ten degrees short of what was ordered and query a certificate that is in fact correct. The document worth querying is the opposite one: a BBR result reported at the grade temperature itself has not been run to M320.

The reason sits in a single fact about bitumen. Its stiffness is governed by two variables, not one. Temperature is the obvious variable; loading time is the other, and the two trade against each other along a shift factor that can be measured for any given binder. Holding a load for longer stiffens the measured response in the same way that chilling the specimen does. Because that trade is systematic rather than approximate, a measurement made under one combination of time and temperature can be converted into the answer for another. That principle is time–temperature superposition, and without it a low-temperature acceptance test would not be practical to run at all.

The failure being controlled is a cold snap, not an impact. A pavement takes hours to cool towards its annual minimum, and across those hours thermal stress accumulates while the binder tries to relax it. An honest simulation would hold the beam at the grade temperature for something on the order of two hours per specimen — unusable as routine acceptance work, and long enough that the deflection drifts towards the limit of what the instrument resolves. SHRP applied the shift instead: 60 seconds of loading at L + 10 °C gives the same stiffness as approximately two hours of loading at L. The beam therefore sits at L + 10 °C, both values are read at the 60-second mark, and the result is judged against the criterion for the grade. The measurement is taken at L + 10; the property it describes is how the binder behaves at L over a realistic cooling period. That is why the BBR column of the test-temperature table further down this page carries a different figure from the grade name in every row.

Creep stiffness S and the m-value

Two numbers come off the same beam, and both must pass:

• Creep stiffness S, maximum 300 MPa. How much stress the binder generates as the pavement contracts. A stiff binder builds stress quickly and cracks.
• m-value, minimum 0.300. The slope of the log stiffness against log time curve at 60 seconds — how fast the binder relaxes the stress it has built. A binder can be soft enough and still crack if it cannot relax.

The distinction matters in service because oxidative ageing attacks the m-value before it attacks stiffness. On badly aged binders the m-value is usually the criterion that fails first, which is why a specification that reported only S would miss the failures it was written to prevent.

M320 permits one substitution on the low-temperature side. Where creep stiffness falls between 300 and 600 MPa but the m-value still passes, the direct tension test (AASHTO T314 / ASTM D6723) may be used instead, at the same L + 10 °C temperature, with a failure strain of at least 1.0 %.

Air temperature is what a weather station records. Pavement temperature is what the binder experiences, and the two are not the same. A dark surface under sun runs far hotter than the air above it, and at night it radiates to a sky colder than the air. Performance grading therefore converts air temperature into a pavement design temperature before any grade is selected.

The high-temperature side

The high pavement design temperature is the average seven-day maximum temperature at 20 mm below the surface. The depth is 20 mm because that is where shear stress under a wheel is at its worst for rutting. It is a seven-day average rather than a single annual peak because rutting accumulates through a heat spell, not in one afternoon. The SHRP algorithm is:

T20mm = (Tair − 0.00618 × Lat² + 0.2289 × Lat + 42.2) × 0.9545 − 17.78

where Tair is the seven-day average maximum air temperature in °C and Lat is the site latitude in degrees. Latitude enters the equation because it governs solar angle: two towns recording the same air temperature at different latitudes do not reach the same pavement temperature.

The low-temperature side

SHRP originally assumed the minimum pavement surface temperature equals the minimum air temperature, which is deliberately conservative. The later LTPP model refined it to Tpav = 0.859 × Tair + 1.7 °C, which normally produces a slightly warmer and therefore less demanding low-temperature requirement. Agencies differ over which model they apply, and the difference is frequently a full grade step. It is worth establishing which model sits behind the grade written into your project specification, because it changes what you have to buy.

Reliability

The temperatures produced by those equations are means of the annual extremes. That is 50 % reliability — the design temperature will be exceeded in roughly half of all years. Most agencies specify 98 % reliability, which adds two standard deviations of the annual extremes to the high side and subtracts two from the low side. Because year-to-year variation in the annual minimum is much larger than variation in the annual maximum, moving from 50 % to 98 % typically costs one grade step at the top and one or two at the bottom. Reliability is the single largest lever on binder cost in the entire selection process, and it is a policy decision made by the road authority rather than a technical property of the site.

Worked example: a Gulf coastal site

Latitude 25 °N, seven-day average maximum air temperature 45 °C, annual minimum air temperature 5 °C.

• 0.00618 × 25² = 3.86 and 0.2289 × 25 = 5.72
• 45 − 3.86 + 5.72 + 42.2 = 89.06
• 89.06 × 0.9545 = 85.01, then 85.01 − 17.78 = 67.2 °C
• The next standard grade at or above 67.2 °C is PG 70.
• On the low side an air minimum of 5 °C sits above the least severe standard low grade, so −10 is used.
• Selection at 50 % reliability: PG 70-10, useful temperature interval 70 − (−10) = 80 °C, comfortably inside unmodified territory.

Worked example: a continental inland site

Latitude 40 °N, seven-day average maximum air temperature 32 °C, annual minimum air temperature −20 °C.

• 0.00618 × 40² = 9.89 and 0.2289 × 40 = 9.16
• 32 − 9.89 + 9.16 + 42.2 = 73.47
• 73.47 × 0.9545 = 70.13, then 70.13 − 17.78 = 52.3 °C, giving PG 52
• Low side under the LTPP model: 0.859 × (−20) + 1.7 = −15.5 °C, giving −16
• Low side under the original SHRP assumption: −20 °C, giving −22
• Selection: PG 52-16 or PG 52-22, depending entirely on which low-temperature model the authority applies. The useful temperature interval is 68 °C or 74 °C respectively — both unmodified.

Both examples are worked at 50 % reliability and both ignore traffic. Traffic is a separate axis: historically it was handled by bumping the high grade upwards, and it is now handled by the MSCR traffic designations described further down this page.

What grade bumping was, and why it was a compromise

Under M320, a project carrying heavy traffic or slow-moving loads — a bus stop, a port apron, a climbing lane, the approach to a signalised intersection — was specified one or two grades above the climatic requirement. A site whose climate called for PG 64-22 might be specified PG 70-22 or PG 76-22 instead. The logic was straightforward: a stiffer binder resists rutting, so make it stiffer.

There were three problems with that. First, the bumped grade became a fiction. PG 76 states that the binder was tested at 76 °C, but the pavement never reaches 76 °C, so the grade no longer means what the specification says it means and the climatic information in the name is lost. Second, bumping the high grade often drags the low grade with it, so the project buys low-temperature performance it does not need and pays for it. Third, and most seriously, G*/sin δ cannot distinguish a stiff binder from an elastic one. It is measured at very small strain, inside the linear viscoelastic region, at a single loading frequency. A hard unmodified binder and a properly polymer modified binder can return the same G*/sin δ and then behave completely differently under a truck wheel, which loads the binder repeatedly and well outside the linear region.

What the MSCR test does — AASHTO T350 / ASTM D7405

The multiple stress creep recovery test runs on the same dynamic shear rheometer, on RTFOT residue, at the climatic high-temperature grade — not a bumped one. The specimen is put through ten creep-and-recovery cycles at a shear stress of 0.1 kPa, then ten more at 3.2 kPa. Each cycle is one second of constant stress followed by nine seconds of recovery. Two results come out of it:

• Jnr, non-recoverable creep compliance, in kPa⁻¹. The strain that does not come back, divided by the applied stress. It is a direct measure of the permanent deformation the binder contributes to rutting. Lower is better.
• R, percent recovery. The proportion of strain recovered during the nine-second rest. This is the practical measure of elastic response, and it is what separates a genuinely modified binder from one that is merely hard.

Running the test at two stress levels also yields Jnr,diff, the percentage increase in Jnr between the 0.1 kPa and 3.2 kPa stages. M332 limits it to 75 %, which catches binders whose performance collapses as soon as the applied stress becomes realistic — exactly the behaviour the single-point G*/sin δ measurement was blind to.

Traffic designations S, H, V and E

M332 keeps the climatic high-temperature number and adds a letter for traffic. The grade is written PG 64S-22, PG 64H-22, PG 64V-22 or PG 64E-22. Every one of those is tested at 64 °C, because 64 °C is what the climate actually delivers. The traffic requirement is carried by the Jnr limit instead of by a fictional temperature:

• S — Standard traffic. Fewer than 10 million ESALs with traffic moving faster than 70 km/h. Jnr at 3.2 kPa maximum 4.0 kPa⁻¹.
• H — Heavy traffic. 10 to 30 million ESALs, or slow-moving traffic between 20 and 70 km/h. Jnr at 3.2 kPa maximum 2.0 kPa⁻¹.
• V — Very heavy traffic. More than 30 million ESALs, or standing traffic below 20 km/h. Jnr at 3.2 kPa maximum 1.0 kPa⁻¹.
• E — Extremely heavy traffic. More than 30 million ESALs combined with standing traffic. Jnr at 3.2 kPa maximum 0.5 kPa⁻¹.

Where elastic response has to be demonstrated, M332 applies the recovery criterion as a curve rather than a fixed number: for binders with Jnr at 3.2 kPa of 2.0 kPa⁻¹ or below, percent recovery must lie above 29.371 × (Jnr at 3.2 kPa)^−0.2633. In plain terms, the softer a binder is permitted to be, the greater the proportion of its deformation it must recover.

What M332 keeps from M320

M332 is not a replacement of the whole specification, and it is a common misunderstanding that it is. The rotational viscosity limit, the flash point minimum, the original-binder DSR minimum of 1.00 kPa, the intermediate-temperature fatigue criterion of 5000 kPa maximum and both bending beam rheometer criteria all carry across unchanged. What M332 replaces is a single line — G*/sin δ minimum 2.20 kPa on RTFOT residue — and with it, the practice of bumping the grade.

What this means commercially

Both systems are in active use and neither is going away soon. North American agencies have largely moved to M332. Export specifications outside North America are still written predominantly in M320 terms, and most refinery certificates report M320 parameters because that is what the market asks for. The practical consequence for a buyer is simple: if your project specification carries a traffic designation letter, ask specifically for the AASHTO T350 report showing Jnr and percent recovery at the stated temperature. An M320 certificate does not contain those values and they cannot be derived from it.

• PG 76-10 — the hottest paving conditions on earth, and why a PG 76 high grade normally means polymer
• PG 76-16 — a 92 degree span, wide enough that modification is effectively certain