Bitumen Performance Grading Guide: How the PG System Works

The Strategic Highway Research Program (SHRP) ran in the United States from 1987 to 1993, with a research budget set aside for exactly this problem: specifications that a failing pavement had passed. Its binder work produced the Superpave system — Superior Performing Asphalt Pavements — and AASHTO issued the binder half of it first as a provisional document and subsequently as the standing specification M320. ASTM publishes a technically equivalent standard as ASTM D6373, and export contracts cite the two interchangeably. Beside them sits AASHTO R29, which is not a specification at all but the practice describing how a laboratory arrives at a grade and how it verifies one that has been claimed.

Where the older systems ran out of information

The objection is not that penetration and viscosity grading are poor tests. Both are quick, inexpensive and repeatable, and both remain in daily commercial use for good reasons. The objection is that neither collects the information a pavement engineer needs, and careful laboratory work cannot add it after the fact:

• The wrong temperature. A road neither ruts nor cracks at 25 °C. It ruts when the top of the layer approaches or exceeds 60 °C and it cracks when the surface falls well below zero. Penetration reports a single point between those extremes, absolute viscosity a single point near the upper one, and neither reports the extremes themselves.
• The wrong material. Both grade the binder as delivered. The binder that carries wheels has been through a plant at around 160 °C and has then oxidised in place for years. Oxidation is not a gentle drift along the same scale — it changes which criterion the binder is closest to failing.
• The wrong kind of number. A penetration reading is an index expressed in tenths of a millimetre of needle travel. Nothing can be multiplied by it, no pavement response model accepts it, and it cannot be set against a stress that traffic actually applies.

What SHRP changed

Four choices carry the entire system, and each answers one of the defects above:

1. The grade names the temperature, so the specification travels. One document serves Basra and Winnipeg, because each criterion is applied wherever the local climate puts it rather than at a fixed laboratory setting.
2. Each failure mode is judged on the ageing state that causes it. Pumping and mixing behaviour on the binder as supplied; rutting on rolling thin film oven residue, because rutting is an early-life failure; fatigue and thermal cracking on pressure ageing vessel residue, because those arrive late.
3. Everything is reported in units that can be calculated with. G* in pascals, δ in degrees, creep stiffness in megapascals, non-recoverable creep compliance in reciprocal kilopascals.
4. Nothing is averaged. Rutting, fatigue and thermal cracking are each given a parameter of their own, measured at a temperature of their own on an ageing state of their own, and each verdict stands alone.

Why the fourth choice is the one that matters at the negotiating table

M320 has no total, no weighting and no compensation. An excellent rheometer result at the hot end buys nothing at the cold end, because the two describe mechanisms that occur six months apart. That is the structural difference between a PG certificate and a penetration-grade certificate, and it is where the label earns its value: a grade is not a description of the material, it is the record of an envelope the material was demonstrated to survive. It is also why the grade cannot be reasoned backwards from a simpler test. A two-number label compresses eight pass-or-fail criteria applied across three deliberately produced ageing states, and no single measurement holds that much.

Both halves of PG H-L are pavement temperatures, and each is a claim somebody has tested. H is the average seven-day maximum pavement design temperature at the site; L is the minimum pavement design temperature. So PG 64-22 reads: this binder satisfied the high-temperature criteria for a pavement whose hottest week averages 64 °C, and the low-temperature criteria for one that falls to minus 22 °C. Neither number is an air temperature. A dark surface under summer sun sits well above the air over it, and on a clear winter night it radiates to the sky and falls below it, so both design temperatures are derived from air records rather than read straight off them.

The ladder has 6 °C rungs on both sides — 46, 52, 58, 64, 70, 76 and 82 °C at the top, −10, −16, −22, −28, −34, −40 and −46 °C at the bottom — and no designation exists between rungs. A binder whose measured limits land at 66.8 °C and −24.6 °C therefore ships as PG 64-22, carrying margin at both ends that the label says nothing about.

The four test temperatures

Once H and L are known, every temperature in the M320 sequence is fixed. Three of them are calculated, one is constant:

• 135 °C — rotational viscosity. Fixed for every grade. It is a pumpability and workability check, not a performance criterion, so it does not move with the climate.
• H — the rutting DSR temperature. Applied twice, once to the original binder and once to the RTFOT residue.
• (H + L) ÷ 2 + 4 — the intermediate, or fatigue, DSR temperature. Applied to PAV residue.
• L + 10 °C — the bending beam rheometer temperature. Applied to PAV residue.

The intermediate temperature, worked through

The fatigue criterion is applied in mid-range conditions: warm enough that the binder is not brittle, cool enough that it is stiff and that repeated flexing under wheel loads accumulates damage. M320 fixes that temperature by formula rather than by judgement.

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

For PG 58-16: (58 + (−16)) ÷ 2 + 4 = 42 ÷ 2 + 4 = 21 + 4 = 25 °C. For PG 70-10: (70 + (−10)) ÷ 2 + 4 = 60 ÷ 2 + 4 = 30 + 4 = 34 °C. For PG 64-22: (64 + (−22)) ÷ 2 + 4 = 42 ÷ 2 + 4 = 21 + 4 = 25 °C. Note that PG 58-16 and PG 64-22 share an intermediate temperature of 25 °C while sharing neither of their grade numbers — the formula depends on both halves, so grades with the same H do not share an intermediate temperature and grades with the same L do not either.

Why the bending beam always runs 10 °C warm

BBR test temperature = L + 10 °C

On a PG 64-22 report the bending beam line reads −12 °C. On PG 58-16 it reads −6 °C. On a −28 grade it reads −18 °C. The offset is ten degrees at every rung, it is written into the method deliberately, and a certificate carrying it is the normal case rather than a short-measured one. The report worth challenging is the reverse: a beam run at the grade temperature itself was not run to M320, whatever standard the header cites.

What licenses the offset is that stiffness in bitumen has two inputs rather than one. Chill the specimen and it stiffens; shorten the time the load is held on it and the measured stiffness rises the same way. Warm it, or hold the load on it longer, and the measured stiffness falls either way. Because the two variables are interchangeable along a shift factor that can be measured for any particular binder, a warmer test held for a shorter time reports the same stiffness as a colder test held longer, and a result obtained at one pairing of time and temperature converts arithmetically into the result for another. That is time–temperature superposition, and without it low-temperature acceptance testing would not be practical to run at all.

The size of the offset then follows from what the test has to represent. Thermal cracking is not an impact event. The pavement sheds heat through the evening, sits near its annual minimum for hours in the dark, and tensile stress accumulates in a restrained layer across that whole period while the binder tries to shed it. Reproducing that honestly at L would mean loading every beam for something like two hours — unusable as routine acceptance work, and at that temperature the beam moves so little that the deflection approaches the resolution of the instrument. SHRP applied the shift instead. Sixty seconds of loading at L + 10 °C is the accepted equivalent of roughly two hours of loading at L. The beam is conditioned at L + 10 °C, both readings are taken at the 60-second mark, and they are judged against the criteria the grade carries. Short and warm in the bath; slow and cold in the road.

The dynamic shear rheometer — AASHTO T315 / ASTM D7175 / EN 14770

A thin disc of binder is held between two parallel plates. The lower plate is fixed, the upper plate oscillates back and forth through a small angle, and the instrument records the torque required and the lag between the applied strain and the resulting stress. From those it computes the complex shear modulus G*, the total resistance to deformation, and the phase angle δ, which says how much of that resistance is elastic and how much is viscous. A perfectly elastic solid returns δ = 0 degrees; a perfectly viscous liquid returns 90 degrees. Bitumen sits between, and moves along that range with temperature.

The geometry changes with the temperature. High-temperature work uses 25 mm plates at a 1 mm gap, because the binder is soft and a larger specimen is needed to produce a measurable torque. Intermediate-temperature work on PAV residue uses 8 mm plates at a 2 mm gap, because the aged binder is far stiffer and a large specimen would overload the transducer. Loading is applied at 10 rad/s, which is 1.59 Hz — a frequency chosen to approximate the loading pulse a pavement receives from traffic moving at roughly highway speed. Strain amplitudes are set by the method to keep the measurement inside the linear viscoelastic region: about 12 % on original binder, 10 % on RTFOT residue and 1 % on PAV residue at the intermediate temperature. Plate temperature is controlled to a tenth of a degree, because G* changes steeply with temperature.

Two parameters come out of the same instrument and they pull in opposite directions. G*/sin δ is the rutting parameter and carries a minimum limit: at high temperature you want the binder stiff and elastic, so that the energy put in by a wheel is returned rather than left as permanent deformation. G*·sin δ is the fatigue parameter and carries a maximum limit: at intermediate temperature you want the binder neither too stiff nor too dissipative, because the energy lost per loading cycle is what drives crack growth.

The rolling thin film oven — AASHTO T240 / ASTM D2872 / EN 12607-1

This is the plant simulation. Roughly 35 g of binder goes into each of eight glass bottles, the bottles are mounted in a vertical carousel rotating at 15 revolutions per minute inside an oven at 163 °C, and a jet of heated air at about 4,000 mL per minute is blown into each bottle as it passes the nozzle. The rotation keeps a continuously renewed film of binder on the bottle wall, so oxygen reaches fresh material rather than a skin. After 85 minutes the bottles are weighed to give the mass loss, which is capped at 1.00 % because volatile loss is what hardens a binder in the mixer, and the residue is collected for the next stages.

The pressure ageing vessel — AASHTO R28 / ASTM D6521 / EN 14769

This is the service simulation, and it is the step that separates the PG system from every specification that preceded it. About 50 g of RTFOT residue is poured into each shallow 140 mm pan, giving a film roughly 3.2 mm thick. The pans are stacked in a pressure vessel held at 2.10 MPa of air pressure for 20 hours, at 100 °C for standard practice, with 90 °C and 110 °C available where a specification calls for them. The elevated pressure drives oxidation far faster than temperature alone could without changing the chemistry of the reaction, and the condition produced is generally taken to represent something in the order of five to ten years of in-service oxidative ageing. The residue is degassed under vacuum before testing, because dissolved air would otherwise produce bubbles in the rheometer specimen.

The bending beam rheometer — AASHTO T313 / ASTM D6648 / EN 14771

A small beam of PAV residue, 127 mm long by 12.7 mm wide by 6.35 mm thick, is supported at both ends in a cold fluid bath and conditioned there for about an hour. A constant load of 980 mN is then applied at midspan for 240 seconds while the deflection is recorded. Two numbers are read at the 60-second mark: the creep stiffness S, capped at 300 MPa, and the m-value, floored at 0.300.

Both numbers are required, and they answer opposite halves of one question. S is about how much stress a contracting pavement can raise in the binder: high stiffness means the stress arrives fast and a restrained layer tears. The m-value is the slope of the log-stiffness against log-time curve at that same 60-second point, so it measures whether the stress is still draining away or has stopped draining. A binder can sit comfortably under 300 MPa and still crack, because softness is no use if the softness does not move. Field experience is that oxidation degrades relaxation before it degrades stiffness, which is why marginal aged material almost always fails on m rather than on S — and why m is the line to read first on a low-temperature report.

The useful temperature interval is simply H − L. PG 64-22 spans 86 °C. PG 70-10 spans 80 °C. PG 76-22 spans 98 °C. It matters because a straight-run binder is a single material being asked to be soft enough at one end of the year and stiff enough at the other, and there is a limit to how far one material can be stretched.

The working rule across the industry is that a span of up to about 92 °C is normally achievable with an unmodified binder from a suitable crude source. Beyond about 92 °C, polymer modification is usually required. It is a rule of thumb and not a line in any standard: the feedstock decides where the real ceiling sits, and two refineries working to the same grade will not have the same one. It is nonetheless accurate enough to tell a buyer which kind of product an enquiry is really about, before anybody quotes on it.

There is one systematic exception, and it belongs to the high end rather than to the span. Anything graded 76 °C or higher at the top end arrives as a polymer modified product in practice, span notwithstanding, because very few straight-run paving residues reach 76 °C on G*/sin δ at all without being taken to a hardness that costs them elsewhere in the sequence. PG 76-10 spans only 86 °C on paper — comfortably inside the neat envelope — and is a modified product in practice all the same. Treat a high grade of 76 as a modification trigger in its own right, independently of the interval.

What a modified binder adds to the certificate

Once a polymer is involved, the acceptance list grows beyond M320. The tests that normally appear alongside it are:

• Elastic recovery — ASTM D6084 or EN 13398, measured in a ductilometer on a stretched and severed specimen. It is the classic demonstration that a polymer network is present and functioning.
• Storage stability, or separation — ASTM D7173 or EN 13399. A sealed tube of binder is held vertically at elevated temperature and the top and bottom sections are then compared, usually by softening point. A large difference means the polymer is separating from the bitumen, which is a shipping and storage problem rather than a laboratory curiosity.
• Percent recovery from the MSCR test — AASHTO T350, described in the next section. This has largely replaced elastic recovery in North American practice because it is measured on the same rheometer as everything else.

Modified binders also change the logistics. Most require agitated or circulated storage, they have a narrower acceptable temperature window than a straight-run binder, and prolonged static storage at high temperature can degrade the polymer. If an offer moves from an unmodified grade to a modified one, the storage arrangements at destination need to be confirmed before the cargo is fixed, not after it arrives.

Safety and handling of PG binders

Two points specific to performance graded material deserve attention, because they differ from the penetration-grade norms many buyers carry over.

The flash point floor is lower than you may expect. AASHTO M320 requires a minimum of 230 °C by Cleveland open cup (ASTM D92). Many Middle East penetration-grade export data sheets quote 250 °C, and operations staff who set tank ceilings from habit rather than from the certificate may assume the higher figure applies. Read the actual flash point on the batch Certificate of Analysis and set the storage and heating ceiling from that number, with margin.

Beyond that, the standing rules for hot bitumen apply without modification:

• Water is the principal hazard. A small volume of water trapped beneath hot bitumen flashes instantly to steam and can foam a tank over the top. Drain condensate from tanks, lines and hoses before hot product is introduced, and never load hot bitumen into a wet or unverified compartment.
• Heating coils must stay submerged. An exposed coil carbonises the binder against the metal, ruins the batch and creates a genuine fire risk. Heat gradually and uniformly, and never apply an open flame to a dry drum or tank wall.
• Burns from hot binder need water, not removal. Bitumen at working temperature adheres to skin and continues to burn. Cool the affected area immediately with clean cold water for at least ten minutes and keep cooling on the way to medical attention. Do not attempt to peel or solvent-strip the material off — that removes skin with it.
• Personal protective equipment is not optional at the sampling point. Face shield, heat-resistant gauntlets with sleeves worn over them, and closed footwear. Sampling a hot line is the single most common point of injury in bitumen handling.
• Work from the Safety Data Sheet for the specific product. A modified binder and a straight-run binder can have different recommended storage temperatures and different maximum static storage periods.