What is the definition of absolute maximum ratings?
So, short story is the definition of absolute maximum rating does vary from manufacturer to manufacturer.
Most every manufacture (of those sampled) will give a variation of this warning:
Stresses beyond those listed under Absolute Maximum Ratings may cause permanent damage to the device.
Then proceed to tell you that this might happen over time:
Exposure to absolute maximum rating
conditions for extended periods may affect device reliability.
If international standards count for anything here is IEC60134 which "standardizes" the definition of absolute maximum ratings:
IEC60134 paragraph 4: Absolute maximum rating system This section states:
“Absolute maximum ratings are limiting values of operating and environmental conditions applicable to any electronic device of a specified type as defined by its published data, which should not be exceeded under the worst probable conditions.
These values are chosen by the device manufacturer to provide acceptable serviceability of the device, taking no responsibility for equipment variations, environmental variations, and the effects of changes in operating conditions due to variations in the characteristics of the device under consideration and of all other electronic devices in the equipment.”
Here is where it gets really boring, but I threw this up for comparison
Analog says that their absolute maximum ratings won't kill the part instantly but it will degrade over time:
Stresses above those listed under Absolute Maximum Ratings may cause permanent damage to the device. This is a stress rating only; functional operation of the device at these or any other conditions above those indicated in the operational section of this specification is not implied. Exposure to absolute maximum ratings for extended periods may affect device functionality.
They also say on many of their chips its process based.
The maximum supply voltage that can be applied to an op amp is determined by the fabrication process. It refers to the instantaneous value, not the average or final value. Low voltage CMOS op amps from Analog Devices, Inc., are typically limited to 6 V, whereas high voltage bipolar parts are limited to 36 V.
Source: Analog App Note MS-2551
Linear Technology gives a standard warning and reliability:
Stresses beyond those listed under Absolute Maximum Ratings may cause permanent damage to the device. Exposure to any Absolute Maximum Rating condition for extended periods may affect device reliability and lifetime.
Maxim Electronics gives the standard warning and that it will affect reliability:
Stresses beyond those listed under “Absolute Maximum Ratings” may cause permanent damage to the device. These are stress ratings only, and functional operation of the device at these or any other conditions beyond those indicated in the operational sections of the specifications is not implied. Exposure to absolute maximum rating conditions for extended periods may affect device reliability.
Texas Instruments is the same:
Stresses beyond those listed under absolute maximum ratings may cause permanent damage to the device. These are stress ratings only, and functional operation of the device at these or any other conditions beyond those indicated under recommended operating conditions is not implied. Exposure to absolute-maximum-rated conditions for extended periods my affect device reliability.
The exception is Rohm and they say you can't exceed the ratings ever or you'll cause damage:
Absolute maximum ratings are conditions that should never be exceeded, even momentarily. For example, supplying a voltage over the maximum rating and/or using in environments outside of the temperature range may cause deterioration of IC characteristics or even damage.
Source: Rohm Absolute Maximum Ratings
Semiconductor manufacture and specification is not exact, there is an element of probability involved. Any component type will have a spread of values for which it 'blows up' according to some definition.
There are three broad groups of users, which have different interpretations of the risk involved in component failure.
There is the 'no risk' group, like automotive, military, aviation, medical.
There is the 'reasonable commercial risk' group, most industrial and commercial manufacturers, and hobbyists who just want to build stuff that works and keeps working.
Then there is the 'turn it up till it blows up, then back off a bit' group of overclockers and Tesla coil builders, who know what they're doing, for whom component failure is part of the normal operating envelope.
AMRs, Absolute Maximum Ratings, are for the first two groups.
If you want your stuff to keep working without trouble, then plan to keep all parameters within the AMRs. If your recommended rail voltage is 15v, and the AMR is 18v, then set your rail crowbar at 17v. You may have one IC that kept working at 20v, lucky you. Or did it keep working, was there some degradation that wasn't immediately apparent? If you plan to exceed any AMRs, then plan to fault-find, and replace.
The AMR (Absolute Maximum Ratings) are a set of parameters that define the boundary of reliable device operations. The primary purpose of them is to avoid manufacturer'r liability. The reliability is formulated in terms of certain probability of device failure, which depends on class of the device (consumer, industrial, automotive, MIL-883, medical, etc.), and usually is not disclosed to public.
The set of AMR parameters is determined for every product by device reliability engineers, who carefully examine all operational parameters over the entire designated range of operational conditions, such as mechanical/thermal fatigue, electromigration, etc., and come up with numbers. Then this set of numbers is validated on real device samples using accelerated aging methods. Then technical marketing people fudge this number to meet customer's expectations and avoid potential liabilities.
The parameters are absolute in the sense that they are defined under the worst combination of all parameters, over all process corners. As such, if a device temperature is kept at or below nominal, exceeding maximum voltage wouldn't kill it, and exceeding device operational frequency might result in a fully functional device. That's why overclocklers do exist. However, the distribution function of these inter-dependencies are usually unknown and would take too much engineering effort to determine, so a manufacturer simply prefer to void any guarantee if one of parameters is exceeded in customer's application.
For really big customers there is some room for negotiations. If, say some reliability parameter was defined based on 1,000,000 hours of operations (~ 100 years), but the customer plans for only 5 years of service, the rating can be changed. But it takes a special effort to come up with reliability PDF to determine this and approve the application.