The mechanism of oil lubrication is either hydrodynamic, elastohydrodynamic or boundary, depending on operating conditions. The lubrication mechanism of greases, however, will be different, due to their different, gel-like or semi-solid structure.
There is no consensus on the lubricating mechanism in grease lubrication.
The most widely used model to describe grease lubrication is that grease acts as an oil reservoir where the oil is slowly released (bleeds) into the interacting surfaces.
However, based on the observation of thickener layers on the interacting surfaces, another model postulates that the interacting surfaces are covered by a thin layer of soap & the film is formed by base oil thickened with broken thickener fibres, from the progressive destruction of the soap matrix by over-rolling. This destruction releases oil, which provides free lubricant for replenishment, i.e. oil is not released by bleeding but by the destruction of the thickener.
There is overall agreement that grease lubricated bearings are generally running under starved lubrication conditions. Measurements have indicated that the film thickness initially is about 25% greater than the value expected in case of fully flooded oil lubrication. However, after a few hours the film thickness decreases to about 40% below the fully flooded (oil) lubricated value. At this point the bearing runs under starved conditions.
As a rule of thumb approximately 30% of the free volume of the bearing should be initially filled with grease. It will be clear that this is much more than required to provide the bearing with a (fully flooded) lubricant film. In the beginning, excessive grease churning, or grease flow, takes place, which is responsible for the high temperature peak caused by the churning component of the friction torque. The initially thick lubricant film in the beginning indicates that at least during this initial bearing operation thickener enters the contact.
Surface tension, Centrifugal forces, Capillary action, Scraping & Redistribution by the bearing cage, Temperature, Evaporation rate, etc are also expected to play a part in the lubricant replenishment in the interacting surface.
The various hypotheses on the mechanisms of grease lubrication, based on observations & measurements, indicate that there may be no unique mechanism. As an example, at low temperatures oxidation and evaporation will not give a significant contribution to “grease aging.” At high temperatures oxidation will dominate. Some metals catalyze oxidation (brass cages!). This may be one explanation why there is no consensus on a single mechanism.
It is certain that initial filling plays a major role. Too much grease leads to excessive churning, high temperatures, and severe grease degradation. If the bearing is properly filled, two phases can be distinguished; i.e., a churning phase where excessive grease will be pushed to the shoulders of the bearing onto the seals/covers. This process is determined by the flow properties of the grease (rheology).
Prediction of flow is very complex due to the nonlinear grease rheology and the two/three phases involved (thickener, oil, air). The remaining grease inside the bearing will be over-rolled, where the thickener structure will be broken down, releasing oil, and where the thickener material could form a thin layer, or a highly viscous layer. After this phase the “bleeding” phase takes place. This phase is characterized by starvation. The lubricant film thickness is initially larger than calculated using the base oil viscosity. Side flow reduces the film thickness. Electrical resistance measurements in bearings confirm the occurrence of starvation.

The grease reservoir may be formed by grease under the cage and/or on the bearing shoulders. However, there are also hypotheses that the bearing simply runs on the initial layer throughout its life-time and will not be replenished at all. Even here, the grease on the shoulders plays a major role in providing a long grease life, due to its role in providing excellent sealing. Such sealing may indeed prevent side flow and reduce starvation. This would also explain why some investigators have reported an increase in oil content of grease close to the raceway. At least for line contact bearings it has been clearly proven that grease located under the cage bars is bleeding oil. It is not clear, though, to what extent this bleeding contributes to replenishing the lubricant film between the ineracting surface.
Prediction of film thickness in grease lubricated bearings is very complex. The film is a result of feed and loss mechanisms where bleeding, or grease creep flow, is the feed mechanism and where side flow (starvation), oxidation, polymerization, and evaporation are the loss mechanisms. Side flow may be hindered by the excellent sealing from grease located on the shoulders of the bearing. For some bearing types an additional loss mechanism is formed by pumping, which takes place due to the tangential component of the centrifugal forces on roller and rings. At low shear rates grease creeps, so it may be that grease very slowly flows into the track. An additional complexity is the dynamic behavior of grease lubrication. The grease lubricated bearing shows an inherent “self-healing” mechanism where replenishment may happen due to film breakdown resulting in metal-to-metal contact, local heat development, and release of grease into the raceways. This means that the life of the grease in bearings cannot easily be predicted based on film thickness calculations only. Ultimately, knowledge on multi-phase flow, nonlinear rheology, EHL theory, and chemistry need to be pumping effect, which reduces the available lubricant in the running track significantly. The models are only applicable if the bearing operates in the temperature range for which the grease has been designed. The lower temperature limit is usually determined by the bleeding properties of the grease or the base combined to develop predictive models for grease lubrication in rolling bearings.

Balance between feed and loss of lubricant ultimately determining the lubricant film thickness
Grease life models have been developed mainly by the bearing manufacturers, are mainly empirical; i.e., based on grease life testing. The main parameters determining grease life are ndm, (the product of rotational speed and mean bearing diameter) and temperature. This ndm parameter translates into a peripheral speed. All models assume a temperature-Arrhenius behavior. The effect of load is generally less pronounced, which may be due to the weak relation between load and EHL film thickness.
Deep-groove ball bearings (DGBBs) are “easier” to lubricate than other bearing types. This is related to more pronounced replenishment, or is sometimes ascribed to ball spin. Most models are normalized for this bearing type and correction factors are applied for the other bearing types. In addition to the starvation effect, some bearings, such as tapered and spherical roller bearings, and also angular contact ball bearings, show an inherent pumping effect, which reduces the available lubricant in the running track significantly.
The models are only applicable if the bearing operates in the temperature range for which the grease has been designed. The lower temperature limit is usually determined by the bleeding properties of the grease or the base oil’s pour point, whereas the high temperature limit is determined by the dropping point of the grease; i.e., the point at which a droplet falls from a standardized cup. This point is accepted as the maximum temperature at which the grease can be exposed without losing its structure. For safety reasons this is reduced by 15–20◦C.