As shown in figure left, a Sensitive Volume (SV) is an artificially defined geometry volume inside a device, which is supposed to be sensitive to radiations, as suggested by its name. It is defined that whether an SEE happens is decided by the deposited electric charge within the SV (caused by the incident particle), compared to a certain value (called Critical Charge):
Thus the concept of SV helps to locate the exact place where the SEE happens. Information of SV is one of the ways of describing a device for SEE mechanism research or rate prediction. The comparison of SV with other describing ways and their proper usages are explained here.
For a typical 6-T SRAM cell, there are usually two sensitive parts, as shown in figure right. If any one of these two parts get hit (electric charge exceeds its critical charge), the cell may flip, which is called Single Event Upset (SEU), a common type of SEE.
So there will be two SVs in every one of the cells of this type of SRAM. Notice that these two volumes are not in the same size. Some rate prediction models like IRPP, which works with the assumption that all sensitive parts are the same size, doesn't fit this SRAM.
All parts of an SV are sensitive, but some are more sensitive than others. This phenomenon can be measured by micro beam experiment, and also can be shown by Fully Physical Simulation. Thus a simple cube as shown in the figure at the top of this page is simplified in some degree. An improved version of the simple cube SV is to divide the cube into layers (as shown in figure right), each layer with a certain weight. The total deposited charge Qtotal is the weighted summary of the charges in all layers:
Qtotal = Q1W1 + Q2W2 + Q3W3 + ...
Weighted SV is also called Composite SV in some literature.
DICE or TMR hardened logic devices use more transistors to reduce SEE rate. In these kinds of devices, sensitive volumes are in associating pairs. Only when both SVs in a pair are hit simultaneously, SEE will happen.
For a certain device, to define its SV is a tricky job. It depends on how much do you know about the device and what kind of measurement apparatus do you have: