I've done my best to read through Wikipedia explanations of floating gate transistors which are used in flash memory. It looks like these transistors have a special region where the charge is "stored" and user circuit has to do certain electrical manipulations to "unlock" the region and alter the charge.

Okay, but how does that charge stay there forever such that the transistor has no need for charge "renewal"? AFAIK memory that uses capacitors for data storage has to be "refreshed" because capacitors slowly discharge over time.

How do floating gate transistors store the charge forever and have no need for "refreshing"?


2 Answers 2


The short answer is that they do not store the charge indefinitely, although the data retention time is in decades, usually.

Floating gate devices are charged across a quantum tunnelling oxide at a relatively high voltage. In simple terms. a high voltage is applied to the control gate, and electrons travel across the tunnel oxide to the floating gate. There are a number of subtleties I have not dealt with here, but the basic idea is correct.

In modern parts the charge pump is within the IC, whereas the older crowd will remember devices that had a 'Vpp' pin.

The tunnel oxide is very thin, incidentally. The process node for a device may be 65nm (about the smallest practical node for NOR flash, incidentally) but the tunnel oxide is typically 5nm to 10nm thick.

The charge on the floating gate eventually discharges to the un-programmed state the same way it was programmed; by quantum tunnelling.

Note that floating gate devices are highly susceptible to discharge under X-Ray (a commonly used process in manufacturing to inspect solder joints, particularly under BGA devices), so it is prudent to have in-system programming capability.


  • 1
    \$\begingroup\$ So is it true that charge actually gets leaked, just rather slowly, and so flash memory loses the stored data after some years (dozens of years) of laying idle? \$\endgroup\$
    – sharptooth
    Jun 16, 2015 at 14:11
  • 1
    \$\begingroup\$ That is correct. Other effects can cause loss of data - primarily the programming threshold levels shifting within the device. This can be caused by a number of things. \$\endgroup\$ Jun 17, 2015 at 6:37
  • \$\begingroup\$ Could you please add that into the answer? \$\endgroup\$
    – sharptooth
    Jun 17, 2015 at 13:06

The difference between DRAM (which stores bits in small capacitors, and 'senses' the voltage on those capacitors to read out the bits) and EPROM/EEPROM/FLASH is that the latter uses floating capacitors: capacitors that have one side not connected.

So how to do the read-out of a capacitor without a connection? The capacitor is the gate of a mosfet, and hence the charge on the capacitor/gate influences the conductivity of the mosfet.

And how is the charge on such a capacitor changed? A 'high' voltage is used to tunnel electrons into the gate. This process is somewhat destructive, hence these memories have a limited number of erase/write cycles.

Do such memories store their content forever? Well, yes, but for small values of forever :) Think of ~ 10 years.

  • 3
    \$\begingroup\$ +1 Perhaps worth mentioning that the value of forever also generally halves for every 10 degrees C you heat it. That's one of the few ways retention can be directly verified. \$\endgroup\$ Jun 16, 2015 at 14:34

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.