I'm wondering what is the need for charge pumps in flash memories.

  • 3
    \$\begingroup\$ This question requires a reference. \$\endgroup\$ – Eugene Sh. Oct 3 '17 at 18:00
  • 2
    \$\begingroup\$ Flash memories require a voltage higher than the power supply voltage to produce the tunneling needed to write data to the memory cells. Some devices required an external high voltage supply for writing. \$\endgroup\$ – Kevin White Oct 3 '17 at 18:05

Flash memory requires multiple high voltages for programming and erasing, on the order of 5-15V. In a low-power digital system that's probably running on 3.3V, it's convenient to have the charge pump and regulators built into the memory instead of making the external system provide 3-5 different high voltages. The high voltage supplies don't need to provide much current, and they're integrated into an IC, so a charge pump is the most reasonable converter to use.

If you want to know more about the processes that require the high voltages, look up hot-carrier injection and Fowler-Nordheim tunneling.


In floating gate based memories

Programming in floating gate memories can occur in two ways:

  • hot electron injection (if the cells are arranged with the NOR topology)
  • Fowler-Nordheim tunneling of electrons toward the floating gate (if the cells are arranged with the NAND topology).

Erase, instead, occurs only by Fowler-Nordheim tunneling (of electrons of the floating gate to the substrate).

Hot carriers injection:

This first process requires "medium" voltages at the drain and gate, and relatively high currents (several hundreds \$\mu A\$ per cell). These voltages are still higher than the typical 1.8-3.3 \$V_{DD}\$ supply. Electrons are accelerated by the drain-to-source voltage, to overcome the silicon to silicon dioxide barrier height (about 3.1 eV). The positive gate to channel voltage attracts the electrons toward the floating gate.

Fowler-Nordheim tunneling:

This second process requires a very high positive or negative (for program or erase, respectively) gate-to-substrate voltage (typically 15 V, but this is technology dependent), but the current is very small. Instead of injecting "high energy" (hot) electrons, they just cross the energy barrier (represented by the silicon dioxide). This quantum mechanical effect has a very small probability at low fields, that's why the retention time is typically 10 or more years. It's enhanced at very high fields, and it allows programming the cells in few hundreds of microseconds (up to some ms).

In discrete-storage memories

Some modern FLASH memories do not use floating gate anymore, but a trap-enriched dielectric layer or a layer of discrete nanocrystals.

Electrons are no more stored in a conductive monolithic medium, but in several electrically insulated dots (traps or nanocrystals).

This has the advantage that a single critical defect will discharge only those storage media close to the defect, preserving the stored information (which depends on the threshold voltage, which, in turn, depends on the "average" stored charge). A single critical defect, instead, could discharge the entire floating gate.

The discrete storage approach, thus, allows to reduce the oxide thickness (required for the scaling of the memory cell). The trap-enriched dielectric also brings additional advantages, such as:

  • the ability of creating 3D structures (Samsung use this on NAND FLASH).
  • Being dielectric, no patterning is necessary: unlike the floating gate it can be deposited in the whole area. It simply won't have any effect outside the MOSFET active region.

Still program/erase occur in the same way of floating gate cells.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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