Flash memory systems store information as one and zeros on arrays of floating gate transistors. Before we get into the details, this means that the information is stored in electrical form: the presence or absence of electrons on a certain part of the transistor (the floating gate) change its conducting properties. Since the conductivity can be assessed by putting a voltage across the transistor and seeing how much current goes through it as a result (Ohm’s law), this means the presence or absence of electrons on the floating gate can be determined. This is the physical basis for representing information (1 or 0) with flash memory.
MOSFET transistor
You should read the long section on MOSFET transistors to fully appreciate how the normal transistor works. You can see a schematic in the figure on the right.
 MOSFET schematic |
In brief, the current flows between the source and drain through the channel. But how much current flows is controlled by the voltage one applies to the gate. By changing the voltage on the gate, one can change the conductivity of the channel and thus turn the transistor on or off (i.e. make it conduct well or poorly an thus make for a large or small current from source to drain). Physically, the electric field created by the gate repels or attracts the conducting charges in the channel: hence the “field-effect” part of the MOSFET name. Basically, we have an elementary switch.
Most critically, the MOSFET device is volatile: it forgets its state if power is removed and the gate voltage is not maintained. If the gate voltage is removed, the transistor forgets that it was applied. Therefore, as such, the MOSFET can not be used to store information permanently (or semipermanently) without continuous power. This is where the floating gate transistor comes in!
Floating gate transistor
 Floating gate transistor schematic (from http://daedalus.caltech.edu/~julie/research.htm) |
To the right is a schematic image of a MOSFET floating gate transistor. The only difference with the standard MOSFET is the addition of a new gate, called the floating gate, between the original gate and the channel. The original gate (topmost) is now called the control gate. The floating gate is an isolated conducting island: it is surrounded on all sides by oxide insulator. But the transistor is operated (mostly) in the standard way in that source-drain current is monitored and it is controlled by the control gate voltage.
What this means is that electrons that are put onto the floating gate can stay there for a long time, typically years for normal operating conditions, as there is no direct path for them to flow away from the floating gate (see below for why they leave at all). This means that one can use the charge state of the floating gate as an information bit (a zero or one): presence or absence of electrons on the floating gate is the binary bit. But having information is not the same as being able to read it out… So how is this accomplished?
 p-doped floating gate transistor: no control gate voltage and no electrons on the floating gate |
 p-doped floating gate transistor: no control gate voltage but electrons are placed on the floating gate; this attracts the holes |
Notice that the floating gate is placed between the control gate and the channel. Therefore, when extra electrons are present (or absent) from the floating gate, their presence and thus the electric fields and forces they create modify the action of the gate voltage onto the channel. For example, as shown on the right in the figure panel above, when electrons are present on the floating gate, they attract the holes in the p-type channel and body to the channel region right below the floating gate. The enhanced density of holes changes the conductivity of the channel region (see the description of MOSFET transistors for more details.) Or viewed differently, the electrons on the floating gate change the value of the gate voltage needed to achieve some desired channel conductivity. When the floating gate is devoid of electrons, shown to the left above, the control gate voltage will need a more negative value to get the same conductivity. This entire setup translates the charge state of the floating gate into a source-drain current level that can be measured. With proper choices of voltages on the control gate and charges on the floating gate, one can ensure a strong or weak current depending on the charging state of the floating gate. Therefore by measuring the conductivity (source-drain current) one can infer whether electrons are on the floating gate or not.
How do the electrons get on and off the floating gate, or equivalently how can one write and erase the information? To get the electrons off the floating gate (erase mode) involves the curious process of quantum tunneling. The region between the floating gate and the channel is insulating: that means that electrons in the floating gate must overcome an energy barrier to make it across (“jump over”) to the channel and they do not have the requisite energy. However, in quantum mechanics, electrons have wave behavior meaning that they can always “leak” away over barriers; the probability is small for a high barrier but is never zero; and the probability can be enhanced by lowering the barrier.
 p-doped floating gate transistor: negative control gate voltage pushes electrons in floating island away, reduces barrier to tunneling (indicated by purple tunneling process for an electron) |
So there is always a current of electrons that leave a charged floating gate — it is just that under normal conditions the current is so slow that the floating gate takes years to discharge by quantum tunneling. However, the process can be greatly accelerated by lowering the energy barrier in the insulator. This is accomplished by putting large negative voltage on the control gate so that electrons in the floating gate are repelled from above and attracted to the channel region below: this greatly reduces the barrier and strongly enhances the tunneling (the oxide is still an insulator but just an insulator presenting a lower barrier). In this way the floating gate can be discharged in a short time.
Charging the floating gate is accomplished by either tunneling in the other direction (so one reverses the voltages) or via hot-electrons injection: here electrons that have enough kinetic energy (“are hot”) can jump across from channel to floating gate. To get hot electrons that move fast enough, one must put a relatively large voltage between source and drain to sufficiently accelerate the electrons going through the channel region.