I'd like to discuss a different type of detector based on a photodiode. This is based on a reverse biased p-n or p-i-n junction. These devices are attractive due to high quantum efficiency and bandwidth. If we think about a p-n junction, we can draw the I-V characteristics. So, here if I plot the I-V characteristics of a p-n junction. It looks something like this. So, when V is greater than zero for a forward bias, and a forward bias of current is dominated by diffusion because the bias lowers the potential barrier between the p and the n regions. In reverse bias, we increase the potential barrier, and so this chokes off the diffusion current. The reverse bias is dominated by the generation recombination current. The generation of excess carriers in the depletion region enhances the reverse bias photocurrent. We can use this to reverse bias photocurrent for photodetection. Essentially, as we apply a reverse bias, we increase the width of the depletion region, which leads to enhanced response of the light. So, the way to think about this is here's your p-n junction. Remember, all the action really happened right at the vicinity of the p-n junction. So, essentially here's my depletion region, and if I apply more reverse bias, maybe I get a depletion region that looks like this. Okay. So, really the key point is that the light really needs to be absorbed in the depletion region. We can get around this problem nicely and basically improve our device by going to a p-i-n structure, where essentially i is an intrinsic region, and now all of a sudden we essentially have this region where we can have light absorbed. So, the electrons and the holes that are generated are separated by the action of the field across this depletion region, and when the reverse bias is large enough, the carriers acquire enough velocity to undergo Impact Ionization. As I mentioned, all the action really happens in the depletion region. So it's really the carriers in the depletion region that contribute to the current. So, the bandwidth of the photodiodes is inversely proportional to the temporal response, and it's affected by three main factors. One is the carrier diffusion towards the depletion region edge. If the carriers are generated outside the depletion region, they must first diffuse to reach the depletion region edge. Once they reach the depletion region, they're swept away by the electric field. The second thing that limits the temporal response is the drift time within the depletion region. It depends on the velocity of the carriers and the width of the depletion region. So, in order to enhance absorption in the depletion region, you have to make it large, and so this will lead to long drift times and decrease the bandwidth. For this reason, the fast photodiodes like sensitivity. The third thing to think about and that affects the bandwidth is the capacitance of the depletion region. So, we can write down an expression for the junction capacitance. The junction capacitance is nothing more than Epsilon_s A over W, where W is the width of the depletion region. W is equal to X_n plus X_p, which is equal to the square root of Epsilon_s over q, charge on the electron Vbi, built-in voltage minus V, the applied bias, and then the doping NA plus ND over NA ND. So, essentially from this, we can write down an expression for the capacitance. If we assume that C0 is the capacitance at zero bias, we end up with C equals C0 over one minus Vbi over V to the mth power, where m is somewhere between one-third to one-half. So one-half corresponds to an abrupt junction and one-third corresponds to a graded junction. So, an addition, you can see that a large reverse bias will increase the junction capacitance and slow down the temporal response. So, photodiodes can be a very attractive choice, but again, you really have to be aware of the design trends to get what you want. If you want a fast photodiode, you will sacrifice responsivity. If you want something that's very responsive then it will be slow.