No Thumbnail Available
Files
Vandermolen_74841600_2017.pdf
Closed access - Adobe PDF
- 5.67 MB
Details
- Supervisors
- Faculty
- Degree label
- Abstract
- For many years the transistor dimensions were reduced at each technology generation to increase the circuit performances. However, the transistors have reached dimensions so small that certain hypotheses initially made on the electronic transport inside the channel of the transistor have to be questioned. On the other hand, the computation of the effective mobility is of the utmost importance since it allows to know the quality of the transport inside those devices. The goal of this work was to extract the effective mobility of those transistors from S-parameters measurement, which are known to be free of the self-heating effect for sufficiently high frequencies, while taking into account the modifications on the electronic transport, known as the ballistic effects. Two transistor architectures were tested: the FDSOI and the LP bulk nMOSFETs. To do so, an analytical model taking into account ballistic effects was elaborated. The effective mobility was reconstructed from the small-signal circuit elements of the transistor, extracted during the S-parameters measurement. From those results, it was seen that the ballistic effects become important when the channel length is lower than 100 nm. In that case, the error on the effective mobility by ignoring ballistic effects is equal to 10% and reaches 20% for a channel length of 25 nm. Concerning the errors made on the extraction of the mobility, three sources of errors have been emphasized: errors on the computation of the effective channel length, the series resistances and the inversion charge. By computing plausible errors for those elements, it was shown that the error made on the extraction of the effective mobility due to those elements varies from 17 to 30%. Finally, it was pointed out that, in order to have more reliable values for the effective mobility, the extraction method for the series resistances needs to be improved and the parasitic contribution of the inner fringing capacitance has to be minimized for the computation of the inversion charge.