90-nm-Device Topology Description:

This is a 90 nm Leff NMOSFET in bulk silicon with super steep retrograde (SSR) channel doping and source/drain halo. The process technology has been described by Hu et al. in [1].

Refer to the SCHEMATIC image for a two dimensional cross section of the fabricated device. The n+ polysilicon height is 300 nm. The physical tox is 4.5 nm and the active doping is 7e19 cm-3, both extracted using the method in [2]. Also included is a gate to bulk CV_PLOT image and digitized data of a 49.3 µm wide and 5.08 µm long channel device from the same lot showing the model fit to the data. Leff (as defined where the source and drain dopings fall to 2e19 cm-3) was extracted from inverse modeling and found to be about 90 nm. We have correlated Leff from the Cgds method [3] and found the source/drain overlap under the gate, delta-L, to be 46 nm; that is, there is 23 nm of overlap on each of the source and drain sides. Thus, Lpoly = Leff + delta-L is about 136 nm.

Extensive inverse modeling using the subthreshold I-V curves of this device was performed using the method in [4]. Because the device is symmetric, the origin for the lateral grid is at the middle of the channel (x = 0); the interface between the gate oxide and bulk is chosen as the depthwise origin (y = 0). In the simulations, only the portion of the MOSFET up to the source/drain extension regions for a lateral distance, Lsd = 92 nm, from edge of gate to model boundary is used. The source and drain contacts are one- node-thick electrodes at the top of the silicon over a length of Lsd/2 from the simulation edge. The particular electrode to silicon contact resistances for this model were assumed to be zero. For comparison between simulated and measured data at high currents, external source/drain lumped resistances of 95 Ohm-µm per side should be added in the simulation. This is because the simulation does not include the complete source/drain contact, including the correct electrode to silicon contact resistance. The above values were fitted by inverse modeling as described in [5] and [6].

A 2-D doping profile (files are placed in the doping directory) was extracted using inverse modeling as in [4]. The doping profile is given in three files that can easily reconstruct a MEDICI [7] doping grid; i.e., they are all on the same x-y grid. The file ssr.doping contains a 1-D profile in the depth or y-direction only and represents the channel dopings Na(y) which is the sum of a SSR plus a uniform background of 3.4e15 cm-3. Two files, source.doping and drain.doping, are mirror symmetric about x = 0 and contain the combined channel halo and source/drain doping profiles Nd(x,y) - Na(x,y). For reference a sample MEDICI file inputfile90 is also included.

The experimentally measured Id vs. Vgs curves (PREVIEW) are stored in the idvg directory. The device width is approximately 9.4 µm, but the values of current have been normalized per micron. There are a variety of I-V characteristics text files labeled according to their drain bias and bulk bias; for example, d0.01bm2 has a Vds of 0.01 V and a Vbs of -2 V ("m" is for minus), and d2.01bp04 has a Vds of 2.01 V and a Vbs of 0.4 V ("p" is for positive). Each file has an initial header row and then two columns of data: the first is Vgs in 0.1 V steps and the second is Id (A/µm).

Additionally, there are experimental data for drain, gate, and substrate current in the ivd directory. The data come from a W = 49.4 µm device with the same nominal Leff, but are also normalized by W. Although this is a different device from the inverse modeled one, their normalized Id-Vd-Vg characteristics agree to better than 10%. So these data can be safely used for simulation evaluation using the extracted doping profiles. Files are labeled according to the gate bias used (which varies from 2.01 V to 3.01 V); here, Vbs = 0 V. The first row is a header; the first column gives values of Vds which vary from 0.01 V to 3.01 V in 0.1 V steps. The subsequent columns represent normalized current data (in A/µm): the second column is Igate, the third is Idrain, the fourth is Isubstrate.


[1] H. Hu et al., IEEE TED-42(4), p. 669, 1995
[2] R. Rios et al., IEDM proceedings, p. 613, 1994
[3] C. Huang et al., IEEE TED-43(6), p. 958, 1996
[4] Z. Lee et al., IEDM proceedings, p. 683, 1997
[5] Z. Lee, Ph.D. Thesis, MIT, 1998
[6] Z. Lee et al., "Two-Dimensional Doping Profile Characterization of MOSFETs by Inverse Modeling using I-V Characteristics in the Subthreshold Region," IEEE Tran. Electron Devices, to appear, 1999
[7] MEDICI manual, Technology Modeling Associates