Desktop version

Home arrow Computer Science arrow Power Microelectronics. Device and Process Technologies

Source

Simulation and Process Parameter Determination of PFVDMOS Device

Although process flow gives individual process steps, practical process parameters to achieve the desired doping profiles and oxide thicknesses can only be obtained through simulations. Process simulation using TSUPREM4 (Synopsys) was first carried out using the above process flow. Then, device simulator MEDICI (Synopsys) was used to obtain both on- and off-state characteristics of PFVDMOS. The input file with explanations is first described to indicate model choice and input deck setup followed by results of these studies.

Process Parameters and Doping Profiles

The novel device as proposed in Fig. 10.1 was simulated. The device was designed to support a breakdown voltage of 250 V. The smallest possible n- and p-layer width, that met the requirement that all the source contacts, channels must be seated inside the n-epi layer of 5 дш, respectively was chosen. Using the superjunction theory equations given in Chap. 6, the doping concentration in p- and n-layers were calculated to be at 7 x 1015 cm-3. Using these values, both process and device simulation (Synopsys) were used to study and compare the DC characteristics of PFVDMOS and conventional DMOS.

This is the input file for process simulator with detailed explanations. Important process steps were optimized by changing the process parameters to obtain the best results. The simulation commands are in capital letters and descriptions are in mixed case.

$ TSUPREM4 - PFVDMOS Process Simulation file. Time is in minutes and temperatures in degree centigrade.

PFVDMOS specification (P Poly: Dose = 8E12, tilt = 10; P Poly drive-in: time = 500, temp = 1150; Source implant: Energy = 200)

$ Specify x mesh. In the ohmic n+ region, specify large spacing of 1 дш. In the device region, specify spacing of 0.25 ^m for a reasonable grid. Since program has adaptive triangular grid, it will introduce grid points where there is a lot of change. Y-direction is all ohmic n+ drain region, so the spacing is large.

LINE X LOCATION = -5 SPACING = 1.0 LINE X LOCATION =-3.75 SPACING = 0.25 LINE X LOCATION = 0 SPACING = 1.0

LINE X LOCATION = 3.75 SPACING = 0.25

LINE X LOCATION = 5.0 SPACING = 1.0

LINE Y LOCATION = -3.5 SPACING = 1.0

LINE Y LOCATION =-3 SPACING = 1.0

$Initialize the structure (n+ sub) with low resistivity of 0.02 G ? cm. INITIALIZE PHOSP = 0.02 RESIST

$ n-epi deposition for n-type superjunction drift region. The region is phosphorus doped with desired concentration and has thickness of 18.5 дш and 40 grid spaces within. The thickness is finalized at the minimum value at which desired breakdown voltage can be attained from device simulator MEDICI.

DEPOSIT SIL THICK = 18.5 SPAC = 40 PHOS = 7E15 DY = 0.1 $ Masking Oxide

DEPOSIT OXIDE THICK = 0.5 SPAC = 2

$ p-type Polysilicon Trench definition. First oxide and then epi-silicon is etched to create this trench. Structure is ideal as TSUPREM4 has only simple lithography, etching, and deposition models.

ETCH OXIDE START X =-2.5, Y =-30

ETCH CONTINUE X =-2.5, Y =-7.0

ETCH CONTINUE X = 2.5, Y =-7.0

ETCH DONE X = 2.5, Y =-30

ETCH SIL START X =-2.5, Y =-30

ETCH CONTINUE X =-2.5, Y =-7.0

ETCH CONTINUE X = 2.5 Y =-7.0

ETCH DONE X = 2.5 Y =-30

$ Plot grid and profiles for complete structure PLOT.2D SCALE GRID

$ Sidewall oxidation to grow oxide of sufficient thickness that acts as an interdiffusion barrier. Since edges are straight, compress model is adequate as there is no nitride. If stress calculation is needed, VISCOELA (Viscoelastic) model can be used.

METHOD COMPRESS

DIFFUSE TIME = 100 TEMP = 1000 DRYO2

$ Etch bottom oxide. This removes only the oxide at the bottom of the trench completely. Original top oxide being thicker, the sidewall oxide is not affected as the etch is only in the vertical direction.

ETCH OXIDE THICK = 0.1

$ Poly deposition of 1 ^m thickness which does not fill the trench. This is done so that poly can be doped in the depth of the trench using tilted implant later. If trench is filled at this stage, the p-type dopants will be only in the top region and may not be able to reach the bottom poly even after prolonged diffusion. DEPOSIT POLY THICK = 1 SPAC = 5

$ Plot grid and profiles for complete structure PLOT.2D SCALE GRID

$ P-poly Boron tilted implant to dope it with the right dose to give correct final boron concentration. Dose may need to be optimized. To avoid shadow region issues with 11°, implant, use accurate Monte Carlo model that individually tracks each ion profile.

IMPLANT BORON DOSE = 8E12 ENERGY = 90 TILT = 11 MONTE N.ION = 10000

IMPLANT BORON DOSE = 8E12 ENERGY = 90 TILT =-11 MONTE N.ION = 10000

$Save intermediate output file so that simulation can be resumed from this point. This is needed to find adequate diffusion time that makes boron concentration uniform in p-type poly-trench in the final structure.

SAVEFILE OUT.FILE = PPOLY

$ Now do isotropic poly deposition to fill up the trench. Dopants in previous 1 ^m poly are distributed everywhere uniformly in the next thermal cycle. DEPOSIT POLY THICK = 3 SPAC = 10

$ Cap oxide deposition to prevent out-diffusion of dopants toward surface. DEPOSIT OXIDE THICK = 0.5 SPAC = 2

$ P-poly drive-in to achieve dopant uniformity. Basically, diffusion is long enough to remove any concentration gradients. Since there are no concentration gradients, it becomes uniform. Minimum time to achieve this must be selected as n+ drain region will diffuse into n-epi reducing its effective thickness if the time is too long. The simplest model with simple treatment of point defects is used (PD.FERMI). This is adequate as diffusions are long enough that point defects reach their equilibrium concentration quite early. For short diffusions, PD.TRANS should be used, and for VLSI devices with ultra-shallow junctions, PD.FULL/ACT.FULL should be used for high accuracy.

DIFFUSE TIME = 500 TEMP = 1150

$ Now the top cap layer of oxide is etched so that p-body and source regions can be formed later.

ETCH OXIDE THICK = 0.7

$ Etch top poly-layer so that poly is only left in the trench.

ETCH POLY THICK = 4.5

$ Etch original masking oxide that was used to form trench.

ETCH OXIDE THICK = 0.7

$ This completes the formation of p- and n-alternating columns.

$ Plot grid and profiles for complete structure. Different materials are given different colors. 2D doping contours are plotted between 1014 and 1018 with five lines in each decade of concentration.

SELECT Z = DOPING TITLE = “MOSFET”

PLOT.2D SCALE

COLOR SILICON COLOR = 7

COLOR OXIDE COLOR = 5

COLOR POLY COLOR = 3

COLOR ALUMINUM COLOR = 2

FOREACH X (14 TO 18 STEP 0.2)

CONTOUR VALUE = (10л X) COLOR = 4 LINE.TYP = 2 CONTOUR VALUE = (-(10" X)) COLOR = 2 LINE.TYP = 5 END

$ Vertical profile at X = 1 г-m is plotted to check uniformity near the center of poly. To avoid interpolation, it is better to use grid line at x = 1 instead of using any x.

SELECT Z = LOG10(doping) TITLE = “Vertical Profiles at x = 1” LABEL = “Log(conc)”

PLOT.1D X.V = 0 COLOR = 2

$ Again save profile so that simulation can be resumed without doing 5 hours diffusion simulation.

SAVEFILE OUT.FILE = PDIFF

$ Simulate only half of the structure. This is because there are two DMOS devices symmetrically placed around the center at X = 0. No need to simulate both. This saves half simulation time.

STRUCT TRUNCATE LEFT X = 0

METHOD COMPRESS

$ Grow gate oxide. Time is decided by the desired thickness and ambient of dry oxygen is dictated by high gate oxide breakdown voltage needed in power devices and good quality of the oxide.

DIFFUSE TIME = 76 TEMP = 1000 DRYO2

$0.6 г-m Gate polysilicon deposition and pattern using gate mask. Poly remains to the right of source/p-body region. Poly should be thick enough to block source implant so that only lateral source profile defines the channel length. Also, this assures that channel length of the device is the same despite misalignment of gate poly mask. Source and p-body implants are automatically self-aligned with the poly gate edge as resist defines poly feature.

DEPOSIT POLY THICK = 0.6 SPACES = 1 ETCH POLY P1.X = 3.5 LEFT

$ Plot grid and profiles for complete structure PLOT.2D SCALE GRID

$ Since p-body implant is deeper, it is better to leave resist on top of the poly. Hence this step is added to emulate resist as there is no detailed lithography simulation in TSUPREM4. Gate mask is patterned.

DEPOSIT PHOTORESIST THICK = 1.2 SPACES = 2 ETCH PHOTORESIST LEFT P1.X = 3.5

$ P-body implant with boron at high energy and reasonable dose is done. Dose is determined by the level of p-type doping needed to get desired MOS threshold voltage.

IMPLANT BORON DOSE = 4E13 ENERGY = 90

$ Remove resist before anneal to drive-in boron as the resist cannot stand high temperature.

ETCH PHOTORESIST ALL

$ P-body boron drive-in is now done to anneal implant damage and get enough lateral diffusion of boron in the n-epi to get reasonable channel length of the device.

DIFFUSE TIME = 20 TEMP = 1100 SAVEFILE OUT.FILE = PBODY

$ Source resist deposit and pattern is done so that the implant does not enter other regions not in the device. The mask opens poly in some area so that alignment of source to the gate edge is assured.

DEPOSIT PHOTORESIST THICK = 1.2 SPACES = 2

ETCH PHOTORESIST START X = 2.75 Y = -30 ETCH CONTINUE X = 2.75 Y =-3.0

ETCH CONTINUE X = 3.5 Y =-3.0

ETCH DONE X = 3.5 Y =-30

ETCH OXIDE START X = 2.75 Y =-30

ETCH CONTINUE X = 2.75 Y =-3.0

ETCH CONTINUE X = 3.5 Y =-3.0

ETCH DONE X = 3.5 Y =-30

$ Source implant is now done with high energy and high dose to give low resistivity source region.

IMPLANT AS DOSE = 5E15 ENERGY = 200

$ Remove resist before anneal.

ETCH PHOTORESIST ALL

$ Source anneal to cure implant damage.

DIFFUSE TIME = 30 TEMP = 900

SAVEFILE OUT.FILE = VDMOS0

$ 1 ^m thick BoroPhospoSilicateGlass is deposited and contact holes are now etched to connect all source, p-poly and p-body regions using aluminum. BPSG reflow step can be added here if needed to planarize the structure.

DEPOSIT OXIDE THICK = 1 ETCH OXIDE START X = 0 Y =-30

ETCH CONTINUE X = 0 Y =-21.75

ETCH CONTINUE X = 2.9 Y =-21.75

ETCH DONE X = 2.9 Y =-30

$ Aluminum metallization to connect regions.

DEPOSIT ALUMINUM THICK = 1.7

ETCH ALUMINUM RIGHT P1.X = 3.5

$ Passivation oxide is now deposited. Since only device is simulated here, bonding pad formation steps can actually be skipped here as pads never overlap device areas.

DEPOSIT OXIDE THICK = 0.1 SPAC = 1 SAVEFILE OUT.FILE = VDMOS1

$ Reflect to obtain complete structure so that full structure with both DMOS devices is generated.

STRUCT REFLECT RIGHT

$ Define electrodes. A single point electrode is adequate as region connected to the point will become an electrode. Since metal connects to two poly sidewalls at X = 4.25 ^m and X = 14 ^m at oxide-poly interface, put an electrode there. Also, put source electrode at the same level at the left end where Al connects source. All of Al automatically becomes source. Also connect p-poly column to source and p-body at X = 8 and X = 18. Add drain contact at the bottom n+ region.

ELECTROD NAME = Gate X = 4.25 Y =-22.5 ELECTROD NAME = Gate X = 14 Y =-22.5

ELECTROD NAME = Source X = 8

ELECTROD NAME = Source X = 0 Y =-22.5

ELECTROD NAME = Source X = 18

ELECTROD NAME = Drain BOT

$ Final device details are saved for further device simulation in MEDICI. SAVEFILE OUT.FILE = VDMOS.spu4 MEDICI ELEC.BOT

$ Plot grid and profiles for complete structure

SELECT Z = DOPING TITLE = “SOI-MOSFET”

PLOT.2D SCALE

COLOR SILICON COLOR = 7

COLOR OXIDE COLOR = 5

COLOR POLY COLOR = 3

COLOR ALUMINUM COLOR = 2

FOREACH X (14 TO 21 STEP 0.2)

CONTOUR VALUE = (10л X) COLOR = 4 LINE.TYP = 2

CONTOUR VALUE = (-(10" X)) COLOR = 2 LINE.TYP = 5 END

PLOT.2D GRID SCALE C.GRID = 2

The simulated final doping concentrations of both conventional VDMOS and PFVDMOS drift region of 18 дш thickness are shown in Fig. 10.3. From the figure, it is clear that the doping concentration of p- and n-layers in PFVDMOS are matched around the value of 7 x 1015 cm-3. The p-poly layer concentration is uniform in the region of interest due to long thermal cycle used to remove any concentration gradients. The “bulge” in the doping profile near the edge of the p-poly actually arose due to the difference in incident angles of the tilted implants on the sidewall and the bottom wall. From the figure, it is also apparent that the doping concentration of the n-drift region in PFVDMOS was nearly an order of magnitude higher than the n-drift region in conventional VDMOS (7 x 1015 cm-3 vs 8 x 1014 cm-3). This would obviously translate into much reduced Ron in PFVDMOS structure. Clearly, process uncertainties can give rise to variations in these uniform concentrations. However, these should be less than 10% as both the epi-growth and implant dose are fairly well controlled. After achieving the desired doping profiles using process simulators such as TSUPREM4 or ATHENA, the next step is to evaluate off-state and on-state performance of the device using device simulators such as MEDICI or ATLAS (Synopsys) (Silvaco).

Doping profiles in conventional DMOS and novel PFVDMOS device

Fig. 10.3. Doping profiles in conventional DMOS and novel PFVDMOS device.

 
Source
Found a mistake? Please highlight the word and press Shift + Enter  
< Prev   CONTENTS   Next >

Related topics