main menu | module menu | << previous section | next section >>

Sentaurus Device
7. Sentaurus Device at Runtime

7.1 Running Sentaurus Device
7.2 Log File
7.3 Version and Host Information
7.4 Active Math Parameters
7.5 Active Physical Models
7.6 Device Structure Files
7.7 Box Volume Statistics
7.8 List of Changed Parameters
7.9 Solve Report
7.10 Runtime Statistics

Objectives

To run Sentaurus Device and to understand its output.

7.1 Running Sentaurus Device

To start the Sentaurus Device default version from the command line, use:
```
> sdevice [command_file_name]
```
where the optional command_file_name is a valid command file of Sentaurus Device. For example:
```
> sdevice pp1_des.cmd
```
To list the Sentaurus Device command-line options, use:
```
> sdevice -h
```
To list all of the installed Sentaurus Device releases, use:
```
> sdevice -releases
```
To check the release version, use:
```
> sdevice -v
```
To run Sentaurus Device from a particular release, use:
```
> sdevice -rel T-2022.03
```

7.2 Log File

For each simulation, Sentaurus Device opens and exports to a log file all the runtime information relevant to the simulation. The file is named after the input command file with the extension .log. The information includes the Sentaurus Device version and the host machine, the device structure definition files, the active models and parameters applied, the solution report, and possibly error messages.

7.3 Version and Host Information

The Sentaurus Device version and host machine information is recorded at the beginning of the log file:

****************************************************************************
***                           Sentaurus Device                           ***
***                          Version U-2022.12                           ***
***                      (0.6958654, x86_64, Linux)                      ***
***                                                                      ***
***                       Copyright (C) 1994-2022                        ***
***                            Synopsys, Inc.                            ***
***                                                                      ***
***  This software and the associated documentation are confidential     ***
***  and proprietary to Synopsys, Inc.  Your use or disclosure of this   ***
***  software is subject to the terms and conditions of a written        ***
***  license agreement between you, or your company, and Synopsys, Inc.  ***
****************************************************************************

Running on machine with the following configuration:
Host Name: blaisun1
Operating System: Linux rel. 3.10.0-1127.10.1.el7.x86_64 ver. #1 SMP Wed Jan 5 \
  14:28:03 UTC 2022
Machine Type: x86_64
Process ID: 26284
Number of processors: 20
Domain Name: synopsys.com
Date: Mon Feb 14 11:39:15 2022  (CEST)

7.4 Active Math Parameters

The active math parameters can be found under "Numerical parameters" in the log file. The parameters define the error limits and error reference values used in the simulation:

Numerical parameters:
  Absolute error : 
    Poisson : 1.0000e-03
    eqPoisson : 1.0000e-03
    Electron : 1.0000e-05
    Hole : 1.0000e-05
    Photon-Rate-Equation : 1.0000e-07
    Photon-Phase-Equation : 1.0000e-05
    Rate-Stabilize-Equation : 1.0000e-07
    QW-Scattering-Equations : 1.0000e-05
    QW-eScattering-Equation : 1.0000e-05
    QW-hScattering-Equation : 1.0000e-05
    Optical Problem : 1.0000e-03
    Wavelength : 1.0000e-03
    Bandstructure : 1.0000e-03
...

7.5 Active Physical Models

Information about the physical models that were applied follows the "Physical models" record:

Physical models:
With SRH-Recombination
Without field dependent lifetimes
With doping dependent lifetimes
With temperature dependent lifetimes (power law)
Without thermal resistance interfaces
Without distributed resistance interfaces for electrons
Without distributed resistance interfaces for holes
Fermi Statistic
Intrinsic density models:
default bandgap model
Bandgap narrowing model: OldSlotboom with bandgap narrowing (Fermi)
Electron mobility:
Doping dependence:
Default
...

7.6 Device Structure Files

Information about the loaded device structure file is reported under the "Reading grid" record:

Reading grid 'n1_msh.tdr' ...
coordinate system: UCS, 3d_sprocess (x is device down direction)
use coordinate system as is (no transformation)
TDR format
Number of grid points is 3548.
done.

7.7 Box Volume Statistics

The box method volume statistics reported under the "non-Delaunay elements" record helps to analyze the mesh quality in terms of the number of poor-quality mesh elements and a BoxMethodVolume deviation from an actual physical device volume:

/-------- Region non-Delaunay elements \
  ------------------------------------------
Region                  Volume      BoxMethodVolume  DeltaVolume  Elements \
  non-Delaunay      non-DelaunayVolume
name                    [um2]          [um2]           [%] \
  Elements              [um2]   [%]
------------------------------------------------------------
R.Gateox             2.0400000e-04  2.0400000e-04     3.7e-13       664 \
  0 ( 0.00 %)   0.0000000e+00 ( 0.0000)
R.PolyReox           1.0000000e-03  1.0000000e-03     8.5e-14       157 \
  0 ( 0.00 %)   0.0000000e+00 ( 0.0000)
R.PolyReox_mirrored  1.0000000e-03  1.0000000e-03     1.2e-13       157 \
  0 ( 0.00 %)   0.0000000e+00 ( 0.0000)
R.Substrate          3.7000000e-01  3.7000000e-01     2.4e-12      4706 \
  0 ( 0.00 %)   0.0000000e+00 ( 0.0000)
R.Polygate           5.0000000e-03  5.0000000e-03     1.2e-12       742 \
  0 ( 0.00 %)   0.0000000e+00 ( 0.0000)
R.Spacer             4.4956237e-03  4.4956237e-03     4.3e-13       237 \
  0 ( 0.00 %)   0.0000000e+00 ( 0.0000)
R.Spacer_mirrored    4.4956237e-03  4.4956237e-03     5.6e-13       237 \
  0 ( 0.00 %)   0.0000000e+00 ( 0.0000)
............................................................
Total                3.8619525e-01  3.8619525e-01     1.9e-13      6900 \
  0 ( 0.00 %)   0.0000000e+00 ( 0.0000)
\------------------------------------------------------------
done.

7.8 List of Changed Parameters

New model parameters other than the defaults are reported under the "Reading parameters" record:

Reading parameter file 'pp2_des.par' ...
---------------------------------------------------
 Reading parameters for material "Silicon"
---------------------------------------------------

	Differences compared with default parameters:
	Scharfetter(elec): tau_max = 3.0000e-08, instead of: 1.0000e-05 [s]
	Energy relaxation time: tau_w_ele = 0.2, instead of: 0.3 [ps]
	Energy relaxation time: tau_w_hol = 0.2, instead of: 0.25 [ps]

...

7.9 Solve Report

This section discusses the different information available in the solve report.

7.9.1 Stepping Information

In Quasistationary sweeps, the t value is the virtual time, ranging between 0 and 1, that represents how much of the sweep goal has been reached. For example, t=0.352967 indicates that approximately 35% of the goal has been completed. In Transient sweeps, however, the t value refers to the real time measured in seconds:

Computing step from t=0.0000e+00 to t=0.1 (Stepsize: 0.1) :

The printed Stepsize is in terms of t.

7.9.2 Equations List

When solving the task for each time step, the following report is produced:

Computing step from t=0.0000e+00 to t=0.1 (Stepsize: 0.1) :
Computing Coupled( 5 circuit-equation(s) , 1 poisson-equation(s) , 
                   1 current-contact-equation(s) , 1 electron-equation(s) , 
                   1 hole-equation(s) ) 
using Bank/Rose nonlinear solver.

Iteration   |Rhs|      factor     |step|     error   #inner  #iterative  time
------------------------------------------------------------------------------
    0      3.73e+10                                                      0.07
    1      5.84e+09   1.00e+00   3.33e+00   1.14e+07    0         5      0.23
    2      4.41e+08   1.00e+00   9.62e-01   9.43e+04    0         5      0.38
    3      8.30e+07   1.00e+00   6.51e-01   9.95e+05    0         5      0.54
    4      6.08e+04   1.00e+00   3.53e-01   1.08e+04    0         5      0.69
    5      1.30e+00   1.00e+00   1.41e-05   7.33e-01    0         5      0.84
Finished, because...
Error smaller than 1 ( 7.3312E-01 ).

Accumulated times:
Assembly time:        0.53 s
Solve time:      0.30 s
Total time:      0.84 s

Device Instance 'nmos1':
Contact              Voltage              Electron      Hole         Conduction 
                 outer       inner        current       current      current
 drain        0.000E+00    0.000E+00      5.378E-16     9.642E-28    5.378E-16
 gate        -3.000E-01   -3.000E-01     -4.627E-36     4.627E-36    0.000E+00
 source       0.000E+00    0.000E+00      5.378E-16    -7.889E-29    5.378E-16

where:

The Iteration column counts the number of Newton iterations.
The |Rhs| column shows the evolution of the normalized residual right-hand side. For good converging iterations, the |Rhs| value becomes smaller with each iteration.
The factor column shows the applied damping factor. For most cases, damping is not recommended. Ensure that, in the Math section, the selected number of NotDamped iterations is greater than the number of Iterations (see Section 1.2.5 Math Section).
The |step| column shows the normalized step size between two Newton iterations. For good converging iterations, the step size decreases steadily.
The error column shows the normalized error estimate. For good-converging iterations, the error decreases steadily.
The #inner and #iterative columns are relevant for plugin-type iterations only, which are usually unnecessary.
The time column shows the accumulated computation time in seconds.

The Newton iterations are considered to have converged if the normalized error estimate becomes smaller than 1 or the normalized residual |Rhs| value becomes smaller than 10^-5. This value can be changed using the keyword RhsMin in the Math section.

Newton iterations are considered to have failed to converge if any of these conditions are met:

The number of iterations exceeds the preset Iterations value (see Section 1.2.5 Math Section).
The right-hand side increases by more than a factor of 1e10. This value can be changed in the Math section using the keyword RhsFactor.
During the iterations, the lattice or carrier temperature at a node becomes higher than a certain upper limit.

If the Newton iterations failed to converge, the step size is reduced by default by a factor of 2. The step size is reduced further as needed, until the step size becomes smaller than the minimum step size (see Section 1.2.6 Solve Section). Then, the Quasistationary or Transient process terminates.

Sentaurus Device will start the next command in the Solve section, unless the flag ExitOnFailure is set in the Math section (see Section 1.2.5 Math Section). In this case, Sentaurus Device terminates the entire run.

If a Coupled statement outside of a Quasistationary or Transient statement fails, Sentaurus Device continues to the next command in the Solve section or terminates, as previously discussed.

If the Newton iterations converge within considerably fewer iterations than set by Iterations (see Section 1.2.5 Math Section), the step size increases by Increment until the MaxStep size is reached (see Section 1.2.6 Solve Section). In Transient, the Increment value exactly as specified is used; while for Quasistationary, the number of iterations, performed on a previous step, is considered for Increment computation.

7.10 Runtime Statistics

Information about runtime and memory use is located at the end of the log file:

******************************************************************************
Sentaurus Device peak memory usage: 248 megabytes
Sentaurus Device simulation times:
  wallclock: 71.71 s (0 h:01 m:11 s)
  total cpu: 71.65 s (0 h:01 m:11 s)
Sentaurus Device simulation finished (Date: Mon Feb 14 11:40:27 2022  (CEST)).
********************************* Good Bye ! *********************************

The reported memory use refers to the maximum memory size over the whole simulation.

The wallclock time records the elapsed time between the launch of the Sentaurus Device job and its completion. The total cpu time records the time that the job was occupying a CPU. If other jobs are running on the computer at the same time, then the wallclock time might be considerably larger than the total cpu time. It is recommended to specify the Wallclock option in the Math section to produce the correct timing report when running a simulation in parallel mode on multiple processors:

Math {
  ...
  Wallclock
}