QuantumATK Forum

QuantumATK => General Questions and Answers => Topic started by: ziand on March 18, 2014, 18:26

Title: Initial densty NeutralAtom
Post by: ziand on March 18, 2014, 18:26: I use ATK 12.8.2 and the Hückel method to calculate a CNT-Device with metal electrodes.
Electrode calculation finishes in 1 step (bulk 5x5x3 metal atoms). This is okay.
Now, to save some time, I thought with Hückel we can choose
DeviceAlgorithmParameters(initial_density_type = NeutralAtom()).

What comes out at first step is (I deleted large portions of the output, but important stuff is there).
Quote
+------------------------------------------------------------------------------+
| Density Matrix Report DM DD |
+------------------------------------------------------------------------------+
| 0 Al [ 0.716 , -0.827 , 1.169 ] 3.95611 0.95611 |
| 1 Al [ 2.148 , -3.306 , 1.169 ] 3.95619 0.95619 |
| 2 Al [ 3.579 , -5.786 , 1.169 ] 3.95611 0.95611 |
| 3 Al [ 5.011 , -8.266 , 1.169 ] 3.95653 0.95653 |
| 4 Al [ 6.443 , -10.746 , 1.169 ] 3.95653 0.95653 |
| 5 Al [ 2.148 , 1.653 , 1.169 ] 3.95653 0.95653 |
| 6 Al [ 3.579 , -0.827 , 1.169 ] 3.95601 0.95601 |
| 7 Al [ 5.011 , -3.306 , 1.169 ] 3.95601 0.95601 |
| 8 Al [ 6.443 , -5.786 , 1.169 ] 3.95653 0.95653 |
| 9 Al [ 7.875 , -8.266 , 1.169 ] 3.95684 0.95684 |
| 10 Al [ 3.579 , 4.133 , 1.169 ] 3.95653 0.95653 |
| 11 Al [ 5.011 , 1.653 , 1.169 ] 3.95616 0.95616 |
| 12 Al [ 6.443 , -0.827 , 1.169 ] 3.95722 0.95722 |
| 13 Al [ 7.875 , -3.306 , 1.169 ] 3.95616 0.95616 |
| 14 Al [ 9.306 , -5.786 , 1.169 ] 3.95653 0.95653 |
| 15 Al [ 5.011 , 6.613 , 1.169 ] 3.95611 0.95611 |
| 16 Al [ 6.443 , 4.133 , 1.169 ] 3.95601 0.95601 |
| 17 Al [ 7.875 , 1.653 , 1.169 ] 3.95722 0.95722 |
| 18 Al [ 9.306 , -0.827 , 1.169 ] 3.95722 0.95722 |
| 19 Al [ 10.738 , -3.306 , 1.169 ] 3.95601 0.95601 |
| 20 Al [ 6.443 , 9.093 , 1.169 ] 3.95619 0.95619 |
| 21 Al [ 7.875 , 6.613 , 1.169 ] 3.95619 0.95619 |
| 22 Al [ 9.306 , 4.133 , 1.169 ] 3.95601 0.95601 |
| 23 Al [ 10.738 , 1.653 , 1.169 ] 3.95616 0.95616 |
| 24 Al [ 12.170 , -0.827 , 1.169 ] 3.95601 0.95601 |
| 25 Al [ 2.148 , 0.000 , 3.507 ] 7.27705 4.27705 |
| 26 Al [ 3.579 , -2.480 , 3.507 ] 7.27734 4.27734 |
| 27 Al [ 5.011 , -4.960 , 3.507 ] 7.27705 4.27705 |
| 28 Al [ 6.443 , -7.440 , 3.507 ] 7.27609 4.27609 |
| 29 Al [ 7.875 , -9.919 , 3.507 ] 7.27609 4.27609 |
| 30 Al [ 3.579 , 2.480 , 3.507 ] 7.27728 4.27728 |
...
| 70 Al [ 6.443 , 10.746 , 5.845 ] 10.84910 7.84910 |
| 71 Al [ 7.875 , 8.266 , 5.845 ] 10.85587 7.85587 |
| 72 Al [ 9.306 , 5.786 , 5.845 ] 10.84910 7.84910 |
| 73 Al [ 10.738 , 3.306 , 5.845 ] 10.85313 7.85313 |
| 74 Al [ 12.170 , 0.827 , 5.845 ] 10.85313 7.85313 |
| 75 Al [ 0.716 , -0.827 , 8.183 ] 13.62751 10.62751 |
| 76 Al [ 2.148 , -3.306 , 8.183 ] 13.63058 10.63058 |
| 77 Al [ 3.579 , -5.786 , 8.183 ] 13.62751 10.62751 |
| 78 Al [ 5.011 , -8.266 , 8.183 ] 13.62573 10.62573 |
| 79 Al [ 6.443 , -10.746 , 8.183 ] 13.62573 10.62573 |
| 80 Al [ 2.148 , 1.653 , 8.183 ] 13.62573 10.62573 |
...
| 120 Al [ 7.875 , 9.919 , 10.521 ] 15.65670 12.65670 |
| 121 Al [ 9.306 , 7.440 , 10.521 ] 15.65670 12.65670 |
| 122 Al [ 10.738 , 4.960 , 10.521 ] 15.62076 12.62076 |
| 123 Al [ 12.170 , 2.480 , 10.521 ] 15.61755 12.61755 |
| 124 Al [ 13.602 , 0.000 , 10.521 ] 15.62076 12.62076 |
| 125 C [ 10.225 , 0.000 , 12.435 ] 15.38028 11.38028 |
| 126 C [ 5.525 , 0.000 , 12.435 ] 15.38028 11.38028 |
| 127 C [ 6.700 , -2.035 , 12.435 ] 15.38028 11.38028 |
| 128 C [ 9.050 , -2.035 , 12.435 ] 15.38028 11.38028 |
| 129 C [ 6.700 , 2.035 , 12.435 ] 15.38028 11.38028 |
| 130 C [ 9.050 , 2.035 , 12.435 ] 15.38028 11.38028 |
| 131 C [ 9.910 , -1.175 , 13.146 ] 14.72651 10.72651 |
| 132 C [ 5.839 , 1.175 , 13.146 ] 14.72929 10.72929 |
| 133 C [ 5.839 , -1.175 , 13.146 ] 14.72651 10.72651 |
| 134 C [ 7.875 , -2.350 , 13.146 ] 14.72929 10.72929 |
| 135 C [ 9.910 , 1.175 , 13.146 ] 14.72929 10.72929 |
| 136 C [ 7.875 , 2.350 , 13.146 ] 14.72651 10.72651 |
| 137 C [ 7.875 , -2.350 , 14.566 ] 14.72651 10.72651 |
| 138 C [ 9.910 , -1.175 , 14.566 ] 14.72929 10.72929 |
| 139 C [ 9.910 , 1.175 , 14.566 ] 14.72651 10.72651 |
| 140 C [ 7.875 , 2.350 , 14.566 ] 14.72929 10.72929 |
...
| 270 Al [ 7.875 , 8.266 , 26.543 ] 3.95684 0.95684 |
| 271 Al [ 9.306 , 5.786 , 26.543 ] 3.95653 0.95653 |
| 272 Al [ 10.738 , 3.306 , 26.543 ] 3.95601 0.95601 |
| 273 Al [ 12.170 , 0.827 , 26.543 ] 3.95601 0.95601 |
+------------------------------------------------------------------------------+
| 0 E = -759.204 dE = 1.113262e+00 dM = 7.343037e+00 dH = 3.330028e+02 |
+------------------------------------------------------------------------------+

This looks like a very bad first guess and clearly far away from neutral atoms.
(The electrode copy are the first and last 75 atoms.)

Why is the starting guess so far off?

It still convergeres towards a reasonable result (--> Density matrix of Al about 3 and for C about 4) but it takes more steps than expected/necessary.
Quote
| 250 Al [ 2.148 , -1.653 , 26.543 ] 2.99953 -0.00047 |
| 251 Al [ 3.579 , -4.133 , 26.543 ] 2.99953 -0.00047 |
| 252 Al [ 5.011 , -6.613 , 26.543 ] 3.00118 0.00118 |
| 253 Al [ 6.443 , -9.093 , 26.543 ] 2.99914 -0.00086 |
| 254 Al [ 2.148 , 3.306 , 26.543 ] 2.99913 -0.00087 |
| 255 Al [ 3.579 , 0.827 , 26.543 ] 3.00023 0.00023 |
| 256 Al [ 5.011 , -1.653 , 26.543 ] 3.00036 0.00036 |
| 257 Al [ 6.443 , -4.133 , 26.543 ] 3.00023 0.00023 |
| 258 Al [ 7.875 , -6.613 , 26.543 ] 2.99914 -0.00086 |
| 259 Al [ 3.579 , 5.786 , 26.543 ] 3.00118 0.00118 |
| 260 Al [ 5.011 , 3.306 , 26.543 ] 3.00023 0.00023 |
| 261 Al [ 6.443 , 0.827 , 26.543 ] 3.00006 0.00006 |
| 262 Al [ 7.875 , -1.653 , 26.543 ] 3.00006 0.00006 |
| 263 Al [ 9.306 , -4.133 , 26.543 ] 3.00023 0.00023 |
| 264 Al [ 5.011 , 8.266 , 26.543 ] 2.99953 -0.00047 |
| 265 Al [ 6.443 , 5.786 , 26.543 ] 2.99953 -0.00047 |
| 266 Al [ 7.875 , 3.306 , 26.543 ] 3.00036 0.00036 |
| 267 Al [ 9.306 , 0.827 , 26.543 ] 3.00006 0.00006 |
| 268 Al [ 10.738 , -1.653 , 26.543 ] 3.00036 0.00036 |
| 269 Al [ 6.443 , 10.746 , 26.543 ] 2.99953 -0.00047 |
| 270 Al [ 7.875 , 8.266 , 26.543 ] 3.00052 0.00052 |
| 271 Al [ 9.306 , 5.786 , 26.543 ] 2.99953 -0.00047 |
| 272 Al [ 10.738 , 3.306 , 26.543 ] 3.00023 0.00023 |
| 273 Al [ 12.170 , 0.827 , 26.543 ] 3.00023 0.00023 |
+------------------------------------------------------------------------------+
| 27 E = -275.907 dE = 4.153771e-06 dM = 1.982099e-05 dH = 1.505746e-05 |
+------------------------------------------------------------------------------+
| Calculation Converged in 27 steps |
+------------------------------------------------------------------------------+

If I increases the length of the central CNT it takes more and more steps to converge and at more than 10 CNT unit cells it refuses to converge at all.

If I do an equivalent bulk calculation, the output of step 0 looks much better.
(Ans this is what I do from now on.)
Title: Re: Initial densty NeutralAtom
Post by: Anders Blom on April 10, 2014, 14:10: There is no simple solution to this, although we are working on some algorithms that make the first few steps more well behaved. Yes, NeutralAtoms is a very bad starting guess for the open system - that's why EquivalentBulk helps.

There is a trick, that can help you reduce the calculation time though. You can give a separate set of IterationControlParameters to EquivalentBulk, so you can for instance set a max_steps of just 5 or 10, or a high convergence criterion (often also a higher damping_factor). Even with a reasonable output from EB, not necessarily fully converged but closer to selfconsistent than NeutralAtoms, the device part will converge much better.