Bus error

[albatross]

  I am using ATK11.2 to simulate dual-gated FET model. When I run the script, the job exited after several hours. The out file shows the error below:

+------------------------------------------------------------------------------+
|   0 E = -815.297 dE =  1.000000e+00 dH =  3.274853e-02                       |
+------------------------------------------------------------------------------+

                            |--------------------------------------------------|
Calculating Eigenvalues    : ==================================================
Calculating Density Matrix : ==================================================

/home/zxgao/QuantumWise/atk-11.2.0/atkpython/bin/atkpython: line 3:  1196 Bus error               PSEUDOPOTENTIALS_PATH=$EXEC_DIR/../share/pseudopotentials GPAW_SETUP_PATH=$EXEC_DIR/../share/gpaw-setups/ PYTHONHOME=$EXEC_DIR/.. PYTHONPATH= LD_LIBRARY_PATH=$EXEC_DIR/../lib $EXEC_DIR/atkpython_exec $*

So what's the problem and how can I solve it?
Thank the administrator very much.

[Anders Blom]

First of all you should upgrade to 11.2.2, although it might be best to wait for 11.2.3 which will be out soon.

"Bus error" is an unusual error, it's something the operating system throws. Perhaps a hardware failure? Which Linux are you using? Do you have enough memory? Perhaps the disk is getting full?

[albatross]

  Thank you for your reply. My linux version is centos 5.0. My memory size is 70 G, and my harddisk size is 60G. I think this is enough for my model containing 520 atoms. I use LDA approximation and SZ orbital.
  OK, I will upgrade the version to 11.2.3 to have a try.

[Anders Blom]

Yes, clearly your system is large enough to handle the model (provided you do have a 64-bit operating system?).

As I indicated in my original reply, "Bus error" is an unusual error, and this discussion confirms: http://stackoverflow.com/questions/212466/what-is-a-bus-error. Thus the suspicion of a hardware problem cannot be ruled out.

Is the error reproducible? Did you try many times? It may have been a spurious thing, let's hope so, because there is very little chance that ATK is really causing this, or that there is anything we could do to fix it.

[albatross]

Yes, I use a 64-bit operating system. The error is reproducible, and sometimes the fault is a segmentation fault. I found that the fault disappeared after I changed the size of our model smaller.

[Anders Blom]

Segmentation fault can certainly occur as a result of not having enough memory, so perhaps the bus error was just a masked "out of memory segfault".

If you have problems with memory while using gated structures, the reason is usually that the mesh cut-off of the DFT model produces large grids. If possible, try using the Huckel method instead, it's highly accurate for graphene (I'm just guessing that's your system) as long as you use the Cerda graphite basis set. Contact me by mail for license options, we can provide a temporary ATK-SE license if you don't have one already; be sure to include information about your affiliation (preferably customer number). If you have at demo/trial license, you already have the ATK-SE feature.

But you really have 70 GB of RAM? It's hard to imagine that you run out of memory with SZ on that, so if you could please send your structure (by mail, if you don't want to post it publicly) so we can run some tests with it.