Open the job's results page and read the Summary tab. It explains most errors. Then
find the message below; the exact text from the output file is shown for each program.
To search the full output, open Output File and use your browser's Find
(Ctrl+F / Cmd+F). The last error near the end is usually the real one.
The sections are in order of how often each error happens on ChemCompute. See the
overview for the numbers.
| Program | What the output says |
|---|---|
| GAMESS | BUT YOU SELECTED MULTIPLICITY MULT= 1, *** CHECK YOUR INPUT CHARGE AND MULTIPLICITY ***, *** SCFTYP=RHF MUST HAVE MULT=1 *** |
| Psi4 | Inconsistent or unspecified chg/mult |
| PySCF | Electron number 17 and spin 0 are not consistent |
This is the single most common error on ChemCompute, about a quarter of all failed
quantum-chemistry jobs. The job stops within seconds, before any real calculation.
The rule: count the electrons (add up the atomic numbers, then subtract the charge).
An even count needs an odd multiplicity: 1 for a normal closed-shell molecule. An
odd count needs an even multiplicity: 2 for a radical with one unpaired electron.
What usually went wrong, in order:
SCFTYP=RHF MUST HAVE MULT=1). RHF pairs everyFix: use Do More Calculations on the failed job, correct the structure, charge or
multiplicity, and submit again.
| Program | What the output says |
|---|---|
| GAMESS | SCF IS UNCONVERGED, TOO MANY ITERATIONS or FAILURE TO LOCATE STATIONARY POINT, SCF HAS NOT CONVERGED |
| Psi4 | Could not converge SCF iterations in 100 iterations. |
| PySCF | SCF not converged. |
The self-consistent field (the step that finds the orbitals) didn't settle, so any energy
the job reports is meaningless. In order of what to try:
Semi-empirical methods (AM1, PM3, PM6) also fail to converge fairly often on unusual
structures. Try a small basis set such as 3-21G instead.
| Where | What it says |
|---|---|
| Output | CANCELLED AT … DUE TO TIME LIMIT, TIME LIMIT EXCEEDED |
| Summary | Your calculation went past the time limit. |
| Status | TIMEOUT |
The job needed longer than the time limit you chose. The results page usually offers to
resubmit with a longer limit, and tells you how long nine in ten similar jobs take. You
can also:
Geometry optimizations with B3LYP/6-31G* are the most common timeouts, simply because
they're the most common jobs. Large molecules with IR (a frequency calculation) take much
longer than an optimization alone.
| Program | What the output says |
|---|---|
| GAMESS | ILLEGAL EXTENDED BASIS FUNCTION REQUESTED. or ERROR! THERE ARE NO BASIS FUNCTIONS DEFINED. |
| Psi4 | BasisSetNotFound: BasisSet::construct: Unable to find a basis set for atom … |
| PySCF | BasisNotFoundError: Basis set not found for Br |
The basis set has no functions for one of your elements. Most of these on ChemCompute
come from iodine with 6-31G*, which only covers hydrogen through krypton. The form now
warns you about this before you submit.
Fix: pick a basis set that covers every element. See
Which basis sets cover which elements. For iodine: SPK-DZP or
3-21G (GAMESS), def2-SVPD (Psi4, PySCF).
| Program | What the output says |
|---|---|
| GAMESS | FAILURE TO LOCATE STATIONARY POINT, TOO MANY STEPS TAKEN |
| Psi4 | Could not converge geometry optimization in 200 iterations. |
| PySCF | Geometry optimization did not converge or Nuclear gradients of … not converged |
The optimizer used up its steps without finding a minimum (or a transition state, for a
saddle point search). The structure in the output is usually closer than where it
started.
Uploaded GAMESS input files hit this far more often than guided jobs, because GAMESS's own
step limit is low. See own input files.
The job had something the program couldn't read. The commonest kinds:
The job had no atoms. GAMESS says ERROR, NO $DATA GROUP WAS FOUND. Psi4 says
Dimension mismatch among fragment quantities: sep + 1 (1), chg (0), and mult(0). The
molecule never made it into the input. Before you press Submit Job, make sure the
molecule shows in the 3D viewer, and wait for the cleanup and symmetry steps to finish.
If it keeps happening, try another browser (Chrome or Firefox) and
tell us.
Uploaded GAMESS files: TYPING ERROR IN $CONTRL INPUT,
ERROR READING VARIABLE … CHECK COLUMN, ILLEGAL POINT GROUP and others. See
Using your own input file.
Method combinations that don't exist:
ROHF reference for DFT is not available (Psi4): use UHF.MP2 is not supported on Jetstream2 using 1 core: choose another cluster.| Where | What it says |
|---|---|
| GAMESS | MEMORY REQUEST EXCEEDS AVAILABLE MEMORY, NOT ENOUGH MEMORY |
| Any program | oom_kill event … Some of the step tasks have been OOM Killed, MemoryError |
| Summary | Your job ran out of memory. Allocation is about 2 GB / core. Try increasing the number of cores |
Each core brings about 2 GB of memory, so more cores means more memory. The results
page offers to resubmit with more cores. Jetstream2 jobs have 1 core and limited memory,
so for big jobs choose Expanse, Bridges-2 or Delta. Larger basis sets (especially with
diffuse functions, the + and aug- sets), frequency calculations and MP2 need the most
memory.
| Program | What the output says |
|---|---|
| GAMESS | THERE ARE ATOMS LESS THAN 0.100 APART, QUITTING... |
| Psi4 | Following atoms are too close: … |
Two atoms sit on top of each other: typically a structure loaded twice, pasted
coordinates with a duplicate line, or atoms dragged together. Look at the 3D view with
Label Atoms on, delete the extra atoms, and press Optimize Structure in 3D.
| What you see | What to do |
|---|---|
There was a problem with your symmetry, BUT ONLY … ATOMS WERE PRESENT IN YOUR $DATA |
The coordinates aren't exactly symmetric. Do More Calculations, untick Use Symmetry, submit |
There was a problem with the use of symmetry |
Same fix: run without symmetry |
Atom Positions have Changed -- click Resymmetrize to use symmetry (before submitting) |
Press Resymmetrize, or untick Use Symmetry |
THERE ARE NOT 6 OR 5 TRANS/ROT MODES (GAMESS): the frequencies were calculated onGRADIENT OUT OF RANGE (GAMESS): the starting structure is too far from a minimum.GAMESS: DDI Process … trapped a segmentation fault or a Summary saying
Something went wrong but we don't know what. Psi4: forrtl: severe … SIGSEGV.
PySCF: The PySCF process crashed before finishing.
Resubmit once with Do More Calculations. Crashes are sometimes a problem with the
node. If it happens again, try without PCM solvent, on another cluster, or with fewer
cores, and open a support ticket with the job number. A PySCF job that
ends in a Python error not listed here is worth a ticket too.
CANCELLED AT … without DUE TO TIME LIMIT means the job was cancelled, by you, by a
resubmit (which cancels the original), or by the cluster. If you didn't cancel it, just
resubmit.