smallbin double linked list corrupted; largerTransferred Power flux

[ZhihaoJia16]

Hi, I install scuff-em-Overhaul201512, trying to reproduce the example Thermal radiation, heat transfer, and non-equilibrium Casimir forces between silicon dioxide spheres about two spheres.

• 1. get error

*** Error in `scuff-neq': malloc(): smallbin double linked list corrupted: 0x000000000217d100 ***

• 2. The Transferred Power flux computed between two SiO2 spheres is larger 4-6 orders of magnitude than that of reference.

The details are follows: I run,

#!/bin/bash
for N in 501 
do
  ARGS=""
  ARGS="${ARGS} --geometry SiO2Spheres_${N}.scuffgeo"
  ARGS="${ARGS} --OmegaFile OmegaFile"
  ARGS="${ARGS} --Power --ZForce"
  ARGS="${ARGS} --DSIRadius 5"
  ARGS="${ARGS} --DSIPoints 302"
  scuff-neq ${ARGS}
done

it crashes after computing of one or several omega every time, with error

*** Error in `scuff-neq': malloc(): smallbin double linked list corrupted: 0x000000000217d100 ***

I do % export SCUFF_INTERPOLATION_TOLERANCE=1e-2, it does not work. Anyway, I finally get SiO2Spheres_501.SIFlux.DSI302 file. The Transferred Power flux between two SiO2 spheres is

But, numerical values computed is larger 4-6 orders of magnitude than that of reference.

Best. Thank you in advance.

[ZhihaoJia16]

I still get error received signal 11 when I reproduce Flux spectrum of the gold cone-plate configuration at a single separation d = 0.2L with fine surface meshes in Fluctuating-surface-current formulation of radiative heat transfer: Theory and applications. (For coarse surface meshes, it's ok )

then I do % export SCUFF_INTERPOLATION_TOLERANCE=1e-3, it does not work, get

...
03/28/16::17:53:06:   80 % (2318/2898)... 
03/28/16::17:53:09:   90 % (2608/2898)... 
03/28/16::17:53:11:  ...done! 
03/28/16::17:53:11: Computing neq quantities at omega=0.113722... 
03/28/16::17:53:11:  Assembling self contributions to T(0)... 
03/28/16::17:53:11: Assembling BEM matrix block (0,0) 
/home/jia/scuff-em-installation/lib/libscuff.so.0(+0x717d3)[0x7ff4742a77d3]
/lib/x86_64-linux-gnu/libc.so.6(+0x36d40)[0x7ff473687d40]
/usr/lib/x86_64-linux-gnu/libgomp.so.1(+0x8a63)[0x7ff472365a63]
/home/jia/scuff-em-installation/lib/libscuff.so.0(_ZN5scuff29GetSurfaceSurfaceInteractionsEPNS_15GetSSIArgStructE+0x227)[0x7ff474266547]
/home/jia/scuff-em-installation/lib/libscuff.so.0(_ZN5scuff11RWGGeometry22AssembleBEMMatrixBlockEiiSt7complexIdEPdP7HMatrixPS5_iiPvbiS6_S3_+0xb00)[0x7ff4742644a0]
scuff-neq[0x404e49]
scuff-neq[0x4036f0]
/lib/x86_64-linux-gnu/libc.so.6(__libc_start_main+0xf5)[0x7ff473672ec5]
scuff-neq[0x4038f8]

**<><> received signal 11 

Best. Thank you in advance.

[HomerReid]

I've made some improvements to SCUFF-NEQ. Please pull from the Overhaul201512 branch and rebuild and re-run your tests.

Here are a couple of comments:

• SCUFF_INTERPOLATION_TOLERANCE is only used for periodic geometries. It has no effect for geometries involving compact objects.
• In addition to (or instead of) --DSIPoints, I encourage you to use --EMTPFT, which is much faster and should yield equivalent or more accurate results. For the transferred power, --OPFT should also be accurate. In general it is a good idea to use --EMTPFT and --OPFT for all frequencies and --DSIPFT for a subset of frequencies just as a sanity check.
• Note that the value of --DSIRadius must be strictly larger than the size of your objects. The bounding sphere should be at least one panel radius (ideally more) away from the surface of any object. For two spheres separated by (say) 3 microns I would use --DSIRadius 1.5.

[ZhihaoJia16]

Thank you for your answer. I will try it.