Density scan (update)

[dlg0]

@diemsj @parkjm @asontag

(Posted by @dlg0 on behalf of @murakamim - copy of earlier comments placed in ips-atom issues with minor modifications)

Hi Steffi, JM, David and Aaron,
I have repeated Steffi’s APS2015/fastran_work/iterss_benchmark/run03 by copying all inputs from her area (I hope that is the current one). I believe that this case is exact iterss_benchmark case (from the JM (public) template) rather than the EPED-IPS edge modified case, but I am not certain yet.
These repeated runs (but NSTEP=21 for TGLF model as well as GLF23 model) were done on edison in:

/project/projectdirs/atom/users/murakami/aps2015/iterss_benchmark/run01/

/project/projectdirs/atom/users/murakami/aps2015/iterss_benchmark/run01_glf23/

and their IDL analysis were done in GA idl by copying results files in GA computer (since nersc idl is too slow), besides more outputs are available than attached:

/task/ips/murakami/aps15_diems/iterss_benchmark/run01/simulation_results/

/task/ips/murakami/aps15_diems/iterss_benchmark/run01_glf23/simulation_results/

Quick observations:

• In TGLF modeling, iteration 4 (NSTEP=5) is not enough, you need at least NSTEP=11 or NSTEP=21, which takes ~5 hours. The attempt for GLF23 model with NSTEP=21 did only finish iteration 16, so running TGLF is faster than GLF23 for this BM case.
• In TGL23 outputs (vs iteration#) is attached as a PDF file. As you see, while Te0, Teav almost reached steady state (SS), Tio and Tiav have not saturated yet. The values of q0, qmin, and f_NI and V_surf have not achieved SS either, which are important for the study. TGLF SS results are better than GLF23 (IOS benchmark) , achieving fNI ~110% and Vsurf =-4mV.
• To get faster and smooth converging solutions in TGLF model probably would require tweeting (by JM) the relaxation factor or longer dt_relax_J or NUBEAM call (NSTEP=10), as in my previous ITB cases. Although the ‘seed-current’ and/or tweaking CURRAY J_multi may help, I don’t like these methods myself.
• The real challenge is to use EPED-IPS edge profiles (which I have not seen, yet) and satisfying ITER SS condition (Q=5, f_NI=100% simultaneously) with allowable ITER H&CD capabilities is not that easy.

I will be in this week at ORNL, and then I will be visiting GA for the whole next week.

Masanori

FI_conv_aps15sjd_itBM_MM_r01_i0-20.pdf