Sylvain G. Korzennik | About | Research | CV & LoP | HPC@SI

Medium-ℓ Tables

Table of Content

• What’s New
• Introduction
• Temporal Coverage and Time-series Lengths
• Available Tables Start Dates
• Organization
• Leakage Matrices
• Plots
• Direct Links to the Tables
• At JSOC and JSOC2

What’s New

• Jan 13, 2025

⚠️ The JSOC database is back up and running, hence all the new results have been ingested. Unfortunately the JSOC is not yet accessible from the outside, but all the files, including the m10qr ones, are available via my Google drive, follow the links to GONG or HMI below and select 64z.jsBo=0 or 64z.sym.jsBo=0, etc.

• Dec 18, 2024

  • New tables for fitting time-series of various lengths up to the latest GONG observations:
    • up to GONG month 298 (2024-08-31);
    • time-series for 36d, 01e, 01e.sym, 02e, 04e, 08e, 10e, 16e & 32e have been fitted;

• Dec 6, 2024

  • Tables for 64x 72d long time series for HMI and coeval GONG (aka 64z & 64z.sym):
    • starting at #6400 or 2010.07.11
      • i.e. 2nd 72d long segment of HMI observations (even progression wrt to #1216);
    • tables for fitting asymmetric and symmetric mode profile are available.
  • New Comparison plots of GONG and HMI co-eval results for 32x and 64z have been added.

• Nov 26, 2024

  • More HMI tables for 10e, and 01e, 01e.sym, 02e, 04e, 08e, 16e & 32e time-series.
  • Most of the plots have been updated.

Introduction

The tables resulting from my fitting to the GONG, HMI and, MDI observations at low and medium degrees (i.e., 0 ≤ ℓ ≤ 200|300) for all the data available to date are:

• on my Google drive, and
• at the JSOC (until it went down 😱),

See: Korzennik, 2005, 2008a, 2008b, 2023; [ADS: 2005ApJ...626..585K, 2008AN.…329..453K, 2008JPhCS.118a2082K, 2023FrASS…931313K].

Available Files and Format

The primary tables list the individual mode frequencies, line-widths, asymmetries and amplitudes.

Derived tables list multiplets mode frequencies, line-widths, asymmetries and amplitudes as well as Clebsch-Gordan coefficients, after performing a fit to 6, 9, 18 or 36 coefficients.

All the tables are plain ASCII files, the file formats are described in the following documents:

• singlets: singlets.txt
• multiplets: multiplets.txt
• cg-coefs & js-coefs: coefs.txt
• m10qr: see “Mode Parameters” under “Data Formats” at JSOC’s Global Helioseismology Products.

The ASCII files are also consolidated in tar-compressed sets (up to 20 files in each .tgz set) for convenience (fewer & faster downloads).

JSOC Availability

‼️ The JSOC is still not accessible from the outside following Stanford Solar Group’s machine room flooding on 26 Nov 2024: Most JSOC services remain unavailable until further notice.

Almost all the tables have been converted to the Stanford format and ingested into the JSOC, after reformatting to match the m10qr file format.

Tables using the same leakage matrix as the MDI and HMI projects’ pipe-line are publicly available at the JSOC; the tables using my leakage matrices are accessible via the password protected JSOC interface (aka JSOC2) or by logging into the JSOC machines.

These data series can be queried using the su_sylvain.* prefix, while the m10qr files are also available on my Google drive.

Temporal Coverage and Time-series Lengths

Time-series of various lengths have been fitted, namely 36, 72, 144, etc., up to 128x 72 day-long ones, as well as 5x 72 and 10x 72 day-long ones.

The 36, 72 and 360 day-long series are offset in time by a multiple of their respective lengths (no overlap), while all the other ones (144, 288, etc., up to 128x 72 day-long and the 720 day-long) are offset by a multiple of half their respective lengths (50% overlap).

Almost all the time-series start at a time aligned with respect to MDI mission day #1216 (1996.05.01), hence the same epochs are always fitted for each instrument. For the very long time-series (32x and 64x 72d), additional time series with offsetted starting time have also been fitted.

Available Tables Start Dates

• First & last available start dates - as of Dec 18 2024

• Numbers preceeded by # are MDI mission day no. (#1216 is 1996.05.01).
• The 1x 5x & 64x 72d have been fitted using symmetric mode profiles as well as asymmetric ones (see below).
• Some cases have also been fitted using different leakage matrices (see below).
• For the 32x 72d and 64x 72d long time series, GONG and HMI observations have been also fitted:
  • using offsetted starting times, i.e., using a 8x 72d offset instead of the 16x or 32x 72d offset
  • aka 32x or 64z instead of 32e or 64e, see below.

Organization

Files are organized as follows:

Instrument: GONG, MDI or HMI
    ↳ time-series length – [peak profile] – leakage matrix
        ↳ type of tables
            ↳ number of CG coefficients fitted

Time-series length – [peak profile] – leakage matrix

Types of Tables

Number of Coefficients

Leakage Matrices

GONG

I have used several leakage matrices:

1. The one provided by J. Schou, for GONG and for Bo=0 [jsBo=0].
2. Two variants I computed, for GONG, and for Bo=0:
   1. without any PSF [skBo=0].
   2. with a PSF, based on the mean MTF derived by HGEOM [skBo=0+psf].
3. Two variants I computed, for GONG, and for Bo=√<|Bo|2>:
   1. without any PSF [skBo=a2].
   2. with a PSF (as above). [skBo=a2+psf].

In the case of my leakage matrices, the upgrade of the GONG detector from a medium-resolution CCD, with rectangular pixels, to a high-resolution one with square pixels is taken into account. Since not all instruments were upgraded simultaneously, some epochs were fitted with a weighted average of the leakage matrix corresponding to pre and post detector upgrade.

MDI

I have used three leakage matrices:

1. The one provided by the MDI team (JS) for the Structure data, and for Bo=0 [jsBo=0].
2. One I computed, for the Structure data, and Bo=0 [skBo=0].
3. One I computed, for the Structure data, and the mean Bo for that epoch [skBo=a].

The skBo=a leakage matrix was used only for the 02e & 04e time-series, since for longer ones, the mean Bo is almost 0.

HMI

I have used three leakage matrices:

1. The one provided by the HMI/MDI team for full disk data, and for Bo=0 [jsBo=0].
2. Two I computed, for HMI:
   1. for Bo=0 [skBo=0].
   2. for Bo=√<|Bo|2> [skBo=a2].

All HMI tables correspond to VERSION=1 of the spherical harmonic decomposition performed by Stanford (i.e., includes the Carrington elements correction).

Notes:

Leaks

The skBo=a cases are not equivalent to the skBo=a2 cases:

• skBo=a means using a leakage matrix computed for the mean value of Bo over the time-series for each epoch - computed for 144-day and 244-day long MDI only.
• skBo=a2 means using a leakage matrix computed for the mean value of Bo=√<|Bo|2> over one year - computed for the 32x 72d cases for GONG & HMI.

New executables

• As of June 2024 and because of changes in (i) operating system (CO7.9 to RL8.9) and (ii) the version of the compiler (and its nature, i.e., LLVM), I am using new executables, but the same source code.
• To validate this, I fitted with the new executables two sets that were fitted previously, one for a short time series (i.e., 36-day long) and one for a long time series (i.e., 720-day long).
• Because of the nature of the algorithm, there are differences at the 1E-05 level (not 1E-09, the numerical noise level), but no systematic differences.
• In the process I discovered an embarrassing coding error (in a logic block termination) in the part that adds the so-called mode contamination (leakage from a different n in the fitting spectral range). This error added an invalid contamination at the level of 1E-300, which is, we can all agree, totally negligible. The code has since been corrected and this invalid contamination is no longer present. May we all ever only make coding errors at the 1E-300 level ;-)
• Plots of comparisons are available for one 36-day long segment, and one 720-day long segment.

Plots

I've produced some plots, that illustrate the content of these tables:

• Statistics on fitting, i.e., the number of modes fitted for each segment, and gaps info:
  • GONG, MDI & HMI fitting;
  • MDI & HMI fitting (or a subset);
  • GONG fitting, (gap-filled).
• Comparison of GONG and HMI co-eval fitting
  • plotted for 32x and 64z
• Mean frequency changes (raw or scaled) with time and activity, for GONG, MDI & HMI tables, resulting from my fitting (SGK) and the respective project fitting methods (NSO/SU), using either a symmetric or asymmetric peak profile:
  • <δν>, i.e., unscaled frequency changes; (or limited to ℓ ≤ 150);
  • <δν Q>, where Q is the relative mode mass; (or limited to ℓ ≤ 150);
  • <δν Q / Γ>, where Γ is the mode line-width (or limited to ℓ ≤ 150);
• The range in degree, ℓ, for the averaging is either the fitted range (ℓ ≤ 150, 200, 300) depending on the data set and the origin of the fitting, or has been limited to the range fitted by all (ℓ ≤ 150).
• The numbers within parentheses on the lower panels are the intercept and slope (x 1000) of the regression with respect to the solar activity index (F10.7).

Direct Links to the Tables

I compiled a set of pages that allows you to jump directly to the tables:

Remember, Google Drive allows you to select which files you want to download, or you can use the tgz sets to speed up the download.

At JSOC and JSOC2

‼️ The JSOC is still not accessible from the outside following Stanford Solar Group’s machine room flooding on 26 Nov 2024: Most JSOC services remain unavailable until further notice.

• Names of series and number of sets for each fitted length and mode profile type:

• Other available series, using SGK’s leakage matrices (JSOC2 only):