"There is a very negative time step detected" warnings

Quite regularly (e.g. reported by @Nikos in the last general meeting), shower simulations throw the warning:

There is a very negative time step detected: -1.42207e-17 s. This is not physical and may easily crash subsequent modules. Set to 0_s, but CHECK AND FIX.

From a first debugging, this happens especially for low-energy electrons/positrons. Furthermore, the warning seems to vanish if one turns of multiple scattering, so this seems to be related.

However, I don't yet understand why this is related, because next_intersect() (where the error is thrown) is called from getTrack(), but getTrack() is called in Cascade.inl before the continuous processes (and therefore multiple scattering) is applied.

added EM label

changed milestone to %2023 Workshop Milestone

I have one question about the code in Intersect.inl which throws the error message:

    if (minTime < 0_s) {
      if (minTime < 1e-8_s) {
        CORSIKA_LOG_ERROR(
            "There is a very negative time step detected: {}. This is not physical and "
            "may "
            "easily crash subsequent modules. Set to 0_s, but CHECK AND FIX.",
            minTime);
      }
      minTime = 0_s; // LCOV_EXCL_STOP
    }

If the first condition minTime < 0_s is true, the second condition minTime < 1e-8_s always true as well, right? This is why warning such as

[corsika:error   (Intersect.inl:114)] There is a very negative time step detected: -1.27084e-17 s. This is not physical and may easily crash subsequent modules. Set to 0_s, but CHECK AND FIX.

is being printed. Is it possible that the code was intended to have the condition std::abs(minTime) < 1e-8_s? So maybe this error message isn't supposed to be thrown at all...

My guess: there's a missing minus sign, i.e. if (minTime < -1e-8_s)

Probably more intuitive writte this way

Another comment: This code/warning has been introduced with commit a8d6de9e. The lines have been written to account for numerical instabilities, which I concluded came from the tracking algorithm being inaccurate at border transitions (see #429 (closed)).

This is the debug output before the error message is printed:

Click to expand

[corsika:debug   (Cascade.inl:149)] total_lambda=20.120715497270293 g/cm2, next_interact=5.990979328724709 g/cm2
[corsika:debug   (Decay.inl:125)] Pythia::Decay: checking if particle: e- can decay in PYTHIA? false 
[corsika:debug   (Cascade.inl:159)] total_lifetime=inf ns, next_decay=inf ns
[corsika:debug   (TrackingLeapFrogCurved.inl:44)] TrackingLeapfrog_Curved pid: e- , E = 0.004478759580410741 GeV 
	Tracking pos: (69.7681 4.88081 6.381e+06) m 
	Tracking   p: (0.00428743 -0.000904804 -0.000795651) [] GeV 
	Tracking   v: (2.86986e+08 -6.05644e+07 -5.32581e+07) m/s
[corsika:debug   (TrackingLeapFrogCurved.inl:96)] gyroradius 300.96 m, steplimit: 117.199 m = 3.93151e-07 s
[corsika:debug   (Intersect.inl:30)] volumeNode=0x600002f28000, numericallyInside=true 
[corsika:debug   (CubicSolver.inl:311)] niter reached max iterations 100
[corsika:debug   (Intersect.inl:64)] search intersection times with child volume 0x600002f28050:
[corsika:debug   (Intersect.inl:69)] children t-entry: 1.04434e-08 s, t-exit: 9.28906e-05 s, smaller? true 
[corsika:debug   (Intersect.inl:106)] next time-intersect: 1.04434e-08 s, node 0x600002f28050 
[corsika:debug   (Cascade.inl:200)] transport particle by : 3.113192364151778 m Medium transition after: 3.113192364151778 m Decay after: inf m Interaction after: 140.3219753487111 m Continuous limit: 5.987235429132918 m 
[corsika:debug   (ParticleCut.inl:94)] ParticleCut: checking e- (11), E_kin= 0.003734381351945659 GeV, E=0.00424538135194566 GeV, m=0.000511 GeV
[corsika:debug   (Cascade.inl:231)] discrete process before geometric limit ? no
[corsika:debug   (Cascade.inl:239)] volume boundary crossing to 0x600002f28050
[corsika:debug   (Cascade.inl:149)] total_lambda=21.23865104911402 g/cm2, next_interact=0.7658464434918187 g/cm2
[corsika:debug   (Decay.inl:125)] Pythia::Decay: checking if particle: e- can decay in PYTHIA? false 
[corsika:debug   (Cascade.inl:159)] total_lifetime=inf ns, next_decay=inf ns
[corsika:debug   (TrackingLeapFrogCurved.inl:44)] TrackingLeapfrog_Curved pid: e- , E = 0.00424538135194566 GeV 
	Tracking pos: (72.7681 4.26308 6.381e+06) m 
	Tracking   p: (0.00408182 -0.000791183 0.000681909) [] GeV 
	Tracking   v: (2.88242e+08 -5.58703e+07 4.81538e+07) m/s
[corsika:debug   (TrackingLeapFrogCurved.inl:96)] gyroradius 243.624 m, steplimit: 94.8715 m = 3.18864e-07 s
[corsika:debug   (Intersect.inl:30)] volumeNode=0x600002f28050, numericallyInside=false 
[corsika:debug   (Intersect.inl:50)] intersection times with currentLogicalVolumeNode: entry=-0.006942525965514321 s and exit=-1.2708429079741811e-17 s
[corsika:debug   (Intersect.inl:64)] search intersection times with child volume 0x600002f280f0:
[corsika:debug   (Intersect.inl:106)] next time-intersect: -1.27084e-17 s, node 0x600002f28000 
[corsika:error   (Intersect.inl:114)] There is a very negative time step detected: -1.27084e-17 s. This is not physical and may easily crash subsequent modules. Set to 0_s, but CHECK AND FIX.

created branch 590-there-is-a-very-negative-time-step-detected-warnings to address this issue

mentioned in merge request !508 (merged)

mentioned in commit 07bf5d2e

closed with merge request !508 (merged)