Geant4Converter bug in dd4hep::Polyhedra

[ianna]

Having Geant4Converter successfully converted geometry to Geant4:

Geant4Converter  INFO  +++  Successfully converted geometry to Geant4.

a run-time error is reported by G4:

**************************************************** 
*  A runtime error has occured :                     
*    
-------- EEEE ------- G4Exception-START -------- EEEE -------
*** G4Exception : GeomMgt0002
      issued by : G4SmartVoxelHeader::BuildNodes()
PANIC! - Overlapping daughter with mother volume.
         Daughter physical volume hcalbarrelalgo:HBLayer0In1_3001
         is entirely outside mother logical volume hcalbarrelalgo:HBLayer0InSide !!
-------- EEEE -------- G4Exception-END --------- EEEE -------

*  the program will have to be terminated - sorry.   
**************************************************** 

Looking at Root geometry, the shape in question (in green) is not outside of its mother (in blue):

*** Shape hcalbarrelalgo:HBLayer0InSide_shape_0x7faac6540780: TGeoPgon ***
    Nedges = 1
*** Shape hcalbarrelalgo:HBLayer0InSide_shape_0x7faac6540780: TGeoPcon ***
    Nz    = 3
    phi1  =   -10.00000
    dphi  =    20.00000
     plane 0: z=    0.00000 Rmin=  177.50000 Rmax=  178.70000
     plane 1: z=  308.23896 Rmin=  177.50000 Rmax=  178.70000
     plane 2: z=  310.32582 Rmin=  178.70000 Rmax=  178.70000
 Bounding box:
*** Shape hcalbarrelalgo:HBLayer0InSide_shape_0x7faac6540780: TGeoBBox ***
    dX =     3.32668
    dY =    31.50963
    dZ =   155.16291
    origin: x=  178.13006 y=    0.00000 z=  155.16291

*** Shape hcalbarrelalgo:HBLayer0In1_shape_0x7faacf055600: TGeoBBox ***
    dX =     0.60000
    dY =     7.88440
    dZ =   154.11948
    origin: x=    0.00000 y=    0.00000 z=    0.00000

[ianna]

@cvuosalo and @bsunanda - FYI

[ianna]

Another message that looks like from G4:

**************************************************** 
*  A runtime error has occured :                     
*    Failed to handle unknown solid shape:pixfwdInnerDiskZminus:PixelForwardInnerDiskInnerRing_uni2_left of type TGeoCtub
*  the program will have to be terminated - sorry.   
**************************************************** 

[MarkusFrankATcernch]

This printout here:

*** Shape hcalbarrelalgo:HBLayer0InSide_shape_0x7faac6540780: TGeoPgon ***
    Nedges = 1

looks very strange. 1 edge should be impossible. There should be 3 or more to span a plane - no?

[ianna]

well, I'm not inventing it :-)

[ianna]

they look quite reasonable:

[MarkusFrankATcernch]

Bad argument: The problem must be more subtle.
Polygons are used in every cylindrical calorimeter - but only in your case a problem arises.

[ianna]

Ideally the blue shapes should be hidden by the red ones. I guess, it's a precision problem - the inner faces are almost identical. See the swap in red when a cursor is over the shape in the second snapshot:

[MarkusFrankATcernch]

This is a good candidate for the problem. I agree.

[ianna]

Here is produced GDML:

    <polyhedra aunit="deg" deltaphi="10" lunit="mm" name="hcalbarrelalgo_HBLayer0InSide_shape_0x7f5e06777d800x7f5e01857400" numsides="1" startphi="350">
      <zplane rmax="1787" rmin="1775" z="0"/>
      <zplane rmax="1787" rmin="1775" z="3082.38960788158"/>
      <zplane rmax="1787" rmin="1787" z="3103.25824748416"/>
    </polyhedra>

    <box lunit="mm" name="hcalbarrelalgo_HBLayer0In1_shape_0x7f5e01df3e800x7f5e01807680" x="12" y="157.68801299901" z="3082.38960788158"/>

    <volume name="hcalbarrelalgo_HBLayer0In10x7f5e01506480">
      <materialref ref="materials_Air0x7f5e0a642100"/>
      <solidref ref="hcalbarrelalgo_HBLayer0In1_shape_0x7f5e01df3e800x7f5e01807680"/>
      <physvol copynumber="1" name="hcalbarrelalgo_HBLayer0In1Plastic_1_10x7f5e00fb89c0">
        <volumeref ref="hcalbarrelalgo_HBLayer0In1Plastic_10x7f5e014ea2a0"/>
        <position name="hcalbarrelalgo_HBLayer0In1Plastic_1_10x7f5e00fb89c0_pos" unit="mm" x="-5" y="0" z="0"/>
      </physvol>
      <physvol name="hcalbarrelalgo_HBScintillatorLayer0In1_00x7f5e00fb8a10">
        <volumeref ref="hcalbarrelalgo_HBScintillatorLayer0In10x7f5e014ea340"/>
        <position name="hcalbarrelalgo_HBScintillatorLayer0In1_00x7f5e00fb8a10_pos" unit="mm" x="0.499999999999999" y="0" z="0"/>
      </physvol>
      <physvol copynumber="1" name="hcalbarrelalgo_HBLayer0In1Plastic_2_10x7f5e00fb8a60">
        <volumeref ref="hcalbarrelalgo_HBLayer0In1Plastic_20x7f5e014ea3e0"/>
        <position name="hcalbarrelalgo_HBLayer0In1Plastic_2_10x7f5e00fb8a60_pos" unit="mm" x="5.5" y="0" z="0"/>
      </physvol>
    </volume>
    
    <volume name="hcalbarrelalgo_HBLayer0InSide0x7f5e01559b80">
      <materialref ref="materials_Air0x7f5e0a642100"/>
      <solidref ref="hcalbarrelalgo_HBLayer0InSide_shape_0x7f5e06777d800x7f5e01857400"/>
      <physvol copynumber="1" name="hcalbarrelalgo_HBLayer0InSideSpacer_10x7f5e01007260">
        <volumeref ref="hcalbarrelalgo_HBLayer0InSideSpacer0x7f5e015063e0"/>
      </physvol>
      <physvol copynumber="3001" name="hcalbarrelalgo_HBLayer0In1_30010x7f5e010072b0">
        <volumeref ref="hcalbarrelalgo_HBLayer0In10x7f5e01506480"/>
        <position name="hcalbarrelalgo_HBLayer0In1_30010x7f5e010072b0_pos" unit="mm" x="1781" y="234.13638425802" z="1541.19480394079"/>
      </physvol>
      <physvol copynumber="3002" name="hcalbarrelalgo_HBLayer0In1_30020x7f5e01007300">
        <volumeref ref="hcalbarrelalgo_HBLayer0In10x7f5e01506480"/>
        <position name="hcalbarrelalgo_HBLayer0In1_30020x7f5e01007300_pos" unit="mm" x="1781" y="-234.13638425802" z="1541.19480394079"/>
        <rotation name="hcalbarrelalgo_HBLayer0In1_30020x7f5e01007300_rot" unit="deg" x="-180" y="0" z="0"/>
      </physvol>
    </volume>

[ianna]

OK, here is what ROOT to G4 converter should produce as a G4 geometry. The HBLayer0In1_3002 rotation should be different: note the sign of x="180"

    <volume name="HBLayer0InSide0x7f732daaf1a0">
      <materialref ref="Air0x7f733243d3c0"/>
      <solidref ref="HBLayer0InSide0x7f732dbfed00"/>
      <physvol copynumber="1" name="HBLayer0InSideSpacer0x7f732dbf4b20">
        <volumeref ref="HBLayer0InSideSpacer0x7f732dab7780"/>
      </physvol>
      <physvol copynumber="3001" name="HBLayer0In10x7f732dbf4bc0">
        <volumeref ref="HBLayer0In10x7f732dab7820"/>
        <position name="HBLayer0In10x7f732dbf4bc0_pos" unit="mm" x="1781" y="234.13638425802" z="1541.19480394079"/>
      </physvol>
      <physvol copynumber="3002" name="HBLayer0In10x7f732dbf4c60">
        <volumeref ref="HBLayer0In10x7f732dab7820"/>
        <position name="HBLayer0In10x7f732dbf4c60_pos" unit="mm" x="1781" y="-234.13638425802" z="1541.19480394079"/>
        <rotation name="HBLayer0In10x7f732dbf4c60_rot" unit="deg" x="180" y="-3.2461146750243e-14" z="2.32458640765866e-14"/>
      </physvol>
      <auxiliary auxtype="gammaECut" auxunit="MeV" auxvalue="0.00099"/>
      <auxiliary auxtype="electronECut" auxunit="MeV" auxvalue="0.00099"/>
      <auxiliary auxtype="positronECut" auxunit="MeV" auxvalue="0.00099"/>
      <auxiliary auxtype="protonECut" auxunit="MeV" auxvalue="0.1"/>
    </volume>

[ianna]

Left G4 geometry produced with CMS DD, right G4 geometry produced with DD4hep G4 converter:



Original ROOT geometry passed to DD4hep:

Set visible containers for ROOT geometry and G4 geometry produced from it by DD4hep:

[ianna]

HBLayer0 ROOT geometry on the left and G4 geometry produced from it on the right:

[ianna]

ROOT printout of the shape:

*** Shape hcalbarrelalgo:HBLayer0: TGeoPgon ***
    Nedges = 1
*** Shape hcalbarrelalgo:HBLayer0: TGeoPcon ***
    Nz    = 3
    phi1  =   -10.00000
    dphi  =    20.00000
     plane 0: z=    0.00000 Rmin=  177.50000 Rmax=  185.15000
     plane 1: z=  308.23896 Rmin=  177.50000 Rmax=  185.15000
     plane 2: z=  321.54272 Rmin=  185.15000 Rmax=  185.15000
 Bounding box:
*** Shape hcalbarrelalgo:HBLayer0: TGeoBBox ***
    dX =     6.60143
    dY =    32.64694
    dZ =   160.77136
    origin: x=  181.40481 y=    0.00000 z=  160.77136

[ianna]

GDML:

*** Shape hcalbarrelalgo_HBLayer0_shape_: TGeoPgon ***
    Nedges = 1
*** Shape hcalbarrelalgo_HBLayer0_shape_: TGeoPcon ***
    Nz    = 3
    phi1  =   350.00000
    dphi  =    10.00000
     plane 0: z=    0.00000 Rmin=  177.50000 Rmax=  185.15000
     plane 1: z=  308.23896 Rmin=  177.50000 Rmax=  185.15000
     plane 2: z=  321.54272 Rmin=  185.15000 Rmax=  185.15000
 Bounding box:
*** Shape hcalbarrelalgo_HBLayer0_shape_: TGeoBBox ***
    dX =     5.52693
    dY =    16.13689
    dZ =   160.77136
    origin: x=  180.33031 y=  -16.13689 z=  160.77136

[MarkusFrankATcernch]

I think this is it -- insert for GetPhi1() "-10" degrees....:

  template <> G4VSolid* convertShape<TGeoPgon>(const TGeoShape* shape)  {
    const TGeoPgon* sh = (const TGeoPgon*) shape;
    double phi_start = sh->GetPhi1() * DEGREE_2_RAD;
    double phi_total = (sh->GetDphi() + sh->GetPhi1()) * DEGREE_2_RAD;
    vector<double> rmin, rmax, z;
    for (Int_t i = 0; i < sh->GetNz(); ++i) {
      rmin.emplace_back(sh->GetRmin(i) * CM_2_MM);
      rmax.emplace_back(sh->GetRmax(i) * CM_2_MM);
      z.emplace_back(sh->GetZ(i) * CM_2_MM);
    }
    return new G4Polyhedra(sh->GetName(), phi_start, phi_total, sh->GetNedges(), sh->GetNz(), &z[0], &rmin[0], &rmax[0]);
  }

But I will verify this suspicion.

[MarkusFrankATcernch]

Hi Yana,
please test.
P.S. Thanks for these last 2 dumps. This really clarified the problem!
You apparently were the first ones where start_phi was != NULL....

[petricm]

@MarkusFrankATcernch you have used BoundingLimits, but G4VSolid does not have this member until Geant4 10.4

[ianna]

fixed, thanks!