source: trunk/Documentation/geant4/UserDocumentation/UsersGuides/ForApplicationDeveloper/html/ch03s07.html@ 901

Last change on this file since 901 was 901, checked in by garnier, 17 years ago

Add Geant4 Documentation at 8.12.2008
File size: 36.3 KB
Line 
1<html><head><meta http-equiv="Content-Type" content="text/html; charset=ISO-8859-1"><title>3.7.  Event Biasing Techniques</title><link rel="stylesheet" href="../xml/XSLCustomizationLayer/G4HTMLStylesheet.css" type="text/css"><meta name="generator" content="DocBook XSL Stylesheets V1.71.1"><link rel="start" href="index.html" title="Geant4 User's Guide for Application Developers"><link rel="up" href="ch03.html" title="Chapter 3.  Toolkit Fundamentals"><link rel="prev" href="ch03s06.html" title="3.6.  Event Generator Interface"><link rel="next" href="ch04.html" title="Chapter 4.  Detector Definition and Response"><script language="JavaScript">
2function remote_win(fName)
3{
4 var url = "AllResources/Detector/geometry.src/" + fName;
5 RemoteWin=window.open(url,"","resizable=no,toolbar=0,location=0,directories=0,status=0,menubar=0,scrollbars=0,copyhistory=0,width=520,height=520")
6 RemoteWin.creator=self
7}
8</script></head><body bgcolor="white" text="black" link="#0000FF" vlink="#840084" alink="#0000FF"><div class="navheader"><table width="100%" summary="Navigation header"><tr><th colspan="3" align="center">3.7. 
9Event Biasing Techniques
10</th></tr><tr><td width="20%" align="left"><a accesskey="p" href="ch03s06.html"><img src="AllResources/IconsGIF/prev.gif" alt="Prev"></a> </td><th width="60%" align="center">Chapter 3. 
11Toolkit Fundamentals
12</th><td width="20%" align="right"> <a accesskey="n" href="ch04.html"><img src="AllResources/IconsGIF/next.gif" alt="Next"></a></td></tr></table><hr></div><div class="sect1" lang="en"><div class="titlepage"><div><div><h2 class="title" style="clear: both"><a name="sect.EvtBias"></a>3.7. 
13Event Biasing Techniques
14</h2></div></div></div><div class="sect2" lang="en"><div class="titlepage"><div><div><h3 class="title"><a name="sect.EvtBias.ScorImpRoul"></a>3.7.1. 
15Scoring, Geometrical Importance Sampling and Weight Roulette
16</h3></div></div></div><p>
17Geant4 provides event biasing techniques which may be used to save
18computing time in such applications as the simulation of radiation
19shielding. These are <span class="emphasis"><em>geometrical splitting
20</em></span> and <span class="emphasis"><em>Russian roulette</em></span>
21(also called geometrical importance sampling), and
22<span class="emphasis"><em>weight roulette</em></span>. Scoring is carried out by <span class="emphasis"><em>G4MultiFunctionalDetector</em></span> (see <a href="ch04s04.html#sect.Hits.G4Multi" title="4.4.5. 
23G4MultiFunctionalDetector and
24G4VPrimitiveScorer
25">Section 4.4.5</a> and <a href="ch04s04.html#sect.Hits.G4VPrim" title="4.4.6. 
26Concrete classes of G4VPrimitiveScorer
27">Section 4.4.6</a>) using the standard Geant4 scoring technique.
28Biasing specific scorers have been implemented and are described within
29<span class="emphasis"><em>G4MultiFunctionDetector</em></span> documentation. In this chapter, it is assumed that
30the reader is familiar with both the usage of Geant4 and the
31concepts of importance sampling. More detailed documentation may be
32found in the documents
33<a href="http://geant4.web.cern.ch/geant4/collaboration/working_groups/geometry/index.shtml" target="_top">
34'Latest development in importance sampling and scoring'
35</a>.
36A detailed description of different use-cases which employ the sampling
37and scoring techniques can be found in the document
38<a href="http://geant4.web.cern.ch/geant4/collaboration/working_groups/geometry/index.shtml" target="_top">
39'Scoring and geometrical importance sampling use cases'
40</a>.
41</p><p>
42The purpose of importance sampling is to save computing time by
43sampling less often the particle histories entering "less
44important" geometry regions, and more often in more "important"
45regions. Given the same amount of computing time, an
46importance-sampled and an analogue-sampled simulation must show
47equal mean values, while the importance-sampled simulation will
48have a decreased variance.
49</p><p>
50The implementation of scoring is independent of the implementation
51of importance sampling. However both share common concepts.
52<span class="emphasis"><em>Scoring and importance sampling apply to particle types chosen
53by the user</em></span>, which should be borne in mind when interpreting the
54output of any biased simulation.
55</p><p>
56Examples on how to use scoring and importance sampling may be found
57in <code class="literal">examples/extended/biasing</code> and
58<code class="literal">examples/advanced/Tiara</code>.
59</p><div class="sect3" lang="en"><div class="titlepage"><div><div><h4 class="title"><a name="sect.EvtBias.ScorImpRoul.Geom"></a>3.7.1.1. 
60Geometries
61</h4></div></div></div><p>
62The kind of scoring referred to in this note and the importance
63sampling apply to spatial cells provided by the user.
64</p><p>
65A <span class="bold"><strong>cell</strong></span> is a physical volume (further specified
66by it's replica number, if the volume is a replica). Cells may be defined
67in two kinds of geometries:
68
69</p><div class="orderedlist"><ol type="1" compact><li><p>
70 <span class="bold"><strong>mass geometry</strong></span>: the geometry setup of the
71 experiment to be simulated. Physics processes apply to this geometry.
72 </p></li><li><p>
73 <span class="bold"><strong>parallel-geometry</strong></span>: a geometry constructed
74 to define the physical volumes according to which scoring and/or importance
75 sampling is applied.
76 </p></li></ol></div><p>
77</p><p>
78The user has the choice to score and/or sample by importance the
79particles of the chosen type, according to mass geometry or to
80parallel geometry. It is possible to utilize several parallel
81geometries in addition to the mass geometry. This provides the user
82with a lot of flexibility to define separate geometries for
83different particle types in order to apply scoring or/and
84importance sampling.
85</p><p>
86</p><div class="note" style="margin-left: 0.5in; margin-right: 0.5in;"><h3 class="title">Note</h3>
87 Parallel geometries should be constructed using the implementation as
88 described in <a href="ch04s07.html" title="4.7. 
89Parallel Geometries
90">Section 4.7</a>.
91
92 There are a few conditions for parallel geometries:
93
94 <div class="itemizedlist"><ul type="disc" compact><li><p>
95 The world volume for parallel and mass geometries must be identical copies.
96 </p></li><li><p>
97 Scoring and importance cells must not share boundaries with the world volume.
98 </p></li></ul></div></div><p>
99</p></div><div class="sect3" lang="en"><div class="titlepage"><div><div><h4 class="title"><a name="sect.EvtBias.ScorImpRoul.ChgSamp"></a>3.7.1.2. 
100Changing the Sampling
101</h4></div></div></div><p>
102Samplers are higher level tools which perform the necessary
103changes of the Geant4 sampling in order to apply importance
104sampling and weight roulette.
105</p><p>
106Variance reduction (and scoring through the
107<span class="emphasis"><em>G4MultiFunctionalDetector</em></span>)
108may be combined arbitrarily for chosen particle types and may be applied to the
109mass or to parallel geometries.
110</p><p>
111The <code class="literal">G4GeometrySampler</code> can be applied equally to mass or
112parallel geometries with an abstract interface supplied by <code class="literal">G4VSampler</code>.
113<code class="literal">G4VSampler</code> provides
114<code class="literal">Prepare...</code> methods and a <code class="literal">Configure</code>
115method:
116
117<a name="anchor_EvtBias_G4VSampler"></a>
118</p><div class="informalexample"><pre class="programlisting">
119 class G4VSampler
120 {
121 public:
122 G4VSampler();
123 virtual ~G4VSampler();
124 virtual void PrepareImportanceSampling(G4VIStore *istore,
125 const G4VImportanceAlgorithm
126 *ialg = 0) = 0;
127 virtual void PrepareWeightRoulett(G4double wsurvive = 0.5,
128 G4double wlimit = 0.25,
129 G4double isource = 1) = 0;
130 virtual void PrepareWeightWindow(G4VWeightWindowStore *wwstore,
131 G4VWeightWindowAlgorithm *wwAlg = 0,
132 G4PlaceOfAction placeOfAction =
133 onBoundary) = 0;
134 virtual void Configure() = 0;
135 virtual void ClearSampling() = 0;
136 virtual G4bool IsConfigured() const = 0;
137 };
138</pre></div><p>
139</p><p>
140The methods for setting up the desired combination need specific
141information:
142
143</p><div class="itemizedlist"><ul type="disc" compact><li><p>
144 Importance sampling: message <code class="literal">PrepareImportanceSampling</code>
145 with a
146 <a href="ch03s07.html#anchor_EvtBias_G4VIStore">
147 <code class="literal">G4VIStore</code></a>
148 and optionally a <code class="literal">G4VImportanceAlgorithm</code>
149 </p></li><li><p>
150 Weight window: message <code class="literal">PrepareWeightWindow</code> with the
151 arguments:
152 </p><div class="itemizedlist"><ul type="circle" compact><li><p>
153 <span class="emphasis"><em>*wwstore</em></span>: a
154 <code class="literal">G4VWeightWindowStore</code> for retrieving the lower
155 weight bounds for the energy-space cells
156 </p></li><li><p>
157 <span class="emphasis"><em>*wwAlg</em></span>: a
158 <code class="literal">G4VWeightWindowAlgorithm</code> if a customized algorithm
159 should be used
160 </p></li><li><p>
161 <span class="emphasis"><em>placeOfAction</em></span>: a
162 <code class="literal">G4PlaceOfAction</code> specifying where to perform the
163 biasing
164 </p></li></ul></div><p>
165 </p></li><li><p>
166 Weight roulette: message <code class="literal">PrepareWeightRoulett</code> with the
167 optional parameters:
168 </p><div class="itemizedlist"><ul type="circle" compact><li><p>
169 <span class="emphasis"><em>wsurvive</em></span>: survival weight
170 </p></li><li><p>
171 <span class="emphasis"><em>wlimit</em></span>: minimal allowed
172 value of weight * source importance / cell importance
173 </p></li><li><p>
174 <span class="emphasis"><em>isource</em></span>: importance of the source cell
175 </p></li></ul></div><p>
176 </p></li></ul></div><p>
177</p><p>
178Each object of a sampler class is responsible for one particle
179type. The particle type is given to the constructor of the sampler
180classes via the particle type name, e.g. "neutron". Depending on
181the specific purpose, the <code class="literal">Configure()</code> of a sampler will
182set up specialized processes (derived from <code class="literal">G4VProcess</code>) for
183transportation in the parallel geometry, importance
184sampling and weight roulette for the given particle type. When
185<code class="literal">Configure()</code> is invoked the sampler places the processes in
186the correct order independent of the order in which user invoked
187the <code class="literal">Prepare...</code> methods.
188</p><p>
189</p><div class="note" style="margin-left: 0.5in; margin-right: 0.5in;"><h3 class="title">Note</h3><p>
190 </p><div class="itemizedlist"><ul type="disc" compact><li><p>
191 The <code class="literal">Prepare...()</code> functions may each only be invoked
192 once.
193 </p></li><li><p>
194 To configure the sampling the function <code class="literal">Configure()</code>
195 must be called <span class="emphasis"><em>after</em></span> the
196 <code class="literal">G4RunManager</code> has been initialized and the PhysicsList has
197 been instantiated.
198 </p></li></ul></div><p>
199 </p></div><p>
200</p><p>
201The interface and framework are demonstrated in the
202<code class="literal">examples/extended/biasing</code> directory, with the main changes being to the
203G4GeometrySampler class and the fact that in the parallel case the WorldVolume
204is a copy of the Mass World.
205
206The parallel geometry now has to inherit from
207<span class="emphasis"><em>G4VUserParallelWorld</em></span> which also has the <span class="emphasis"><em>GetWorld()</em></span> method
208in order to retrieve a copy of the mass geometry WorldVolume.
209
210</p><div class="informalexample"><pre class="programlisting">
211 class B02ImportanceDetectorConstruction : public G4VUserParallelWorld
212 ghostWorld = GetWorld();
213</pre></div><p>
214
215</p><p>
216The constructor for <span class="emphasis"><em>G4GeometrySampler</em></span> takes a pointer to
217the physical world volume and the particle type name (e.g. "neutron") as arguments.
218In a single mass geometry the sampler is created as follows:
219
220</p><div class="informalexample"><pre class="programlisting">
221 G4GeometrySampler mgs(detector-&gt;GetWorldVolume(),"neutron");
222 mgs.SetParallel(false);
223</pre></div><p>
224
225
226Whilst the following lines of code are required in order to set up the sampler for the
227parallel geometry case:
228
229</p><div class="informalexample"><pre class="programlisting">
230 G4VPhysicalVolume* ghostWorld = pdet-&gt;GetWorldVolume();
231
232 G4GeometrySampler pgs(ghostWorld,"neutron");
233
234 pgs.SetParallel(true);
235</pre></div><p>
236
237
238Also note that the preparation and configuration of the samplers has to be
239carried out <span class="emphasis"><em>after</em></span> the instantiation of the UserPhysicsList
240and after the initialisation of the <span class="emphasis"><em>G4RunManager</em></span>:
241
242</p><div class="informalexample"><pre class="programlisting">
243 pgs.PrepareImportanceSampling(&amp;aIstore, 0);
244 pgs.Configure();
245</pre></div><p>
246
247Due to the fact that biasing is a process and has to be inserted after all the
248other processes have been created.
249</p></div><div class="sect3" lang="en"><div class="titlepage"><div><div><h4 class="title"><a name="sect.EvtBias.ScorImpRoul.ImpSamp"></a>3.7.1.3. 
250Importance Sampling
251</h4></div></div></div><p>
252Importance sampling acts on particles crossing boundaries
253between "importance cells". The action taken depends on the
254importance values assigned to the cells. In general a particle
255history is either split or Russian roulette is played if the
256importance increases or decreases, respectively. A weight assigned
257to the history is changed according to the action taken.
258</p><p>
259The tools provided for importance sampling require the user to
260have a good understanding of the physics in the problem. This is
261because the user has to decide which particle types require
262importance sampled, define the cells, and assign importance values
263to the cells. If this is not done properly the results cannot be
264expected to describe a real experiment.
265</p><p>
266The assignment of importance values to a cell is done using an
267importance store described below.
268</p><p>
269An "importance store" with the interface <code class="literal">G4VIStore</code> is
270used to store importance values related to cells. In order to do
271importance sampling the user has to create an object (e.g. of class
272<code class="literal">G4IStore</code>) of type <code class="literal">G4VIStore</code>. The samplers may be
273given a <code class="literal">G4VIStore</code>. The user fills the store with cells and
274their importance values.
275</p><p>
276An importance store has to be constructed with a reference to
277the world volume of the geometry used for importance sampling. This
278may be the world volume of the mass or of a parallel geometry.
279Importance stores derive from the interface <code class="literal">G4VIStore</code>:
280
281<a name="anchor_EvtBias_G4VIStore"></a>
282</p><div class="informalexample"><pre class="programlisting">
283 class G4VIStore
284 {
285 public:
286 G4VIStore();
287 virtual ~G4VIStore();
288 virtual G4double GetImportance(const G4GeometryCell &amp;gCell) const = 0;
289 virtual G4bool IsKnown(const G4GeometryCell &amp;gCell) const = 0;
290 virtual const G4VPhysicalVolume &amp;GetWorldVolume() const = 0;
291 };
292</pre></div><p>
293</p><p>
294A concrete implementation of an importance store is provided by
295the class <code class="literal">G4VStore</code>. The <span class="emphasis"><em>public</em></span>
296part of the class is:
297
298</p><div class="informalexample"><pre class="programlisting">
299 class G4IStore : public G4VIStore
300 {
301 public:
302 explicit G4IStore(const G4VPhysicalVolume &amp;worldvolume);
303 virtual ~G4IStore();
304 virtual G4double GetImportance(const G4GeometryCell &amp;gCell) const;
305 virtual G4bool IsKnown(const G4GeometryCell &amp;gCell) const;
306 virtual const G4VPhysicalVolume &amp;GetWorldVolume() const;
307 void AddImportanceGeometryCell(G4double importance,
308 const G4GeometryCell &amp;gCell);
309 void AddImportanceGeometryCell(G4double importance,
310 const G4VPhysicalVolume &amp;,
311 G4int aRepNum = 0);
312 void ChangeImportance(G4double importance,
313 const G4GeometryCell &amp;gCell);
314 void ChangeImportance(G4double importance,
315 const G4VPhysicalVolume &amp;,
316 G4int aRepNum = 0);
317 G4double GetImportance(const G4VPhysicalVolume &amp;,
318 G4int aRepNum = 0) const ;
319 private: .....
320 };
321</pre></div><p>
322</p><p>
323The member function <code class="literal">AddImportanceGeometryCell()</code> enters
324a cell and an importance value into the importance store. The
325importance values may be returned either according to a physical
326volume and a replica number or according to a
327<code class="literal">G4GeometryCell</code>. The user must be aware of the
328interpretation of assigning importance values to a cell.
329If scoring is also implemented then this is attached to logical volumes, in
330which case the physical volume and replica number method should be used for
331assigning importance values. See <code class="literal">examples/extended/biasing
332B01</code> and <code class="literal">B02</code> for examples of this.
333</p><p>
334</p><div class="note" style="margin-left: 0.5in; margin-right: 0.5in;"><h3 class="title">Note</h3><p>
335 </p><div class="itemizedlist"><ul type="disc" compact><li><p>
336 An importance value must be assigned to every cell.
337 </p></li></ul></div><p>
338 </p></div><p>
339</p><p>
340The different cases:
341
342</p><div class="itemizedlist"><ul type="disc" compact><li><p>
343 <span class="emphasis"><em>Cell is not in store</em></span>
344 </p><p>
345 Not filling a certain cell in the store will cause an
346 exception.
347 </p><p>
348 </p></li><li><p>
349 <span class="emphasis"><em>Importance value = zero</em></span>
350 </p><p>
351 Tracks of the chosen particle type will be killed.
352 </p><p>
353 </p></li><li><p>
354 <span class="emphasis"><em>importance values &gt; 0</em></span>
355 </p><p>
356 Normal allowed values
357 </p><p>
358 </p></li><li><p>
359 <span class="emphasis"><em>Importance value smaller zero</em></span>
360 </p><p>
361 Not allowed!
362 </p><p>
363 </p></li></ul></div><p>
364</p></div><div class="sect3" lang="en"><div class="titlepage"><div><div><h4 class="title"><a name="sect.EvtBias.ScorImpRoul.SampAlgor"></a>3.7.1.4. 
365The Importance Sampling Algorithm
366</h4></div></div></div><p>
367Importance sampling supports using a customized importance
368sampling algorithm. To this end, the sampler interface
369<a href="ch03s07.html#anchor_EvtBias_G4VSampler">
370<code class="literal">G4VSampler</code></a>
371may be given a pointer to the interface
372<code class="literal">G4VImportanceAlgorithm</code>:
373
374</p><div class="informalexample"><pre class="programlisting">
375 class G4VImportanceAlgorithm
376 {
377 public:
378 G4VImportanceAlgorithm();
379 virtual ~G4VImportanceAlgorithm();
380 virtual G4Nsplit_Weight Calculate(G4double ipre,
381 G4double ipost,
382 G4double init_w) const = 0;
383 };
384</pre></div><p>
385</p><p>
386The method <code class="literal">Calculate()</code> takes the arguments:
387
388</p><div class="itemizedlist"><ul type="disc" compact><li><p>
389 <span class="emphasis"><em>ipre, ipost</em></span>: importance
390 of the previous cell and the importance of the current cell,
391 respectively.
392 </p></li><li><p>
393 <span class="emphasis"><em>init_w</em></span>: the particles weight
394 </p></li></ul></div><p>
395</p><p>
396It returns the struct:
397
398</p><div class="informalexample"><pre class="programlisting">
399 class G4Nsplit_Weight
400 {
401 public:
402
403 G4int fN;
404 G4double fW;
405 };
406</pre></div><p>
407
408</p><div class="itemizedlist"><ul type="disc" compact><li><p>
409 <span class="emphasis"><em>fN</em></span>: the calculated
410 number of particles to exit the importance sampling
411 </p></li><li><p>
412 <span class="emphasis"><em>fW</em></span>: the weight of the particles
413 </p></li></ul></div><p>
414</p><p>
415The user may have a customized algorithm used by providing a
416class inheriting from <code class="literal">G4VImportanceAlgorithm</code>.
417</p><p>
418If no customized algorithm is given to the sampler the default
419importance sampling algorithm is used. This algorithm is
420implemented in <code class="literal">G4ImportanceAlgorithm</code>.
421</p></div><div class="sect3" lang="en"><div class="titlepage"><div><div><h4 class="title"><a name="sect.EvtBias.ScorImpRoul.WeightWin"></a>3.7.1.5. 
422The Weight Window Technique
423</h4></div></div></div><p>
424The weight window technique is a weight-based alternative to
425importance sampling:
426
427</p><div class="itemizedlist"><ul type="disc" compact><li><p>
428 applies splitting and Russian roulette depending on space
429 (cells) and energy
430 </p></li><li><p>
431 user defines weight windows in contrast to defining importance
432 values as in importance sampling
433 </p></li></ul></div><p>
434</p><p>
435In contrast to importance sampling this technique is not weight
436blind. Instead the technique is applied according to the particle
437weight with respect to the current energy-space cell.
438</p><p>
439Therefore the technique is convenient to apply in combination
440with other variance reduction techniques such as cross-section
441biasing and implicit capture.
442</p><p>
443A weight window may be specified for every cell and for several
444energy regions: <span class="emphasis"><em>space-energy cell</em></span>.
445
446</p><div class="figure"><a name="fig.EvtBias.WeightWindow"></a><div class="figure-contents"><div class="mediaobject" align="center"><img src="./AllResources/Fundamentals/wwconcept.gif" align="middle" alt="Weight window concept"></div></div><p class="title"><b>Figure 3.2. 
447 Weight window concept
448</b></p></div><p><br class="figure-break">
449</p><h5><a name="id374547"></a>
450Weight window concept
451</h5><p>
452 The user specifies a <span class="emphasis"><em>lower weight bound W_L</em></span>
453 for every space-energy cell.
454
455 </p><div class="itemizedlist"><ul type="disc" compact><li><p>
456 The upper weight bound W_U and the survival weight W_S are
457 calculated as:
458 </p><p>
459 W_U = C_U <span class="emphasis"><em>W_L</em></span> and
460 </p><p>
461 </p><p>
462 W_S = C_S <span class="emphasis"><em>W_L</em></span>.
463 </p><p>
464 </p></li><li><p>
465 The user specifies C_S and C_U once for the whole problem.
466 </p></li><li><p>
467 The user may give different sets of energy bounds for every cell
468 or one set for all geometrical cells
469 </p></li><li><p>
470 Special case: if C_S = C_U = 1 for all energies then weight
471 window is equivalent to importance sampling
472 </p></li><li><p>
473 The user can choose to apply the technique: at boundaries, on
474 collisions or on boundaries and collisions
475 </p></li></ul></div><p>
476</p><p>
477The energy-space cells are realized by <code class="literal">G4GeometryCell</code>
478as in importance sampling. The cells are stored in a weight window
479store defined by <code class="literal">G4VWeightWindowStore</code>:
480
481</p><div class="informalexample"><pre class="programlisting">
482 class G4VWeightWindowStore {
483 public:
484 G4VWeightWindowStore();
485 virtual ~G4VWeightWindowStore();
486 virtual G4double GetLowerWeitgh(const G4GeometryCell &amp;gCell,
487 G4double partEnergy) const = 0;
488 virtual G4bool IsKnown(const G4GeometryCell &amp;gCell) const = 0;
489 virtual const G4VPhysicalVolume &amp;GetWorldVolume() const = 0;
490 };
491</pre></div><p>
492</p><p>
493A concrete implementation is provided:
494
495</p><div class="informalexample"><pre class="programlisting">
496 class G4WeightWindowStore: public G4VWeightWindowStore {
497 public:
498 explicit G4WeightWindowStore(const G4VPhysicalVolume &amp;worldvolume);
499 virtual ~G4WeightWindowStore();
500 virtual G4double GetLowerWeitgh(const G4GeometryCell &amp;gCell,
501 G4double partEnergy) const;
502 virtual G4bool IsKnown(const G4GeometryCell &amp;gCell) const;
503 virtual const G4VPhysicalVolume &amp;GetWorldVolume() const;
504 void AddLowerWeights(const G4GeometryCell &amp;gCell,
505 const std::vector&lt;G4double&gt; &amp;lowerWeights);
506 void AddUpperEboundLowerWeightPairs(const G4GeometryCell &amp;gCell,
507 const G4UpperEnergyToLowerWeightMap&amp;
508 enWeMap);
509 void SetGeneralUpperEnergyBounds(const
510 std::set&lt;G4double, std::less&lt;G4double&gt; &gt; &amp; enBounds);
511
512 private::
513 ...
514 };
515</pre></div><p>
516</p><p>
517The user may choose equal energy bounds for all cells. In this
518case a set of upper energy bounds must be given to the store using
519the method <code class="literal">SetGeneralUpperEnergyBounds</code>. If a general set
520of energy bounds have been set <code class="literal">AddLowerWeights</code> can be used
521to add the cells.
522</p><p>
523Alternatively, the user may chose different energy regions for
524different cells. In this case the user must provide a mapping of
525upper energy bounds to lower weight bounds for every cell using the
526method <code class="literal">AddUpperEboundLowerWeightPairs</code>.
527</p><p>
528Weight window algorithms implementing the interface class
529<code class="literal">G4VWeightWindowAlgorithm</code> can be used to define a
530customized algorithm:
531
532</p><div class="informalexample"><pre class="programlisting">
533 class G4VWeightWindowAlgorithm {
534 public:
535 G4VWeightWindowAlgorithm();
536 virtual ~G4VWeightWindowAlgorithm();
537 virtual G4Nsplit_Weight Calculate(G4double init_w,
538 G4double lowerWeightBound) const = 0;
539 };
540</pre></div><p>
541</p><p>
542A concrete implementation is provided and used as a default:
543
544</p><div class="informalexample"><pre class="programlisting">
545 class G4WeightWindowAlgorithm : public G4VWeightWindowAlgorithm {
546 public:
547 G4WeightWindowAlgorithm(G4double upperLimitFaktor = 5,
548 G4double survivalFaktor = 3,
549 G4int maxNumberOfSplits = 5);
550 virtual ~G4WeightWindowAlgorithm();
551 virtual G4Nsplit_Weight Calculate(G4double init_w,
552 G4double lowerWeightBound) const;
553 private:
554 ...
555};
556</pre></div><p>
557</p><p>
558The constructor takes three parameters which are used to:
559calculate the upper weight bound (upperLimitFaktor), calculate the
560survival weight (survivalFaktor), and introduce a maximal number
561(maxNumberOfSplits) of copies to be created in one go.
562</p><p>
563In addition, the inverse of the maxNumberOfSplits is used to
564specify the minimum survival probability in case of Russian
565roulette.
566</p></div><div class="sect3" lang="en"><div class="titlepage"><div><div><h4 class="title"><a name="sect.EvtBias.ScorImpRoul.WeigRoul"></a>3.7.1.6. 
567The Weight Roulette Technique
568</h4></div></div></div><p>
569Weight roulette (also called weight cutoff) is usually applied
570if importance sampling and implicit capture are used together.
571Implicit capture is not described here but it is useful to note
572that this procedure reduces a particle weight in every collision
573instead of killing the particle with some probability.
574</p><p>
575Together with importance sampling the weight of a particle may
576become so low that it does not change any result significantly.
577Hence tracking a very low weight particle is a waste of computing
578time. Weight roulette is applied in order to solve this
579problem.
580</p><h5><a name="id374768"></a>
581The weight roulette concept
582</h5><p>
583Weight roulette takes into account the importance "Ic" of the
584current cell and the importance "Is" of the cell in which the
585source is located, by using the ratio "R=Is/Ic".
586</p><p>
587Weight roulette uses a relative minimal weight limit and a
588relative survival weight. When a particle falls below the weight
589limit Russian roulette is applied. If the particle survives,
590tracking will be continued and the particle weight will be set to
591the survival weight.
592</p><p>
593The weight roulette uses the following parameters with their
594default values:
595
596</p><div class="itemizedlist"><ul type="disc" compact><li><p>
597 <span class="emphasis"><em>wsurvival</em></span>: 0.5
598 </p></li><li><p>
599 <span class="emphasis"><em>wlimit</em></span>: 0.25
600 </p></li><li><p>
601 <span class="emphasis"><em>isource</em></span>: 1
602 </p></li></ul></div><p>
603</p><p>
604The following algorithm is applied:
605</p><p>
606If a particle weight "w" is lower than R*wlimit:
607
608</p><div class="itemizedlist"><ul type="disc" compact><li><p>
609 the weight of the particle will be changed to "ws = wsurvival*R"
610 </p></li><li><p>
611 the probability for the particle to survive is "p = w/ws"
612 </p></li></ul></div><p>
613</p></div></div><div class="sect2" lang="en"><div class="titlepage"><div><div><h3 class="title"><a name="sect.EvtBias.PhysBias"></a>3.7.2. 
614Physics Based Biasing
615</h3></div></div></div><p>
616Geant4 supports physics based biasing through a number of general
617use, built in biasing techniques. A utility class,
618G4WrapperProcess, is also available to support user defined
619biasing.
620</p><div class="sect3" lang="en"><div class="titlepage"><div><div><h4 class="title"><a name="sect.EvtBias.PhysBias.BuiltInOpt"></a>3.7.2.1. 
621Built in Biasing Options
622</h4></div></div></div><div class="sect4" lang="en"><div class="titlepage"><div><div><h5 class="title"><a name="sect.EvtBias.PhysBias.BuiltInOpt.PrimPart"></a>3.7.2.1.1. 
623Primary Particle Biasing
624</h5></div></div></div><p>
625Primary particle biasing can be used to increase the number of
626primary particles generated in a particular phase space region of
627interest. The weight of the primary particle is modified as
628appropriate. A general implementation is provided through the
629G4GeneralParticleSource class. It is possible to bias position,
630angular and energy distributions.
631</p><p>
632G4GeneralParticleSource is a concrete implementation of
633G4VPrimaryGenerator. To use, instantiate G4GeneralParticleSource in
634the G4VUserPrimaryGeneratorAction class, as demonstrated below.
635
636</p><div class="informalexample"><pre class="programlisting">
637 MyPrimaryGeneratorAction::MyPrimaryGeneratorAction() {
638 generator = new G4GeneralParticleSource;
639 }
640
641 void
642 MyPrimaryGeneratorAction::GeneratePrimaries(G4Event*anEvent){
643 generator-&gt;GeneratePrimaryVertex(anEvent);
644 }
645</pre></div><p>
646</p><p>
647The biasing can be configured through interactive commands.
648Extensive documentation can be found in
649<a href="http://reat.space.qinetiq.com/gps/" target="_top">
650Primary particle biasing</a>. Examples are also distributed
651with the Geant4 distribution in
652<span class="bold"><strong>examples/extended/eventgenerator/exgps</strong></span>.
653</p></div><div class="sect4" lang="en"><div class="titlepage"><div><div><h5 class="title"><a name="sect.EvtBias.PhysBias.BuiltInOpt.RadDcy"></a>3.7.2.1.2. 
654Radioactive Decay Biasing
655</h5></div></div></div><p>
656The G4RadioactiveDecay class simulates the decay of radioactive
657nuclei and implements the following biasing options:
658
659</p><div class="itemizedlist"><ul type="disc" compact><li><p>
660 Increase the sampling rate of radionuclides within observation
661 times through a user defined probability distribution function
662 </p></li><li><p>
663 Nuclear splitting, where the parent nuclide is split into a
664 user defined number of nuclides
665 </p></li><li><p>
666 Branching ratio biasing where branching ratios are sampled with
667 equal probability
668 </p></li></ul></div><p>
669</p><p>
670G4RadioactiveDecay is a process which must be registered with a
671process manager, as demonstrated below.
672
673</p><div class="informalexample"><pre class="programlisting">
674 void MyPhysicsList::ConstructProcess()
675 {
676 ...
677 G4RadioactiveDecay* theRadioactiveDecay =
678 new G4RadioactiveDecay();
679
680 G4ProcessManager* pmanager = ...
681 pmanager -&gt;AddProcess(theRadioactiveDecay);
682 ...
683 }
684</pre></div><p>
685</p><p>
686The biasing can be controlled either in compiled code or through
687interactive commands. Extensive documentation can be found in
688
689<a href="http://reat.space.qinetiq.com/septimess/exrdm/" target="_top">
690Radioactive decay biasing example
691</a>
692 and
693<a href="http://www.space.qinetiq.com/geant4/rdm.html" target="_top">
694Radioactive decay biasing
695</a>.
696</p><p>
697Radioactive decay biasing examples are also distributed with the Geant4
698distribution in
699<span class="bold"><strong>examples/extended/radioactivedecay/exrdm</strong></span>.
700</p></div><div class="sect4" lang="en"><div class="titlepage"><div><div><h5 class="title"><a name="sect.EvtBias.PhysBias.BuiltInOpt.HadLeadPar"></a>3.7.2.1.3. 
701Hadronic Leading Particle Biasing
702</h5></div></div></div><p>
703One hadronic leading particle biasing technique is
704implemented in the G4HadLeadBias utility. This method keeps only
705the most important part of the event, as well as representative
706tracks of each given particle type. So the track with the highest
707energy as well as one of each of Baryon, pi0, mesons and leptons.
708As usual, appropriate weights are assigned to the particles.
709Setting the <span class="bold"><strong>SwitchLeadBiasOn</strong></span>
710environmental variable will activate this utility.
711</p></div><div class="sect4" lang="en"><div class="titlepage"><div><div><h5 class="title"><a name="sect.EvtBias.PhysBias.BuiltInOpt.HadCrsSect"></a>3.7.2.1.4. 
712Hadronic Cross Section Biasing
713</h5></div></div></div><p>
714Cross section biasing artificially enhances/reduces the cross
715section of a process. This may be useful for studying thin layer
716interactions or thick layer shielding. The built in hadronic cross
717section biasing applies to photon inelastic, electron nuclear and
718positron nuclear processes.
719</p><p>
720The biasing is controlled through the
721<span class="bold"><strong>BiasCrossSectionByFactor</strong></span> method
722in G4HadronicProcess, as demonstrated below.
723
724</p><div class="informalexample"><pre class="programlisting">
725 void MyPhysicsList::ConstructProcess()
726 {
727 ...
728 G4ElectroNuclearReaction * theElectroReaction =
729 new G4ElectroNuclearReaction;
730
731 G4ElectronNuclearProcess theElectronNuclearProcess;
732 theElectronNuclearProcess.RegisterMe(theElectroReaction);
733 theElectronNuclearProcess.BiasCrossSectionByFactor(100);
734
735 pManager-&gt;AddDiscreteProcess(&amp;theElectronNuclearProcess);
736 ...
737 }
738</pre></div><p>
739</p><p>
740More details can be found in
741<a href="http://www.triumf.ca/geant4-03/talks/03-Wednesday-AM-1/03-J.Wellisch/biasing.hadronics.pdf" target="_top">
742Hadronic cross section biasing
743</a>.
744</p></div></div><div class="sect3" lang="en"><div class="titlepage"><div><div><h4 class="title"><a name="sect.EvtBias.PhysBias."></a>3.7.2.2. 
745G4WrapperProcess
746</h4></div></div></div><p>
747G4WrapperProcess can be used to implement user defined event
748biasing. G4WrapperProcess, which is a process itself, wraps an
749existing process. By default, all function calls are forwared to
750the wrapped process. It is a non-invasive way to modify the
751behaviour of an existing process.
752</p><p>
753To use this utility, first create a derived class inheriting
754from G4WrapperProcess. Override the methods whose behaviour you
755would like to modify, for example, PostStepDoIt, and register the
756derived class in place of the process to be wrapped. Finally,
757register the wrapped process with G4WrapperProcess. The code
758snippets below demonstrate its use.
759
760</p><div class="informalexample"><pre class="programlisting">
761 class MyWrapperProcess : public G4WrapperProcess {
762 ...
763 G4VParticleChange* PostStepDoIt(const G4Track&amp; track,
764 const G4Step&amp; step) {
765 // Do something interesting
766 }
767 };
768
769
770 void MyPhysicsList::ConstructProcess()
771 {
772 ...
773 G4LowEnergyBremsstrahlung* bremProcess =
774 new G4LowEnergyBremsstrahlung();
775
776 MyWrapperProcess* wrapper = new MyWrapperProcess();
777 wrapper-&gt;RegisterProcess(bremProcess);
778
779 processManager-&gt;AddProcess(wrapper, -1, -1, 3);
780 }
781</pre></div><p>
782</p></div></div></div><div class="navfooter"><hr><table width="100%" summary="Navigation footer"><tr><td width="40%" align="left"><a accesskey="p" href="ch03s06.html"><img src="AllResources/IconsGIF/prev.gif" alt="Prev"></a> </td><td width="20%" align="center"><a accesskey="u" href="ch03.html"><img src="AllResources/IconsGIF/up.gif" alt="Up"></a></td><td width="40%" align="right"> <a accesskey="n" href="ch04.html"><img src="AllResources/IconsGIF/next.gif" alt="Next"></a></td></tr><tr><td width="40%" align="left" valign="top">3.6. 
783Event Generator Interface
784 </td><td width="20%" align="center"><a accesskey="h" href="index.html"><img src="AllResources/IconsGIF/home.gif" alt="Home"></a></td><td width="40%" align="right" valign="top"> Chapter 4. 
785Detector Definition and Response
786</td></tr></table></div></body></html>
Note: See TracBrowser for help on using the repository browser.