GENXICC2.1: An improved version of GENXICC for hadronic production of doubly heavy baryons

详细信息    查看全文

• 作者：Xian-You Wang^a ; ^b ; ^{xywang@cern.ch} ; Xing-Gang Wu^a ; ^{wuxg@cqu.edu.cn}
• 关键词：Hadronic production ; Doubly heavy baryon
• 刊名：Computer Physics Communications
• 出版年：2013
• 出版时间：March, 2013
• 年：2013
• 卷：184
• 期：3
• 页码：1070-1074
• 全文大小：241 K

文摘

We present an improved version of GENXICC, which is a generator for hadronic production of the doubly heavy baryons , and and has been introduced by C.H. Chang, J.X. Wang and X.G. Wu [Comput. Phys. Commun. 177 (2007) 467; Comput. Phys. Commun. 181 (2010) 1144]. In comparison with the previous GENXICC versions, we update the program in order to generate the unweighted baryon events more effectively under various simulation environments, whose distributions are now generated according to the probability proportional to the integrand. One Les Houches Event (LHE) common block has been added to produce a standard LHE data file that contains useful information of the doubly heavy baryon and its accompanying partons. Such LHE data can be conveniently imported into PYTHIA to do further hadronization and decay simulation, especially, the color-flow problem can be solved with PYTHIA8.0.

NEW VERSION PROGRAM SUMMARY

Title of program: GENXICC2.1

Program obtained from: CPC Program Library

Reference to original program: GENXICC

Reference in CPC: Comput. Phys. Commun. 177, 467 (2007); Comput. Phys. Commun. 181, 1144 (2010)

Does the new version supersede the old program: No

Computer: Any LINUX based on PC with FORTRAN 77 or FORTRAN 90 and GNU C compiler as well

Operating systems: LINUX

Programming language used: FORTRAN 77/90

Memory required to execute with typical data: About 2.0 MB

No. of bytes in distributed program: About 2 MB, including PYTHIA6.4

Distribution format:?.tar.gz

Nature of physical problem: Hadronic production of doubly heavy baryons , and .

Method of solution: The upgraded version with a proper interface to PYTHIA can generate full production and decay events, either weighted or unweighted, conveniently and effectively. Especially, the unweighted events are generated by using an improved hit-and-miss approach.

Reasons for new version: Responding to the feedback from users of CMS and LHCb groups at the Large Hadron Collider, and based on the recent improvements of PYTHIA on the color-flow problem, we improve the efficiency for generating the unweighted events, and also improve the color-flow part for further hadronization. Especially, an interface has been added to import the output production events into a suitable form for PYTHIA8.0 simulation, in which the color-flow during the simulation can be correctly set.

Typical running time: It depends on which option is chosen to match PYTHIA when generating the full events and also on which mechanism is chosen to generate the events. Typically, for the dominant gluon-gluon fusion mechanism to generate the mixed events via the intermediate diquarks in and states, setting IDWTUP=3 and unwght =.true., it takes 30?min to generate 10⁵ unweighted events on a 2.27?GHz Intel Xeon E5520 processor machine; setting IDWTUP=3 and unwght =.false. or IDWTUP=1 and IGENERATE=0, it only needs 2?min to generate the 10⁵ baryon events (the fastest way, for theoretical purposes only). As a comparison, for previous GENXICC versions, if setting IDWTUP=1 and IGENERATE=1, it takes about 22 hours to generate 1000 unweighted events.

Keywords: Event generator; Doubly heavy baryons; Hadronic production.

Summary of the changes (improvements): (1) The scheme for generating unweighted events has been improved; (2) One Les Houches Event (LHE) common block has been added to record the standard LHE data in order to be the correct input for PYTHIA8.0 for later simulation; (3) We present the code for connecting GENXICC to PYTHIA8.0, where three color-flows have to be correctly set for later simulation. More specifically, we present the changes together with their detailed explanations in the following:

?

Unweighted events generation. For theoretical studies, e.g. to derive the total baryon production cross-section or various differential distributions, one can directly use the fastest way, e.g. setting the PYTHIA parameter IDWTUP=3 and unwght =.false. or setting IDWTUP=1 and IGENERATE=0 (in these cases, xmaxup should be set as 0), to generate the baryon events?. By using GENXICC? in this way, some interesting properties for hadronic production of , and have been found in the literature, cf. Refs. . While, for the events simulation in detector conditions, it is necessary to get the unweighted events. In previous GENXICC versions, the unweighted events are generated by setting IDWGTUP=1 and IGENERATE=1; i.e., the events are generated according to PYTHIA¡¯s inner mechanism, the so-called hit-and-miss approach (von Neumann algorithm), to reject those unsatisfied events and output the allowed events. But, as is well-known, the original hit-and-miss approach is really time-consuming. Some alterations must be made to improve its efficiency.

As an intermediate step, in BCVEGPY2.1a? we have suggested a practical trick to increase the efficiency of generating unweighted events (BCVEGPY is a generator for hadronic production mesons?). In this trick, other than choosing the maximum differential cross-section as a reference weight in the hit-and-miss approach, we directly select an effective differential cross-section, which is smaller than the maximum one, as the reference weight?. This treatment can greatly improve the generation efficiency without affecting the total cross-section of the process. However, in using this trick to generate unweighted events such as for CMS detector simulation, one will incidentally find a false peak in the distributions. This is caused by the fact that sometimes the same event will be stored a (false) large number of times in the hit-and-miss process. Then, we are facing a dilemma: such a false peak can be avoided by raising the effective reference weight to a value approaching the maximum weight, but, conversely, a larger reference weight will surely lead to a much longer running time.

One observes that by using the VEGAS algorithm?, the SPRING-BASES program? performs the integration in using the BASES subroutines and generates events with a probability proportional to the integrand in using the SPRING subroutines. After each iteration of VEGAS running, the integration result and the maximum value of the function will be stored in a file for each cell of the adaptive mesh. In the generation stage, a cell is chosen with a probability proportional to the corresponding integral, and then a point in the cell is generated using the hit-and-miss approach. This method is highly efficient, but it has the disadvantage that the required amount of storage space grows exponentially with the integration dimension.

Next, in the POWHEG program? the authors have developed a new method MINT? to replace the SPRING-BASES package. This MINT package also use the VEGAS algorithm to perform the integration. What¡¯s different is that it does not store the value of the integral but stores the upper bound value for each cell. The multidimensional stepwise function that equals the upper bound of the function to be integrated in each cell is in fact an upper bound for the whole function, which is the wanted upper bound for BCVEGPY2.1a or PYTHIA. So, the program is to find the upper bound grid for those cells. And next, by using again the hit-and-miss technique in each cell, one can generate the points according to the original distribution.

Based on these methods, as a further improvement, we present an ultimate solution to generate unweighted events in the present new GENXICC version. We adopt the MINT algorithm but with certain alterations to do the simulation. For this purpose, we change the VEGAS subroutine as follows. Three new variables have been added in the original VEGAS subroutine, where xint is the integral value for the integrand fxn after a ndim-dimensional integration, the xmax array records the upper bounding envelope of the integrand in all cells, imode is a flag:

vegas(fxn,ndim,ncall,itmx,nprn,xint,xmax,imode)

-

When called with imode=0, vegas performs the integration over the integrand fxn, and stores the answer in a common block parameter vegsec.

-

xmax stands for a (nvegbin,ndim) dimensional array, where nvegbin denotes the bin number for each coordinate, and ndim stands for the integration dimension. When called with imode=1, vegas will first initialize all the elements of xmax to be , where xint equals the value of vegsec that has been derived from a previous VEGAS running with imode=0. During the following sampling iteration, when the calculated integral value is larger than the initial xmax(nvegbin,ndim) value in a specific cell, then the value of xmax(nvegbin,ndim) for this cell will be increased by a fixed factor . After a sufficiently large number of calls, the values of xmax(nvegbin,ndim) will be stabilized for all cells. Such a final xmax array will be stored in the same grid file as that of the importance sampling function in order to do the final simulation.

Comparing to the previous GENXICC versions, in doing the initialization (subroutine evntinit), we will call vegas twice with imode=0 and imode=1 accordingly to generate the upper bound grid xmax and also a more precise importance sampling function. Practically, the user can directly use the existing grid file derived by previous VEGAS running to generate events by setting methodevnt=2 or methodevnt=3 without running VEGAS again, which is the same as in the older GENXICC versions.