Professional Documents
Culture Documents
Comment On Non-Mean-Field Behavior of The Contact Process On Scale-Free Networks''
Comment On Non-Mean-Field Behavior of The Contact Process On Scale-Free Networks''
-β/ν||
0.4240
-1.2 0.4250
Recently, Castellano and Pastor-Satorras [1] used the -1.4 -1.33
finite size scaling (FSS) theory to analyze simulation -1.6
data for the contact process (CP) on scale-free networks -1.8
0.001 0.01 0.1
(SFNs) and claimed that its absorbing critical behavior is
1/t
not consistent with the mean-field (MF) prediction.
Furthermore, they pointed out large density fluctuations FIG. 2 (color online). Effective exponents =jj for 2:75
at highly connected vertices as a possible origin for non- with N 107 . All samples are alive when data are obtained.
MF critical behavior. In this Comment, we propose a
scaling theory for relative density fluctuations in the spirit
of the MF theory, which turns out to explain simulation with N at criticality, one may assume a critical
data perfectly well. We also measure the value of the scaling relation as
critical density decay exponent =k , which agrees well rk N 1==2 fkN 1= ; (3)
with the MF prediction. Our results strongly support that
1=2 1=2
the CP on SFNs still exhibits a MF-type critical behavior. where fx x as x ! 0 and fx x as x !
In contrast to equilibrium models with thermal noise, 1. In Fig. 1, our simulation results at criticality for
nonequilibrium models exhibiting absorbing critical phe- 2:75 and 4.0 with various sizes N 103 ; . . . ; 106 show
nomena involve so-called multiplicative noise. In particu- excellent agreement with our prediction. Note that the
lar, the noise amplitude is proportional to the square root of case of 4:0 where the -independent ordinary MF
the particle density in models belonging to the directed appears also shows the similar behavior in rk .
percolation class such as CP [2]. Therefore, at the MF level We also observe the density decay dynamics for
neglecting spatial fluctuations, the particle number fluctua- 2:75 in Fig. 2, where the effective decay exponents =k
tion should be proportional to the square root of the total are plotted against 1=t. Our best estimate is =k
number of particles. It is easy to derive the relative density 1:345, which agrees well with the MF prediction of
fluctuation on vertices of degree k as k 1 and 1= 2 for < 3.
Finally, we comment on the FSS exponent =.
rk k =k Nk Pk 1=2 ; (1) Vertices of degree k > N 1= are practically irrelevant.
Accordingly, the cutoff in k bigger than N 1= cannot in-
where k is the partial density on vertices of degree k, N is
fluence critical FSS which should be determined by the
the total number of vertices, and Pk is the degree distribu-
competition between correlated volume and system size.
tion. As k k near criticality (: total density), we find,
We conjecture that 1= 2 from the
for reasonably large k,
hyperscaling-type argument [3]. Our numerical estimate
rk N1=2 k1=2 ; (2) = 0:592 for 2:75 agrees well with our conjec-
ture. In fact, the results of Ref. [1] also agree reasonably
where is the degree exponent. well with our conjecture. In conclusion, we believe that the
For large enough k > N 1= where the degree distribution CP on SFNs exhibits the MF-type critical behavior.
becomes almost flat in practice [NPk O1], we expect This work was supported in part by the KRF Grant
rk k1=2 . With these two limiting results and also (MOEHRD) (No. R14-2002-059-01000-0).
γ=2.75-10
3 Meesoon
1
Ha,1,2 Hyunsuk Hong,2 and Hyunggyu Park1
1045 School of Physics
10 -0.5 Korea Institute for Advanced Study
6
10 0.875
γ=4.0-103 Seoul 130-722, Korea
rk N -y
1 2
104 Department of Physics and RIPC
105 -0.5 Chonbuk National University
6
10 1.5 Jeonju 561-756, Korea