Parameter Uncertainties on the Predictability of Periodicity and Chaos

Parameter Uncertainties on the Predictability of Periodicity and Chaos

E. S. Medeiros esm@if.usp.br Institute of Physics, University of São Paulo, Rua do Matão, Travessa R 187, 05508-090, São Paulo, Brazil Institute for Complex Systems and Mathematical Biology, SUPA, University of Aberdeen, AB24 3UE Aberdeen, United Kingdom    I. L. Caldas Institute of Physics, University of São Paulo, Rua do Matão, Travessa R 187, 05508-090, São Paulo, Brazil    M. S. Baptista Institute for Complex Systems and Mathematical Biology, SUPA, University of Aberdeen, AB24 3UE Aberdeen, United Kingdom
July 19, 2019
Abstract

Nonlinear dynamical systems, ranging from insect populations to lasers and chemical reactions, might exhibit sensitivity to small perturbations in their control parameters, resulting in uncertainties on the predictability of tunning parameters that lead these systems to either a chaotic or a periodic behavior. By quantifying such uncertainties in four different classes of nonlinear systems, we show that this sensitivity is to be expected because the boundary between the sets of parameters leading to chaos and to periodicity is fractal. Moreover, the dimension of this fractal boundary was shown to be roughly the same for these classes of systems. Using an heuristic model for the way periodic windows appear in parameter spaces, we provide an explanation for the universal character of this fractal boundary.

The topology of solutions of nonlinear dynamical systems can be severely affected by small perturbations in their control parameters Kuznetsov (2004). The so-called parameter sensitivity has been experimentally observed in dynamical and nonlinear models of systems in different areas of knowledge Costantino et al. (1997); Ren et al. (1997); Valling et al. (2007); Kolokolov and Monovskaya (2013). The cause of this sensitivity is the existence of bifurcations, such as the ones leading to crisis Grebogi et al. (1983) where chaotic attractors abruptly bifurcate into periodic ones (or vice-versa). The most profund consequence of this parameter sensitivity is to limit the ability of someone to set a parameter of a system that surely places it into either a chaotic or a periodic behavior.

In parameter spaces and bifurcation diagrams of discrete and continuous-time nonlinear dynamical systems, the set of parameters leading to chaotic behavior is intertwined with hierarchical structures of sets of parameters leading to periodic stable behavior, the complex periodic windows (CPWs) Baptista et al. (2003); Façanha et al. (2013). The structure of CPWs describes a scenario for the way that periodicity and chaos appear in a large variety of nonlinear dynamical systems: lasers Bonatto et al. (2005), electronic circuits Albuquerque and Rech (2012), population dynamics Slipantschuk et al. (2010), nonlinear oscillators Medeiros et al. (2013), etc Medeiros et al. (2010); Baptista and Caldas (1996); Gallas (2010); Bonatto and Gallas (2008a). These periodic structures were numerically shown to have self-similar-like properties, i.e., the structure of the CPWs is preserved for any scale of the parameter space Belair and Glass (1983). Moreover, the CPWs appear aligned in infinite torsion and period-adding sequences Bonatto and Gallas (2007); Medeiros et al. (2013); Bonatto and Gallas (2008a). Recently, many researchers have been carrying out additional theoretical, experimental, and numerical works to find mechanisms to explain the existence, the genesis, and the organization of CPWs. Important results were found in two-dimensional parameter spaces of dynamical systems for which the Shilnikov theorem Gaspard and Nicolis (1983); Gaspard et al. (1984) can be applied, and homoclinic orbits converge to saddle-focus equilibrium points. It was found that CPWs are connected to each other forming spiral-like structures emerging from the homoclinic bifurcation points, i.e, the parameters corresponding to the saddle-focus for which homoclinic orbits converge. Moreover, sets of homoclinic bifurcation points are aligned forming homoclinic bifurcation curves. From each point in these curves an entire spiral-like structures emerge Medrano-T and Caldas (2010); Vitolo et al. (2011); Barrio et al. (2011, 2012); Bonatto and Gallas (2008b); Albuquerque et al. (2008). This configuration is suggesting that these spiral-like structures are accumulating in a fractal way in two-dimensional parameter spaces. These spiral-like structures of CPWs have also been observed in real-world experiments Maranhão et al. (2008); Stoop et al. (2010); Cabeza et al. (2013). Additionally, in Ref. Hunt and Ott (1997) it has been argued that the width (Lebesgue measure) of the CPWs decreases exponentially with the period of the attractor and the topological entropy of the surrounding chaotic region [See supplementary material].

Parameter sets corresponding to CPWs not only appear in all scales of parameter spaces (self-similar), but they also have positive Lebesgue measure. Self-similar sets with non-zero Lebesgue measure are called fat Cantor sets. These sets are topologically equivalent to the usual Cantor set, but their properties are different, specially, their capacity dimension. In the case of fat Cantor sets, the capacity dimension is equal to the dimension of the embedding euclidean space Eykholt and Umberger (1986); Farmer (1985). Therefore, the dimension of CPWs in two-dimensional parameter spaces would be . CPWs form fat Cantor sets. Consequently, the likelihood of CPWs being experimentally found in any scale of parameter spaces is high. In fact, both periodicity and chaos in the asymptotic limit are likely to be found by either making a controlled tunning of the parameters or by taking a random sample of parameters. However, if CPWs are self-similar, it could lead to uncertainties for the predictability of tunning the parameters to produce either periodic or chaotic behaviors, since that, even though possessing integer dimensions, self-similar sets can have fractal boundaries. Consequently, if CPWs of a nonlinear dynamical system have fractal boundaries, predictability in the setting of parameters that would surely take the system to either a periodic or a chaotic behavior could be severely compromised.

Even though chaotic regions appearing in many bifurcation diagrams of one-dimensional systems Jakobson (1981) are known to be self-similar fat cantor sets, the kind of self-similarity present in CPWs appearing in two-dimensional parameter spaces was so far an open problem. Visual inspections of sucessive enlargements of parameter spaces regions Gallas (1993); Lorenz (2008) have suggested that CPWs are self-similar. Recently new evidences are pointing out that CPWs are self-similarly organized, occurring for parameters in the center of spiral-like structures. Gaspard and Nicolis (1983); Gaspard et al. (1984); Medrano-T and Caldas (2010); Vitolo et al. (2011); Barrio et al. (2011, 2012).

In this work, we indeed show that CPWs are self-similar. Self-similarity appears not only in the parameter widths of the CPWs but also in the boundaries between them and the chaotic regions. We show that the functionality of the self-similarity of CPWs regarding their widths can be both, power-law or even exponential as proposed in Hunt and Ott (1997) and observed in Viana et al. (2010). We numerically estimate the capacity dimension of the boundaries between parameters corresponding to CPWs and parameters leading to chaos, showing that they are usual skinnies fractals possessing a non-integer exterior capacity dimension. Consequently, in any scale in two-dimensional parameter spaces of a large class of nonlinear systems, there are always uncertainties associated with the predictability for tunning the parameters that surely lead the system to either a periodic or a chaotic behavior. In our simulations, the capacity dimension of these boundaries seem to be universal for different classes of nonlinear dynamical systems. We then developed an heuristic model for the appearance of CPWs and showed that the self-similarity of the widths of the CPWs appearing in this model produce fractal boundaries. This was quantified by the relation between the capacity dimension of the parameter boundaries and the decreasing rate of CPWs widths along accumulating sequences. The capacity dimension of the boundaries predicted by the model agrees with the values obtained in our simulations. So, the universal character of these boundaries is attributed to the way CPWs appear. They appear in sequences organized by their ”order” and have parameter widths that decay as a power-law with their order Not ().

To calculate the dimension of the boundary of two sets, a special capacity dimension has been defined. Defining the set as a boundary between two regions, considering () a new set formed by all points within a distance from , then, defining , the exterior capacity dimension Grebogi et al. (1985) was defined by:

(1)

where is the volume of the set . This operation is however difficult to be calculated directly. An alternative way is done by considering the uncertain exponent.

Along a direction transversal to , we take three parameter values -distant to each other. Count the number of uncertain triplets, i.e., we take three neighboring parameter values and check whether these parameters do not lead to an unique type of behavior (chaos or periodicity). From the uncertain parameters one can calculate the uncertain fraction of parameters Grebogi et al. (1985). It has been verified for certain one-dimensional quadratic maps that the uncertain fraction varies as a power-law with , i.e., , where the factor was believed to be dependent of the parameter range considered, and the exponent has been called uncertainty exponent, and believed to be independent of the considered parameter interval Grebogi et al. (1985).

The uncertainty exponent can be directly related to the exterior dimension of a set McDonald et al. (1984). It has been heuristically demonstrated by considering as the minimum number of -dimensional cubes of side required to cover the boundary of the set McDonald et al. (1984). It is well-known that scales with the cube side as . Setting to be equal to the parameter error , the uncertain region will be of the order of the total volume of all -dimensional cubes required to cover the boundary. The volume of one uncertain cube is given by , so the total uncertain volume is given by . Assuming that the uncertain fraction is proportional to the uncertain volume, then:

(2)

Our numerical results are based on simulations of four different classes of dynamical systems. We consider parameter spaces regions for which the parameters corresponding to CPWs are hierarchically distributed giving self-similar features to the region.

We consider the Rössler oscillator for which the Shilnikov theorem can be applied. This system is described by the following set of nonlinear differential equations:

(3)

We investigate an extension of the parameter plane , where the complex periodic structures emerge from homoclinic orbits and are spiral-shaped organized in sequences. The other parameter of Eq. (3) is fixed at Gallas (2010).

The class of nonlinear forced oscillators are represented by the Morse oscillator which is governed by the following nonlinear differential equation Scheffczyk et al. (1991):

(4)

in the parameter plane , the CPWs are aligned in sequences of period and torsion-adding.

We also work with a loss-modulated CO laser described by a rate-equation with a time-dependent parameter:

(5)

where . We investigate a complex periodic sequence in the parameter plane. All other parameters are fixed: , , , and Bonatto et al. (2005).

Finally, we consider a sequence of CPWs in the parameter spaces of the well-known circle map described by the discrete-time equations Jensen et al. (1983):

(6)

In the two-dimensional parameter spaces of those systems, we select pairs (, ) of random parameters uniformly space distributed and compute the largest Lyapunov exponents to determine if the correspondent state is periodic () or chaotic (). Then, each pair of parameters is perturbed by an error in both orientations along one parameter. This process generates pairs (, ) of parameters. We also obtain the Lyapunov exponent of states corresponding to the perturbed parameter pairs. We compare only the unperturbed chaotic parameters (that produces chaotic attractors, i.e., ) to their two correspondent perturbed pairs along the horizontal direction. If at least one of them is not chaotic, the pair (, ) is counted as an uncertain pair for the error value . We record the uncertain fraction for an error interval.

In Figure 1(Left), we show the parameter spaces for the four systems considered, for which and consequently are calculated. The black regions indicate the set of parameters leading to chaos, while the white regions correspond to parameters leading to periodic stable behavior. In these figures, CPWs are aligned along sequences accumulating in periodic regions of parameter spaces. Here, a specific sequence of CPWs is ordered by its characteristic period-adding rule. The CPWs where attractors have theirs periods added along the sequence are identified by their order . The window with the largest width of a sequence has the lowest order, . For dynamical systems where torsion and rotation numbers are defined, sequences can also be identified by those parameters where frequency locking occurs Medeiros et al. (2013). There exists infinite sequences with different period-adding rules and CPWs sizes. For this work, we measure the width of periodic windows beloging only to the main sequences (larger width) identified by the filled circles in Fig. 1(Left). In Figure 1(Center), for the correspondent parameter space shown in Fig. 1(Left), we show the fraction of uncertain periodic parameters as a function of the error . The straight line is a power-law fitting between and which provides the uncertainty exponent . We observe that the exponent is in the same confidence interval given by for the different classes of dynamical systems considered here. The standard deviation of has been obtained by considering that the occurrence of uncertain parameters are random events. From in Eq. (2) we obtain that the dimension of the boundary between the chaotic and periodic parameter sets is given by . In Figure 1(Right), we show the width of the first CPWs as a function of the order it appears along the main accumulation sequence. We fit a power-law of the form to the way that CPW widths decrease Not ().

Figure 1: (Left) Two-dimensional parameter spaces of the four considered dynamical systems. Black regions represent the chaotic parameter set. White regions represent the periodic parameter set. (Center) The uncertain fraction of the chaotic sets shown in (Left) scales as power law with the error . (Right) The width of the periodic windows shown in (Left) versus their order . The window width is obtained by taking an one-dimensional cut of the two-dimensional parameters space and measuring the distance from the initial saddle-node bifurcation to the final crisis of a peridic windows Yorke et al. (1985).

The measurements of Fig. 1 indicate that the exterior capacity dimension of the boundaries between parameters corresponding to CPWs and parameters leading to chaos is universal for different classes of dynamical systems. To understand why that would be so, we formulate an heuristic model for the appearance of CPWs and chaotic regions where the observed decreasing of the CPWs width is considered. In our model, CPWs are created by removing pieces of a one-dimensional cut of the two-dimensional parameters space, Fig. . The gaps represent a CPW and the remaining intervals represent chaotic regions. One begins with a chaotic interval of length and removes the amount correspondent to the width of a CPW of first order, , leaving a chaotic interval of length . Next, one removes from the amount correspondent to the width of a CPW of second order, , leaving a chaotic interval of length . Continuing in this manner, ad infinitum, we will obtain a set of elements corresponding to the chaotic parameters from which the intervals corresponding to CPWs of one sequence have been removed. To decide if this asymptotic set is a fat fractal, we obtain its uncertainty exponent , the set will be a fat fractal for Eykholt and Umberger (1986); Farmer (1985). Moreover, for one-dimensional quadratic maps, the parameters corresponding to chaotic behavior have been demonstrated to have nonzero Lebesgue measure (fat fractal) Jakobson (1981).

Figure 2: (a) A two-dimensional parameter space, the white regions represent parameters corresponding to CPWs. (b) Schematic of the formation of CPWs that have fractal boundaries. For and , a periodic window is created by the removal of of the initial line. For , a periodic window is created by removal of the line, and so on. The remaining pieces of the line represent chaotic regions. Secondary accumulation of CPWs can be created by removing two intervals at each iteration. is the number of uncertain points of each iteration.

In order to obtain the uncertainty exponent of the heuristic model, we count the number of uncertain parameters as the CPWs width decreases. The number of uncertain parameters of each iteration is given by , a number representing the numbers of boundaries between the remaining pieces and the gaps. The width of the CPW decreases at each iteration by , Fig. 2. The uncertain fraction of parameter is written as function of the error , i.e., . So, the number of uncertain parameters of the model can be written as . Considering that the model error is inversely proportional to the CPW width, , using that , we derive the uncertainty exponent, , of the heuristic model:

(7)

In this heuristic model we consider only the main sequence of peridic windows, numbered in Fig. (1). We believe that the uncertain exponent, , is independent of the number of sequences considered, once that, for a larger number of sequences the fraction of uncertain elements, , per iteration is higher, but the error, , will be lower, so the limit of Eq. (7) must be the same despite of the number of sequences considered by the model. In fact, we obtain that (see Table 1), an evidence that smaller sequences do not contribute much for . Using Eq. (2) in Eq. (7), we obtain the exterior capacity dimension of the model:

(8)

where is the exponent of the width decreasing and is the dimension of the embedding euclidean space (for two-dimensional parameters space, ). Substituting the values of , obtained in Fig. 1, in Eq. (8), we obtain the exterior capacity dimension expected for the boundary between CPWs and chaos according to the heuristic model. The standard deviations of , and are obtained by propagating the uncertainties of the window width measured in bifurcations diagrams.

In Table 1, for all dynamical systems considered here, we compare the measurements from the simulations with the results provided by the model when the measured exponents for the decreasing of CPWs are given. We verify that the exterior capacity dimension provided by the model agrees with the exterior capacity dimension obtained in our simulations shown in Fig. 1.

Dynamical System
Rössler Oscillator
Morse Oscillator
CO Laser
Circle Map
Table 1: In this table, we show in the first column the value of the exponent obtained from the fitting in Fig. 1(right), in the second column the values of calculated from the data in Fig. 1(center), and in the third column, the values of using Eq. (8). In fourth column, the values of and fifth column the values of .

In conclusion, we have shown that the boundaries between parameters corresponding to CPWs and parameters leading to chaos in two-dimensional parameter spaces are fractals. Consequently, the ability of tunning a parameter of a nonlinear dynamical system to set its behavior to be surely either chaotic or periodic is seriously compromised by this fine structure of the boundary in all scales of the parameter space. In our simulations, we found that the capacity dimension of such boundaries seems to be universal for different classes of dynamical systems treated in this work. From an heuristic model for the appearance of CPWs, we deduce a relation between the exterior capacity dimension and the exponent of decreasing of the CPWs widths along sequences. The dimension deduced from the model agrees with values observed in our simulations. This fact strongly suggests that the universality observed for this boundary between chaos and periodicity is a consequence of the power-law fashion with which periodic windows decrease their sizes as a function of their order. We also remark that the decreasing of the width of periodic structures immersed in parameters corresponding to quasi-periodic behavior, called Arnold tongues Glass and Perez (1982), has been observed to have a power-law dependence on its period Ecke et al. (1989). These facts give support to our main claim that the decreasing of the window width of the CPWs in period-adding sequences can be described by a power-law function. However, CPWs can also have their sizes exponentially decreasing with their order, when the accumulation of CPWs forms the spiral-like structures due to the homoclinic bifurcation scenario (see supplementary material). These spiral-like structures are however not predominant all over the domain of the parameter space, but coexist with accumulation sequences where CPWs have their widths decreasing in a power-law fashion.

We would like to thank the partial support of this work by the Brazilian agencies FAPESP (process: 2011/19296-1), CNPq and CAPES.

References

  • Kuznetsov (2004) Y. A. Kuznetsov, Elements of applied bifurcation theory (Springer, United States of America, 2004).
  • Costantino et al. (1997) R. F. Costantino, R. A. Desharnais, J. M. Cushing,  and B. Dennis, Science 275, 389 (1997).
  • Ren et al. (1997) W. Ren, S. J. Hu, B. J. Zhang, F. Z. Wang, Y. F. Gong,  and J. X. Xu, Int. J. of Bif. and Chaos 7, 1867 (1997).
  • Valling et al. (2007) S. Valling, B. Krauskopf, T. Fordell,  and A. M. Lindberg, Opt. Commun. 271, 532 (2007).
  • Kolokolov and Monovskaya (2013) Y. Kolokolov and A. Monovskaya, Int. J. Bif. and Chaos 23, 1350063 (2013).
  • Grebogi et al. (1983) C. Grebogi, E. Ott,  and J. A. Yorke, Physica D 7, 181 (1983).
  • Baptista et al. (2003) M. S. Baptista, C. Grebogi,  and E. Barreto, Int. J. of Bif. and Chaos 13, 2681 (2003).
  • Façanha et al. (2013) W. Façanha, B. Oldeman,  and L. Glass, Phys. Lett. A 377, 1264 (2013).
  • Bonatto et al. (2005) C. Bonatto, J. C. Garreau,  and J. A. C. Gallas, Phys. Rev. Lett. 95, 143905 (2005).
  • Albuquerque and Rech (2012) H. A. Albuquerque and P. C. Rech, Int. J. Circ. Theor. Appl. 40, 189 (2012).
  • Slipantschuk et al. (2010) J. Slipantschuk, E. Ullner, M. S. Baptista, M. Zeineddine,  and M. Thiel, Chaos 20, 045117 (2010).
  • Medeiros et al. (2010) E. S. Medeiros, S. L. T. de Souza, R. O. Medrano-T,  and I. L. Caldas, Phys. Lett. A 374, 2628 (2010).
  • Baptista and Caldas (1996) M. S. Baptista and I. L. Caldas, Chaos, Solitons and Fractals 7, 325 (1996).
  • Gallas (2010) J. A. C. Gallas, Int. J. Bif. and Chaos 20, 197 (2010).
  • Bonatto and Gallas (2008a) C. Bonatto and J. A. C. Gallas, Phil. Trans. R. Soc. A 366, 505 (2008a).
  • Belair and Glass (1983) J. Belair and L. Glass, Phys. Lett. A 96, 113 (1983).
  • Bonatto and Gallas (2007) C. Bonatto and J. A. C. Gallas, Phys. Rev. E 75, 055204(R) (2007).
  • Medeiros et al. (2013) E. S. Medeiros, R. O. Medrano-T, I. L. Caldas,  and S. L. T. de Souza, Phys. Lett. A 377, 628 (2013).
  • Gaspard and Nicolis (1983) P. Gaspard and G. Nicolis, J. Stat. Phys. 31, 499 (1983).
  • Gaspard et al. (1984) P. Gaspard, R. Kapral,  and G. Nicolis, J. Stat. Phys. 35, 697 (1984).
  • Medrano-T and Caldas (2010) R. O. Medrano-T and I. L. Caldas, ArXiv e-prints  (2010), 1012.2241 .
  • Vitolo et al. (2011) R. Vitolo, P. Glendinning,  and J. A. C. Gallas, Phys. Rev. E 84, 016216 (2011).
  • Barrio et al. (2011) R. Barrio, F. Blesa, S. Serrano,  and A. Shilnikov, Phys. Rev. E 84, 035201(R) (2011).
  • Barrio et al. (2012) R. Barrio, F. Blesa,  and S. Serrano, Phys. Rev. Lett. 108, 214102 (2012).
  • Bonatto and Gallas (2008b) C. Bonatto and J. A. C. Gallas, Phys. Rev. Lett. 101, 0541011 (2008b).
  • Albuquerque et al. (2008) H. A. Albuquerque, R. M. Rubinger,  and P. C. Rech, Phys. Lett. A 372, 4793 (2008).
  • Maranhão et al. (2008) D. M. Maranhão, M. S. Baptista, J. C. Sartoreli,  and I. L. Caldas, Phys. Rev. E 77, 037202 (2008).
  • Stoop et al. (2010) R. Stoop, P. Benner,  and Y. Uwate, Phys. Rev. Lett. 105, 074102 (2010).
  • Cabeza et al. (2013) C. Cabeza, C. A. Briozzo, R. Garcia, J. G. Freire, A. C. Marti,  and J. A. Gallas, Chaos, Solitons and Fractals 52, 59 (2013).
  • Hunt and Ott (1997) B. Hunt and E. Ott, J. Phys. A: Math. Gen. 30, 7067 (1997).
  • Eykholt and Umberger (1986) R. Eykholt and D. K. Umberger, Phys. Rev. Lett. 57, 2333 (1986).
  • Farmer (1985) J. D. Farmer, Phys. Rev. lett. 55, 351 (1985).
  • Jakobson (1981) M. V. Jakobson, Commun. Math. Phys. 81, 39 (1981).
  • Gallas (1993) J. A. C. Gallas, Phys. Rev. Lett. 70, 2714 (1993).
  • Lorenz (2008) E. N. Lorenz, Physica D 13, 1689 (2008).
  • Viana et al. (2010) E. R. Viana, R. M. Rubinger, H. A. Albuquerque, A. G. de Oliveira,  and G. M. Ribeiro, Chaos 20, 0231101 (2010).
  • (37) The window width has been proposed to be dependent of the orbital period and of the topological entropy of the nearby chaotic attractors as an exponential law for quadratic maps Hunt and Ott (1997), we remark that this previous result is for periodic windows independent of any choice of accumulation sequences. In addition, notice that change along the sequence. The functional form of as function of the parameters along the accumulation sequence is unknown. Therefore, our measurements do not exclude its validity for the dynamical systems considered here.
  • Grebogi et al. (1985) C. Grebogi, S. W. McDonald, E. Ott,  and J. A. Yorke, Phys. Lett. A 110, 01 (1985).
  • McDonald et al. (1984) S. W. McDonald, C. Grebogi, E. Ott,  and J. A. Yorke, Physica D 17, 125 (1984).
  • Scheffczyk et al. (1991) C. Scheffczyk, U. Parlitz, T. Kurz, W. Konp,  and W. Lauterborn, Phys. Rev. A 43, 6495 (1991).
  • Jensen et al. (1983) M. H. Jensen, P. Bak,  and T. Bohr, Phys. Rev. Lett. 50, 1637 (1983).
  • Yorke et al. (1985) J. A. Yorke, C. Grebogi, E. Ott,  and L. Tedeschini-Lalli, Phys. Rev. Lett. 54, 1095 (1985).
  • Glass and Perez (1982) L. Glass and R. Perez, Phys. Rev. Lett. 48, 1772 (1982).
  • Ecke et al. (1989) R. E. Ecke, J. D. Farmer,  and D. K. Umberger, Nonlinearity 2, 175 (1989).
Comments 0
Request Comment
You are adding the first comment!
How to quickly get a good reply:
  • Give credit where it’s due by listing out the positive aspects of a paper before getting into which changes should be made.
  • Be specific in your critique, and provide supporting evidence with appropriate references to substantiate general statements.
  • Your comment should inspire ideas to flow and help the author improves the paper.

The better we are at sharing our knowledge with each other, the faster we move forward.
""
The feedback must be of minimum 40 characters and the title a minimum of 5 characters
   
Add comment
Cancel
Loading ...
321902
This is a comment super asjknd jkasnjk adsnkj
Upvote
Downvote
""
The feedback must be of minumum 40 characters
The feedback must be of minumum 40 characters
Submit
Cancel

You are asking your first question!
How to quickly get a good answer:
  • Keep your question short and to the point
  • Check for grammar or spelling errors.
  • Phrase it like a question
Test
Test description