Filename
stringlengths 22
64
| Paragraph
stringlengths 8
5.57k
|
|---|---|
Processed_A_Game-Theoretic_Analysis_of_the_Off-Switch_Game.txt | This work grew out of a MIRIx workshop, with Owen Cameron, John Aslanides, Huon Puertas also attending. Thanks to Amy Zhang for proof reading multiple drafts. This work was in part supported by ARC grant DP150104590. |
Processed_A_Game-Theoretic_Analysis_of_the_Off-Switch_Game.txt | Armstrong, S. (2015). Motivated Value Selection for Artificial Agents. Twenty-Ninth AAAI Conference on Artificial Intelligence, pages 12–20. |
Processed_A_Game-Theoretic_Analysis_of_the_Off-Switch_Game.txt | Everitt, T., Filan, D., Daswani, M., and Hutter, M. (2016). Self-modificication of Policy and Utility Function in Rational Agents. In Artificial General Intelligence, pages 1–11. Springer. |
Processed_A_Game-Theoretic_Analysis_of_the_Off-Switch_Game.txt | Soares, N. and Fallenstein, B. (2014). Aligning Superintelligence with Human Interests: A Tech- nical Research Agenda. Technical report, Machine Intelligence Research Institute (MIRI). |
Processed_Core_properties_of_alpha_Cen_A_using_asteroseismol.txt | Context. A set of long and nearly continuous observations of α Centauri A should allow us to derive an accurate set of asteroseismic constraints to compare to models, and make inferences on the internal structure of our closest stellar neighbour. Aims. We intend to improve the knowledge of the interior of α Centauri A by determining the nature of its core. Methods. We combined the radial velocity time series obtained in May 2001 with three spectrographs in Chile and Australia: CORALIE, UVES, and UCLES. The resulting combined time series has a length of 12.45 days and contains over 10,000 data points and allows to greatly reduce the daily alias peaks in the power spectral window. Results. We detected 44 frequencies that are in good overall agreement with previous studies, and found that 14 of these show possi- ble rotational splittings. New values for the large (∆ν) and small separations (δν02, δν13) have been derived. Conclusions. A comparison with stellar models indicates that the asteroseismic constraints determined in this study (namely r10 and δν13) allows us to set an upper limit to the amount of convective-core overshooting needed to model stars of mass and metallicity similar to those of α Cen A. |
Processed_Core_properties_of_alpha_Cen_A_using_asteroseismol.txt | During the last decade, the visual binary stellar system α Centauri turned out to be a very interesting asteroseismic tar- get to observe and model because of its proximity and of the similarity of these stars to the Sun. Moreover, the mass of the primary component is very close to the limit above which stellar models predict the onset of convection in the energy-generating core. This makes the theoretical study of α Cen A particularly valuable for testing the poorly-modelled treatment of convection and extra-mixing in the central regions of low-mass stars. |
Processed_Core_properties_of_alpha_Cen_A_using_asteroseismol.txt | nation of the surface gravities: log gA = 4.307 ± 0.005 and log gB = 4.538 ± 0.008, an accuracy rarely reached in stars other than the Sun. Moreover, thees brightness of both components of the system (VA = −0.01 and VB = 1.33) allows the acquisition of extremely high-quality spectra. |
Processed_Core_properties_of_alpha_Cen_A_using_asteroseismol.txt | Asteroseismic data have been obtained by several teams. The first unambiguous detection of p-modes in α Cen A was made by Bouchy & Carrier (2001, 2002) during a 13-night campaign with the spectrograph CORALIE, confirming the earlier claimed de- tection made by Schou & Buzasi (2000) with the WIRE satellite. Bouchy & Carrier (2002) detected 28 modes with angular de- grees ℓ = 0, 1 and 2. At the same time, another team (Butler et al. 2004; Bedding et al. 2004) made observations during five nights from Chile with UVES and from Australia with UCLES, and detected 42 frequencies with ℓ = 0, 1, 2 and 3. Kjeldsen et al. (2005) also provided a value of the mode lifetime for α Cen A of 2.3+1.0 −0.4 days. Fletcher et al. (2006) then carried out a re−analysis of the WIRE observations, resulting in additional asteroseismic data. In particular, they suggested two values of the rotational frequency, 0.54 ± 0.22µHz and 0.64 ± 0.25 µHz, using two differ- ent analysis methods, and a mode lifetime of 3.9±1.4 days. More recently, Bazot et al. (2007) detected 34 modes with the HARPS spectrograph in Chile, and suggested five rotational splittings for ℓ = 2 modes. Asteroseismic data have also been obtained for the B component (Carrier & Bourban 2003; Kjeldsen et al. 2005). |
Processed_Core_properties_of_alpha_Cen_A_using_asteroseismol.txt | Fig. 1. Combined time series of CORALIE, UVES and UCLES. One can see that the UCLES data, which were taken in Australia, fill several gaps in the time series of the CORALIE and UVES data, taken in Chile. This will allow a better detection of p-modes frequencies by reducing the daily aliases in the spectrum of the star (see text for details). |
Processed_Core_properties_of_alpha_Cen_A_using_asteroseismol.txt | Fig. 2. Comparison of the spectral windows of the time series of the CORALIE data (top panel) and the one of the combined time series of the CORALIE, UVES and UCLES data with standard weights (middle panel) and sidelobe-optimised weights (bot- tom panel). The daily aliases are already much reduced in the case of the combined time series with standard weights, as ex- pected from the fact that UCLES nights fill several gaps in the CORALIE and UVES time series. |
Processed_Core_properties_of_alpha_Cen_A_using_asteroseismol.txt | The α Cen system has been also extensively mod- elled. Guenther & Demarque (2000), Morel et al. (2000) and Noels et al. (1991) performed a calibration based only on non−asteroseismic constraints. They found that two kinds of models, one with a radiative core and the other with a convec- tive one, could satisfy these constraints. Subsequently, the high- quality non−asteroseismic constraints together with asteroseis- mic constraints stimulated new calibrations of the stellar sys- tem (Th´evenin et al. 2002; Thoul et al. 2003; Eggenberger et al. 2004; Miglio & Montalb´an 2005). Although these authors suc- ceeded to fit some of the asteroseismic constraints, they were not in a position to draw a definite conclusion on the convective ver- sus radiative nature of the core of α Centauri A. To perform such a discrimination, all teams suggested that more accurate astero- seismic constraints were needed. |
Processed_Core_properties_of_alpha_Cen_A_using_asteroseismol.txt | In the present paper, the May 2001 data of Bouchy & Carrier (2002) from CORALIE and of Bedding et al. (2004) from UVES and UCLES have been unified to compute a combined velocity time series in order to reduce the effect of daily aliases in the power spectrum. The objective is to provide a more accurate set of asteroseismic constraints that allows a better comparison with theoretical models and a clear discrimination between them. |
Processed_Core_properties_of_alpha_Cen_A_using_asteroseismol.txt | Section 2 describes the different data sets and their main features, along with the method used to analyse the acoustic spectrum. Section 3 presents the frequencies and amplitudes de- tected, along with new values for the large and small separations and a rotational frequency. In the section 4, we compare these new observational constraints with the models of α Cen A pre- sented by Miglio & Montalb´an (2005). Section 5 is dedicated to the conclusions. |
Processed_The_faint_source_population_at_15.7_GHz_-_III._A_h.txt | ABSTRACT A complete sample of 96 faint (S > 0.5 mJy) radio galaxies is selected from the Tenth Cam- bridge (10C) survey at 15.7 GHz. Optical spectra are used to classify 17 of the sources as high-excitation or low-excitation radio galaxies (HERGs and LERGs respectively), for the remaining sources three other methods are used; these are optical compactness, X-ray obser- vations and mid-infrared colour–colour diagrams. 32 sources are HERGs and 35 are LERGs while the remaining 29 sources could not be classified. We find that the 10C HERGs tend to have higher 15.7-GHz flux densities, flatter spectra, smaller linear sizes and be found at higher redshifts than the LERGs. This suggests that the 10C HERGs are more core dominated than the LERGs. |
Processed_The_faint_source_population_at_15.7_GHz_-_III._A_h.txt | Lower-frequency radio images, linear sizes and spectral indices are used to classify the sources according to their radio morphology; 18 are Fanaroff and Riley type I or II sources, a further 13 show some extended emission, and the remaining 65 sources are compact and are referred to as FR0 sources. The FR0 sources are sub-divided into compact, steep-spectrum (CSS) sources (13 sources) or GHz-peaked spectrum (GPS) sources (10 sources) with the remaining 42 in an unclassified class. FR0 sources are more dominant in the subset of sources with 15.7-GHz flux densities <1 mJy, consistent with the previous result that the fainter 10C sources have flatter radio spectra. |
Processed_The_faint_source_population_at_15.7_GHz_-_III._A_h.txt | The properties of the 10C sources are compared to the higher-flux density Australia Tele- scope 20 GHz (AT20G) survey. The 10C sources are found at similar redshifts to the AT20G sources but have lower luminosities. The nature of the high-frequency selected objects change as flux density decreases; at high flux densities the objects are primarily quasars, while at low flux densities radio galaxies dominate. |
Processed_The_faint_source_population_at_15.7_GHz_-_III._A_h.txt | In the first two papers in this series (Whittam et al. 2013, 2015, referred to as Papers I and II respectively) we studied the properties of a sample of sources selected from the Tenth Cambridge (10C; AMI Consortium: Davies et al. 2011; AMI Consortium: Franzen et al. 2011) survey at 15.7 GHz in the Lockman Hole. The 10C survey is complete to 0.5 mJy, making it the deepest high-frequency radio survey to date. In Paper II we found that the vast majority ((cid:62) 94 per cent) of the sources in the 10C sample are radio galaxies; the 10C sample is therefore the ideal starting point for a study of the properties of faint radio galaxies. |
Processed_The_faint_source_population_at_15.7_GHz_-_III._A_h.txt | It has been known for some time (Hine & Longair 1979) that the properties of radio galaxies are not fully explained by the con- ventional model of an AGN, consisting of an accretion disk sur- rounded by a dusty torus (Antonucci 1993). Based on this model, we would expect radio-loud objects viewed close to the jet axis to show both broad and narrow optical emission lines, while radio- loud objects viewed at larger angles to the jet would only show nar- row lines and would have a clear mid-infrared signature of the dusty torus. However, many radio-loud AGN lack the expected narrow- line optical emission and do not display evidence of an obscuring torus (e.g. Whysong & Antonucci 2004). |
Processed_The_faint_source_population_at_15.7_GHz_-_III._A_h.txt | Subsequent studies have suggested that there are two funda- mentally distinct accretion modes, known as ‘cold mode’ and ‘hot mode’ (see Best et al. 2005a; Hardcastle et al. 2007) which could be responsible for these differences (these modes are sometimes referred to as ‘quasar’ and ‘radio’ modes respectively). Cold-mode accretion occurs when cold gas is accreted onto the central black hole through a radiatively efficient, geometrically thin, optically thick accretion disk (e.g. Shakura & Sunyaev 1973) and gives rise to the traditional model of an AGN (Antonucci 1993). These ob- jects therefore show high-excitation lines in their optical spectra, so are often referred to as high-excitation radio galaxies (HERGs). ‘Hot mode’ sources, however, are thought to be fuelled by the ac- cretion of warm gas through advection-dominated accretion flows (e.g. Narayan & Yi 1995) and lack many of the typical signatures of AGN, such as strong optical emission lines. These objects are often referred to as low-excitation radio galaxies (LERGs). LERGs typically show no evidence for a dusty torus (Ogle et al. 2006) or for accretion-related X-ray emission (Hardcastle et al. 2006). There are also differences in the host galaxies of the two populations, with HERGs found in less massive and bluer galaxies than LERGs (Best et al. 2005b; Tasse et al. 2008; Herbert et al. 2010, 2011; Willi- ams & Röttgering 2015). Although both HERGs and LERGs are found across the full range in radio luminosities, LERGs seem to dominate at lower luminosities and HERGs at higher luminosities (Best & Heckman 2012). There is a substantial overlap between the FRI/II classification and the HERG and LERG classes, with most FRI sources being LERGs and most FRIIs being HERGs (Best & Heckman 2012). There are, however, differences in the two classi- fications, in particular a significant population of FRII LERGs has been found (Laing et al. 1994). |
Processed_The_faint_source_population_at_15.7_GHz_-_III._A_h.txt | Best & Heckman (2012) showed that low-redshift HERGs and LERGs have distinct accretion rates; HERGs typically accrete at between 1 and 10 per cent of their Eddington rate, while LERGs generally have accretion rates much less than 1 per cent of their Eddington rate. Fernandes et al. (2015) also find evidence for this at z ∼ 1 using mid-infrared data. Therefore a picture is emerging where HERGs accrete cold gas at a relatively high rate (Best & Heckman 2012) and radiate efficiently across the whole electro- magnetic spectrum (e.g. Elvis et al. 1994). This causes them to produce a stable accretion disc and therefore display the typical properties of an AGN. The cold gas leads to star formation (Virdee et al. 2013; Hardcastle et al. 2013), causing the host galaxies of HERGs to be relatively blue (e.g. Kauffmann et al. 2003). In ad- dition, HERGs are more prevalent at earlier cosmic epochs, where higher rates of mergers and interactions provided a steady supply of cold gas. |
Processed_The_faint_source_population_at_15.7_GHz_-_III._A_h.txt | The fundamentally different accretion modes of HERGs and LERGs cause them to have different radio properties. Mahony et al. (2011) studied the properties of high flux density (S 20 GHz > 40 mJy) HERGs and LERGs in the Australia Telescope 20 GHz (AT20G; Murphy et al. 2010) sample and found that while both ac- cretion modes display a range of radio properties, a higher fraction of HERGs are extended and have steep spectra. They suggest that this is because HERGs are accreting more efficiently than LERGs, and therefore have a greater chance of producing more luminous jets and lobes. They also find that HERGs display different prop- erties depending on their orientation, with objects displaying broad emission lines tending to be flat spectrum, and objects with narrow lines tending to be steep spectrum, as predicted by orientation mod- els (Antonucci 1993; Urry & Padovani 1995). LERGs, however, display no orientation effects. These results are consistent with the scenario in which HERGs have a typical AGN accretion disk and torus while LERGs do not. |
Processed_The_faint_source_population_at_15.7_GHz_-_III._A_h.txt | Most studies of HERGs and LERGs have focussed on high flux density (e.g. Mahony et al. 2011) or low redshift (e.g. Best & Heckman 2012) sources, with the exception of Fernandes et al. (2015), who studied a small sample at z ∼ 1. In this work, we focus on a sample of high-frequency selected mJy and sub-mJy sources. In Section 2 we outline the data used in this study. In Section 3 we describe how this data is used to separate the radio galaxies into HERGs and LERGs and the properties of these HERGs and LERGs are explored in Section 4. In Section 5 the sample is split into differ- ent radio morphological classes and the properties of these classes are discussed and compared to the HERG and LERGs classifica- tions. The sample is compared to the higher-flux density AT20G sample in Section 6 and the conclusions are presented in Section 7. |
Processed_The_faint_source_population_at_15.7_GHz_-_III._A_h.txt | The 10C survey was observed with the Arcminute Microkelvin Im- ager (AMI; Zwart et al. 2008) and covered a total of 27 deg2 com- plete to 1 mJy across ten different fields. A further 12 deg2, con- tained within these fields, is complete to 0.5 mJy. In Paper I we se- lected a sample of 296 sources from two fields of the 10C survey in the Lockman Hole. This sample was matched to a number of lower- frequency (and higher resolution) catalogues available in the field; a deep 610 MHz Giant Meterwave Radio Telescope (GMRT) im- age (Garn et al. 2008, 2010), a 1.4 GHz Westerbork Synthesis Ra- dio Telescope (WSRT) image (Guglielmino et al. 2012), two deep Very Large Array (VLA) images at 1.4 GHz which only cover part of the field (Biggs & Ivison 2006; Owen & Morrison 2008), the National Radio Astronomy Observatory (NRAO) VLA Sky Survey (NVSS; Condon et al. 1998) and the Faint Images of the Radio Sky at Twenty centimetres (FIRST; White et al. 1997). For full details of the catalogues used and the matching procedure see Paper I. These data enabled spectral indices and angular sizes to be found for the 10C sources. |
Processed_The_faint_source_population_at_15.7_GHz_-_III._A_h.txt | served by Chandra (Wilkes et al. 2009) and one by XMM-Newton (Brunner et al. 2008). Fig. 1 illustrates the positions of these two surveys. The Chandra Lockman Hole field covers 0.7 deg2 and de- tected 775 X-ray sources to a limiting broadband (0.3 to 8 keV) flux ∼ 4 × 10−16 erg cm−2 s−1. The XMM-Newton field covers ∼0.2 deg2 and detects 409 sources with a sensitivity limit of 1.9, 9 and 180 ×10−16 erg cm−2 s−1 in the 0.5 to 2.0, 2.0 to 10.0 and 5.0 to 10.0 keV bands respectively. |
Processed_The_faint_source_population_at_15.7_GHz_-_III._A_h.txt | The two X-ray catalogues were matched to the 10C sources using a 5 arcsec match radius (a procedure similar to that described in Section 3.2 in Paper II was carried out to determine an appropri- ate match radius). For those 10C sources with a match to the Data Fusion catalogue, the SERVS position from the Data Fusion cata- logue was used. For the remaining sources the radio position was used, with the radio position chosen in the following order of pref- erence: FIRST, GMRT, BI2006/OM2008, WSRT, 10C. The results of this matching are summarised in Table 1. In total, out of 32 10C sources in the two X-ray survey areas, 15 (47 per cent) have an X-ray counterpart. |
Processed_Design_optimisation_of_a_multi-mode_wave_energy_co.txt | Figure 1: Geometry and parameters of the three-tether wave energy converter: (a) 3D view, (b) front view, (c) top view. |
Processed_Design_optimisation_of_a_multi-mode_wave_energy_co.txt | turing such a system, the buoy should be built as large as possible (20 m radius and 30 m height). |
Processed_Design_optimisation_of_a_multi-mode_wave_energy_co.txt | used as an objective function for shape optimisation of various WECs in [5, 6, 7, 8]. |
Processed_Design_optimisation_of_a_multi-mode_wave_energy_co.txt | metric to assess energy investments, its calculations for wave energy devices are full of assumptions and uncertainties. |
Processed_Design_optimisation_of_a_multi-mode_wave_energy_co.txt | been extended in [11] by including the cost of materials and manufacturing processes into the optimisation procedure. |
Processed_Design_optimisation_of_a_multi-mode_wave_energy_co.txt | The majority of geometry optimisation studies consider WECs that absorb power from one hydrodynamic mode (e.g. |
Processed_Design_optimisation_of_a_multi-mode_wave_energy_co.txt | of the buoy is equal to half the displaced mass of water mb = 0.5ρwV (ρw = 1025 kg/m3, and V = πa2H). |
Processed_Design_optimisation_of_a_multi-mode_wave_energy_co.txt | states with a total probability of 99% are chosen to represent the deployment climate (outlined by a black line). |
Processed_Design_optimisation_of_a_multi-mode_wave_energy_co.txt | Figure 2: The wave climate at the Albany deployment site located in Western Australia (117.7547◦E, 35.1081◦S, 33.9 kW/m mean annual wave power resource) [16]. |
Processed_Convergence_in_$L_p([0,T])$_of_wavelet_expansions_.txt | Will appear in Methodology and Computing in Applied Probability. The final publication is available at link.springer.com. DOI: 10.1007/s11009-013-9346-7. |
Processed_Convergence_in_$L_p([0,T])$_of_wavelet_expansions_.txt | Abstract The article presents new results on convergence in Lp([0, T ]) of wavelet expansions of ϕ-sub-Gaussian random processes. The convergence rate of the ex- pansions is obtained. Specifications of the obtained results are discussed. |
Processed_Convergence_in_$L_p([0,T])$_of_wavelet_expansions_.txt | Multiresolution analysis of deterministic signals by the use of wavelets has been extensively studied in recent years. However, in the context of stochastic processes, general wavelet approximations has not yet been fully investigated. In the majority of cases developed deterministic methods and used error measures and metrics may not be appropriate to investigate wavelet representations of stochastic processes. It indicates the necessity of elaborating special stochastic techniques. |
Processed_Convergence_in_$L_p([0,T])$_of_wavelet_expansions_.txt | From a practical point of view, multiresolution analysis provides an efficient framework for the decomposition of random processes. Wavelet representations could be used to convert the problem of analyzing a continuous-time random process to that of analyzing a random sequence, which is much simpler. This approach is widely used in statistics to estimate a curve given observations of the curve plus some noise, in time series analysis for smoothing functional data, in simulation studies of various functionals defined on realizations of a random process, etc. |
Processed_Convergence_in_$L_p([0,T])$_of_wavelet_expansions_.txt | Recently, a considerable attention was given to wavelet orthonormal series representations of stochastic processes. Numerous results, applications, and refer- ences on convergence of wavelet expansions of random processes in various spaces can be found in Atto, Berthoumieu 2010, Bardet, Tudor 2010, Cambanis, Masry 1994, Clausel et al. 2012, Didier, Pipiras 2008, Kurbanmuradov, Sabelfeld 2008, Kozachenko et al. 2011, 2013, Kozachenko, Polosmak 2008, Zhang, Waiter 1994, just to mention a few. |
Processed_Convergence_in_$L_p([0,T])$_of_wavelet_expansions_.txt | Figures 1 and 21 illustrate wavelet expansions of stochastic processes. A sim- ulated realization of the process X(t) and its two wavelet reconstructions with different numbers of terms are plotted in Figure 1. Figure 2 displays boxplots of mean-square approximation errors for 500 simulated realizations of X(t) for each reconstruction. Figure 2 suggests that empirical probabilities of large errors become smaller when the number of terms in the wavelet expansion increases. |
Processed_Convergence_in_$L_p([0,T])$_of_wavelet_expansions_.txt | Although the mentioned effect is well-known for deterministic functions, it has to be established theoretically for different stochastic processes and probability metrics. Numerical simulation results need to be confirmed by theoretical analysis. It is also important to obtain theoretical estimations of the rate of convergence for various stochastic wavelet expansions. |
Processed_Convergence_in_$L_p([0,T])$_of_wavelet_expansions_.txt | Our focus in this paper is on convergence in Lp([0, T ]) of wavelet expansions of ϕ-sub-Gaussian random processes. The paper extends the recent results on uniform convergence in the papers Kozachenko et al. 2011, 2013, Kozachenko, Polosmak 2008 to new classes of stochastic processes and probability metrics. |
Processed_Convergence_in_$L_p([0,T])$_of_wavelet_expansions_.txt | The analysis and the approach presented in the paper contribute to the in- vestigations of wavelet expansions of random processes in the former literature. The approach is of a special interest if p > 2, as it extends the available L2 results. In that sense, Theorems 4-7 are of special importance. The results are obtained under simple assumptions which can be easily verified. The paper deals with the most general class of wavelet expansions in comparison with particular cases considered by different authors, see, for example, Cambanis, Masry 1994, Kurbanmuradov, Sabelfeld 2008. |
Processed_Convergence_in_$L_p([0,T])$_of_wavelet_expansions_.txt | The organization of the article is the following. In the second section we intro- duce the necessary background from the theory of ϕ-sub-Gaussian random vari- ables and processes. Section 3 discusses wavelet expansions and approximations of stochastic processes. In 4 we present the main results on convergence in Lp([0, T ]) § of wavelet expansions of ϕ-sub-Gaussian random processes. In this section we also obtain the rate of convergence of the wavelet expansions and discuss some speci- fications for which the assumptions in the theorems can be easily verified. |
Processed_Convergence_in_$L_p([0,T])$_of_wavelet_expansions_.txt | In this section, we review the definition of ϕ-sub-Gaussian random processes and their relevant properties. |
Processed_Convergence_in_$L_p([0,T])$_of_wavelet_expansions_.txt | The space of ϕ-sub-Gaussian random variables was first introduced in the pa- per Kozachenko, Ostrovskyi 1985. More information about the space of ϕ-sub- Gaussian random variables and processes can be found in Buldygin, Kozachenko 2000, Giuliano Antonini et al. 2003, Kozachenko, Kamenshchikova 2009. |
Processed_Convergence_in_$L_p([0,T])$_of_wavelet_expansions_.txt | Definition 6 (Kozachenko, Kovalchuk 1985) ϕ-sub-Gaussian random process X(t), is t strictly Subϕ(Ω). The determinative constant of this family is called a determinative constant of the process and denoted by CX . |
Processed_Convergence_in_$L_p([0,T])$_of_wavelet_expansions_.txt | Remark 3 Gaussian centered random process X(t), t process, where ϕ(x) = x2/2 and the determinative constant CX = 1. |
Processed_Convergence_in_$L_p([0,T])$_of_wavelet_expansions_.txt | In this section we introduce wavelet representations and approximations of non- random functions and stochastic processes. |
Processed_Convergence_in_$L_p([0,T])$_of_wavelet_expansions_.txt | However, the majority of random processes does not possess the required prop- erty. For example, sample paths of stationary processes are not in the space L2(R) (a.s.). However, in many cases it is possible to construct a representation of type (3) for X(t). |
Processed_Convergence_in_$L_p([0,T])$_of_wavelet_expansions_.txt | In this section we present the main results on convergence in Lp(T), T = [0, T ], T > 0, of the wavelet expansions of ϕ-sub-Gaussian random processes. The rate of convergence in the space Lp([0, T ]) is obtained. We also present some specifications of the general results for which the assumptions can be easily verified. |
Processed_Convergence_in_$L_p([0,T])$_of_wavelet_expansions_.txt | Theorem 3 Let X(t), t X(t) is continuous in the norm τϕ( satisfy the assumptions of Theorem 2. |
Processed_Convergence_in_$L_p([0,T])$_of_wavelet_expansions_.txt | Proof. First, we will show that the random variables ξ0k and ηjk are from the space Subϕ(Ω). We prove it only for ηjk. The case of ξ0k can be dealt with similarly. We will need the following generalization of the Minkowski inequality. |
Processed_Convergence_in_$L_p([0,T])$_of_wavelet_expansions_.txt | for arbitrary ε > 0 and sufficiently large n, m inequality (11) holds true. By defi- (cid:16) nitions 1 and 2 the right-hand side of the inequality vanishes when cm 0, which implies (10). |
Processed_Convergence_in_$L_p([0,T])$_of_wavelet_expansions_.txt | Corollary 1 Let X(t), t tive constant CX . Then the condition (5) holds true if the integral is convergent. |
Processed_Convergence_in_$L_p([0,T])$_of_wavelet_expansions_.txt | Proof. The statement of Corollary 1 follows from Theorem 4, Definition 6, and the fact that a linear closure of a family of strictly Subϕ(Ω) random variables is a (cid:3) family of strictly Subϕ(Ω), see Kozachenko, Kovalchuk 1985. |
Processed_Convergence_in_$L_p([0,T])$_of_wavelet_expansions_.txt | Proof. Inequality (14) is a simple modification of an inequality from Hardle et al. 1998. Therefore we only prove (15). |
Processed_Convergence_in_$L_p([0,T])$_of_wavelet_expansions_.txt | Remark 7 Conditions of Theorems 6 and 7 on the random process X(t) are formu- lated in terms of its spectral density. These conditions are related to the behavior of the high-frequency part of the spectrum. Such assumptions are standard in the convergence studies of stochastic approximations. |
Processed_Convergence_in_$L_p([0,T])$_of_wavelet_expansions_.txt | Remark 8 If the assumptions of Theorem 6 or 7 are satisfied, then (7) holds true for c∞ n given by (17). All the terms in (17) can be easily computed in practice for specific stochastic processes and wavelet bases. |
Processed_Convergence_in_$L_p([0,T])$_of_wavelet_expansions_.txt | The obtained results may have various practical applications for the approxima- tion and simulation of random processes. The analysis of the rate of convergence provides a constructive algorithm for determining the number of terms in the wavelet expansions to ensure the approximation of stochastic processes with given accuracy. |
Processed_Convergence_in_$L_p([0,T])$_of_wavelet_expansions_.txt | The developed methodology and results are important extensions of the recent findings in the wavelet approximation theory of stochastic processes to the space Lp([0, T ]) and the class of ϕ-sub-Gaussian random processes. This class plays an important role in generalizations of various theoretical properties of Gaussian pro- cesses. In addition to classical applications of ϕ-sub-Gaussian random processes in signal processing, the results can also be used in new areas, like compressed sens- ing and actuarial modelling, consult, for example, Labate et al. 2013, Vershynin 2012, Yamnenko 2006. |
Processed_Transitions_between_turbulent_and_laminar_superflu.txt | We investigate the global transition from a turbulent state of superfluid vorticity (quasi-isotropic vortex tangle) to a laminar state (rectilinear vortex array), and vice versa, in the outer core of a neutron star. By solving numerically the hydro- dynamic Hall-Vinen-Bekarevich-Khalatnikov equations for a rotating superfluid in a differentially rotating spherical shell, we find that the meridional counterflow driven by Ekman pumping exceeds the Donnelly-Glaberson threshold through- out most of the outer core, exciting unstable Kelvin waves which disrupt the In the turbulent state, the rectilinear vortex array, creating a vortex tangle. torque exerted on the crust oscillates, and the crust-core coupling is weaker than in the laminar state. This leads to a new scenario for the rotational glitches ob- served in radio pulsars: a vortex tangle is sustained in the differentially rotating outer core by the meridional counterflow, a sudden spin-up event (triggered by an unknown process) brings the crust and core into corotation, the vortex tangle relaxes back to a rectilinear vortex array (in <∼ 105 s), then the crust spins down electromagnetically until enough meridional counterflow builds up (after <∼ 1 yr) to reform a vortex tangle. The turbulent-laminar transition can occur uniformly or in patches; the associated time-scales are estimated from vortex filament the- ory. We calculate numerically the global structure of the flow with and without an inviscid superfluid component, for Hall-Vinen (laminar) and Gorter-Mellink (turbulent) forms of the mutual friction. We also calculate the post-glitch evo- lution of the angular velocity of the crust and its time derivative, and compare the results with radio pulse timing data, predicting a correlation between glitch activity and Reynolds number. Terrestrial laboratory experiments are proposed to test some of these ideas. |
Processed_Transitions_between_turbulent_and_laminar_superflu.txt | Timing irregularities in a rotation-powered pulsar, such as discontinuous glitches (Lyne et al. 2000; Zou et al. 2004; Shabanova 2005) and stochastic timing noise (Hobbs 2002; Scott et al. 2003), provide an indirect probe of the internal structure of the star. The physical processes usually invoked to explain these phenomena are (un)pinning of Feynman-Onsager vortices in the crystalline inner crust (Anderson & Itoh 1975), starquakes (Ruderman 1976), and thermally driven vortex creep (Alpar et al. 1984b; Link et al. 1993). Less attention has been directed at the global hydrodynamics of the superfluid, except within the context of the spin-up problem in cylindrical geometry (Anderson et al. 1978; Reisenegger 1993; Carter et al. 2000). The importance of the global hydrodynamics was demonstrated by Tsakadze & Tsakadze (1980), who simulated pulsar rotational irregularities in the laboratory by impulsively accelerating rotating containers of He II, obtaining qualitative agreement with radio timing data (e.g. glitch amplitudes and post-glitch relaxation times). |
Processed_Transitions_between_turbulent_and_laminar_superflu.txt | In this paper, we examine how the global flow pattern of superfluid in the outer core of a neutron star affects the rotation of the star. We focus on the outer core for simplicity: vortex pinning is thought to be weak or non-existent there (Alpar et al. 1984a; Donati & Pizzochero 2003, 2004), and the fluid is mainly isotropic (1S0 Cooper pairing) (Sedrakian & Sedrakian 1995; Yakovlev et al. 1999), reducing the problem to a hydrodynamic one in a spherical shell. Even with this simplification, the calculation remains numerically challenging: the spherical Couette problem for a superfluid was solved for the first time only recently (Peralta et al. 2005), generalizing previous work on the cylindrical Taylor-Couette problem for a superfluid (Henderson et al. 1995; Henderson & Barenghi 2004) and the spherical Couette problem for a classical viscous fluid (Marcus & Tuckerman 1987a; Dumas & Leonard 1994). |
Processed_Transitions_between_turbulent_and_laminar_superflu.txt | component of the spherical Couette flow (SCF) in the interior. The mutual friction force changes dramatically during transitions between a vortex array and a vortex tangle (Gorter & Mellink 1949; Vinen 1957; Swanson et al. 1983; Schwarz 1985), affecting the rotational evolution of the star. |
Processed_Transitions_between_turbulent_and_laminar_superflu.txt | In this paper, we propose a phenomenological model for timing irregularities in radio pulsars based on the creation and destruction of a vortex tangle — superfluid turbulence (Barenghi et al. 1995; Vinen 2003) — in the outer core of a rotating neutron star. In our scenario, a glitch comprises the following sequence of events. (i) Differential rotation between the outer core and crust of the star, built up over time through electromagnetic spin down, generates a meridional Ekman counterflow in the outer core. We show that the axial counterflow exceeds the DGI threshold, creating a vortex tangle throughout the outer core. The mutual friction in this turbulent state takes the isotropic Gorter-Mellink (GM) form and is much weaker than the mutual friction associated with a rectilinear vortex array. (ii) When the glitch occurs, triggered by an unknown mechanism, the outer core and inner crust suddenly come into corotation and the vortex tangle decays, ultimately converting into a rectilinear vortex array. The decay process lasts <∼ 105 s, depending on the drag force acting on the vortex rings, after which the mutual friction takes the anisotropic Hall-Vinen (HV) form (Hall & Vinen 1956a,b) and increases by ∼ 5 orders of magnitude, precipitating a “torque crisis”. (iii) After the glitch, differential rotation builds up again between the outer core and the crust due to electromagnetic spin down. When the axial counterflow exceeds the DGI threshold, after <∼ 1 yr, the vortex array breaks up again into a tangle and the mutual friction drops sharply. Similar transitions from turbulent to laminar flow in a superfluid have been observed in laboratory experiments where He II, cooled to a few mK, flows around an oscillating microsphere (Niemetz et al. 2002; Schoepe 2004). |
Processed_Transitions_between_turbulent_and_laminar_superflu.txt | The paper is organized as follows. HVBK theory is briefly reviewed in §2, together with the pseudospectral numerical method which we use to solve the HVBK equations in a differentially rotating spherical shell. The physics of the turbulent-laminar transition in a generic glitch scenario is elaborated in §3. The response of the stellar crust to a turbulent- laminar transition in the outer core is calculated numerically in §4. Finally, the results are summarized and applied to observational data in §5; a fuller observational analysis will be carried out in a future paper. |
Processed_Transitions_between_turbulent_and_laminar_superflu.txt | with dn,s/dt = ∂/∂t + vn,s · ∇, where vn (vs) and ρn (ρs) are the normal fluid (superfluid) velocities and densities respectively. Effective pressures ps and pn are defined by ∇ps = ∇p − 1 ns), with vns = vn − vs. Note that, in neutron star applications, the gravitational potential can be subsumed into p without loss of generality, in the incompressible limit. |
Processed_Transitions_between_turbulent_and_laminar_superflu.txt | with ˆωs = ωs/|ωs|. Here, νs = (κ/4π) ln(b0/a0) is the stiffness parameter, κ = h/2mn is the quantum of circulation, mn is the mass of the neutron, a0 is the radius of the vortex core, and b0 is the intervortex spacing. Note that νs, which has the dimensions of a kinematic viscosity, controls the oscillation frequency of Kelvin waves excited on vortex lines (Henderson et al. 1995). |
Processed_Transitions_between_turbulent_and_laminar_superflu.txt | where A′ = B3ρ2 2 is a dimensionless temperature-dependent coefficient, related to the original GM constant (usually denoted by A in the literature) by A′ = Aρκ. Here, χ1 and χ2 are dimensionless constants of order unity (Vinen 1957). The HV and GM forces are in the ratio ∼ 105 under the physical conditions prevailing in the outer core of a neutron star (Peralta et al. 2005). This implies that the normal and superfluid components of the star (and hence the crust and core, through viscous torques) are effectively uncoupled most of the time (when enough differential rotation has built up), but become tightly coupled in the immediate aftermath of a glitch when a turbulent-to-laminar transition occurs (Peralta et al. 2005). It is important to bear this in mind when reading §3. |
Processed_Transitions_between_turbulent_and_laminar_superflu.txt | We model the global hydrodynamics of the outer core by considering an HVBK super- fluid enclosed in a differentially rotating spherical shell, i.e. superfluid spherical Couette flow. The inner (radius R1) and outer (radius R2) surfaces of the shell rotate at angular velocities Ω1 and Ω2, respectively, about a common axis. All quantities are expressed in dimensionless form using R2 as the unit of length and Ω−1 1 as the unit of time. We adopt spherical polar coordinates (r, θ, φ), with the z direction along the axis of rotation. |
Processed_Transitions_between_turbulent_and_laminar_superflu.txt | The boundary conditions satisfied by the normal and superfluid components are sub- tle. Neither component can penetrate the boundaries, implying (vn)r = (vs)r = 0. The (vn)θ = 0, tangential normal fluid velocity satisfies the no-slip Dirichlet condition, viz. (vn)φ = R1,2Ω1,2 sin θ. The behavior of the superfluid at the boundaries is influenced by the quantized vortices, which can either slide past, or pin to, irregularities on the boundaries. The former scenario implies (ωs)θ = (ωs)φ = 0, while the latter implies (ωs)rωs × vL = 0 and requires a knowledge of the vortex line velocity vL locally. As the HVBK model pro- vides no information about vL, we adopt the widely used no-slip compromise (vs)θ = 0, (vs)φ = R1,2Ω1,2 sin θ, which in the spherical case does not restrict the orientation of the vortex lines at the surface (Henderson et al. 1995; Henderson & Barenghi 1995, 2004). Note that, initially, vn and vs must be divergence-free, with ∇ × vs 6= 0; the Stokes solution with vs = vn satisfies this constraint (Landau & Lifshitz 1959). However, the numerical results presented in §3 and §4 are obtained after several Ekman times (and hence rotation periods) have elapsed, by which time the flow has forgotten its initial conditions, as in most astrophysical applications. |
Processed_Transitions_between_turbulent_and_laminar_superflu.txt | spherical Couette flow differs from its cylindrical counterpart in two important ways. First, in a cylinder, the principal flow is toroidal, whereas, in a sphere, the principal flow is an axisymmetric combination of toroidal flow and meridional circulation for all Re (Tuckerman 1983). 2 Second, it is misleading to approximate the equatorial region of a sphere by a cylin- der, even for R1 ≈ R2; the curvature, although slight, causes significant differences in the critical Taylor number at which vortices appear (Soward & Jones 1983; Stuart 1986). The differences between cylindrical and spherical Couette flow are equally prominent in classical fluids and superfluids (Peralta et al. 2005). |
Processed_Transitions_between_turbulent_and_laminar_superflu.txt | Spherical Couette flow of a viscous fluid is controlled by three parameters: the Reynolds number Re = Ω1R2 2/νn, the dimensionless gap width δ = (R2−R1)/R2, and the angular shear ∆Ω = Ω2 − Ω1. If a superfluid is also introduced, νs emerges as an additional parameter. In the laminar regime (Re <∼ 104), the global flow can be classified according to the number of cells of meridional circulation in each hemisphere. In a narrow gap (δ < 0.11), the merid- ional circulation is slow and can be approximated by (vn)θ/(R2∆Ω) ≈ 10−2δ2(R2/R1)Re (Yavorskaya et al. 1986). As Re increases, a centrifugal instability generates toroidal Taylor vortices near the equator until, at the onset of turbulence (Re >∼ 104), a helical traveling wave develops. In a wide gap (δ > 0.4), by contrast, all secondary flows are nonaxisymmetric, and they oscillate when Re exceeds a threshold (Yavorskaya et al. 1986). In classical fluids, these flows have been thoroughly investigated numerically and experimentally (Yavorskaya et al. 1975, 1977; Belayev et al. 1978; Yavorskaya et al. 1986; Marcus & Tuckerman 1987a,b; Junk & Egbers 2000). |
Processed_Transitions_between_turbulent_and_laminar_superflu.txt | The flow pattern in Figure 1 changes with temperature just like in cylindrical Couette flow. The superfluid behaves like a classical, viscous fluid near the critical temperature. The Taylor vortices elongate axially, with a more complex pattern of eddies and counter-eddies emerging, as the temperature is lowered (Henderson et al. 1995; Henderson & Barenghi 1995). For Re <∼ 268, anomalous modes emerge in the normal component, as in classical axisymmetric flow (Lorenzen & Mullin 1985), characterized by Ekman cells rotating in the opposite sense to the classical flow (Henderson & Barenghi 2000) except near the critical temperature. For Re >∼ 268, the tension becomes less important than the mutual friction and the superfluid component comes to resemble the normal component (Henderson & Barenghi 1995). |
Processed_Transitions_between_turbulent_and_laminar_superflu.txt | In this section, we investigate how the vorticity state of the outer core changes before, during, and after a glitch, in the context of a generic glitch scenario where the trigger mechanism of the glitch is unspecified. |
Processed_Transitions_between_turbulent_and_laminar_superflu.txt | The axial component of the Ekman counterflow in the outer core of a typical neutron star generally exceeds the instability threshold (6). This is illustrated by Figure 2, a greyscale plot of (vns)z/vDG; the DGI is active in regions with (vns)z/vDG ≥ 1. Figure 2a corresponds to HV mutual friction (initially rectilinear vortex array); Figure 2b corresponds to GM mutual friction (after a tangle forms). In both figures, the DGI is active in most of the computational domain, implying that an initially rectilinear array is disrupted and, once disrupted, stays that way. This result is extracted empirically from the simulations. The inclusion of compressibility and hence stratification (Abney & Epstein 1996) restricts the DGI region to a thin boundary layer, but the overall conclusion is the same (see §4.5). |
Processed_Transitions_between_turbulent_and_laminar_superflu.txt | where t is the time elapsed since the last glitch (Lyne & Graham-Smith 2006; Lyne et al. 2000). The differential rotation induces meridional circulation (Greenspan 1968; Reisenegger 1993): an Ekman boundary layer, with (vn)θ ∼ R∗∆Ω, develops on a time-scale ∼ 2π/Ω∗, and grows radially to a thickness dE ≈ Re −1/2R∗ cm on a time scale tE ≈ Re 1/2(Ω∗)−1, spinning up the interior fluid (Andersson 2003; Andersson et al. 2005). Only the normal fluid is Ekman pumped directly; the superfluid is spun up by the normal fluid through the mutual friction force (Adams et al. 1985; Reisenegger 1993). |
Processed_Transitions_between_turbulent_and_laminar_superflu.txt | approaches the average vortex separation, whereupon the vortex lines continuously reconnect to form loops (Jou & Mongiov`ı 2004). Reconnection stops once the counterflow ceases, as happens immediately after a glitch, when the outer core and inner crust come into corotation (Ω1 = Ω2). The decay time-scale τd equals, to a good approximation, the reconnection time- scale just before the counterflow ceases. |
Processed_Transitions_between_turbulent_and_laminar_superflu.txt | where vns is the counterflow speed immediately before the differential rotation shuts off, given by (8). |
Processed_Transitions_between_turbulent_and_laminar_superflu.txt | where xp = ρ/ρp is the proton fraction, and mp and m∗ p are the bare and effective masses of the proton respectively (Andersson et al. 2006). Equation (13) includes modifications from entrainment (Mendell 1991; Andersson et al. 2006). In the outer core, with ρ = 2.8 × 1014 g cm−3 and xp = 0.038 (Mendell 1991), we obtain B = 2.7 × 10−4, approximately the value quoted by Sauls et al. (1982) and Sedrakian & Cordes (1998). |
Processed_Transitions_between_turbulent_and_laminar_superflu.txt | In laboratory and numerical experiments, it is observed that the vortex tangle is polar- ized; the average vorticity projected along the rotation axis is not zero (Finne et al. 2003; Tsubota et al. 2004; Tsubota & Kasamatsu 2005). This is because, under these experimen- >∼ β4vns/(κΩ2)1/2 in equation (10), where β1 is the polarization- tal conditions, one has β1 inducing term. Naively, one might expect the tangle to be even more polarized in a rapidly rotating neutron star, but in fact the opposite is true: the tangle is less polarized, because we have β1 ≪ β4vns/(κΩ2)1/2 from Figure 2. A thorough study of this issue, including whether the large-Ω2 limit considered by Jou & Mongiov`ı (2004) in deriving equation (10) is applicable for vns ≫ (κΩ2)1/2, lies outside the scope of this paper. |
Processed_Transitions_between_turbulent_and_laminar_superflu.txt | density regime, where the relaxation time-scale equals the age of the pulsar [Sedrakian & Cordes (1998); A. Sedrakian 2006, private communication]. |
Processed_Transitions_between_turbulent_and_laminar_superflu.txt | whereupon a vortex tangle develops again in the outer core, via the DGI. This triggers a switch from HV to GM mutual friction, weakening the coupling between the normal fluid and superfluid components. |
Processed_Transitions_between_turbulent_and_laminar_superflu.txt | where (16) follows from (15) by substituting (6). Hence, for typical neutron star parameters, the tangle is reestablished over ∼ 1 d after ∼ 1 yr elapses following a glitch, less than the inter-glitch interval observed in most pulsars (Shemar & Lyne 1996; Lyne et al. 2000). Note that τg does not equal τd; the time required for a tangle to grow from a rectilinear array is shorter than the time required for an existing tangle to decay back to a rectilinear array, provided that t < 8.0ttan (and longer otherwise). Note also that vns just before the glitch (∼ R∗∆Ω), which appears in (11), is typically greater than vns at ttan (∼ vDG), which appears in (15), provided that t > ttan. |
Processed_Transitions_between_turbulent_and_laminar_superflu.txt | Equation (16) gives the minimum time for a tangle to reform, assuming it does so simultaneously everywhere in the outer core of the star. In practice, vns does not exceed vDG everywhere simultaneously. Regions where the DGI is activated are interspersed with regions where it is not, even at r = R1 and R2, so that the transition from HV to GM friction is more gradual than equation (16) suggests. This important issue, which is also relevant to the interpretation of the laboratory experiments performed by Tsakadze & Tsakadze (1980), is explored briefly in §5. However, a thorough numerical study lies outside the scope of this paper. |
Processed_Transitions_between_turbulent_and_laminar_superflu.txt | In this section, we calculate numerically the torque acting on the stellar crust during a transition from a state of turbulent vorticity (tangle) to laminar vorticity (rectilinear array), by solving the hydrodynamic HVBK equations in a differentially rotating spherical shell. Global simulations of this type are necessary to calculate the observable response of a neutron star, e.g. Ω(t) and ˙Ω(t), to the internal physics elaborated in §3. |
Processed_Transitions_between_turbulent_and_laminar_superflu.txt | We solve equations (1) and (2) using a pseudospectral collocation method to discretize the problem (Boyd 2001; Canuto et al. 1988) and a time-split algorithm to step forward in time (Canuto et al. 1988), as described by Peralta et al. (2005). The velocity fields are expanded in a restricted Fourier series in θ and φ and a Chebyshev series in r (Orszag 1974; Boyd 2001), with (Nr, Nθ, Nφ) = (120, 250, 4) modes required to fully resolve the flow. The initial time-step ∆t = 10−4 is lowered to ∆t = 10−5 during the GM → HV transition to prevent spurious oscillations in the torque (Peralta et al. 2005). Instabilities arising from the sensitivity of spectral methods to the boundary conditions (Peralta et al. 2005), and oscillations due to the Gibbs phenomenon (Gottlieb et al. 1984; Canuto et al. 1988), are smoothed using a low-pass spectral filter (Don 1994), a common practice (Osher 1984; Mittal 1999). |
Processed_Transitions_between_turbulent_and_laminar_superflu.txt | We adopt parameters as close to those of a realistic neutron star as our computational In the outer core, we take ρs/ρ = 0.99 and ρn/ρ = 0.01 (Baym et al. resources permit. 1992), and hence B = 1.5, B′ = 0.9, and A′ = 1.0 × 10−4 at the corresponding scaled temperature T /Tc in He II (Barenghi et al. 1983; Donnelly & Barenghi 1998). These friction coefficients are ∼ 104 times greater than those used in the analytic estimates in Section 3. We adopt the higher values deliberately. Otherwise, the effects introduced by B and B′ would take too long to build up; computational limitations prevent us from simulating more than ∼ 10 rotation periods. (For the same reason, our ∆Ω/Ω is unrealistically large.) Realistic Reynolds numbers, estimated from equation (9), are too high to be simulated directly, so we restrict ourselves to the range 103 <∼ Re <∼ 105; the most stable evolution occurs for Re ≈ 3 × 104, corresponding to an Ekman number (Re)−1 ≈ 3 × 10−5 that is two orders of magnitude smaller than in a typical neutron star (Abney & Epstein 1996). The tension force is dominated by mutual friction, except in relatively old (> 104 yr) neutron stars (Greenstein 1970), with νs ∼ 10−18 in our dimensionless units. For 10 ≤ A ≤ 104, we find empirically that the torque depends weakly on the GM force, so we use 10 ≤ A ≤ 102 to obtain identical results at lower computational cost. |
Processed_Transitions_between_turbulent_and_laminar_superflu.txt | before the transition lies in the range 0.1 ≤ ∆Ω ≤ 0.3, for gaps in the range 0.2 ≤ δ ≤ 0.5. Note that we are forced to choose ∆Ω unrealistically large, compared to typical neutron star values, in order to allow the inner and outer surfaces of the shell to “lap” each other within a reasonable run time.8 Various combinations of ∆Ω and δ affect the viscous torque on the inner and outer surfaces during steady differential rotation in different ways, as discussed below. |
Processed_Transitions_between_turbulent_and_laminar_superflu.txt | Streamline snapshots for a run with δ = 0.5 and ∆Ω = 0.2 are presented in Figures 3a–3f, with the turbulent-laminar transition GM → HV, Ω2 → Ω1 occurring at t = 20. The HV mutual friction, with |FHV|/|FGM| ∼ 105, couples the normal and superfluid components strongly. 9 Both components exhibit a quasi-periodic motion, with a secondary circulation cell filling most of the shell, and two or three smaller vortices emerging intermittently. An oscillatory polar Ekman cell also exists initially, disappearing at t = 140. |
Processed_Transitions_between_turbulent_and_laminar_superflu.txt | We can compare this behaviour to the flow without a turbulent-laminar transition. Figures 4a–4f display a sequence of streamline snapshots for δ = 0.5 and ∆Ω = 0.2, with GM mutual friction at all times. At t = 18, three polar cells are observed in both components, together with a secondary cell at mid-latitudes. The latter cell elongates after t = 20 and is present only in the superfluid component at later times. At the equator, the flow switches between one and three cells, but this behavior persists after t = 100 only in the superfluid component, while the normal fluid decouples increasingly. |
Processed_Transitions_between_turbulent_and_laminar_superflu.txt | The axial counterflow is significant at all times in Figures 4 and 5. Scaling to neutron star parameters, we find vns/vDG ∼ 103(∆Ω/10−4 rad s−1)(Ω∗/102 rad s−1) near the equator, and vns/vDG ∼ 106(∆Ω/10−4 rad s−1)(Ω∗/102 rad s−1) near the poles, at t = 200. |
Processed_Transitions_between_turbulent_and_laminar_superflu.txt | followed by a rapid exponential decay and persistent, small-amplitude oscillations. We find an e−1 relaxation time of τ ∼ 2, 1.6, and 1.6 for ∆Ω = 0.1, 0.2, and 0.3 respectively, with τ essentially independent of δ. In contrast, in the inverse experiment, where the mutual friction changes from HV to GM, Peralta et al. (2005) observed that ∆ ˙Ω is almost constant after the post-glitch transient. The period of the oscillations does not change with t for t >∼ 100. |
Processed_Transitions_between_turbulent_and_laminar_superflu.txt | The effect of the superfluid on the evolution can be analyzed with the help of Figure 7a, which plots the torque on the outer sphere as a function of time (solid curve) in the absence of spin up (δ = 0.5, Re = 3 × 104, ∆Ω = 0.1, GM friction). For comparison, the dashed curve plots the torque for a classical viscous fluid with the same parameters mentioned above. The evolution is qualitatively similar in both cases, displaying oscillations with the same periodicity and similar amplitude, but the curves diverge for t >∼ 20 as the viscous fluid evolves more rapidly to a stationary state, driven by classical Ekman pumping. Classical theory predicts a time-scale tE ≈ 173 in the limit where the Rossby number ∆Ω/Ω vanishes; an exponential fit to the dashed curve gives tE ≈ 243, which agrees well given that we have ∆Ω/Ω = 0.1 in our numerical experiments. In a superfluid, on the other hand, the torque decays faster initially, with time constant ∼ 50, then starts to increase again for t ∼ 240, as part of a long-period oscillation whose ultimate fate is unknown as it occurs over a time-scale beyond our computational limit. |
Processed_Transitions_between_turbulent_and_laminar_superflu.txt | Figure 7b shows the evolution of the fractional change in angular velocity after spin up (Ω2 = 0.9 → 1.0 at t = 20) in three scenarios: instantaneous change in mutual friction, GM → HV (solid curve); constant GM friction (dashed curve); and a classical viscous fluid (dashed dotted curve). The classical viscous fluid and constant GM friction evolve similarly. The torque decays exponentially on a time-scale ∼ 2.15, then decays almost linearly with t. By contrast, in the GM → HV transition, the torque decays exponentially, on a time-scale ∼ 2.20, then oscillates with peak-to-peak amplitude ≈ 0.4 (in units of ρR5 1) and period ∼ 6. In other words, the oscillations are sustained by the HV friction, which is much weaker than its GM counterpart. Note that the oscillations do not correspond to Tkachenko waves, since the HVBK equations assume that the free energy of the vortex array depends only on the vortex line density, not on vortex lattice deformations (Chandler & Baym 1983, 1986; Donnelly 1991). |
Processed_Transitions_between_turbulent_and_laminar_superflu.txt | The meridional streamlines for the transitions described in the previous paragraph evolve as in Figure 8. Figures 8a–8b show the streamlines for a classic viscous fluid before (t = 20) and after (t = 22) the glitch. There is a rapid redistribution of vorticity near the outer sphere, with two circulation cells replaced by one elongated meridional shell, while the flow near the inner sphere hardly changes, with one circulation cell near the equator and one near the pole. Figures 8c–8d show the meridional streamlines for the normal component of the superfluid. The flow resembles a classical viscous fluid (cf. Figure 8a), although it does become more complex after the spin up. As shown in Figure 8d, the large cell near the outer sphere is replaced by one primary circulation cell and three secondary cells (including one near the poles). Note that the greatest contributions to the torque come from the mid- latitude regions where the structure of eddies and counter-eddies is richer, as in Figure 8d. Finally, in Figures 8e–8f, we see the meridional streamlines for the inviscid component of the superfluid. The pattern before the jump (Figure 8e) differs from the normal fluid; the GM friction creates additional eddies near the outer sphere and closer to the poles. However, after the transition to HV friction, the inviscid component comes to resemble the normal component, as can be seen by comparing Figure 8f with Figure 8d. This occurs because the two components are coupled more strongly by HV friction (|FHV|/|FGM| ∼ 105), so that the superfluid is dragged along by the normal fluid. |
Processed_Transitions_between_turbulent_and_laminar_superflu.txt | Gravitational stratification can strongly suppress Ekman pumping, as the Ekman layer is squashed close to the outer sphere (Clark et al. 1971; Abney & Epstein 1996). Nevertheless, the Ekman layer, however thin, always exists, in order to satisfy the boundary conditions at r = R2. It contains a meridional counterflow given by equation (8). Consequently, the transition from turbulent to laminar vorticity (and vice versa), which relies on such a meridional counterflow, still occurs in a stratified star, and its dramatic effect on the torque is the same. The Ekman layer, no matter how thin, acts like a film of “oil” between two sliding surfaces (here, the outer core and inner crust), whose coupling strength (“stickiness”) changes abruptly when the meridional speed of the “oil” exceeds a threshold. In other words, equations (8), (12), (14), and (16), and the conclusions that follow from them, are unaltered by stratification; even though the volume of the outer core occupied by the vortex tangle (which does not affect the torque on the crust directly) does change. |
Processed_Transitions_between_turbulent_and_laminar_superflu.txt | It is conceivable that the DGI is excited in shells at certain radii where vns peaks, so that we end up with a sequence of shells containing alternating laminar and turbulent superfluid vorticity. In this scenario, the detailed study of which lies outside the scope of this paper, the stability of the vorticity configuration is restricted by the Richardson criterion, which states that the configuration is Kelvin-Helmholtz unstable for N 2|∂vφ/∂r|−2 < 1/4 (Mastrano & Melatos 2005). In a neutron star, the Brunt-V¨ais¨ala frequency is N ≈ 5×102 s−1 (Reisenegger & Goldreich 1992). |
Processed_Transitions_between_turbulent_and_laminar_superflu.txt | The effects of stratification are not considered in detail in this paper because they cannot be studied properly with our numerical method; the pressure projection step only works with divergence-free velocity fields (Bagchi & Balachandar 2002; Peralta et al. 2005). However, a crude approach to get a feel for the effects is to numerically suppress vr using a low-pass exponential filter (Don 1994), viz. vr → exp[−(k/Nr)γ ln ǫ]vr, with 0 ≤ |k| ≤ Nr, where ǫ = 2.2 × 10−16 is the machine zero, and γ is the order of the filter. If we ramp up γ with r as γ = r(n2 − n1)/(R2 − R1) + n1 − R1(n2 − n1)/(R2 − R1), with n1 = 2 and n2 = 12, then vr is weakly suppressed near R2 (γ = 12) and strongly suppressed near R1 (γ = 2), so that the flow is confined approximately to concentric shells. Preliminary results, for a viscous Navier-Stokes fluid, show that the filtering reduces the meridional circulation, flattening the streamlines radially. |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.