Filename
stringlengths 22
64
| Paragraph
stringlengths 8
5.57k
|
|---|---|
Processed_Fused_Spatial_Point_Process_Intensity_Estimation_w.txt | noticeable observations. First, a denser graph such as the 5-NN graph produces a more accurate estimation result compared with that of a sparser graph such as the MST or 3-NN graph. The bottom panel of Figure 3 illustrates an example of the estimated coefficients using the 5-NN graph when m = 2000, nd2 = m, which demonstrates the capability of SVCI in capturing the cluster structure in the regression coefficients and detecting the abrupt changes across the boundaries of adjacent clusters. However, there is clearly a trade off between the estimation accuracy and computation efficiency when using different graphs; as reported in the left panel of Figure 4, the computation time (in seconds) using the 5-NN or Delaunay triangulation graph is roughly 1.5 times of the computation time using the MST. Second, the parameter estimation is more accurate when m grows larger, as evidenced by the decreasing value of MISEβ. Finally, SVCI-LRL produces comparable results with those from SVCI-PL when m = 800 or using the MST graph, but it notably outperforms SVCI-PL when m ≥ 1600. This is consistent with the findings in Baddeley et al. (2014), which showed that for datasets with a large number of points or a highly structured point pattern, the logistic likelihood method produces a less biased estimator than its Poisson counterpart. |
Processed_Fused_Spatial_Point_Process_Intensity_Estimation_w.txt | We further examine the performance of the SVCI model in terms of recovering the intensity function. Given ˆβ(u), we obtain the estimate of the log intensity function by log ˆρ(u) = zT (u) ˆβ(u) for u ∈ D. The right panel of Figure 4 compares the MISE of log ˆρ(u) from SVCI-PL, SVCI-LRL and LGCP, respectively. It is noted that SVCI-LRL maintains its superior performance when predicting the intensity function in comparison with SVCI-PL. Besides, both versions of SVCI produce more accurate estimates than LGCP when estimating the intensity function with clustered regression coefficients. |
Processed_Fused_Spatial_Point_Process_Intensity_Estimation_w.txt | Next we examine the performance of SVCI in recovering the clusters of coefficients. Table 2 reports the Rand index for each of the SVCI estimates ˆβ1, ˆβ2 and ˆβ0 averaged over 100 simulations, using the 5-NN graph and setting nd2 = m. Rand index measures the proportion of pairs consisting of a true parameter and the corresponding estimated parameter that agree by virtue of belonging either to the same cluster or to different clusters. Overall, SVCI achieves an accurate cluster recovery result, evidenced by the relatively high Rand index value ranging from 0.73 to 0.93 in all settings. We also find that SVCI-LPL surpasses SVCI-PL in detecting spatial clusters. Finally, an interesting observation is that ˆβ0 has a lower Rand index value than both ˆβ1 and ˆβ2, which might be the consequence of having more clusters in the true function of β0. |
Processed_Fused_Spatial_Point_Process_Intensity_Estimation_w.txt | Table 1: Scenario 1: Mean integrated squared error of β (MISEβ) averaged over 100 simulations for different values of m with nd2 = m, different connection graphs, and the Poisson-based SVCI-PL method and the logistic regression based SVCI-LRL method. |
Processed_Fused_Spatial_Point_Process_Intensity_Estimation_w.txt | Table 2: Scenario 1: Rand index of the estimates ˆβ1, ˆβ2 and ˆβ0 (averaged over 100 Monte Carlo simula- tions) for SVCI-PL and SVCI-LRL for different values of m, using nd2 = n and 5-NN connection graphs. |
Processed_Fused_Spatial_Point_Process_Intensity_Estimation_w.txt | Table 3: Scenario 1: Comparing the MISEβ (averaged over 100 simulations) and computation time (in seconds) between SVCI-PL and SVCI-LRL for different numbers of dummy points, using m = 1600 and 5-NN connection graphs. |
Processed_Fused_Spatial_Point_Process_Intensity_Estimation_w.txt | the road network in an area of Chicago, USA near the University of Chicago (Baddeley and Turner, 2005). We bound the linear network in a window D = R × [0, 1]2 and increase R to expand D, so that the linear network that resides in D grows with D at the same rate to obtain an increasing number of realizations on the network. |
Processed_Fused_Spatial_Point_Process_Intensity_Estimation_w.txt | We first consider a simplified case where there is no covariate available. We focus on the estimation of intensity function whose true value is a piece-wise constant function, that is, the intensity function ρ(u) = exp{β0(u)}, and the true value of β0(u) has a clustered pattern as shown in the left panel of Figure 5. The original graph is constructed following the method described in the last paragraph of Section 3.1. The upper part of Table 4 presents the MISE of log ρ, i.e., the log intensity function for each value of m. In general, we obtain similar findings on the linear network as on a planar window presented in Scenario 1; MISE from both the Poisson based and logistic based SVCI models show a convergence tendency as the domain expands and m goes up, and the logistic likelihood based method achieves a slightly more accurate estimation than the Poisson based method with a large number of points. It is clear from Table 4 that both the SVCI-PL and SVCI-LRL models outperform the KDE based and resample-smoothed Voronoi based intensity estimation methods in almost all settings. Previous studies (Barr and Schoenberg, 2010) show that KDE estimators may suffer from the problem of having substantial bias and high variance when there are abrupt changes in the intensity. Both SVCI and Voronoi.lpp are designed to alleviate this problem, as evidenced by their improved performance over the two KDE methods in Table 4. Nevertheless, SVCI seems to be more effective than Voronoi.lpp to capture abrupt changes or clustering patterns. We illustrate an example in the right panel of Figure 5, which plots the true and the estimated log intensity along a selected road segment from one simulation. It clearly shows that SVCI captures the intensity with discontinuities more efficiently than KDE.lpp. |
Processed_Fused_Spatial_Point_Process_Intensity_Estimation_w.txt | Left: Computation time to solve the optimization for one tuning parameter using different Figure 4: connection graphs; Right: the boxplots of MISEβ for SVCI-PL, SVCI-LRL with 5-NN graphs and LGCP. Reported results are averaged over 100 Monte Carlo simulations under Scenario 1. |
Processed_Fused_Spatial_Point_Process_Intensity_Estimation_w.txt | We also compare the computation time of each method and report the detailed results in Appendix Table A2 for various values of m. Taking m = 2400 as an example, to get one estimate, KDE.lpp requires 4.95 seconds, KDEQuick.lpp requires 0.084 seconds, and Voronoi.lpp requires 4.95 seconds. In contrast, SVCI-LRL need 0.93 seconds to construct the connection graph and 1.11 seconds to get an estimate. Although SVCI is not the fastest among the compared methods, overall, its computation is still reasonable and competitive, especially considering its superior performance in intensity estimations. |
Processed_Fused_Spatial_Point_Process_Intensity_Estimation_w.txt | We then consider the case with an intercept and two covariates as described in Scenario 2(b). The true regression coefficients are plotted in the subfigures (a-c) of Figure 6. The subfigures (d-f) of Figure 6 give the estimated coefficients from SVCI on the chicago network. The results demonstrate that our method is also capable of capturing the clustered coefficient patterns on a linear network. In addition, the log intensity estimation results presented in the lower part of Table 4 are in general consistent with the findings presented in Scenario 1 and Scenario 2(a), that is, the performance of the SVCI model with the logistic regression likelihood or with a larger m is more preferable. Table 4 displays the results from LGCP as a comparison, which indicate a clear improvement of using SVCI over LGCP in terms of estimation accuracy. |
Processed_Fused_Spatial_Point_Process_Intensity_Estimation_w.txt | Figure 5: Left: The true log intensity β0 on the chicago network in Scenario 2(a); Right: The true and the estimated log intensity along one road segment corresponding to the line between the two black arrows in the left panel of Figure 5. |
Processed_Fused_Spatial_Point_Process_Intensity_Estimation_w.txt | Table 4: Scenario 2: mean integrated squared error of log intensity (MISElog ρ) averaged over 100 simula- tions for different values of m. We compare the two estimating equations, the Poisson likelihood (PL) and the logistic regression likelihood (LRL) with their competitors. |
Processed_Fused_Spatial_Point_Process_Intensity_Estimation_w.txt | Figure 6: The top three panels (a-c): the spatial structures of true coefficients β1, β2 and β0 in Scenario 2(b); The upper three panels (d-f): the estimated coefficient surfaces from in one simulation using SVCI- LRL with n = 2400. |
Processed_Fused_Spatial_Point_Process_Intensity_Estimation_w.txt | We consider two real data examples to illustrate the performance of the proposed method. The first Toronto Homicide data example has a moderate data size with 1398 points and three explanatory variables on a domain with irregular boundaries. And the second Western Australia Traffic Accidents data has a larger data size with 14, 562 points on a linear road network. In both studies, we use SVCI-LRL and nd2 ≈ m, due to their favorable performance in our simulation studies. |
Processed_Fused_Spatial_Point_Process_Intensity_Estimation_w.txt | We apply the proposed SVCI model to the analysis of the Toronto Homicide dataset. The raw dataset contains the information of 1398 homicides occurred in Toronto, Canada during 1990 to 2014, recording the locations of murder scenes, homocide types and infor- mation of victims obtained from the Toronto Star Newspaper (http://www.thestar.com/ news/crime/torontohomicidemap.html). We select the more recent years since 2000 and delete the data which have duplicated locations. There remains 764 homicide cases for the final analysis. Figure 1 shows the entire Toronto city and the locations of the selected cases within a 42 × 31 km rectangle window. Notably, the old Toronto region in the middle of the coast has more frequent occurrences of homicides. |
Processed_Fused_Spatial_Point_Process_Intensity_Estimation_w.txt | The data also contains the records of average income, night lights and population density of Toronto city in 2006, and we use them as the explanatory variables. Figures 7 (a-c) show the observations of the three variables. As can be seen, there is a large spatial variation of these variables across the city. We scaled and centered each spatial covariates before running our point process models. |
Processed_Fused_Spatial_Point_Process_Intensity_Estimation_w.txt | where β1, β2 and β3 are the constant-coefficients across the domain and φ(u) is a spatial Gaussian process with a zero mean and a Matern correlation function. After centering and scaling the covariates, we obtain the parameter estimates from LGCP as β1 = −0.918, β2 = 0.378, β3 = 0.207 and β0 = 1.096. These estimates imply that the homicides are more likely to occur in the area with a lower average income, a better lights condition and a denser residential population. |
Processed_Fused_Spatial_Point_Process_Intensity_Estimation_w.txt | Here βk(u), k = 0, 1, 2, 3 are spatially varying coefficients, whose estimates are shown in Figure 7(d-f). It takes about 0.072 seconds to construct the 5-NN graph and 0.69 seconds to get an estimate of β for each tuning parameter. Clearly, the results of SVCI reveal more details about the effects of covariates than those from LGCP. Overall, the signs of βk(u) are the same as the results of LGCP. For example, the estimates of β1(u) range from -2 to 0, indicating a negative relationship between income and homicide occurrence as is expected. Such a negative relationship is most prominent in the western region of Toronto whereas a weaker relationship is observed near the upper east corner of Toronto City. For both β2 and β3, we observe a small cluster at the Old Toronto region, which has the most concentrated homicide cases. It is notable that the relationships between light intensity/population density and homicides occurrence in the Old Toronto region differ significantly from the rest of Toronto city; a stronger positive relationship is observed for both variables. |
Processed_Fused_Spatial_Point_Process_Intensity_Estimation_w.txt | In this section, we study the traffic accidents data in the state of Western Australia for the year 2011, as shown in Figure 1. The data were originally provided by the Western Australian State Government Department of Main Roads and are made publicly available as part of the Western Australian Whole of Government Open Data Policy. The data can also be accessed from the R package spatstat.Knet. It consists of 14, 562 locations of accidents on a road network with 115, 169 road segments constrained in a [217.4, 1679.1] × [6114.9, 7320.6] km rectangle window. |
Processed_Fused_Spatial_Point_Process_Intensity_Estimation_w.txt | The grey lines in Figure 1 represent the traffic network of Western Australia and each red point denotes an accident spot. It is clear from this Figure that accidents are highly concentrated around the Perth metropolitan area located in the western coastal region. This region contains nearly 75% of the population in Western Australia. By contrast, the remote eastern region has a much sparser road network and a smaller number of traffic accidents. Our goal is to estimate the intensity function over this network to investigate the spatial variation of accident occurrences. |
Processed_Fused_Spatial_Point_Process_Intensity_Estimation_w.txt | takes about 1.75 minutes to construct a connection graph using the graph construction method in Section 3.1 and takes on average 5.01 seconds to get an estimate of β for each tuning parameter. Figure 8 plots the estimated log intensity log ˆρ(u) on the road network. We notice that ˆρ(u) has a large spatial variation, ranging from 0 per kilometer in some remote eastern areas to nearly 50 accidents per kilometer in some busy roads in the Perth metropolitan area. |
Processed_Fused_Spatial_Point_Process_Intensity_Estimation_w.txt | Figure 9: (a): Intensity estimates map for the accidents in the metropolitan Perth Area; (b): Intensity estimates map by zooming into the red circle area in the left panel. |
Processed_Fused_Spatial_Point_Process_Intensity_Estimation_w.txt | estimated intensity function appears near the southwest end of the road. After verifying with the Google satellite image, we confirm that the northeast part of the road passes through a large residential area, whereas the southwest part is a commercial and public service area (restaurants/shops/school/church/hospital) that is expected to have a higher rate of accidents. |
Processed_Fused_Spatial_Point_Process_Intensity_Estimation_w.txt | competing methods such as the KDE-based methods (McSwiggan et al., 2017; Rakshit et al., 2019) and the resample-smoothing Voronoi intensity estimation method (Moradi et al., 2019), when intensity functions exhibit discontinuous changes on linear networks. The computation of SVCI is in general reasonable compared to its competitors considered in this paper. The method is applied to identify spatially heterogeneous patterns in the determinants of Toronto crime events and the intensity of traffic accidents in Western Australia. |
Processed_Fused_Spatial_Point_Process_Intensity_Estimation_w.txt | Moving forward, this work could be further refined in several ways. First, SVCI only considers a small fixed number of covariates. However, in practice, practitioners may face a large number of available covariates but lack a strong theory to inform variable selection. There is a need of a more general model that allows researchers to undergo variable selection and spatial cluster identification simultaneously for point processes. Second, the SVCI estimator does not come with an uncertainty measure that makes it hard for statistical inference, a common issue shared by many regularization based approaches. We may consider a Bayesian version of the method or a bootstrapping based approach to address the inference problem. Third, an interesting research direction is to extend the finite dimensional graph regularization based method to an infinite dimensional process-based clustered coefficient model such that spatial predictions can be done in a more rigorous way. Moreover, the method can be extended by considering a weighted composite log- likelihood to further improve statistical efficiency for non-Poisson point processes (Guan and Shen, 2010). Finally, we acknowledge that further investigations are needed on how to verify and relax some of the assumptions that are used to establish the theoretical results. Tighter error bounds may be obtained following the method of entropy bound similar as those in Ortelli and van de Geer (2019). In addition, infilling domain asymptotic theoretical framework might be more reasonable for point patterns on bounded domains or linear networks. However, these are not trivial theoretical questions, and we will leave those for future research. |
Processed_Fused_Spatial_Point_Process_Intensity_Estimation_w.txt | The proof of Lemma 1 can be found in Guan and Shen (2010) and Thurman et al. |
Processed_Fused_Spatial_Point_Process_Intensity_Estimation_w.txt | For the quadrature approximation in (4), we need to divide the domain into small subdivisions or quadrats. For the 2-D square domain, it is easy to add dummy points and draw regular equal-sized rectangular quadrats on the domain using dummy.ppm and quad.ppm in R library spatstat. For planar domains with irregular boundary, we follow the routine quadrature approximation methods for irregular domains in R2 (Shen et al., 2009), where the domain is masked by regular pixel grids, and we expand the domain slightly to include the whole pixel if it has intersections with domain boundary. See the left panel of Figure A1 for an example. |
Processed_Fused_Spatial_Point_Process_Intensity_Estimation_w.txt | As for the partition on a linear work, Chapter 9 of Okabe and Sugihara (2012) discussed the implementation of both equal-length and unequal-length network cells. Furuta et al. (2008); Shiode (2008) proposed computational methods for dividing a network into equal- length network cells. However, their methods are not a guarantee of success if we want to insert enough dummy points. We propose to randomly draw dummy points from a homogeneous Poisson point process on the linear network with intensity function δ(u) = (M − m)/|D|. Then we obtain the quadrats on linear networks using the network Voronoi tessellation method (Chapter 4 of Okabe and Sugihara, 2012), each of which contains one dummy point as its centroid. See the right panel of Figure A1 for an example. |
Processed_Fused_Spatial_Point_Process_Intensity_Estimation_w.txt | Figure A1: Illustration of quadrat schemes for Toronto city (left) and a toy linear network example (right); Left panel: boundary lines (red) of Toronto city and the grids (grey) covering the irregular domain; Right panel: dummy points (red nodes) and their corresponding Voronoi quadrats (marked by different shades of grey). |
Processed_Fused_Spatial_Point_Process_Intensity_Estimation_w.txt | The results generally agree well with the findings based on Rand index and the MISEβ, in the sense that SVCI-LRL achieves a slightly smaller bias and variance compared with SVCI-PL. |
Processed_Fused_Spatial_Point_Process_Intensity_Estimation_w.txt | Under the setting of Scenario 1, we examine the performance of SVCI-PL based on K-NN graphs with K = 3, 4, 5, 6, 9 and r-NN graphs with r = 0.02R, 0.03R. Figure A2 shows the MISEβ versus the number of edges for each graph. We notice that MISE does not always decrease as the graph includes more neighbors by increasing K of K-NN or r of r-NN. Both K-NN and r-NN lose some estimation accuracy if there are too many neighbors. It also seems that K-NN slightly outperform r-NN in terms of MISE when K and r are chosen such that the two graphs have comparable number of edges. |
Processed_Fused_Spatial_Point_Process_Intensity_Estimation_w.txt | Figure A2: Illustration of MISEβ versus the number of edges, under the setting of Scenario 1, with n = 2400 and nd2 = m. (cid:15)R indicates the SVCI-PL method based on an r-NN graph with a radius (cid:15) × R. |
Processed_Fused_Spatial_Point_Process_Intensity_Estimation_w.txt | Table A1: Scenario 1: The integrated squared bias and variance (in parentheses) of ˆβ1, ˆβ2 and ˆβ0 respec- tively, in the case that m = 2400, and nd2 = m based on 5-NN connection graphs. |
Processed_Fused_Spatial_Point_Process_Intensity_Estimation_w.txt | the 3-NN graphs constructed based on the shortest-path distance and Euclidean distance metrics, respectively. Overall, we observe very similar results between these two choices of distance metrics, especially when m and nd2 go up. For example, when n = 2400, the MISEs of SVCI-LRL using shortest-path distance and Euclidean distance metrics are 0.121 and 0.124, respectively, which are very close to the result reported in Table 4 obtained using the graph constructed by connecting natural neighbors on the Chicago network. |
Processed_Fused_Spatial_Point_Process_Intensity_Estimation_w.txt | Table A3 reports the computation time of different methods with one tuning parameter, under the setting of Scenario 2(a). We notice that the computation time of KDE.lpp and Voronoi.lpp vary notably with their tuning parameter, so the computation times are calculated by averaging 20 repeats with a sequence of tuning parameters. |
Processed_Fused_Spatial_Point_Process_Intensity_Estimation_w.txt | For the Toronto Homicide data considered in Section 5.1, we compare the estimated log intensity surfaces obtained by SVCI and LGCP in Figure A3. Both methods seem to be capable of capturing the inhomogeneity pattern in intensities and agree well with each other in most areas. The most notable difference between the two methods occurs near Toronto islands, where we observe more variations in intensity estimations by LGCP than those by SVCI possibly due to the piece-wise homogeneity assumptions made in the latter. |
Processed_Fused_Spatial_Point_Process_Intensity_Estimation_w.txt | Baddeley, A., Chang, Y.M., Song, Y., Turner, R., 2012. Nonparametric estimation of the dependence of a spatial point process on spatial covariates. Statistics and its interface 5, 221–236. |
Processed_Fused_Spatial_Point_Process_Intensity_Estimation_w.txt | Beck, A., Teboulle, M., 2009. Fast gradient-based algorithms for constrained total variation image denoising and deblurring problems. IEEE Transactions on Image Processing 18, 2419–2434. |
Processed_Fused_Spatial_Point_Process_Intensity_Estimation_w.txt | Boyd, S., Parikh, N., Chu, E., Peleato, B., Eckstein, J., et al., 2011. Distributed op- timization and statistical learning via the alternating direction method of multipliers. Foundations and Trends in Machine learning 3, 1–122. |
Processed_Fused_Spatial_Point_Process_Intensity_Estimation_w.txt | Choiruddin, A., Coeurjolly, J.F., Letu´e, F., et al., 2018. Convex and non-convex regular- ization methods for spatial point processes intensity estimation. Electronic Journal of Statistics 12, 1210–1255. |
Processed_Fused_Spatial_Point_Process_Intensity_Estimation_w.txt | Choiruddin, A., Coeurjolly, J.F., Waagepetersen, R., 2021. Information criteria for inho- mogeneous spatial point processes. To appear in Australian & New Zealand Journal of Statistics . |
Processed_Fused_Spatial_Point_Process_Intensity_Estimation_w.txt | Gelfand, A.E., Kim, H.J., Sirmans, C., Banerjee, S., 2003. Spatial modeling with spatially varying coefficient processes. Journal of the American Statistical Association 98, 387–396. |
Processed_Fused_Spatial_Point_Process_Intensity_Estimation_w.txt | Gon¸calves, F.B., Gamerman, D., 2018. Exact Bayesian inference in spatiotemporal Cox processes driven by multivariate Gaussian processes. Journal of the Royal Statistical Society: Series B (Statistical Methodology) 80, 157–175. |
Processed_Fused_Spatial_Point_Process_Intensity_Estimation_w.txt | Guan, Y., Jalilian, A., Waagepetersen, R., 2015. Quasi-likelihood for spatial point pro- cesses. Journal of the Royal Statistical Society. Series B (Statistical methodology) 77, 677. |
Processed_Fused_Spatial_Point_Process_Intensity_Estimation_w.txt | Leininger, T.J., Gelfand, A.E., et al., 2017. Bayesian inference and model assessment for spatial point patterns using posterior predictive samples. Bayesian Analysis 12, 1–30. |
Processed_Fused_Spatial_Point_Process_Intensity_Estimation_w.txt | Lindgren, F., Rue, H., Lindstr¨om, J., 2011. An explicit link between Gaussian fields and Gaussian Markov random fields: the stochastic partial differential equation approach. Journal of the Royal Statistical Society: Series B (Statistical Methodology) 73, 423–498. |
Processed_Fused_Spatial_Point_Process_Intensity_Estimation_w.txt | Moradi, M.M., Cronie, O., Rubak, E., Lachieze-Rey, R., Mateu, J., Baddeley, A., 2019. Resample-smoothing of Voronoi intensity estimators. Statistics and Computing 29, 995– 1010. |
Processed_Fused_Spatial_Point_Process_Intensity_Estimation_w.txt | Moradi, M.M., Rodr´ıguez-Cort´es, F.J., Mateu, J., 2018. On kernel-based intensity es- timation of spatial point patterns on linear networks. Journal of Computational and Graphical Statistics 27, 302–311. |
Processed_Fused_Spatial_Point_Process_Intensity_Estimation_w.txt | Padilla, O.H.M., Sharpnack, J., Scott, J.G., Tibshirani, R.J., 2018. The DFS fused Lasso: linear-time denoising over general graphs. Journal of Machine Learning Research 18, 176–1. |
Processed_Fused_Spatial_Point_Process_Intensity_Estimation_w.txt | Pinto Junior, J.A., Gamerman, D., Paez, M.S., Fonseca Alves, R.H., 2015. Point pattern analysis with spatially varying covariate effects, applied to the study of cerebrovascular deaths. Statistics in Medicine 34, 1214–1226. |
Processed_Fused_Spatial_Point_Process_Intensity_Estimation_w.txt | Rakshit, S., Davies, T., Moradi, M.M., McSwiggan, G., Nair, G., Mateu, J., Baddeley, A., 2019. Fast kernel smoothing of point patterns on a large network using two-dimensional convolution. International Statistical Review 87, 531–556. |
Processed_Fused_Spatial_Point_Process_Intensity_Estimation_w.txt | Shen, L., Young, D., Lo, D., Sun, C., 2009. Local differential quadrature method for 2-D flow and forced-convection problems in irregular domains. Numerical Heat Transfer, Part B: Fundamentals 55, 116–134. |
Processed_Fused_Spatial_Point_Process_Intensity_Estimation_w.txt | Wahlberg, B., Boyd, S., Annergren, M., Wang, Y., 2012. An ADMM algorithm for a class of total variation regularized estimation problems. IFAC Proceedings Volumes 45, 83–88. |
Processed_Effect_of_Incoming_Solar_Particle_Radiations_on_Th.txt | Mars Exospheric Neutral Composition Analyzer (MENCA) of Mars Orbiter Mission (MOM) measures the in-situ neutral upper atmospheric constituents of Mars. Martian lower atmosphere predominated by the presence of CO2 which photo-dissociates into atomic oxygen (O) in higher altitudes much near the exobase. Atomic O plays a significant role in invoking stronger presence of O+ 2 in the Martian ionosphere. Primary photo-dissociative species CO2, crossover its neutral abundance with atomic O in the collisonless hetergenous atmosphere with varying local solar con- ditions. Initial measurements from Neutral Gas and Ion Mass Spectrometer (NGIMS) instrument on Mars Atmosphere and Volatile Evolution (MAVEN) estimated these crossover/transition al- titude wavering between ≈225 km to 240 km during solar maximum conditions with peak solar illuminations. MENCA sampled the neutral atmospheric species, below the exobase upto peri- areion of ≈160 km, under low solar active conditions during June 2018. Observations of partial pressures of CO2 and O in subsequent orbits reveals that solar inputs are crucial in quantifying these crossing points, where [O]/[CO2] remain unity, alongside the influences from temperature. The multi-spacecraft measurements of the direct influences of solar wind charged particle fluxes and velocity on the daily variation of neutral thermospheric/exospheric compositions were ob- served on the local evening hours of Mars and presented. It marks the first-ever direct in-situ observation of interaction between the energetic solar particle radiations on Martian exospheric compositions, potentially contibuting for the steady escape and differing population of atomic [O] in the exosphere. |
Processed_Effect_of_Incoming_Solar_Particle_Radiations_on_Th.txt | on 20 July 1976 at 22.5◦N, 48◦W and 03 September 1976 at 48◦N, 22◦W respectively. These in- situ measurement of atmospheric profile below 200 km confirmed the presence of abundant CO2 supporting earlier Ultra-Violet Spectroscopic (UVS) observation using Mariner 6, 7 and 9 missions (Barth et al., 1971, 1972; Stewart et al., 1972). Alongside CO2, the presence of N2, Ar, and trace measurements of atomic and molecular oxygen (O and O2), CO, N O, Kr, N e and Xe were reported (Nier and McElroy, 1976, 1977; Owen and Biemann, 1976; Owen et al., 1977). The complete set of near surface meteorological data obtained from Viking Landers (1976) to the Curiosity Rover (2012) has been analysed and results have been consolidated in terms of diurnal, seasonal and inter-annual variations of meteorological parameters including dust storms over a span of more than 20 Martian years (Martínez et al., 2017). Though, the Viking missions succesfully identified the major and minor species of Martian atmosphere to surface, it is to be noted that this observational data provides paucity on the study pertaining to seasonal and solar activity related variations of the Martian thermosphere and exosphere. A similar analysis to characterise the thermosphere/exosphere system has not been possible due to the lack of observational data on the atmospheric neutral/ionised gas constituents covering the exospheric region (Bougher et al., 2015a). |
Processed_Effect_of_Incoming_Solar_Particle_Radiations_on_Th.txt | Indian Space Research Organisation (ISRO) has launched its first interplanetary mission to Mars viz. Mars Orbiter Mission (MOM) on 05 November 2013, reached Mars on 24 September 2014 during Martian nighttime. MOM arrived at Mars with its initial primary objective concerning studies on planet’s morphological features, detection of Methane (CH4), estimating the ratio between Deuterium and atomic Hydrogen (D/H), coverage of spatio-temporal profiles of various neutral atmospheric constituents and cartographic events including dust storms, atmospheric clouds and so-on (Kumar and Chauhan, 2014). As of August 2020, the MOM spacecraft remains healthy in performing the extended mission objectives. |
Processed_Effect_of_Incoming_Solar_Particle_Radiations_on_Th.txt | Utilizing the opportunity that Mars approaches Earth closely, for every ≈2 Gregorian years, in collateral to the Mars Orbiter Mission, National Aeronautics and Space Administration (NASA) launched the Mars Atmosphere and Volatile Evolution (MAVEN) mission on 18 November 2013 reaching Mars on 22 September 2014. MAVEN mission dedictated to study the structure, composition and dynamics of the upper atmosphere of Mars (above ≈150 km), also investigated the role of solar wind plasma and habitability aspects of Mars (Jakosky et al., 2015). The ExoMars Trace Gas Orbiter (ExoMars TGO) mission launched on 14 March 2016, as an joint venture between the European Space Agency (ESA) and Roscosmos (Russian Space Agency), inserted intially into a periareion of ≈400 km exospheric orbit at an inclination of 74◦, much similar to MAVEN’s orbit inclination, for observing biologically relevant trace constituents such as Methane (CH4) and potential organic gases (Olsen et al., 2017). |
Processed_Effect_of_Incoming_Solar_Particle_Radiations_on_Th.txt | Based on the observations using MENCA payload between December 2014 and May 2015, analysis to extract spatial and temporal distribution of the atmospheric composition of the thermosphere- exosphere region of Mars have been carried out by Nagaraja et al. (2020). The results obtained from NGIMS-MAVEN data have provided valuable information about the spatial variation of the thermospheric neutral/ion constituents delineating their vertical and latitudinal distribution, and the effect of solar zenith angle (Mahaffy et al., 2015a). MAVEN Deep Dip campaign to sample the sub-solar collisional homosphere and magnetic field structure in April 2015 demonstrated the orbit-to-orbit variation of thermosphere and ionosphere of Mars (Bougher et al., 2015b). Direct measurements of atmospheric neutrals from MENCA-MOM and NGIM-MAVEN has been used in our study to examine the influence of solar activity driven changes on Martian upper atmosphere. The results from both the spacecrafts agrees that the energetic particle fluxes deposited over the exosphere plays a major role in steering the daily variations on the exospheric compositions of Mars. |
Processed_Effect_of_Incoming_Solar_Particle_Radiations_on_Th.txt | Mars Orbiter Mission spacecraft inserted into Martian orbit on 24 September 2014, initially with a periareion of ≈400 km. MOM had a orbital period of ≈72 hours due to its highly eccentric orbit inclined at 150◦ focusing observations over the Martian equator. Comet Siding Spring (C/2013 A1) had a closest approach by Mars on 19 October 2014, within a month of arrival of MOM and MAVEN spacecrafts at Mars. Operations of all spacecrafts were ceased to protective mode to prevent from the meteor shower caused during the passage of the comet (Schneider et al., 2015). MOM occulted behind Mars through orbital manoeuvres, later brought down its periareion altitudes to around 262 km during December 2014, more feasible region to study the Martian exosphere. |
Processed_Effect_of_Incoming_Solar_Particle_Radiations_on_Th.txt | Figure 1: Mars Orbiter Mission’s periareion altitudes between October 2014 and October 2020, as projected by the MOM team at Jet Propulsion Laboratory. The figure forecasts the periareion altitudes reached by MOM immediate after orbital maneuvers due to the closest encounter of Comet Siding Spring (C/2013 A1) by Mars on 19 October 2014 (Helfrich et al., 2015). |
Processed_Effect_of_Incoming_Solar_Particle_Radiations_on_Th.txt | Mars Exospheric Neutral Composition Analyser (MENCA), a dedicated atmospheric suite of MOM, sampled exospheric neutrals near the exobase from December 2014 through May 2015 com- prising of 88 orbits. MENCA measures the total atmospheric pressure and partial pressures of various atmospheric constituents covering the mass range from 1-300 amu, which is further programmable to limit its operation under mass-sweep and trend modes. MENCA consists of a built-in electron impact ionizer operated at ≈70 eV to ionize the atmospheric neutrals, a set of four quadrupole rods and a detector assembly. The detector comprises of Faraday Cup and Channel Electron Multiplier measures the partial pressures of gas constituents with a mass resolution of 1 amu. Part of postively ionized atmospheric neutrals inside the ionization chamber known as the source grid region drags the ions inside the quadrupole mass filter system. Another part of neutrals ionized outside the source grid were collected by the Bayard-Alpert gauge, which is calibrated to measure the total pressure. The detailed description of the instrumentation, calibration factors, complete working mechanism, sensitivity and measurement limitations were given by Bhardwaj et al. (2016). |
Processed_Effect_of_Incoming_Solar_Particle_Radiations_on_Th.txt | Indian Space Science Data Centre (ISSDC) at Bengaluru, India archives the MOM observa- tional data and disseminate to scientific users. This archived data consists of total pressure and partial pressure values in the units of Torr with variable time resolution of ≈12 to 30 s. The data is used for scientific studies after incorporating calibration, correction and normalization factors, and time tagging the processed data with ancillary information such as latitude, longitude, altitude and solar zenith angle (SZA) corresponding to the orbital phase of MENCA-MOM using relevant SPICE/Ephemeris kernels (Nagaraja et al., 2020). The data sets are identified and arranged with respect to different orbit numbers of MOM. However due to the oscillatory nature of periareion height of MOM, good coverage of lower and crucial exospheric altitudes has been obtained only for a small number of orbits (in spite of some orbit lowering excercises) during 5-13, June 2018. |
Processed_Effect_of_Incoming_Solar_Particle_Radiations_on_Th.txt | Figure 1 shows the periareion altitude forecast of MOM spacecraft from October 2014 through October 2020. It is clear that after 2014-2015 there has been only a very short period during 2018- 2019 when useful data could be collected closer to the exobase and above. The predicted minimum periareion altitude reached by MOM will be 112.9 km on 08 July 2029 and can traverse above 100 km (Helfrich et al., 2015). The analysis was carried out for the additional MENCA data available during June 2018 and results obtaiend are discussed in detail. |
Processed_Effect_of_Incoming_Solar_Particle_Radiations_on_Th.txt | The Mars Atmosphere and Volatile EvolutioN (MAVEN) mission inserted into a highly elliptical orbit, focuses on the search for past history of Martian atmosphere and to understand its present cli- mate. The Neutral Gas and Ion Mass Spectrometer (NGIMS) instrument of the MAVEN spacecraft studies the structure and composition of the upper neutral atmosphere of planet Mars, measures iso- tope ratios, and measures thermal and supra-thermal ions (Mahaffy et al., 2015b). NGIMS measures in a mass range of 2-150 amu. NGIMS science operation starts below 500 km to periapsis (inbound) and periapsis to 500 km (outbound) during each orbit lasting for ≈600 s for each leg, with spatial resolution of 1 km. These observation covers both exosphere as well as thermosphere of Martian atmosphere that peers through the exobase at ≈200 km, falls within NGIMS limits. The level-2 (version 08 and revision 01) data-sets of NGIMS-MAVEN has been retrieved from MAVEN Science Data Center at Laboratory for Atmosphere and Space Physics (LASP), during the period June 2018 were used for this study. NGIMS-MAVEN observation has been been chosen from 03 June 2018 through 15 June 2018 in parallel to MENCA-MOM observations. |
Processed_Effect_of_Incoming_Solar_Particle_Radiations_on_Th.txt | Figure 2 shows the analysis of atomic H ion current data on 5 June 2018 covering the time period 10 to 13 h UTC having both temporal and spatial variations. The time period of closest approach i.e. near periareion, of MOM towards Mars has been shown along with altitude coverage. The top panel shows altitude profile derived from SPICE kernels provided along with the observational data and ion current of raw data. The middle panel shows the ion current of atomic H with background corrected at ≈500 km. The bottom panel shows the calibrated partial pressure of atomic H for the actual time sampled values with background correction. The pressure values have been converted to Torr units by using the calibration procedure prescribed in Bhardwaj et al. (2016). The termainal panel plot clearly shows that the maximum pressure values occur much near to the periareion altitudes though it can vary from orbit-to-orbit. The figure also shows a small height interval near pariareion with increasing pressure much above the noise levels of the instrument. This is the region of useful data to be used for further analysis. |
Processed_Effect_of_Incoming_Solar_Particle_Radiations_on_Th.txt | Figure 2: Spatio-temporal variation of raw ion current and calibrated partial pressure of atomic Hydrogen (H) observed from MENCA-MOM on 05 June 2018. |
Processed_Effect_of_Incoming_Solar_Particle_Radiations_on_Th.txt | of change in solar zenith angle may be insignificant. However, for long-term variabilty, the seasonal changes in solar zenith angles need to be corrected for studying any effect of solar activity related variations in exospheric partial pressures. |
Processed_Effect_of_Incoming_Solar_Particle_Radiations_on_Th.txt | Figure 3: Variation of atmospheric CO2 and O abundances observed on 05, 08 and 13 June 2018 by MENCA-MOM. Corresponding variation in local solar time (LST) in hours and solar zenith angle (SZA) in degrees has been shown. |
Processed_Effect_of_Incoming_Solar_Particle_Radiations_on_Th.txt | Figure 3 shows details of the variation of atmospheric CO2 and O partial pressures near the periareion region between ≈160 km to 500 km, subtracting the background noise at 500 km. This covers the Martian upper thermosphere, its exobase and lower exosphere regions for ascertaining relative effects of gaseous transition and escape. While the pattern of variation of partial pressure is similar on 05 June 2018 and 08 June 2018, there is a shift of about few minutes in the peak density observed on 13 June 2018, where maximum of pressure has not occurred at the lowest height of observation in both CO2 and atomic O. This anomaly needs to be examined by studying the condition of atmospheric dynamics. It is also observed that the rate of change of pressure values for inbound and outbound trajectories are generally symmetric. It can be seen that the SZA variation is from the nighttime and moving towards the evening hours of Mars. This transition does not affect the density profiles appreciably for short duration of observations. |
Processed_Effect_of_Incoming_Solar_Particle_Radiations_on_Th.txt | Figure 4: Relative abundance of major atmospheric species measured from MENCA-MOM on 05 June 2018. Increased N2 + CO/amu28 has contributions from CO2/amu44. Variation of water vapour (H2O) is mainly due to the outgassing of MOM spacecraft and eliminated during background correction. |
Processed_Effect_of_Incoming_Solar_Particle_Radiations_on_Th.txt | Figure 4 shows the relative partial pressure of major atmospheric constituents. It is noted that the CO2 shows the maximum density values. According to the known pattern of variations, the other constituents have relatively lower pressures compared to CO2. In addition, the water vapor density, product of outgassing, in MOM has reduced considerably compared to the values observed during December 2014. However, the temporal profile of H2O follows similar pattern of other species due to reduction in further degassing. The escaping remanant water vapour may be maintained through the hydrostatic equilibrium. |
Processed_Effect_of_Incoming_Solar_Particle_Radiations_on_Th.txt | Figure 5: Inbound (top) and Outbound (bottom) altitude profiles of CO2, O, CO + N2 and Ar on 05, 08 and 13 June 2018 from MENCA of Mars Orbiter Mission. |
Processed_Effect_of_Incoming_Solar_Particle_Radiations_on_Th.txt | Figure 6: Vertical profiles of upper atmospheric neutral gas constituents for four days viz., 03, 04, 08 and 13 June 2018. The corresponding variations of SZA during the period of each profiles are also shown. |
Processed_Effect_of_Incoming_Solar_Particle_Radiations_on_Th.txt | the track/traverse. The daily average values of parameters are computed from the available data during the 24 h of each Earth day. Figure 6 shows the results of the altitude profiles of a few selected Martian atmospheric constituents in the units of number density per cubic centimeter for four days of June 2018. |
Processed_Effect_of_Incoming_Solar_Particle_Radiations_on_Th.txt | Figure 7: Vertical profiles of daily mean conentrations of CO2 and O densities on 03, 04, 08 and 13 June 2018. [CO2] steadily increase with decline in [O] during this period as observed from NGIMS-MAVEN. |
Processed_Effect_of_Incoming_Solar_Particle_Radiations_on_Th.txt | Figure 8: Ion and electron fluxes signifying the solar-wind variations as measured from the SEP instrument (forward and rear-view) of MAVEN during 03-13, June 2018. Steady reduction in the incoming solar flux, immediate after 05 June 2018, shows significant contribution to the decrease of the photodissociative product, atomic oxygen (O). |
Processed_Effect_of_Incoming_Solar_Particle_Radiations_on_Th.txt | from the solar wind particle velocity and flux variations. Figure 7 shows the mean daily profiles of CO2 and O. There exists a striking relationship between the decrease in O and increase in CO2 densities during the period from 03 June 2018 to 13 June 2018. |
Processed_Effect_of_Incoming_Solar_Particle_Radiations_on_Th.txt | The Solar Energetic Particle (SEP) instrument onboard MAVEN measured the solar-wind electron and ion fluxes over various energy levels during the same period. The SEP sensors are positioned on two corners to ensure that the field of view (FOV) adequately cover the canonical Parker spiral direction around which solar energetic particle distributions are normally centered. SEP provides measurements with ≈1 h time resolution. The data of the electron and ion fluxes have been analysed from 03-13, June 2018 and are shown in Figure 8. The energy fluxes of both ion and electron were reduced from 09 June 2018. This reduction in solar heating of the upper atmosphere resulted in the fall of photoionization of CO2. Further, the dissociation of CO2 under this condition has limited the prodution of O, and the same is reflected. |
Processed_Effect_of_Incoming_Solar_Particle_Radiations_on_Th.txt | The upper atmosphere of Mars has been observed from MENCA-MOM and NGIMS-MAVEN instru- ments during June 2018. Interesting, the solar wind particle fluxes show a steady decrease between 09 June 2018 and 13 June 2018, providing the possible clues of charged particle interactions with Martian atmospheric constituents playing a major role in the day-to-day variation of O and CO2 concentrations. This study shows the increase in concentration of O and decrease in CO2 during 03-08, June 2018 indicates an enhanced production of O from the dissociation of CO2 corresponding to higher energy and fluxes of solar wind particles. Hence, it marks the first-ever observation that in absence of a global magnetic field of Mars, the direct interaction of solar wind charged parti- cles affect the daily variation of thermosphere/exosphere gaseous concentrations, contributing to the steady escape of O with enhanced solar activity. |
Processed_Discrete_product_systems_and_twisted_crossed_produ.txt | Abstract. A product system E over a semigroup P is a family of Hilbert spaces {Es : s ∈ P } together with multiplications Es×Et → Est. We view E as a unitary-valued cocycle on P , and consider twisted crossed products A⋊β,EP involving E and an action β of P by en- domorphisms of a C ∗ -algebra A. When P is quasi-lattice ordered in the sense of Nica, we isolate a class of covariant representations of E, and consider a twisted crossed product BP ⋊τ,EP which is universal for covariant representations of E when E has finite-dimensional fibres, and in general is slightly larger. In particular, when P = N and dim E1 = ∞, our algebra BN ⋊τ,E N is a new infinite analogue of the Toeplitz-Cuntz algebras T On. Our main theorem is a characterisation of the faithful representations of BP ⋊τ,EP . |
Processed_Discrete_product_systems_and_twisted_crossed_produ.txt | Crossed products of C ∗-algebras by semigroups of endomorphisms have been profitably used to model Toeplitz algebras [2, 1, 13] and the Hecke algebras arising in the Bost-Connes analysis of phase transitions in number theory [14, 3, 11]. There are two main ways of studying such a crossed product. First, one can try to embed it as a corner in a crossed product by an automorphic action of an enveloping group, and then apply the established theory. The algebra on which the group acts is typically a direct limit, and the success of this approach depends on being able to recognise the direct limit and the action on it [7, 23, 17]. Or, second, one can use the techniques developed in [5, 2, 13] to deal directly with the semigroup crossed product and its representation theory. Here the goal is a characterisation of the faithful representations of the crossed product, and such characterisations have given important information about a wide range of semigroup crossed products [2, 13, 14, 3]. |
Processed_Discrete_product_systems_and_twisted_crossed_produ.txt | Date: November 19, 2018. 1991 Mathematics Subject Classification. Primary 46L55. This research was supported by the Australian Research Council. It was completed while the second author was visiting Dartmouth College; he thanks Dana Williams and his colleagues for their warm hospitality and their excellent facilities. |
Processed_Discrete_product_systems_and_twisted_crossed_produ.txt | In this paper we investigate a phenomenon which arises only for semi- groups: crossed products twisted by unitary cocycles acting on Hilbert spaces of varying dimension. Such cocycles were introduced by Arveson under the name of product systems [4]. The idea is to associate to each ele- ment s of the semigroup S a Hilbert space Es, and then the cocycle describes a multiplication from Es × Et to Est; a scalar cocycle ω : S × S → T deter- mines such a system by taking Es = C for all s and using (w, z) 7→ ω(s, t)wz as the product from Es × Et to Est. Because dim Est = dim Es × dim Et, product systems with fibres of dimension other than 1 cannot exist on groups (at least in a naive sense), so the possibility of twisting crossed products by product systems is appropriate only for actions of semigroups. It is not an entirely new idea: the crossed products of multiplicity n of Stacey [23] are twisted crossed products by actions of the semigroup N in which the product system E has dim E1 = n. |
Processed_Discrete_product_systems_and_twisted_crossed_produ.txt | Of the various kinds of semigroups studied in the literature, we have cho- sen to work with the quasi-lattice ordered semigroups of Nica [19]; these include the totally ordered groups considered in [15, 2, 10], the direct sums Nk, and the free products considered in [13]. For a product system E over such a semigroup P , one can define a natural notion of covariant represen- tation generalising that of [19, 13]: loosely speaking, a representation φ of E is a family of isometric maps φs : Es → B(H) such that each φs(v) is an isometry and φst(uv) is the composition of the operators φs(u) and φt(v), and φ is covariant if the projections on the ranges φs(Es) are aligned in a manner compatible with the ordering on P . The motivating example is the trivial product system on N2, where the representations are given by two commuting isometries and the covariant representations by two ∗-commuting isometries. |
Processed_Discrete_product_systems_and_twisted_crossed_produ.txt | The main results of [19] and [13] concern the C ∗-algebra, here denoted C ∗ cov(P ), which is universal for covariant isometric representations of the quasi-lattice ordered group (G, P ). In [13], C ∗ cov(P ) is viewed as a semigroup crossed product BP ⋊τ P , where BP is the C ∗-subalgebra of ℓ∞(P ) spanned by the characteristic functions 1x := χxP , and τt(1x) = 1tx. Here we aim to view the universal C ∗-algebra C ∗ cov(P, E) for covariant representations of E as a twisted crossed product BP ⋊τ,EP , and use techniques like those of [13] to characterise their faithful representations. However, carrying out this program has raised some intriguing issues. |
Processed_Discrete_product_systems_and_twisted_crossed_produ.txt | representations of BP ⋊τ,EP rather than C ∗ cov(P, E), and thus achieves our goal only for systems with finite-dimensional fibres. We plan to return to the topic of systems with infinite-dimensional fibres in a sequel. |
Processed_Discrete_product_systems_and_twisted_crossed_produ.txt | We have organised our work as follows. We begin with an introductory section on product systems and their representations, giving a variety of examples and constructions. General twisted crossed products are discussed only in §2: as in [13], we are mainy interested in the specific crossed products BP ⋊τ,EP which capture the covariance condition on representations of E. The covariance condition itself is modelled on that of Nica, and only makes sense for product systems on quasi-lattice ordered semigroups. In §3 we discuss it and its connection with covariant representations of the system (BP , P, τ, E). We can then prove that C ∗ cov(P, E) embeds naturally in the semigroup crossed product BP ⋊τ,EP (Theorem 4.3). |
Processed_Discrete_product_systems_and_twisted_crossed_produ.txt | Our main theorem is our characterisation of faithful representations of BP ⋊τ,EP . There are two main steps. First, under an amenability hypothe- sis, we follow the procedure pioneered by Cuntz, which reduces the problem to proving an estimate concerning the deletion of off-diagonal terms. The details are necessarily different, but the general plan of [13, §3] carries over under a spanning hypothesis on the product system which holds in the in- teresting examples. Second, we have to verify the amenability hypothesis in a reasonable number of situations. It is automatic, for example, if the enveloping group of P is amenable, or if P is a free product of such semi- groups and the product system satisfies a modest-looking spanning condi- tion. Both the spanning conditions we have mentioned are satisfied if E has finite-dimensional fibres, so our main theorem applies to all such product systems on Nk or on free products of subsemigroups of amenable groups. |
Processed_Discrete_product_systems_and_twisted_crossed_produ.txt | Definition 1.1. Suppose P is a semigroup with identity and p : E → P is a family of nontrivial complex Hilbert spaces whose fibre over the identity is one-dimensional. Write Et for the fibre p−1(t) over t ∈ P . We say that E is a (discrete) product system over P if E is a semigroup, p is a semigroup homomorphism, and for each s, t ∈ P the map (u, v) ∈ Es × Et 7→ uv ∈ Est extends to a unitary isomorphism Us,t of Es ⊗ Et onto Est. |
Processed_Discrete_product_systems_and_twisted_crossed_produ.txt | for r, s, t ∈ P . Thus product systems over P can be viewed as unitary 2-cocycles acting on a varying but coherent system of Hilbert spaces. |
Processed_Discrete_product_systems_and_twisted_crossed_produ.txt | Lemma 1.3. Suppose E is a product system over a semigroup P with iden- tity e. Then E has an identity Ω such that p(Ω) = e and kΩk = 1. |
Processed_Discrete_product_systems_and_twisted_crossed_produ.txt | so zx = λx. Similarly we have xz = λx, and thus Ω = λz is an identity for E. We have p(Ω) = e because z ∈ Ee, and kΩk = 1 because z is a unit vector. |
Processed_Discrete_product_systems_and_twisted_crossed_produ.txt | Examples 1.4. (E1) The trivial product system over P is the trivial bundle P × C with multiplication given by (s, w)(t, z) = (st, wz). |
Processed_Discrete_product_systems_and_twisted_crossed_produ.txt | alternatively, one might say µ is a 2-cocycle on P with values in T. Let Eµ = E, pµ = p, and define multiplication by (u, v) 7→ µ(p(u), p(v))uv. Then Eµ is a product system over P ; we say that Eµ is E twisted by µ. |
Processed_Discrete_product_systems_and_twisted_crossed_produ.txt | If ν is another multiplier on P , then Eµ is isomorphic to Eν iff [µ] = [ν] as elements of the second cohomology group H 2(P, T); the automorphism group of Eµ is Hom(P, T). |
Processed_Discrete_product_systems_and_twisted_crossed_produ.txt | w1 ⊗ · · · ⊗ wnv1 ⊗ · · · ⊗ vm if λn = µ1 w1 ⊗ · · · ⊗ wn ⊗ v1 ⊗ · · · ⊗ vm otherwise. |
Processed_Discrete_product_systems_and_twisted_crossed_produ.txt | Proposition 1.5. Suppose E and F are product systems over N. Then E and F are isomorphic iff E1 ∼= F1. |
Processed_Discrete_product_systems_and_twisted_crossed_produ.txt | Corollary 1.6. For each d ∈ {1, 2, . . . , ℵ0} there is, up to isomorphism, a unique product system Ed over N whose fibre over 1 is d-dimensional. |
Processed_Discrete_product_systems_and_twisted_crossed_produ.txt | Remarks 1.8. (1) Condition (2) implies that every operator in the range of φ is a multiple of an isometry, and that φ is linear on the fibres of p; see [4, p.8]. It also implies that φ is isometric, hence injective; thus each vector space φ(Et) has a Hilbert space structure in which the inner product is given by hS, T iI = T ∗S, and the corresponding Hilbert space norm on φ(Et) agrees with the operator norm. |
Subsets and Splits