Filename
stringlengths 22
64
| Paragraph
stringlengths 8
5.57k
|
|---|---|
Processed_A_generalised_multi-factor_deep_learning_electrici.txt | Based on the weekly periodicity in load, calendar effects have been applied to some load forecasts. However, little research has compared the different ways of using calen- dar labels and their impact on load forecasts. [21] consid- ered working days and non-working days as non-time se- ries characteristics to assist in a Time Convolutional Net- work (TCN) and GRU combined model to forecast load for an industrial building. [22] generated six binary fea- tures and a zero feature to encode the calendar index, rep- resenting Monday to Sunday, respectively. - Different ways of applying calendar labels on load fore- cast are compared to maximise the forecast accuracy ad- vancement attributed to calendar label employment. |
Processed_A_generalised_multi-factor_deep_learning_electrici.txt | models, such as temperature, humidity, wind speed, pre- cipitation, etc. In [23], the daily high & low tempera- tures and the daily high & low dew temperatures were set as the input features to help forecast the daily electricity consumption of a residential building in Shanghai, China. [22] used Pearson Correlation Coefficient to assess the re- lationship between climate variables and load, indicating that the maximum, minimum, and average temperature, air pressure and water vapour pressure are strongly cor- related to load in a subtropical monsoon climate region. As one of the most significant climate variables in load forecast, most existing studies directly use the daily peak, mean, or instantaneous temperature [17], [23], [24], [25]. However, the relationship between load and temperature varies from region to region. There may be better vari- ables to help with the load forecast than the peak and mean temperature. - The inner relationship between energy use and different types of lead temperatures is investigated, significantly fa- cilitating load forecast performance with the novel flexi- ble temperature conditions designed in our paper. |
Processed_A_generalised_multi-factor_deep_learning_electrici.txt | Extreme weather events like wildfires raise the uncertainty and complexity of load patterns. Hence, the load behaviour analysis and forecast in regions encountering frequent ex- treme weather events are more worth studying than in the normal area [26]. - The wildfire seasons in 2015-2020 are chosen as the tem- poral scale of our study, filling the gap of load forecast under low-probability high-risk climate events. |
Processed_A_generalised_multi-factor_deep_learning_electrici.txt | While load forecasts at the DN level have been applied in some IEEE standardised networks, e.g., IEEE 33 bus model [27], [28], [29], the DN load forecast research based on real grids are relatively fewer conducted. - Our paper developed the load forecast model based on authentic DN operations under extreme climates. Com- monalities and differences in forecast performances in 8 real DNs are discussed and interpreted based on the re- gional multi-feature visualisation analysis. |
Processed_A_generalised_multi-factor_deep_learning_electrici.txt | Few studies have delved into the robustness of the load forecast model to extreme weather events. Research con- sidering climate robustness is indispensable to propel progress in mitigating and adapting to Climate Change. - Sensitivity and robustness analysis are conducted to prove the model reliability to data sets with large-scale climate variability. |
Processed_A_generalised_multi-factor_deep_learning_electrici.txt | Most existing research evaluated forecast performance solely based on technical metrics. However, the economic bene- fits calculated based on real network case studies can pro- vide more intuitive and comprehensive guidance for sys- tem decision-makers and the authorities. - A cost-benefit analysis is discussed to demonstrate the feasibility and sustainability of our model. |
Processed_A_generalised_multi-factor_deep_learning_electrici.txt | and main model structure. The forecast performance evalu- ation metrics, correlation analysis methods to form flexible temperature conditions, and techniques to visualise the data dispersion level are then discussed in detail. Section 3 in- troduces the background of our Case Study. Data sources and processing methods are explained for the original en- ergy and climate input variables. Based on the forecast per- formance advancement, the generalised model is developed through four steps in Section 4, taking into consideration of input data structures, calendar labels and the flexible temper- ature condition assistance. In addition, robustness analysis and cost-benefit analysis are also discussed to prove model stability during extreme weather events and significant eco- nomic benefits. Section 5 summarises the key findings, ad- vising potential additional features that could strengthen the load forecast model in the future. |
Processed_A_generalised_multi-factor_deep_learning_electrici.txt | In this Section, the main steps to set up the multi-factor load forecast model are explained in detail. Then, the evalu- ation metrics are introduced to quantify the forecast perfor- mances. In addition, correlation analysis for energy demand and temperature are studied to explore how the climate fac- tors affect the energy consumption behaviour so as to ascer- tain the optimal method that climate factors may be used to improve load forecast accuracy. Finally, methods to assess and visualise dispersion level of energy and climate data sets are discussed to help with the model generalisation step. |
Processed_A_generalised_multi-factor_deep_learning_electrici.txt | and weight matrix can be found in [8] to further understand the LSTM cell architecture and theory. As displayed in Figure 2b, GRU is a variant of LSTM with only two gates - the reset gate and the update gate. In 2014, GRU was proposed by [32]. The cell and hidden states are combined as one state ℎ𝑡 . The reset gate determines how much past information should be forgotten in the current time step. The update gate generates the current hidden state and the candidate ℎ𝑡 hidden state. Detailed equations and weight matrix forma- tion can be studied in [8] to further understand the GRU cell memory principle. While LSTM and GRU both belong to RNN, they have ad- vantages over each other in certain cases. GRU was invented later than LSTM, and the simplified gate structure uses less memory and is faster than LSTM. Also, it is proven that GRU structure is more immune to gradient exploding and vanishing issues [? ]. But for data sets with long and high- complexity sequences, LSTM has shown a better forecast performance [34]. Thus, both LSTM and GRU structures are tested and compared in our paper. The main structure of the LSTM and GRU deep learning models are displayed in Figure 3. First, the input matrix is put into the multiple LSTM/GRU layers, where the layer and neuron numbers can be set differently in various con- ditions. In general, the deterministic number of layers and nodes cannot be analytically calculated. There is a trade- off between forecast accuracy and computational efficiency: each data set has its optimal architecture. The appropriate perceptron structure can be explored by referring to existing deep learning models and adjusting with experimentation. Referring to the system in [35] and [36], three LSTM/GRU layers with either 32 or 64 neurons are used to predict load with good accuracy and acceptable computational complex- ity in our paper. Then, multiple dense layers are connected to further process the forecast data. Dense layers are full con- nection layers with adjustable neurons. In our model, two dense layers are connected, with 16 and 1 neurons, respec- tively. Finally, the forecast load is output as a sequence. The forecast performance is evaluated on the testing data set. |
Processed_A_generalised_multi-factor_deep_learning_electrici.txt | In our paper, Mean Squared Error (MSE) is used to cal- culate the loss at each epoch to adjust the deep learning model training direction [37]. Also, MSE is used as one metric to evaluate the forecast performance in the Results and discus- sion. |
Processed_A_generalised_multi-factor_deep_learning_electrici.txt | where: y is the actual value, ̂𝑦 is the predicted value, and n is the number of data points. Mean Absolute Percentage Error (MAPE) is another com- monly used metric to assess forecast accuracy [38]. While MSE is a scale-based metric, the unit of MAPE is per cent, making it comparable among data sets with different units and scales. |
Processed_A_generalised_multi-factor_deep_learning_electrici.txt | where: 𝜎2 is the error variance, X is the average half-hourly error, 𝜇 is the mean of the average half-hourly error over a day, and N is the total sampling points for a day. The stability of forecast accuracy within a day is worth studying to figure out the weak point of the forecast. The daily error variance is calculated to measure the model stability: the smaller the variance, the better robustness of the forecast model [39]. And the total forecast model performance can be improved by assessing the hard-to-predict period in a day. |
Processed_A_generalised_multi-factor_deep_learning_electrici.txt | Correlation analysis is conducted to find the optimal tem- perature relationship that can improve the load forecast ac- curacy. For each DN, the Pearson Correlation Coefficient of the half-hourly load and the half-hourly instantaneous, max- imum and average temperatures for lead time from 0 to 24 hours ahead are assessed to find the most strongly-correlated conditions. |
Processed_A_generalised_multi-factor_deep_learning_electrici.txt | and the 2𝑛𝑑 order 𝑟2 value are computed as other methods to study the relationship between load and leading tempera- tures. To assess the dispersion level of temperature and load data sets at different DNs, each data set’s Coefficient of Quartile Variation (CQV) and Standard Deviation (SD) are calculated and displayed in the data set visualisation plots in the Results and discussion. As the CQV is a unitless metric, it is selected to measure the spread of data sets at different scales [41]. |
Processed_A_generalised_multi-factor_deep_learning_electrici.txt | where 𝑄𝑖 is the in- terquartile range. In the data dispersion level visualisation plots, the 5𝑡ℎ, 25𝑡ℎ, 50𝑡ℎ, 75𝑡ℎ, and 95𝑡ℎ percentile lines are plotted to assist in analysing the daily data distribution. |
Processed_A_generalised_multi-factor_deep_learning_electrici.txt | This section describes the main steps of designing a gen- eralised multi-factor deep learning load forecast model. Crit- ical steps are explained in the context of the State of Victoria case study. Relevant data sources and processing methods for the load and temperature data used in our paper are pre- sented in this section. |
Processed_A_generalised_multi-factor_deep_learning_electrici.txt | In Step 1, the input sequence length is varied to find a proper input data structure with acceptable low error and computational burden. Then, the performances of LSTM- based and GRU-based forecast models are compared, and the better one is selected as the main DL structure in the fol- lowing steps. In Step 2, the effect of applying different cal- endar classification methods to input data is studied to im- prove the load forecast accuracy. In Step 3, correlation anal- ysis is conducted to learn the relationship between tempera- ture and load and to find the strongly-correlated temperature conditions. Some selected leading temperature conditions with high correlation coefficients are input into our model to improve the forecast performance as a novel method using flexible temperature conditions. In Step 4, we further apply the model to the other 7 DNs in Victoria, Australia, to study the regional difference and generalise our model. The temperature tends to correlate stronger with load, play- ing a more essential role in load forecast under extreme weather conditions. Thus, regions with frequent extreme climate sea- sons are preferred to be the research region. Victoria, Aus- tralia, is selected as the research region for our paper due to its annual extreme wildfire season and data availability. In Step 3, one of the Victoria DNs, Horsham, is chosen to anal- yse the potential temperature function in improving the DN load forecast. As Horsham is located in the mid-latitude of Victoria State and has good data availability, we start with this location before generalising to other DNs in Victoria. |
Processed_A_generalised_multi-factor_deep_learning_electrici.txt | POWERCOR is one of the five main energy distributors in Victoria, Australia, covering western and central Victo- ria. As the mean summer wildfire risk (Forest Fire Danger Index) is generally higher in the western than in the eastern, POWERCOR coverage is selected as the research area in our paper [42]. The energy consumption data set is downloaded through the Powercor Zone Substation Reports [43]. The load at different substations is recorded every 30 minutes in the POWERCOR database. Temperature data is obtained from the Copernicus Climate data store. The 2m temperature netCDF data set is accessi- ble from [44]. The original temperature data has an hourly temporal resolution and 0.25°×0.25°spatial resolution. As the temporal resolution of load data is half-hourly, the tem- perature data set is temporally interpolated to half-hourly us- ing Climate Data Operator (CDO) [45] to keep it consistent with the load data set. In order to position the DN more pre- cisely, the temperature data set is spatially interpolated into 0.01°×0.01°resolution, again using CDO. The temperature data of each DN is then extracted according to the city’s ge- ographical location at the spacial accuracy of 0.01°×0.01°. The consistent temporal resolution of input variables and a relatively high spatial resolution to locate the DN together contribute to a more accurate and smoother forecast process. |
Processed_A_generalised_multi-factor_deep_learning_electrici.txt | In this Section, we discuss the forecast performance en- hancement at each step, determining the best design to de- velop the final generalised model. In Step 1, the effect of using different input data lengths is studied, and the forecast performances based on LSTM and GRU structures are com- pared for our case study. Various classifications of calen- dar labels are applied to the forecast model input to improve the sequential forecast accuracy in Step 2. And then, the relationship between temperature and load behaviour is dis- cussed to explore flexible temperature conditions that may enhance the load forecast performance in Step 3. The oper- ating cases with selected temperature conditions are tested over five wildfire seasons. The temperature condition that best contributes to the load forecast is determined based on three primary performance metrics. In Step 4, our forecast model is applied to 8 DNs in Victoria, forming a generalised and standardised forecast model. |
Processed_A_generalised_multi-factor_deep_learning_electrici.txt | with the lowest error are marked with a triangle sign. MSE and MAPE decrease as the input length increases and reach a stable level when the input is around 16 hours for both LSTM and GRU cases. It is demonstrated that the GRU-based model outperforms the LSTM-based model on MSE and MAPE by 8.18% and 22.81%, respectively. So, we primarily develop the model based on GRU structure and keep the data input length as pre-16-hour (32 sampling points) in the following steps. In later steps, the LSTM-based model is also tested as a com- parison with results shown in the Appendix. |
Processed_A_generalised_multi-factor_deep_learning_electrici.txt | The second step studies the effect of considering calen- dar factors in our model input. In this step, three operat- ing conditions are tested: no-day-type-label, three-day-type- label, and eight-day-type-label. For the three-day type label conditions, the day types are classified into weekday, week- end, and holiday. The eight-day-type-label conditions are classified into Monday to Sunday, plus a holiday label. As a comparison group, the no-day-type-label condition is tested with the original energy consumption data as the only input. In Figure 5a, the real power versus three forecast powers using the GRU-based model during the testing period (the wildfire seasons in 2018-2020) are visualised and compared. The three forecast power are strongly correlated with the real power. Among all three cases, the two with calendar labels perform better than the one without day-type tags on the cor- relation coefficient, especially at energy consumption peaks. Quantitatively, the MAPE and the error Kernel Density Esti- mate (KDE) plots of the three forecast cases are displayed in Figure 5b and Figure 5c. The MAPE of the eight-day-type- label case is 4.23% lower than not using the day-type-label and 1.01% lower than the three-day-type-label case. In the KDE distribution and cumulative plots, it is demonstrated that the majority of error points for the eight-day-type-label case gather at a lower value than the other two cases. |
Processed_Thermodynamics_in_2+1_flavor_QCD_with_improved_Wil.txt | c(cid:13) Copyright owned by the author(s) under the terms of the Creative Commons Attribution-NonCommercial-ShareAlike Licence. |
Processed_Thermodynamics_in_2+1_flavor_QCD_with_improved_Wil.txt | Recent progress of computer algorithms and high performance computers enabled us to per- form realistic simulations of 2 + 1 flavor QCD with the physical quark masses [1]. Since the equa- tion of state (EOS) is one of the most expensive calculations in lattice QCD, realistic simulations for the EOS has been performed only with staggered-type quarks [2, 3, 4, 5], whose theoretical basis such as locality and universality are, however, not well established. Therefore, to check the validity of these results it is important to compare with those obtained using theoretically sound lattice quarks, such as the Wilson-type quarks. To this end we are pushing forward a project to study the thermodynamics of 2+1 flavor QCD, using a nonperturbatively improved Wilson quark action coupled to a RG-improved Iwasaki gauge action [6]. |
Processed_Thermodynamics_in_2+1_flavor_QCD_with_improved_Wil.txt | In this report we present the results of EOS, Polyakov loop, and chiral condensate in 2 + 1 flavor QCD using the fixed-scale approach. Calculation of the beta-functions is updated from the previous reports [9, 10]. Chiral condensate with Wilson-type quarks is known to have a severe di- vergence due to the explicit chiral symmetry breaking of the Wilson term [11, 12]. We renormalize the Polyakov loop and chiral condensate taking the advantage of the fixed-scale approach that the renormalization is common to all temperatures. |
Processed_Thermodynamics_in_2+1_flavor_QCD_with_improved_Wil.txt | The lattice set-up is given in the next section. Details of beta-functions and results of EOS are presented in Sec.3 and 4. The Polyakov loop and the chiral condensate are discussed in Sec.5 and 6. The report is summarized in the last section. Part of the results including the details of the EOS calculation was published recently in [8]. |
Processed_Thermodynamics_in_2+1_flavor_QCD_with_improved_Wil.txt | where cSW is non-perturbatively determined as a function of b [14]. Among their simulation points, we choose b = 2.05, k ud = 0.1356 and k s = 0.1351, which is the smallest lattice spacing [a ≃ 0.07 fm with a scale set by r0 = 0.5 fm] and the lightest u and d quark masses [mp /mr = 0.6337(38)]. The s quark mass corresponds to mh ss/mf = 0.7377(28). The lattice size is 283 × 56. |
Processed_Thermodynamics_in_2+1_flavor_QCD_with_improved_Wil.txt | Using the same coupling parameters as the zero-temperature simulation, we have generated finite temperature configurations on 323 × Nt lattices with Nt = 16, 14, · · ·, 4 [8]. Our range of Nt corresponds to the range T = 174–696 MeV. The pseudo-critical point is expected to be Nt ≈ 14. |
Processed_Thermodynamics_in_2+1_flavor_QCD_with_improved_Wil.txt | To evaluate the trace anomaly, we need the information of beta functions a(db /da) and a(dk f /da) ( f = ud and s). In this study, we define the lines of constant physics (LCP’s) by mp /mr and mh ss/mf at T = 0. The beta functions are determined nonperturbatively through the coupling parameter dependence of zero-temperature observables on a LCP. We use the data of amr , mp /mr and mh ss/mf at 30 simulation points of the CP-PACS+JLQCD zero-temperature configurations [13] to extract the beta functions. |
Processed_Thermodynamics_in_2+1_flavor_QCD_with_improved_Wil.txt | We have first tried to evaluate the beta functions by ‘the inverse matrix method’, i.e., fit the data of amr , mp /mr and mh ss/mf as functions of the coupling parameters b , k ud and k s, and then invert the matrix of the slopes to obtain the beta functions [9]. It turned out that a high statistics and many data points are required to suppress the errors in a(dk f /da) because these beta functions have much smaller magnitude than a(db /da) and thus the error in the latter contaminates those of the formers through the matrix inversion procedure. |
Processed_Thermodynamics_in_2+1_flavor_QCD_with_improved_Wil.txt | Figure 2: Trace anomaly (e − 3p)/T 4, energy density e /T 4, and pressure 3p/T 4 in 2+1 flavor QCD. The thin and thick vertical bars represent statistic and systematic errors, respectively. The curves are drawn by the Akima spline interpolation. |
Processed_Thermodynamics_in_2+1_flavor_QCD_with_improved_Wil.txt | To avoid the matrix inversion, we thus adopt ‘the direct method’, in which we fit the coupling parameters, b , k ud and k s as functions of the observables amr , mp /mr and mh ss/mf . Consulting the overall quality of the fits, we have refined the fit function and the fit range from our previous report [10]. We now fit with the third order polynomial [8]. |
Processed_Thermodynamics_in_2+1_flavor_QCD_with_improved_Wil.txt | at our simulation point, where the first brackets are for statistic errors, and the second brackets are for systematic errors estimated by the variation of the scale setting. |
Processed_Thermodynamics_in_2+1_flavor_QCD_with_improved_Wil.txt | Using the results of the beta functions, we calculate the trace anomaly (e − 3p)/T 4. See Ref.[8] for the details of the calculation. In Fig. 2, we show the results of the trace anomaly by a thick red curve. |
Processed_Thermodynamics_in_2+1_flavor_QCD_with_improved_Wil.txt | Carrying out the T -integration of the trace anomaly according to Eq. (1.1), we obtain the pressure p shown in Fig. 2. The energy density e is then calculated from p and e − 3p. Although our errors are still large, our EOS shown in Fig. 2 is roughly consistent with recent results with highly improved staggered quarks near the physical point [3, 5]. |
Processed_Thermodynamics_in_2+1_flavor_QCD_with_improved_Wil.txt | Figure 3: Renormalized Polyakov loop and its susceptibility as functions of T , obtained with the fixed-scale approach. The left panel shows the results for 2 + 1 flavor QCD. The same quantities obtained in quenched QCD [20] are shown in the right panel. |
Processed_Thermodynamics_in_2+1_flavor_QCD_with_improved_Wil.txt | x4 U4(x) measures the quark free energy F(T ). To renormalize F(T ), we have to add an additive renormalization constant, which is independent of T and thus common for all T ’s in the fixed-scale approach. However, renormalization of L depends on T due to the relation hLi ∼ e−F/T . We adopt the renormalization scheme of Ref.[17]. The renormalized Polyakov loop is then given as Lren = ZNt tenL, where we set Zren = 1.4801(90) from our potential data obtained at T = 0 [18]. |
Processed_Thermodynamics_in_2+1_flavor_QCD_with_improved_Wil.txt | Our results for hLreni are plotted in the left panel of Fig. 3. As discussed in [10], the T - dependence in these quantities are largely influenced by the renormalization factor. We note that our hLreni agree well with a result from the p4 staggered quark action in the fixed-Nt approach at Nt = 8 [19], as shown by shaded symbols in the figure, although our quarks are heavier than theirs. In the same figure, the results for the susceptibility c Lren are also shown. We expect the pseudo- critical temperature around 200 MeV. We do not see a clear peak in c L, although a peak at T ∼ 180– 200 MeV is not excluded due to the large errors. For comparison, we show, the results for the same quantities obtained in quenched QCD adopting the fixed scale approach in the right panel of Fig. 3 [20]. In this case, we confirm a clear peak of the susceptibility at the weakly 1st order deconfinement transition temperature. We thus suspect that a peak is hidden in the large errors and/or between the two data points. Some reasons of the lower resolution of T = (Nta)−1 in the full QCD study are that Nt is restricted to be even due to the CPS simulation code we used, and that the lattice spacing a is coarser than that in the quenched study. We also note that the crossover may be weak due to the large quark mass, large Nt, small aspect ratio Ns/Nt, etc. We reserve investigation of these points for future study at lower quark masses. |
Processed_Thermodynamics_in_2+1_flavor_QCD_with_improved_Wil.txt | Figure 4: Bare chiral condensates (left) and their susceptibilities (right) as a function of temperature. |
Processed_Thermodynamics_in_2+1_flavor_QCD_with_improved_Wil.txt | studied the Polyakov loop, chiral condensate, and their susceptibilities. With the current precision of the simulation, however, we could not find a clear signal of the pseudo critical point. To improve the situation, it is important to achieve a higher resolution in T , e.g., by performing simulations at odd Nt and/or by combining data at different scales. We are planning to test these ideas in a study at the physical point, using the on-the-physical-point configurations generated by the PACS-CS Collaboration [1]. |
Processed_Thermodynamics_in_2+1_flavor_QCD_with_improved_Wil.txt | We thank the members of the CP-PACS and JLQCD Collaborations for providing us with their high-statistics 2+1 flavor QCD configurations. This work is in part supported by Grants-in-Aid of the Japanese Ministry of Education, Culture, Sports, Science and Technology, (Nos.22740168, 21340049 23540295 20340047 ) and the Grant-in-Aid for Scientific Research on Innovative Areas (Nos.20105001, 20105003 , 23105706 ). This work is in part supported also by the Large Scale Simulation Program of High Energy Accelerator Research Organization (KEK) Nos. 09/10-25, 10-09, (T)11-13 and 12-14. |
Processed_Photonic_force_optical_coherence_elastography_for_.txt | size, and scattering cross-section of the bead in relation to the characteristics of the forcing beam (e.g. |
Processed_Photonic_force_optical_coherence_elastography_for_.txt | the beam was comparable to the bead diameter (Fig. 1b); the force was lower as the two deviated. |
Processed_Photonic_force_optical_coherence_elastography_for_.txt | comparable diameters, which can be attributed to a higher backscattering cross-section of a larger bead. |
Processed_Photonic_force_optical_coherence_elastography_for_.txt | denote the radial and axial (depth) coordinates, defined w.r.t. the focus of the beam, respectively. |
Processed_Photonic_force_optical_coherence_elastography_for_.txt | stirred by hand to ensure all particles were dispersed before injecting into the pre-made glass chamber. |
Processed_Photonic_force_optical_coherence_elastography_for_.txt | were cut into a disk with diameter of 40 mm and thickness of 2 mm. |
Processed_Photonic_force_optical_coherence_elastography_for_.txt | the function generator output at the mth A-scan in each frame to reconstruct the full PF drive waveform. |
Processed_Photonic_force_optical_coherence_elastography_for_.txt | spatial pixels in each 3D dataset that constitute each bead, as described in Method 4. |
Processed_Photonic_force_optical_coherence_elastography_for_.txt | Bowman, R. W. & Padgett, M. J. Optical trapping and binding. Rep. Prog. Phys. 76, 026401 (2013). |
Processed_Photonic_force_optical_coherence_elastography_for_.txt | Block, S. M., Blair, D. F. & Berg, H. C. Compliance of Bacterial Flagella Measured with Optical Tweezers. |
Processed_Photonic_force_optical_coherence_elastography_for_.txt | Forties, R. A. & Wang, M. D. Discovering the Power of Single Molecules. Cell 157, 4-7 (2014). |
Processed_Photonic_force_optical_coherence_elastography_for_.txt | Ashkin, A. Acceleration and trapping of particles by radiation pressure. Phys. Rev. Lett. 24, 156-159 (1970). |
Processed_Photonic_force_optical_coherence_elastography_for_.txt | Dholakia, K., MacDonald, M. P., Zemanek, P. & Cizmar, T. in Methods Cell Biol. 82, 467-495 (2007). |
Processed_Photonic_force_optical_coherence_elastography_for_.txt | McInnes, C. R. in Solar Sailing Astronomy and Planetary Sciences Ch. 2, 32-55 (Springer, 1999). |
Processed_Photonic_force_optical_coherence_elastography_for_.txt | Discher, D. E., Janmey, P. & Wang, Y. L. Tissue cells feel and respond to the stiffness of their substrate. |
Processed_Photonic_force_optical_coherence_elastography_for_.txt | Paszek, M. J. et al. Tensional homeostasis and the malignant phenotype. Cancer Cell 8, 241-254 (2005). |
Processed_Photonic_force_optical_coherence_elastography_for_.txt | Trepat, X. et al. Physical forces during collective cell migration. Nat. Phys. 5, 426-430 (2009). |
Processed_Photonic_force_optical_coherence_elastography_for_.txt | Tambe, D. et al. Collective cell guidance by cooperative intercellular forces. Nat. Mater. 10, 469-475 (2011). |
Processed_Photonic_force_optical_coherence_elastography_for_.txt | Legant, W. R. et al. Measurement of mechanical tractions exerted by cells in three-dimensional matrices. |
Processed_Photonic_force_optical_coherence_elastography_for_.txt | Kotlarchyk, M. A. et al. Concentration Independent Modulation of Local Micromechanics in a Fibrin Gel. |
Processed_Photonic_force_optical_coherence_elastography_for_.txt | Scarcelli , G. & Yun, S. H. Confocal Brillouin microscopy for three-dimensional mechanical imaging. Nat. |
Processed_Photonic_force_optical_coherence_elastography_for_.txt | Kennedy, B. F., Wijesinghe, P. & Sampson, D. D. The emergence of optical elastography in biomedicine. |
Processed_Photonic_force_optical_coherence_elastography_for_.txt | Larin, K. V. & Sampson, D. D. Optical coherence elastography - OCT at work in tissue biomechanics. |
Processed_Photonic_force_optical_coherence_elastography_for_.txt | Chang, E. W., Kobler, J. B. & Yun, S. H. Subnanometer optical coherence tomographic vibrography. Opt. |
Processed_Photonic_force_optical_coherence_elastography_for_.txt | Klein, T. & Huber, R. High-speed OCT light sources and system. Biomed. Opt. Exp. 8, 828-859 (2017). |
Processed_Photonic_force_optical_coherence_elastography_for_.txt | Hale, G. M. & Querry, M. R. Optical Constants of Water in the 200-nm to 200-um Wavelength Region. Appl. |
Processed_Photonic_force_optical_coherence_elastography_for_.txt | Pernodet, N., Maaloum, M. & Tinland, B. Pore size of agarose gels by atomic force microscopy. |
Processed_Photonic_force_optical_coherence_elastography_for_.txt | axis in the BM-mode acquisition scheme resembled a pulse-train excitation with a 20-Hz sinusoidal envelope (∙∙∙∙). |
Processed_Photonic_force_optical_coherence_elastography_for_.txt | a parallel-plate shear rheometer. a, Storage moduli (―) and loss moduli (---) as a function of oscillation frequency. |
Processed_Photonic_force_optical_coherence_elastography_for_.txt | obtain the change in temperature due to optical absorption by water molecules in the agarose hydrogels. |
Processed_Photonic_force_optical_coherence_elastography_for_.txt | the reflected confocal response from a bead as the monolayer phantom was translated to various depths. |
Processed_Photonic_force_optical_coherence_elastography_for_.txt | of the PF forcing beam used in the experiments in this paper is shown in Supplementary Figure 1. |
Processed_Photonic_force_optical_coherence_elastography_for_.txt | by this variation in the magnitude of force due to the scanning of the PF forcing beam. |
Processed_Photonic_force_optical_coherence_elastography_for_.txt | Mao, R., Tang, J. & Swanson, B. G. Water holding capacity and microstructure of gellan gels. Carbohydr. |
Processed_Photonic_force_optical_coherence_elastography_for_.txt | Hale, G. M. & Querry, M. R. Optical Constants of Water in the 200-nm to 200-um Wavelength Region. Appl. |
Processed_Photonic_force_optical_coherence_elastography_for_.txt | Hawkes, J. B. & Astheimer, R. W. The Temperature Coefficient of the Refractive Index of Water. J. Opt. |
Processed_Photonic_force_optical_coherence_elastography_for_.txt | Chang, E. W., Kobler, J. B. & Yun, S. H. Subnanometer optical coherence tomographic vibrography. Opt. |
Processed_Photonic_force_optical_coherence_elastography_for_.txt | Pernodet, N., Maaloum, M. & Tinland, B. Pore size of agarose gels by atomic force microscopy. |
Processed_Unusual_shock-excited_OH_maser_emission_in_a_young.txt | of highly energetic processes (e.g. strong radiation fields and shocks), which are required to invert the energy lev- els involved in these transitions. |
Processed_Unusual_shock-excited_OH_maser_emission_in_a_young.txt | Circumstellar envelopes of low and intermediate mass (0.8-8M⊙) stars usually exhibit prominent OH maser emission at 1612 MHz. The spectra of these masers dur- ing the AGB phase tend to show double-horned pro- files, with two peaks indicating expansion velocities typ- ically between 10 to 30 km s−1, tracing the expansion of the circumstellar envelope (e.g. Sevenster et al. 1997). Also, some evolved stars show maser emission at 1665 and 1667 MHz (e.g. Hu et al. 1994; Deacon et al. 2004), the velocities of which are close to that of the 1612 MHz masers (Deacon et al. 2004). The 1720 MHz OH masers are best known to be associated with shocks created by SNRs interacting with surrounding molecular clouds (Frail et al. 1996; Green et al. 1997). The 1720 MHz maser transition has never been detected in AGB stars and is extremely rare in other evolved sources, i.e. it has only been detected in one post-AGB star (OH009.1−0.4; IRAS 18043−2116; Sevenster & Chapman 2001) and one Planetary Nebula (PN) (K 3−35; G´omez et al. 2009). Sevenster & Chapman (2001) argued that the 1720 MHz masers from OH009.1−0.4 might arise in the shock be- tween the AGB superwind and the fast post-AGB winds. Walsh et al. (2009) also detected extremely high veloc- ity H2O masers towards this source and suggested they originated in a jet. |
Processed_Unusual_shock-excited_OH_maser_emission_in_a_young.txt | phase. Compared to the copious OH masers in the AGB phase, the prevalence of OH masers decays significantly in subsequent phases, and they are rarely found in PNe. So far, only six PNe have been confirmed to harbour OH masers (OH-maser-emiting PNe, hereafter referred to as OHPN/OHPNe; Zijlstra et al. 1989; Uscanga et al. 2012). They all show maser emission with strong asym- metry, i.e. preferentially blue-shifted emission with re- spect to the stellar velocity (Uscanga et al. 2012). One of them, K 3−35, shows OH maser emission at all four ground-state transitions (1612, 1665, 1667, 1720 MHz; G´omez et al. 2009), as well as excited-state masers at 6035 MHz (Desmurs et al. 2010). It is so far the only PN known to exhibit OH maser lines at 1720 MHz. The ve- locity and location of the 1720 MHz OH maser emission are also close to those of 22 GHz H2O masers, which in- dicates that these two kinds of masers may be excited by the same shock (G´omez et al. 2009). |
Processed_Unusual_shock-excited_OH_maser_emission_in_a_young.txt | The Southern Parkes Large-Area Survey in Hydroxyl (SPLASH) is a survey of the OH radical across the inner Galaxy to high sensitivity (Dawson et al. 2014). In order to determine accurate positions for OH masers detected in SPLASH, we have used the Australia Telescope Com- pact Array (ATCA), which allows us to obtain arcsec- ond positional accuracy of the masers. In this paper, we report on one OH maser source (G336.644−0.695; here- after referred to as IRAS 16333−4807), seen at 1720 MHz as well as 1612 and 1667 MHz, that is found towards a previously-known PN, thus making it the second exam- ple of this very small class of 1720 MHz OH maser sources after K 3−35. |
Processed_Unusual_shock-excited_OH_maser_emission_in_a_young.txt | In the case of 1.1-3.1 GHz broad-band (continuum) data, we also used the Miriad package to reduce the data. Dirty maps of the continuum data were obtained with the task invert, applying frequency synthesis and a robust parameter of 0.5. Beam sizes at the two epochs were 7.3′′ × 3.9′′ and 12.1′′ × 4.6′′, respectively. Images were subsequently deconvolved with the task mfclean to perform the CLEAN algorithm. Continuum data at distances < 3 kλ in the uv-plane were discarded in the second epoch to filter out extended Galactic emission. |
Processed_Unusual_shock-excited_OH_maser_emission_in_a_young.txt | The flux density of the source shows significant varia- tion across the observed 1.1-3.1 GHz band. Fig.1 shows the flux density in 0.2-GHz intervals across this band. At both epochs the flux density increases with fre- quency, but this trend is steeper for frequencies ν . 2.4 GHz. The derived spectral indices α (with Sν ∝ να, where Sν is the flux density) are 1.39±0.11 (first epoch) and 1.39±0.08 (second epoch) between ≃ 1.4 and 2.4 GHz, and 0.18±0.07 (first epoch) and 0.20±0.19 (sec- ond epoch) between ≃ 2.4 and 3.0 GHz. The derived spectral indices are consistent, within errors, for both epochs. Since these observations were obtained with dif- ferent array configurations, we believe the flux density trend is a robust result and not the consequence of ex- tended emission being missed at the higher frequencies. We discuss the spectral index of the source in more detail in Section 4.1. |
Processed_Unusual_shock-excited_OH_maser_emission_in_a_young.txt | Fig. 1.— Variation of flux density of radio continuum emission as a function of frequency. Red symbols and blue symbols are from the first epoch and second epoch, respectively. The dot-dashed lines are the fitted results from 1383 MHz to 2407 MHz with spectral indices of 1.39±0.11 (first epoch) and 1.39±0.08 (second epoch), respectively. The dashed lines are the fitted results from 2407 MHz to 3022 MHz with spectral indices of 0.18±0.07 (first epoch) and 0.20±0.19 (second epoch), respectively. Error bars are 1-σ errors. |
Processed_Unusual_shock-excited_OH_maser_emission_in_a_young.txt | Fig. 2.— OH maser spectra of IRAS 16333−4807 at different transitions and epochs. The shaded velocity range shows the emis- sion channels for each maser spot (spectral component). The blank channels (from about −45 km s−1 to −35 km s−1) in the 1667 MHz spectrum taken in October 2013 were due to RFI and prompted us to re-observe this source. |
Processed_Unusual_shock-excited_OH_maser_emission_in_a_young.txt | The spectrum of the radio continuum emission in IRAS 16333−4807 is consistent with free-free emission in an ionized region. At low frequencies, a spectral index α ≃ 1.39 indicates partially optically thick free-free emission. Above the turnover frequency of ≃ 2.4 GHz, the emission is optically thinner, with a spectral index α ≃ 0.2, which is close to the optically thin limit in an ionized region (≃ −0.1). We note, however, that the flux density at 22 GHz reported by Uscanga et al. (2014) is not consistent with the spectrum in our observation. A crude extrapolation with α ≃ 0.2 to 22 GHz would give a flux density of ≃ 27 mJy. Therefore, there is an excess of ≃ 70 mJy in the Uscanga et al. (2014) observations. This discrepancy may be due to source variability. Note that we do not expect the spectral index of free-free emission to be ≃ 0.2 all the way to 22 GHz. At high frequencies this emission will become optically thin, and it will flat- ten up to α ≃ −0.1. In that case, the excess of emission at 22 GHz will be even higher. Alternatively, it is possible that the emission at 22 GHz includes a contribution ad- ditional to free-free radiation from the photoionized neb- ula, such as a central ionized disk (Tafoya et al. 2009), a collimated wind (Reynolds 1986), or circumstellar dust (Cerrigone et al. 2008). A more complete frequency cov- erage above 3 GHz is necessary to better ascertain the nature of the radio continuum emission. |
Processed_Unusual_shock-excited_OH_maser_emission_in_a_young.txt | data show wide absorption and emission features and we were unable to use them to obtain additional informa- tion on variability of the OH maser emission. These de- tections are very unusual for a PN. In particular, they confirm that IRAS 16333−4807 is the seventh OHPN, the second OHPN with 1720 MHz OH maser emission, and the only evolved object in whose circumstellar enve- lope the 1720 MHz maser is the strongest OH transition. Table 1 shows the relevant parameters of each maser spot at both epochs. The emission at 1612 and 1667 MHz shows a single spectral component at both epochs, peaked at ≃ −45 km s−1 in all cases. This velocity is close to that of the H2O maser emission (Uscanga et al. 2014, two components at ≃ −41.8 and −43.9 km s−1) and it is likely to be close to the stellar velocity. However, the 1720 MHz spectra show spectral components significantly offset, both red- and blue-shifted from that velocity. We find five maser spots in the October 2013 epoch, with velocities of approximately −60.7, −55.3, −49.0, −24.7 and −22.9 km s−1. During the May 2015 observation, we only find two maser spots at velocities of ∼ −55.1 and −23.2 km s−1. |
Processed_Unusual_shock-excited_OH_maser_emission_in_a_young.txt | Notes: Columns 1 and 2 are fitted positions (RA and DEC). Column 3 is the frequency. Columns 4, 5 and 6 are minor axis uncertainty, major axis uncertainty and position angle (measured from north to east). Columns 7 and 8 are peak flux density and peak velocity. Column 9 is line width. Column 10 is integrated flux density. The positions and relative positional uncertainty are derived from the Miriad using imfit task. The peak flux density, peak velocity, line width and integrated flux density are derived from Gaussian fitting in Class. |
Processed_Unusual_shock-excited_OH_maser_emission_in_a_young.txt | Fig. 3.— The LHC and RHC polarization spectra of the 1720 MHz transition at two epochs. The velocity resolution is 0.18 km s−1. |
Processed_Unusual_shock-excited_OH_maser_emission_in_a_young.txt | Fig. 4.— Distribution of the maser spots with respect to the lobes of the PN. Contours represent the radio continuum emission of IRAS 16333−4807 at 1.3 cm from Uscanga et al. (2014). The H2O maser components are represented with green filled diamonds. The 1612 and 1667 MHz OH maser spots from this paper are rep- resented with green filled triangle and green filled square. The blue and red filled circles are blue-shifted and red-shifted 1720 MHz OH maser spots compared to the velocities of H2O, 1612 and 1667 MHz maser spots. Note that only the OH maser spots from the first epoch are shown in this figure because they have higher absolute positional accuracy than those from the second epoch. The posi- tional uncertainties of maser spots are not plotted but are listed in Table 1. |
Processed_Unusual_shock-excited_OH_maser_emission_in_a_young.txt | positional uncertainty of the ATCA observations. Most maser spots are close to the center of the PN and they do not seem to be aligned along the lobes, but rather trace a structure perpendicular to them. The only exception could be the 1720 MHz maser spot at ≃ −22.9 km s−1, which seems to be located towards the northern lobe of the PN. |
Processed_Unusual_shock-excited_OH_maser_emission_in_a_young.txt | TABLE 2 Magnetic field properties of the 1720 MHz OH maser components in IRAS 16333−4807. |
Processed_Unusual_shock-excited_OH_maser_emission_in_a_young.txt | Notes: Column 1 is the local standard of rest (LSR) velocity of the component peak, which is derived from least-squares fitted Gaussian components with Class. Column 2 and 3 are the polarisation and peak flux density of the component. Column 4 is the numbered Zeeman pairs. Column 5 is the magnetic field strength with errors, which are all at 1-σ level. We adopt the common field convention with RHC at a lower velocity than LHC representing a negative magnetic field strength, directed towards us. The last column is the reliability for the Zeeman pairs, which is defined according to Green et al. (2015), with class ‘A’ being the most reliable. Note that we did not obtain the magnetic field strength of the weak maser component at the velocity of −55.3 km s−1 from the first epoch, as the signal to noise ratio for this maser is low, and as such the splitting has low statistical significance. |
Processed_Unusual_shock-excited_OH_maser_emission_in_a_young.txt | high emission measure could be a signature of youth (Kwok et al. 1981). Therefore, assuming that the central stars in the group of PNe with both OH and H2O masers have similar characteristics, the turnover frequency sug- gests that IRAS 16333−4807 is somewhat more evolved than K 3−35 and IRAS 17347−3139. |
Processed_Unusual_shock-excited_OH_maser_emission_in_a_young.txt | epoch within the errors (Table 2), which suggests that they are tracing the same structure. |
Processed_Unusual_shock-excited_OH_maser_emission_in_a_young.txt | It is believed that PNe are shaped by jets ejected dur- ing the post-AGB phase, which carve cavities in the cir- cumstellar envelope (Sahai & Trauger 1998). These jets obviously create shocks, which can power maser emis- sion. This is clearly seen in the case of the post-AGB stars called “water fountains”, in which H2O masers trace collimated jets (Imai 2007). If the OH masers at 1720 MHz in IRAS 16333−4807 were pumped by jets, we would expect them to be associated with the lobes of the PN. This could be the case for the 1720 MHz maser at ≃ −22.9 km s−1. However, most of these masers, and the overall distribution of all OH and H2O masers (Fig.4) seem to trace an equatorial struc- ture, perpendicular to the lobes. This could be re- lated to the low-velocity equatorial flows seen in some “water fountains” (Imai 2007) and the toroidal dis- tribution of H2O masers in other maser-emitting PNe (de Gregorio-Monsalvo et al. 2004; Uscanga et al. 2008). The spatial distribution of the 1720 MHz OH maser spots and the presence of clear Zeeman pairs suggest that OH emission at 1720 MHz in IRAS 16333−4807 arises from an energetic ejection of material along the equatorial plane, probably in the form of an expanding, magnetized torus. Note that K 3−35, the other PN showing maser emission at 1720 MHz, has also been inferred to host a magnetized disk (Miranda et al. 2001; G´omez et al. 2009). These equatorial ejections could be created, for instance, by binary interactions (Nordhaus & Blackman 2006; Huggins 2007). Magnetized equatorial structures traced by maser emission have also been found in post- AGB stars (Bains et al. 2003, 2004), with similar val- ues for the derived magnetic field as the ones found in PNe (G´omez et al. 2009, and this paper). This sug- gests that the characteristics of the magnetic field do not significantly evolve in the transition to the PN phase (Vlemmings & van Langevelde 2008). |
Processed_Unusual_shock-excited_OH_maser_emission_in_a_young.txt | The presence of OH emission at 1720 MHz could be the hallmark of a particular phase of PN evolution, char- acterized by energetic equatorial ejections. The extreme rarity of this emission in evolved stars suggests that this phase is very short, or that energetic equatorial ejections only happen in the form of recurrent, but short-lived events during the early PN phase. |
Processed_Unusual_shock-excited_OH_maser_emission_in_a_young.txt | natively, it could be a projection effect, if the OH masers at 1720 MHz tend to trace the expansion along the line of sight, and the other maser transitions are seen in re- gions expanding along the plane of the sky. This could happen if the velocity gradient is very high in the central regions, which would not favour the maser amplification along the line of sight. Moreover, the magnetic field is stronger in the region emitting at 1720 MHz than for the other OH masers, which suggests that these transitions are pumped in regions with significantly different mag- netic properties. This suggests that the different velocity patterns in these OH transitions are not merely projec- tion effects, but that the 1720 MHz masers are selectively tracing the energetic ejection of an equatorial magnetized structure. Further observations should be undertaken to obtain a better understanding of the source. |
Processed_Unusual_shock-excited_OH_maser_emission_in_a_young.txt | PNe. Alternatively, it is possible that they reflect a time variation in intrinsically similar processes, such as recur- rent episodes of mass loss. Monitoring of the OH emis- sion in these sources may help to clarify this issue. |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.