Our Work

Este livro-texto foi escrito para cursos de graduação em Engenharia Agrônoma, Ciências Agronômicas, Zootecnia, Administração Rural, Agronegócio e &…

Este livro-texto foi escrito para cursos de graduação em Engenharia Agrônoma, Ciências Agronômicas, Zootecnia, Administração Rural, Agronegócio e áreas afins — e também para o produtor, o técnico e o gestor rural que queiram estudar por conta própria. Não pressupomos nenhum conhecimento prévio de contabilidade. Partimos do zero, com a linguagem do campo, e chegamos à análise completa de demonstrações financeiras, com índices, prazos médios e necessidade de capital de giro.

Final Version of Record: https://doi.org/10.1021/jacs.5c10198 https://pubs.acs.org/jacsat/article/147/36/32982/3738411/Developing-a-Highly-Reducing-Heterogeneous-Nickel    This document is t…

Final Version of Record: https://doi.org/10.1021/jacs.5c10198 https://pubs.acs.org/jacsat/article/147/36/32982/3738411/Developing-a-Highly-Reducing-Heterogeneous-Nickel    This document is the Accepted Manuscript version of a Published Article that appeared in final form in the Journal of the American Chemical Society, copyright © 2026 American Chemical Society. To access the final published article, see ACS Articles on Request.   While intricate designs have been implemented to transform noble-metal complexes and sophisticated organic molecules into reactive photocatalysts for the single-electron reduction of unactivated alkyl halides, the development of highly reducing photocatalytic systems based on earth-abundant metals is still in its early stages. Herein, we show that a simple terpyridine-ligated nickel(II) complex, integrated into an imine-linked covalent organic framework, can be readily reduced by photoexcited Hantzsch esters to produce a heterogeneous nickel(I) photocatalyst with a fairly low excited-state oxidation potential of approximately −3.5 V (referenced to the saturated calomel electrode). The efficient electron transfer processes within the framework support and the confinement effects that stabilize the nickel(I) species contribute to the improved performance of this innovative heterogeneous nickel photocatalysis in comparison to its homogeneous counterparts. Notably, the current system allows for substantial reductions in catalyst loadings and exhibits high recyclability in promoting reductive couplings between challenging organohalide electrophiles and allyl sulfones. Our research indicates that employing framework-based heterogenization strategies can facilitate the development of new classes of robust photoredox catalysts with earth-abundant metals for sustainable and practical organic synthesis.

Bug fixes New data appended to an AMDA dataset (e.g. ACE) now shows up. Before, a request reaching past the end of a dataset cached the missing part as empty, for good, because AMDA does not change a…

Bug fixes New data appended to an AMDA dataset (e.g. ACE) now shows up. Before, a request reaching past the end of a dataset cached the missing part as empty, for good, because AMDA does not change a dataset's version when it only appends data. Fragments past the end of a dataset's coverage are now kept for one hour only. Empty AMDA fragments cached by Speasy 1.8.6 or older are fetched again once. New features Data providers register themselves: a provider declares NAME / ALIASES on its class and decorates it with @register_provider. Adding one no longer means editing Speasy's core. Packages can ship providers through a speasy.providers entry point. dir() on speasy and on the inventories lists every enabled provider, started or not. Behavior changes file is now a full alias of the archive provider: get_data("file/...") works, list_providers() includes it, and disabled_providers = file disables the archive. speasy.amda and the other provider attributes are no longer copied into the module namespace at import. They always reflect the current provider, including after a successful update_inventories() retry. vars(speasy) no longer lists them. Providers start on first use. import speasy no longer contacts any web service or builds any inventory. The first get_data(), speasy.amda or speasy.inventories.tree.cda access starts that provider, so it takes longer. SPEASY_SKIP_INIT_PROVIDERS now has no effect, and update_inventories() still starts every enabled provider. list_providers() lists the enabled providers, started or not. A provider whose web service is down is still listed. Using it raises an error that says it failed to start and how to retry. Code that walks speasy.inventories.tree.__dict__ (or vars()) only sees providers that have started. Use attribute access or dir(), which start providers as needed. Initialize providers on first use when SPEASY_SKIP_INIT_PROVIDERS is set by @Beforerr in https://github.com/SciQLop/speasy/pull/389 Self-registering data providers and a speasy.providers entry point by @jeandet in https://github.com/SciQLop/speasy/pull/391 Start providers on first use by default by @jeandet in https://github.com/SciQLop/speasy/pull/393

Três décadas de expansão do agronegócio produziram um mercado de derivativos agropecuários com 34 toneladas por contrato de soja FOB Santos, 330 arrobas por contrato …

Três décadas de expansão do agronegócio produziram um mercado de derivativos agropecuários com 34 toneladas por contrato de soja FOB Santos, 330 arrobas por contrato de boi gordo e 450 sacas por contrato de milho na B3 — e ainda assim pouquíssima literatura em português que mostre, passo a passo, como um real entra e sai de uma conta de margem. Este livro preenche esse vazio usando como espinha dorsal o curso de Mercados Futuros Agropecuários da UFSCar, reorganizado em 15 capítulos com arquitetura idêntica e dados reais verificáveis.

Unreleased Require at least two Local+Global station P/S pairs per final AI-PAL event by default. Make second-class duplicate thresholds inclusive: OT gap <= tolerance, pick overlap >= thresho…

Unreleased Require at least two Local+Global station P/S pairs per final AI-PAL event by default. Make second-class duplicate thresholds inclusive: OT gap <= tolerance, pick overlap >= threshold, and failed-pick fraction >= threshold. Training NPY cutting now resumes completed station-dates, regenerates incomplete shards, and preserves randomized splits across restarts. Fix negative training-sample cutting for station-days with no catalog picks. 2026-10-01: PAL range exports include one integrated association-rate CSV beside phase/catalog outputs, ready for training sample preparation. Add configurable second-class duplicate suppression for final AI-PAL events: ranked P-to-S overlap links, including descendants of discarded events, with per-link diagnostic CSVs. Available in local, AWS, and realtime workflows. Native realtime pickers now reject waveform glitches before initial association, consistent with local/AWS picking. Reference PHN-SB catalogs no longer receive PAL post-association glitch filtering. Initial-QC rejection bars are removed from monitoring; amplitude/magnitude measurement remains. 2026-09-30: Final event repicking can prefer Global (default) or Local timing using repick_timing_preference in the shared configuration. 2026-09-30: Unit-Gain Placeholder Treat gain 1.0 as missing calibration in local/AWS PAL and AI-PAL realtime processing, including legacy gain layouts. Retain waveform picking but output NaN station amplitude and exclude it from magnitude calculation if a selected component is uncalibrated. Existing amplitudes require reprocessing. AI-PAL repick version is advanced to invalidate previous repick completion records. Regression tests cover unit gain on all or one component and both AI-PAL/PALM calibration implementations. 2026-09-30: PAL Magnitude QC Require three distinct valid station magnitude estimates and population std <= 1.0, configurable with mag_min_stations and mag_max_std. Replace unconditional worst-station removal with a spread check followed by the median. Failed magnitude is -1 without discarding arrival picks or events; missing-gain amplitudes remain NaN. Magnitude merging excludes the -1 sentinel and retains other negative values. This supersedes the earlier NaN event-mag convention for PAL; MFT magnitude estimation is unchanged. Shared helper and parameter propagation cover local/AWS PAL and AI-PAL realtime/reassociation. Repick completion checks include magnitude thresholds. Tests cover the reported M6.56 case, missing amplitudes, distinct stations, negative magnitudes, threshold boundaries and AI-PAL/PALM source parity. Final AI-PAL events additionally require two quality-0 station P/S pairs by default (final_event_min_quality0_picks); zero disables this requirement. Final AI-PAL events now require at least one Local+Global station P/S pair by default across local, AWS and realtime; the minimum is configurable. 2026-09-29: Centralized packaged inference checkpoints in Pre-trained_models; removed verified duplicate input copies. CEED phase extraction now supports per-HDF5 multiprocessing and resumable completed-file outputs in both the packaged and SoCal workflows. Packaged CEED retraining now matches the workdir's separate positive-Zarr build (2.1) and resumable annual-negative transfer (2.2), using example paths. CEED training-shard preparation now adds randomly scaled synthetic Gaussian noise to copies after the first, following Local augmentation semantics. Validation remains unaugmented by noise; existing datasets must be rebuilt. Repicking now merges P/S pairs using distinct-window support and the shared multi-picker clustering algorithm, independent of predicted arrival ranking. Random windows retain the full phase buffer. Final duplicate merging matches NET.STA while preserving instrument selectors and recalculating support/quality. Added configurable negative-window training loss weighting for all four pickers, with separate group-loss diagnostics and unchanged full validation. Updated CEED retraining and SoCal case batch/weight recipes. Final waveform-measured phase and QC rows retain the selected channel band and location (NET.STA.BAND.LOC); locator readers accept these identifiers. Repick quality 0 now also includes three strong pickers within either Local or Global, configurable through pick_quality_code0_min_pickers. Event reassociation now uses all QC-accepted Local and Global repicks, including single-group picks. Removed the later residual-matched supplementation pass; existing QC and quality codes remain unchanged. All four picker trainers now evaluate the full negative validation set even when the negative training batch is zero, including metrics and balanced validation loss used for best-checkpoint selection. Default training batches are now [128, 32] for SAR/FT and [128, 8] for PHN/RUN. The dedicated CEED training recipe retains its smaller 8/4 negative batches. Picker architecture is inferred from the name prefix (e.g. SAR_CEED -> SAR). Removed redundant model keys from local/AWS/realtime picker settings. Fixed event repicking for dataset-qualified model names such as SAR_CEED and FT_CEED: architecture-specific input framing and inference are now selected using the configured model type, not its display name. Missing instrument gains no longer remove arrival-time picks. Same-band location/epoch fallback is allowed; no cross-band gain borrowing. Uncalibrated amplitudes are nan and excluded from event magnitude. 2026-09-25: realtime processing skips stations with missing instrument/epoch gains with a warning instead of aborting the whole segment; calibration remains strict and uncalibrated picks are not published. 2026-09-25: refreshed bundled SoCal 2020-2025 picker checkpoints and realtime example copies; training batches are [128, 32] for SAR/FT and [128, 16] for PHN/RUN. 2026-09-24: PAL amplitude QC no longer crashes on incomplete component windows. Unavailable checks are recorded as nan and bypassed; available failing checks still reject picks. 2026-09-24: Retired the combined offline AI-PAL launcher. Local and AWS inference now use stages 1 (pick), 2 (associate), and 3 (repick/reassociate). Existing output paths and resume manifests are unchanged; realtime is unchanged. AI picking batch size is configured only in the workflow config; the local example defaults to 256 without a pick-only launcher override. Local AI pick-only step 2.1 supports spawned date blocks (default 2), retaining per-process preprocessing concurrency (4). All picker device defaults use GPU 0, with batch size 256, private configs, per-block logs and GPU peak reporting. Move CEED-specific preprocessing/negative-transfer helpers and tests into Pre-trained_models/CEED, keeping dataset-specific workflows out of PAL_src. Clarify window-inference modules as picker_window.py; NPY-shard benchmark runners/adapters now live in the SoCal benchmark workdir instead of the source picker folders. Production repicking uses the shared window engines. Refresh bundled CEED inference checkpoints across local, AWS, and realtime workflows. Checkpoints use ceed_<model>_best.ckpt; CEED picker configs use config_<model>_global_ceed.py consistently, including optional training. Local picking exposes threads_per_worker for native numerical libraries, matching the AWS control; example launchers default to 2. Local PAL picking now uses independent date-block processes instead of station threads. Worker count controls date blocks; daily outputs and resume checks are unchanged, with per-block logs and aggregate console progress. Simplified local PAL entry points to 1_run_pal_pick_eg.py and 2_run_pal_assoc_eg.py; removed the redundant combined launcher. Local picking resumes without reading waveform tails for every skipped day; previous-day context is loaded only when picking actually resumes. Fixed daily trigger-count ownership at buffered boundaries: pre-QC candidates now use refined P time like accepted picks, avoiding false count failures. Local picking retains waveforms with missing gain locations/epochs, using warned same-band gain fallback or uncalibrated counts as a last resort. Local PAL picking now prints live day/station progress and a 30-second heartbeat while retaining detailed output in a line-buffered log. Daily PAL association now shares buffered picks across all subnets inside parallel contiguous date blocks, eliminating per-subnet pick-file rereads. PAL configs explicitly use a 30 s association buffer. Local PAL examples now enable cross-day association and pick halos, with bounded per-worker daily pick caching. Association resumes detect buffer-policy changes. Align positive training batches to each annual Zarr store's physical chunks for all four pickers, avoiding unnecessary cross-chunk reads. Negative training now reuses decoded Zarr chunks across minibatches, with disjoint worker partitions and traversal continuing across positive epochs. Positive sampling, batch ratios and validation are unchanged. Separated optional CEED preparation/retraining from local picker training; numbered each workflow independently and shared CEED waveform settings across preparation and training. Local users can reuse the supplied Global models. Renamed raw waveform cleaning control to to_clean, distinct from filtering; legacy to_prep configs and reader keyword arguments remain supported. Final pick quality counts votes at the configured strong-vote threshold inclusively (>=); a ratio of exactly 0.5 now qualifies by default. Default picker trigger thresholds to 0.3 across Local/Global models; move final quality settings into post-processing and standardize waveform option names to singular (save_filtered_event_waveform). Aligned the SoCal realtime case wrapper with Local/Global continuous picking, dataset-qualified CEED models, and current source configuration/CLI names. Compact final phase rows to 13 columns: picker identities are included in vote ratios, SNR components share one field, and probability spreads move to the QC companion. Shared readers retain legacy-format support. Fixed FT training failing after its first update because the validation scheduler referenced an undefined batch-count variable. Separated CEED preprocessing from case-specific SeisBench benchmark utilities; the CEED fixed-window builder no longer depends on the SeisBench builder. All four trainers validate the entire positive and negative splits separately, select best checkpoints using class-balanced validation loss, report both class losses, and always validate at the final training step. Training mode now follows batch_size alone: [128, 0] trains and validates using positives only. Removed the separate positive-only CLI switch; converters select source classes with --sample_types instead. The historical benchmark train_pos/3_train_pos_pickers.py now trains on CEED positives plus imported local negatives, with reduced negative batches. Annual CEED negative transfers now checkpoint and resume interrupted copies; legacy interrupted transfers have an explicit, positive-preserving restart. Continuous picker support now accepts [min_local, min_global, min_total], defaulting to [0, 0, 1]; all three minima must pass. Global picker identities now include their training dataset, such as SAR_CEED, independently of the architecture used to load each model. Renamed inference picker groups to Local (formerly POS_NEG) and Global (formerly POS), including config keys, global config filenames, and QC labels. Historical phase/QC and timing labels remain readable; checkpoint files are not renamed. Continuous AI-PAL picking now combines local POS_NEG and mixed-trained CEED POS SAR/PHN models. picker_group_min_picker_support defaults to one vote across the combined ensemble; local, AWS, and realtime workflows are aligned. Existing CEED positive Zarr stores can now import negatives from annual local archives without rebuilding positives, using the special-purpose benchmark_picker/train_pos/2.3_copy_negatives_ceed.py launcher. Added separate local and CEED training stages; CEED training now mixes CEED positives with local negatives using explicit per-model batches. The positive-only benchmark remains unchanged. Training batches now specify [positive, negative] counts directly: SAR/FT [128, 32], PHN/RUN [128, 16]. Removed ratio-based negative reduction; only requested negative windows are loaded. POS-only training uses the first count; validation sampling remains unchanged. Added standalone polarity_RUN: a residual U-Net for 2.5-second vertical waveforms with [Noise, Up, Down] soft labels and strict probability > 0.5 decoding. Source layout follows phase RUN; training-data construction and executable workflows are deferred. POS picker trigger thresholds now default to 0.6 for SAR/FT and 0.3 for PHN/RUN. SoCal AWS FT/RUN configs use the current compact architectures; matching checkpoints are required. Unified configurable band/location selection (band first by default) and fullfed-derived, time-dependent NET.STA.BAND.LOC gain inventories across local and AWS PAL, AI-PAL realtime, and PALM MFT. Simplified files remain supported; routine post-merge station-file reconciliation is unnecessary. Documented station-file formats and workflow restrictions; corrected AWS inference examples to use SCEDC epoch metadata and matching example dates. Added per-picker P/S probability histogram inspection from final QC tables, with POS/POS_NEG selection, time filtering and downloadable bin counts. Offline final-event PNGs now use event_waveform_plot/, distinct from filtered waveform data in event_waveforms/; existing folders are retained. Published AI-PAL final phases now separate detailed per-picker QC into matching .picker_qc.csv files, including median P/S probabilities across random-window votes. The phase row drops the packed picker-uncertainty field; built-in readers support both old and new formats. Unified offline initial/final range phase and catalog outputs across local and AWS combined/split workflows. Offline initial results use 2.1_phase_init_AI-PAL; new intermediate files live under _internal. Legacy resume paths remain supported; realtime numbering is unchanged. Fixed PAL event merging that discarded valid negative magnitudes. Missing magnitudes now use nan, not the physically valid value -1. Existing affected catalogs require magnitude recalculation; they are not auto-repaired. Added realtime preferred/reference catalog inspection: FMD, side-by-side event maps, matched phase-count plots, residual-ranked one-to-one event matching, and waveform copies for detections without eligible counterparts. Competing assignments are reported separately rather than labeled as misses. Event waveform filenames now include OT, latitude, and longitude to avoid overwriting distinct co-OT detections; QC also supports legacy filenames.

This dataset contains dated ancestral recombination graphs (ARGs), stored as tskit tree sequences, for chromosomes 17 and 20 from the 1000 Genomes Project 30x high-coverage dataset (3,202 individuals)…

This dataset contains dated ancestral recombination graphs (ARGs), stored as tskit tree sequences, for chromosomes 17 and 20 from the 1000 Genomes Project 30x high-coverage dataset (3,202 individuals). They were created as part of the research described in "Tracing the evolutionary histories of ultra-rare variants using variational dating of large ancestral recombination graphs". ARGs were inferred from the phased panel using tsinfer. Singletons from the unphased genotypes were then added, and the ARGs were dated using tsdate v0.2 (variational_gamma) with a mutation rate of 1.29e-8 per base per generation. Each file covers one chromosome arm or sub-arm region (GRCh38). It can be loaded with tszip, and its top-level metadata describes the data sources and the region. Population labels are stored in the individual metadata.

The papers released after Complementarity-First Foundational Release III developexplicit constructions in which a retained record, a comparison frame, an analytic domain, or apreparation map determine…

The papers released after Complementarity-First Foundational Release III developexplicit constructions in which a retained record, a comparison frame, an analytic domain, or apreparation map determines what can be composed and inferred. This review synthesizes fourteenresearch papers in five branches: autonomous source refresh; structured magnetic-core completionand local monopole dynamics; orientation-twisted composition and reciprocal quantum cells; non-linear superconducting response; and Feynman composition, configuration histories, and low-energyinterfaces. The source-refresh studies distinguish an exact spatial exchange from a reusable appara-tus, and an intended smooth model from its exact piecewise numerical source. The magnetic trilogyseparates extension of a structured bundle, observable-specific exterior recovery, energy coercivity,kinetic threshold behavior, and regular local evolution. The quantum-cell papers distinguish coher-ent changes of presentation from physically admitted processes, geometric cells from quantum statespaces, and boundary-word identities from unrestricted refinement invariance. The superconduct-ing comparison separates restricted density equivalence, a specified nonlinear field contrast, andexperimental identification. The four composition papers connect conditional amplitude rules tocontrolled one-dimensional histories, coherent coarse memory, induced metrics, common-windowpreparation, and explicit spectral-selection costs. Cross-branch comparisons identify genuine im-ports without assigning a common physical carrier to mathematically different models. Analyticarguments, finite certificates, historical replays, and partial formalization retain separate eviden-tiary roles. The contribution is a source-attributed release map and a synthesis of compositionand inference boundaries, not a new primitive derivation of quantum theory, matter, or gravity.Publication confirmation, access to the consulted edition, and deposited-file identity are recordedseparately.

Aerial rope systems are currently widely used for carrying out a variety of transport and logistics operations in a wide range of sectors of the modern economy. Along with the traditional use of ropew…

Aerial rope systems are currently widely used for carrying out a variety of transport and logistics operations in a wide range of sectors of the modern economy. Along with the traditional use of ropeways in stationary design, the direction of using mobile rope transport systems has been actively developing in recent years. This makes it possible to effectively solve important transport and logistics tasks within a limited time limit and in difficult natural, climatic and/or operational conditions, in particular, such tasks as carrying out transport operations in remote and inaccessible areas, carrying out loading and unloading operations on an unequipped coast, carrying out special rescue and transport operations, carrying out transport works on the surface of other cosmic bodies, etc. This article discusses the task of developing intermediate towers for mobile ropeways, characterized by a minimum mass while meeting the necessary strength requirements. The developed mathematical optimization model and the computer program implementing it are presented. Five design variants of the cross-section of the intermediate tower of the truss structure with supporting vertical thin-walled struts are considered. When minimizing the mass of the metal structure of the tower, the influence and mass of the anchor ropes providing a given spatial position of the intermediate tower are taken into account. Some results of optimization calculations for the basic version of the rope system are also presented.

Machine-learning screening of 1,066 stable perovskite-family oxides forvisible-light photocatalysis, with a grouped-model ablation and a literaturevalidation tier. This archive contains the training d…

Machine-learning screening of 1,066 stable perovskite-family oxides forvisible-light photocatalysis, with a grouped-model ablation and a literaturevalidation tier. This archive contains the training descriptors, the screeningpool, every generated data product, all figures, and the four scripts thatregenerate them.