Twelve deeply authored native research-methods units on claim-evidence maps, formal reading, independent proofs, computational reproducibility, discretization and extrapolation, inverse problems, uncertainty, research graphics, ethics, preprints and falsifiable proposals. Fully rendered mathematical developments and worked checks appear on QuantumSpark itself, with original instructor videos used only where relevant.
Published syllabus
A literature map begins with the research question and separates seminal results, current benchmarks, contrary evidence, independent replications and review papers. Citation counts alone do not determine validity. Query strategies should use multiple terms and follow backwards/forwards references, recording exact papers and their limitations.
Read a paper by identifying its precise theorem, model, experimental method, dataset or benchmark, then trace notation into the main conclusion. Skimming the abstract is not enough to assess error models or extrapolations. A useful reading record states the strongest valid claim and the assumptions under which it holds.
Reproducing a derivation requires reconstructing steps, units, operator domains and mathematical conventions. Restate a key theorem independently, prove it in a minimal example and identify a nearby counterexample when hypotheses fail. This distinguishes learning from rearranging published algebra.
A reproducible numerical result records code version, package environment, random seeds, input configuration, hardware if relevant and expected output tests. Separate computational reproducibility from scientific replicability, which may require a different apparatus or independent implementation.
Simulation design starts with the physical model, controlled parameter regime and the numerical discretization. Conservation laws, symmetry sectors and known limits provide better checks than visually plausible trajectories. Plan computational cost before sweeping arbitrary multidimensional grids.
Finite lattice approximations replace an infinite theory by tractable matrix or Monte Carlo models. Boundary conditions can create spurious states and change the topology of configuration space. Extrapolating system size requires observable-specific scaling and estimates of truncation and statistical error.
A theory-to-experiment comparison must specify instrument response, calibration, noise, nuisance parameters, sample selection and observational uncertainty. Predicted ideal quantities may differ from measured readout counts after detector convolution. The model should state which measurements could falsify it.
A research notebook should distinguish explanatory prose, source code, derived results and machine configuration. LaTeX source documents should compile reliably, store equations as semantic source and include bibliographic identifiers. Archiving a static PDF without reproducible code may impede later corrections.
Plots need explicit axes, units, uncertainty estimates and clear distinctions between simulation and empirical data. Logarithmic scales, normalized ratios and color maps can mislead without limits and conventions. A claim about a statistically significant difference requires an appropriate comparison test or confidence interval.
Scientific ethics includes accurate source citation, author contribution, consent where relevant and preservation of negative results or failed qualification gates. AI assistance cannot replace human responsibility for verifiable results. Separate predictions from demonstrations and do not present conceptual diagrams as experimental images.
Preprints establish a public versioned record but do not automatically carry peer-review endorsement. Peer review tests clarity and soundness without guaranteeing correctness or novelty. Revision should respond to substantive methodological critiques and retain a traceable history of changed claims.
A capstone research proposal should pose a falsifiable question, characterize competing baselines, identify data or computational resources, budget time, and declare stop criteria. The scope should be narrow enough to yield a finished result with clear uncertainty and limitations rather than a giant list of unrelated frontier topics.