Engineering Research Excellence
Conceptual design, CAD, FEA, thermal and fluid analysis, optimisation and validation.
Read more
Theoretical modelling, experimental data interpretation, computation and validation.
Physics research fails in two places: a model that is elegant but not testable, and data that is real but analysed without regard for its uncertainty. Both are avoidable.
The six services on the poster span the work: theoretical modelling, experimental analysis, computational physics, data analysis and visualisation, research validation, and publication support. The research areas listed alongside them — classical mechanics, electromagnetism, quantum physics, thermodynamics, optics and photonics, solid state physics, and astrophysics and cosmology — are the areas we can support with someone who knows the field.
Theoretical work means deriving the model, checking its limits, and producing numerical solutions where analytical ones do not exist. Experimental work means treating measurement properly: propagating uncertainty, distinguishing statistical from systematic error, fitting with the right weighting, and reporting a result with an honest interval around it.
The example analyses shown on the artwork — a harmonic oscillator displacement trace, a linear fit with its coefficient of determination, a spectrum with resolved peaks, a field distribution and a quantum well energy-level diagram — are the everyday outputs of this work, and each one has a right and a wrong way to be produced.
Six service areas from the poster, and what each involves in practice.
A measurement without an uncertainty is not a measurement. This is treated as obvious in physics and yet a surprising number of theses report values to five decimal places from an instrument accurate to two, or quote a fitted parameter without the standard error the fit produced.
We build an error budget: which contributions are statistical and reduce with repetition, which are systematic and do not, how they combine, and which one dominates. That last point is the useful one, because it tells you where to spend effort. Improving a contribution that is an order of magnitude below the dominant term is wasted work, and knowing that early saves months in the laboratory.
Fitting deserves the same care. Least squares assumes the residuals are normally distributed with constant variance, and where they are not — Poisson counting data, heteroscedastic instrument noise — the fit needs weighting or a different likelihood. A reduced chi-squared far from one is telling you something, and it is worth reading rather than ignoring.
The question stated precisely enough that a specific measurement or calculation can answer it.
What is already established, what is contested, and where the actual open question sits.
The theoretical model, its assumptions, its limiting cases and the regime in which it is expected to hold.
Numerical solution or simulation, with convergence and stability verified rather than assumed.
Comparison against experiment, an analytical limit or published data, with uncertainty carried through.
Manuscript, figures and equations prepared to the standard of the target journal, usually in LaTeX.
A simulation result that cannot be reproduced is not evidence. We deliver code that runs, with the parameters in a configuration file rather than buried in the source, a fixed random seed where stochastic methods are used, and a short note on the machine and library versions used to produce the published numbers.
Convergence has to be demonstrated, not assumed. That means grid or time-step refinement studies for differential equation solvers, sample-size convergence for Monte Carlo, and energy conservation checks for dynamics. These plots belong in the thesis, and they answer questions an examiner would otherwise ask.
Where analytical solutions exist for a simplified case, we run the code against them first. Matching a known result in a limiting case is the cheapest and most convincing verification available, and its absence is the first thing a careful reviewer looks for.
| What you receive |
|---|
| Derivations written out with assumptions stated |
| Working, documented simulation or analysis code |
| Convergence and verification studies |
| Reduced data with a full error budget |
| Fits with parameters, uncertainties and residual plots |
| Publication-quality figures with correct error representation |
| LaTeX manuscript with typeset equations |
| Interpretation and discussion of limits and validity |
Classical mechanics and dynamics, electromagnetism, quantum mechanics, thermodynamics and statistical physics, optics and photonics, solid state and nanomaterials, and astrophysics and cosmology. If your topic falls outside what we can support properly we say so at the first call.
Yes. LaTeX is the norm in physics and we prepare manuscripts, equations, figures and bibliographies in it, using the journal's own class file where one exists.
With the instrument and the question. What was measured, with what resolution, under what conditions, and what claim do you want to make. From there we build the reduction pipeline and the error budget, and only then choose the fitting or statistical method.
Yes. Dimensional consistency, limiting cases, sign conventions and the algebra itself — a second careful reader finds errors that the author cannot see after weeks with the same pages.
Often the answer is both. An analytical solution for a simplified case gives you physical insight and a check on the code; the numerical model handles the real geometry or the non-linear term the analysis cannot. We usually recommend deriving the limiting case first, because it costs little and it verifies everything that follows. Where the two disagree, that disagreement is itself a finding worth investigating.
Conceptual design, CAD, FEA, thermal and fluid analysis, optimisation and validation.
Read more
From raw data to real insight — descriptive, inferential, regression and multivariate work.
Read more
Structural, thermal, CFD, modal, impact and multi-physics analysis with optimisation.
Read moreSend your topic, your dataset or one draft chapter. We will tell you honestly what it needs — before you pay anything. The first consultation is free.