The Laboratory Notebook and the Report - Modern Physics Laboratory

These are two different documents with two different purposes, and the most common mistake is to write one when the other was wanted.

The notebook is a contemporaneous record. Its test is: could someone repeat this measurement from what is written here, including the mistakes?

The report is an argument. Its test is: does a reader who was not there believe the number, and know how much to believe it?

The notebook¶

Use a bound notebook with numbered pages — not a spiral pad, not loose sheets, not a laptop file. Write in ink. When you make a mistake, draw a single line through it so that it remains legible, and write what was wrong beside it. The crossed-out entry is often the most useful thing on the page three weeks later.

What goes on the page¶

Date, partner, experiment, and the instrument’s identity. “Tektronix TBS1052B, bench 3, serial ...074” is worth writing down; when two groups get different answers, the instrument is a suspect and you will want to know which one you had.

A sketch of the setup, with distances. Thirty seconds of drawing saves an hour of reconstruction. Label what you actually measured, with the origin marked. Photographs supplement the sketch; they do not replace it, because a photograph does not tell the reader which distance you called $L$ .

Raw readings, in the units the instrument displays. Do not convert in your head and record the converted value. If the counter reads 4127 counts in $60\ \text{s}$ , write “4127 counts, 60 s”, not “68.8 s $^{-1}$ ”. Conversions are recoverable; the raw reading is not.

The uncertainty of each reading, decided at the time you take it. Was it the last-digit resolution of the display? The spread over three repeats? The width over which you could not tell the fringe had moved? You know this at the bench and you will be guessing at your desk.

Tables drawn before they are filled. A table with a header row and labeled units, drawn empty, is a plan. A column of numbers written down the margin is a puzzle.

What went wrong, and what you did. “First run: photodiode saturated above $3\ \text{mm}$ aperture — added ND1 filter, retook from step 4.” This sentence is worth more to your report’s discussion section than any of the good data.

A running conclusion. Before you leave, write one sentence: what number did you get, and does it look right? This forces the twenty-minute in-room fit described in How to Use This Manual.

What the instructor checks¶

Notebooks are initialled at the end of every period — a signature that the work in front of them was done that day — and graded four times a semester against the rubric. Grading is for completeness and contemporaneity, not for neatness. A messy, honest, complete notebook scores above a tidy one written up afterwards, and the difference is usually detectable.

The report¶

Reports follow the structure of a short experimental paper. Length: two to three pages for a short report, five to eight for a full report, in both cases excluding figures.

Structure¶

Title and abstract. The abstract is three or four sentences and contains the number. “We measured the wavelength of a helium–neon laser by counting interference fringes in a Michelson interferometer as one mirror was translated, obtaining $\lambda = 633.4 \pm 0.8\ \text{nm}$ , consistent with the accepted $632.816\ \text{nm}$ .” An abstract that promises the reader will find a result, without stating it, has failed.

Introduction (full reports only). What physical question is at stake and why the measurement bears on it. Two paragraphs. Cite the textbook chapter and at least one outside source.

Theory. The derivation of the equation you will fit, starting from something the reader already accepts. Include the equation you actually use, numbered, with every symbol defined. Do not reproduce the whole chapter.

Apparatus and method. Enough that a competent reader could repeat it. A labeled figure — your sketch, redrawn, or a photograph with annotations — does most of this work. Give model numbers and the quantities you controlled.

Results. The data, reduced. A figure showing the data with error bars and the fitted curve; a table of fitted parameters with their uncertainties. Every figure has a caption that states what is plotted and what the reader should notice. Every number has a unit and an uncertainty.

Discussion. The part that is actually graded hardest. It must answer:

Is the discrepancy from the accepted value consistent with the uncertainty? Quote the discrepancy in units of the combined uncertainty, $t = |x_{\text{meas}} - x_{\text{acc}}| / \sqrt{\sigma_{\text{meas}}^2 + \sigma_{\text{acc}}^2}$ . A $t$ of 0.4 suggests no notable discrepancy under the stated uncertainty model; a $t$ of 6 calls for investigating the model, the measurement, and the uncertainty budget. This interpretation assumes independent estimates with comparable standard uncertainties; see Uncertainty and Curve Fitting.
Which uncertainty dominates, and what would you change to reduce it? “Random” and “human error” are not answers. “The dominant term is the $\pm 0.5\ \text{mm}$ ruler reading on $L = 2.14\ \text{m}$ ; a laser rangefinder would reduce it tenfold and make the photodiode rise time the new limit” is an answer.
What systematic effects could shift the result in a particular direction, and did you bound them?

Conclusion. One paragraph. The number, its uncertainty, and what it means.

References. Any consistent style, used consistently.

On voice and tense¶

Write in the past tense for what you did (“the mirror was translated in $50\ \mu\text{m}$ steps”) and the present for what is true (“the fringe spacing is inversely proportional to the slit separation”). First person plural is fine and often clearer than the passive. Avoid “the experiment was a success” — the reader will decide that.

On figures¶

A figure that is not referred to in the text should be deleted. Axes are labeled with quantity and unit. Data are points with error bars; fits are lines without markers. Do not use a spreadsheet’s default styling, do not connect data points with straight segments, and do not plot a fit outside the range of the data.

On significant figures¶

Round the uncertainty to one significant figure (two if the leading digit is 1), then round the value to the same decimal place. $\lambda = 633.4 \pm 0.8\ \text{nm}$ , never $633.427 \pm 0.8113\ \text{nm}$ . This single habit accounts for a surprising fraction of the difference between reports that read as competent and reports that do not.