Physics IA topic ideas: eight models you can test
Compare eight Physics IA topic ideas by measurement, model and uncertainty. See how a falling light curve can still leave a systematic prediction error.

Choose an outcome you can measure
A Physics investigation needs a clear connection between the question, model and measurements. These original seeds are starting points that require adaptation and teacher approval. Check the school's equipment, safe operating arrangements and available collection time. Use ordinary school apparatus within approved limits; the examples do not provide experimental instructions.
Name the variable you will vary or compare and the physical quantity you will measure. Decide what the model predicts so you can select an informative range rather than collect arbitrary readings.
Mechanics investigations
- Pendulum amplitude: How does release amplitude relate to measured period for one fixed pendulum? Define amplitude and pilot whether timing precision can reveal the expected differences.
- Spring loading: How does extension relate to applied load within a teacher-approved range? Establish a consistent reference position and examine whether loading and unloading agree.
- A rolling ramp: How does ramp angle relate to the travel time of one rolling object over a fixed distance? Keep release conditions consistent and consider friction and rotation.
Energy, light and sound investigations
- Cooling and covering: How does one approved insulating layer affect a measured cooling curve? Match starting conditions and monitor the surroundings rather than assuming identical heat exchange.
- Light and distance: How does a sensor reading vary with distance from one suitable light source? Check background light, geometry and the sensor's operating range.
- Resonance and geometry: How does the length of an approved air-column setup relate to a measured resonance frequency? Define the endpoint method and use school-supported sound levels and apparatus.
Fields and published-data investigations
- A coil and magnetic response: How does a school-approved low-voltage circuit condition relate to an observable magnetic response in a fixed setup? Check alignment, background fields and safe operating limits with the teacher.
- Orbital data: How well do published orbital periods and orbital dimensions follow a stated model for a carefully selected system? Check parameter definitions, host properties and measurement uncertainties.
Compare quantities, models and error checks
These are possible model choices for the eight original seeds, not fixed IA requirements. Derive or source the relationship you use and explain why its assumptions fit your setup. Some extensions may need mathematics beyond your current course.
On a small screen, swipe inside the table to read all four columns.
| Topic | Quantities to define | Model prediction to examine | Main error or assumption check |
|---|---|---|---|
| Pendulum amplitude | Release angle and period, with length fixed | At small amplitudes, the simple-pendulum approximation predicts nearly constant period; larger amplitudes can depart from it. | Timing resolution, length definition and changing amplitude. Can the expected difference be resolved? |
| Spring loading | Applied force and extension from a defined zero | Within the elastic, approximately linear regime, extension is proportional to force. Compare loading with unloading. | Zero offset, parallax and a changing spring response. A straight segment does not justify extrapolating beyond the tested range. |
| Rolling ramp | Ramp angle and travel time over a fixed distance | For the same object rolling without slipping from rest in an ideal fixed-distance model, time decreases as sin(angle) increases. | Release motion, slipping, timing and angle measurement. Include rotation when choosing a prediction. |
| Cooling and covering | Temperature and time, with ambient temperature recorded | A simple Newton-cooling model predicts exponential decay of temperature excess when its assumptions apply. Compare fitted cooling constants. | Changing room temperature, sensor response and unequal starting conditions. One exponential is not suitable for every interval. |
| Light and distance | Source–sensor distance and calibrated sensor response | A point-source approximation predicts an inverse-square source contribution. Background adds a separate signal. | Distance zero, extended-source geometry, background and saturation. Sensor output must represent the quantity being compared. |
| Resonance and geometry | Air-column length and resonance frequency | For the same mode and boundary conditions, frequency varies inversely with effective length; end effects may matter. | Finding the same mode, length reference, temperature and frequency resolution. Do not mix open and closed boundary models. |
| Coil and magnetic response | Current and a stated field component at fixed geometry | For a fixed air-core geometry, the source-field component is proportional to current in the simple model. | Sensor alignment, background field and changing current. Applied voltage is not a substitute for measuring current. |
| Orbital data | Orbital period and semi-major axis in consistent units | For a shared dominant central mass, a two-body approximation gives period squared proportional to semi-major axis cubed. | Mixing host masses, radius versus semi-major axis, uncertain catalogue values and inconsistent units. |
Choose a question for which a pilot can expose a real decision: change the measurement range, improve the quantity definition, or revise the model. Do not choose an unfamiliar equation simply because it looks more advanced.
Pilot the measurement that limits the question
Shortlist two seeds and identify the weakest step. In the light-distance example, a pilot can show whether background light dominates at larger distances or the sensor saturates nearby. That information should guide the usable range. In the spring example, a pilot can reveal an unclear zero position or a response outside the intended regime.
Record the pilot conditions and unexpected observations. Repeats should help you understand variability; they do not repair an unsuitable range or a poorly defined measurement. Discuss realistic adjustments before full data collection.
Define the mathematical comparison
Write the relationship you expect, the variables' units and the graph or calculation that can test it. Explain any transformation and whether its assumptions apply to the setup. For database work, inspect actual records and documentation before committing; a published number can have a definition or uncertainty that changes its use.
Choose the question with a defensible route from raw observations to an answer. Bring the pilot and analysis plan to your teacher, confirm the course requirements and reserve time for collection, processing and evaluation within the school deadline.
A falling curve can still disagree with the model
Suppose a light signal decreases as the sensor is moved farther away. That direction is consistent with many models; it does not by itself establish an inverse-square relationship.
Use a fixed reference prediction of 100/r², where r is the distance in metres and the signal is expressed in arbitrary units. The coefficient is chosen for this illustration, not estimated from an experiment. The invented readings add exactly 6 units to every prediction.
Difference = observed signal − reference prediction
1. The overall trend
2. The prediction differences
| Distance / m | Prediction / a.u. | Invented reading / a.u. | Difference / a.u. |
|---|---|---|---|
| 1.0 | 100.00 | 106.00 | +6.00 |
| 2.0 | 25.00 | 31.00 | +6.00 |
| 3.0 | 11.11 | 17.11 | +6.00 |
| 4.0 | 6.25 | 12.25 | +6.00 |
What changes in the conclusion? You can say that the signal falls with distance, but this fixed source-only prediction leaves an unexplained positive offset. Background light or a sensor zero offset would be possibilities to investigate, not causes established by the graph.
What would you check next? Obtain an independent background or zero check, examine the source–sensor distance definition, and confirm the sensor is operating in its usable range. A model with a background term may be justified if the physical checks support it. Simply subtracting 6 because it makes these invented points match would not be a defensible experimental method.
In real work, compare differences with measurement uncertainty and repeat variability, and fit model parameters transparently where needed. Residuals from a fitted model mean observed minus fitted values; the graphs here use a deliberately fixed reference to make the subtraction easy to follow. Real residuals will rarely be identical, and four invented points with no uncertainty estimates cannot validate a physical model.
The NIST guide to model validation explains why residual patterns provide information that a single fit statistic can miss.
Your next-step checklist
Checklist choices stay in this browser.
References and course guidance
This is an original practical guide from IBvia. The examples are illustrative; your subject guide, assessment year and school instructions determine the requirements.

