Gains:
- Ability to test the accuracy of a formula by dimensionally equalizing its two sides without taking into account numerical calculations
- Ability to convert inputs into a single unit system and verify compound and energy unit conversions with unit-aware tools or manually
- Ability to catch silent conversion errors by routinely adding unit, order and boundary condition checking to each account
The fastest, most powerful and most neglected way to test the accuracy of a result in physics is dimensional analysis. Dimensional analysis is checking whether two sides of an equation have the same physical dimension (length, time, mass, energy…). Because the two sides of a physical equation must be dimensionally equal before they can be numerically equal: no formula that equates an energy to a force can be correct. In this unit, you'll learn how to use artificial intelligence (AI) in dimensional analysis, unit conversion, and overall result validation, and why you should make these checks a routine for every account. This unit is the “verification backbone” of the entire module.
Size and unit: basic concepts and the role of AI
Dimension is a type of quantity: length [L], time [T], mass [M]. It has the dimension of speed [L/T], force [M·L/T²], energy [M·L²/T²]. The unit is the scale on which that dimension is measured: length can be metres, feet or light years. AI is helpful in performing dimensional analysis, establishing unit conversions, and checking the dimensional consistency of a formula. But AI can make silent errors in unit conversions — especially in compound units (km/h to m/s, eV to J). Therefore, it is necessary to perform the conversions manually, if possible, with a unit-aware library (for example, Python's pint library; it carries the quantities with their units and controls the conversion itself).
size
Size
SI unit
frequently confused unit
force
M·L/T²
newton (N)
dyne, kgf
Energy
M·L²/T²
joule (J)
eV, calories, kWh
pressure
M/(L·T²)
pascal (Pa)
bar, atm, mmHg
temperature
Θ
kelvin (K)
centigrade, fahrenheit
angle
dimensionless
radian
degree
Step by step: adding size controls to each account
1. Rewrite the size of the desired result. Are you looking for a speed [L/T] or an energy [M·L²/T²]? Knowing the target will instantly ensure a wrong sized result.
2. Make the two sides of the formula dimensionally equal. The right side should be the same size as the left side. If it doesn't, the formula is wrong — without even going into numerical calculation.
3. Convert all inputs to SI, work in a single system. Confusing km and m, hours and seconds in a calculation is the most common physics mistake. Convert everything to SI first.
4. Provide the unit and rank of the result. Is the result in the correct unit? Is the rank physically reasonable?
5. Test limit states. Does the formula give the expected behavior as a variable goes to zero or infinity?
Tip: Even if you don't remember a formula, dimensional analysis often predicts the correct format. For example, if the period of a pendulum depends on length L and gravity g, the only way to obtain dimension [T] is √(L/g) — since [√(L/(L/T²))] = [√(T²)] = [T]. Dimensional analysis does not give the constant factor of the formula (here 2π) but confirms its structure; This is a great way to test a formula given by the AI.
three mini cases
Case 1 — Catching the size error. A student asked the AI for a formula for kinetic energy and got E = ½mv (v instead of v²). He did a size check before calculating the number: the size of ½mv is M·L/T, whereas the energy is M·L²/T². It didn't fit the size. The correct formula was E = ½mv²; dimensional analysis showed the error at a glance.
Case 2 — Silent unit interference. An engineer calculated the time by multiplying the speed of a vehicle in km/h by the distance in m and got an absurd result. While making the calculation, YZ did not convert km/h to m/s. When reinstalled with pint, the library flagged the unit incompatibility from the beginning, and with the correct conversion, the result was reasonable.
Case 3 — eV/joule trap. A teacher asked the AI for the energy of a particle. YZ gave 5 eV, then used it directly like joules in a heat calculation. The teacher remembered the conversion 1 eV = 1.602 × 10⁻¹⁹ J and provided it manually; The AI had skipped a multiplier of 19 orders of magnitude. Once the transformation was corrected, the result was physical.
Four copyable templates
1) Size consistency check:
Check whether the following formula [formula] is dimensionally consistent. Write the size of each variable in [M], [L], [T], find the size of both sides separately and show whether they are equal. If not, indicate which term is problematic.
2) Unit-aware conversion (by pint):
Write a code that performs the following calculation using Python's pint library, defining all quantities with their units: [problem]. Enter the inputs in their units (e.g. km/h, eV), let pints do the conversion, and print the result in the desired SI unit. If there is a unit mismatch, it should give an error.
3) Formula estimation from dimensional analysis:
That physical quantity [target] depends on the following variables: [variables and their dimensions]. Using only dimensional analysis, derive the possible formula form of this quantity (other than a dimensionless constant factor). Also note that dimensional analysis cannot give the constant factor.
4) Providing general results:
Make 4 independent checks (redo the calculation) for the following result: (1) is the unit correct, (2) is the order physically plausible, (3) are the boundary states as expected, (4) is the sign correct. For each, evaluate the result in a sentence. Result: [here]
Weak prompt / Strong prompt
Weak: "Calculate the energy with this formula and give the result."
Result: A number with no unit or size control; Any possible conversion or size error remains hidden.
Strong: "Before calculating the energy with the formula E = ½mv², first satisfy the formula by equalizing the size of the two sides in [M],[L],[T]. Then, for m = 2 kg and v = 36 km/h, first convert v to m/s, do the calculation and evaluate the result with its order in joules."
Result: A reliable result that has passed the dimensional check, its unit is clearly converted, and its rank is ensured.
Common mistakes
- Skipping dimensional analysis. Not using the fastest error catching tool in physics; However, it takes seconds and eliminates the wrong formulas from the beginning.
- Confusing unit systems. Using cm and m, hour and seconds in the same calculation; Everything must first be converted to a single system (SI).
- Risking compound unit conversion manually. forgetting the multipliers in the transition from km/h to m/s, from eV to J; The unit-aware library reduces this risk.
- Radian/degree confusion. Not checking whether the angle is in radians or degrees in trigonometric calculations; The code generally expects radians.
- Not maintaining the rank. Not checking what power of 10 the result is; A major conversion error only appears here.
Note: Dimensional analysis does not necessarily prove that a formula is correct (it may miss a dimensionless constant factor or angle), but it does definitively show that a formula is incorrect. A formula that does not hold dimensionally is absolutely wrong. That's why dimensional analysis should be the first audit of every account; It is cheap and catches a lot of errors. Do not use any formula given by AI without providing the dimension.
In summary
Dimensional analysis and unit verification are the cheapest and most powerful tools for controlling AI output in physics. If the dimensions of both sides of a formula do not match, the formula is incorrect — without calculating any numbers. In unit conversions, especially compound and energy units, AI silently makes mistakes; It is necessary to provide them with a unit-aware library or manually. Checks that should be routine for every account are: rewrite target dimension, make two sides dimensionally equal, convert to SI, ensure unit and order, test limit states. In the next unit, we will cover how to accurately present these validated results with scientific visualization.
Application task
Choose a physics calculation in which at least one unit conversion is required (for example, kinetic energy with a speed in km/h, or a temperature relationship with an energy in eV). First, have the AI provide the size of the formula with the 1st template, then make the unit-aware calculation in pint with the 2nd template and run it. Check if there is a size or unit error. Write down in 5-6 sentences: what error did the dimensional analysis catch or confirm?
checklist
- [ ] I determined the size of the desired result from the beginning.
- [ ] I made the two sides of the formula dimensionally equal.
- [ ] I converted all the entries into a single system of units (SI).
- [ ] I verified compound unit conversions with the library or manually.
- [ ] I physically provided the unit and order of the result.
- [ ] I checked the radian/degree compatibility in trigonometric calculations.