This article outlines the core competencies required for experimental design, data collection, uncertainty propagation, graphical analysis, and evaluation in IB Physics. It aligns strictly with the assessment criteria and external examination standards.

1. Classification of Experimental Errors and Basic Measurement Uncertainties

Every physical measurement possesses an inherent experimental error, which represents the deviation between the measured value and the true value. Experimental errors are categorized into random errors and systematic errors, each possessing distinct characteristics and remediation methods.

• Random Errors: These cause a statistical scatter of readings around the mean value, affecting the precision (reproducibility) of the measurement. They originate from unpredictable, uncontrollable fluctuations in environmental conditions or observer reading techniques. Random errors can be mitigated by conducting repeated trials and calculating the arithmetic mean.
• Systematic Errors: These cause a constant offset in readings in a continuous direction relative to the true value, affecting the accuracy (closeness to the true value) of the measurement. They are caused by poorly calibrated instruments, zero offsets, or structural flaws in the experimental design. Repeating measurements does not reduce systematic errors; they can only be minimized or corrected via instrument calibration or procedural modifications.

Determining Uncertainty in Single Measurements ($\Delta x$)

• Analog Instruments: The absolute uncertainty is estimated by the experimenter based on the readability of the scale and the experimental environment, typically taken as $\pm 0.5$ or $\pm 1$ of the smallest scale division.
• Digital Instruments: The absolute uncertainty is taken as the smallest readable unit (the resolution) of the digital display.
• Distance and Interval Measurements: For a measurement determined by the difference between an initial and a final position reading (such as a length interval $L = x_2 - x_1$), the absolute uncertainties of both individual readings must be added directly.
  • Standard Case: For a standard millimeter ruler where each position reading has an uncertainty of $\pm 0.5\text{ mm}$, the total absolute uncertainty of the measured interval is the sum of the two individual uncertainties, yielding $\pm 1\text{ mm}$.

Determining Uncertainty in Repeated Measurements

When a variable is measured across multiple independent trials to mitigate random variations, the absolute uncertainty in the mean value ($\bar{x}$) is estimated using half the range of the repeated readings:

$\Delta x = \frac{x_{\text{max}} - x_{\text{min}}}{2}$

Resolution Constraint: If the value calculated by half the range is smaller than the instrumental resolution, the inherent resolution of the instrument must be adopted as the absolute uncertainty.

Quantitative Validation of Theoretical Models

To evaluate the validity of a theoretical relationship, a quantitative comparison must be conducted between the experimental value ($E$) with its absolute uncertainty ($\Delta E$) and the accepted literature value ($A$):

• Model Validated: If the absolute discrepancy between the experimental value and the literature value is less than or equal to the absolute uncertainty:$|E - A| \le \Delta E$This condition is mathematically equivalent to the accepted value $A$ falling within the experimental error bar interval $[E - \Delta E, E + \Delta E]$. The discrepancy is attributed entirely to random uncertainty, validating the theoretical model within experimental limits.
• Model Invalidated: If the absolute discrepancy exceeds the absolute uncertainty:$|E - A| > \Delta E$This indicates that the accepted literature value lies outside the experimental error bounds, confirming either the presence of an uncorrected systematic error or a fundamental limitation of the theoretical model under the given experimental boundary conditions.

2. Propagation of Uncertainties

When processing raw data to calculate a derived physical quantity, the propagation of uncertainties adheres to the worst-case scenario principle. Within the standard examination framework, algebraic linear addition models are utilized rather than calculus-based approaches.

Fundamental Propagation Equations

• Addition and Subtraction ($y = a \pm b$): Absolute uncertainties add directly. $\Delta y = \Delta a + \Delta b$
• Multiplication and Division ($y = \frac{ab}{c}$): Fractional or percentage uncertainties add directly. $\frac{\Delta y}{y} = \frac{\Delta a}{a} + \frac{\Delta b}{b} + \frac{\Delta c}{c}$
• Power Functions ($y = a^n$): The fractional uncertainty is multiplied by the absolute value of the power index. $\frac{\Delta y}{y} = |n| \frac{\Delta a}{a}$

Logarithmic Transformations

For non-linear relationships, equations can be linearized by applying natural logarithms. The uncertainties on logarithmic axes are determined via linear approximation.

• Logarithmic Uncertainty Equation: The absolute uncertainty in a logarithmic quantity $\ln x$ corresponds directly to the fractional uncertainty of its argument $x$: $\Delta(\ln x) = \frac{\Delta x}{x}$

Error Model Uniformity: The IB Physics syllabus enforces a worst-case linear addition model for uncertainty propagation in both written examinations and Internal Assessments (IA). While advanced statistics employs the Root Sum Square (RSS) model for independent random errors, the linear addition model provides a more conservative upper bound and is the exclusive method accepted for assessment.

3. Data Table Standards and Conventions

Data tables must reflect both the resolution of the apparatus used and mathematical consistency during data processing.

• Logarithmic Table Headers: Because the argument of a logarithmic function must be a dimensionless ratio, variables on logarithmic axes or table headers must be divided by their respective units, written strictly as: $\ln(V / \text{V})$ or $\ln(t / \text{s})$. Notations with isolated units, such as $\ln V\ (\text{V})$, are mathematically incorrect.
• Significant Figures and Decimal Alignment: * For raw data, all data points within a single column must be recorded to the same number of decimal places, aligning precisely with the resolution of the measuring instrument.
  • For processed data, the number of significant figures in the calculated result should be equal to, or one additional than, the fewest significant figures in the relevant raw data points, while remaining consistent with precision alignment rules.

Treatment of Anomalous Data

An anomaly (outlier) is a data point that deviates significantly from the trend established by the remaining data set.

• Procedural Standard: Anomalous data points caused by an identifiable human error or instrumental malfunction during collection must be omitted before calculating the mean value or plotting lines of fit. However, a data point must never be deleted purely because it does not align with the theoretical model, as it may indicate an unmodeled physical phenomenon or a systematic limitation of the apparatus.

4. Linearization of Non-Linear Equations

Linearization Principle

Non-linear physics equations must be rearranged into the standard straight-line form $Y = mX + C$, where compound variables are plotted on the axes to extract physical constants from the gradient ($m$) or vertical intercept ($C$).

Linearization of Common Physical Models

Equation	Dependent Variable (Y-Axis)	Independent Variable (X-Axis)	Gradient ($m$)
$T = 2\pi\sqrt{\frac{L}{g}}$	$T^2$	$L$	$\frac{4\pi^2}{g}$
$P = \frac{k}{V}$	$P$	$\frac{1}{V}$	$k$
$v^2 = u^2 + 2as$	$v^2$	$s$	$2a$

*Note: For the kinematical model $v^2 = u^2 + 2as$, the vertical intercept represents $u^2$. For the other two proportional models, the theoretical vertical intercept is strictly zero.

Error Bars for Derived Quantities

When plotting calculated or compound variables, the lengths of the error bars must be determined via fractional propagation rules prior to plotting:

$$\begin{align*} \frac{\Delta(T^2)}{T^2} &= 2\frac{\Delta T}{T}\\ \frac{\Delta(1/V)}{1/V} &= \frac{\Delta V}{V} \end{align*}$$

5. Graphical Analysis: LOBF and LOWF

Error Bars and Error Boxes

Graphs must include horizontal ($\pm \Delta x$) and vertical ($\pm \Delta y$) absolute uncertainty error bars for each data point, defining a two-dimensional uncertainty region (Error Box) around the coordinates.

Lines of Best and Worst Fit

• Line of Best Fit (LOBF): A single straight line or curve drawn to pass evenly through the error boxes of all data points, representing the underlying physical trend with random variations smoothed out.
• Lines of Worst Fit (LOWF): The two limiting straight lines drawn through the error boxes to quantify the experimental uncertainty in both the gradient and the intercept. They represent the steepest acceptable gradient ($m_{\text{max}}$) and the shallowest acceptable gradient ($m_{\text{min}}$) that can still pass through the majority (or all) of the error boxes. Do not mechanically force these lines to connect the extreme corners of the first and last data points, as localized random anomalies can distort the true boundary limits.

Calculating Parameter Uncertainties

$\Delta m = \frac{m_{\text{max}} - m_{\text{min}}}{2} \quad \text{and} \quad \Delta c = \frac{c_{\text{max}} - c_{\text{min}}}{2}$

Critical Constraint: The intercept uncertainty $\Delta c$ must be calculated algebraically using coordinates ($c = y - mx$), never estimated visually from a truncated or broken axis.

6. Data Rounding Conventions

Significant Figure and Decimal Alignment

• Uncertainty Rounding: Calculated absolute uncertainties ($\Delta m, \Delta c$) must be rounded to one significant figure using standard rounding rules. If the leading digit of the uncertainty is 1, it is acceptable to retain two significant figures.
• Precision Alignment: The final values of the parameters ($m, c$) must be rounded so that their last significant digits align strictly with the decimal place value of their corresponding absolute uncertainty.

Intercept Evaluation for Systematic Shifts

If a theoretical model predicts a proportional relationship ($y=kx$), but the calculated experimental vertical intercept range $c \pm \Delta c$ does not include zero, a systematic error is mathematically confirmed.

Analysis of Non-Linear Graphs

• Instantaneous Rate: Extracted by calculating the gradient of a geometric tangent line constructed to touch the curve evenly at that specific continuous point.
• Rate Uncertainty: Determined by constructing the steepest and shallowest possible acceptable tangents that touch the data point without crossing the curve.

7. Experimental Methodology: Variable Management

A valid scientific investigation requires an explicitly detailed variable control structure:

• Independent Variable: State precisely how the independent variable is altered, requiring at least 5 distinct values across a defined physical range to establish a mathematically reliable relationship.
• Quantification of Qualitative Variables: Qualitative properties must be converted into quantifiable, measurable metrics (e.g., replacing descriptive labels like “rough/smooth” with “sandpaper grit size rating”).
• Dependent Variable: Specify the apparatus used to measure the dependent variable, its resolution, and the number of repetitions (minimum of 3 independent trials per independent value) to calculate a mean and address random errors.
• Controlled Variables: General descriptions like “keep temperature constant” are insufficient. The exact physical method of control must be specified (e.g., “submerging the container in a thermostatically controlled water bath and monitoring with a digital thermometer”).
• Limitations of Control Variables: In complex physical systems, students must identify the single most critical uncontrolled variable, assess its quantitative impact on the dependent variable, and suggest a physical mechanism to actively stabilize it.

8. Safety and Protocol Improvements

Risk Assessment Structure

Safety and environmental statements must follow a clear causal chain:

$\text{Hazard} \rightarrow \text{Associated Risk} \rightarrow \text{Specific Control Measure}$

• Standard Case: High-voltage laser (Hazard) $\rightarrow$ Permanent retinal damage (Associated Risk) $\rightarrow$ Wear certified laser safety goggles and insert a matte beam dump block (Specific Control Measure).

Specificity of Experimental Improvements

Suggested modifications must directly address the specific physical mechanism causing the systematic error, rather than merely proposing the use of more precise instruments:

• Thermal Experiments: Add an insulation jacket and apply a cooling curve correction to determine and add back lost heat energy.
• Mechanical Experiments: Replace manual stopwatches with photogates and digital data loggers to eliminate human reaction time error ($\pm 0.2\text{ s}$).
• Electrical Experiments: Use a pulsed-current switch to restrict current flow to active measurement intervals, preventing resistance drift due to Joule heating.

Appendix A. Mathematical Foundations of Uncertainty Propagation (Advanced Reference)

Note: This section outlines the multivariable calculus origin of the linear propagation rules used in IB Physics. It serves as theoretical depth for Internal Assessments (IA); partial differentiation is not tested in written examinations.

In a multivariable physical system where a derived quantity $y$ depends on multiple independent variables, $y = f(x_1, x_2, \dots, x_n)$, the total differential $dy$ represents the net change in the function due to minor variations in its components:

$dy = \sum_{i=1}^{n} \frac{\partial f}{\partial x_i}dx_i$

In experimental physics, the differentials $dx_i$ correspond to the absolute uncertainties $\Delta x_i$. To establish the most conservative upper boundary for error limits, the IB framework assumes a worst-case scenario where all individual measurement errors occur in the same direction simultaneously (perfect positive correlation). To prevent opposing terms from canceling each other out, the absolute value of each partial derivative term is taken, yielding the linear propagation formula:

$\Delta y = \sum_{i=1}^{n} \left| \frac{\partial f}{\partial x_i} \right| \Delta x_i$

Derivation of Standard Rules:

• Addition ($y = a + b$):

  $\frac{\partial y}{\partial a} = 1,\quad \frac{\partial y}{\partial b} = 1 \implies \Delta y = \Delta a + \Delta b$

• Division ($y = \frac{a}{b}$):

  $dy = \frac{1}{b}da - \frac{a}{b^2}db$

  Dividing both sides by $y$ yields:

  $\frac{dy}{y} = \frac{da}{a} - \frac{db}{b}$

  Applying absolute values for the worst-case scenario results in the linear fractional uncertainty equation:

  $\frac{\Delta y}{y} = \frac{\Delta a}{a} + \frac{\Delta b}{b}$

In formal research, independent random errors are propagated using the Root Sum Square (RSS) model based on a Gaussian distribution:

$\Delta y_{\text{RSS}} = \sqrt{\sum_{i=1}^{n} \left(\frac{\partial f}{\partial x_i}\Delta x_i\right)^2}$

The RSS model accounts for the statistical probability that random errors will partially cancel one another out. However, to maintain algebraic simplicity and a uniform standard of conservative error estimation, the IB Physics syllabus enforces the linear addition model.

Show replies