Logarithmic graphs for power and exponential fits
Introduced| English | Meaning |
|---|---|
| intercept/ˌɪntəˈsept/ | The coordinate at which a graph crosses a specified coordinate axis. |
Doubling on a straight line, 1965
- In April 1965, Gordon Moore, later a founder of Intel, predicted that the number of transistors on a chip would keep doubling at a steady rate. Plotted normally, the counts shoot upward. Plotted as $\log N$ against the year, they lie close to a straight line: the Intel 4004 of 1971 had 2,300 transistors, and chips of 2021 had tens of billions. Logarithms turn exponential and power models into straight lines.
- With a straight line, the gradient and intercept 截距 give the constants of the model.
Exponential models: $y = ab^x$
- Take logs: $\log y = \log a + x\log b$.
- Plot $\log y$ against $x$: gradient $\log b$, intercept $\log a$.
Power models: $y = ax^n$
- Take logs: $\log y = \log a + n\log x$.
- Plot $\log y$ against $\log x$: gradient $n$, intercept $\log a$.
Logarithmic graphs for power and exponential fits
For positive observations and parameters, y=ax^p with x>0 gives ln y=ln a+p ln x
State the valid input, plotted variable or time unit before using the logarithmic/exponential relationship.
A plot of $\log y$ against $\log x$ is a line with gradient 1.5 and intercept 0.3. Find $n$ in $y = ax^n$.
Difficulty 2/5 · 140 coins + a small bonus on first completion today
The gradient is $n$.
In the same graph, find $a$ to 3 s.f.
Difficulty 2/5 · 140 coins + a small bonus on first completion today
$10^{0.3} = 1.995$.
Worked example: Moore's law as a line
Transistor counts are modelled by $N = ab^t$, $t$ years after 1971. The line of $\log_{10}N$ against $t$ has intercept 3.36 and gradient 0.150. Find $a$ and $b$, and predict $N$ for 2021.
- Constants: $a = 10^{3.36} = 2290$ (about the 4004's 2,300); $b = 10^{0.150} = 1.41$, so the count grows about 41% a year — doubling every 2 years, since $1.41^2 \approx 2$.
- Predict: $t = 50$: $\log_{10}N = 3.36 + 7.5 = 10.86$, so $N \approx 7.2 \times 10^{10}$.
- Compare: Apple's M1 Max chip of 2021 had about $5.7 \times 10^{10}$ transistors — the same order of size.
A plot of $\log_{10}y$ against $x$ has gradient 0.0414. Find $b$ in $y = ab^x$, to 3 s.f.
Difficulty 2/5 · 140 coins + a small bonus on first completion today
$10^{0.0414} = 1.100$.
Which plot to use
- If $\log y$ against $x$ is straight, the model is exponential.
- If $\log y$ against $\log x$ is straight, the model is a power law.
Match each model to the graph that is a straight line.
Difficulty 2/5 · 140 coins + a small bonus on first completion today
Take logs of each model.
Using natural logarithms
- With $y = Ae^{kx}$: $\ln y = \ln A + kx$, so the gradient of $\ln y$ against $x$ is $k$ directly.
- Base 10 or base $e$ both work; use the same base throughout.
For $y = 5e^{0.3x}$, what is the gradient of $\ln y$ against $x$?
Difficulty 1/5 · 100 coins + a small bonus on first completion today
$\ln y = \ln 5 + 0.3x$.
The gradient of a semi-log graph is $\log b$, not $b$. In the example, the gradient is 0.150, but $b = 10^{0.150} = 1.41$. Likewise, the intercept is $\log a$, so $a = 10^{\text{intercept}}$.
If a semi-log graph has gradient 0.2, then $b = 0.2$ in $y = ab^x$.
Difficulty 1/5 · 100 coins + a small bonus on first completion today
$b = 10^{0.2}$.
For $y = ab^x$, plot $\log y$ against $x$ (gradient $\log b$, intercept $\log a$). For $y = ax^n$, plot $\log y$ against $\log x$ (gradient $n$, intercept $\log a$). Undo the logs to recover the constants, and judge predictions by the range of the data.
Kepler-style data for planets give $\log_{10}T$ against $\log_{10}r$ as a line through $(0, 0)$ and $(0.7, 1.05)$. Find the model $T = ar^n$ and interpret $n$.
A power law from a log–log line.