The Gold Standard of Spectroscopy | Equal Peak Area, Different Peak Heights

Two spectral peaks can have the same total baseline-corrected area and still look very different on a graph. This lesson puts numbers to the difference between peak height, Gaussian width and integrated response.

Chapter index

Chapters 1–3

1. Decide which spectral quantity you are comparing · 2. Specify the Gaussian model and its units · 3. Choose two peaks with the same area

Chapters 4–6

4. Check the identity rather than trusting the chart · 5. Reverse the question: equal height is not equal area · 6. Separate physical width from instrumental broadening

Chapters 7–9

7. Understand baseline removal and integration windows · 8. Distinguish an isolated peak from overlapping signals · 9. Write an audit trail for a peak comparison

Chapters 10–12

10. Answer questions readers frequently ask · 11. Practise with a different width and a finite window · 12. Connect the result to wider spectroscopy and measurement

A spectrum is not a single number. A line has a position along its measurement axis, a profile, an observed width, a maximum signal above background and an integrated signal across a chosen range. These quantities answer different questions.

This chapter compares two ideal Gaussian peaks with exactly equal integrated areas and different widths. Their unequal heights reveal why a graph’s tallest feature need not represent the largest integrated response. Peak height and area have different dimensions when the horizontal axis carries a physical unit.

The examples are original mathematical teaching cases, not spectra of identified chemicals or instruments. Whether an experimental peak area can estimate concentration or intensity depends on the measurement technique, detector response, calibration, background and the physical question.

Contents · Next chapter

Let the baseline-corrected signal be y(x)=H exp[−(x−μ)²/(2σ²)]. The centre is μ, H is the height above baseline, and σ is the standard-deviation width expressed in units of the horizontal axis.

If y is measured in detector counts and x in nanometres, the peak height has count units while the area has count·nanometre units. Integrated area is not merely a taller or shorter version of the height: it multiplies signal by a range along the axis.

For a peak integrated across the full real line, A=∫y(x)dx=Hσ√(2π). The full width at half maximum is FWHM=2√(2ln2)σ≈2.35482σ. These results depend on the ideal Gaussian profile; other line shapes have other width–area relations.

Contents · Previous chapter · Next chapter

Define both ideal peaks to have integrated area A=100 count·nm and the same centre. Let Peak Narrow have σ=1 nm. Its height must be H=100/(1×√(2π))≈39.89 counts. Its FWHM is approximately 2.355 nm.

Let Peak Broad have σ=2 nm. Its height must be H=100/(2×√(2π))≈19.95 counts. Its FWHM is approximately 4.710 nm.

Peak Narrow is twice as tall as Peak Broad, but both have identical integrated area by construction. The broad peak spreads its response across more of the horizontal axis. A graph that ranks these two peaks by their maxima would tell a different story from a calculation that compares their areas.

Contents · Previous chapter · Next chapter

For Peak Narrow, A≈39.894×1×2.50663=100 count·nm. For Peak Broad, A≈19.947×2×2.50663=100 count·nm. The small difference in a displayed result comes from rounding the heights, not from a model difference.

Their heights have ratio 2:1, while their widths have ratio 1:2. Because area in the Gaussian model is proportional to Hσ, the two products are equal. This illustrates a conservation of area under an exact re-scaling of the shape.

Do not extrapolate the result to every observed line. An instrument may truncate a peak, saturate its centre or treat overlapping bands as one feature. Observed numerical integration can then depart from this ideal calculation.

Contents · Previous chapter · Next chapter

Now give both peaks the same height H=40 counts, while retaining σ=1 nm for the first and σ=2 nm for the second. Their respective areas become approximately 100.27 and 200.53 count·nm.

Identical maxima no longer imply identical integrated responses. Peak Broad now contains twice the ideal area because its width is twice as large. Neither the height nor the width can determine an integral by itself without a shape model or an actual integration.

When students are shown a chromatogram, emission spectrum or fitted optical line, ask whether a number called ‘intensity’ refers to a maximum, an area, an integrated flux, a count rate or some other defined observable. Precision begins with the quantity and its units.

Ideal Gaussian quantity	Narrow	Broad
σ	1 nm	2 nm
FWHM	2.355 nm	4.710 nm
Height	39.89 counts	19.95 counts
Full area	100 count·nm	100 count·nm

Contents · Previous chapter · Next chapter

Spectral lines can broaden because of physical processes and because instruments have finite resolution. NIST’s atomic-spectroscopy reference distinguishes line-shape processes including Doppler and pressure broadening.

For a Gaussian physical line convolved with an independent Gaussian instrument response, variances add: σ_observed²=σ_physical²+σ_instrument², if both profiles use consistent units and the instrument response is normalised. This simple rule is not valid for every non-Gaussian broadening model.

A normalised convolution preserves full integrated area when the processing introduces no signal loss and the entire result is integrated. It changes peak height and usually width. Calibration and resolution must therefore be considered before interpreting a difference in plotted maxima.

Contents · Previous chapter · Next chapter

A real spectrum may have a background level, drift, continuum or nearby signal. If the true Gaussian peak has height H above a flat baseline b, its raw maximum is H+b. Reading the raw maximum as the peak height includes the background contribution.

Integrating a constant baseline across a finite window of width W adds an unwanted amount bW to the peak area. With b=5 counts and W=10 nm, that contribution is 50 count·nm. Background subtraction is not a cosmetic choice: it changes the numerical result.

Even after background correction, a finite integration window may exclude more of a broad peak than a narrow one. Our equal-area proof concerns the complete Gaussian integral. Define the window and explain how much of the line is captured before comparing measured areas.

Contents · Previous chapter · Next chapter

Suppose two peaks overlap substantially. The highest point of their sum may no longer equal the maximum of either component. A fitted component area depends on a model for centres, widths, baseline and additional peaks.

Overlapping profiles can produce multiple plausible decompositions when the data lack resolution. Narrow-looking features are not automatically separate species, and a numerical fitting algorithm cannot create information that the instrument did not measure.

NIST’s work on reproducible mass-spectral peak integration makes clear why identification and integration need defined operations. The present Gaussian calculations teach the mathematical distinction, not a universal peak-detection algorithm.

Contents · Previous chapter · Next chapter

A defensible comparison names the horizontal axis and signal units, states whether a baseline has been subtracted, identifies the integration bounds and records the width convention. If a fitted profile was used, report its type and the treatment of residuals.

When comparing instruments, also verify resolution, response calibration, saturation, acquisition settings and whether the same physical quantity is being measured. Equal mathematical area does not guarantee equal chemical concentration or identical material behaviour.

The practical standard is to make the reported number reproducible from the stated data and method. Two observers should know whether they are checking an observed maximum, a model-fitted height or an integrated, baseline-corrected response.

Contents · Previous chapter · Next chapter

Is a taller peak always a larger amount? No. Our ideal broad and narrow peaks have equal area and different heights. Converting any chosen feature into an amount also requires a suitable technique and calibration.

Does instrument broadening always conserve area? The full integral is preserved under a normalised linear convolution without losses. Truncated windows, nonlinear detection or incorrect baselines can break that practical comparison.

Is FWHM the same as σ? No. In a Gaussian, FWHM≈2.35482σ. Other profiles use different relationships.

Can area be negative? A baseline-corrected numerical integral may be negative through noise or modelling choices. That does not mean the experiment measured a negative physical amount.

Contents · Previous chapter · Next chapter

A third Gaussian peak has A=75 count·nm and σ=1.5 nm. Calculate its height: H=75/(1.5√(2π))≈19.95 counts. Its FWHM≈3.532 nm. Notice that it has about the same height as our Broad 100 count·nm peak, but its integrated area is smaller.

Now consider two equal-area Gaussians integrated only from μ−2 nm to μ+2 nm. The narrow σ=1 nm line lies within two standard deviations on each side, capturing approximately 95.45% of its full area. The broad σ=2 nm line lies within one standard deviation, capturing approximately 68.27%.

For A=100, the finite-window areas are therefore about 95.45 and 68.27 count·nm. The original full areas are still equal. The reporting window has changed the measured comparison.

Contents · Previous chapter · Next chapter

Continue with The Gold Standard of Spectroscopy, the subject root for interpreting radiation–matter measurements, line positions and signal evidence. For broader optical concepts visit The Gold Standard of Photonics and The Gold Standard of Atomic Physics.

For authoritative line-shape context consult NIST on spectral line shapes and NIST on mass-spectral integration. They do not supply the invented peaks in this guide.

The common reasoning is to keep measurement definitions attached to their numbers. Return to The Gold Standard hub or How X Works for related subject routes.

Contents · Previous chapter