Chemistry · Lesson

Ultrafast, single-molecule and emerging chemical measurements

Separate an underlying molecular population from temporal response, ensemble weighting and finite entity traces.

By the end, you can

Infer a Gaussian width under an explicit convolution model.
Evaluate an assigned exponential population without equating it with a measured signal.
Compare survival-area and initial-slope summaries for two population components.
Distinguish occupied time, dwell counts and independently observed entities.

A rapid event needs a measured observable and a time model

An experiment can measure a changing population, spectrum or other response after a defined perturbation. The observable may report a selected part of the molecular process rather than directly image every bond or electronic state.

A measured delay, a decay time and the instrument’s response width describe different quantities. The definition and timing reference should be explicit before comparing reported times.

Time summaries need a precisely defined response

The original Gaussian case describes independent normalized response functions whose convolution adds variances. Its width relation follows from those shapes. An exponential population instead has a lifetime parameter; replacing one function with the other would change the inference.

Two separate population components can have a mean survival time that differs from the inverse initial fractional loss rate. The first is the integral of normalized survival probability over time. The second depends on the initial slope. Neither is the time centroid of a measured intensity distribution.

Zewail’s original Nobel lecture provides historical primary context for pump-probe study of molecular change. Only the bounded indexed lecture material was accessible here. The assigned widths and populations below are new teaching cases; no lecture figure, instrument performance claim or measured trace is reproduced.

Instrument response can blur the underlying change

A measured time trace can combine an underlying response with the instrument’s temporal response. The convolution model needs its functions, normalization and conditions stated. An observed width is not automatically the underlying process width.

Deconvolution or fitting can estimate parameters under an assumed response model, but noise and alternative models can limit identification. Dense time sampling alone does not prove finer demonstrated temporal resolution.

Changing population weights changes the visible summary

The two-component population graph changes only the initial weights of assigned 100 fs and 900 fs survival components. Both start at total fraction one. Without exchange or temporal blur, their curves and summaries follow directly from the stated model. An actual optical amplitude need not equal a molecular population fraction.

Single-entity dwells can reveal time occupation that an equal count of state visits conceals. The original trace has complete intervals by definition. Boundary censoring, missed events, state-detection thresholds and the representativeness of the sampled entity would matter in real observations.

Model the link between population, observable and instrument before assigning a mechanism. A smooth fitted trace can constrain a stated comparison while leaving other processes, state mixtures or response models plausible.

Initial weights change a two-component survival population

Retained population fraction

Elapsed time / fs

Fast initial fraction 0.250
Fast initial fraction 0.750
— Assigned mixed survival population

Underlying data

Elapsed time / fs	Retained population fraction
0	1 · Fast initial fraction 0.250
200	0.6343868729967592 · Fast initial fraction 0.250
900	0.27594043332960344 · Fast initial fraction 0.250
1800	0.10150146623495446 · Fast initial fraction 0.250
0	1 · Fast initial fraction 0.750
200	0.30168581315666154 · Fast initial fraction 0.750
900	0.0920624176459256 · Fast initial fraction 0.750
1800	0.03383383223163798 · Fast initial fraction 0.750

Original assigned survival components with lifetimes 100 fs and 900 fs. Circles use fast initial fraction 0.250, squares 0.750 at 0, 200, 900, 1800 fs. The dashed curve evaluates the first assignment. There is no component exchange, instrument blur or measured intensity conversion.

Individual traces and an ensemble average can reveal different features

An ensemble measurement combines many constituents, potentially with different states or histories. A single-molecule approach can reveal variations or transitions that an average conceals, subject to its detection and sampling limitations.

A finite trace can miss rare states, and background or missed events can distort inferred dwell times. Repeated detected events, observation time and independent molecular units belong to different counts.

Emerging measurements still need controls and models

A changing signal can be consistent with more than one process. Compare expected spectral, time and control behaviour rather than assigning a mechanism from one fitted exponential alone. Preserve the relationship between raw observations and fitted parameters.

The original traces below are synthetic. They do not establish a real ultrafast mechanism, the performance of an instrument or a complete molecular movie from an invented time curve.

Worked examples

A measured Gaussian width includes the assigned instrument width

Assign normalized Gaussian underlying and instrument responses whose convolution is Gaussian. Measured full width at half maximum (FWHM) is 190 fs and instrument FWHM is 130 fs. Under this fixed model, find the underlying FWHM.

For Gaussian convolution, variances add: σ_meas² = σ_under² + σ_inst². Each FWHM is 2√(2 ln 2) times its σ, so the same squared-width relation applies to these FWHMs.
Underlying width is √(190² − 130²) = √19200 = 138.564064606 fs. It is smaller than the measured width because the assigned instrument response contributes broadening.
These are assigned Gaussian widths, not an instrument calibration or a measured molecular lifetime. A Gaussian width and an exponential population lifetime describe different functions. A different response shape, uncertain instrument width or measurement noise would require a different inference account.

Underlying Gaussian FWHM 138.564064606 fs under the stated convolution model.

An assigned lifetime predicts a population fraction

Assign underlying population P(t)/P(0) = exp(−t/τ), with τ = 400 fs. Find the retained fraction at 250 fs and express this time in picoseconds. No instrument blur or offset is included.

One picosecond is 1000 femtoseconds, so 250 fs = 0.250 ps. The dimensionless exponent is −250/400 = −0.625.
Retained fraction is exp(−0.625) = 0.535261428519. The remaining fraction depends on the ratio of elapsed time to the assigned lifetime.
This underlying population is supplied rather than inferred from a measured signal. A real observable can involve response factors, backgrounds, state mixtures and temporal convolution. A fitted exponential alone does not identify a unique molecular mechanism.

Retained underlying population fraction 0.535261428519 at 0.250 ps.

A mixed population has more than one time summary

Assign separate normalized population components with lifetimes 100 fs and 900 fs, no exchange or instrument blur, and initial fast fraction w = 0.250. Let P(t) = w exp(−t/100 fs) + (1−w)exp(−t/900 fs). Compare the mean survival time with the inverse initial fractional loss rate.

P(0) = 1. The mean survival time of this initial population is ∫₀∞ P(t)dt/P(0) = w(100 fs) + (1−w)(900 fs) = 700 fs. This is an integrated survival probability, not the time centroid ∫tP(t)dt/∫P(t)dt.
The initial fractional loss rate is −P′(0)/P(0) = 0.250/100 + 0.750/900 = 0.00333333333333 fs⁻¹. Its inverse is 300 fs, different from the mean survival time.
At 200 fs, retained fraction is 0.250 exp(−2) + 0.750 exp(−2/9) = 0.634386872997. The two assigned components make this a non-single-exponential trace.
Initial slope and survival area summarize different features. Neither supplies a unique effective lifetime or mechanism for an unknown signal. The weights here are initial population fractions, without an observable-specific amplitude or response conversion.

Mean survival time 700 fs; inverse initial fractional loss rate 300 fs; retained fraction at 200 fs is 0.634386872997.

Counting dwells differs from counting time

Invent one entity trace with three complete on/off pairs. On dwells are [1, 2, 3] s and off dwells [3, 5, 4] s. Assume no missed events or boundary censoring. Find the mean dwells and fraction of observed time in the on state.

On time totals 6 s over three dwells, giving mean 2 s. Off time totals 12 s over three dwells, giving mean 4 s.
The on-time fraction is 6/(6+12) = 1/3. There are three on and three off dwells, so the fraction of recorded dwells that are on is 1/2. Equal dwell counts do not imply equal occupied time.
This finite, invented trace concerns one entity. Its six dwells are not six independently assigned molecular units. A stationary transition-rate interpretation, representative long-time occupancy or uncertainty estimate would need additional assumptions and sampling evidence.

Mean on/off dwells 2/4 s; observed on-time fraction 1/3.

Try it yourself

Change only measured Gaussian FWHM to 100 fs, keeping the fixed 130 fs instrument response. Is a real nonnegative underlying width compatible with this model?
Show the answer and reasoning
No. The squared underlying width would be 100²−130² = −6900 fs², incompatible with the fixed Gaussian convolution assumptions.
A negative squared width is a model conflict, not an imaginary physical width. Changing the response account or examining uncertainty would require new evidence.
Double only assigned underlying lifetime to 800 fs. Find retained population at the same 250 fs.
Show the answer and reasoning
Retained fraction exp(−250/800) = 0.731615628947; elapsed time remains 0.250 ps.
The exponent now uses the smaller elapsed-time/lifetime ratio; the supplied underlying model retains more population.
Change only initial fast fraction to 0.750. Recalculate mean survival time, inverse initial fractional loss rate and retained fraction at 200 fs.
Show the answer and reasoning
Mean survival time 300 fs; initial fractional loss rate 0.00777777777778 fs⁻¹, inverse 128.571428571 fs; retained fraction 0.301685813157.
Use the new initial weights in the survival area, initial slope and population function. The lifetime components are unchanged, while their weighted summaries change.
Change only on dwells to [1, 2, 6] s. Keep off dwells [3, 5, 4] s. Find the means and on-time fraction.
Show the answer and reasoning
On time 9 s, off time 12 s; mean on/off 3/4 s; observed on-time fraction 9/21 = 3/7. There are still three dwells of each kind.
The longer final on dwell changes occupied time, without changing the dwell counts or number of observed entities.

What to carry forward

Instrument response and underlying population need separate definitions.
A Gaussian width and an exponential lifetime summarize different functions.
A mixed population has distinct survival-area and initial-slope summaries.
Finite dwell counts, occupied time and independent entities are different counts.

References

Teaching resources and reference material supporting this lesson’s explanations and examples.