Sources of Profile Uncertainty

A (Few) Sources of Uncertainty in NDP analysis

Good knowledge of the sample and standards, cleanliness, reproducibility, and using ratios of adequate statistical certainty - These are key attributes to good analyses but there are many ways uncertainty can be introduced into the results.

• Neutron Beam Conditions

Total fluence measured during each spectrum acquisition

This is best done with a monitor prior to sample (inside or outside of chamber)

A fluence monitor should compensate for any change in neutron energy spectrum during and between measurements

Adequate statistical value for fluence

Fluence uniformity (or at least the same pattern) over area of sample being analyzed

(Note: fluence measurement only needs to be a relative number between measurements, not a true neutrons/M2 value

• Electronics

Noise

Variable noise

Detector calibration for the spectrum - Energy/channel

Detector energy resolution

Detector degradation over time with exposure to radiation (signal and noise)

• In the chamber

Adequate vacuum

Contamination of sample/detector from vacuum pump oil

Contamination on sample mask

Detector(s) exposed to excessive/variable radiation levels

• Sample mounting

Geometric variables: Reproducibility of sample position

sample distance from detector

sample's lateral position relative to detector

angle of sample surface to detector surface

Reproducibility in angle formed between sample, detector, and neutron beam

Changing mask/aperture aperture between spectrum acquisition

Mask/aperture orientation on samples - especially if aperture is not round

• Standards

Appropriate standard(s) for sample under study, ideally, but rarely available: It would ...

be of the same material as the unknown (composition, density, distribution, dimensions)

contain a known number of nuclides of the same type that is being studied

(more typically, the best accepted standards are a material from National Labs that contain a known amount of B or ¹⁰B

and measurements are ratio'ed to this "known."

• Sample

Elemental composition (X,Y,Z) needs to be known for depth scale calculation or comparison to other samples

if not spatially uniform, variably changing stopping power in X,Y or Z dimensions

can contain strong neutron absorbers, self absorption

can contain strong neutron scatterers, reaction rate enhancement

volatile components - loss of mass in vacuum

Surface contamination

Surface roughness

NDP references depth from the surface

averages over the aperture/beam defined area (X-Y) making a sharply defined layer look broadened.

Layers – are they parallel with sample surface - otherwise spectrum appears broader

Mass Density – spatial non-uniformity (X,Y,Z) changes depth scale

Isotopic composition – natural or perturbed

Sample area

- aperture may be needed to define sampled area for comparisons to standards and other samples

• Mass density

Variation in mass density through sample (with depth and across area (lateral)

may be due to compositional variation, density variation, voids, etc.

Variation in elemental composition through sample (with depth and across area)

either static variations or more difficult to correct dynamic changes in composition

• Data reduction

Normalization of data to acquisition time (OR more accurately, normalize to the neutron fluence)

Adequate statistics for integrated neutron fluence value

Adequate statistics per channel for sample spectrum

(Actually rigorous statistics are not required per channel, but a sum of channels proportional to the resolution of the detector)

Dead time correction

Neutron integrated fluence normalization

Normalization of data to concentration standard (adequate statistics for each)

Uncertainty in nuclide cross section ratio (if compared to different standard)

Blank subtraction correction (adequate statistics)

Identification of surface channel (ideally to a fraction of a channel)

Uncertainty in stopping power for the sample composition and structure

(Link to the Topic of Spectral Broadening in Specific Detail)

IMPORTANTLY: The easiest and most accurate method to reduce the measurement uncertainty is to conduct the experiment and data processing by normalizing out the various sources of uncertainty. Therefore, actions that can be taken that are made relative to a "known" material/condition/value will cancel out much uncertainty in the final result ... even if the source of uncertainty in the measurement is poorly or not known.