Professional Documents
Culture Documents
net/publication/260823912
Article in Proceedings of SPIE - The International Society for Optical Engineering · June 1998
DOI: 10.1117/12.308758
CITATIONS READS
7 134
1 author:
James E. Potzick
National Institute of Standards and Technology
54 PUBLICATIONS 219 CITATIONS
SEE PROFILE
All content following this page was uploaded by James E. Potzick on 12 January 2015.
While the importance of the concepts of measurement accuracy and measurement traceability
has long been recognized, they have not always been applied in a consistent or rigorous
manner. In 1993 the International Organization for Standards (ISO) published two documents
which establish consistent definitions of these and other metrology terms and provide an un-
ambiguous way to calculate measurement uncertainty. These documents are widely accepted
in the international metrology community. Vendors and buyers of metrology tools and stan-
dards will benefit from the improved communication that results from using this standardized
metrology vocabulary.
Dimensional measurements of importance to microlithography include feature sizes and fea-
ture placement on photomasks and wafers, overlay eccentricities, defect and particle sizes on
masks and wafers, step heights, and many others. A common element in these measurements
is that the object sizes and required measurement uncertainties are often on the order of the
wavelength of light or less. This can lead to interesting challenges for certain applications
where measurement traceability is desirable.
The necessary and sufficient conditions for traceability will be outlined, and some examples
will be given.
Introduction Measurements are an important part base unit may be needed if a part’s function is di-
of product development and manufacturing. They mensionally dependent, for a high confidence level
are needed for developing specifications, for ven- in long term stability, to compare experiment with
dor/buyer communication, and for ensuring that theory, for consistent measurements between dis-
products meet specifications. Measurements are tant manufacturing sites, or for comparing prod-
also needed for process monitoring for either opti- ucts from different manufacturers. It provides a
mization or stability, and for process modeling. basis for comparing measurements that is mutually
All measurements are subject to unknown errors. acceptable to all parties involved in a transaction,
When different measurements are to be compared for resolving differences between buyer and seller,
with each other, such as product measurements and to ensure compliance with legal requirements
those leading to product specifications, they must (such as by government agencies or ISO 9000),
be traceable to a common standard through a chain and for resolving differences between two trace-
of comparisons. This makes the comparison pro- ability chains. It may not be needed, however, for
cess differential and removes or assesses the ef- many manufacturing purposes, such as process
fects of measurement errors. The common stan- monitoring.
dard can be a local artifact standard, a national Measurements of an object’s size (e.g., linewidth)
standard, or a defined base unit (e.g., the meter). and of feature placement (or the center-to-center
Traceability to a national standard or to a defined pitch of two similar objects) are important dimen-
sional measurements in microlithography. This
Proceedings of SPIE 23nd International Symposium on Mi- paper discusses the metrological meaning of mea-
crolithography, vol. 3332-57 (1998). surement traceability as defined by the Internation-
Substrate
FIGURE 2 A Type B uncertainty in measuring linewidth FIGURE 3 Top to bottom edge runout may have a benign
can arise from the fact that the edge of the object is not effect on pitch measurements, but can add serious uncer-
well defined. tainty to linewidth measurements.
ized to unity, Then the measured length could differ from the length
Lstd at some standard temperature Tstd by
∫p(x) dx = 1, - ∞ < x < ∞,
Lmeas = Lstd + ∆T CTE Lstd
then the expectation value of x is Lstd = Lmeas/(1 + ∆T CTE)
〈x〉 = ∫x p(x) dx, ≈ Lmeas(1 - ∆T CTE)
and the variance of the probability distribution is where ∆T = T - Tstd and CTE is the coefficient of ther-
u2(x) = ∫(x - 〈x〉)2 p(x) dx. mal expansion of the substrate. A temperature data set
with a statistical distribution is not available, so the
For example, if one can say only that some error y probability distribution of the temperature must be
is between the bounds ±y0 and no more is known used. If δT is the square root of the variance of the
about it, then the probability distribution is distribution of possible temperatures then the tempera-
ture uncertainty u(T) = δT. The temperature correction
rectangular,
is
p(y) = 1/(2y0) for -y0 < y < y0 ∆L = ∆T CTE Lmeas,
and p(y) = 0 otherwise, and its uncertainty is
2 2
and the variance is u (y) = y0 /3. [Ref 2] u(L)temp = u(T) CTE Lmeas.
A Type A or Type B uncertainty component is the The length is
square root of the variance of the statistical distri- Lstd = Lmeas + ∆L,
bution or probability distribution of its measure- and
ment error respectively. The measurement uncer- U(Lstd) = 2√[σ2 + u(L)scale2 + u(L)temp2 + u(L)other2]
tainty is 2 times the root-sum-square of all of the On the other hand, if a set of temperature measure-
Type A and Type B uncertainty components. If ments Ti had been obtained, then the same procedure
would be used, with T = 〈Ti〉 and u(T) = σ(Ti).
there are k uncorrelated uncertainty components, -6
then the ISO measurement uncertainty is The CTE of quartz is 0.5×10 /°C. If ∆T = (5±2) °C (2σ)
and Lstd = 10 µm, then ∆L = 0.025 nm and U(L)temp =
U = 2√∑uj2, j = 1,...,k. 0.005 nm, which may not be significant, but if Lstd =
The probability distributions of the random and 10 cm, then ∆L = 250 nm and U(L)temp = 50 nm.
systematic errors are treated in the same way. The Sometimes additional measurement uncertainty
only difference is that the random distribution is arises from an inappropriate definition for the mea-
represented by a discrete set of data and the sys- surand (see the definition above for true value) or
tematic distribution is continuous and is inferred imperfections in the artifact being measured. Fig-
from other factors. ure 2 shows an artist’s rendition of a chrome pho-
In the example, suppose the substrate temperature had tomask line as seen in a SEM image. Since there is
not been measured, but it can be estimated to be T±δT. no universally accepted definition for the edge of