semi合集-English.pdf - 第6003页

SEMI P43-0304 © SEMI 2004 12 8.2.6.2 1D metrology t echniques for measuring OPC are not considered within this document’s scope. As a good referenc e the reader is referred to for example Yonekura et al., (T oppan, PMJ20…

SEMI P43-0304 © SEMI 2004 11

the region of interest contains a line-end. Mandatory

information for each of the 4 above:

• actual and nominal width,

• tone, pitch (or surrounding area), and

• orientation.

DEFAULT: isolated feature (line or space).

NOTE 31: As the line-end shape is expected not to be

decisive for its printability, but rather the balance between

area gain and loss, it is recommended to use area difference

for line-end qualification rather than area deviation.

8.2.4.6 line-end pull-back (LEPB) — the distance,

parallel to the line center, between the line-ends of the

nominal and the actual features (see Figure 15). This

distance may be determined along the line center

(center LEPB), or alternatively it may be determined by

the distance between the extreme point of the actual

line and the nominal line-end (minimum LEPB), in

analogy to corner pull-back. Mandatory information is:

• actual and nominal width,

• tone, pitch (or surrounding area),

• orientation, and

• choice of LEPB technique (center or minimum).

DEFAULT: isolated feature.

NOTE 32: Edge roughness may have an important influence

on the line-end pull-back, such that contour averaging may be

necessary to produce a meaningful result. The contour

averaging method is mandatory info for LEPB, if done.

8.2.4.7 equivalent line-end pull-back (ELEPB) —

defined as the negative line-end area difference divided

by the nominal line width (see Figure 16), assuming

accurate 1D control (see Section 8.1).

NOTE 33: This definition actually gives a 1D representation

for a 2D quality assessement, but it is found useful when

comparing mask quality to wafer printing results, which are

typically characterized by 1D measurements, such that a

dimensionless MEEF (mask error enhancement factor) can be

used. As with area based assessment, also this term

disregards the shape at the line-end.

NOTE 34: LEPB and ELES are positive when the actual line

is shorter than the nominal line. A negative sign is added in

case of over-correction, when the actual line becomes longer

than the nominal line.

8.2.5 Specific Case of Contacts and Dots

NOTE 35: This sub-section explains terminology for contacts

in full detail. This terminology can analogously be extended

to dots.

8.2.5.1 contact area — special case of feature area, in

which the feature is a contact (see Figure 18a).

8.2.5.2

contact area gain — special case of feature

area gain, in which the region of interest contains a

contact (see Figure 18b).

8.2.5.3 contact area loss — special case of feature

area loss, in which the region of interest contains a

contact (see Figure 18b).

8.2.5.4 contact area difference — contact area gain

minus contact area loss. As such it becomes a special

case of

feature area difference, in which region of

interest contains a contact.

8.2.5.5 contact area deviation — the sum of contact

area gain and contact area loss. As such it becomes a

special case of feature area deviation, in which the

region of interest contains a contact. Mandatory

information for each of the 4 above:

• nominal width in X and Y (Y not required for

square contacts),

• nominal area (not required for square or

rectangular contacts), and

• pitch (or the surroundings).

NOTE 36: The absolute value of the above qualification

parameters can be normalized to the nominal contact area,

i.e., normalized contact area deviation and normalized

contact area difference.

NOTE 37: As the contact shape is expected not to be decisive

for its printability, but rather the balance between area gain

and loss, it is recommended to use area difference for contact

qualification rather than area deviation.

8.2.5.6 contact X-width (or contact Y-width) — width

in X (or Y) of the smallest rectangle along X (or Y)

encompassing the contact (see Figure18c).

8.2.5.7 contact diagonal widths — widths determined

using the smallest rectangle encompassing the contact

confined along the directions

± arctan(W

Y,nominal

X,nominal

) (see Figure 18d), which is

± 45degrees for square contacts.

NOTE 38: Edge roughness may have an important influence

on the 1D determination of contact width (X-, Y-, diagonal-),

such that contour averaging may be necessary to produce a

meaningful result. The contour averaging method used is

mandatory information for contact width and contact diagonal

width, as its influence is increasingly important for smaller

contacts (see Note 43 in Section 8.5).

8.2.6 Specific Case of Optical Proximity Correction

(OPC)

8.2.6.1 This document recommends to treat OPC’d

features or patterns as a special case of Section 8.2.1

(based on area). OPC fidelity is based on normalized

pattern area deviation and

normalized pattern area

difference.

SEMI P43-0304 © SEMI 2004 12

8.2.6.2 1D metrology techniques for measuring OPC

are not

considered within this document’s scope. As a

good reference the reader is referred to for example

Yonekura et al., (Toppan, PMJ2001, SPIE Proceeding

Vol. 4409 p. 204)

NOTE 39: Edge roughness may have an important influence

on the 1D determination of OPC’d features, such that contour

averaging may be necessary to produce a meaningful result.

(a)

(b)

(c)

(d)

Figure 18

2D Quantification of a contact hole

by contact area,

by area gain and area loss

(dotted line is nominal contact),

by X and Y width (dashed line is smallest rectangle

encompassing the actual contact),

by diagonal width (white dotted lines are the

nominal contact and its diagonals confined by ± α,

where α = arctan(W

Y,nominal

X,nominal

) )

8.2.7 Uniformity of 2D Qualification Parameters

8.2.7.1 In principle all 2D qualification parameters, as

defined above, may vary across the mask. By analogy

to the transition from feature width to feature width

uniformity, any 2D qualification parameter may be

additionally characterized by uniformity. Below, as an

example, corner area difference uniformity is

elaborated.

8.2.7.2 corner area difference uniformity — the spread

of the distribution of the corner area difference of all

mask features selected as described below. To be stated

as mandatory information in addition to that of corner

rounding (Section 8.2.3.1) and also adopting the same

DEFAULTS:

• the criterion used (range, 3-sigma, maximum area

difference, etc.), where sigma stands for standard

deviation.

(Recommendation: before 3-sigma is relevant, the

distribution needs to be “normal” or “near normal”,

and the number of measurements should be > 30.)

• The considered area of interest on the mask.

8.3

Corrected True Values

8.3.1 Some 2D measurements and definitions require

that the mask is on target from a 1D-viewpoint, i.e.,

nominal and actual size coincide for

semi-infinitely

long features present in the region of interest. It is

recommended to limit 2D characterization to features

with approximately the same width, such that 1D-

linearity does not significantly affect the 2D value.

8.3.1.1 As an example, Figure 19 illustrates how a

feature width deviation could affect the line-end.

8.3.1.2 Correction is required for benchmarking

purposes, as described below.

Figure 19

Example where correction of 2D-measurement (here

shown for line-end shortening) for 1D-errors (full

line) provides a different result

8.3.2 2D qualification may be refined by

customer/vendor agreement (or by the user of this

document) by compensating for

• the feature misplacement by alignment (see Section

8.4),

NOTE 40: Pattern alignment is often required and

affects the obtained values. Recommendations on

pattern alignment for 2D qualification will be addressed

in Section 8.4.

• the linewidth deviation (see Figure 20)

b1) by a sizing correction based on the CD mean-

to-target. This correction should be done before

2D-parameter determination,

b2) by a sizing correction based on the feature

mean-to-target of the feature considered. This

correction should be done before 2D-parameter

determination, and

• feature linearity errors (feature proximity errors),

using the same sizing corrections as in 8.3.2b, but

based on specific feature width deviation.

8.3.2.1 What was modified and how it was done, is

mandatory information.

NOTE 41: Scaling is not permitted.

NOTE 42: Sizing algorithm to be mentioned: orthogonally,

by circular brush, perpendicularly, etc. (see Figure 21).

DEFAULT is compensation a and b1.

Figure 20

Illustration of correction for 2D features for 1D-

error, according to techniques described under

Section 8.3.2: b1 (upper) and b2 (lower)

Figure 21

Examples of sizing rules for correction:

from left to right: perpendicularly, by circular

brush, orthogonally

8.4 Alignment

8.4.1 Alignment of nominal and actual feature (or

pattern) is only practical when the ROI includes

features where the actual case is a fairly good

approximation to the nominal one.

8.4.2 Alignment can be done by one or more of the

methods shown below. The result obtained by the

different methods may differ. The method is

mandatory information (see Figure 22).