WO2012114147A1

WO2012114147A1 - Method for detecting defects in a microscopic scale on a surface of a sample, and device implementing this method

Info

Publication number: WO2012114147A1
Application number: PCT/IB2011/001761
Authority: WO
Inventors: Peter Choi
Original assignee: Nano-Uv
Priority date: 2011-02-23
Filing date: 2011-02-23
Publication date: 2012-08-30
Also published as: WO2012114147A9

Abstract

The invention concerns a method for detecting defects on or under a surface of a sample (3), comprising : measuring a signal during a first scan of the surface through a first optical circuit (17, 20), e.g. dark-field scan; detecting, from the signal measured during the first scan, at least one undifferentiated event corresponding to a false defect due to measurement noise or a real defect on or under the scanned surface of the sample; measuring a signal during a second scan of the sample through a second optical circuit (23, 24), i.e. light-field scan; and identifying, among the detected undifferentiated events, and from the signal measured during the second scan, which undifferentiated events correspond to a real defect and which undifferentiated events correspond to a false defect. The invention also relates to a device (2) implementing this method.

Description

« Method for detecting defects in a microscopic scale on a surface of a sample, and device implementing this method »

Technical field

The present invention relates to a method for detecting defects, rably in a microscopic scale, on a surface of a sample.

The invention also relates to a device implementing this method . State of the Art

Method for detecting defects on a surface at a microscopic scale are already known . One major problem for scanning a surface and detecting defects on this surface is shot noise.

Shot noise is a type of noise occurring when a finite number of particles carrying energy (such as electrons or photons) is small enough to give rise to detectable statistical fluctuations in a measurement.

In optics, for a detector generating a signal (electrical current) as a function of the amount of received light, the magnitude of this noise increases with the average magnitude of the current or light intensity. However, since the magnitude of the average signal increases more rapidly than that of the shot noise, shot noise is generally only a problem with small currents or light intensities.

The intensity of a light source yields the average number of photons collected by the detector. The actual number of photons collected by the detector will be more than, equal to, or less than the average, and their distribution about that average will be typically a Poisson distribution . Since the Poisson distribution approaches a normal distribution for large numbers, the photon noise in a signal will approach a normal distribution for large numbers of photons collected . The standard deviation of the photon noise is equal to the square root of the average number of photons. The signal-to- noise ratio (SN R) is then equal to , where N is the average number of

photons collected . When N is very large, the signal-to-noise ratio is also very large. Photon noise becomes more important when the ι photons collected is small .

Thus, to detect a microscopic defect on a surface, the one the art considers that in "good" optical measurement conditions of illumination by the source and/or high power of illuminati source), any signal greater than 2^N^~ has a very high pro correspond to a real defect and not to shot noise.

Nevertheless, these "good" conditions are time consumi energy consuming .

The goal of the invention is to present a method for detecti on a surface of a sample, the method according to the invention time consuming and/or energy consuming than prior art.

Summary of the Invention

The first scan preferably scans all the surface of the sample and the second scan does preferably not scan all the surface but only regions corresponding to the undifferentiated events.

The second optical circuit can be integral with the first optical circuit. The second scan can be slower than the first scan .

The second scan can have a higher resolution power than the first scan .

The sample can be on a same platform during the first and second scans, the sample having not moved from its platform between the first scan and the second scan .

For an average measurement signal equal to Ni, where Ni is the average number of photons collected per pixel during the first scan, every event having an amplitude higher than 2_y[N \N\th respect to this average signal can be identified as being an undifferentiated event.

For an average measurement signal equal to N₂, where N₂ is the average number of photons collected per pixel during the second scan, every event having an amplitude inferior to 2_A/N ^{ca n} be identified as being a false defect and every event having an amplitude superior to 2_A/N ^can be identified as being a true defect, preferably in a dark field scan .

During the first scan, a pixel area can be chosen to be larger than a smallest size of defect.

During the second scan, a pixel area can be chosen to be smaller than a smallest size of defect.

The signal measured during the first scan and the signal measured during the second scan can be measured via a same detector.

The sample can be a mask blank or a patterned mask.

The illumination light can comprise photons during both first and second scans. In this case :

- the sample can be illuminated with EUV light during the first scan and during the second scan, and/or

- the sample can be illuminated with light during the first scan and during the second scan, but the wavelength of the illumination light during the first scan can be different from the wavelength of the illumination light during the second scan. Illumination can comprise photons during the first scan and electrons during the second scan.

If illumination light comprises photons during the first or the second scan, the detected signal can be the back scattered or reflected photons.

If the detected signal comprises photoelectrons, then these photoelectrons are preferably used to create an image of the sample, i.e. to do photoemission electron microscopy (PEEM).

If illumination comprises electrons during the second scan, then these electrons are preferably used to create an image of the sample using a raster scanning process, such as in the well known principle of a scanning electron microscope.

The sample can be illuminated with photons during the first scan and illuminated with electrons during the second scan. An other aspect of the invention concerns a device for detecting defects (preferably in a microscopic scale) on or under a surface of a sample, comprising

- means for measuring a signal during a first scan (for example bright field scan or dark field scan) of the surface through a first optical circuit, - means for detecting, from the signal measured during the first scan, at least one undifferentiated event corresponding to a false defect due to measurement noise or a real defect on or under the scanned surface of the sample,

characterized in that it further comprises:

- means for measuring a signal during a second scan of the sample through a second optical circuit, and

- means for identifying, among the detected undifferentiated events, and from the signal measured during the second scan, which undifferentiated events correspond to a real defect and which undifferentiated events correspond to a false defect.

The first scan can be a dark-field scan.

The second scan can be a bright-field scan. The first scan can be a scan of all the surface and the second scan can be not a scan of all the surface but is only a scan of the regions corresponding to the undifferentiated events.

The second optical circuit can be integral with the first optical circuit. The device according to the invention can be arranged so that the second scan is slower than the first scan .

The device according to the invention can be arranged so that the second scan has a higher resolution power than the first scan .

The device according to on the invention can comprise a single platform arranged for being integral with the sample during the first and second scans, so that the sample does not move from its platform between the first scan and the second scan .

The identifying means can be arranged so that for an average measurement signal equal to Ni, where Ni is the average number of photons collected per pixel during the first scan, every event having an amplitude higher than

respect to this average signal is identified as being an undifferentiated event.

The identifying means can be arranged so that for an average measurement signal equal to N₂, where N₂ is the average number of photons collected per pixel during the second scan, every event having an amplitude inferior to 2^Ή is identified as being a false defect and every event having an amplitude superior to 2^Ή is identified as being a true defect.

The device according to the invention can be arranged so that during the first scan, a pixel area is larger than a smallest size of defect.

The device according to the invention can be arranged so that during the second scan, a pixel area is smaller than a smallest size of defect.

The device according to the invention can comprise a single detector for measuring the signal measured during the first scan and the signal measured during the second scan .

The sample can be a mask blank or a patterned mask.

The device according to the invention can comprise an illumination source for illuminating the sample during the first scan and during the second scan, the illumination source emitting preferably light having a wavelength comprised between 1 nanometres and 120 nanometres, that is preferably:

Extreme Ultraviolet light (also called EUV light) typically from 50 nm down to 5 nm wavelength, preferably around 13.5 nm wavelength, or

Soft X-rays light, typically from 5 nm down to 2 nm wavelength, preferably around 3 nm wavelength .

The device according to the invention can comprise an illumination source for illuminating the sample during the first scan and an other illumination source for illuminating the sample during the second scan . The illuminations sources can be arranged such that illumination light comprises photons during both first and second scans. In this case, the illuminations sources can be arranged such that:

- the sample is illuminated with EUV light during the first scan and during the second scan, and/or

- the sample is illuminated with light during the first scan and during the second scan, but the wavelength of the illumination light during the first scan can be different from the wavelength of the illumination light during the second scan .

The illuminations sources can be arranged such that illumination comprises photons during the first scan and electrons during the second scan .

If the detected signal comprises photoelectrons, then these photoelectrons are preferably used to create an image of the sample, i .e. to do photoemission electron microscopy (PEEM).

If illumination comprises electrons during the second scan, then the device according to the invention can comprise means for using these electrons to create an image of the sample, using a raster scanning process, such as in the well known principle of a scanning electron microscope.

The device according to the invention can comprise : means for using photons as a source of illumination of the sample for the first scan with an associated optical circuit for illumination and detection, and/or

means for using electrons as a source of illumination of the sample for the second scan with an associated optical circuit for illumination and detection, typically implementing the structure of a scanning electron microscope

Detailed description of the figures

and of realization modes of the invention

Other advantages and characteristics of the invention will appear upon examination of the detailed description of embodiments which are in no way limitative, and of the appended drawings in which :

- Figure 1 illustrates a first scan of a surface of a sample implemented in a best realization mode of a method according to the invention,

- Figure 2 illustrates a second scan of a surface of the same sample implemented in a method according to the invention,

- Figure 3 is a schematic side view of a first embodiment of a device 1 according to the invention implementing a method according to the invention,

- Figure 4 is a schematic side view of a second embodiment (best realization mode) of a device 2 according to the invention implementing a method according to the invention, and

- Figures 5 and 6 illustrate two examples of samples and their associated real defects.

We are now going to describe, in reference to figures 1 to 4, a best realization mode of a method according to the invention for detecting defects in a microscopic scale on or under a surface 35 of a sample 3.

By "microscopic scale", it is meant that the invention allows detection of a defect having microscopic or even sub-microscopic dimensions.

This method according to the invention, for detecting defects in a microscopic scale on or under a flat surface 35 of a sample 3, comprises: - measuring a signal during a first scan of the surface through a first optical circuit, this signal being generated by light coming from the scanned surface during the first scan,

- detecting, from the signal measured during the first scan, at least one undifferentiated event corresponding to :

° a false defect due to measurement noise, or

° a real defect on or under the scanned surface of the sample,

- measuring a signal during a second scan of the sample through a second optical circuit, this signal being generated by light coming from the scanned surface during the second scan, and

- identifying, among the detected undifferentiated events, and from the signal measured during the second scan, which undifferentiated events correspond to a real defect and which undifferentiated events correspond to a false defect.

The detecting step is implemented by a unit 25.

The identifying step is also implemented by unit 25.

Unit 25 comprises a microprocessor, and/or an analogical electronic circuit, and/or a digital electronic circuit, and/or a computer.

By "real" defect, it is meant in references to figures 5 and 6 :

- a dust or any other foreign material 27 on this surface 35, or a any defect with respect to the flatness or the smoothness of the surface (for example a scratch or cavity 29 on the surface of the sample, or a bump 28 of material constituting the sample); in this case, the real defect is an amplitude defect; or

- a dust or any other foreign material 31 under this surface 35, typically at an interface 32 between two layers of the sample, or a any defect with respect to the flatness or the smoothness of an interface 32 between two layers of the sample (for example a scratch or cavity 34, or a bump 33 of material constituting the sample) even if the scanned surface 35 of the sample is perfectly flat without any defects; in this case, the real defect is a phase defect.

The sample is typically:

- a mask blank for lithography, or - a patterned mask for lithography.

During the first scan, the sample is illuminated with illumination light 4. The first scan is advantageously a dark-field microscopy scan. By "dark field scan", it is meant that the signal measured during the first scan comes only from light 5 scattered from the sample, while the reflection 6 (in the case of an inverted microscope, as illustrated in figures 3 and 4) or direct transmission 6 (in the case of a non inverted microscope) of the illumination light is omitted to produce the measured signal. In other words, only light 5 scattered from the sample goes on a detector 7 to produce the measured signal, while the reflection 6 or direct transmission 6 of the illumination light is omitted from this detector. Dark field microscopy is a technique well known by the one skilled in the art. A dark field scan is advantageous for the first scan, because it offers a better signal to noise ratio.

The first scan takes place as follows. During the first scan, all the surface Mi of the sample 3 is scanned :

il) a part 9 of the surface of the sample is illuminated with illumination light 4, and this illuminated area 9 of the sample is imaged on the detector 7 during a time t_Ai, Ai being the value of the area 9 of the sample imaged on the detector 7 during time t_Ai, then

iil) an other part 9' of the sample is illuminated with illumination light, and the illuminated area 9' of the sample is imaged on the detector 7 during time t_Ai, Ai being the area of the sample imaged on the detector during step iil) being iterated as many time as necessary in order to illuminate and image all the surface Mi of the sample part by part.

Each step iil) is implemented after moving the platform 10 and the sample 3 with respect to the first optical circuit.

In the illustrated case of figure 1, step iil) is implemented five times. Nevertheless, figure 1 is only a schematic view to simply illustrate the invention. Step iil) is in fact typically implemented hundreds of time.

The detector 7 is a matrix detector, comprising an array of ni*pi pixels. This way, the area Ai imaged on the detector 7 at a same instant is split into ni*pi pixels 11. In figure 1, Ai is split into nine pixels: figure 1 is only a schematic view to simply illustrate the invention; in fact, Ai is typically split into millions of pixels. During the first scan, the value Di of the area of each pixel 11 in the plan of the scanned surface of the sample is chosen to be larger than a smallest size D of defect 12 searched for: D_x > D .

Let's consider that:

- Ai (area 9 with the diagonal lines in figure 1) is the value of the area of the surface part imaged on the detector (unit: square metre); it is typically comprised between 0.04· 10^"7 m² and 4· 10^"7 m²,

- Mi is the value of the total area of the surface of the sample scanned during the first scan, this surface being continuous (unit: square metre) ; it is typically comprised between 0.001 m² and 0.1 m²,

- Di (area 11 with crossing diagonal lines in figure 1) is the value of the area of each pixel 11 during the first scan, measured in the plan of the scanned surface (unit: square metre); it is typically comprised between 0.02-10^"13 m² and 2-10^"13 m²,

- NAI is the amount of incident photons, i.e. the amount of photons from the illumination light, that are incident on the area Ai; it is typically comprised between 4 · 10¹⁰ and 1200 · 10¹⁰,

- Ni is the average number of photons collected per pixel during the first scan; it is typically comprised between 3· 10³ and lOOO-10³ ,

- R is the reflectivity of the surface of the sample; it is typically comprised between 0.3 and 0.8,

- D is the minimum size of a defect that we want to detect (unit: square metre) ; it is typically comprised between 0.04· 10^"16 m² and 4· 10^"16 m², - ni and pi are the number of pixels in respectively a row or column of the array of pixels of the detector 7; each one of ni and pi is typically comprised between 1 · 10⁶ and 20 · 10⁶ ,

- t_Ai is the illumination time of the area Ai during which area Ai is imaged on the detector 7 in each step il) and iil) (unit: seconds) ; it is typically comprised between 0.01 · 10^"3 s and 1 · 10^"3 s,

- ti is total time of the first scan, i.e. the illumination time of the area Mi during which area Mi is imaged on the detector part by part (unit: seconds) ; it is typically comprised between 150 s and 15000 s, - Pi is the irradiance of the illumination light 4 measured on the scanned surface of the sample during the first scan (unit: Watt per square metre) ; it is typically comprised between 1 · 10⁴ W/m² and 100 · 10⁴ W/m² ,

- h is the Planck constant; (6,62606896x l0^"34 J .s)

- vi is the frequency of the photons of the illumination light during the first scan, in the case wherein the illumination light is monochromatic light (unit: Hertz) ; it is typically comprised between 0.2 · 10¹⁶ Hz and 20 · 10¹⁶ Hz,

For the first scan :

The first scan is implemented in very unusual conditions : the first scan is implemented in "unfavorable" conditions for which the probability to determine whether a high signal value of a pixel compared to the mean value of the signal is due to a real defect or to noise is less than 80% . At the same time, this "unfavourable" scan conditions allow complete scanning of the sample in a short time, for example less than one hour for a typical mask blank substrate with dimensions 120 mm x 120 mm .

For the first scan, the following experimental condition is used :

> — , preferably 2 — < ' Al

A_l t_Al R D^z t_l n_x p_x ^{' '} R D^l t_x n_{x x} Α_γ t_Al

or

M, M

R > ¹ hv, , preferably 2 hv, < R < 10_-_^ hv_y

R D^l t_l n_x p_l ^{1 1 '} R D^l ί_γ η_γ p_l ¹ R D t_l η_γ p_l

For an average measurement signal equal to N i, where N i is the average number of photons collected per pixel during the first scan, every event corresponding to a pixel 11 for which the measured signal has an amplitude higher than

respect to this average signal is identified by unit 25 as being an undifferentiated event. The position of each pixel on the surface for which an undifferentiated event is detected is memorized in unit 25 : a region for which an undifferentiated event is detected corresponds to the position of such pixel and of the surroundings pixels. Unit 25 is connected to a Micrometric Positioning Sliding Plate 26, this plate 26 positioning the platform 3 with respect to the first and second optical circuits according to two orthogonal directions X and Y. Unit 25 controls the positioning of this plate during the two scans.

During the second scan, the sample is illuminated with illumination light 4. The second scan is advantageously a bright-field microscopy scan. By "bright field scan", it is meant that the signal measured during the second scan comes mainly from:

the reflection 6 of the illumination light to produce the measured signal, typically reflected from the scanned surface 35 and/or from at least one interface 32 between two layers inside the sample (for example if the illumination light and the detected light are located at a same side of the sample, as illustrated in figures 3 and 4), or

direct transmission 6 (if the illumination light and the detected light are located at two opposite sides of the sample, for example in the case of a non inverted microscope) of the illumination light to produce the measured signal.

In other words, the reflection 6 or direct transmission 6 of the illumination light goes on a detector 8 to produce the measured signal. Bright field microscopy is a technique well known by the one skilled in the art. A bright field scan is advantageous for the second scan, because it allows actually seeing and imaging the defect, in order to determine the nature (for example scratch or dust) of the real defect.

The second scan takes place as follows. The second scan does not scan all the surface Mi, but scans (thanks to unit 25 and plate 26) only the regions for which undifferentiated events have been previously detected and memorized. For each scanned region:

i2) a part 13 of the surface of the sample is illuminated with illumination light 4, and this illuminated area 13 of the sample is imaged on the detector 8 during a time t_A2, A₂ being the value of the area 13 of the sample imaged on the detector during time t_A2, then ii2) only if necessary (for example is a defect is bigger than A₂), an other part 13' of the sample is illuminated with illumination light, and the illuminated area 13' of the sample is imaged on the detector 8 during a time tA2, A₂ being the area of the sample imaged on the detector during time t_A2, step ii2) being iterated as many time as necessary in order to illuminate and image all the region part by part.

Each step ii2) is implemented after moving the platform 10 and the sample 3 with respect to the second optical circuit.

In the illustrated case of figure 2, step ii2) is implemented only once and only for one region . The size of each region is then equal to A₂ or 2*A₂. The total area M₂ of the second scan is equal to four times A₂.

The second optical circuit is integral with the first optical circuit, and the sample is on the same platform 10 during the first and second scans, the sample having not moved from its platform between the first scan and the second scan and during these scans. This way, the second scan can precisely target only the regions for which undifferentiated events have been previously detected during the first scan, with a micrometric or nanometric precision, almost without any mistakes.

The detector 8 is a matrix detector, comprising an array of n₂*p₂ pixels. This way, the area A₂ imaged on the detector 8 at a same instant is split into n₂*p₂ pixels 14. In figure 2, A₂ is split into nine pixels : figure 2 is only a schematic view to simply illustrate the invention; in fact, A₂ is typically split into millions of pixels. During the second scan, the value D₂ of the area of each pixel 14 in the plan of the scanned surface of the sample is chosen to be smaller than a smallest size D of defect 12 searched for: D₂ < D .

Furthermore, one could use the technique of binning (i.e. combining physical pixels into one pixel) in order to reduce the average noise and to enhance the dynamic range.

Let's consider that:

- A₂ is the value of the area of the surface part imaged on the detector

(unit: square metre); it is typically comprised between 1 - 10^"11 m² and 400 - 10^"11 m², - M₂ is the value of the total area of the surface of the sample scanned during the second scan, this surface being not continuous but comprising distant regions for which undifferentiated events have been previously detected during the first scan (unit: square metre) ; it is typically comprised between 0.004· 10^"8 m² and 4· 10^"8 m² ,

- D₂ (area 14 with crossing diagonal lines in figure 2) is the value of the area of each pixel 14 during the second scan, measured in the plan of the scanned surface (unit: square metre); it is typically comprised between 0.25-10^~17 m² and 25-10^"17 m²,

- N_A2 is the amount of incident photons, i.e. the amount of photons from the illumination light, that are incident on the area A₂; it is typically comprised between 3 · 10⁸ and 300 · 10⁸,

- N₂ is the average number of photons collected per pixel during the second scan; it is typically comprised between 40 and 4000,

- n₂ and p₂ are the number of pixels in respectively a row or column of the array of pixels of the detector 8; each one of n₂ and p₂ is typically comprised between 1 · 10⁶ and 20 · 10⁶ ,

- t_A2 is the illumination time of the area A₂ during which area A₂ is imaged on the detector 8 in each step i2) and ii2) (unit: seconds) ; it is typically comprised between 0.01 · 10^"3 s and 1 · 10^"3 s ,

- t₂ is total time of the second scan, i.e. the illumination time of the area M₂ during which area M₂ is imaged on the detector part by part (unit: seconds) ; it is typically comprised between 10^"3 s and 10 s,

- P₂ is the irradiance of the illumination light 4 measured on the scanned surface of the sample during the second scan (unit: Watt per square metre)

; it is typically comprised between 3 · 10⁴ W/m² and 300 · 10⁴ W/m²,

- h is the Planck constant; (6,62606896xl0^"34 J.s)

- v₂ is the frequency of the photons of the illumination light during the second scan, in the case wherein the illumination light is monochromatic light (unit: Hertz) ; it is typically comprised between 0.2 · 10¹⁶ Hz and 20 -10¹⁶ Hz ,

In particular:

A₂ <A_l M₂ <Μ_γ

Preferably, the illumination source 15 emitting the illuminating light 4 is the same source for the first and second scans, but is more "focused" (i.e. is arranged to produce a higher irradiance) for the second scan:

The sample is illuminated with EUV light during the first scan and during the second scan.

Preferably:

n₂ =n_x and p₂= p_x, in particular if the detector 7 is the same one than detector 8.

Advantageously, the second scan is slower than the first scan :

^ A2 ^> ^Al

Advantageously, the second scan has a higher resolution power than the first scan:

P₂ >P_X and/or D₂ <D_X

The second scan is implemented in usual conditions (see "High Performance next Generation EUV Lithography Light Source", P. Choi & al., SPIE symposium on advanced lithography, 22-27 February 2009, San Jose).

For the second scan, the following experimental condition is used:

— N_A2 4 M₂

^-> , or

A₂ t_A2 RD t₂ n₂ p₂

P₂ > ^L- hv₂,

RD t₂n₂ p₂

For an average measurement signal equal to N₂, where N₂ is the average number of photons collected per pixel during the second scan, every event corresponding to a pixel 14 for which the measured signal has an amplitude inferior to

is identified by unit 25 as being a false defect and every event corresponding to a pixel 14 for which the measured signal has an amplitude superior to

is identified by unit 25 as being a true defect. Comparing the invention to prior art.

Let's consider the example of scanning a sample having dimensions equal to 0.104x0.132 m² (i.e. M l = 0.0137 m²), with a smallest size of defect 12 searched for D= 10^"16 m² with ten defects to be searched for, with a reflectivity of the surface of the sample R=0.6, and with a power of the source 15 p= 1.34 mW and photon energy hv_x = 1.47 - 10 ^~17 J and operation frequency around 2 kHz.

According to prior art (see "Dynamic scan operation of actinic EUVL mask blank inspection system with TDI mode", Tsuneo Terasawa et al ., Oct 20^th, 2009 EUVL Symposium, Prague, Czech Rep., and "Actinic dark-field mask blank inspection and defect printability analysis for detecting critical phase defects", Tsuneo Terasawa et al ., Oct 18^th, 2010 EUVL Symposium, Kobe, Japan), only the first scan would be implemented for directly localizing real defects, with the following values:

The total time of the scan according to prior art would be t₀=9152 seconds .

According to a first example of the invention, the sample would be scanned with the following values:

- A₂ = 3.4 - 10^"10 m²,

- M₂ = 10 · Α₂ =34 · 10^"10 m²,

- N_A2 = 1.2 · 10⁹,

- N₂ =200,

- m = n₂ = 2048,

- Pi = p₂ =2048,

- t_A2 = 5-10^"4 s,

- P₂ =1·10⁵ W/m²,

- h vi = h v₂ =1.47· 10^"17 J,

In this first example, it is assumed that the condition of the first scan leads to a very large addition of false positives, being 100000 times larger than the number of real defects, which is taken as 10. Despite this very large false positive rate, it is clear that the total time of the first and second scans ti+t₂=1020 s according to the invention is much shorter than the total time t₀=9152 s according to prior art. According to a second example of the invention, the sample would be scanned with a more focused illumination during the second scan, leading to a higher N₂ and thus requiring a smaller t_A2 , with the following values:

- A₂ = A₂ =3.4 -10^"10 m²,

- M₂ =10·Α₂ =34· 10^"10 m²

- N_A2 =1.2· 10¹⁰

- N₂ =2·10³,

- t_A2 =5· 10^"5 s

- P₂ =1·10⁶ W/m²,

- h vi = h v₂ =1.47· 10^"17 J, It is clear that the total time of the first and second scans ti+t₂=570s according to the invention is much shorter than the total time t₀=9152 s according to prior art. Device according to the invention

In the illustrated embodiments of a device according to the invention, the illumination light 4 is reflected by a mirror 16 towards the sample 3, and the first optical circuit comprises:

a first curved (typically spherical or parabolic) mirror 17 having a revolution symmetry around axis 18 and a hole 19 in its middle,

a second curved (typically spherical or parabolic) mirror 20, the first parabolic mirror 17 being arranged for receiving light 5 scattered by the sample during the first scan and reflecting this light 5 onto the second parabolic mirror 20, the second mirror 20 being arranged for reflecting this scattered light 5 through the hole 19 and onto the detector 7.

In the first embodiment of device 1 illustrated in figure 3, the second optical circuit comprises a beam splitter or mirror 21 arranged for:

- transmitting the illumination light 4 from mirror 16 to the sample 3, and

reflecting the light 6 reflected by the sample during the second scan.

In the first embodiment illustrated in figure 3, the second optical circuit further comprises optics 22 for focalizing light 6 onto detector 8.

In the second embodiment of device 2 illustrated in figure 4, the mirror 16 has a hole in its middle through which the light 6 reflected by the sample goes through. Furthermore, the signal measured during the first scan and the signal measured during the second scan are measured via a same detector 7,8.

In the second embodiment illustrated in figure 4, the second optical circuit comprises a shutter 23 arranged for being : closed during the first scan, for avoiding reflected light 6 to be imaged on detector 7,8 and

open during the second scan, for allowing reflected light 6 to be imaged on detector 7,8.

The shutter 23 is controlled by unit 25.

In the second embodiment illustrated in figure 4, the second optical circuit comprises a shutter 36 arranged for being :

open during the first scan, for allowing scattered light 5 to be imaged on detector 7,8 and

- closed during the second scan, for avoiding scattered light 5 to be imaged on detector 7,8.

The shutter 36 is controlled by unit 25.

In the second embodiment illustrated in figure 4, the second parabolic mirror 20 has a hole in its middle for allowing reflected light 6 to go through mirror 20 towards detector 7, 8, and the second optical circuit further comprises optics 24 for focalizing light 6 onto detector 7, 8.

Of course, the invention is not limited to the examples which have just been described and numerous amendments can be made to these examples without exceeding the scope of the invention .

Claims

1. Method for detecting defects ( 12) on or under a surface of a sample (3), comprising

- measuring a signal during a first scan of the surface through a first optical circuit ( 17, 20),

- detecting, from the signal measured during the first scan, at least one undifferentiated event corresponding to a false defect due to measurement noise or a real defect ( 12) on or under the scanned surface of the sample,

characterized in that it further comprises :

- measuring a signal during a second scan of the sample through a second optical circuit (21, 6; 23, 24), and

2. Method according to claim 1 , characterized in that the first scan is a dark-field scan .

3. Method according to claim 1 or 2, characterized in that the second scan is a bright-field scan .

4. Method according to any one of the previous claims, characterized in that the first scan scans all the surface (Mi) and in that the second scan does not scan all the surface but only regions ( 13, 13') corresponding to the undifferentiated events.

5. Method according to any one of the previous claims, characterized in that the second optical circuit is integral with the first optical circuit.

6. Method according to any one of the previous claims, characterized in that the second scan is slower than the first scan .

7. Method according to any one of the previous claims, characterized in that the second scan has a higher resolution power than the first scan.

8. Method according to any one of the previous claims, characterized in that the sample is on a same platform (10) during the first and second scans, the sample having not moved from its platform between the first scan and the second scan.

9. Method according to any one of the previous claims, characterized in that for an average measurement signal equal to Ni, where Ni is the average number of photons collected per pixel during the first scan, every event having an amplitude higher than 2_^ N \N\ h respect to this average signal is identified as being an undifferentiated event.

10. Method according to any one of the previous claims, characterized in that for an average measurement signal equal to N₂, where N₂ is the average number of photons collected per pixel during the second scan, every event having an amplitude inferior to 2_A/N is identified as being a false defect and every event having an amplitude superior to 2^Ή is identified as being a true defect.

11. Method according to any one of the previous claims, characterized in that during the first scan, a pixel area (Di) is chosen to be larger than a smallest size (D) of defect (12).

12. Method according to any one of the previous claims, characterized in that during the second scan, a pixel area (D₂) is chosen to be smaller than a smallest size (D) of defect (12).

13. Method according to any one of the previous claims, characterized in that the signal measured during the first scan and the signal measured during the second scan are measured via a same detector (7, 8).

14. Method according to any one of the previous claims, characterized in that the sample is a mask blank.

15. Method according to any one of the previous claims, characterized in that the sample is a patterned mask.

16. Method according to any one of the previous claims, characterized in that the sample is illuminated with EUV light during the first scan and during the second scan.

17. Method according to any one of claims 1 to 15, characterized in that the sample is illuminated with photons during the first scan and illuminated with electrons during the second scan.

18. Device for detecting defects on or under a surface of a sample, comprising

- means (7, 17, 20) for measuring a signal during a first scan of the surface through a first optical circuit (17, 20),

- means (25) for detecting, from the signal measured during the first scan, at least one undifferentiated event corresponding to a false defect due to measurement noise or a real defect on or under the scanned surface of the sample,

characterized in that it further comprises:

- means (8, 21, 6; 23, 24) for measuring a signal during a second scan of the sample through a second optical circuit (21, 6; 23, 24), and

- means (25) for identifying, among the detected undifferentiated events, and from the signal measured during the second scan, which undifferentiated events correspond to a real defect and which undifferentiated events correspond to a false defect.

19. Device according to claim 18, characterized in that the first scan is a dark-field scan .

20. Device according to claim 18 or 19, characterized in that the second scan is a bright-field scan .

21. Device according to any one of claims 18 to 20, characterized in that the first scan is a scan of all the surface (Mi) and in that the second scan is not a scan of all the surface but is only a scan of the regions ( 13, 13') corresponding to the undifferentiated events.

22. Device according to any one of claims 18 to 21, characterized in that the second optical circuit is integral with the first optical circuit.

23. Device according to any one of claims 18 to 22, characterized in that it comprises a single platform ( 10) arranged for being integral with the sample during the first and second scans, so that the sample does not move from its platform between the first scan and the second scan .

24. Device according to any one of claims 18 to 23, characterized in that it comprises a single detector (7, 8) for measuring the signal measured during the first scan and the signal measured during the second scan .

25. Device according to any one of claims 18 to 24, characterized in that it comprises means for using photons as a source of illumination of the sample for the first scan with an associated optical circuit for illumination and detection, and means for using electrons as a source of illumination of the sample for the second scan implementing the structure of a scanning electron microscope.