CN101393015B - On-line measurement method and device for micro/nano deep trench structure - Google Patents

Abstract

The invention discloses an online measuring method and an online measuring device for a micro-nano deep groove structure. The method comprises the following steps: linearly polarized infrared beams are projected to the surface of a sample piece which is provided with the deep groove structure, and measured reflection spectra are obtained after interference signals which are formed by reflected light on various interfaces of the groove structure are subjected to filtration and so on; an equivalent optical model of the deep groove structure is established by the polarization-based equivalent medium theory, and theoretic reflection spectra of the equivalent optical model of the deep groove structure are calculated; and the width and the depth of the groove are quickly extracted by the quick parameter extraction method of combination of an artificial neural network and a local search algorithm and through fitting of the theoretic reflection spectra and the measured reflection spectra, and precise online measurement of geometrical shape parameters of the deep groove is realized. The device comprises an infrared source, an infrared polaroid sheet, an interferometer, a plane mirror, two off-axis parabolic mirrors and an infrared detector. The device can realize online measurement of the depth and the width of the deep groove structure with high depth-width ratio in a field effect tube and a dynamic RAM during the manufacturing procedure, and has the characteristics of nondestructiveness, quickness and low cost.

Description

A kind of micro-nano deep groove structure On-line Measuring Method and device

Technical field

The invention belongs to integrated circuit (IC) and MEMS (micro electro mechanical system) (MEMS) device measuring technique, be specifically related to micro-nano deep groove structure On-line Measuring Method and device, this method is particularly useful for the on-line measurement of field effect transistor (MOSFETs) and the middle deep groove structure process degree of depth of dynamic RAM (DRAM) and width.

Background technology

In microelectronics and power semiconductor device design and process for making, extensively adopted intensive three-dimensional structure array at present, as one or more layers membrane structure of deposition in the Silicon Wafer substrate, etching lines groove array, circular hole or other shape groove arrays in substrate again, use packing material backfill groove structure then, repeat above etching, filling step again to form complicated groove array structure.In these groove array structure process, its three-dimensional appearance, particularly gash depth, online, the nondestructive measurement control of geometric properties yardsticks such as width is particularly important.In numerous nondestructive measuring methods, measuring method is particularly suitable for this application demand, reflective spectral measure method for example, scattering spectrometry etc., these methods have been widely used in optical film thickness and composition measurement, have applied it in the measurement of linear grating groove structure in partial monopoly and document.

The applicant " a kind of micro-nano deep groove structure measuring method and device based on Infrared Reflective Spectra " (notification number is CN101131317A) proposed on 09 20th, 2007, this inventive method projects the silicon chip surface that contains deep groove structure with infrared beam, analyzes the interference light that forms from each surface reflections of deep groove structure and obtains measuring reflectance spectrum; Adopt EFFECTIVE MEDIUM THEORY to make up the theoretical reflectance spectrum of this deep groove structure equivalence multilayer film storehouse optical model, utilize simulated annealing and based on the optimized Algorithm of gradient, by theoretical reflectance spectrum this measurement reflectance spectrum is carried out match, and then geometrical characteristic parameters such as the degree of depth of extraction groove and width, realize the accurate measurement of high-aspect-ratio deep trench width and degree of depth equidimension.This method can be measured gash depth, width and film thickness simultaneously.The implement device that this inventive method provides projects article surface to be measured by incident beam is focused on, and a slit diaphragm is set on emitting light path, with the influence of elimination tested sample back parasitic light, thereby measures the accurate reflectance spectrum of groove.

Mentioning measuring method in above-mentioned patent documentation is a kind of reflectance spectrum based on complex light, and the groove Geometric Modeling of this method is based on the EFFECTIVE MEDIUM THEORY of nonpolarized light incident.For different polarization directions, each polarization direction equivalent precision has than big-difference, and therefore, complex light incident is bigger than polarized light incident error.In above open invention, parameter extraction adopts simulated annealing and based on the method for the optimized Algorithm combination of gradient, this method is little to the dependence of initial value than classic method, measures in the unknown also can extract under the situation of initial value to obtain groove parameter to be measured.But this parameter extracting method can't satisfy the requirement of on-line measurement rapid extraction in the several seconds.Implement device in this method proposition, adopt accurate complicated light channel structure design, assurance measure reflectance spectrum accurately, eliminate of the influence of back parasitic light, but structural design that should complexity has brought bigger difficulty also for simultaneously precision installation, debugging and the staking-out work of device to measurement structure.

Summary of the invention

The object of the present invention is to provide a kind of micro-nano deep groove structure On-line Measuring Method, this method can be carried out the online precise monitoring of etching process to geometry patterns such as gash depth, width, have untouchable, non-destructive, high speed and high-precision characteristics, the present invention also provides the device of realizing that this method is more easy, cost is lower.

Micro-nano deep groove structure On-line Measuring Method disclosed by the invention, its step comprises:

The 1st step projected the article surface to be measured that comprises micro-nano deep groove structure with infrared beam, and the wavelength of infrared beam is 2～20um;

The 2nd steps into irradiating light beam after each surface reflection of micro-nano deep groove structure, adopts infrared eye to receive each reflected signal, obtains comprising the interference signal of groove geological information;

The interference signal that the 3rd step obtained the 2nd step carries out Fourier transform, obtains the Infrared Reflective Spectra based on wave number;

The 4th step was carried out low-pass filtering to the 3rd Infrared Reflective Spectra that obtains of step, and filtering determinand article back parasitic light spectrum obtains reflecting the measurement reflectance spectrum of groove structure characteristics;

The 5th step was set up its equivalent multilayer film storehouse optical model according to deep groove structure characteristics to be measured, utilized the reflection coefficient r of the equivalent film storehouse under each wavelength that formula (I) calculates, and utilized reflection coefficient r to obtain the theoretical reflectance spectrum of this groove structure;

$r=\frac{M_{21}}{M_{11}}---\left(I\right)$

Wherein, $\left(\begin{array}{cc}{M}_{11}& M_{12}\\ M_{21}& M_{22}\end{array}\right)=D_0^{-1}\left[\underset{l=1}{\overset{N}{\mathrm{\Π}}}{D}_l{P}_l{D}_l^{-1}\right]D_s---\left(\mathrm{II}\right)$

In its Chinese style (II), M ₁₁, M ₁₂, M ₂₁, M ₂₂Be the every intermediate variable of multilayer optical propogator matrix, D ₀Be the optical signature matrix of environment, D _sBe the optical signature matrix of substrate, P _lBe the matrix function at l layer phase change angle, D _lBe the refractive index of film stack l layer and the matrix function at refraction angle, wherein, D _lCalculate according to formula (III) or (IV):

For the TE polarization direction:

$D_l=\left(\begin{array}{cc}1& 1\\ {\mathrm{<figure-callout\; id="2"\; label="n\; TE"\; filenames="H2008101972791E00022.png"\; state="\{\{state\}\}">n</figure-callout>}}_{\mathrm{TE}}\cos {\mathrm{\&theta;}}_l& -{\mathrm{<figure-callout\; id="2"\; label="n\; TE"\; filenames="H2008101972791E00022.png"\; state="\{\{state\}\}">n</figure-callout>}}_{\mathrm{TE}}\cos {\mathrm{\&theta;}}_l\end{array}\right)---\left(\mathrm{III}\right)$

For the TM polarization direction:

$D_l=\left(\begin{array}{cc}\cos {\mathrm{\&theta;}}_l& \cos {\mathrm{\&theta;}}_l\\ n_{\mathrm{TM}2}& -n_{\mathrm{TM}2}\end{array}\right)---\left(\mathrm{IV}\right)$

Wherein, θ _lBe l layer refraction angle, n _TE2And n _TM2Be the equivalent refractive index of each equivalent layer TE polarization and TM polarization direction, its calculating formula is respectively formula (V) and formula (VI):

${\mathrm{<figure-callout\; id="2"\; label="n\; TE"\; filenames="H2008101972791E00022.png"\; state="\{\{state\}\}">n</figure-callout>}}_{\mathrm{TE}}=\{n_{\mathrm{TE}0}^2+\frac{1}{3}[\frac{\mathrm{\&pi;f}(1-f)p}{\mathrm{\&lambda;}}{]}^2{(n_g^2-n_m^2)}^2{\}}^{1/2}---\left(V\right)$

${\mathrm{<figure-callout\; id="2"\; label="n\; TM"\; filenames="H2008101972791E00022.png"\; state="\{\{state\}\}">n</figure-callout>}}_{\mathrm{TM}}={\{n_{\mathrm{TM}0}^2+\frac{1}{3}{\left[\frac{\mathrm{\&pi;f}(1-f)p}{\mathrm{\&lambda;}}\right]}^2{(\frac{1}{n_g^2}-\frac{1}{n_m^2})}^2{n}_{\mathrm{TM}0}^6{n}_{\mathrm{TE}0}^2\}}^{1/2}---\left(\mathrm{VI}\right)$

Wherein, n _TE0And n _TM0Be respectively the zeroth order equivalent refractive index of TE and TM polarization direction, i.e. Zero-order diffractive in scatterometry; λ is the detecting light beam wavelength, and p is the groove array Cycle Length, and f is each layer dutycycle, n _gAnd n _mBe respectively trench material and packing material refractive index wherein;

The theoretical reflectance spectrum match the 4th that the 6th step utilized for the 5th step obtained goes on foot the measurement reflectance spectrum that obtains, and extracts how much pattern parameters that obtain micro-nano deep groove structure.

Realize the device of said method, it is characterized in that: infrared light supply, the infrared polarization sheet, interferometer, the plane mirror and first off-axis paraboloidal mirror are successively inferior on the same light path, sample stage is positioned on the reflected light path of first off-axis paraboloidal mirror, the angle surperficial at 45 of the first off-axis paraboloidal mirror reflected light and sample stage; Second off-axis paraboloidal mirror and first off-axis paraboloidal mirror are placed with respect to the incidence point symmetry of sample on the sample stage, and infrared eye is positioned on the reflected light path of second off-axis paraboloidal mirror, and computing machine links to each other with infrared eye; Computing machine receives the interference signal after pre-process of infrared eye output, carries out obtaining after Fourier transform is handled the reflectance spectrum of groove structure, handles according to the 4th step to the process in the 6th step, extracts and obtains required groove geometric parameter values.

Compare with current measuring methods, method provided by the present invention can realize online, quick, the high-precision measurement of micro-nano deep groove structure, will have wide practical use in Measurement of Semiconductors and technology controlling and process field.Particularly, the present invention can obtain following effect in the deep trench capacitor structure of DRAM is measured:

(1) realizes the conventional deep trench of DRAM, the tiltedly on-line measurement of typical deep groove structures such as sidewall deep trench, ampuliform deep trench and polysilicon filling groove;

(2) be implemented in that DRAM deep trench defective in situ detection, high-aspect-ratio micro-nano structure etching are monitored in real time, ampuliform groove polysilicon recharges the online detection of groove, whole audience silicon chip CD homogeneity rapid evaluation, and characterize at photoresist and dielectric film.Oxygen dosage control etc. is annotated in thin film epitaxial growth process feedback, Silicon-On-Insulator processing.

Description of drawings

Fig. 1 is a polarization red reflex spectrometry light path synoptic diagram;

Fig. 2 is skew wall deep groove structure and incident beam reflection synoptic diagram;

Fig. 3 is skew wall deep groove structure equivalence optical model and incident beam reflection synoptic diagram;

Fig. 4 is the fast automatic extraction process flow diagram of groove parameter;

Fig. 5 is a BP artificial neural network synoptic diagram;

Fig. 6 is the present invention's one case study on implementation plant system drawing.

Embodiment

Measuring process with the skew wall deep groove structure is an example below, and the principle and the course of work to the inventive method is described in further detail in conjunction with the accompanying drawings:

(1) infrared beam is projected the article surface to be measured that comprises deep groove structure, the wavelength of infrared beam is positioned at the middle infrared wavelength scope, and wavelength is 2～20um;

(2) incident beam adopts infrared eye to receive each reflected signal after each surface reflection of groove structure, obtains comprising the interference signal of groove geological information;

As shown in Figure 1, inclined to one side by the infrared beam that infrared light supply 1 sends by 2 of infrared polarization sheets, obtain linearly polarized light, linearly polarized light enters interferometer 3, after interferometer 3 modulation, converge after mirror reflects, project article to be measured 4 surfaces that comprise groove structure, each surface reflection signal of groove structure is parallel after catoptron 4 reflection goes into to inject infrared eye 6.

As shown in Figure 2, the skew wall deep groove structure comprises mask layer 41 from top to bottom successively, channeled layer 42 and basalis 43.When incident beam 11 projects the groove structure surface, respectively on the mask layer surface, mask layer and channeled layer interface, channel bottom produce reflection, each surperficial folded light beam 12,13,14 produces interference on infrared eye, obtain comprising the interference signal of groove geological information.

(3) interference signal that detector measurement is obtained carries out Fourier transform, obtains the Infrared Reflective Spectra based on wave number;

(4) reflectance spectrum that obtains in the step (3) is carried out low-pass filtering, filtering determinand article back parasitic light spectrum obtains reflecting the measurement reflectance spectrum of groove structure characteristics;

Include the frequency content corresponding to each layer depth of groove structure in the reflectance spectrum, wherein, the back parasitic light shows as high-frequency signal therein, therefore, can eliminate the influence of back parasitic light by low-pass filtering.This filtering mode is compared traditional spatial filter that passes through and is carried out hardware to eliminate the mode of back parasitic light simpler, and assorted better effects if disappears.

(5), set up its equivalent multilayer film storehouse optical model, the optical parametric of approximate description groove structure and infrared external reflection characteristic according to deep groove structure characteristics to be measured;

Deep trench array to be measured can be counted as sub-wavelength grate structure, and therefore the grating Cycle Length, can be multilayer homogeneous film storehouse model with its Approximate Equivalent much smaller than surveying optical wavelength.The concrete modeling process of equivalence multilayer film storehouse optical model can be with reference to disclosed method among the CN101131317A.

As shown in Figure 3, this multi-layer film structure is the equivalent multilayer film storehouse model of skew wall deep groove structure among Fig. 2, comprise mask equivalent layer 511 from top to bottom successively, groove equivalent layer 521, substrate equivalent layer 531 calculates the TE polarization of each equivalent layer and the equivalent refractive index n of TM polarization direction respectively according to formula (1) and (2) _TE2And n _TM2

${\mathrm{<figure-callout\; id="2"\; label="n\; TE"\; filenames="H2008101972791E00022.png"\; state="\{\{state\}\}">n</figure-callout>}}_{\mathrm{TE}}=\{n_{\mathrm{TE}0}^2+\frac{1}{3}[\frac{\mathrm{\&pi;f}(1-f)p}{\mathrm{\&lambda;}}{]}^2{(n_g^2-n_m^2)}^2{\}}^{1/2}---\left(1\right)$

${\mathrm{<figure-callout\; id="2"\; label="n\; TM"\; filenames="H2008101972791E00022.png"\; state="\{\{state\}\}">n</figure-callout>}}_{\mathrm{TM}}={\{n_{\mathrm{TM}0}^2+\frac{1}{3}{\left[\frac{\mathrm{\&pi;f}(1-f)p}{\mathrm{\&lambda;}}\right]}^2{(\frac{1}{n_g^2}-\frac{1}{n_m^2})}^2{n}_{\mathrm{TM}0}^6{n}_{\mathrm{TE}0}^2\}}^{1/2}---\left(2\right)$

Wherein, n _TE0And n _TM0Be respectively the zeroth order equivalent refractive index of TE and TM polarization direction, i.e. Zero-order diffractive in scatterometry.λ is the detecting light beam wavelength, and p is the groove array Cycle Length, and f is each layer dutycycle, n _gAnd n _mBe respectively trench material and packing material refractive index wherein.The equivalent modeling method precision based on complex light that this equivalent modeling method based on polarization is compared classic method is higher.

Utilization multilayer film optical propagation theory, the reflection of polarization spectrum of calculating groove structure equivalence optical model.The reflection coefficient of multilayer film storehouse can use the method for optical propagation matrix to calculate.The optical propagation matrix of multilayer film storehouse is as shown in Equation (3):

$\left(\begin{array}{cc}{M}_{11}& M_{12}\\ M_{21}& M_{22}\end{array}\right)=D_0^{-1}\left[\underset{l=1}{\overset{N}{\mathrm{\&Pi;}}}{D}_l{P}_l{D}_l^{-1}\right]D_s---\left(3\right)$

For the TE polarization direction:

$D_l=\left(\begin{array}{cc}1& 1\\ {\mathrm{<figure-callout\; id="2"\; label="n\; TE"\; filenames="H2008101972791E00022.png"\; state="\{\{state\}\}">n</figure-callout>}}_{\mathrm{TE}}\cos {\mathrm{\&theta;}}_l& -{\mathrm{<figure-callout\; id="2"\; label="n\; TE"\; filenames="H2008101972791E00022.png"\; state="\{\{state\}\}">n</figure-callout>}}_{\mathrm{TE}}\cos {\mathrm{\&theta;}}_l\end{array}\right)---\left(4\right)$

For the TM polarization direction:

$D_l=\left(\begin{array}{cc}\cos {\mathrm{\&theta;}}_l& \cos {\mathrm{\&theta;}}_l\\ n_{\mathrm{TM}2}& -n_{\mathrm{TM}2}\end{array}\right)---\left(5\right)$

Can get the reflection coefficient of film storehouse thus

$r=\frac{M_{21}}{{\mathrm{<figure-callout\; id="11"\; label="M"\; filenames="H2008101972791E00021.png"\; state="\{\{state\}\}">M</figure-callout>}}_{11}}$

M wherein ₁₁, M ₁₂, M ₂₁, M ₂₂The every intermediate variable of multilayer optical propogator matrix, D ₀Be the optical signature matrix of environment, D _sBe the optical signature matrix of substrate, D _lBe the refractive index of film stack l layer and the matrix function at refraction angle, P _lBe the matrix function at l layer phase change angle, θ _lBe l layer refraction angle.Can obtain the theoretical reflectance spectrum of this groove structure by the reflection coefficient of the equivalent film storehouse under above each wavelength that calculates.

(6) the measurement reflectance spectrum that obtains by the theoretical reflectance spectrum match step (4) that obtains based on the equivalent optical model Theoretical Calculation of the groove structure of polarization, and then rapid extraction obtains gash depth, how much pattern parameters such as width.

The etching groove process monitoring needs real-time geometric parameter measurement, and it is most important that speed is found the solution in the inverting of groove parameter.The present invention proposes based on the fast parameter extracting method of artificial neural network, accurately measure in real time to realize the groove parameter in conjunction with the Levenberg-Marquardt local search algorithm.Specifically describe groove parameter extraction step below in conjunction with Fig. 4:

Step 1: according to groove reflective spectral property and parameter amount to be measured, set up a multi-layer artificial neural network (ANN);

As shown in Figure 5, create three layers of feedforward network (BP network), comprise input layer 71, hidden layer 72 and output layer 73.According to the reflective spectral property and the number of parameters to be measured of groove structure to be measured, determine input layer 71 node numbers and output layer 73 node numbers respectively, rule of thumb formula and input again, output layer node number are determined hidden layer 72 node numbers.The input layer number is bigger to training time of BP network and precision influence, can be the step-length of input spectrum sequence with equiwavelength according to the reflective spectral property of multilayer groove structure, determines the input spectrum sequence;

Step 2: according to the theoretical modeling method of the groove structure in the step (5), create BP network training sample set;

According to groove structure characteristics to be measured, selecting a multilayer film storehouse is its equivalent optical model.Determine each layer of groove structure groove depth and groove width scope respectively according to groove structure design and process, thereby determine the training sample set scope.According to multilayer film optical propagation theory, calculate the reflectance spectrum of its equivalent optical model under each groove and groove width, obtain BP network training sample set (O _i, I _i), wherein, O _iBe the reflectance spectrum vector, I _iBe groove geometric parameter vector to be measured, comprise geometric parameters such as gash depth and width, i=1 ..., N, N is for creating sample size.

Step 3: train with the training sample set pair BP network that Step2 creates, training is input as reflectance spectrum vector O _i, be output as groove geometric parameter vector to be measured I _i

For improving the output stability of BP network, in the reflectance spectrum vector, add a certain amount of noise at random, be input with the reflectance spectrum that comprises noise, corresponding groove geometric parameter vector to be measured is trained the BP network for output.

Step 4: with the neural network that the filtered measure spectrum input of step (4) Step3 trains, output is the groove that comprises certain error and measures initial value;

Step 5: be output as local iteration's searching algorithm initial value with Step4, with the measure spectrum after theoretical reflectance spectrum match step (4) Filtering Processing of step (5) calculating, and then extraction obtains high-precision groove geometric parameter values;

Be output as Levenberg-Marquardt local iteration algorithm initial value with the BP network, can in several milliseconds of times, reach predefined iteration precision, solve traditional match iterative algorithm and need choose the difficulty of iterative initial value in advance, guaranteed the quick convergence of iterative algorithm.

As shown in Figure 6, apparatus of the present invention comprise infrared light supply 1, infrared polarization sheet 2, interferometer 3, plane mirror 41, first, second off- axis paraboloidal mirror 42,43, sample stage 91, infrared eye 6, computing machine 8.

Infrared light supply 1, infrared polarization sheet 2, interferometer 3, the plane mirror 41 and first off-axis paraboloidal mirror 42 are positioned on the same light path successively, sample stage 91 is positioned on the reflected light path of first off-axis paraboloidal mirror 42, the angle surperficial at 45 of first off-axis paraboloidal mirror, 42 reflected light and sample stage 91.Second off-axis paraboloidal mirror 43 and first off-axis paraboloidal mirror 42 are placed with respect to the incidence point symmetry of sample on the sample stage 91, and infrared eye 6 is positioned on the reflected light path of second off-axis paraboloidal mirror 43, and computing machine 8 links to each other with infrared eye 6.

The parallel beam that infrared light supply 1 sends enters polaroid 2, obtain the parallel lines light beam, light beam is after interferometer 3 modulation, through plane mirror 41 reflections, converge by first off-axis paraboloidal mirror 42 again, with 45 ° of etching article surfaces to be measured that project in the etching reaction chamber 9, folded light beam is through off-axis paraboloidal mirror 43 reflections, parallel injecting in the infrared eye 6.Infrared eye 6 comprises functions such as signals collecting, amplification, filtering, digital-to-analog conversion.The interference signal that infrared eye collects is sent into computing machine 8 after pre-process.Interference signal carries out obtaining after Fourier transform is handled the reflectance spectrum of groove structure by computing machine 8, by the method for above-mentioned step (4)～(7) measure spectrum is carried out filtering, analysis again, and then extraction obtains the groove geometric parameter values.

Claims (3)

1. micro-nano deep groove structure On-line Measuring Method, its step comprises:

The 1st step projected the article surface to be measured that comprises micro-nano deep groove structure with the linear polarization infrared beam, and the wavelength of infrared beam is 2～20um;

The 2nd steps into irradiating light beam after each surface reflection of micro-nano deep groove structure, adopts infrared eye to receive each reflected signal, obtains comprising the interference signal of groove geological information;

The interference signal that the 3rd step obtained the 2nd step carries out Fourier transform, obtains the Infrared Reflective Spectra based on wave number;

The 4th step was carried out low-pass filtering to the 3rd Infrared Reflective Spectra that obtains of step, and filtering determinand article back parasitic light spectrum obtains reflecting the measurement reflectance spectrum of groove structure characteristics;

The 5th step is according to deep groove structure characteristics to be measured, set up its equivalent multilayer film storehouse optical model, utilize the reflection coefficient r of the equivalent multilayer film storehouse under each wavelength that formula (I) calculates, utilize reflection coefficient r to obtain the theoretical reflectance spectrum of this groove structure;

$r=\frac{M_{21}}{M_{11}}---\left(I\right)$

Wherein, $\left(\begin{array}{cc}{M}_{11}& M_{12}\\ M_{21}& M_{22}\end{array}\right)=D_0^{-1}\left[\underset{l=1}{\overset{N}{\mathrm{\&Pi;}}}{D}_l{P}_l{D}_l^{-1}\right]D_s---\left(\mathrm{II}\right)$

In its Chinese style (II), M ₁₁, M ₁₂, M ₂₁, M ₂₂Be the every intermediate variable of multilayer optical propogator matrix, D ₀Be the optical signature matrix of environment, D _sBe the optical signature matrix of substrate, P _lBe the matrix function at l layer phase change angle, D _lBe the refractive index of equivalent multilayer film storehouse l layer and the matrix function at refraction angle, wherein, D _lCalculate according to formula (III) or (IV):

For the TE polarization direction:

$D_l=\left(\begin{array}{cc}1& 1\\ n_{\mathrm{TE}2}\cos {\mathrm{\&theta;}}_l& -n_{\mathrm{TE}2}\cos {\mathrm{\&theta;}}_l\end{array}\right)---\left(\mathrm{III}\right)$

For the TM polarization direction:

$D_l=\left(\begin{array}{cc}\cos {\mathrm{\&theta;}}_l& \cos {\mathrm{\&theta;}}_l\\ n_{\mathrm{TM}2}& -n_{\mathrm{TM}2}\end{array}\right)---\left(\mathrm{IV}\right)$

Wherein, θ _lBe l layer refraction angle, n _TE2And n _TM2Be the equivalent refractive index of each equivalent layer TE polarization and TM polarization direction, its calculating formula is respectively formula (V) and formula (VI):

$n_{\mathrm{TE}2}={\{n_{\mathrm{TE}0}^2+\frac{1}{3}{\left[\frac{\mathrm{\&pi;f}(1-f)p}{\mathrm{\&lambda;}}\right]}^2{(n_g^2-n_m^2)}^2\}}^{1/2}---\left(V\right)$

$n_{\mathrm{TM}2}={\{n_{\mathrm{TM}0}^2+\frac{1}{3}{\left[\frac{\mathrm{\&pi;f}(1-f)p}{\mathrm{\&lambda;}}\right]}^2{(\frac{1}{n_g^2}-\frac{1}{n_m^2})}^2{n}_{\mathrm{TM}0}^6{n}_{\mathrm{TE}0}^2\}}^{1/2}---\left(\mathrm{VI}\right)$

Wherein, n _TE0And n _TM0Be respectively the zeroth order equivalent refractive index of TE and TM polarization direction, i.e. Zero-order diffractive in scatterometry; λ is the detecting light beam wavelength, and p is the groove array Cycle Length, and f is each layer dutycycle, n _gAnd n _mBe respectively trench material and packing material refractive index wherein;

The theoretical reflectance spectrum match the 4th that the 6th step utilized for the 5th step obtained goes on foot the measurement reflectance spectrum that obtains, and extracts how much pattern parameters that obtain micro-nano deep groove structure.

2. micro-nano deep groove structure On-line Measuring Method according to claim 1 is characterized in that: the 6th step was extracted how much pattern parameters of micro-nano deep groove structure according to following process:

The 6.1st step:, set up a multi-layer artificial neural network according to groove reflective spectral property and parameter amount to be measured;

The 6.2nd step: determine each layer of groove structure groove depth and groove width scope respectively according to groove structure design and process, obtain the training sample set scope; Utilize the equivalent optical modeling method in the 5th step, calculate the theoretical reflectance spectrum of its equivalent optical model under each layer groove depth and groove width, obtain the training sample set (O of multi-layer artificial neural network _i, I _i), wherein, O _iBe the reflectance spectrum vector, I _iBe groove geometric parameter vector to be measured, i=1 ..., N, N is for creating sample size;

The 6.3rd step: train with the training sample set pair multi-layer artificial neural network that the 6.2nd step created, training is input as reflectance spectrum vector O _i, be output as groove geometric parameter vector to be measured I _i

The 6.4th step: filtered measurement reflectance spectrum of the 4th step is input to the multi-layer artificial neural network that the 6.3rd step trained, and the output groove is measured initial value;

The 6.5th step: measuring initial value with the groove of the 6.4th step output is local iteration's searching algorithm initial value, and the measurement reflectance spectrum that the 4th step of theoretical reflectance spectrum match that obtains with the 5th step obtains is extracted and obtained required groove geometric parameter values.

3. device of realizing the described method of claim 1 is characterized in that:

Infrared light supply (1), infrared polarization sheet (2), interferometer (3), plane mirror (41) and first off-axis paraboloidal mirror (42) are positioned on the same light path successively, sample stage (91) is positioned on the reflected light path of first off-axis paraboloidal mirror (42), the angle surperficial at 45 of first off-axis paraboloidal mirror (42) reflected light and sample stage (91); Second off-axis paraboloidal mirror (43) is placed with respect to the incidence point symmetry of the last sample of sample stage (91) with first off-axis paraboloidal mirror (42), infrared eye (6) is positioned on the reflected light path of second off-axis paraboloidal mirror (43), and computing machine (8) links to each other with infrared eye (6); Computing machine (8) receives the interference signal after pre-process of infrared eye (6) output, carry out obtaining after Fourier transform is handled the reflectance spectrum of groove structure, handle according to the 4th step to the process in the 6th step, extract and obtain required groove geometric parameter values.

Images