JPH0580147B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a method for non-destructively diagnosing the properties of an object.

For example, non-destructive diagnosis is used to diagnose failures in combinational logic circuits. In this case, various failure locations and failure modes (1 fixed failure, 0 fixed failure, etc.) are assumed, and what kind of outputs are obtained under various test input signals is maintained as a failure dictionary. A commonly used method is to apply a test input signal to the object to be inspected, observe the output, search a fault dictionary, and find the cause of the fault that matches the input signal. This method requires the creation of a very troublesome fault dictionary prior to diagnosis. As another example, measuring the carrier concentration distribution in the depth direction at a specific location on a wafer in a semiconductor manufacturing process is an extremely important task for controlling the semiconductor manufacturing process. As mentioned in Japanese Patent Application No. 56-17012 previously filed by the present applicant, conventionally the process of etching a certain amount from the surface of a silicon wafer and quantifying the amount of carrier by electrical measurement is repeated. It used a time-consuming procedure. For this example as well, it is important to develop a non-destructive method, and the carrier distribution measurement method using infrared light has been proposed in the above-mentioned prior application. The proposed method takes advantage of the fact that when an inspection site on a silicon wafer is irradiated with infrared light at a certain angle of incidence, the reflected light electric field reflects the shape of the carrier distribution at the inspection site. The purpose is to estimate the shape of the unknown carrier distribution from observation results. In the above invention, a method is shown in which the distribution of carriers in the depth direction is rectangular, but the actual distribution shape is not necessarily rectangular. One possible method is to assume various distribution shapes in advance and create a table that combines these with data representing the reflected light electric field observed when a certain incident light electric field is applied.
Using this table, a method can be considered in which the carrier distribution is inversely estimated from the observed reflected light electric field data. However, this method has the drawback that the table must be created in advance. Furthermore, it is difficult to accurately estimate the shape of a distribution other than the carrier distribution assumed in advance.

The purpose of the present invention is to provide a non-destructive method that is effective when diagnosing the properties of a target F, when methods such as destroying the F are impossible, or even if possible, require a long time. The objective is to provide a diagnostic method.

In the present invention, a model f of the object F is used in order to avoid using direct means such as destroying the object F. f is input i to F, diagnosis target F
Output Q from parameters P and F representing the internal state of
Give the relationship between. Let us suppose that Q=f(i,P)
Write. Now, suppose that when input i ^* is given to F, output Q ^* is obtained. At this time, the internal state of F is estimated by calculating P that satisfies Q ^* =f(i ^* , P).

In the following, examples of the invention are presented.

FIG. 1 shows a non-destructive diagnostic system using the present invention. In the figure, 100 is the diagnosis target (denoted as F), 11
0 is an input signal generator to F, 120 is an output signal receiver from F, and 130 and 140 are the output and input interfaces of the computer 150, respectively.
160 is a non-destructive diagnostic program stored in the memory 165, which is a program for model F (170 in the figure).
Diagnose F using (The processing procedure of 160 is shown in FIG. 2). 180 is a data storage table;
185 is a deviation calculation program between the measured output value and the calculated value, and 190 is an input/output interface between the system and the user, and in this figure, a graphic display device is used.

FIG. 2 shows a processing flowchart of the program 160 in FIG. 200 in the figure is an initial setting process,
An instruction as to whether or not to continue the analysis process is received from the user via the graphic display device 190 in FIG.
Set the input signal generator 110 to the initial state (signal lines are omitted, but in some cases 10
An initial set of 0.120 is also performed). continue,
Clear the data storage table 180 to zero. Also, parameter values (initial values, constants, etc.) used in the processing procedure shown in FIG. 2 are set.

At 210 and 220 in FIG. 2, the input signal generator 1 is connected via the interface 130 in FIG.
10 and an input signal (indicated by 115) i to F.
(θ _k ) is given. Here, θ _k is a parameter for changing the input signal, and 110 receives θ _k from the computer side and generates a predetermined signal i (θ _k ),
Enter in F. Under this input signal, the diagnostic object F produces an output O(θ _k ) (denoted by symbol 118). This output is detected by receiving device 120. Processing 230 in FIG. 2 reads the output O(θ _k ) detected at 120 into the computer. Subsequently, in process 240, information θ _k , i
(θ _k ) and O(θ _k ) are stored in the table 180. After this, it is determined whether or not to repeat the above steps 210 to 240, 245, and if necessary, it is repeated to set a total of M pieces of θ _k , i(θ _k ), and O(θ _k ) within 180 Remember.

The subsequent process is a parameter estimation process using the model f shown at 170 in FIG. Here, the model _O ( _{θ k} )
=f(i(θ _k ), P) to estimate the model parameter P. First, in 250, the parameter P
Generate appropriately. As a generation method, there are a method of generating a random number vector and setting it as P, and a method of adding the variation ΔP to the P value used last time and stored in the computer memory. In the latter case, see Figure 1.18.
A method such as calculating the sensitivity ∂δ/∂P of the function δ shown in 5 with respect to P and taking ΔP in the direction of P where δ becomes smaller is used.

In the process 255, in the previous process 240, FIG.
Each input value i (θ _k ) stored in 80 tables,
Using k=1 to M, the parameter value P generated in step 250 above, and the model f in _FIG . calculate.

In process 260, a deviation value δ between the output value calculated in 255 and the actual output value stored in 240 is calculated. For example, the following equation may be used as δ.

δ=|O ^* ( _θ1 )−f(i ^* ( _θ1 ),P)|+|O ^* ( _θ2
)−f(i ^* ( _θ2 ),P)｜+…+｜O ^* ( _θM )−f(i ^*
(θ _M ), P) | This is the sum of the absolute values of the differences between the calculated output value and the actual output value for all input values.

In process 265, it is determined whether the error δ of the model determined above is small by comparing it with the convergence determination parameter ε given in initial setting process 200. If δ is greater than ε, then 25
Return to process 0. If not, process 27
0, and on the display device 190 of FIG.
Display the determined parameter P. At this time, if the state of diagnosis target F under parameter P can be visualized more clearly, this visualized image ((P)
) are also displayed at 190.

After the above processing, the process returns to the initial processing 200, replaces the diagnosis target F with another one, and repeats the same procedure as many times as necessary.

Next, a failure diagnosis system for a combinational logic circuit using the present invention will be described.

FIG. 3 300 shows a 2-input OR gate 310 and a 2-input OR gate 310.
It is a combinational circuit consisting of an input AND gate 320. This corresponds to diagnosis target F in FIG. The normal operation of this circuit is given by the following equation.

O ₁ = (i ₁ ∨i ₂ ) i ₃ Now, only the part 330 in the figure is 0 or 1.
Assume that there is a possibility that a fixed failure may occur. At this time, the operation of this circuit F including failures can be expressed by the following equation.

O ₁ = ((P ₁ (i ₁ ∨i ₂ ) ∨ ( ₂ ( ₁ ∨ ₂ ))) i ₃ Here, P ₁ and P ₂ are parameters that take 0 or 1. P ₁ = P ₂ = 1, O = (i ₁ ∨i ₂ ) i ₃ , indicating a normal circuit. When _{P 1} = 1, P ₂ = 0,
When P ₁ =0 and P ₂ =1, O 1 =i ₃ and represents a 1-fixed fault in section 330 of FIG. 3, _{O 1 =} 0 and represents a 0-fixed fault. From the above, the above equation can be used as the model f in FIG. 1 170.

Next, the output O ₁ calculated by the model f shown in Fig. 1 170 and the actually observed output
The following formula may be used for the deviation δ from O ₁ ^* .

δ=〓(O ₁ ^* O ₁ ) Here is the EOR symbol, which takes 1 when O ₁ ^* and O ₁ are different, and 0 when they are the same. Σ means the sum over all observed O ₁ ^* . Other than this, δ=〓|O ₁ ^* −O ₁ | may also be used.

Next, the diagnostic procedure processing in this example (Fig. 1 16)
The main parts of 0) will be explained in accordance with FIG. In process 220, a random number that takes a value of 0 or 1 is set to 3.
340, 350, 3 to generate a three-dimensional spectrum consisting of these three elements.
Input to terminal 60. In process 230, terminal 3
Read the output signal from 70. In process 240, the above input and output signals are stored in the table 180 in FIG. Repeat the above operation M times. FIG. 4 shows an example of the contents of the table obtained as described above. In this example, the case where M=3 is shown.

As mentioned at the beginning of this example, the failure state is 0.
Fixed failures and one fixed failure are assumed. As mentioned earlier, these are (P ₁ , P ₂ )=(0,1),
(P ₁ , P ₂ )=(1, 0), respectively. Further, in a normal case, it corresponds to (P ₁ , P ₂ )=(1, 1). Therefore, in the process of FIG. 2 250,
(P ₁ , P ₂ ) = (1, 1), (0, 1), (1, 0) can be generated in order (in this example, the failure part is 1
Since we are using a simple example of only a few locations, the simple method described above will suffice; however, if l failure locations are to be considered, a 2l-dimensional parameter vector may be generated. If there are 0 failed parts, if there is 1 failed part,
Although a case of two locations may occur, since the probability of multiple locations failing at the same time is low, it is sufficient to sequentially generate a 2l-dimensional parameter vector P as shown below.

(P ₁ , P ₂ , P ₃ , P ₄ , ...P _2l-1 , P _2l ) (1, 1, 1, ... 1, 1) (0, 1, 1, ... 1, 1) ( 1,0,1,...1,1) (1,1,0,1,...1,1) (1,1,1,0,...1,1) (1,1,...0 , 1) (1, 1, ... 1, 0) Next, in the process 255 in FIG. 2, the parameter vectors sequentially assumed in the above method and the input signal values in the data table shown in FIG. i ₁ ^* , i ₂ ^* ,
Using i ₃ ^* and the above-mentioned model f, an estimated output value corresponding to each row in FIG. 4 is determined under the assumed parameters. FIG. 5 shows that P=(1,
1), (0, 1), and (1, 0) are shown in correspondence with FIG. 4.

Next, in the process shown in FIG. 2 260, the previously defined deviation function δ is calculated from the estimated output value and the actual measured output value calculated above. This result is also shown in FIG. In the process of 265, the judgment condition parameter ε is
For example, if it is set to 0.5, case 3 in Figure 5
satisfies δ≦ε. Therefore, in process 270,
In addition to displaying P=(P ₁ , P ₂ )=(1,0) in FIG. 1 190, and displaying, for example, FIG. 3 in FIG. 1 190 as a function (P), It is sufficient to display that the point 1 is a fixed failure. Figure 6 shows an example of this. During the program of FIG. 1 160,
Assume that the table shown in FIG. 6a is prepared.
The first, second, and third rows of the table are the parameters P
Place a green mark on the part where the value was found to be (1,1), put a red mark on the part where the value was found to be (0,1), and put a red mark on the part where the value was found to be (1,0).
Indicates adding Yellow. FIG. 6b shows a display example. In the case of the above example, the sixth
The diagram in Figure b is displayed, and the 600 copies are colored yellow, informing the system user that there is a 1 fixed failure.

In the above embodiments, the case where there is only one failure location is shown, but it is easy to extend the system to failures at multiple locations. In this case, FIG.
What is necessary is to change f to a model in which multiple parts fail. Furthermore, although the circuit in Figure 3 has one output, if multiple outputs can be observed, define f in Figure 1 170 for each output signal, and use the data in Figure 4. Prepare the actual output storage columns of the table for the number of output signals, and calculate the deviation δ of 185 in Fig. 1.
As such, the following formula may be used.

δ= 〓 ⁱ 〓 ^k (O _ik ^* O _ik ) where O _ik is the output signal value of the first output terminal under the k-th input signal.

〓 ⁱ 〓 ^k means to create the sum of all output terminals and all input signals.

In some cases, there may be a plurality of parameter vectors P that satisfy the determination 265 in FIG. In this case, all parameter vectors that satisfy the conditions may be investigated, and the diagnosis result may be a list of failure parts corresponding to each parameter vector. At this time, although it is not possible to find a single faulty part through diagnosis, parts that may be faulty are restored.

As can be seen from the above description, when this method is adopted, the task of creating a fault dictionary in advance can be omitted, and non-destructive diagnosis of the combinational circuit can be performed.

Next, an example will be shown in which the present invention is applied to non-contact measurement of carrier distribution within a silicon wafer in a semiconductor manufacturing process. As mentioned above, in Japanese Patent Application No. 56-17012, linearly polarized infrared light is incident on the wafer surface,
A method for measuring carrier concentration distribution has been proposed, which is characterized in that the distribution of carrier concentration in the depth direction within a wafer is determined by measuring the state of change in the polarization of reflected light.

As shown in Fig. 7, when monochromatic linearly polarized light is incident on the wafer surface at an incident angle θ, the incident light electric field and reflected light electric field are E _S , E _P , E _S ′, E _P ′
(where S and P represent the S-polarized light component and the P-polarized light component, respectively), the following relational expression exists between the incident light and the reflected light.

E _S ′ E _P ′＝r _S e ⁱ 〓 ^S 00 r _P e ⁱ 〓 ^P E _S E _PHere , r _S e ⁱ 〓 ^S and r _P e ⁱ 〓 ^P are the reflections of S-polarized light and P-polarized light, respectively. It is the amplitude.

When monochromatic linearly polarized light is incident on the surface of a desired substance and the state of change in polarization of the resulting reflected light is measured, the quantities Δ and Ψ defined by the following equations can be actually measured.

Ψ=tan ^-1 (r _P /r _S ) Δ=φ _P −φ _S This measurement method is called ellipsonmetry.

Now, the above r _S , r _P , φ _S , and φ _P are determined by the dielectric constant ε(ω) of the material into which the polarized light is incident (ω is the angular frequency of the incident light) and the incident angle θ. Therefore, the data Ψ and Δ obtained by ellipsometry are functions of the dielectric constants ε(ω) and θ of the material.

Here, the dielectric constant ε(ω) can be approximately calculated by the following formula when infrared light is used as the incident light beam.

ε(ω)=ε _∞ −4π・Ne ² /m ^* ω ²・1/1+i/
ωτ where ε _∞ = material constant, m ^* = effective mass of carrier, τ
= carrier relaxation time, ω = incident light angular frequency, e
=electronic charge, N=carrier concentration As shown at 710 in FIG. 7, when the carrier concentration is given as a function N(z) of depth z,
The dielectric constant is a function of z. This is ε(ω, N(z))
Write.

To summarize the above explanation, data (Δ, Ψ) obtained by injecting infrared light into a specific point on the wafer and observing the state of change in polarization of the reflected light.
is a function of the angular frequency ω of the incident light, the carrier concentration distribution N(z) at a specific point on the wafer, and the incident angle θ of the light beam. This is tentatively written as follows.

Ψ=f ₁ (ω, N(z), θ) Δ=f ₂ (ω, N(z), θ) The specific calculation method for the above functions f ₁ and f ₂ that give Ψ and Δ is well known. Therefore, I will omit the explanation here.

From the above equation, it can be seen that the measured values of (Δ, Ψ) when ω is kept constant and θ is varied variously reflect N(z). Similarly, it can be seen that the actually measured values of (Δ, Ψ) when θ is kept constant and ω is varied also reflect N(z). A method for determining N(z) using this fact is shown in the earlier application mentioned above. FIG. 8 shows one embodiment of this invention. In the figure, 80
1 is an infrared light beam, 802 is a spectrometer, 803 is a polarizer, 804 is a silicon wafer, 805 is a rotating analyzer, 806 is a rotary encoder, 807 is a detector, 808 is an interface, 809 is a mini computer, and 810 is a large size It's a computer.
In this example, a light beam emitted from an infrared light source 801 is converted into monochromatic light by a spectroscope 802, then linearly polarized through a light polarizer 803 and incident on the surface of a silicon wafer 804. The reflected light from the surface of 804 passes through a rotating analyzer 805 and an infrared detector 8
Detected by 07. The wavelength (λ) of the spectrometer 802 is determined from the minicomputer 809 via the interface 808, and also from the detection output of the detector 807 and the rotary encoder 80.
The angle signal from 6 is sent to minicomputer 809 via interface 808. The obtained ellipso data (Δ(λ), Ψ(λ)) is handled by a minicomputer 809,
The obtained data is processed by a large computer 810 to determine the carrier distribution.

FIG. 9 shows another embodiment of the above invention, in which ellipso data is measured by arbitrarily changing the incident angle, in which a laser beam emitted from an infrared laser 911 passes through a rotating polarizer 912. The light is incident on the silicon wafer 904. The angle of incidence θ is set to a desired value by a rotation stage 913 operated according to instructions from a minicomputer 909. Analyzer 90
5 and an infrared detector 907 are mounted on a rotating stage 91
Linked with 3. The detection output of 907 is sent to a minicomputer 909 via an interface 908 along with an angle signal from a rotary encoder 914.
ellipsodata (Δ(θ), Ψ(θ))
The processing is executed by a large computer 910.

In the above invention, the results of processing a case where the carrier concentration distribution can be approximated by a rectangle using the embodiment in FIG. 9 are also shown. This is shown in FIG. The white circles are actually measured values, and the rectangles are those obtained using a large-scale computer that closely match the actual measured values.

The actual carrier distribution N(z) within the wafer is not necessarily rectangular, but often has a complicated shape. Therefore, a method for improving the accuracy of application of the invention of the prior application using the present invention will be described below. As an example, as ellipso data, (Δ(θ), Ψ
Although the example of FIG. 9 for obtaining (θ)) is used, a similar method can be applied to the example shown in FIG. 8 for obtaining (Δ(λ), Ψ(λ)).

In this case, the functions of the large-sized computer 910 and mini-computer 909 in the prior invention of FIG. 9 are performed by the computer 150 in FIG. 1. The interface 908 in FIG.
This corresponds to 130 and 140 in the figure. The inspection target 100 in FIG. 1 is the wafer 904 in FIG.
3,914, the output signal observation device 120 is 90
5,907.

Now, as the model f of the target F (in this example, the physical structure of the carrier distribution measurement site on the silicon wafer) shown at 170 in FIG. 1, the following f ₁ and f ₂ shown earlier are used. use

Ψ=f ₁ (ω, N(z), θ) Δ=f ₂ (ω, N(z), θ) Here, when the carrier distribution N(z) exhibits various complicated shapes, these In order to approximate with high accuracy, the following formula is used as N(z).

N(z)≡P ₀₁ +〓 ^I _i=1 P _i1 exp [−(z−P _i2 ) ² /σ(P _i3
, P _i4 )] Here, P ₀₁ , P _i1 , P _i2 , P _i3 , and P _i4 are all parameters ≧0, and when α>β, they are selected so that P〓 ₂ ≧P〓 ₂ . shall be. σ(P _i3 , P _i4 ) is z≧P _i2
It is a function that becomes P _i4 when z<P _i3 and P _i3 when z<P i3.

As shown in FIG. 11, the above formula has the median value z=
With P _i2 as the border, I normal distributions with asymmetric spreads P _i3 and P _i4 on the left and right are added with P _i1 as the weight,
It is the sum of the constant P ₀₁ .

Figure 12 shows the parameters for the case I=2.
Concentration distribution N (z, P) when P ₀₁ to _{P 24} are changed variously
(Hereinafter, the parameter may be written as P). As shown in the figure, it can be seen that various distribution shapes can be expressed by changing the parameter P.

By using this method, the previous model f ₁ ,
f ₂ can be changed to a model that uses the parameter P as a variable instead of N(z), as shown in Fig. 1.17.
Use this as 0.

Ψ=f ₁ (ω, P, θ) Δ=f ₂ (ω, P, θ) An example of the data table 180 in FIG. 1 is shown in FIG. In this table, actual measured values of ellipso data at various angles of incidence are stored through the processing up to 245 in FIG.

The following equation is used as the deviation function δ of 185 in FIG.

δ= 〓 〓 〓 [(Ψ ^* (θ) − Ψ (θ)) ² + (Δ ^* (θ) − Δ(θ
)) ² ] This is the actually measured ellipsodata (Δ ^* (θ), Ψ ^*
(θ)) and the data (Δ(θ), Ψ(θ)) calculated by the functions f ₁ and f ₂ , sum of squares for all specified incident angles. .

As the function δ, in addition to the above formula, it is also possible to use, for example, the following one.

δ=max( max 〓|Ψ ^* (θ) − Ψ(θ)|, max 〓|Δ ^* (θ) − Δ(θ)|) This means that for all incident angles, Ψ ^* (θ) and Ψ (θ) and the absolute value of the difference between Δ ^* (θ) and Δ(θ)
In this method, the deviation between the measured data and the calculated data is determined by using the larger value among the maximum absolute values of the difference between the measured data and the calculated data.

FIG. 14 shows the state when the process of FIG. 2 of the present invention is applied to this example. The data read in the processing up to 245 in Fig. 2 (Fig. 13) also reads the incident angular frequency ω shown on the left of Fig. 10,
In this example, since it is a constant value, its display is omitted. 25 in Figure 2
0 processing, the parameter P is assumed, for example, as shown in the case 1 column of the figure. In the process of 255, estimated output Δ=f ₁ using the previous models f ₁ and f ₂
(ω, P ₁ , θ) and Ψ=f ₂ (ω, P ₁ , θ) are calculated for each value of θ. In process 260, the deviation between the measured ellipso data value and the calculated value is calculated using the function δ shown above. With the judgment of 265,
It is checked whether δ is larger than a separately given reference value ε. FIG. 14 shows an example in which δ>ε and the process returns to step 250 and recalculates. The parameter P ₁ assumed in the first calculation has been changed to P ₂ . The following processes are similar, and the above steps are repeated until the determination condition of process 265 satisfies δ≦ε.

FIG. 15 shows the results of process 270. The state of the object to be analyzed, calculated from P along with the parameter P
(P) (In this example, the carrier distribution in the depth direction N(z,
P)...see FIG. 12) is displayed in FIG. 1 190.

The above example shows the application of the present invention to the method shown in FIG.
From the relationship 2πc/ω, a similar method can be used to determine the carrier distribution using ellipsodata, which is equivalent to varying the incident wavelength λ, where c is the speed of light. In this case, Fig.
Ellipso data measured by fixing θ and varying λ may be stored in the data table 80. Model Ψ=f ₁ (ω,
P, θ) and Δ=f ₂ (ω, P ₁ , θ) can be left as they are. However, when calculating Ψ and Δ, θ uses a given fixed value, and ω is calculated from λ using (ω=2πc/λ, c: speed of light).

According to this embodiment, the carrier distribution measurement principle shown in the invention of the prior application can be applied with high accuracy when the carrier distribution within the wafer has a complicated shape.

Note that in this example, a method of expressing the intra-wafer carrier distribution as the sum of a constant and an asymmetric normal distribution function was used, but as another example, it is also possible to use an orthogonal function family such as a Fourier series. . In this case, the definition formula for N(z, P) may be changed.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing the basic configuration of the present invention.
FIG. 2 is a flowchart showing the processing procedure of the program shown in FIG. 1, and FIGS. Explanatory drawings, Fig. 7, Fig. 8, Fig. 9,
Figure 10, Figure 11, Figure 12, Figure 13, Figure 1
4 and 15 are explanatory diagrams when the present invention is applied to carrier distribution measurement. 100...Diagnosis target, 110...Input signal generating device, 120...Output signal receiving device, 130...
Computer output interface, 140... Computer input interface, 150... Computer, 160...
...Processing program, 170...Model f of diagnosis target 100 (output signal calculation program), 180...
...Data storage table, 190...Difference calculation program.

Claims (1)

[Claims] 1. An input means for applying a series of input signals to a diagnosis target, an output means for extracting a series of output signals from the diagnosis target, and a method for determining the diagnosis target based on the series of output signals from the output means. a diagnostic system having a processing means for diagnosing the condition, applying a series of input signals from the input means to the diagnosis target, and thereby outputting a series of output signals from the diagnosis target from the output means. The processing means generates a series of outputs corresponding to the series of input signals of the diagnosis target using a model formula for obtaining an output signal of the diagnosis target based on the input signal to the diagnosis target and parameters representing characteristics of the diagnosis target. obtaining a signal, and determining the deviation between the value of the obtained series of output signals and the value of the series of output signals from the output means by changing the value of the parameter,
A non-destructive diagnostic method characterized by determining the value of a parameter when the deviation satisfies a predetermined condition, and diagnosing the condition of the object to be diagnosed based on that value. 2 The above diagnosis target is a combinational logic circuit, the above input signal is an input signal to the above logic circuit, the above output signal is an output signal from the above logic circuit, and the above parameter is a combinational logic circuit. Used to create a model formula for the above logic circuit so that the signal value of the part can take any one of a normal value, a fault value that is always a value of 1, and a fault value that is always a value of 0. A non-destructive diagnostic method according to claim 1, characterized in that it is an additional logical value. 3. An input signal generating device that injects infrared light onto the surface of the semiconductor wafer, an output signal receiving device that detects reflected light from the semiconductor wafer, and a carrier distribution of the semiconductor wafer based on the detected reflected light. a diagnostic system having a processing device that determines the Changes in the deflection of the reflected light from the surface of the semiconductor wafer are measured by the output signal receiving device to obtain ellipsodata of the reflected infrared light, and the processing device calculates the wavelength or incident angle of the infrared light. Using a model formula for obtaining ellipso data based on the carrier distribution on the semiconductor wafer, the values of ellipso data for a series of wavelengths or angles of incidence of the infrared light are calculated, and the values of the ellipso data and the above are calculated. A non-destructive diagnostic method characterized by determining a deviation from ellipso data obtained by measurement and obtaining a carrier distribution on the semiconductor wafer based on the deviation.