JPH11312719A - Method for detecting iron contamination in cz silicon wafer - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a method for detecting iron contamination in a CZ-Si wafer, in a stat in which iron is captured by oxide deposit which can be applied to any conductive wafer. SOLUTION: This method includes a process for impressing an outside magnetic field to an Si sample 5, and polarizing an electronic spin system, a light irradiating process for irradiating the Si sample 5 with a light, and generating excitation excess carriers, a magnetic field sweeping process for changing the strength of an outside magnetic field to be impressed in a prescribed range, and a resonance detecting process for measuring the resonance change of the resistibility of the Si sample 5 during the magnetic field sweeping process accompanied with the resonance change at the re-combining speed of the excess carriers. Thus, the iron(Fe) contamination in a silicon can be detected.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

The present invention relates to iron (Fe) in silicon (CZ-Si) grown by the Czochralski method.
The present invention relates to a method for detecting contamination, a method for manufacturing a silicon wafer using the same, and a method for manufacturing a CZ-Si semiconductor device in a manufacturing line using the same. Specifically, the present invention provides:
The present invention relates to a method for detecting iron using electrical detection of carrier spin resonance.

[0002]

2. Description of the Related Art The performance of a semiconductor device is greatly affected even by extremely low concentrations of defects and unknowingly introduced impurities. Therefore, detection of such defects and impurities and measurement of the concentration are very important issues. Today's C
The purity levels of the Z-Si ingot and the wafer are extremely high, and the harmful concentrations of impurities and defects are below the level of 10 ¹⁰ cm ^-3 , requiring an ultra-sensitive measurement method.

[0003] Particularly harmful to electronic devices containing Si as a main component is iron that easily mixes into the Si material during the manufacturing process. Recently, several methods have been used to detect iron contamination in silicon wafers.

Electron spin resonance (Electron S)
pin Resonance (ESR) and deep level transient spectroscopy (Deep Level Transient S)
Although methods such as spectroscopy (DLTS) can detect iron in n-type and p-type Si wafers, the sensitivity of the ESR and DLTS methods is not sufficiently high for current contamination levels. Surface light voltage (Surface
Photo Voltage (SPV) measurement method and microwave photoconductive decay (Microwave Photo
Conductive Decay (μPCD) method
It is much more sensitive to Fe impurities than the ESR and DLTS methods, but can only be applied to p-type Si materials.

[0005] One of the possibilities for dealing with iron contamination in the current device manufacturing process is the use of the so-called internal gettering (IG) method. In the internal gettering, three-step annealing is performed on a CZ-Si wafer.
That is, high-temperature annealing, low-temperature annealing, and high-temperature annealing. During the first high temperature anneal, oxygen diffuses from near the surface area of the wafer. This step prevents the formation of oxide precipitates in the vicinity of the surface region, and produces a so-called defect-free region (DZ) near the surface of the wafer. The second annealing is CZ-Si
It is intended to produce oxide precipitates in oxygen-rich areas of the wafer. The third annealing increases the size of the precipitate generated during the second annealing. After the IG process, the CZ-Si wafer is composed of a clean, defect-free surface region and a bulk region having precipitates and other defects generated by the deposition process. The precipitates generated during the IG process have the ability to remove (capture) iron atoms from the surrounding Si and clean the interior of the wafer. Such iron precipitates modified with iron are outside the active region of the device and are not expected to affect device performance.

[0006] Iron atoms once captured by oxygen precipitates do not interact with boron atoms and cannot be detected by conventional SPV or μPCD methods.

As described above, the conventional iron contamination measurement method has various kinds of problems. For example, the ESR method and the DLTS method have insufficient sensitivity. The SPV method and the μPCD method can be applied only to p-type silicon wafers, and cannot detect iron atoms captured by oxide precipitates.

On the other hand, in the device manufacturing process using the IG method, the occurrence of iron contamination of silicon cannot be detected. Thus, it cannot even be determined whether the formation of a bulk with precipitates is a necessary step for good device performance.

[0009]

SUMMARY OF THE INVENTION The present invention has been made in view of the above circumstances, and an object of the present invention is to provide a method for detecting iron in a state in which oxide precipitates capture iron atoms. That is. This detection method can be applied to both conductivity types and has high sensitivity.

[0010]

According to a first aspect of the present invention, a polarization step of applying an external magnetic field to a Si sample to polarize an electron spin system; applying energy to the Si sample to generate excess carriers A microwave excitation step of arranging a Si sample in a microwave cavity irradiated with microwaves; a magnetic field sweeping step in which the intensity of an external magnetic field to be applied is changed; and an excitation carrier during the magnetic field sweeping step. And a method for detecting iron contamination in a silicon wafer.

According to a second aspect of the present invention, a polarization step of applying an external magnetic field to the Si sample to polarize the electron spin system; a light irradiation step of irradiating the Si sample with light to generate excess carriers; A magnetic field sweeping step in which the intensity of an external magnetic field to be changed is changed within a predetermined range; a resonance detecting step of measuring a resonance change in resistivity of a Si sample during a magnetic field sweeping step accompanied by a resonance change in a recombination speed of excited carriers; A step of detecting iron contamination in the silicon wafer.

According to a third aspect of the present invention, a Si wafer is cut (sliced, divided) into a size required for measurement,
A sample preparation step in which resistive contacts are provided at opposite ends of a sample and a lead wire is attached to the contact; an electron spin system polarization step in which an external magnetic field is applied to the Si sample; A light irradiation step of generating carriers; a magnetic field sweeping step performed by changing the intensity of an external magnetic field to be applied within a predetermined range;
A resonance detection step of measuring the resonance change in the resistivity of the Si sample in association with the resonance change in the recombination rate of the excess carrier; and a measurement step of measuring the level of iron contamination in the silicon wafer from the measured resonance intensity. A method is provided for measuring the level of iron contamination in silicon containing silicon.

According to a fourth aspect of the present invention, there is provided a method for irradiating a sample placed in a microwave cavity with microwaves to measure iron contamination in silicon.

According to a fifth aspect of the invention, the measurement is made with the sample temperature below 15K.

According to a sixth aspect of the present invention, a monitor wafer sampling step of slicing monitor samples from various portions of a CZ-Si ingot that has completed all the steps before slicing; A wafer preparation step for generating oxide precipitates inside; a sample preparation step for cutting out a Si wafer to a size required for measurement, providing resistive contacts at opposite ends thereof, and attaching leads to the contacts; A polarization step of an electron spin system performed by applying a magnetic field; a light irradiation step of irradiating a Si sample with light to generate excited excess carriers; a magnetic field sweeping step performed by changing the intensity of an applied external magnetic field within a predetermined range; During a magnetic field sweep step involving a resonance change in carrier recombination, the resonance change in the resistivity of the Si sample was measured, and the A CZ-Si ingot production line, comprising: a resonance detection step of detecting whether or not the nitrer wafer is contaminated with iron; and a step of determining whether the CZ-Si ingot is contaminated with iron based on the resonance detection step. A method is provided for detecting iron contamination in.

According to a seventh aspect of the present invention, a wafer which has been subjected to one or several ordinary steps of one processing line or different processing lines is used as a monitor wafer; A wafer preparation step of annealing a monitor wafer to form an oxide precipitate therein when no precipitate is generated;
A sample preparation step in which a Si wafer is cut out to a size required for measurement, resistive contacts are provided at opposite ends thereof, and leads are attached to the contacts; polarization of an electron spin system performed by applying an external magnetic field to the Si sample A light irradiation step of irradiating the Si sample with light to generate excited excess carriers; a magnetic field sweeping step performed by changing the intensity of an applied external magnetic field within a predetermined range; a magnetic field sweep accompanied by a resonance change in the recombination of excess carriers. During the process, a resonance change in resistivity of the Si sample is measured to detect whether or not the monitor wafer is contaminated with iron; Determining whether or not the semiconductor device is brought in. A method for testing for iron contamination in a semiconductor device manufacturing line is provided.

[0017] According to an eighth aspect of the present invention, a number of commonly used processing steps can be improved, changed, or eliminated based on the results of iron contamination level measurements in monitor wafers. -Si ingots, CZ-Si wafers and methods for manufacturing devices are provided.

[0018]

DETAILED DESCRIPTION OF THE INVENTION In the present invention, spin-dependent recombination (Spin Dependent Recombina) is described.
Tion, SDR) measurement method is used to measure iron contamination. Specifically, an electron paramagnetic resonance electric detection method (Ele
critically DetectedElectro
n Paramagnetic Resonance,
EDEPR or Electrically De
protected Magnetic Resonance
e, EDMR) to detect iron contamination. This method is known per se and exhibits high sensitivity to various defects and contaminants in silicon. The principle of this method lies in the law of conservation of angular momentum. Due to the conservation of angular momentum, the recombination of excess carriers happens to be spin-dependent. That is, before and after recombination, the recombination particle pairs must have the same total spin amount S. The total spin rate of the system after recombination is zero (S = 0). Prior to recombination, the system consists of a singlet state (S _s = 0) and a triplet state (S _tr =
1) is in a mixed state. Due to the conservation of spins, recombination from the S _s state proceeds at a high rate ( _accepted ) and recombination from the S _tr state is slow (disturbed). If the spin system is polarized by an external magnetic field, changing the triplet state of a recombination pair to a singlet state by applying a spin resonance condition to one of the recombination particles (electrons or holes). Can be. The conversion from the triplet configuration to the singlet configuration causes fast recombination of pairs that were not previously recombination pairs, thus increasing the overall recombination rate of the system. In the present invention, utilizing this principle, a non-recombined triplet state of a defect including iron is converted into a fast-coupled singlet state. Such a conversion increases the overall recombination rate of the entire system. Thus, the present invention detects iron contamination using the basic concept of spin-dependent recombination.

The detailed mechanism of SDR has not yet been elucidated, and it is only recently that a prima facie systematic study of the SDR effect in bulk Si samples has been reported. The detection of neutral interstitial iron atoms in heavily doped silicon quantum wells by SDR measurement has also recently been reported.

In the present invention, the SDR measurement method, specifically, E
The signal is detected from a sample contaminated with iron using DMR (EDEPR). The signal obtained in this way, the SDR signal, is the EDR obtained so far from silicon.
It has different characteristics from the PR signal. This signal is generated by the floating zone melting method and similarly contaminated silicon (FZ).
-Si, low oxygen content). This fact also indicates that this signal is due to defects including iron and oxygen.

All the results obtained can be explained consistently on the assumption that the above signal is due to iron-modified oxide precipitates.

[0022]

Hereinafter, measurement examples performed using the present invention will be described.

FIG. 1 shows an example of the measuring apparatus. The device is a standard ES with a TM ₀₁₁ mode resonator cavity 1
R measuring device 1 (using JEOLX band ESR spectrometer)
Based on At the time of measurement, the sample was a continuous helium gas flow cryostat 2 (Oxford Instrument)
uments Inc. ). The microwave power supply means 3 has a resonance frequency f _RES of about 9300 MHz and a maximum output of about 200 mW in the resonator cavity.
Generates electromagnetic waves. The sample holder 4 is made of a high-purity quartz tube. Sample 5 (size is about 3 × 8 ×
0.5 mm ³ ) and a quartz holder are placed in the resonator 1. The quartz tube and the sample can rotate in the direction of the magnetic field. The initial position of the sample is adjusted using laser light reflected from the front surface of the sample placed inside the microwave cavity.

The resonator has a window W for sample irradiation. Irradiation is performed by a white light lamp or laser 6 having a photon energy larger than the Si energy band gap.
This is performed using To minimize the microwave power absorbed in the wiring system, the leads coming out of the sample contacts are passed through a sample holder tube. The fixed magnetic field applied to the sample is modulated by a modulating coil disposed in a resonator at a frequency f _MOD of 100 KMz and an amplitude of 0.1 G. This frequency is used as a reference signal of the lock-in amplifier 7. Can be

The EDMR signal detection circuit includes a current source 8 (battery), a sample, and a resistor 9 connected in series. A signal is taken out of the resistor via the capacitor 10 and the lock-in amplifier 7 (EG & G model 511)
0) is passed to the input. SD generated in resistance during magnetic field sweep
The voltage change (± 50G) close to the R resonance condition is due to the EDMR signal,
That is, a signal related to the SDR process inside the sample. The output of the lock-in amplifier passes through the AD converter
It is connected to the PC 11 and represents the amplitude of the signal, that is, the Y-axis data. Magnetic field control system 12 for providing X-axis data of signals
Is connected to the PC via another AD converter. In this way, the spectrum of the EDMR, ie the dependence of the resistor voltage on the magnetic field, is recorded and used for subsequent calculations.

Next, the samples used in the measurement examples will be described.

Mirror-polished p-type CZ- having (100) plane
An Si wafer was used. Initial resistivity 3Ωcm and 10Ωc
m were used. The oxygen concentration in the wafer is
_CO O1.5 × 10 ¹⁸ cm ^-3 (ASTM F-121
Measured by FTIR method using -76 standard). To form a contaminated state, the wafers were exposed to different concentrations of FeCl _3.
Immersed in an SC-1 solution containing Thereafter, the wafer was spin-dried and annealed at 1000 ° C. for 1 hour in a gas flow of N ₂ : O ₂ = 10 (liter): 0.5 (liter). A wafer that had exactly the same initial state as described above, but was not contaminated, and was annealed under the same conditions, was prepared for comparison. The iron concentration of all wafers was measured by the SPV method according to the established measurement procedure. Some samples with an initial resistivity of 10 Ωcm were annealed at 450 ° C. for 60 hours to generate thermal donors and change conductivity type to n-type.

A sample having a size of 3 × 8 × 0.5 mm ³ was cut out from the wafer. Resistive contacts were bonded to opposite ends along the longitudinal axis of the sample using an ultrasonic soldering iron.

The results of the samples in these groups are not further differentiated, as the initial resistivity of the samples did not differ significantly from the EPR or EDMR assays. Samples annealed at 450 ° C. are referred to as “converted” samples.

Whether the sample is a p-type Si sample and a converted sample
Thus, the EDMR spectra shown in FIGS. 2A and 2B were obtained.
The presented EDMR spectrum shows that the iron atom concentration is C_Fe= 1.5 ×
10 ^FifteenAtom / cm^ThreeRecorded from a sample of
You. The sample is rotated in the <110> crystal direction,
The degree was set to 8K. These spectra are shown in FIG.
And 2B, referred to as KEM-1 and KEM-2, respectively.
Different.

Both KEM spectra are in the X-ray band EPR
The spectrometer could not separate the individual resonance lines. One narrow line can be reliably separated from one KEM-1 spectrum. This line has the largest amplitude in FIG.
Called EM-1T. The width of this line was fitted to 3.8G. The anisotropy of the g factor of the KEM-1T line with respect to the <110> axis of rotation is indicated by a mark in FIG. 3A. S =
The main g value for １ ／ is _gｇ = 1.9956 ± 0.0001 for the <011> type crystal axis,
<100> with respect to the type crystal axis g _|| = 1.9962
It can be fitted as ± 0.0001 (the fit curve is shown as a solid line, the error is within 3 degrees, which is consistent with the sample placement accuracy in the cavity). Thus, the KEM-1T line exhibits trigonal symmetry. Similarly, two lines having the same width can be separated from the KEM-2 spectrum.

These lines are from 30 ° to 50 ° in FIG. 3B.
(KEM-20)
line). The width of these lines was fitted to 2.9G.
The anisotropy of the g-factor of the two lines with respect to the <110> axis of rotation is indicated by the triangles and triangles in FIG. 3A.
The fitting (indicated by the solid curve in FIG. 3A) shows that the signal is due to a similar defect with orthorhombic-I symmetry. The main values are g ₁₁ = 1.9926 ± 0.0001 and g ₂₂ = 1.9962 ± 0.0001 for the <011> type crystal axis, and <10
For the 0> type crystal axis, g ₃₃ = 1.99928 ± 0.
0001 (the fitting error in this case is about 5
Degrees).

The symmetry of the defect obtained is that which is normally found in composites containing transition metals in silicon. This symmetry is also characteristic of oxide precipitates in silicon.

The next interesting characteristic of the EDMR signal was that the amplitude had a strong anisotropy (as can be seen from FIGS. 3A and 3B, there is a strong dependence of the signal amplitude on the rotation angle. ). The amplitude anisotropy of the KEM-1T line is indicated by a triangle in FIG. 3B. The solid curve in the graph is obtained by fitting the cos ² φ dependence to the experimental data. The anisotropy of the integrated intensity of the KEM-1 and KEM-2 spectra can be fitted by a similar dependency, but at low signal-to-noise ratios the accuracy is poor, probably due to the uncertainty of the integration procedure. The observed phenomena is probably not due to changes in the illumination conditions during rotation of the sample (the walls of the microwave resonator are mirror-polished, so that during rotation the sample receives almost the same intensity of illumination. receive). Due to the triplet nature of the recombination center, amplitude anisotropy of the EDMR signal can occur.

KEM-1T line and KEM-1 and KEM
-2 The similarity in the amplitude anisotropy of the entire spectrum is
This suggests that all EDMR lines in the EM spectrum are due to similar defects. To further confirm this assumption, EDMR measurements of the samples were performed at different experimental conditions (microwave power, irradiation intensity, and temperature). Under different experimental conditions (measurements were repeated for several angles of rotation), the KEM-1 and KEM-2 spectral shapes were the same. These observations support our assumption that all signals in the KEM spectrum are due to the same defect.

The microwave power dependence of the EDMR spectrum intensity and A _EDMR of both the KEM-1 and KEM-2 spectra can be fitted by the power law A _EDMR ∝P _MW ^0.8 . The spectrum is not saturated even at the maximum power obtained (200 mW).

The irradiation intensity was controlled by inserting various neutral filters in front of the window of the microwave resonator. Strength EDMR line initially increases in the form of I ^{0. 5} to the irradiation intensity I. In the high irradiation intensity region, the dependence saturates and finally begins to decrease. This decrease in the EDMR signal in the high illumination intensity region can probably be explained by considering that the high intensity illumination generated carriers in the surface region of the sample. In such a case, the sample volume is also partially shielded from the penetration of microwaves and the light itself.

The temperature dependence of the amplitude of the KEM-1 spectrum and the amplitude of the signal from phosphorus in the neutron-irradiated CZ-Si sample is indicated by ■ in FIG. The dependence A∝T ^-5 (plotted with a solid line in FIG. 4) shows KEM for temperature T> 13K.
-1 spectra were obtained by fitting experimental data.

The change in resistivity at the time of magnetic field resonance is proportional to the EDMR signal amplitude (A （Δρ). Low-temperature nonresonant photoconductivity depends on the concentration of excited carriers and varies with temperature according to the σ = 1 / ρ∝T ³ rule. Therefore, the ED shown in FIG.
From the fitted temperature dependence of the MR signal amplitude, the KEM
For measurements on spectra (T> 14K), Δρ /
ρ∝T ^-2 is obtained.

Another interesting feature regarding temperature dependence is:
The noise level in all measurements (see error width for KEM and phosphorus signal in FIG. 4) was lowest around 12K. The noise increased slowly at lower temperatures and sharply at higher temperatures.

The KEM signal was observed only in samples prepared from iron contaminated CZ-Si wafers.
No EDMR signal was observed from the initial wafer and the uncontaminated but annealed comparative wafer. Example spectra from samples of different contamination levels are shown in FIG. 5A. Dependence of EDMR signal intensity of KEM-1 on iron concentration in sample A _EDMR (C _Fe )
Is indicated by a circle in FIG. 5B. The difference in the sample preparation process is that the FeCl
_There was only a concentration of ₃ . Therefore, the KEM signal is CZ
-It can be concluded that it is associated with iron contamination of the Si wafer.

The KEM spectrum is in the EPR measurement mode.
Not observed. This is the KEM EDMR spec.
The upper limit of the concentration of defects that govern the torque is ≦ 10 ¹³cm^-3(E
(The sensitivity limit of the device in the PR mode).
This is well below the level of contamination of iron atoms. Furthermore,
As can be seen from FIG. 5B, the iron concentration of the EDMR signal of KEM
Is not linear and the concentration C_Fe≧ 10¹⁴cm^-3
Then saturates. This behavior is observed in all samples.
Exist at the same concentration and interact with iron atoms
This can be predicted from the defect that is "EDMR activated".

To gain further clues to the description, CZ-
FZ-S contaminated with iron in the same manner as the Si sample
The i-sample was analyzed. No EDMR signal was obtained from iron contaminated samples of FZ-Si.

These facts allow the assumption that the KEM spectrum is due to defects including iron and oxides. It is well known that oxide precipitates (OP) attract iron atoms in Si bulk (so-called internal gettering of iron). Therefore, it can be assumed that the KEM signal is due to such iron-modified OP. To prove this possibility, the concentration of OP in the sample was measured by a standard procedure of etching the sample with a Wright solution. The average OP density calculated based on the observation is C _OP = (1.1 ± 0.8) ×
It was 10 ⁸ cm ^-3 . Dependence on iron concentration C _OP in samples was detected (free of Fe is C _OP was the same even in samples annealed at the same conditions).

EDM for increasing iron concentration in sample
The saturation of the R signal (see FIG. 5B) is within the range of the above assumptions, but suggests that the KEM signal is due to the iron-modified OP interface layer. With this assumption in mind,
The size (radius) of the OP was calculated from the experimental results. The iron density on the surface of one OP is Si on the (001) crystal plane.
From the atomic density, it is considered that σ _Fe ~ 1 × 10 ¹⁵ cm ⁻² .
The data in FIG. 4B was fitted with a Hill-type function to obtain a threshold value A _EDMR (C _Fe ) for saturation of iron atoms on the OP surface. Wherein ^{_{A EDMR = A EDMR max (C}} Fe n) / (C Fe n + C T n) (1) where the maximum level of A _EDMR ^max is EDMR signal, C _T is the saturation threshold, n represents the speed of the saturation process Parameter. The fit dependence (1) is shown by the solid line in FIG. 5B and gives a saturation threshold C _T = 6.4 × 10 ¹³ cm ⁻³ and a saturation velocity n = 5.1. Such a large value of n indicates that the modification process is very fast. The average area value of OP can be estimated from the following equation. _{S OP = (C T) /} (C EMD × σ Fe) = 6 × 10 -10 cm 2

Considering the annealing conditions, it would be expected that in our sample the oxide would precipitate in the form of platelet precipitates (PLP). The PLP can be schematically represented as a plate with a radius R _PLP and a height D _PLP . The thickness of the PLP is considered to be constant in the region where the radius is large, and 5
nm. Therefore, a simple calculation yields an average radius of <R _PLP > ~ 10 nm, which is quite appropriate as the average value of PLP in CZ-Si after annealing. Therefore, EDMR KEM
Our assumption that the signal is due to the iron-modified OP seems reasonable.

The change from a KEM-1 spectrum to a KEM-2 spectrum during prolonged annealing at 450 ° C. may be related to a structural change in the PLP surface. The symmetry of the KEM-1 spectrum is PLP (<
001> type crystal axis), and the symmetry of the KEM-2 spectrum is <110.
> Suggests further arrangement in the type direction. Furthermore, it can be assumed that annealing at 450 ° C., which does not cause growth of the OP, allows for the relaxation and alignment of the OP's surface structure.

Assuming that the origin of the EDMR signal of KEM proposed above is correct, modification by iron atom changes OP to "SD
R activity "defect. Therefore, OP
Before modification with a metal (iron) atom, it can be assumed that the activity is much lower in the process of recombination of excited carriers than after modification. However, it is also possible to propose an assumption that explains the SDR activation of the OP based on the triplet properties of the defect that causes the KEM signal. In other words, it is also believed that iron modification creates a triplet state defect, which causes OP to become SDR active.

The circles in FIG. 5B indicate CZ-S having different iron concentrations.
This shows the noise level when EDMR measurement was performed on the i-contaminated sample. The noise level was calculated from the experimental EDMR spectrum extracted as the standard deviation of the EDMR signal from zero level at a field range of 10 G away from the resonance line (signal from 40 points was analyzed). As is evident from the graph, there is a strong correlation between the noise level during the experiment and the iron concentration in the sample (the dashed line in FIG. 5B indicates the fitted dependence, σ ₄₀ ∝C _Fe ^0.7 ). . Presumably, the level of the signal to noise ratio in the high defect concentration sample is too low to detect the EDMR signal.

E corresponding to the EDMR spectrum of KEM
EDMR measurements appear to be more sensitive than EPR, since no PR signal was detected. Assuming that the KEM signal is due to iron-modified OP, the sensitivity limit is 1
It will be at a level of about 2 × 10 ⁶ per cubic centimeter (estimated from EDMR signal to noise ratio and OP concentration in the sample). Taking into account the size of our sample, it is possible to detect EDMR signals from only 25000 precipitates.

In summary, an EDM called KEM-1
The R signal was detected from a sample cut from an iron-contaminated CZ-Si wafer. Annealing the sample at 450 ° C. for 60 hours caused a change in the initial EDMR spectrum (the KEM-1 spectrum was changed to KEM-2).
Spectra), but the dependence of the EDMR signal on the different experimental conditions did not change. The KEM-1 signal shows trigonal symmetry, while the KEM-2 signal shows orthorhombic-I symmetry. The spectrum shape does not change even under different experimental conditions (temperature, microwave power, irradiation intensity). This also indicates that all EDMR lines in the spectrum are due to the same defect. Both the KEM-1 and KEM-2 signals show strong anisotropy in amplitude, suggesting that recombination occurs through the center excited triplet state.

Dependence of EDMR Signal Strength on Iron Concentration in Wafer A _EDMR (C _Fe ) is not linear, indicating a rapid saturation of EDMR signal strength with increasing iron concentration.
This type of dependence can be explained by the interaction of the iron atom with any defects already present in the sample. Since no EDMR signal was detected from FZ-Si samples contaminated with iron under similar conditions, it can be assumed that the KEM signal is due to oxygen-related defects modified with iron atoms, i.e., oxide plate-like precipitates. did it. From the obtained experimental results,
Within the above assumptions, the average radius of oxide precipitates in our samples was estimated to be <R _PLP > ~ 10 nm. This value is quite reasonable as the radius of the plate-like precipitate in the CZ-Si wafer after annealing.

The noise figure during EDMR measurement increases with the concentration of recombination centers in the sample. The observed effects may explain what was previously reported about the difficulty of EDMR measurements on samples with high concentrations of recombination centers. Temperature dependence of the observed EDMR signal, and Fe in p-type samples with thermal donors
^{0 The} absence of a signal from the recombination center cannot be explained in terms of the existing model of the SDR procedure.

This is an example of measuring iron contamination in a Si sample.

Further, the iron contamination level in the processing line can be measured as follows. The monitor CZ-Si wafer after completing all the steps of the processing line is cut into a size desired for EDMR measurement, and iron contamination measurement described below is performed. Processing can be improved depending on the results obtained for the iron concentration inside the monitor wafer. For example, if the level of Fe contamination is below an acceptable value, the cleaning process and even the second half of the IG process can be omitted. As a result, manufacturing costs can be reduced.

[0056]

As described above, according to the present invention, the CZ-S is obtained by utilizing the SDR method or the EDMR method.
Iron contamination of the i-wafer can be detected with high accuracy.

[Brief description of the drawings]

FIG. 1 is a diagram showing an embodiment of a measuring apparatus used in the measuring method of the present invention.

2A and 2B are EDMR signals of a Si crystal obtained in an embodiment of the present invention, mutual arrangement of fixed magnetic fields,
FIG. 9 is a view showing a result of measurement of a relationship with a crystal orientation.

FIGS. 3A and 3B are diagrams showing the relationship between the crystal orientation toward a fixed magnetic field and the g-factor and intensity of EDMR, obtained in an example of the present invention.

FIG. 4 is a diagram showing the temperature dependence of an EDMR signal obtained in an example of the present invention.

FIGS. 5A and 5B show EDMR signals obtained from samples with different degrees of iron contamination.

[Explanation of symbols]

 REFERENCE SIGNS LIST 1 resonator 2 cryostat 3 microwave power supply means 4 sample holder 5 sample 6 laser 7 lock-in amplifier 8 current source 9 resistor 10 capacitor 11 PC 12 magnetic field control system

Claims (8)

[Claims] 1. A method for detecting iron contamination in silicon, comprising: applying an external magnetic field to a Si sample to polarize an electron spin system of the Si sample; and applying energy to the Si sample to generate excess carriers. Performing a magnetic field sweep by changing the intensity of an applied external magnetic field within a predetermined range; and performing a resonance change at a coupling speed of excited carriers during the magnetic field sweep. Detecting iron contamination in the silicon wafer. 2. A method for detecting iron contamination in silicon, comprising: applying an external magnetic field to a Si sample to polarize an electron spin system of the Si sample; and irradiating the Si sample with light to be excited. Generating an excess carrier; sweeping the magnetic field by varying the strength of the applied external magnetic field within a predetermined range; and resonating the resistivity of the Si sample during the magnetic field sweep with a resonant change in the recombination rate of the excess carrier. A method comprising: detecting resonance by measuring a change; and detecting iron contamination in a silicon wafer. 3. A method for measuring the level of iron contamination in silicon, comprising cutting a Si wafer into a size required for measurement, providing resistive contacts at opposite ends thereof, and connecting leads to the contacts. Forming a sample; applying an external magnetic field to the Si sample to polarize the electron spin system of the Si sample; irradiating the Si sample with light to generate excited excess carriers; and Sweeping the magnetic field by varying the intensity in a predetermined range; and detecting the resonance by measuring the resonant change in the resistivity of the Si sample during the magnetic field sweep with a resonant change in the recombination rate of excess carriers; Calculating the iron contamination level in the silicon wafer from the resonance intensity measured in the resonance detection step. 4. The method for detecting iron contamination in silicon according to claim 3, wherein the sample placed in the microwave cavity is irradiated with microwaves. 5. The method for detecting iron contamination in silicon according to claim 4, wherein the measurement is performed at a temperature lower than 15K. 6. A method for detecting iron contamination in a Si ingot production line, comprising: collecting monitor wafers from various portions of a CZ-Si ingot having completed all processes before slicing; and annealing the monitor wafer. Preparing oxide wafers in the wafer by preparing oxide wafers, and cutting out the prepared Si wafer to a size required for measurement,
Forming resistive contacts at opposite ends thereof and attaching leads to the contacts to form a sample; applying an external magnetic field to the Si sample to polarize the electron spin system of the Si sample; A step of irradiating light to generate excited excess carriers, a step of sweeping a magnetic field by changing an applied external magnetic field intensity within a predetermined range, and a step of sweeping a magnetic field accompanied by a resonance change in a recombination rate of excess carriers. Detecting the resonance by measuring the change in the resonance of the resistance of the CZ-Si ingot, and determining whether the CZ-Si ingot is contaminated with iron based on the detection. Method. 7. A method for inspecting iron contamination in a semiconductor device manufacturing line, wherein a wafer which has completed at least one normal processing step in one processing line or a different processing line is used as a monitor wafer; If oxide precipitates have not yet formed in at least one processing step, a step of annealing the monitor wafer to generate such precipitates in the wafer; and cutting out a Si wafer to a size required for measurement. Providing resistive contacts at opposite ends and attaching leads to the contacts to form a sample; applying an external magnetic field to the Si sample to polarize the electron spin system of the sample; irradiating the Si sample with light To generate excited excess carriers by changing the strength of the applied external magnetic field within a predetermined range. During the magnetic field sweeping step, which involves the resonance change in the recombination rate of the excess carriers, by measuring the resonance change in the resistivity of the Si sample to determine whether the monitor wafer is contaminated with iron. A method comprising the steps of: detecting a resonance to detect; and, from the resonance detection step, determining whether or not the processing and the processing line have caused iron contamination on the wafer. 8. A CZ-Si ingot, CZ-C, characterized in that at least one processing step which is normally performed is improved, changed or eliminated based on the result of the iron contamination level measurement in the monitor wafer. Manufacturing method of Si wafer and device.