Method and apparatus for in-situ determination of the temperature of a wafer
By adjusting the spectral intensity distribution of the reflection measurement to align effective wavelengths, the method addresses residual oscillations in emissivity-corrected pyrometry, providing accurate temperature control in coating processes with thick, transparent layers.
Patent Information
- Application Number
- DE102020111293
- Authority / Receiving Office
- DE · DE
- Patent Type
- Patents
- Current Assignee / Owner
- Priority Date
- 2019-04-24
- Filing Date
- 2020-04-24
- Publication Date
- 2024-11-07
- Estimated Expiration
- 2040-04-24
Smart Images

Figure 00000000_0000_ABST
Abstract
Description
[0001] The invention relates to a method and a device for in-situ determination of the temperature of a semiconductor substrate (hereinafter referred to as wafer), in particular a method and a device for determining the temperature of a wafer in coating processes by means of emissivity-corrected pyrometry (ECP).
[0002] Methods for determining the temperature of a sample and temperature measuring devices with a pyrometer and an evaluation unit connected to the pyrometer for signal transmission are well known. Modern process pyrometers for thin-film processes are emissivity-corrected. This means that, in addition to the pyrometer detection, a reflection measurement is also integrated at the same wavelength. This reflection measurement is used to determine the current emissivity of the wafer during the process, which is modified by a layer structure on the surface, and thus to appropriately correct the pyrometer signal, enabling a more precise temperature determination.
[0003] DE 44 19 476 C2 discloses a corresponding method and device for determining the layer thickness and the substrate temperature during the coating of semiconductor substrates (wafers). The core of the emissivity-corrected pyrometry method is that, using a single, combined apparatus at a suitable wavelength λ P both a pyrometric measurement of the thermal radiation E(T,λ P ) as well as a reflection measurement R(λ P ) at the wafer surface. Since under certain, generally well-fulfilled conditions, the reflection R(λ P ) directly the emissivity ε(λ P ) = 1 - R(λ P ) of the wafer surface can be calculated, thus the true temperature of the wafer T Wafer by a simple algorithm f ECP which is directly derived from Wien's approximation of the Planck function P(λ p ,T) results in: TWafer=fECP{E(T,λP)}; E(T,λP)=P(T,λP)*ε(λP)
[0004] Through this procedure, the true temperature of the wafer T Wafer no longer distorted by layer thickness effects (e.g. Fabry-Pérot interference within the layer structure), which always significantly interfere with a “simple” pyrometry measurement that only evaluates the thermal radiation P(λ) of the wafer.
[0005] However, in practical applications of the ECP method, it has been found that for certain coating processes of multilayers on semiconductor wafers and where very high accuracy requirements are required, the established method is no longer sufficient to eliminate all Fabry-Pérot interference artifacts from the wafer temperature measurement with sufficient accuracy. This particularly applies to coating processes where • particularly thick layers or layer systems are applied, which also consist of layer materials which, at the pyrometer wavelength λ P are completely transparent (k = 0) and / or • Fabry-Pérot oscillations (FPO) with particularly large amplitude in the emissivity must be corrected.
[0006] The Fig. Figure 1 shows, as an example of such a process where the challenges for emissivity correction are particularly high, the reflection and pyrometry measurement data during HEMT (High Electron Mobility Transistor) epitaxy. In Fig. 1a clearly shows that without emissivity correction, oscillation artifacts are superimposed on the temperature signal, which are in exact antiphase to the corresponding reflection Fabry-Pérot oscillations. Fig. In contrast, Figure 1b shows that such apparent temperature oscillations, which are exactly in antiphase to the reflection, can be completely eliminated by an emissivity correction (ECP).
[0007] The challenge of emissivity correction is particularly high in such HEMT processes because the Fabry-Pérot oscillations in the reflection can have a very large amplitude (up to 35%). This is because the GaN layer material, which is transparent at the pyrometer wavelength, acts as an anti-reflection layer for the silicon substrate material. As a result, the emissivity ε(λ) of the wafer varies between 100% and 65%. Therefore, the emissivity correction ε(λ p ) = 1 - R(λ P ) does not have to correct a “small disturbance”, but an error influence must be correctly compensated, which modifies up to 35 % of the thermal emission radiation P(λ).
[0008] In Fig. 2 shows that under such conditions, particularly at the end of the HEMT coating process, i.e., after the growth of several micrometers of the transparent GaN layer material, residual oscillations occur in the ECP temperature signal according to the state of the art, which are no longer fully compensated by the conventional ECP process. These residual oscillations are no longer exactly in antiphase to the reflection Fabry-Pérot oscillations, but rather are shifted by approximately 1 / 4 of the oscillation phase relative to the reflection Fabry-Pérot oscillation when they first occur (from a certain larger total layer thickness). This corresponds to a so-called "derivative signature," since the oscillation phase of the temperature residual oscillation corresponds to that of the first derivative of the reflection oscillations. These temperature residual oscillations are generally relatively small (±1 K ... ±4 K), but can severely disrupt temperature control processes.For such control processes, the occurring temperature residual oscillations should always remain below about ±0.5 K.
[0009] The significantly increased demands in coating system control technology and new coating materials and structures also make it necessary to measure the wafer temperature with extremely low noise and absolute accuracy. Accordingly, in order to increase the intensity of the detected thermal radiation, the use of spectrally broader bandpass filters in pyrometers has become necessary for some applications. Spectrally broader bandpass filters not only transmit the narrow (±0.5 nm) spectral range around λ P , but are in a wider spectral range (e.g. λ P±5 nm) transparent. This spectral transparency window of the bandpass filters is referred to below as the bandpass spectral range. However, spectrally wider bandpass filters lead to residual oscillations in the temporal course of the measured wafer temperature in certain technologically important processes. Experiments have shown that these temperature residual oscillations are not a constant disturbance in conventional emissivity-corrected pyrometry, but rather can vary in amplitude with the total layer thickness, the layer material, the substrate material, and the wafer temperature.
[0010] US 6,398,406 B1 describes a method for determining the temperature of a surface during a coating process using ECP. DE 198 32 833 C2 relates to a method for the thermographic examination of a workpiece and a thermographic device therefor. US 6,299,346 B1 discloses a method and a device for the active pyrometric measurement of the temperature of a body whose emissivity varies with wavelength. DE 693 12 894 T2 relates to a pyrometer with an emission meter, wherein the emission meter comprises a radiation source providing measurement radiation and a measurement radiation detector for converting the measurement radiation reflected from the object surface into an electrical signal.
[0011] However, none of these patents solves the above-described problem of residual temperature oscillations, which arises in dynamic emissivity-corrected pyrometry when measuring optically thicker layers with the highest temperature precision. Therefore, the object of the invention is to overcome or at least mitigate the problems of the prior art and to provide a solution for eliminating the residual temperature oscillations that remain when using a conventional ECP method.In particular, the object is to provide a method and a device for determining the temperature of a wafer in coating processes by means of emissivity-corrected pyrometry, which also provide a correct, oscillation-free ECP temperature signal when they are used (to improve the signal-to-noise ratio) with a wider bandpass filter in layer growth processes for very thick layer stacks consisting of layers at λ. P transparent materials and lead to reflection Fabry-Pérot oscillations with very large amplitude.
[0012] According to the invention, this object is achieved by an improved method according to claim 1 for residual oscillation-free emissivity-corrected in-situ pyrometry.
[0013] The starting point for the improved ECP method is the reflection measurement on the surface of the wafer known from the conventional ECP method (e.g. according to DE 44 19 476 C2), where the reflectance R(λ P ) is determined from R(λP)=IR(λP) / I0(λP), ie, from the ratio of the reflected intensity I R (λ P ) to the intensity of the incident sample radiation I0(λ P). However, for the improved ECP method presented here, the spectral intensity distribution I0(λ) of the sample radiation can be specifically adjusted within the bandpass spectral range and is not, as in the conventional ECP method, fixedly predetermined by the reflected light source used. The sample radiation, whose spectral profile can be adjusted according to the invention, is preferably optical radiation from a near infrared (NIR) range between 0.78 µm and 3 µm, more preferably from the IR-A range between 0.78 µm and 1.4 µm. Particularly preferred is a sample radiation around λ P = 950 nm. A typical half-width (HWB) of the sample radiation is between 100 nm and 1 nm, more preferably between 75 nm and 25 nm. A sample radiation with a HWB of around 50 nm is particularly preferred.
[0014] The spectral intensity distribution I0(λ) indicates the dependence of the intensity of the sample radiation I o0of wavelength λ in the vicinity of the bandpass spectral range. The radiation I reflected from the wafer R (λ) (reflected sample radiation) and thermal radiation P(λ) emanating from the wafer are filtered by a spectral filter element with bandpass characteristics (i.e., with a corresponding form of the filter function). An ideal bandpass filter of a given half-width (HWB, in nm) has an exactly rectangular 100% transmission characteristic between (λ P - 1 / 2 HWB) and (λ P + 1 / 2 HWB). A real bandpass characteristic results from the real transmission curve of the bandpass filter (and possibly other optical components in the beam path of the ECP pyrometer). In the following, it is shown that the effective effective wavelengths λ e = λ A eff and λ P = λ B effthe reflection measurement (A) and the pyrometry measurement (B) despite using the common bandpass filter. The effect of this difference between λ A eff and λ B eff was in Fig. 2b: the period duration of the FPOs of both measurement methods (and thus also the resulting layer growth rates) are not identical. If one determines the period duration of the 950.0 nm reflection Fabry-Pérot oscillation (Δt R = 223.19 s) the growth rate, the resulting value is 0.8394 nm / s. The period of the Fabry-Pérot oscillation in the pyrometer raw signal, however, is barely noticeably longer (Δt pyro = 223.28 s). This means that the effective wavelength of the pyrometer measurement is not 950.0 nm, but λ P = 950.7 nm. This slight difference of 0.7 nm between the effective reflection wavelength λ Rand the effective pyrometer wavelength λ P During the several hours of growth of the several micrometer thick GaN layer, phase differences accumulate, which lead to the Fig. 2a shown residual oscillations in the temperature signal of the conventional ECP system.
[0015] For an ECP measuring system with a wider bandpass filter of transmission characteristic T F (λ) the pyrometry signal is calculated according to E(λP)=∫λminλmax{TF(λ)∗P(T,λ)∗ε(λ)}dλ and the reflection signal according to IR(λR)=∫λminλmax{TF(λ)∗R(λ)∗I0(λ)}dλ.
[0016] The effective wavelengths of the reflection measurement (λ R ) and pyrometry measurement (λ P ) are characteristic quantities of the measuring system that are independent of the sample properties. Without loss of generality, they can therefore be determined using a “grey radiator” (GS) with ε(λ)=0.5=1−R;→R=0.5 be determined.
[0017] For a GS the pyrometry signal is: EGS(λGSP)=0.5*∫λminλmax{TF(λ)∗P(T,λ)}dλ and the reflection signal: IGSR(λAeff)=0.5∫λminλmax{TF(λ)∗I0(λ)}dλ
[0018] The effective wavelengths λ A eff and λ B eff of the reflection measurement (A) and the pyrometry measurement (B) are those wavelengths which lead to equally large partial area integrals between λ min and λ eff and between λ eff and λ max lead (compare also Fig. 6). This shows that λ A eff and λ B eff are identical in any case if the spectral profiles of the reflected sample radiation I0(λ) and the Planck spectrum P(T,λ) are proportional to each other within the bandpass spectral range. λ A eff = λ B effcan also be achieved with other I0(λ) curves, but then the bandwidth attenuation of the FPO amplitudes of both partial measurement methods would be different. Setting the optimal spectral intensity distribution I0(λ) thus means: the spectral curve of the Planck function is reproduced within the bandpass spectral range for I0(λ), so that both the result λ R = λ P and the bandwidth attenuation of the FPO amplitudes affects both partial measurement methods as equally as possible. λ R and λ P are defined as follows: ∫λminλPTF(λ)∗P(T,λ) dλ=∫λPλmaxTF(λ)∗P(T,λ) dλ ∫λminλRTF(λ)∗I0(λ) dλ=∫λRλmaxTF(λ)∗I0(λ) dλ.
[0019] For the spectral filter element T F(λ) is a real bandpass filter, the transmission characteristics of which can be slightly modified by additional optical elements in the beam path. The preferred half-width (HWB) of the bandpass filter is between 5 nm and 30 nm, with an HWB of around 10 nm being particularly preferred. When applying the method according to the invention to extremely thick layer stacks, the bandpass HWB of the filter element can also be chosen to be very wide in order to attenuate the Fabry-Pérot interference by incoherent superposition (in reflection and pyrometry alike). This can reduce the effort required to match the two effective wavelengths λ A eff and λ B eff (A: reflection and B: pyrometry). In this case, the preferred HWB of the bandpass is between 100 nm and 300 nm, more preferably between 150 nm and 250 nm. A filter bandwidth (HWB) of around 200 nm is particularly preferred.
[0020] According to the invention, the effective effective wavelength λ is determined according to the above definition (Eq. (7) and the explanation therein). P = λ B eff the pyrometry measurement from the spectral intensity distribution of the Planck function P(λ,T) at a predetermined temperature T of the wafer and the bandpass characteristic T F (λ) of the spectral filter element.
[0021] Alternatively, λ B eff can also be determined experimentally from the increase N of the log 10 [P(λ,T)]=M+N*(1 / T) - calibration line of the pyrometer, which results directly from the Planck function.
[0022] Alternatively, λ B eff can also be determined experimentally from the period of a layer growth FPO in the pyrometer raw signal with known growth rate and known refractive index (see also Fig. 2b and explanations thereto).
[0023] Subsequently, the spectral intensity distribution I0(λ) of the reflection sample radiation is specifically adjusted to set the effective effective wavelength λ A eff the reflection measurement to the previously determined effective effective wavelength λ B eff of the pyrometry measurement. This is achieved by appropriately fitting the I0(λ) spectral response of the reflected light source to that of the Planck function. A quality criterion for the success of this fitting is the resulting agreement of both the period duration and the bandwidth attenuation of the film growth FPOs in the reflected and raw pyrometer signals.
[0024] Finally, a measurement of the total intensity of the filtered radiations at the same effective effective wavelength λ A eff , λ B effand a determination of the current temperature T2 of the wafer using emissivity-corrected in-situ pyrometry (ECP) from the measured total intensities of the filtered radiation.
[0025] To determine a temperature profile, these process steps can be repeated multiple times, whereby the current temperature T2 determined in the last step can be used as the starting temperature T1 for the next run. For example, an initial temperature determination can be carried out in advance on a wafer section that is not coated or not yet coated, or using another measurement method. Since the FPOs of the pyrometer measurement and the reflection measurement are identical (in period duration and amplitude attenuation) during the growth process, even for the wider bandpass filter, the algorithms of emissivity-corrected in-situ pyrometry (ECP) are fully applicable and lead to wafer temperature measurements without residual oscillations, even with very thick and transparent layer structures.
[0026] If the residual oscillations in the ECP temperature measurement are to be completely avoided in several temperature ranges, the effective effective wavelength λ A eff the reflection test radiation can be adjusted to suit the reflection measurement.
[0027] One idea of the invention is thus that during an ECP reflection measurement, an additional static or dynamic adaptation of the spectral intensity distribution I0(λ) of the reflection sample radiation to the spectral profile of the Planck curve within the bandpass spectral range is carried out.
[0028] Conventional ECP methods (e.g., according to DE 44 19 476 C2) require both the thermal radiation and the reflected radiation to pass through the same relatively narrowband (e.g., ±1 nm) filter. For narrowband filters, this already ensures that an identical wavelength (the center wavelength of the filter) is used for both measurements. However, the practical implementation of known ECP methods shows that spectral bandwidths between 5 nm and 30 nm are necessary to enable a sufficiently low-noise pyrometry signal even at the relevant low process temperatures (i.e., at low thermal radiation intensities) and / or high temporal resolution for temperature measurement.
[0029] However, with conventional ECP methods, when using bandpass filters with a HWB between 5 nm and 30 nm, the wavelength of the reflection measurement and the pyrometry measurement are generally not exactly the same. This has already been demonstrated with the data in Fig. 2 and the corresponding calculation of the effective reflection and pyrometer wavelengths. Only through the inventive precise adjustment of the reflection effective wavelength λ A eff to the pyrometry effective wavelength λ B eff (e.g., if the effective pyrometry wavelength of the thermal emission measurement is shifted by 0.7 nm to 950.7 nm instead of the assumed filter center wavelength of 950.0 nm), the accumulating phase differences between reflection FPOs and the FPOs in the pyrometer raw signal can be avoided, and the residual oscillations in the ECP signal can be significantly suppressed. This advantage of the improved ECP method ensures that the temperature control loops in the coating process remain unaffected by such residual temperature oscillations.
[0030] To achieve this, an inventive static or dynamic adjustment of the spectral intensity distribution I0(λ) of the reflection sample radiation to adjust the effective effective wavelength λ A eff the reflection measurement to a predetermined effective wavelength λ B eff of the pyrometry measurement. The resulting amount of this necessary adjustment Δλ A eff of λ A eff depends on the slope and the mean value of the Planck function P(λ,T) in the bandpass filter spectral range and thus varies with the wafer temperature T Wafer (cf. Fig. 4a). Although the absolute slope of the Planck function P(λ,T) in the spectral range of the bandpass filter increases with temperature T, the relative slope – which is responsible for the shift Δλ of the pyrometric effective wavelength – decreases with increasing temperature T (cf. Fig. 4b).
[0031] The effect of shifting the effective wavelength λ B eff The thermal radiation is therefore dependent on the temperature. The difference in the effective wavelengths of the pyrometry measurement and the reflection measurement completely and clearly explains the Fig. 2a shows the phase behavior of the temperature residual oscillations (“derivative signature” with a still relatively small phase shift between pyrometry and reflection Fabry-Pérot oscillations due to the wavelength difference Δλ).
[0032] With the method according to the invention, the temperature residual oscillations always remain below ±0.5 K even under extreme conditions (e.g., when Fabry-Pérot oscillations with particularly large amplitudes in the emissivity are to be corrected) and are thus also suitable for sensitive temperature control purposes. Furthermore, the improved method allows the wafer temperatures to be determined precisely, especially when particularly thick layers or layer systems are applied, each of which consists of layer materials that are radiant at the pyrometer wavelength λ. B eff are completely transparent. The inventive method thus represents a significant improvement over conventional ECP methods.
[0033] To distinguish the filtered radiation, the sample radiation is preferably modulated so that it can be separated from the thermal radiation using a modulation method. This is particularly relevant when, during the step of measuring the total intensity of the filtered radiation, the effective effective wavelength λ A eff , λ B eff These are to be measured together (e.g., using a single-channel detector). Preferably, suitable modulation methods are used to separate the reflected radiation from the thermal radiation (either direct light source modulation, mechanical light chopper, or electro-optical modulators).
[0034] In a first embodiment of a method according to the invention, the test radiation is emitted by an LED as the light source, which can be directly electrically modulated. LEDs have the additional advantage that their spectral intensity distribution I0(λ) can be modified by deliberately changing the operating temperature and current supply to the LED. This is a common method for changing the emission wavelength of an LED beyond its operating temperature ("temperature tuning"). As the operating temperature increases, the mean emission wavelength increases, the maximum intensity decreases, and the HWB of the LED emission spectrum increases. This allows the spectral intensity distribution I0(λ) to be adjusted and stabilized in a simple and established manner and to be adapted to the shape of the Planck function within the limited bandpass spectral range.This procedure for adjusting and stabilizing the LED light source is not only of great advantage for the ECP process, but also for any LED-based in-situ reflectance measurement on long-duration thin-film processes.
[0035] In a second embodiment of a method according to the invention, the sample radiation is emitted by a thermal white light source as the light source, and the spectral intensity distribution I0(λ) is adjusted by changing the radiation temperature of the white light source. The spectrum of a thermal white light source approximately corresponds to that of a blackbody radiator and can thus also be described by the temperature-dependent Planck function P(λ,T). The spectral intensity profiles of the sample radiation and thermal radiation emanating from the wafer then largely match, so that only a minor adjustment of the spectral intensity distribution I0(λ) of the sample radiation may be necessary.In particular, the curves of the respective Planck function P(λ,T) of sample radiation and thermal radiation can be largely approximated within the filter bandwidth, so that even very wide bandpass filters can be used for high-precision pyrometry on thicker layer stacks with total layer thicknesses > 20 µm.
[0036] In a third embodiment of a method according to the invention, the sample radiation is emitted by a light source containing several individual light sources whose optical output power can be separately controlled within the bandpass spectral range, with the bandwidth of the individual light sources being narrow compared to the width of the bandpass spectral range. Adaptation of I0(λ) to the profile of the Planck spectrum can be achieved by changing the intensity of the individual light sources. Thus, a combination of several narrowband light sources with their optical output power separately controllable, different emission wavelengths within the filter width of the bandpass filter is achieved (e.g., using LEDs, lasers, or sLEDs of different wavelengths at a fixed operating temperature and individually controllable light intensity).
[0037] The individual intensities of these (compared to the width of the bandpass filter) narrowband light sources can then be adjusted for any desired temperature range to form the overall intensity distribution I0(λ) such that the profile of the respective Planck function P(λ,T) is best approximated within the filter bandwidth. A method according to the invention is therefore, in principle, also suitable for thick IR-transparent layer stacks with total layer thicknesses > 100 µm, since a spectrally individually controllable "white light" source (consisting of multiple light sources) can also be used in combination with very wide bandpass filters.
[0038] A further aspect of the invention relates to a device according to claim 6 for carrying out such a method according to the invention. The device has, in particular, all the features required for carrying out the method. The advantages resulting from the individual device features, as well as their conceptual interpretation, can be directly derived from the sections of this description of the invention relating to the respective method.
[0039] A device according to the invention comprises: • a light source for irradiating a sample radiation onto a surface layer of the wafer, wherein the spectral intensity distribution I0(λ) of the sample radiation is adjustable; • a spectral filter element with a bandpass characteristic for filtering the sample radiation I reflected from the wafer R(λ) as well as the thermal radiation P(λ,T) emitted by the wafer and transmitted by the surface layer, whereby for each radiation an effective effective wavelength λ A eff and λ B eff which is the bandpass integral of the radiation intensities I R (λ) and P(λ,T) are each decomposed into two segments of equal partial intensity; • a suitable detector for measuring the total intensity of the radiation behind the bandpass filter; • a control unit, configured to measure the spectral intensity distribution I0(λ) of the reflection sample radiation with its effective wavelength λ A eff to be precisely adjusted to the effective effective wavelength λ B effof the pyrometry measurement. This is done within the bandpass spectral range by appropriately adapting the spectral intensity distribution I0(λ) to the shape of the Planck curve. The determination of the effective effective wavelength λ B eff The thermal radiation is determined, for example, from its spectral intensity distribution P(λ,T) of the wafer or from the slope of the Planck calibration line Log 10 [P(λ,T)]=f(1 / T) (both preferably in the expected process temperature range or at a predetermined or otherwise determined temperature T1); • and an evaluation unit configured to determine the current temperature T2 of the wafer using ECP from the measured total intensity of the filtered radiation.
[0040] Preferably, the effective effective wavelengths λ A eff and λ B effwithin the control unit. Furthermore, the effective effective wavelengths can also be determined via the evaluation unit and transferred to the control unit.
[0041] The separation into a control unit and an evaluation unit should be interpreted in a logically abstract way. The control unit and the evaluation unit can also be designed as a single module; parts of the control unit can also be integrated into a module with the evaluation unit (or vice versa).
[0042] Preferably, a device according to the invention comprises an optical modulator which is designed to detect the sample radiation (reflection radiation I R (λ)) from thermal radiation using a modulation method.
[0043] Embodiments of a device according to the invention provide that the light source is an LED which, at a first operating temperature, has a maximum light intensity at the center wavelength λ BP of the spectral filter element. The spectral intensity distribution I0(λ) of this LED light source can be specifically adjusted by changing the operating temperature of the LED. The light source can preferably also be a thermal white light source, in which case the spectral intensity distribution I0(λ) can be adjusted by changing the radiation temperature of the white light source.
[0044] In a further embodiment, the light source comprises a plurality of individual light sources with different emission wavelengths within the bandpass spectral range, the optical output of which can be separately controlled. The bandwidth of the individual light sources is narrow compared to the width of the bandpass spectral range, the individual light sources form a common spectrum within the bandpass spectral range, and the spectral intensity distribution I0(λ) is adjusted by changing the intensity of the individual light sources. Depending on the number of individual light sources forming the light source, an individual emission spectrum that approximately corresponds to the bandwidth of the bandpass spectral range of the spectral filter element divided by the number of individual light sources forming the spectral range is preferably considered "narrow compared to the width of the bandpass spectral range."
[0045] Further preferred embodiments of the invention result from the features mentioned in the subclaims.
[0046] The various embodiments of the invention mentioned in this application can be advantageously combined with one another, unless otherwise stated in the individual case.
[0047] The invention is explained below in exemplary embodiments with reference to the accompanying drawings. They show: Fig. 1 Illustrations of an exemplary determination of the temperature of a wafer in the coating process using a) simple pyrometry and b) conventional ECP; Fig. 2 a) a detailed representation of residual oscillations in the temperature profile of conventional ECP and b) an analysis of the corresponding FPO period in the reflection and pyrometry raw signal; Fig. 3 a schematic representation of an embodiment of a device according to the invention; Fig. 4 a representation of a comparison of the Planck function P(λ,T) for two different temperatures T; Fig. 5 a representation of the superposition of the spectral response of a Planck function P(λ,T) and an LED with an ideal bandpass filter spectrum, and Fig. 6 a schematic representation of the inventive determination of the effective wavelengths λ A eff and λ B eff .
[0048] Fig. Figure 1 shows representations of an exemplary determination of the temperature of a wafer during the coating process using a) simple pyrometry and b) conventional ECP. This involves a temperature measurement during a coating with a III-nitride multilayer structure on silicon using metal-organic vapor phase epitaxy (MOCVD). The data for this so-called HEMT (high-electron mobility transistor) process were kindly provided by the Otto von Guericke University in Magdeburg (Prof. Armin Dadgar). In a), the reflection measurement of a 950 nm pyrometer (gray curve) is compared with the temperature output (black curve) of a non-ECP-corrected pyrometer (standard 950 nm pyrometer). It is clearly visible that the temperature measurement is distorted by additional oscillation (Fabry-Pérot artifacts).In b), the reflection measurement of the 950 nm pyrometer (gray curve) is compared with the temperature output (black curve) of an emissivity-corrected pyrometer (ECP 950 nm pyrometer according to DE 44 19 476 C2). This ECP-corrected measurement provides the correct wafer temperature and no longer shows any Fabry-Pérot oscillations detectable at this scale.
[0049] Fig. Figure 2a shows a detailed representation of residual oscillations in the temperature profile during conventional ECP. This represents the temperature signal (gray curve) of an ECP-corrected pyrometer (ECP 950 nm pyrometer according to DE 44 19 476 C2) at the end of a III-nitride epitaxy on silicon compared to the simultaneous reflection measurement of a 950 nm pyrometer (black curve). The temperature scale of the ordinate has been significantly expanded to emphasize the residual oscillations (approximately ±1 K amplitude). The amplitude of these residual oscillations noticeably increases with increasing growth time and total layer thickness. Fig. Figure 2b shows the Fabry-Pérot growth oscillations (FPO) during a GaN epitaxial growth process, both in the 950 nm reflection signal and in the 950 nm pyrometry raw signal. The different period lengths of the respective FPOs indicate the difference in the effective wavelengths of the two processes.
[0050] Fig. Figure 3 shows a schematic representation of an embodiment of an apparatus according to the invention for carrying out an inventive method for residual oscillation-free, emissivity-corrected in-situ pyrometry on a wafer 20 during a coating process. This arrangement completely solves, in particular, the problem of temperature residual oscillations for total optical thicknesses nd (average refractive index x total layer thickness) in layer stacks with total layer thicknesses < 20 µm. However, the apparatus can also be readily applied to larger layer thicknesses according to the invention.
[0051] The device includes: • a light source 10 for radiating a sample radiation A0 onto a surface layer 22 of a wafer 20, wherein the spectral intensity distribution I0(λ) of the sample radiation A0 is controllable; • a spectral filter element 30 with a bandpass characteristic for filtering the sample radiation A reflected by the wafer 20 and a thermal radiation B emanating from the wafer 20 and transmitted by the surface layer 22, wherein for the individual radiations A, B an effective effective wavelength λ A eff , λ B eff which decomposes the bandpass spectral integral for the individual filtered radiations (A', B') into two surface segments (A1, A2) or (B1, B2) with the same partial intensity (more details on the decomposition are shown in Fig. 6); • a detector 40 for measuring the total intensity of the filtered radiations A', B'; • a control unit 50, configured to use the spectral intensity distribution I0(λ) of the sample radiation A0 to set the effective effective wavelength λ A effof the reflected sample radiation A to the effective effective wavelength λ B eff of the thermal radiation B, whereby a determination of the effective effective wavelength λ A eff the intensity distribution I0(λ) of the reflection sample radiation A0 from its spectral profile and the bandpass characteristic and a determination of the effective effective wavelength λ B eff the thermal radiation B0 from its spectral intensity distribution according to the Planck function P(λ,T) at a suitable temperature T1 of the wafer 20 and the bandpass characteristic; • and an evaluation unit 60 configured to determine the current temperature T2 of the wafer 20 using ECP from the measured total intensity of the filtered radiation A', B'. The control unit 50 and the evaluation unit 60 can also be configured as a combined control and evaluation unit 50, 60.
[0052] An LED can be used as the light source 10, which at a first operating temperature has a maximum light intensity at the center wavelength λ BP of the spectral filter element 30. By changing the operating temperature of the LED, the spectral intensity distribution I0(λ) of the light source 10 can be adjusted. The light source 10 can also be a thermal white light source, whereby the spectral intensity distribution I0(λ) is adjusted by changing the radiation temperature of the white light source. Using a white light source, an advantageously adapted Planck profile P(λ,T) for the reflection sample radiation A0 can be set.
[0053] As shown in the illustration, the light source 10 can also comprise several individual light sources 10', 10'', 10''', each with its own optical output power, each with different emission wavelengths within the bandpass spectral range, wherein the bandwidth of the individual light sources 10', 10'', 10''' is narrow compared to the spectral width of the bandpass spectral range. The individual light sources 10', 10'', 10''' generate a common spectrum within the bandpass spectral range. The spectral intensity distribution I0(λ) can be adjusted by changing the intensity of the individual light sources 10', 10'', 10'''. Such a light source 10 has the advantage that the spectral intensity distribution I0(λ) of the light source 10 can be influenced immediately and directly. Therefore, spectral intensity distributions I0(λ) with a particularly advantageous curve can also be generated and used as sample radiation A0.
[0054] The illustrated device further comprises an optical modulator 12 configured to separate the filtered sample radiation A' from the filtered thermal radiation B' using a modulation method. The optical modulator 12 can preferably be a mechanical light chopper or an electro-optical or acousto-optical modulator. However, direct electronic modulation of the output power of the light source is particularly preferred. Direct modulation can significantly simplify the beam path within the device.
[0055] Fig. 4 compares the Planck function P(λ,T) for two different temperatures T. It shows a) the spectral response of the Planck function P(λ,T) for an ideal black body radiator at 1000 °C and 1300 °C in a wide spectral range window between 750 nm - 3000 nm and b) the spectral response of the two Planck functions P(λ,T) scaled to the same relative scale in the bandpass spectral range of a spectral filter element in an exemplary selected range between 910 nm - 990 nm (filter width: 80 nm, λ BP = 950 nm).
[0056] The temperatures of 1000 °C and 1300 °C are typical process temperatures, for example, for III-nitride epitaxy. The spectral segment typically used for pyrometric temperature measurement on silicon wafers (as substrates for III-nitrides) is marked by the box in a). A typical center wavelength λ BPfor a spectral filter element used for ECP at 950 nm.
[0057] The comparison shows that although the absolute slope of the Planck functions P(λ,T) in the spectral range of the bandpass filter increases with temperature T, the relative slope - which is responsible for the shift Δλ of the pyrometric effective wavelength λ B eff responsible for - decreases with higher temperature T. The difference Δλ between a given effective wavelength λ A eff the reflected sample radiation A (reflection measurement) and the effective effective wavelength λ B eff the thermal radiation B depends on the course of the Planck function P(λ,T) in the spectral range of the bandpass filter and thus varies with the temperature T of the wafer.
[0058] The difference between the spectral curves of the Planck function for different temperatures on the one hand and the spectral curve of the reflected light source on the other hand makes it understandable why, despite using the same bandpass filter as spectral filter element, the effective wavelengths λ A eff , λ B eff of reflection measurement (A) and pyrometry (B) are usually slightly different. The thermal radiation B' transmitted by the bandpass filter is significantly higher at 955 nm than at 945 nm. Thus, the effective effective wavelength λ B eff a pyrometry measurement no longer at the center wavelength λ BP of the spectral filter element of 950 nm, but is shifted by a Δλ to longer wavelengths λ. The exact value of this Δλ can be determined mathematically by simple means of integral calculus (see Fig. 6).
[0059] Fig. Figure 5 shows the superposition of the spectral response of a Planck function P(λ,T) and an LED (at two LED operating temperatures) with a typical bandpass spectral range of an ideal spectral filter element. For this purpose, the Fig. 4b shows the spectral section for the spectral response of the Planck function P(λ,T) for an ideal black body radiator at 1000 °C with a typical spectral response of the bandpass spectral range of an ideal bandpass filter (filter width: 10 nm, λ BP = 950 nm). Also shown is a resulting shift of the effective effective wavelength λ B eff of the thermal radiation B by Δλ above the center wavelength λ BP of the spectral filter element.
[0060] Furthermore, the influence of an adjustment of the spectral intensity distribution I0(λ) of the reflection sample radiation A0 carried out according to the invention for setting the effective effective wavelength λ A eff of the reflected sample radiation A to the effective effective wavelength λ B eff of the transmitted thermal radiation B. In particular, the light source can be an LED which, at a first operating temperature, has a maximum light intensity at the center wavelength λ BP of the spectral filter element and whose spectral intensity distribution I0(λ) is specifically modified by changing the operating temperature. It is therefore a wavelength λ A eff adjustable reflection light source, which is adapted to a shift of the effective effective wavelength λ B effthe thermal radiation B can be adjusted.
[0061] In this example, a 950 nm LED light source was first selected to provide a maximum light intensity at the bandpass filter center wavelength λ at a suitable first operating temperature (reference operating temperature). BP emitted. This selection would already lead to almost residual oscillation-free measurements of the wafer temperature at very high wafer temperatures (e.g., T > 1300 °C). For any other (lower) temperature range, however, the effective wavelength of the reflection measurement must be shifted by a suitable value Δλ to ensure complete oscillation-free measurement. "Suitable" means exactly corresponding to the shift of the effective wavelength λ B effaccording to pyrometry. The LED emission wavelength can be adjusted by carefully and precisely adjusting the reference operating temperature of the 950 nm LED using a suitable heating or cooling module within the LED light source to shift the peak wavelength of the LED from 950 nm (at room temperature, 20 °C) to suitable wavelengths. Typical values are +0.3 nm wavelength shift of the peak wavelength per 1 K increase in the operating temperature of an LED.
[0062] The developed method for controlling the LED operating temperature can also be applied to stabilize any reflectance measurement (e.g., with respect to the ambient temperature), regardless of whether these LED-based reflectance measurements are used for emissivity correction or for general process control.
[0063] Fig. 6 shows a schematic representation for the inventive determination of the effective wavelengths λ A eff and λ B eff The figure largely reflects the Fig. 5 already presented relationships, but the definition of the individual effective effective wavelengths λ A eff , λ B effThe individual radiations A and B are discussed. In addition to an exemplary spectral profile of the Planck function P(λ,T) at a specific temperature (curve "Pyro"), the emission profile of an LED (curve "LED") as an example of a light source and a typical profile of the bandpass spectral range (curve "Filter") of a spectral filter element are also shown. The resulting spectral profile of the individual filtered radiations, i.e., the filtered reflected radiation A' (broken curve) and the filtered thermal radiation B' (black curve), is also shown in the figure.
[0064] For the individual radiations A, B, an effective effective wavelength is determined at the wavelength λ A eff , λ B effwhich divides the bandpass filter transmission integral of the reflection sample intensity A0 into two equally sized area segments A1, A2 and the Planck function "Pyro" into two equally sized area segments B1 and B2. The individual area segments can be analyzed analytically and / or numerically. The respective partial area segment follows as an integral of the respective spectral intensity curve within the bandpass spectral range limits, i.e. from the area below the respective curve sections in the range of the bandpass spectral range. In particular, the integral of the intensity from the lower limiting wavelength of the bandpass spectral range to the effective effective wavelength λ eff equal to the integral of the intensity of this radiation from the effective effective wavelength λ effup to the upper limiting wavelength of the bandpass spectral range. Even with non-ideal bandpass transmission curves, the effective effective wavelength λ eff the radiation can be determined. List of reference symbols 10 Light source (reflection measurement) 12 optical modulator 14 beam splitters 20 wafers (semiconductor substrate with layer system) 22 Surface layer / layer system 30 spectral filter element 40 detector 50 control unit 60 evaluation unit A0 test radiation (reflection measurement) I0(λ) spectral intensity curve (of the sample radiation A0) A reflected sample radiation (reflection beam) A' spectrally filtered reflected sample radiation B0 Planck radiation (from wafer 20 before transmission to surface layer 22) B thermal radiation (pyrometry measurement) B' spectrally filtered thermal radiation A1 LED first surface segment of the bandpass intensity integral < λ A eff A2 LED second surface segment of the bandpass intensity integral > λ A eff B1 Pyro first surface segment of the bandpass intensity integral < λ B eff B2 Pyro second surface segment of the bandpass intensity integral > λ B eff
Claims
[1] Method for residual oscillation-free emissivity-corrected in-situ pyrometry on a wafer (20) during a coating, comprising the following steps: - irradiating a sample radiation (A0) onto a surface layer (22) of the wafer (20), wherein the spectral intensity distribution I0(λ) of the sample radiation (A0) is adjustable; - filtering both the sample radiation (A) reflected by the wafer (20) and the thermal radiation (B) emanating from the wafer (20) and transmitted by the surface layer (22) by a spectral filter element (30) with bandpass transmission characteristics, wherein for each of the individual radiations (A, B) an effective effective wavelength λ A eff , λ B eff which decomposes the bandpass spectral integral for the individual filtered radiations (A', B') into two surface segments (A1, A2) or (B1, B2) each with the same partial intensity; - Determination of the effective effective wavelength λ B eff the thermal radiation from the spectral intensity distribution of the Planck function at a process-typical temperature T1 of the wafer (20) and the spectral bandpass characteristic; - targeted modification of the spectral intensity distribution I0(λ) of the sample radiation (A0) for precise adjustment of the effective effective wavelength λ A eff of the reflected sample radiation (A) to the effective effective wavelength λ B eff of the thermal radiation (B), whereby a determination of the effective effective wavelength λ A eff the reflected sample radiation (A) is determined from the spectral intensity distribution of the reflected sample radiation (A) and the spectral bandpass characteristic; - Measurement of the total intensities of the filtered radiations (A', B') at the same effective wavelength λ A eff , λB eff , and - Determination of the current temperature T2 of the wafer (20) by means of emissivity-corrected in-situ pyrometry, ECP, from the measured total intensities of the filtered radiations (A', B'). [2] The method of claim 1, wherein the bandwidth of the spectral filter element (30) is 5 nm - 30 nm. [3] Method according to claim 1 or 2, wherein, in order to distinguish the filtered radiations (A', B'), the sample radiation (A0) is made distinguishable from the thermal radiation (B) by a modulation method. [4] Method according to one of claims 1 to 3, wherein the sample radiation (A0) is emitted by an LED as light source (10), wherein at a first operating temperature a maximum spectral intensity at the center wavelength λ BPof the spectral filter element (30) occurs and a targeted change in the spectral intensity distribution I0(λ) occurs by adjusting the working temperature of the LED. [5] Method according to one of claims 1 to 3, wherein the sample radiation (A0) is emitted by a thermal white light source as the light source (10) and an adjustment of the spectral intensity distribution I0(λ) is carried out by changing the radiation temperature of the white light source. [6] Method according to one of claims 1 to 3, wherein the sample radiation (A0) is emitted by a light source (10), wherein the light source comprises a plurality of individual light sources (10', 10'', 10''') of different emission wavelengths within the bandpass spectral range, the optical output power of which can be separately controlled, wherein the bandwidth of the individual light sources (10', 10'', 10''') is narrow compared to the width of the bandpass spectral range, the individual light sources (10', 10'', 10''') form a spectrum within the bandpass spectral range and an adjustment of the spectral intensity distribution I0(λ) is carried out by changing the intensity of the individual light sources (10', 10'', 10'''). [7] Apparatus for carrying out a method according to any one of the preceding claims, comprising: - a light source (10) for radiating a sample radiation (A0) onto a surface layer (22) of a wafer (20), wherein the spectral intensity distribution I0(λ) of the sample radiation (A0) is adjustable; - a spectral filter element (30) with a spectral bandpass characteristic for filtering the sample radiation (A) reflected by the wafer (20) and a thermal radiation (B) emanating from the wafer (20) and transmitted by the surface layer (22), wherein for each of the individual radiations (A, B) an effective effective wavelength λ A eff , λ B eff which decomposes the bandpass spectral integral for the individual filtered radiations (A', B') into two surface segments (A1, A2) or (B1, B2) each with the same partial intensity; - a detector (40) for measuring the respective total intensities of the filtered radiations (A', B'); - a control unit (50) configured to adjust the spectral intensity distribution I0(λ) of the sample radiation (A0) to set the effective effective wavelength λ A eff of the reflected sample radiation (A) to the effective effective wavelength λ B eff of the thermal radiation (B), whereby a determination of the effective effective wavelength λ A eff of the reflected sample radiation (A) from the spectral intensity distribution I0(λ) of the reflected sample radiation (A) and the spectral bandpass characteristic and a determination of the effective effective wavelength λ B eff the thermal radiation (B) is determined from the spectral intensity distribution of the Planck function at a process-typical temperature T1 of the wafer (20) and the spectral bandpass characteristic; and - an evaluation unit (60) configured to determine the current temperature T2 of the wafer (20) by means of ECP from the measured total intensities of the filtered radiations (A', B'). [8] Apparatus according to claim 7, further comprising an optical modulator (12) arranged to make the sample radiation (A0) distinguishable from the thermal radiation (B) by a modulation method. [9] Device according to claim 7 or 8, wherein the light source (10) is an LED which, at a first operating temperature, has a maximum light intensity at the center wavelength λ BP of the spectral filter element (30) and whose spectral intensity distribution I0(λ) can be specifically adjusted by changing the operating temperature of the LED; or the light source (10) is a thermal white light source and the spectral intensity distribution I0(λ) is adjusted by changing the radiation temperature of the white light source. [10] Device according to one of claims 7 to 9, wherein the light source (10) comprises a plurality of individual light sources (10', 10'', 10''') of different emission wavelengths within the bandpass spectral range, the optical output power of which can be separately controlled, wherein the bandwidth of the individual light sources (10', 10'', 10''') is narrow compared to the width of the bandpass spectral range, the individual light sources (10', 10'', 10''') form a spectrum within the bandpass spectral range, and an adjustment of the spectral intensity distribution I0(λ) is carried out by changing the intensity of the individual light sources (10', 10'', 10''').
Citation Information
Patent Citations
Method for thermographic examination of a workpiece and device therefor
DE19832833C2
Method and device for determining the layer thickness and the substrate temperature during coating
DE4419476C2
pyrometer with emission meter
DE69312894T2
Active pyrometry with emissivity extrapolation and compensation
US6299346B1
Temperature determination using pyrometry
US6398406B1
