It is the time of the year that the annual AVS International Conference on Atomic Layer Deposition combined with the International Atomic Layer Etching Workshop takes place. The location of 18^th edition of the ALD conference (ALD 2018) and the 5^th edition of the ALE workshop (ALE 2018) is Incheon, South Korea (July 29 – August 1, 2018). The primary reason for this blog post is the tutorial session that traditionally kicks off the event on Sunday. During this session, a tutorial on “In situ Studies of ALD Processes & Reaction Mechanisms” is on the program. And guess what? We are responsible for this tutorial. I write “we” as this tutorial is put together by myself and by Erwin Kessels, who will be the actual presenter of the tutorial.

During the preparation of the tutorial slides (which can be downloaded here), we noted that it was not possible to adequately cover the topic extensively in a time span of only 45 minutes. Some tough decisions had to be made regarding the content. The aim was to give an overview of methods that can be used for in situ studies ALD processes and reaction mechanisms while also providing some insight into these processes and mechanisms. However given the time available, a comprehensive overview of the methods was not feasible and neither the techniques could be explained in large detail. Therefore, we decided to cover only a few methods in the first part of the presentation and focus on what can be learned from these techniques, briefly describe their pros and cons and also give some practical comments. The methods covered were chosen on the basis of the fact that they are relatively widely employed, are fairly basic and can give important insight into the ALD processes and reaction mechanisms. Moreover, the choice was also based on the fact that we have experience with these methods ourselves as it is always easier to talk about methods that you have been working with yourself. It does not make much sense to just repeat what others have written in papers when not being able to add some personal experience. Nonetheless, even with these limitations, there has been a lot of information in the tutorial and therefore we decided to write some things down in this blog such that it can read back. Below, we cover some aspects about six methods in more detail:

• Quartz crystal microbalance
• Spectroscopic ellipsometry
• Quadrupole mass spectrometry
• Gas phase infrared spectroscopy
• Surface infrared spectroscopy
• Optical emission spectroscopy

In addition, there are some other aspects and thoughts we want to share. We’ll do this next first.

When talking about in situ studies of ALD processes, when can distinguish several reasons to do so. First of all, it is helpful to obtain information about the film properties already during film growth. With this information, one can make a quick link between the process settings of the ALD reactor and the film properties resulting. Secondly, it can also be very useful to monitor the film growth directly. It can be very helpful to see how the film thickness increases as a function of the number of ALD cycles and also to establish how the film properties change with thickness. Furthermore, it is desirable to verify ALD growth in a simple and fast way, for example by testing the self-limiting nature of the surface reactions without the need to deposit separate films for every condition to be tested. Saturation curves can be quickly established when the thickness evolution can be measured over a number of cycles for one condition before switching to another condition in which the thickness evolution is again measured over another set of cycles, etc. Thirdly, it is also important to study reaction mechanisms such that a better understanding of ALD film growth can be obtained. For example, reaction products can be measured to learn about the surface reactions taken while investigation of the surface groups present after every half-cycle yields insight in the surface chemistry as well.

Regarding the methods and techniques that can be used for in situ studies of ALD processes, it is possible to define a few metrics that can be used to classify them. For example, one can consider the “adequacy”, “feasibility”, “complexity” and “costs” of the methods and techniques. A method can for example be well-suited to study a process but require a dedicated ALD reactor, be pretty advanced and also be expensive. Another method can yield only limited information, be straightforwardly applicable in virtually all types of ALD reactors and be easy to use at low cost. These are just hypothetical examples but clearly such aspects are important when considering the use of certain methods. Furthermore, the ALD reactor available might set boundary conditions as well. Industrial ALD reactors provide evidently much less possibilities than advanced research-type ALD reactors.

Another topic addressed in the tutorial (in the second part) is related to the key merits of ALD: precise growth control, excellent uniformity and unparalleled conformality. These merits can require some attention too as it is important to know whether they are fulfilled whereas they can also provide insight in the reaction mechanisms as well. For example, the conformality can be verified by depositing films in high-aspect ratio trenches and subsequent SEM or TEM inspection. However, the conformality can also be investigated by special test structures which can also provide information on the sticking probability of the precursor and co-reactant. A nice example of such test structures are those developed by Finish institute VTT (see Ylilammi et al., J. Appl. Phys. 123, 205301 (2018) and Gao et al., J. Vac. Sci. Technol. A 33, 010601 (2015)) although it has to be noted that similar, less sophisticated experiments were also reported by others (Gordon et al., Chem. Vap. Deposition 9, 73 (2003), Dendooven et al., J. Electrochem. Soc. 157, G111, (2010)). Insight into the sticking probabilities of the precursor and co-reactant is important for establishing whether the ALD growth is in the reaction-limited or diffusion-limited case for 3D structures with a certain aspect ratio (Elam et al., Chem. Mater. 15, 3507 (2003).) Moreover, quantitative information on the sticking probabilities is important as input for simulations related to ALD film growth.

Investigating the uniformity can also provide insight into surface reactions taking place. A nice example is the work by Knoops et al. when working in the group of Jeff Elam at the Argonne National Lab (Knoops et al., Chem. Mater. 23, 2381 (2011).). For the case of ALD of ZnO by Zn(C₂H₅)₂ and O₃, it was found that the non-uniformity of the ZnO films deteriorated when reaching saturation for the O₃ step. This is somewhat counterintuitive but it can be explained by surface recombination of O₃ which depends on the nature of the surface: surface recombination of O₃ turned out to be higher on a pristine ZnO surface than on a ZnO surface covered by C₂H₅ groups.

ALD is known for its precise growth control. In principle this does not require any in situ monitoring or detailed evaluation once the growth-per-cycle is known for the particular ALD conditions employed: the growth-per-cycle should be extremely reproducible. However, the situation is different in the initial growth regime when a thin film material is deposited on a “foreign” substrate. This brings up another topic, the topic of initial growth. This topic requires more attention, certainly considering the current trend towards more ultrathin films (a nucleation delay has much more impact when trying to deposit a 1 nm thick film than when trying to deposit a 30 nm thick film) and the growing interest in area-selective ALD (in which a difference in nucleation delay on the growth and non-growth surface) is often exploited. To better understand initial growth dedicated in situ studies will be required and the implementation of advanced in situ methods might be involved.

The ALD community would benefit from analytical methods that would be new to the field. Two examples addressed in the tutorial are broadband sum-frequency generation (Vandalon et al., Appl. Phys. Lett. 108, 011607 (2016), Vandalon and Kessels, J. Vac. Sci. Technol. A 35, 05C313 (2017) ) and adsorption calorimetry (Lownsbury et al., Chem. Mater. 29, 8566 (2017)). The latter method is highly relevant as it yields insight into the reaction heats associated with the half-cycles. This kind of information has not been available until now while it would provide additional thermodynamic and mechanistic insight into the ALD reactions. It would also allow for a direct comparison with results obtained by first-principle calculations as so far calculated reaction heats have remained untested with respect to experiment. Hopefully the field of ALD will also develop in that direction in the next years!

 

---

Quartz crystal microbalance

Schematic representation of an ALD reactor equipped with two quartz crystal microbalances (QCM). One is located at the position of the substrate stage while the other one is located at the position of the reactor wall.

In a nutshell:

A quartz crystal microbalance (QCM) is a device that can be used to measure very small changes in mass when depositing or etching a material. This is done by probing the resonance frequency of a quartz crystal resonator covered with the material to be studied. The method is very well suited to monitor film growth as a function of the number of ALD cycles and can be used to obtain information about the “growth-per-cycle”. It is also sufficiently sensitive to measure the mass gain or loss during the half-cycles and can therefore be used to study reaction mechanisms.

QCMs specifically designed for ALD are commercially available.

Photo of two possible configurations of a QCM. The gold colored disk in the middle is the actual quartz crystal. The photo is taken from the Inficon website, www.inficon.com.

Some pros and cons:

+	QCM is a rather inexpensive device that can relatively easily be implemented in a wide variety of reactors.
+	The mass gain or loss is directly measured in a quantitative manner.
+	The simplicity of the technique enables fast process development.
–	The quartz resonator is very sensitive to variations in pressure, gas flows, and temperature. It is important to have well-controlled reactor conditions and a QCM can be best used in a hot-wall reactor. Using a QCM in a PE-ALD reactor is also not straightforward.

Further reading:

Viscous flow reactor with quartz crystal microbalance for thin film growth by atomic layer deposition, J.W. Elam, M.D. Groner and S.M. George, Rev. Sci. Instrum. 73, 2981 (2002).

This is an early paper on the use of QCMs in ALD reactors. It has contributed to QCMs specifically designed for ALD.

Temperature-induced apparent mass changes observed during quartz crystal microbalance measurements of atomic layer deposition, M.N. Rocklein and S.M. George, Anal. Chem. 75, 4975 (2003).

This paper describes how temperature fluctuations in the ALD reactor can lead to apparent mass changes and how it can cause misinterpretation of QCM data.

Typical example of a graph obtained by QCM measurements during ALD of Al₂O₃. The graph shows the mass deposited (right axis) which can be converted in film thickness when knowing the mass density of the film prepared (the 3.5 g/cm² is the bulk mass density of Al₂O₃ and therefore an overestimation of the mass density of the ALD film). The data in the circle is a zoom-in to clearly show the staircase behaviour of the data.

Figure adapted from: Viscous flow reactor with quartz crystal microbalance for thin film growth by atomic layer deposition, J.W. Elam, M.D. Groner and S.M. George, Rev. Sci. Instrum. 73, 2981 (2002).

---

Spectroscopic ellipsometry

Schematic representation of an ALD reactor equipped with a spectroscopic ellipsometer. On the left side light from a lamp with a broad spectrum is coupled into the reactor after passing through optical elements such as a polarizer. After reflection on the sample, the light is probed by a detector after passing through another set of optical elements (including a polarizer). The light is spectrally analyzed by passing it through a spectrograph before reaching the detector. The preferred angle of incidence of the light on the sample is 70 degrees with respect to normal incidence. Protective valves need to be employed to avoid deposition of material on the windows of the ellipsometry viewports.

In a nutshell:

Ellipsometry measurements are based on the change in polarization of a light beam upon reflection from a sample. For a spectroscopic ellipsometer, this change in polarization is measured for a range of wavelengths of light. When comparing the measurement to an optical model of the sample (substrate with one or more layers of material), it is possible to extract optical properties of the sample as well as thickness information for the layer(s). When measuring the thickness as a function of number of ALD cycles, the film growth can be monitored and “growth-per-cycle” data can be extracted. From the optical properties, information about the refractive index, extinction coefficient, band gap, etc. can be extracted. For metallic films, it is even possible to extract electrical properties of the films by considering the free carrier absorption in the infrared part of the spectrum.

Spectroscopic ellipsometers are commercially available from a range of suppliers.

Picture of an Oxford Instruments FlexAL reactor with a spectroscopic ellipsometer attached.

 

Some pros and cons:

+	Directly measures thickness, very helpful for (fast) process development as e.g. saturation curves can be made without preparing separate samples with varying thickness or under various dosing conditions
+	Spectroscopic ellipsometry also yields insight into many other properties than only the thickness, including optical and electrical properties.
–	Optical modeling is required to extract the quantitative information, which can be challenging for some layers/materials.
–	SE is rather expensive and requires special ports on the reactor of optical access.

 

Further reading:

In situ spectroscopic ellipsometry as a versatile tool for studying atomic layer deposition,  E. Langereis, S.B.S. Heil, H.C.M. Knoops, W. Keuning, M.C.M. van de Sanden and W.M.M. Kessels, J. Phys. D: Appl. Phys. 42, 073001 (2009).

A review paper that describes the information that can be extracted about the ALD process and the material prepared from in situ spectroscopic ellipsometry measurements. The paper only addresses oxides and nitrides.

In situ spectroscopic ellipsometry during atomic layer deposition of Pt, Ru and Pd, N. Leick, J.W. Weber, A.J.M. Mackus, M.J. Weber, M.C.M. van de Sanden and W.M.M. Kessels Phys. D: Appl. Phys. 49, 115504 (2016).

A paper describing which information can be extracted about ALD of metals (Pt, Ru, Pd) using in situ spectroscopic ellipsometry. This paper is complementary to the review paper by Langereis et al.

Example of thickness monitoring as a function of number of cycles, here for the case of ALD of Pt. While the thickness increases linearly with the number of cycles eventually, in the first 150 cycles the thickness remains zero and does not increase at all. This is the so-call nucleation delay which often takes place when depositing a material on a “foreign” substrate material. The “growth-per-cycle” can be deduced from the slope of the plot at higher number of cycles. Note that the “growth-per-cycle” value would be erroneous if calculated from the final film thickness (14 nm) and the total number of cycles (550). This demonstrates the importance of monitoring the film thickness during ALD. The nucleation delay reveals that Pt film growth by ALD is inhibited on the Al₂O₃ starting surface. As shown by the top view TEM images included in the figure, the Pt starts growing on defects sites on the Al₂O₃ leading to island formation (which can be exploited for the deposition of nanoparticles). Subsequently the islands grow after which coalescence takes place leading to film closure. Although the ellipsometry data shows a zero thickness in the first 150 cycles, this reveals that ellipsometry cannot capture what exactly happens in the nucleation phase. This is especially due to the assumptions made in the optical model used to extract the information.

Figure adapted from: Influence of Oxygen Exposure on the Nucleation of Platinum Atomic Layer Deposition: Consequences for Film Growth, Nanopatterning, and Nanoparticle Synthesis, A.J.M. Mackus, M.A. Verheijen, N. Leick, A.A. Bol and W.M.M. Kessels, Chem. Mater. 25, 1905 (2013).

---

Quadrupole mass spectrometry

Schematic representation of an ALD reactor to which a quadrupole mass spectrometer has been fitted. The mass spectrometer needs to operate at pressures below 10^-5 Torr and is therefore differentially pumped. Gas is sampled from the ALD reactor (with pressures up a few Torr) through a small pinhole. The mass spectrometer itself consists of an ionizer, a quadrupole mass filter and a detector. Note that it is also possible to mount the mass spectrometer in the exhaust line of the reactor to probe the gas composition (with reaction products) downstream.

In a nutshell:

A quadrupole mass spectrometer (QMS) is an instrument that can be used to measure the composition of a gas in terms of atoms and molecules present. To do so, a fraction of the gas is ionized by energetic electrons (electron energy is typically 70 eV) generated by a hot filament. Subsequently, the ions generated are focused into the quadrupole mass filter which filters the ions on the basis of their mass-to-charge ratio. The ions that pass through the mass filter are finally measured by a detector. The mass spectrometer can be used to measure a mass spectrum by scanning the mass filter or it can be used to measure certain mass-over-charge ratios over time. To measure species in the ALD reactor (e.g., precursor and co-reactant molecules as well as the gaseous reaction products created), gas from the ALD reactor is sampled through a small pinhole.

Mass spectrometers (often called “residual gas analyzers”) are commercially available from a range of suppliers.

Two reactors with a quadrupole mass spectrometer attached. In the left picture the differential pumping is clearly visible.

 

Some pros and cons:

+	QMS is relatively easy to implement on ALD reactors. At higher reactor pressures it has to be verified that the gas sampled by the QMS is representative of the gas composition above the substrate region.
+	The mass-over-charge ratio can be quickly scanned which means that different types of molecules within the mass range of the QMS (typically 1-200 amu) can be easily measured. In practice, however, the sensitivity for heavier ions (> 50 amu) is relatively low.
+ –	The QMS will probe all gas phase reaction products created in the ALD reactor, not only those created at the substrate surface but also those created at the reactor walls. This can be an advantage in terms of a higher signal but it can also be a disadvantage as the reaction products created at the reactor wall are different from those created at the substrate surface (as could be the case, for example, for a warm wall reactor as opposed to a hot wall reactor).
–	In the ionizer the molecules are cracked into fragment ions which can make it difficult to determine which parent molecule is being probed. Typically parent molecules have a clear “fingerprint” when being ionized but in case there are a multitude of possible parent molecules (e.g., precursor with various reaction products), the interpretation of the data can become rather difficult.

 

Further reading:

Reaction mechanisms of atomic layer deposition of TaN_x from Ta(NMe₂)₅ precursor and H₂ -based plasmas, H.C.M. Knoops, E. Langereis, M.C.M. van de Sanden and W.M.M. Kessels, J. Vac. Sci. Technol. A 30, 01A101 (2012).

This paper describes a procedure that can be used to probe reaction products generated during an ALD process despite a limited signal-to-noise ratio. It also reports on some important practical aspects when monitoring certain mass-over-charge ratios over a period of time.

A typical example of how reaction products created during an ALD process can be probed and monitored over time by a quadrupole mass spectrometer. This particular case shows data for the thermal ALD process of Al₂O₃ from trimethylaluminum (TMA, Al(CH₃)₃) and water (H₂O). The reaction product during both half-cycles is CH₄ as is probed here. A signal due to CH₄ is measured at a mass-over-charge ratio 16 (corresponding to the CH₄⁺ ion) every time when Al(CH₃)₃ or H₂O is dosed in the reactor. The signal due to H₂O (measured at a mass-over-charge ratio 18 corresponding to the H₂O⁺ ion) is also shown. Probing the Al(CH₃)₃ was already more complicated. The main ions created by ionizing Al(CH₃)₃ have a quite high mass-over-charge ratio for which the QMS used was not very sensitive. The best signal was obtained at a mass-over-charge ratio of 27 (corresponding to the Al⁺ ion).

Figure adapted from: Revisiting the growth mechanism of atomic layer deposition of Al₂O₃ : A vibrational sum-frequency generation study , V. Vandalon and W.M.M. Kessels, , J. Vac. Sci. Technol., A 35, 05C313 (2017).

---

Gas-phase infrared spectroscopy

Schematic representation of an ALD reactor through which a beam of infrared light is passed in order to measure gas phase species by infrared spectroscopy. Infrared transparent windows are used to transmit the infrared beam and these windows are protected against film deposition by valves. On one side of the reactor the interferometer is located which includes an infrared light source. The interferometer is typically a Fourier transform infrared (FTIR) interferometer which explains why this method is often referred to as “FTIR”. On the other side of the reactor an external infrared detector is located. Note that it is also possible to measure the gas composition (with reaction products) in the exhaust line of the ALD reactor, preferably using a multipass cell for enhancing the signal-over-noise ratio. Infrared measurements are not restricted to certain pressures and can also be used at 1 atmosphere (which can be of interest for spatial ALD applications).

In a nutshell:

Infrared spectroscopy or Fourier transform infrared spectroscopy (FTIR) is a technique that measures the absorption of infrared light by vibrational transitions of molecular species. When these species are present in the gas-phase, one refers to gas-phase infrared spectroscopy. Typically a set of two sequential measurements are performed, one for an evacuated reactor without absorbing species (baseline spectrum) and one with the gas phase species present. When dividing both spectra, the transmittance can be calculated which can be converted in “absorbance”. One often refers to “differential spectrum” because the absorbance spectrum shows the difference in species present when comparing the two spectra. Identification of the species present takes place by comparing the vibrational absorption features with literature spectra often reported in databases.

Interferometers for FTIR spectroscopy are commercially available as well as sensitive infrared detectors. The beam path to and from the ALD reactor needs to be purged with dry air or N₂ such that a purge box needs to be build.

Home-built ALD reactor with a FTIR system. On the left one can see the interferometer. On the right, the N₂ purge box is visible with a liquid-N₂ cooled infrared detector inside.

 

Some pros and cons:

+	Gas-phase FTIR can easily yield absolute densities of reaction products after calibration experiments with individual gases.
+ –	Due to characteristic rovibrational fingerprints, some gases/reaction products can easily be identified and measured. However, gas phase infrared spectroscopy can only be used to probe species with a permanent electric dipole moment (e.g., gas such as O2, N2 and H2 cannot be measured)
+ –	Similar to QMS, gas phase FTIR will probe gas phase reaction products created in the ALD reactor independent from the surface they are formed. Again, this can be an advantage in terms of a higher signal but it can also be a disadvantage as the reaction products created at the reactor wall are different from those created at the substrate surface.
–	For sufficient signal-to-noise ratios, typically data should be averaged over several scans of the interferometer. It can therefore be necessary to confine the reaction products in the reactor by closing off the reactor from the pump by a valve.

 

 

Further reading:

Time-resolved Fourier transform infrared spectroscopy of the gas phase during atomic layer deposition, B.A. Sperling, W.A. Kimes, J.E. Maslar, and P.M. Chu, J. Vac. Sci. Technol. A 28, 613 (2010).

Paper that describes the use of FTIR to measure gas phase species during ALD. The primary focus of the paper is the time-resolved measurement in order to capture that gas-phase dynamics.

Surface reactions during atomic layer deposition of Pt derived from gas phase infrared spectroscopy, W.M.M. Kessels, H.C.M. Knoops, S.A.F. Dielissen, A.J.M. Mackus and M.C.M. van de Sanden, Appl. Phys. Lett. 95, 2 (2009).

Paper that describes how gas phase FTIR spectroscopy can be used to determine information about the reaction products including their absolute densities. The calibration procedure is briefly described.

 

Gas-phase infrared absorption spectra for the Al(CH₃)₃ and H₂O half-cycles during ALD of Al₂O₃. The positive peaks represent species present in the reactor. Both the Al(CH₃)₃ and H₂O is dosed in the reactor in three sequential micropulses and the spectra show the species in the reactor per micropulse. For the first Al(CH₃)₃ dose, absorption peaks due to both Al(CH₃)₃ and CH₄ can be distinguished. This indicates that CH₄ is created as a reaction product during the first Al(CH₃)₃ micropulse. During the second and third Al(CH₃)₃ dose, no signal due to CH₄ can be distinguished revealing that the surface reactions have already saturated. This is different when dosing the H₂O. The first H₂O dose shows the highest CH₄ signal revealing that most reaction products are produced during this step. However, in the subsequent H₂O doses a CH₄ signal can still be discerned despite the fact that sufficient H₂O is dosed already in the first microdose as a large signal of H₂O is present (i.e., the H₂O is not reacted fully away). This suggests that the reaction of H₂O is with –CH₃ surface groups is slower than the reaction of Al(CH₃)₃ with –OH surface groups. 

Figure adapted from: Revisiting the growth mechanism of atomic layer deposition of Al₂O₃ : A vibrational sum-frequency generation study , V. Vandalon and W.M.M. Kessels, , J. Vac. Sci. Technol., A 35, 05C313 (2017).

---

Surface infrared spectroscopy

Schematic representation of an ALD reactor suited for surface infrared spectroscopy. In this case, many aspects are the same as for gas phase infrared spectroscopy apart from the fact that the beam of infrared light is passed through the substrate which should be infrared transparent. Such substrate can be low-resistivity crystalline silicon. The substrate can for example be mounted on a manipulator and should preferably allow for substrate heating (crystalline silicon can be heated resistively by passing a current through it) and temperature control when using a warm wall ALD reactor. In this configuration shown, the same setup also allows for gas phase infrared spectroscopy when taking appropriate differential spectra.

 

In a nutshell:

Surface infrared spectroscopy is similar to gas phase infrared spectroscopy apart from the fact that now vibrational transitions of surface groups are measured instead of rovibrational transitions of gas phase molecules. Typically a spectrum is taken before and after carrying out a half-cycle such that the change in surface groups due to the half-cycle can be probed. When expressed in terms of absorbance, a positive absorbance peak with respect to the baseline indicates species generated at the surface while a negative absorbance peak indicates removal of the surface species. Again, identification of the species present takes place by comparing the vibrational absorption features with literature spectra often reported in databases. One should note that the wavenumber position of the vibrational modes by a surface species is often affected by the backbonds of the surface species.

Surface infrared spectroscopy uses the same type of interferometers as gas phase infrared spectroscopy. A highly sensitive, liquid-N₂ cooled infrared detector is however key for surface infrared spectroscopy as surface groups with a sub-monolayer coverage need to be measured. This typically also means that long integration times per spectrum are required. Integration times up to 10 minutes are no exception. In order to improve the signal-to-noise ratio the absorbance can be increased by depositing the ALD films on particles pressed into a grid (increased surface area) or by using an attenuated total reflection (ATR) crystal. In the latter case the infrared beam probes the film multiple times due to multiple reflections from the surface.

Configurations of surface infrared spectroscopy employing a high surface area created by pressed particles (top) and multiple surface reflections by an ATR crystal (bottom).

 

Some pros and cons:

+	The key merit of surface infrared spectroscopy is that enables a direct measurement of surface groups created at, removed from or incorporated into the film when carrying out ALD (half-)cycles.
+ –	Due to the low signal generated by the sub-monolayer coverage of the surface species, differential spectra are typically required. This means that it directly yields insight into the surface species that are changing every half-cycle. A drawback is that surface species that do not change in the half-cycle reactions are not detected.
–	Typically very long integration times are required to obtain a sufficiently high signal-to-noise ratio.
–	Surface infrared spectroscopy requires a dedicated reactor with optical access through the substrate or it requires an optical path that allows for the use of an ATR crystal. Specifically-designed sample holders are necessary and the substrate material as well as the material to be deposited itself should be sufficient infrared transparent.

 

Further reading:

Surface infrared spectroscopy, Y.J. Chabal, Surf. Sci. Rep. 8, 211 (1988).

Extensive review paper with theoretical and practical aspects related to surface infrared spectroscopy. The paper does not refer to in situ ALD studies but is highly valuable anyway.

  

Study of surface groups involved in thermal and plasma-enhanced ALD of Al₂O₃ by surface infrared spectroscopy. The top spectra show the differential spectra for the Al(CH₃)₃ and H₂O half-cycles. When Al(CH₃)₃ is dosed, -CH₃ surface groups appear and –OH surface groups disappear. The situation is reversed when dosing H₂O. This confirms the well-known reaction mechanism proposed for ALD of Al₂O₃ by thermal ALD. For plasma-enhanced ALD the situation is very similar revealing that also in this case the surface is –OH covered after the co-reactant step. A small difference between the plasma-enhanced ALD and thermal ALD case is a small feature related to CO at 1662 cm^-1. This can be attributed to incomplete combustion of the ligands during the O₂ plasma exposure. The effect becomes more apparent at lower substrate temperatures.

 

Figure adapted from: Surface chemistry of plasma-assisted atomic layer deposition of Al₂O₃ studied by infrared spectroscopy, E. Langereis, J. Keijmel, M.C.M. van de Sanden and W.M.M. Kessels, Appl. Phys. Lett. 92, 231904 (2008).

 

---

Optical emission spectroscopy

Schematic representation of an ALD reactor equipped with optical emission) spectroscopy (OES). Two options are shown. In option 1, the emission is simple collected by an optical fiber connected to the reactor through a small view port. The collected light is spectrally resolved by using a simple spectrograph. In option 2, the emitted light is collected by a lens and imaged on a higher-resolution spectrograph. This allow in principle for spatially resolved measurements. Optical windows have to be protected from deposition by using windows with protection values.

 

In a nutshell:

Optical emission spectroscopy (OES) is only possible during the plasma step of plasma-enhanced ALD. In OES the (visible) radiation emitted by electronically-excited species (atoms and molecular fragments) is measured. This radiation is emitted when the excited species decay radiatively to a lower energy state (not necessarily the ground state). The emission lines measured by OES gives an indication of the (excited) species present in the plasma. The dependence of the intensity of the emission lines on the plasma conditions provides information about the processes taking place in the plasma.

Many systems for OES are commercially available. The systems range from simple, miniature spectrometers to very high-resolution monochromators with advanced detectors.

Picture of an Oxford Instruments FlexAL reactor with a compact spectrometer probing the optical emission from the plasma via an optical fiber.

Some pros and cons:

+	Ideally suited for processing monitoring of plasma-assisted ALD processes. Due to its simplicity and fast time resolution, it is also well suited for process control (it is e.g., used as endpoint technique during plasma etching).
+	In a basic configuration, OES is easy to implement and relatively cheap
–	OES spectra only contain information about electronically-excited species that decay back to a lower energy level. Ground state species cannot be measured. OES provides therefore no direct information about the most abundant species.
–	OES typically yields only indirect and qualitative information related to the ALD process,

Further reading:

Optical Emission Spectroscopy as a Tool for Studying, Optimizing, and Monitoring Plasma-Assisted Atomic Layer Deposition Processes, A.J.M. Mackus, S.B.S. Heil, E. Langereis, H.C.M. Knoops, M.C.M. van de Sanden and W.M.M. Kessels,  J. Vac. Sci. Technol. A, 28, 77 (2010).

Paper that describes the many ways in which OES can provide relevant information about the plasma-enhanced ALD process.

The top spectrum shows the optical emission from a regular O₂ plasma revealing some emission lines characteristic for electronically-excited O and O₂ present in the plasma. The lower spectrum shows the optical emission of the O₂ plasma immediately after plasma ignition when being part of the plasma-enhanced ALD process of Al₂O₃. Additional emission lines are observed which can be attributed to electronically-excited OH, CO and H. These species are created when the reaction products liberated from the surface become excited in the plasma by electron collisions.

 

 Figure adapted from: Plasma-Assisted ALD of Al₂O₃ at Low Temperatures: Reaction Mechanisms and Material Properties, E. Langereis, M. Bouman, J. Keijmel, S. Heil, M.C.M. van de Sanden and W.M.M. Kessels, ECS Trans. 16, 247, (2008).