Structure of high-reflectance La / B-based multilayer mirrors with partial La nitridation

We investigate a hybrid thin film deposition procedure that significantly enhances reflectivity of La/B based nanoscale multilayer structures to be used as Extreme UV mirrors at 6.7 nm wavelength and beyond. We have analyzed the La-nitridation process in detail, and proposed a growth mechanism and deposition procedure for full, stoichiometric passivation of La, avoiding the formation of optically unfavorable BN formation at the LaN-on-B interface. A partial nitridation was applied and studied as a function of the nitridation delay. © 2016 Author(s). All article content, except where otherwise noted, is licensed under a Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/). [http://dx.doi.org/10.1063/1.4968175]


INTRODUCTION
The La/B-based multilayers have the potential to serve as mirrors for 6.x nm wavelength EUV lithography 1 and various applications, for instance, EUV telescopes for space research, 2,3 ultrasensitive analysis of materials by x-ray fluorescence 4 or optics for high-intensity free electron lasers (FEL). 5,6Different researchers achieved notable reflectivity at 6.x nm, dealing with different methods to control the interlayer quality.Examples include a LaN/B 4 C system showing 58.1% at 6.645 nm at off-normal AOI 10 deg., 7 and the deposition of carbon barriers on the La-on-B4C interface yielded 58.6% at 6.66 nm, measured at 20.9 deg off-normal AOI. 8 For LaN/B reflectivity of 57.3% was achieved at 6.65 nm, measured already at near-normal incidence of AOI=1.5 deg off-normal. 9 significant step towards the application-desired reflectivity of ∼70% has been made by the so called delayed (partial) nitridation of La developed and demonstrated in Ref. 10.The obtained 64.1% reflectivity at AOI=1.5 • off-normal, λ = 6.65 nm is a record value so far.The aim of this paper is gaining a deeper insight into the processes taking place at the interfaces during the partial nitridation.We first explore conditions required for the complete passivation of La layers.Then we focus on the formation of compounds during the full nitridation process of La layers.Finally the partial nitridation processes is investigated in terms of the formed compounds and their influence on 6.x nm reflectivity as a function of the nitridation delay.

EXPERIMENTAL
The deposition of multilayers was performed using DC magnetron sputtering onto natively oxidized super-polished Si substrates (RMS ∼2.0 ± 0.1 Å) in a vacuum chamber with a base pressure of 1 × 10 8 mbar.The coater had no load lock.In order to protect the targets, especially, La, from atmosphere they were kept in vacuum-sealed packages and as quickly as possible installed inside the chamber.Right before each deposition the targets were cleaned by pre-sputtering till stable discharge parameters (current, voltage) and stable deposition rates measured by Quartz-Crystal Microbalances (QCM) were achieved.The working gas pressure was ∼2 ×  0.03 nm/s for La, 0.05 nm/s for LaN, and 0.03 nm/s for B. All multilayers were covered with a 2.0 nm B layer to prevent oxidation.For all experiments, including EUV reflectivity measurements, 50-period multilayers were used (instead of 220-period full stacks).Thermo Scientific Theta Probe instrument was used for XPS measurements employing monochromatic Al-Kα radiation.Two types of XPS measurements were done: employing the ion depth-profiling and non-destructive measurement (angle resolved or AR XPS) of as-deposited samples.The XPS depth-profiling, using 0.5 keV Ar ions, was aimed to assess the in-depth distribution of chemical elements.The non-destructive measurement was used to avoid changes in the structure that provoke formation of compounds that initially were not present in the structure.For that, we performed angular resolved measurements with electron take-off angles from 27 • to 79 • relative to the surface normal.In the XPS measurement, the probing depth is in the order of 5 nm, so only the top multilayer period was analyzed.The measured top layers were representative for the deeper layers in the multilayer.Indeed, sputter-cleaning of the samples (removing about 0.5 nm top contaminated part of the B layer) with subsequent analysis of XPS spectra revealed that there is no contamination of B or LaN by O or C within the XPS detectability.However, it is important to notice that a small part of the top B layer would still react with N 2 from atmosphere, with small amount of BN being formed at the surface.The signal from BN compound was present also for the elemental B reference sample, which set the lowest limit of BN detectability in the multilayer itself.
For the determination of the multilayers period, the grazing incidence X-ray reflectivity (GIXRR) was measured, employing PANalytical X'Pert X-ray diffractometer with a four-bounce monochromator (Cu-Kα radiation, 0.154 nm).

SYNTHESIS OF FULLY-PASSIVATED LaN
To minimize the chemical interaction with B layers in a multilayer, La should be fully passivated, i.e. as many La bonds as possible should take part in LaN compound formation.The passivation of La is implemented by the reactive magnetron sputter deposition of La with the addition of nitrogen gas into the chamber. 9To find out the nitrogen pressure required during the deposition, modelling of the reactive sputtering process could be employed, for instance as developed by S. Berg et al. 11 However, without precise values of the La and LaN sputtering yields (reactive deposition), the sticking coefficient of nitrogen to lanthanum, without taking into account nitrogen ion implantation into the target, etc., accurate calculations seem to be impossible.Moreover, in the plasma N • , N + , and N + 2 are produced via various electron-impact reactions, 12 and positively charged species impinge the La target, neutralize (with certain probability) and reflect back, in the direction of the growing La layer, 13 performing additional nitridation of deposited La.Modern advanced models of reactive deposition, for instance 14 require extraction of certain parameters by fitting the experimental data, 15 which is time consuming.Therefore in this work we applied an experimental approach to determine conditions required for the complete passivation of La.
The required partial pressure of N 2 for this was found empirically by varying the pressure and employing XPS analysis of the atomic percentage of N in the multilayer.In Fig. 1 XPS depth-profiles of nitrogen are plotted.For the case of 6 sccm N 2 , saturation of La with N clearly was not reached, implying that B would still react with La on both interfaces due to the favorable LaBx formation, with the enthalpy of formation of the stoichiometric LaB 6 compound being ∆H(LaB 6 )=-130 kJ/mol. 16he presence of LaB 6 in non-passivated La/B multilayers was confirmed in 17 by wide-angle x-ray diffraction analysis.For the under-saturated LaN, the measured soft x-ray reflectivity around 6.7 nm wavelength at AOI of 1.5 • off-normal was only about 14.0±0.5% for a 50-period multilayer stack, where the indicated error bar covers the typical reflectance reproducibility.The relatively low value is explained by the formation of La bonds with B, enhancing the interface layer thickness and reducing the optical contrast.Further increase of the N 2 flow higher than 12 sccm seems to result in the saturation of the N content in La (same N percentage in La on Fig. 1 for 12-21 sccm).This could be explained by the fact that La will not take more N during the reactive sputter deposition than needed to form a stoichiometric nitride.It was expected that a fully saturated LaN would result in an improved optical contrast due to reduced chemical interaction between the adjacent layers.Indeed for the N 2 flows 12-18 sccm the formation of a fully saturated La nitride resulted in a gain of the reflectance up AIP Advances 6, 115117 ( 2016 to 18.5±0.5% at the same wavelength.However, further increase in N 2 flow resulted in a reflectivity loss.Namely, for an N 2 flow of 21 sccm the reflectivity dropped to ∼12.5% and this is even lower than it was for the under-saturated case.Markedly, for the N 2 flow of 21 sccm nitrogen is already detected in-between the La layers at the XPS depth-profile (see Fig. 1, B layer), indicating that the nitrogen atoms are present in the boron layers and/or at the interfaces.Note that due to significant alteration of the structure during depth-profiling by ion intermixing and probable sputter-induced compound formation the accurate composition of the samples could not be determined from this measurement.

BN FORMATION
In order to reveal the compounds formed in the deposited stack, a AR-XPS measurements of as-deposited samples was employed.The measured spectra of B1s clearly showed the energy shift typical for the presence of the BN compound.The intensity of the corresponding BN peaks increases with the higher N 2 flows used for LaN deposition.For the adequate comparison of the BN amount of the multilayers with somewhat different periods, it had to be taken into account that the thicker B and LaN layers absorb more radiation of the probing X-ray beam in the XPS analysis.Moreover, a different degree of oxidation of the cap and the first LaN layer could make comparison of the absolute atomic percentage of BN unreliable.As a solution, the peak area ratios of BN to total B (elemental B1s and the B-compound(s)) were calculated.Their comparison is represented in Fig. 2a.The empirically obtained ratio for elemental B is also plotted as a reference where BN is present at the surface due to the exposure of the sample to atmosphere.As observed in Fig. 2a, the peak area ratio BN/total B increases in the whole N 2 flow range explored, indicating the amount of BN increasing with N 2 flow, from a certain onset value on.This can be explained by the fact that at higher N 2 flows there is more N available which interacts with B at the initial stage of LaN deposition.
There are two "types" of interfaces in LaN/B multilayers: B-on-LaN and LaN-on-B.In order to reveal at which of them the most of BN is formed, AR-XPS measurements were done.The XPS spectra collected when the probing beam was at a more grazing angle clearly demonstrated less BN amount.That means that most BN is formed on the deeper-situated, LaN-on-B interface.The supportive schematic representation is sketched in Fig. 2b.
It is also possible to follow the trend of the amount of the formed BN at the interfaces by determining an effective expansion or compaction of the period due to the compound formation for various deposition conditions.For that a dedicated set of samples was deposited varying N2 flow values.In every deposition run a few stacks were deposited on top of each other, each one having a different LaN thickness.From GIXRR measurements periods of the stacks in every individual deposition run were determined.The extracted period values were plotted versus the LaN deposition times used for different stacks deposited in the one deposition run.An example of such analysis for a deposition with N2 flow of 14 sccm is presented on FIG. 3.
The points were fitted with a linear regression.The slope of the line, about 0.04, determines the deposition rate of LaN.Extrapolation of the line to zero yields a value (∼2.3nm) corresponding to a B layer thickness plus a certain value related to interface compaction or expansion.The latter was obtained by subtracting a known from a separate calibration deposited B thickness.In our experiment this value is positive that means an effective expansion of the multilayer period.It is important to note that this approach does not allow to judge about the width of the interface, because the depth of interdiffusion and real densities of the formed materials are not known.
On Fig. 4 the determined above effective multilayer period expansion versus nitrogen flow used for La nitridation is plotted.The expansion is clearly not linear with the N 2 flow, and can be considered as two processes as indicated by the two lines with different slopes at the graph on Fig. 4. To explain this we consider a factor of competition between different atomic interactions during La deposition on B in N 2 environment.During this, two simultaneous processes are expected to occur: the interaction of B with La and interaction of B with N. The enthalpies of formation of the corresponding compounds are 130 kJ/mol (LaB 6 ) and 253 kJ/mol (BN), respectively. 16Therefore, here we disregard the La -B interaction which is less favorable from thermodynamics point of view.At the initial stage of LaN layer deposition, if the N 2 flow is low, La takes significant part of the overall amount of N to form LaN. The remaining part is consumed by B to form BN, which causes the multilayer period to expand.The combination of these two processes results in the observed period expansion rate at low N 2 flows on Fig. 4. At higher N 2 flows, we observe a saturation in the amount of N in La.As shown above, La cannot take more N than needed to form stoichiometric LaN compound.As a result, after LaN is fully saturated, all the remaining (due to further increased N 2 flow) nitrogen species are available to react with B to form BN. At this stage the BN formation is greatly enhanced, that corresponds to the steeper slope on the Fig. 4.

Special scheme of La nitridation ("delayed" nitridation)
In Ref. 10 the effect of a thin BN interlayer on the LaN-on-B interface was calculated, and a significant drop of 6.x nm reflectivity due to this BN was shown.This loss can be explained by comparing the optical constants of B and BN at 6.70 nm wavelength.The absorption coefficients for B and BN are β(B) ∼ 4.1e-4, β(BN) ∼ 8.9e-4, 18 which means that BN absorbs 6.7 nm radiation more than 2 times stronger than B. The refraction coefficient for boron nitride (δ(BN) ∼ 5.4e-3) is ∼4 times higher than that for B (δ(B) ∼ 1.3e-3).This means that instead of working as a spacer, BN reduces the optical contrast with the reflector layer, La(N), which also contributes to reflectivity losses.So, from the application point of view, it is important to notice that the standard reactive deposition of LaN in nitrogen atmosphere cannot in practice yield the highest possible reflectance of LaN/B multilayers.The formation of optically unfavorable boron nitride is inevitable even at relatively low nitrogen flows.But the latter should be high enough to obtain, as already mentioned above, a fully-passivated (saturated) LaN, i.e. to have it chemically inert to the maximum possible degree.
To prevent interaction of nitrogen species with B atoms of the underneath layer, we synthesized structures in such a way to avoid direct contact of N with the underlying B layer at the LaN-on-B interface.The simplest way to realize that was to delay the process of lanthanum nitridation.Effectively this should introduce a pure La layer on the B layer.Below we investigate the effect of this layer on the underlying interface and its impact on the reflective properties of the multilayers.A range of La thicknesses of 0.1 to 0.8 nm was taken for this experiment before starting the La nitridation.The actually deposited amount of La was a little larger to compensate for the interaction of La with the underlying B. To determine the moment when this interaction is complete, compaction of the period was taken into account as described above.For the deposition of LaN we selected the conditions of the fully-passivated LaN (N 2 flow 12 sccm).
To determine if there is a reduction in the BN formation on the LaN-on-B interface, AR-XPS analysis was performed.The ratio of the fitted XPS B1s BN to total B1s peaks was taken for comparison (as described above).The resulting BN/total B ratio as a function of the La thickness is shown in Fig. 5.
The small error bars in the Figure 5 account for reproducibility of the peak fit with fixed constraints on the peaks.The bigger error bars stay for different fitting results with varied constraints, i.e. the uncertainty of the peak fitting model.The dashed line represents a reference BN/B ratio measured for a single layer boron film, where BN is present due to the exposure of the sample to atmosphere.The probed depth corresponds up to approximately 5 nm 19 of the top part of the structure.Strictly, the measured for this top part may not be exactly multiple of the bilayer thickness.So, the fitted percentage cannot be attributed to the composition of a bilayer.
As seen in Fig. 5 for La thickness of 0.1 nm the BN/total B ratio has the same value within the error bars as the "reference" LaN/B multilayer with fully passivated La layers.This means that 0.1 nm La did not noticeably reduce the BN formation.This is in agreement with the reflectivity measurement of this sample that showed 16.2±0.5%versus 17.2±0.5% % for the multilayer with fully passivated La.Starting from 0.3 nm of La, the BN/total B ratio becomes close to the elemental B reference, which suggests that these La thicknesses lead to a significant reduction of the amount of BN formed at the interface.This result is supported by a sharp increase in the reflectance of the multilayers to 21.5±0.5% at 6.75 nm.From this it becomes obvious that the La thicknesses of 0.3 nm and higher are sufficient to form a structure with a closed layer, which prevents interaction of B with the nitrogen species.
It should be mentioned that introducing the lanthanum layer directly on the B layer will cause a chemical interaction of La with the underlying B layer, resulting in the formation of a LaB X interlayer.Our calculations showed that reflectance of the multilayer would clearly decrease with the increase of LaBx interlayer thickness. 10Therefore it could be expected that the reflectivity of the multilayers would also decrease with the increase of the non-passivated La thickness, or larger delay in nitridation.This is provided the entire non-nitrided La would interact with B. However, like mentioned above, the multilayers with partial nitridation having La thickness in the range 0.3-0.8nm demonstrated the same reflectivity within run-to-run reproducibility of ±0.5%.This points to the fact that as soon as a certain very thin LaBx interlayer is formed, it does not grow anymore with the increase of the deposited La thickness.And as shown in Ref. 10 such a LaBx interlayer is optically more favorable than the BN formed in case of complete La nitridation, resulting in a significantly enhanced reflectance.
To find out what is in fact formed on the interface as a result of the applied delayed nitridation, XPS analysis was employed.In Fig. 6 a fit of La3d measured spectra for one of the samples fabricated with delayed nitridation is plotted together with the spectra from La and LaN references.All peaks of the fitted spectrum for the sample with delayed nitridation coincide with the ones of LaN sputtercleaned reference, but the exact fit depends on the constraints set on the fitting procedure.To fit the mentioned La3d spectrum from the partial nitridation sample the width of Gaussians had to be bigger than for fitting the spectrum from LaN reference.
Fig. 6 shows that the main La3d peaks for the elemental La reference has a position in-between the La peaks from the delayed nitridation sample.Since all La XPS peaks of the multilayer samples are rather broad, it is not possible to perform a unique peak fitting to prove or disprove the presence of elemental La.Therefore, fitting of the XPS spectra solely cannot be employed for tracing the elemental La, in this particular case.Below on Fig. 8 La3d spectra from the samples with ∼0.2 and ∼0.8 nm thick La are compared.
In the XPS measurements no differences in the position of peaks and their intensity could be observed.This means that the chemical composition and the amount of materials are the same within the accuracy of the XPS measurement.It is anticipated that elemental La is mostly nitridized by the subsequent deposition of LaN.The reasons for nitridation of La by LaN deposition on top of it could be the following.First, nitridation is chemically-driven by direct interaction with reactive nitrogen species.Experiments with a thick La layer 20 revealed that the exposure of a La layer to N 2 gas already leads to the formation of a LaN stoichiometric overlayer with a thickness of 0.5±0.2nm.The second reason, in the case of LaN reactive magnetron sputter deposition, is the presence of relatively high-energy nitrogen neutrals 13 reflected from the La target which serve as an additional source of nitridation of the deposited elemental La, as mentioned above.The remaining question is why the nitrogen species do not penetrate deeper inside the B layer.We suggest that all arriving nitrogen species are trapped inside the elemental La with the subsequent formation of LaN.As soon as LaN is formed, it serves as a good diffusion barrier against further penetration of nitrogen species.
We also made an attempt to trace the anticipated on La-on-B interface LaB 6 by XPS.The La3d spectrum from a LaB 6 crystal was collected for the purpose of reference (not shown).The main fitted La3d 5/2 peak for a sputter-cleaned (oxide and other surface contaminants removed) LaB 6 crystal appears at 837.5 eV, which is in agreement with literature values: 837.3 eV in Ref.21, or 837.5 eV in Ref. 22.The positions of those peaks almost coincide (within ∼0.2 eV) with the peaks in the La3d spectrum of the partial nitridation samples (and by so also with the peaks from our LaN reference, see Fig. 6).This makes tracing LaB 6 in the particular samples impossible if analysing XPS spectra from La levels.Due to that the XPS spectra of B1s were also analysed.In Fig. 7 the B1s spectrum from a sputter-cleaned LaB 6 crystal is presented together with the oxidized (as-is) LaB 6 crystal to demonstrate that the binding energy for the main (highest intensity) LaB 6 peak decreases just by ∼0.2 eV due to ion surface treatment to sputter the oxide.This confirms no significant structure and composition changes due to this cleaning procedure.The main B1s peak from this LaB 6 reference has a binding energy in-between B1s elemental reference value (see Fig. 7) and sub-oxide of B (fitted at about 189 eV peak for B1s reference in Fig. 7), and, in principle, may even overlap with boron sub-oxide of a certain stoichiometry.However, sub-stoichiometric boron oxides are inevitably present in the multilayer samples (due to exposure to atmosphere).Therefore, a possible contribution of LaB 6 to the B1s spectrum cannot be traced by spectrum fitting in our particular case.The side peak fitted in the B1s spectrum from LaB 6 reference at ∼187.4 eV, has a larger separation from the B1s elemental peak (∼187.9eV), but the low intensity of this side peak even for a pure LaB 6 reference, in AIP Advances 6, 115117 (2016) FIG. 7. Fitted B1s spectra from as-introduced (oxidized) LaB6 reference crystal (top), the latter measured after sputter-cleaning (middle) and elemental B reference (down).combination with the broad nature of S peaks, suggests that the contribution from a few-monolayerthick LaB 6 layer below other layers will not be visible in the measured spectrum at all.Moreover, the real stoichiometry of the formed lanthanum boride is unknown, so the formed compound may have deviation of binding energy from the expected for the stoichiometric one.So, additional experiments are required.

SUMMARY
We have investigated a hybrid thin film deposition procedure that significantly enhances the reflectivity of La/B based multilayer structures to be used as reflecting elements at 6.7 nm wavelength and beyond.The procedure is a refinement of using fully nitrided La layers.It was found that La will not take more N than needed for the formation of a stoichiometric LaN compound, however excessive N 2 during the La growth results in the formation of optically unfavorable BN at the LaNon-B interface.To avoid this, the so-called delayed (partial) nitridation at the initial stage of La growth was applied and studied as a function of the nitridation delay.A range of 0.1 to 0.8 nm thicknesses of the non-nitrided elemental La was explored.For La thicknesses ∼0.3 nm and thicker clear reduction of the BN content on LaN-on-B interface was observed by AR-XPS.This observation was correlated with a noticeable improvement of the 6.x nm reflectivity of the 50-period multilayer structures from 17.2±0.5% to 21.5±0.5%,λ≈6.75 nm.For full (220 periods) stacks, as reported, 10 the increase was from 57.3% 9 to 64.1%, AOI=1.5 • , λ≈6.66 nm.The calculated 10 theoretical maximum about 80% for an ideal multilayer structure leaves a significant space for future improvements.The mechanism of the partial nitridation is explained as the prevention of N to interact with the underlying B layer by forming a closed layer of elemental La.It has been concluded that a part of this non-passivated La layer forms a thin LaBx compound at the interfaces with B, though this LaBx is proven to be optically more favorable at LaN-on-B interface than BN.

FIG. 2 .
FIG. 2. (a)Ratios of fitted XPS peaks (B1s BN/total B1s) versus the N2 flow employed during the deposition of LaN.The mentioned ratio for a deposited layer of elemental B is plotted with a dash line (for reference).(b) Schematic representation of the multilayer top layers as analyzed by AR-XPS.

FIG. 3 .
FIG. 3. Periods of the stacks for different deposition time of LaN layers in the stacks.The LaN layers were deposited with 14 sccm N 2 flow.

FIG. 5 .
FIG. 5. Ratio of the fitted B1s BN to total B1s XPS peak for different La thickness in the multilayers with delayed nitridation prepared at 12 sccm of N 2 .