Towards quantitative modelling of magnetron sputter deposition

Size: px

Start display at page:

Download "Towards quantitative modelling of magnetron sputter deposition"

Julia Atkinson
5 years ago
Views:

1 Towards quantitative modelling of magnetron sputter deposition D. Depla Dedicated Research on Advanced Films and Targets Ghent University Magnetron sputtering: components A magnetron discharge is a magnetically enhanced glow discharge. Key components are the magnets located below the cathode. 1 1

showing the hopping of the electron on the

2 Magnetrons Planar magnetrons: target usage

2 Magnetron sputtering : the discharge Simulation of an electron trajectory above a cylindrical magnetron Schematic of the electron movement above a planar magnetron, showing the hopping of the electron on the target, and the path defined by the ExB drift. 2 Magnetrons Planar magnetrons: target usage Optimalisation of the magnet configuration 23% 46% 65-90% 3 2

Together with the pressure change, we notice changes in discharge voltage and deposition rate. And who would care if the deposition rate stayed constant?

3 Together with the pressure change, we notice changes in discharge voltage and deposition rate. And who would care if the deposition rate stayed constant? Reactive magnetron sputtering Metallic mode High deposition rate No fully oxidized layers Reactive mode Low deposition rate Fully oxidized layers 4 The processes The processes Sputtering Chemisorption Knock on and direct implantation (Re)deposition 5 3

4 Sputtering (1) Astrongneedforsputteryieldsnotonlyforoxidesbutalsoformetals. Quantification of target erosion/deposition rate Measuring the mass difference and discharge current at constant discharge voltage Difficulties 1. Electron yield 2. Neutrals 3. Ionenergy(seef ion ) On the effective sputter yield during magnetron sputter deposition D. Depla NIMB 328 (2014) 65 6 Sputtering (2) For oxides : data has been obtainedfory 2 O 3 andal 2 O 3 See later Neutrals Charge exchange mechanism Window based on literature I.Y. Burmakinskii et al. Tech. Phys., 49 (1) (2004), pp E. Bultinck, et al., New J. Phys., 11 (2009) Consistent with advanced simulations E. Bultinck, et al., New J. Phys., 11 (2009) M.J. Goeckner, J.A. Goree, T.E. Sheridan IEEE Trans. PlasmaSci. 19 (1991),

5 Chemisorption : target (1) After exposing the target to oxygen the target is sputter cleaned. The sputter cleaning time allows to calculate the thickness. Discharge voltage (V) discharge on power supply on discharge current stable discharge -0.2 discharge voltage current time after the start of the recording (ms) Discharge current (A) 8 Chemisorption : target (2) Good agreement with literature on the oxidation rate So, no influence of the bombardment prior to the oxygen exposure But the plasma does not contain only O 2. See later This story however shows another important message. C. Palacio, J.M. Martinez-Duart Thin SolidFilms90 (1982), J.M. Sanz, S. Hofmann, J. Less-Common Met. 92 (1983), P.B. Sewell, D.F. Mitchell, M. Cohen, Surf. Sci. 29 (1972)

6 Chemisorption : target (3) How clean is a target during sputtering? R. Schelfhout, K. Strijckmans, F. Boydens, D. Depla Applied Surface Science 355(2015) Method: Chemisorption : substrate (1) In corporation coefficient = # depositions at same conditions oxygen flow argon pressure pumping speed discharge current T-S distance chamber geometry # depositions WITH fixed, small oxygen flow # depositions WITHOUT fixed oxygen flow to account for the background gases EPMA Amount in the layer Amount arriving Mass Spectrometry at the position of the substrate under the same conditions as the depositions W.P. Leroy, S. Mahieu, R. Persoons, D. Depla Thin Solid Films518 (2009) 1527 W.P. Leroy, S. Mahieu, R. Persoons, D. Depla Plasma Processes and Polymers 6 (2009) S

7 T-S distance Ar pressure Chemisorption : substrate (2) Aluminium 15 cm 0.34 Pa O 2 /Ar 1.5/20 I discharge p eff 0.47 A 5.26 x 10-4 Pa ± 0.44 x 10-4 Pa α ± T-S distance Ar pressure Yttrium 8cm 0.34 Pa O 2 /Ar 1/20 I discharge p eff 0.47 A 5.54 x 10-5 Pa ± 1.15 x 10-5 Pa α ± T-S distance Ar pressure Magnesium 10cm 0.34 Pa O 2 /Ar 2.9/20 I discharge p eff 0.47 A 2.53 x 10-4 Pa ± 0.67 x 10-4 Pa α ± T-S distance Ar pressure Copper 15cm 0.34 Pa O 2 /Ar 1.5/20 I discharge p eff 0.47 A 1.57 x 10-2 Pa ± 0.75 x 10-2 Pa α ± status status status voltage Knock-on and direct implantation V O2 V oxar V Ar argon Oxygen magnetron A similar scheme as for chemisorption on the target, but now the target is oxidized by exposure to an oxygen plasma. time t This time scale is only understandable when oxygen is implanted. 13 7

Knock on and direct implantation 8.3 7.3 6.4 5.5 4.6 3.7 2.8 1.8 0.9 0 monolayers Weassumethatallnitrogenis bonded to titanium.

8 Knock on and direct implantation monolayers Weassumethatallnitrogenis bonded to titanium. The calculated thicknesses fit nicely with implantation range of nitrogen ions into titanium at energies defined by the discharge voltage Guttler D, Abendroth B, Grotzschel R, Moller W, Depla D Applied Physics Letters 85 (2004) (Re)deposition SIMTRA is particle trajectory code. New version with improved GUI is available for free 15 8

$(Re)deposition The fraction of sputtered material returning to the target scales almost linearly with the argon pressure.$

9 (Re)deposition The fraction of sputtered material returning to the target scales almost linearly with the argon pressure. Target redeposition is high in the racetrack but this is not important as it is balanced by the erosion process. However outside the racetrack oxide formation will occur. 16 The model The model Target description Substrate description How to use the model?: GUI 17 9

Target description The transport is defined by sputtering and redeposition As we know ions get implanted, we describe the ion implantation in the bulk

We describe explicitly the chemical reaction kn r n m Surface processes are described as a balance between removal and addition K. Strijckmans and D.

10 Target description The transport is defined by sputtering and redeposition As we know ions get implanted, we describe the ion implantation in the bulk Initially the implanted atoms are not chemically bonded. We describe explicitly the chemical reaction kn r n m Surface processes are described as a balance between removal and addition K. Strijckmans and D. Depla Journal of Physics D: Applied Physics 47 (2014) Target surface description DerivationbasedonconservationofparticlesMandMR z One example Transport metal compound chemisorbed oxygen Chemisorption knock-on Sputtering Redeposition Redeposition 19 10

Substrate condition metal compound chemisorbed oxygen Similar story for the substrate RSD2013= critical revision of RSD2009 Two discrete layer + subsurface Saturation limit Redefining surface

11 Substrate condition metal compound chemisorbed oxygen Similar story for the substrate RSD2013= critical revision of RSD2009 Two discrete layer + subsurface Saturation limit Redefining surface equations Simulations are possible in steady-state and time Multi-cell approach for both target and substrate 20 The Chinese menu card approach Solution: Steady Time Uniform deposition C,D,E A,B Deposition profile C,D,E A,B Uniform deposition No redeposition D,E A,B,C Redeposition D,E A,B,C No redeposition Uniform current B,C,D, E A Non-uniform current B,C,D,E A Uniform current Nosaturation limit E A,B,C, D Saturation limit E A,B,C,D Nosaturation limit 21 11

12 Running RSD2013 Command line Graphical user interface (GUI) Manual Free download download SRIM download SIMTRA download RSD K. Strijckmans and D. Depla Journal of Physics D: Applied Physics 47 (2014) The parameters The parameters What do we know? What we don t know! Fitting experiments What can we learn? 23 12

Input parameters : the quest Problem:2 unknown parameters (α t, k) Reason:experimentally or by simulation hard to retrieve (even for Y c ) Goal: fit freedom and material dependency Solution: fitting

13 Input parameters : the quest Problem:2 unknown parameters (α t, k) Reason:experimentally or by simulation hard to retrieve (even for Y c ) Goal: fit freedom and material dependency Solution: fitting RSD model to experiments 24 Experimental setup Sputter conditions: Target Alor Y(D = 2 ) Process gas Ar Reactive gas O 2 S = 55 L/sor 112 L/s P base = ~10-4 Pa P Ar = 0.45 Paor 0.37 Pa I = 0.4 A, 0.5 A or 0.6 A Hysteresis experiment = stepwise in/decreasing the O 2 flow while collecting steady state values V, I and P tot 25 13

Fitting procedure Fitcriteria = 6 critical O 2 flow values goodness of the fit = worst match out of 6 fits are good if critical points fall within acceptancetolerancef a (=experimental resolution)

14 Fitting procedure Fitcriteria = 6 critical O 2 flow values goodness of the fit = worst match out of 6 fits are good if critical points fall within acceptancetolerancef a (=experimental resolution) Simulation includes measured V and I variation changing target oxidation Y m and I ion oxygen implantation profile I ion (by Y SEE ) 26 Scan algorithm Serial implementation illustrated for a 2-D parameter space (X,Y) Goal finding all (xi,yi) combinations that pass the fit criteria Ingredients starting point (Start) found by (slightly) modified version step size (Δx, Δy) parameter boundaries fit procedure + acceptancetolerancef a (>1) three lists: rejected, accepted and unfinished Limitation only for a connected region K. Strijckmans, W.P. Leroy, R. De Gryse, D. Depla Surface and Coatings Technology 206 (2012)

Fitting results : correlations one-cell target multi-cell

oxide Y c highly correlated PowerlawY c =b k a ratioy Alc /Y

15 Fitting results : correlations one-cell target multi-cell target (f a = 1.5) experimental Y c (f a = 3) experimental Y c 28 Fitting results : analysis Reaction coefficient k and sputter yield oxide Y c highly correlated PowerlawY c =b k a ratioy Alc /Y Yc andcorrelationindependentof reaction coefficient k one-cell or multi-cell target implantation profile 29 15

16 Fitting : some remarks (1) Using the experimental oxide sputter yields, we get a quite similar reaction rate for Y andalwithimplantedoatoms. Restricting the possible reaction rate constants to the same values and within the range of the experimental oxide sputter yields, we get a value for the sticking coefficient on the target. Al : 0.08±0.046 Y : 0.53 ±0.14 How this compares to literature? Aluminium: (average value over 6 different papers) Yttrium: 0.44(combination of 2 papers) The fitted values are much higher : ion stimulated chemisorption, but mainly the presence of atomic oxygen(sticking coefficient 1) 30 Fitting : some remarks (2) Atomic flux to the molecular flux is order 0.1 So, not only the sputter yields fit, but also the sticking coefficients. Al: 0.08±0.046 Y:0.53 ±

The examples The examples Discharge voltage behaviour Target rotation Argon pressure influence Critical point prediction 32 Discharge voltage behaviour Flow 120 s Duty cycle between

17 The examples The examples Discharge voltage behaviour Target rotation Argon pressure influence Critical point prediction 32 Discharge voltage behaviour Flow 120 s Duty cycle between 100% and 0% Target: Al(406.4x127 mm) Reactivegas:O 2 Constantcurrent:2A Argonflow:50sccm Measurement performed at ENSAM (Cluny, France) With cooperation of A. Besnard, N. Martin 33 17

increase on reactive gas addition 34 Details We see this slight increase when the oxygen pressure change is small.

18 Details Discharge voltage for a poisoned targetinar/o 2 Small decrease when oxygen is removed Simple : effective ionisation energy of Ar is lower as compared to oxygen Discharge voltage for a metal target in pure argon Small increase on reactive gas addition 34 Details We see this slight increase when the oxygen pressure change is small. Simulation : the coverage by chemisorption results in a voltage increase, while compound formation results in voltage decrease

19 Target rotation (1) Hysteresis shifts to lower oxygen flow on increasing the rotation speed D. Depla, J. Haemers, G. Buyle, R. De Gryse, J. Vac. Sci. Technol. Science A 24 (2006) Target rotation (2) 37 19

20 Target rotation : sputter cleaning (1) Sputtering a stationary target, and then sputter cleaning in pure argon while rotating X.Y. Li, D. Depla, W.P. Leroy, J. Haemers, R. De Gryse J. Phys. D: Appl. Phys. 41 (2008) (6pp) 38 Target rotation : sputter cleaning (2) More rounds 39 20

21 Target rotation : sputter cleaning (3) In poisoned mode: faster sputter cleaning than in metallic mode Reason : the deposited layer is much thinner because the deposition rate is much lower 40 Including deposition enables to mimick the sputter cleaning experiments 41 21

22 t out s t in S enters plasma region fully oxidized increaserpm smallert in =lesstimeforoxideremoval critical gas flow decreases with increasing RPM 42 Interpretation Without deposition, we would expect a similar behaviour, i.e. 1/RPM behaviour But with deposition Lower rotation speed means a thicker layer, butthereismoretimetosputterthelayer Higher rotation speed means a thinner layer, butthereislesstimetosputterthelayer t out s 43 22

23 Argon pressure (1) E. Särhammer, K. Strijckmans, T. Nyberg, S. Van Steenberge, S. Berg, D. Depla Surface and CoatingsTechnology 232 (2013) Argon pressure (2) Low sticking coefficient High sticking coefficient 45 23

Prediction of the first critical point The first critical point is defined as the lowest flow before entering the poisoned regime Mainly defined by the reaction of the reactive gas with the deposited

24 Prediction of the first critical point The first critical point is defined as the lowest flow before entering the poisoned regime Mainly defined by the reaction of the reactive gas with the deposited metal 46 Prediction and agreement The RSD2013 model predicts not one but two hysteresis curves. Recently (paper just accepted in APL) this is now experimentally proven. Origin : 1. The hysteresis originates from the avalanche mechanism related to the getter activity of the deposited material. 2. Two conditions : the transition from metallic mode to poisoned mode results in avalanche mechanism due to implanted, non-reactive reactive ions

Conclusions The processes are relatively well understood and described.

N. Mason in The plasma 2012 roadmap Developing such models and gaining a detailed understanding of the physical and chemical

status of the database for characterizing electron, ion and photon interactions with those atomic and molecular species within

25 Conclusions The processes are relatively well understood and described. The model is (too) simple but gives quantitative results, But the parameters are hard to get, and needs dedicated experiments. N. Mason in The plasma 2012 roadmap Developing such models and gaining a detailed understanding of the physical and chemical mechanisms within plasma systems is intricately linked to our knowledge of the key interactions within the plasma and thus the status of the database for characterizing electron, ion and photon interactions with those atomic and molecular species within the plasma and knowledge of both the cross-sections and reaction rates for such collisions, both in the gaseous phase and on the surfaces of the plasma reactor. 52 Acknowledgements Further development of the RSD model Simulations Koen Strijckmans Measurements of incorporation coefficients and sputter yields Wouter Leroy Further developments of the SiMTRA code Francis Boydens Oxide formation on the target Roeland Schelfhout 53 25

26 Learn more and enjoy December 1 st and 2 nd Short course on Nov. 30 th 2016 Downtown Ghent Invited speakers : H. Urbassek, R. Smith, E. Lewin, Z. Barber, and C. Schiffers Sarton Chair Lecture by J. Greene The History of Ancient Metallurgy Leading to Modern Materials Science 54 26

Reactive magnetron sputter deposition: a journey from target to substrate

Reactive magnetron sputter deposition: a journey from target to substrate D. Depla Dedicated Research on Advanced Films and Targets Ghent University Setting the scene Who? Some keywords Main features of