Lecture 16 Chemical Mechanical Planarization

Transcription

1 Lecture 16 Chemical Mechanical Planarization 1/75

2 Announcements Term Paper: The term paper should be handed in today: Tuesday 21 st November. The term paper will be returned to you in class on Tuesday 28 th November. 2/75

3 Announcements Homework 4/4: Will be online later today (Tuesday 21 st ). It will be a normal length homework (like HW 1 & 2). 25 marks. Due Tuesday 28 th November. I will return it after the final exam. I will post the solutions immediately after I receive all the homeworks. 3/75

4 Announcements Final Exam: Gleeson 100. Tuesday December 5 th Scheduled: 14:00 to 15: minutes (1 hour 50 minutes). 3 out 4 questions. Otherwise format will be the same as midterm. Closed book and closed notes. All constants and most formulae will be given. Review lecture November 28 th will go through examples. No lecture November 30 th I will upload some sample questions. 4/75

5 Lecture 16 Chemical Mechanical Planarization Overview. Chemical Mechanical Planarization Process. Quantifying Chemical Mechanical Planarization. Chemical Mechanical Planarization in Industry. Overview of Electronic Materials Processing. The Future of Electronics. 5/75

6 Chemical Mechanical Planarization Overview 6/75

7 Planarization In VLSI we need to carry out many vertical depositions. Many depositions are conformal. But we often need access to buried features. 7/75

8 Planarization Similarly, many processes lead to non-uniform features. CVD. Oxidation. Etching. Photolithography. 8/75

9 CMP Chemical Mechanical Planarization (CMP) is a process for smoothing surfaces. It is a hybrid technique: Chemical etching. Abrasive polishing. 9/75

10 CMP Why do we need to combine mechanical grinding and chemical etching? Chemistry alone cannot attain planarization because most chemical reactions are isotropic. Mechanical grinding alone may theoretically achieve planarization but the surface damage is high as compared to CMP. 10/75

11 CMP Pressure = 5 10 psi Cross section Wafer rotation, ω w = 60 rpm W afer Slurry Pad Want enhanced removal from high points in surface Pad rotation, ω p = 30 rpm 11/75

12 Key Issues of CMP Polish rate. Planarity. Selectivity. Uniformity. Post CMP ease of cleaning. 12/75

13 Chemical Mechanical Planarization Process 13/75

14 Planarization Local Planarity: Surfaces are flat locally but the surface height varies across the die. Global Planarity: Surfaces are flat across the entire wafer (or stepper field). Local Planarization Global Planarization Chemical Mechanical Planarization is the only IC process which can achieve global planarity. 14/75

15 Wafer-Pad Contact Modes Hydrodynamic contact mode: Chemicals Velocity Down pressure High velocity Low down pressure Abrasives Pad asperity Pad Three body abrasion Inactive Solid-solid contact mode: Velocity Down pressure Low velocity High down pressure Chemicals Abrasives Pad Two body abrasion Inactive 15/75

16 Pad Materials: polyurethane + filler, polyurethane impregnated felts. Key to transfer of mechanical forces to surfaces being polished. W afer Slurry Pad The pad contacts the higher areas of the wafer. 16/75

17 Pad Harder pad: higher removal rate & better within die uniformity. Softer pad: better within wafer uniformity. Pad break-in is carried out on dummy wafer to stabilize pad performance before running product. The properties of the pad will depend on the surface material of the wafer. Different pads used for different stages in the VLSI process. 17/75

18 Slurry Solution The slurry solution consists of abrasive materials and also contributes some chemical reaction. Abrasives provide mechanical action: E.g. SiO 2, Al 2 O 3, CeO 2 : W afer Slurry Pad We want a uniform size distribution ( nm). Isoelectric Point: the ph when the abrasive surface has no net charge (neutral). 18/75

19 Slurry Solution The slurry solution consists of abrasive materials and also contributes some chemical reaction. Composition (basic, acidic) provides chemical action. Electrochemical process to form oxides and hydroxides (oxidation/reduction). ph is very important. Metal slurries are more chemically active than oxide slurries. W afer Slurry Pad 19/75

20 Slurry What do the chemicals do? Modify the polishing process. Add control and speed. Prevent defects. Remove byproducts. 20/75

21 Slurry Solution In reality the slurry consist of many components: Water. Abrasive nanoparticles. Acids and bases. Anti-coagulating agents. Corrosion inhibitors. Surfactants. Oxidizers. Buffers. Bactericides and fungicides. For mechanical abrasion For controlling ph To ensure high dispersion To stop corrosion pits in metals To neutralize charged surfaces Metals are often oxidized before being polished To control reaction rate To stop solution spoiling 21/75

22 Nanoparticles Various abrasives are employed in the slurry: Silica (SiO 2 ) Alumina (Al 2 O 3 ) Ceria (Ce 2 O 3 ) Polymers 22/75

23 Nanoparticles Surface area is important for abrasion: Surface area = 15 m 2 /g Surface area = 100 m 2 /g 23/75

24 SiO 2 CMP Chemistry Glass polishing for preparation of optical lenses is well-established. Basic solutions: High ph. H 2 O provides the chemical component of the polish: H 2 O enters the SiO 2 and weakens it by breaking Si-O bonds: Si Si O Si + H 2 O 2 Si OH O Si +4H 2 O Si(OH) 4 Si(OH) 4 is highly soluble in H 2 O at ph > 10. Stress of the abrasive particles accelerates these reactions. 24/75

25 Mechanism of Metal CMP Certain deposition techniques are conformal: Oxide Metal Substrate Formation of passivation oxide film (via water or oxidizer in slurry): Oxide Metal Substrate 25/75

26 Mechanism of Metal CMP Depassivation (abrasion): Oxide Metal Substrate Dissolution takes place via corrosion (chemically): Oxide Metal Substrate 26/75

27 Mechanism of Metal CMP Repassivation. Oxide Metal Substrate Planarization by repetitive cycles: Oxide Metal Substrate 27/75

28 Quantifying CMP 28/75

29 Measures of Planarity Planarity is quantified via step properties: φ R SH R Global Planarization 100 m 0.5 o Local Planarization m 0.5 o - 30 o Surface Smoothing m 30 o Degree of planarization: Step Height after CMP DP = 1 SH postcmp SH precmp Step Height before CMP 29/75

30 Preston s Equation For purely mechanical polishing, the Preston Equation was developed empirically: R = Δh Δt = K PPV pad Where: R is the polishing rate. Δh is the change in height. Δt is the time the polish is applied for. K P is the Preston coefficient. P is the pressure applied to the pad. V pad is the velocity of the pad relative to wafer. 30/75

31 R Preston s Equation In CMP the observed rate deviates from the predictions of Preston s Equation: PV 31/75

32 R R Preston s Equation In Preston s Equation: Material removal rate typically over-estimated. Model derived for glass polishing (hard pad). A threshold pressure often exists. Modifications are necessary to Preston s Equation: PV P m V m 32/75

33 Theoretical CMP Models Preston s Equation (1926 based on glass polishing): R = K P PV pad Tseng and Wang (1997 hydrodynamic boundary layer, Hertzian penetration model): R = K P P 5/6 1/2 V pad Shi (1998 particle penetration, pad elasticity): R = K P P 2/3 V pad 33/75

34 Chemical Mechanical Planarization in Industry 34/75

35 CMP Systems Examples of real CMP systems. 35/75

36 Pressurized Polishing Head 36/75

37 Linear Polisher Configuration Slurry dispenser ω w Rotating wafer carrier Wafer (upside down) air-bearing platen for uniformity control Continuous belt polish pad linear velocity, v p 37/75

38 Linear Polisher Configuration 38/75

39 CMP Defect Modes There are many potential problems with CMP: 39/75

40 Particle Accumulation 40/75

41 Macro Scratches 41/75

42 Defects Due to Trenching 42/75

43 Micro Scratches 43/75

44 Corrosion 44/75

45 Corrosion Corrosion is where a pit will be formed due to chemical processes: Abrasive Particle Inhibitor H 2 O 2 Metal Oxide Substrate Oxide Substrate Oxide Substrate Inhibitors in the slurry can be applied to protect porous metal layers. The more inhibitor, the lower the corrosion, but too much can stop the polishing process all together. 45/75

46 Defects Due to Slurry fumed colloidal 46/75

47 Additives Sometimes surfaces can become charged in the CMP - - process: This can accelerate or decelerate the process unpredictably. Charged surfactants can be included in the slurry to neutralize charges. 47/75

48 Advantages of CMP Global Planarization can be realized. Universal: all types of materials can be planarized. Allows alternative method to pattern a material that cannot be readily plasma etched (e.g., Cu). Potential to be a low cost process. Low complexity. 48/75

49 Disadvantages of CMP New process optimized. Lack of fundamental understanding. Most models are empirical. Potential defect mechanisms: Scratches from abrasive. Residual abrasive particles. Delamitation at weak interfaces. Corrosive attack from slurry chemicals. 49/75

50 Overview of Electronic Materials Processing 50/75

51 What we Have Covered Sand to Ingot Wafer Dicing 51/75

52 What we Have Covered Photolithography Ion Implantation 52/75

53 What we Have Covered Etching Deposition 53/75

54 What we Have Covered Electrodeposition CMP 54/75

55 What We Did Not Cover Circuit Design Packaging 55/75

56 The Future of Electronics 56/75

57 The End of Moore s Law From The Economist 2016: 57/75

58 The End of Moore s Law Peter Lee (Microsoft): There s a law about Moore s law, The number of people predicting the death of Moore s law doubles every two years. 58/75

59 The End of Moore s Law 59/75

60 The End of Moore s Law We know the lattice spacing of Si is 5.43Å. So there is a fundamental limit Even before we get there, the electron wavefunction will not be contained in the device. Amplitude Probability 60/75

61 The End of Moore s Law There are economic reasons: 61/75

62 Short Term Approaches Extreme UV lithography (EUV) should get us down to < 10nm. However there are many challenges: Absorption of light by many components of system. New photoresists are required. Mirrors and sources need to be carefully designed / selected. 62/75

63 Short-Term Approaches High κ-dielectrics allow a higher capacitance with the same thickness. C = capacitance. C = κε 0A x κ = relative permittivity. ε 0 = vacuum permittivity. A = device area. x = dielectric thickness. 63/75

64 Short-Term Approaches High κ-dielectrics allow a higher capacitance with the same thickness. C = κε 0A x With a higher κ dielectric, a thicker gate dielectric layer might be used which can reduce the leakage current flowing through the structure as well as improving the gate dielectric reliability. 64/75

65 Short-Term Approaches FinFETs can be used to apply a field from 3- sides rather than 1: Capability to get better performance from same space. 65/75

66 Long-Term Approaches Optical computing: processing information with light rather than charge, 66/75

67 Long-Term Approaches Tunneling and quantum transistors. Switching time for quantum-mechanical tunneling << than for field-effect transistors. 67/75

68 Long-Term Approaches 2-dimensional materials. Graphene was the first, but is not semiconducting. Many more exist now. 68/75

69 Long-Term Approaches 2-dimensional materials. 69/75

70 Long-Term Approaches Metal-Insulator Transition. Some materials exhibit discreet change in electronic properties at certain temperatures. Can we induce phase-transition using electrical signals? /75

71 Long-Term Approaches Quantum computing. Rather than processing bits (1) or (0) we can process superpositions of states. These are essentially mixtures of 1 s and 0 s. ȁψ =ȁ1 ȁψ =ȁ0 ȁψ = ȁ 1 +ȁ0 For certain, applications quantum computers can be greatly superior to electronic computer: Cracking RSA encryption. Travelling salesman problem. Certain AI applications. 2 71/75

72 Long-Term Approaches Quantum computing Require low-temperatures (mk). Still in development. 72/75

73 Long-Term Approaches Quantum computing Commercial quasi-quantum computers do exist. 73/75

74 Long-Term Approaches Quantum computing You can already run calculations: 74/75

75 Long-Term Approaches Neuromorphic Computing. Computation strategy is based on the way neurons processing information in brains. Potential applications in artificial intelligence. 75/75