Wafer fab operation techniques

1. Dr. Len Mei
2023/10
1

2. Introduction
2

3. Characteristics of wafer fab
 Complexity
 Process steps >1,000
 Cycletime > 100 days
 Specification < 10 nm
 Defect sensitive
 Hundreds of high purity raw materials – liquid
chemicals, specialty gases, bulk gases, DI water, solids
 Hazard environment – toxic gases, explosive gases,
liquid wastes, powder
3

4. Characteristics of wafer fab
 Capital intensive
 High value WIP
 Sophisticated automation systems
 Highly sensitive data
 Multiple discipline working together – engineers
(process, product, testing, device, integration,
equipment, production, quality, utility, IT), operators
 Cleanroom – vibration, particle, humidity controlled
environment
4

5. Cooling water system for cleanroom
5

6. Power panels for equipment
6

7. Exhaust system
7

8. 8

9. Nitrogen plant
9

10. Cross section
of an IC
transistor
Minimum feature size
10

11. 3D view of integrated circuit
Oxide removed
Silicon substrate
Poly silicon
Metal conductor
11

12. 18 layers
1 Tb
12

13. Conditions of successful semiconductor
wafer fab
 Robust technology （process and design) – follow the
Design for Manufacturability (DFM) rules
 Efficient operation (from equipment layout to
automation system design, to daily operation etc.)
 Built-in quality in the operation (quality degradation
wastes resources.)
 Scrap
 Downgrade
 Lost yield
13

14. Operation  cost
 Operation can determine chip cost to a large extent.
 Chip cost breakdown:
 Wafer cost (equipment depreciation, labor, utility, fab
depreciation, raw materials, automation system)
 Cost of quality (scrap, downgrade)
 Yield (wafer, die, package)
 Backend cost
 Balance quality cost
14

15. 6 major challenges
15

16. Challenge 1 - dynamic
 Every manufacturing operation is unique.
 Manufacturing is dynamic that is constantly changing.
 Managing manufacturing has large momentum, a decision
can change the course and make it difficult to turn around.
16

17. Challenge 2 - organizational
 Manufacturing is a large organization > 5,000 people/ fab
 Multi-fab manufacturing is even more complicated
 Complicated organization – formal (department) and
informal organization (project, task force); many
disciplines
 Indecision or conflict of directions waste resources.
 A good operation requires precipitation of every member in
the factory with clear understanding of his daily, weekly,
monthly goals.
 Goals for each member, each team/ task force/department
must be coherent.
17

18. Challenge 3 – systematic
 Managing by system rather than managing by personal
judgment (mostly biased)
 Personal judgment introduces many variations.
 Consequences of these variations may not be traceable.
 System based management decisions, based on the most
updated and accurate data, make the operation consistent,
independent of the personal decisions
 Operation theory can help.
 A good manufacturing system will reduce complex
operations into easy to follow goals at different levels
18

19. Challenges 4 – resources conflict
 Production, engineering (equipment, process and
product), quality and R&D are often competing for
the same resources.
 Priority setting
19

20. Challenges 5 - constrains
 Manufacturing operation is not an isolated
operation. It has to exist and operate under the
business context.
 Factory priority is often set by business priority
and financial requirements
20

21. Challenge 6 – escalating cost
Cost escalating because:
1. Equipment is more
expensive
2. Process becomes longer
3. Economic scale
becomes larger
21

22. Benefits of improved productivity
22

23. Manufacturing efficiency
 Productivity is measured by number of wafers
processed per day (DGR – daily going rate, or moves)
and inventory turn rate
 Example: 60k/month capacity, 1,000 steps process
 60k/mth *1,000 steps/30 days = 2,000 k moves per day
 if each equipment can process 100 w/hr, and works 20
hours a day, we need 1,000 pieces of equipment
 Turn = moves/ WIP or WIP = 2,000 k / turn
 CT = total process steps/ turn
 If turn =10, WIP = 200 k, CT = 1,000/10 = 100 days
Productivity = added value/ cost
23

24. Improving productivity
24
Daily running
time
Throughput
Wafers/
hour
Total capacity
20 22
Hours/day
110
100
w/hr
If each equipment
can process 110
wafers/hour, and can
work 22 hours a day.
Total moves will
increase from
100* 20 =2,000 /day
to
110* 22= 2,420 /day

25. Cost saving of improved
productivity
25

26. More output or shorter CT
26

27. WIP inventory value
Assume average WIP value is $1,000
Wafer start cost $200
Wafer finish cost $1,800
27

28. How to improve productivity
28

29. Principle
 Improve the efficiency of all resources (man, machine,
method, materials)
 Organization
 Tool management
 Process management
 WIP management
 Improve the efficiency of their interaction
 Operating Curve Management (OCM)
 Resources synchronization
29

30. Tools and systems that can help
 Production system (OCM, workflow)
 Automation (factory, equipment, data)
 ISO 9000
30

31. OCM (OPERATING CURVE MANAGEMENT)
31
Derived from
Queuing theory and
Operational Research
in the Industrial
Engineering
(raw process time)

32. OCM
 Use operation techniques to reduce α and increases
capa
32
Wafer fab loading

33. How to improve α
 To reduce α, one must reduce the variability in the line.
 Achieve line balance using line control methodology:
 WIP management
 Accurately forecast WIP for bottleneck equipment
 Establishalert for coming WIP too high or too low
 Using dispatch rules to divert WIP flow
 Priority setting consistent with reducing variability
 Addressing bottleneck issues
 Minimize disturbance
33

34. Cycle time
 Flow factor (FF) is the normalized cycle time
 Too long cycle time is bad:
 Increase cost of WIP
 Decrease quality
 The curve (FF vs. UU) can be represented by
FF=α*[UU/(1-UU)]+1
 Where FF is the flow factor [=CT/RPT] and UU is
the loading factor and α is the manufacturing
variability
34

35. Cycle time breakdown
 Raw processing time
 Queue time (waiting for tools, operator, WIP)
 Hold time (waiting for engineers/ process)
 Transport time
35
Raw Queue time Transport Hold
Process time time time
cycletime
FF = 1 +

36. Synchronize resources
36
4M

37. Production partners
37
All 4 are available

38. Capacity loss due to lacking labor
38

39. Improve efficiency of each partner
 Manpower
 Process
 Equipment
 WIP
39

40. Improve efficiency of human
resources
40
diffusion etch
photo
Thin film
operator
Eq eng
Proc eng
cell
department
A working cell involves operators, process engineers, equipment
engineers, integration. The communication and cooperation among
the working cell members are critical to the performance of the cell

41. Sharing of resources
41
Available
resources
p[production p[R&D p[Engineering
Non available
resources

42. Improve organization efficiency
 Streamline workflow
 Lean organization
 System driven operation
 Management by objectives
 ISO 9000 compliance
42

43. Improve process efficiency
 Process standardization
 Process simplification
 Yield improvement (excursion wafers require large
resources to manage)
 Robust process design
 Unnecessarily tight process spec reduces efficiency
without buying more quality or yield
43

44. Improve tool efficiency
 “Tools” is the most expensive partner
 Two components
 For 40 w/m capacity fab making 32 nm DRAM product:
 Depreciation ~ 60% of wafer cost
 Maintenance ~ 14% of wafer cost
44

45. Equipment cost (12”)
45

46. Economic of scale
When fab capacity is larger, it is more economical per wafer.
46

47. Reason for the economic scale
0
2
4
6
8
10
12
equipment 1 equipment 2 equipment 3 equipment 4
Excess capacity
Identifying and improving the capacity of bottleneck
equipment is a sure way to increase overall fab capacity
Bottleneck equipment
47

48. Factors determining depreciation
per unit output
 Equipment purchasing price
 Capacity balance
 Fab utilization
 Equipment utilization (including uptime and capacity
balance)
 Throughput
48

49. Depreciation bench mark
 The equipment of a fab with 60K/m for 15 nm
technology costs $ 6 B
 6 years (72 months) depreciation
 Depreciation cost per wafer: $6,000,000,000
/72/60,000=$1,389
 The total wafer cost is around $2,315 (assuming
depreciation cost is 60%)
 Wafer selling price $4,000
49

50. Equipment management
 Equipment performance is defined by SEMI standards
 SEMI E10 defines the equipment state of operation
 SEMI E35 defines Cost of Ownership (COO)
 SEMI E79 defines Overall Equipment Efficiency (OEE)
 SEMI E58 defines Automated Reliability, Availability,
Maintainability Standard (ARAMS)
 SEMI E124 defines Factory Level Productivity
50

51. Equipment utilization
51
Maximize
useful
capacity
Minimize
non useful
capacity

52. Two major down times
 Scheduled down
 Make scheduled down effective
 Unscheduled down
 Reduce the frequency of unscheduled down
 Minimize the time spent on unscheduled down
52

53. Scheduled down is the preventive maintenance
 Daily, weekly, monthly, quarterly, yearly
 How long takes to do maintenance
 Effectiveness of scheduled down (If preventive
maintenance is done correctly, there should be no
unscheduled down.)
 Review necessity of parts change
 Bench mark different equipment of the same
equipment type
53

54. Unscheduled down time
 Unscheduled down usually causes product damage. Its loss is
four folds-
 Depreciationand maintenance
 Productscrap, downgrade or rework
 Extra resources to sort out damaged wafers, additional testing
and inspection
 Delayed output
54
Step n Step n+1 Step n+2 Step n+3
Problem
happened
Problem
detected
Damaged
wafers
t2
t1
Metrology step

55. Minimize unscheduled down time
 Unscheduled down maintenance requires two steps:
 Diagnosis
 Repair
 Diagnosis time can be shortened by equipment
automation. EAP extracts all data from equipment in
real time.
 Repair time can be shortened by proper maintenance
kit.
55

56. Monitoring tool performance
 Compile trend charts for MTBF (Mean Time
Between Failure) and MTTR (Mean Time to Repair)
 Compile trend chart data for the total equipment
maintenance cost (parts/ labor for both internal and
outsourcing)
 Add maintenance cost to the equipment cost to
determine the cost of ownership
56

57. Predictive maintenance
 Predictive maintenance is to monitor the equipment
status in real time. Any deviation will send alert to do
potential predictive maintenance.
 Predictive maintenance requires extensive automation
(requires Fault Detection System).
57

58. Overall Equipment Efficiency (OEE)
– SEMI E79
58

59. WIP efficiency: Line balancing
 To avoid capacity loss due to lack of WIP
 According to OCM, too much WIP will slow down
cycletime.
 The ideal WIP should be set up according to the daily going
rate of the equipment
 In average, WIP should not spend more than 1 hour in
waiting before processing
 Balance line has the same WIP turn rate at each process
step
59

60. How to do line balancing
 Line balance improves α
 Line balance is achieved by using effective WIP
dispatching
 Both too much WIP and too little WIP are not
desirable
 WIP arriving rate to an equipment should match its
throughput
 Effective WIP dispatching needs to forecast WIP
distribution at least 3 hours ahead. This is done by
simulation with dispatch rules using the real-time WIP
distribution data
60

61. Simulation
61
dispatch
Simulation

62. Make sure there is steady flow of
WIP from EQ 1 to EQ 2
EQ 1
EQ 2
AMHS
62

63. Wafer fab processes centered
around photomasking
photo
implant
diffusion
deposition
etching
oxidation
63

64. Dynamic capacity bottleneck
 Many events create disturbance in line balance.
 Unscheduled equipment down creates a
temporarily bottleneck.
 This bottleneck is dynamic because it always
changes.
 Many dynamic bottlenecks can happen at the
same time.
 Identifying and resolving the dynamic bottlenecks
are the major task of the wafer fab operation.
64

65. How to find dynamic bottleneck
 Determine WIP turn rate for each equipment or
process step
 Turn rate = total moves in 24 hours/ average WIP
 Rank turn rate
 Process steps with smallest turn rates are dynamic
bottleneck
 Priority of equipment maintenance should be given to
the dynamic bottleneck equipment
65

66. Daily management of WIP
日常在制品管理
66
bottleneck
bottleneck
This should be done by process steps, each equipment, equipment types and
equipment groups and process areas.

67. To improve line balance
 Bottleneck equipment should be given highest priority for
maintenance
 Relationship between line balancing and production
efficiency (α) should be carefully monitored
 Manufacturing system should flash alert for the equipment
type with inventory turn trending down for prolong period
or below certain value
 Using dispatching to avoid sending WIP to the
downstream bottleneck
 Avoid equipment, operator and process recipe dedication if
possible
67

68. Gap Analysis
one shot view of operation status
68
Gap too large
FF too large
efficient production
In-efficient production
cycletime
Unused capacity
target

69. A Gap Trend Chart
69
date

70. Link projects to performance index
Production
smoothness
index
Daily going rate
 Production engineering department is responsible for line balance.
70

71. Loss of productivity due to quality
issue
 Excursion
 Wafer scrap, downgrade
 Yield loss
 New lots need to be issued
 Engineering resources to debug the problem
 Experiment
 Equipment re-qualification
 Process re-qualification
 Quality problem is more than just quality. It is also a
productivity problem.
71

72. Other production improvement
techniques
 Push vs. Pull
 Just-in-time
 Kanban
 Lean manufacturing
 Toyota manufacturing system
72

73. Automation
 Factory automation
 Equipment automation
 Data automation
73

74. Schematics of automation system
74
AHMS: Automatic Materials handling System
MES: Manufacturing Execution System
PDM: Production Database Management
SPC: Statistical Process Control
APC: Advanced Process Control
RMS: Recipe Management System
EAP: Equipment Automation Program
PKD: Process Knowledge Database
EDA: Engineering Data Analysis
PKD
Qua
lity
sys
Customer
Service
sys
EDA
Data
Automation
EQ
automation

75. Advanced Process Control
•Fault Detection System (FDS)
•Statistical Process Control
•Process Recipe Management
•Run to Run Control
•Tool Preventive Maintenance
•Equipmentdown diagnsis
Tool
Equipment
Automation
Interface
Equipment Data Collection
New sensor :
plasma sensor
75

76. Fault Detection System
76

77. Data correlation between tool
signal and metrology data
77

78. Real time process performance
monitoring
Tool Interface
Sensor signals
Filter
Qualified data
Database
Model
Process Performance
Disqualified data
Data adaptation
Discard
N
y
78

79. Yield enhancement system
79
A yield enhancement
system includes:
• data collection
• data analysis
• algorithm to determine
yield loss mechanisms
• corrective actions

80. Conclusion
 There are many theories and applied techniques for
the manufacturing operation.
 These theories and techniques need to be adapted to
fit different situations of each factory.
 Sophisticated automation systems are required to
handle quickly changing production environment.
 Modern wafer fab manufacturing system is a prelude
to the industry 4.0.
80