role of the hdi in hdd 2 - idema | international disk ... · role of the head disk ... slider-disk...
TRANSCRIPT
Role of the Head Disk Interface in HDD Role of the Head Disk Interface in HDD ReliabilityReliability
George W. TyndallReliability Technology
Samsung Information Systems AmericaSan Jose, CA
IDEMA Reliability Symposium 2
IntroductionIntroductionMotivation: Designing a reliable HDD requires a rigorous understanding of the
most prevalent failure modes/mechanisms, and a means to predict their impact on end-user failure rates.
HDD reliability information can be obtained from a variety of sources End user – Reliability information from customers can be the hardest to obtain, is often times the most
ambiguous, and can be difficult to feedback into product design.Customer qualification – Reliability Demonstration Test (RDT) on well controlled sample distribution to
determine intrinsic reliability of product (AFR% / MTBF) . Sample size can be limited.Internal qualification – Highly Accelerated Stress Testing (HAST) during product development to
investigate design margins. Drives may not be sufficiently mature, but a substantial number are available. Subsystem testing – component-level testing (H/M, platform, firmware, channel) to evaluate product
feasibility and uncover design limitationsFundamental studies – Research and development into the scientific underpinnings of current
technology, and evaluation of new technologies / alternative designs to meet future product specifications
Failure rates from drive-level testing (end-user, customer qualification, and internal qualification) are used to identify the predominant failure mechanisms
Failure of the head-disk interface is found to be the dominant factor impacting HDD reliability
A course-grained model for HDD reliability is developed to address the head-disk interface and the influence of a few vital parameters.
Results of drive level testing are presented to vailidate the modelModel Refinement – results from component level and fundamental studies
IDEMA Reliability Symposium 3
Failure Modes SummaryFailure Modes SummaryHighly Accelerated Stress testing results
Conducted during product development (>10,000 HDD’s)Dominant failure modes (64%) are HDI – related
RDT testing results (~100 HDD’s)
Majority of failures (75%) are HDI – related
Field Returns – Generic failure pareto (1000’s HDD)
Large fraction of “real” HDD failures are HDI-related
64%
18%
18%
HDI-relatedCND
other
Failure Modes
HDI-relatedCND
Other
767K1.14%2F - Disk defects1F - Weak write1F - Head Degradation
848K1.03%1F - CND2F - Disk Scratch1F - Head Degradation
80GB/platter (3.5”)
1.11 M0.79%1F - PCBA2F - Disk Scratch1F - Head Degradation
MTBF(hrs)
AFR(%)Failure modesProduct
6%
28%
49%
11%6%PCBA
CND
TA / scratch
WW
HDI-related
Head Degradation
Sample HDD Failure Mode Pareto
Scratch
TA
Head degradation
Grown defect
Motor
PCBMishandling
CND
High-fly writeWrite abort
HDI-related
35.4% CND
Mishandling
All results point to the head-disk interface as the dominant contributor to HDD reliability
Can we develop a deterministic model to predict these failures?
IDEMA Reliability Symposium 4
Dominant HDI-related failure mechanismsand approximate failure rates
Failure Analysis Failure Analysis –– HDI FailuresHDI Failures
Scratch (50%)
Weak write / modulation (15%)
Head degradation (30%)
6%
28%
49%
11%6%PCBA
CND
Head Degradation
TA / scratch
WW
-2 0 2 4 6 8 10
x 10-5
-0.1
-0.08
-0.06
-0.04
-0.02
0
0.02
0.04
0.06
0.08Read Signal
Time
Vol
tage
Particle-induced scratch (5%)
Head-disk contact (45%)
90 – 95% of scratching results from direct contact between flying head and diskMajority of head degradation failures result from head-disk contacts
100% of all modulation failures result from flying too low - “near”contact
Majority of HDI-related failures result from the head and disk coming into contact, or near contact
IDEMA Reliability Symposium 5
Clearance is key to HDI ReliabilityClearance is key to HDI ReliabilityClearance, space between the slider and disk, is the primary factor for HDI reliability.
Small clearance results in an increased risk of head-disk contact
Zero or negative clearance results in eventual failure due to scratch or head degradationFailure rates increase exponentially with decreasing clearance
Air flow
Lubricant
Slider
Head-diskclearance
slider
+ + +
Magnetic layer
Under layers + substrate
COC layer
(b)
Population FH Distribution
DISK
DISK GLIDE AVALANCHE
Mean Clearance
No head-disk contact
Clearance (Z)
DISK
DISK GLIDE AVALANCHE
∆ Ζ
Head-disk interference-4 -3 -2 -1 0 1 2 3 4
0.0
0.2
0.4
0.6
0.8
1.0
16% Failure Rate
50% Failure Rate
σCum
ulat
ive
Failu
re R
ate
Z-value
Reduced HDI failure rate requires minimizing head-disk contacts by maximizing clearance
IDEMA Reliability Symposium 6
HDI reliability modelHDI reliability modelObjective: Provide reliability design criteria by modeling clearance, and predicting failure rates, under any given set of operating conditions.
AssumptionsFinite clearance 0% failure rateZero clearance 100% failure rateNot a dynamic model, i.e. no explicit time dependence is incorporated into model
Parameters identified as affecting drive clearance (z) include:Incoming clearance distribution (zo)TemperatureAltitude Humidity Duty cycle: Clearance loss during seek (not covered)
Sensitivity of clearance on each parameter measured independently
IDEMA Reliability Symposium 7
4000 m altitude
2
3
4
5
6
7
25 35 45 55 65 75 85
Drive temperature (C)
Cle
aran
ce (n
m)
Temperature sensitivity = -0.035nm/ºC
Drive temperature (ºC)
Cle
aran
ce (n
m)
Engineering Sample Data
HeadHead--Disk Clearance vs. Altitude and TemperatureDisk Clearance vs. Altitude and Temperature
60708090100
0
2
4
6
8
Pressure (kPa)
Cle
aran
ce (n
m)
Touch-down
Altitude sensitivity = 0.13nm/kPa
Disk
Head
Cle
aran
ce d
rop
(nm
)
Clearance also decreases with increasing temperature
Consistent with temperature deratingdiscussed above.
Clearance decreases w/ increasing altitudeLess lift is generated by the ABS as the pressure drops
InitialClearance
IDEMA Reliability Symposium 8
Theoretical model developed to describe the effect of humidity on clearance Compression under ABS can result in local supersaturation of water vaporSupersaturated water vapor is incapable of supporting ABSEffective ABS lift drops – fly height decreases
Clearance is Sensitive to HumidityClearance is Sensitive to Humidity
The head-disk clearance decreases with drive humidity
0.1453
Po (atm)T (ºC)
0.1453
Po (atm)T (ºC)
MeasurementSimulation
-1
0
1
2
3
4
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14Partial pressure of water (atm)
Cle
aran
ce d
rop
(nm
)
For this HDI design, clearance decreases with water vapor pressure as – 0.2 nm/kPa(comparable to altitude effect)
For example, clearance at 60 C drops by 3. 5 nmwhen the humidity is increased from 0 90%RH
(Strom et al., IEEE Trans. Magn. (2007), 43, 3301)
-5
-4
-3
-2
-1
0
0 20 40 60 80 100
Ambient Humidity RH (%)
FH C
hang
e (n
m)
-80
-60
-40
-20
0
20
Pitc
h/R
oll C
hang
es (u
rad)
Gap FH (nm)Pitch (urad)Roll (urad)
ABS Compression SupersaturationFly height drop
IDEMA Reliability Symposium 9
Factors impacting failure rates:Initial clearanceTemperature sensitivityAltitude sensitivityHumidity sensitivity
Data input into Monte Carlo simulations
)()( 0 RHfPbTaznmz −∆−∆−=
Reliability / failure rate modelReliability / failure rate model
Model structure
0
50
100
150
200
250
300
-1.01 -0.03 0.96 1.95 2.94
nm
Freq
uenc
y
0
50
100
150
200
250
0.84 1.71 2.59 3.46 4.33
nm
Freq
uenc
y
Head-disk contact
T=65ºC, RH=30% (Altitude=sea level) T=65ºC, RH=80%
05 0.1 0.15 0.2Altitude sensitivity (nm/kPa)
2 4 6 8
Prob
abili
ty
Incoming clearance (nm) RH sensitivity
5.3 nm
0 0.02 0.04 0.06 0.08Temperature sensitivity (nm/oC)
0.0035nm/C 0.13 nm/kPa clea
ranc
e
No failure 6.2% failure rate
Model Output
RH
The failure predictions were then compared directly with results obtained from standard HDD reliability testing conducted under the specified temperature / altitude / humidity conditions
IDEMA Reliability Symposium 10
)()( 0 RHfPbTaznmClearance −∆−∆−=
4 6 8 10 12 14 16
0
20
40
60
80
100
T = 0 C T = 60C
Failu
re R
ate
(%)
Altitude (ft)
Model Verification Model Verification –– Failure rate Failure rate vsvs Altitude and TemperatureAltitude and Temperature
)()( 0 RHfPbTaznmClearance −∆−∆−=21.2%
63.7%
94.3% 99.5% 100.0%
100.0%100.0%100.0%
66.7%
16.7%0%
20%
40%
60%
80%
100%
120%
8 10 12 14 16Altitude (kft)
Failu
re ra
te (%
)
Simulation
Experiment
3.0%24.4%
68.9%
95.0%99.5%
100.0%
95.2%
66.7%
28.6%
0.0%0%
20%
40%
60%
80%
100%
120%
8 10 12 14 16
Altitude (kft)
Failu
re ra
te (%
)
Simulation
Experiment
Engineering Sample Data
Conditions: Variable altitude test conducted at two temperatures
Quantitative agreement found between model and experiment
Failure rate increases with increasing tempFailure rate increases with increased altitude
Experimental vs Model results
T = 25CT = 60C
∆ T
IDEMA Reliability Symposium 11
Conditions: Fixed Temperature (T = 70oC) / variable humidity ( 5 – 95%)
Model failure rate predictions vs Experimental Data
Excellent agreement observedIndicates validity of approach for clearance / humidity dependence
120 140 160 180 200 220 240Water Vapor Pressure (torr)
Cum
ulat
ive
Failu
re R
ate
0.0
0.2
0.4
0.6
0.8
1.0
Humidity impact on driveHumidity impact on drive reliabilityreliability
Predicted
)()( 0 RHfPbTaznmClearance −∆−∆−=
Relative Humidity Predicted Failure Rate Experiment
data 60 0.0% 0.0%65 6.3% 3.9%70 12.2% 12.0%75 25.5% 30.6%80 66.4% 65.8%90 99.0% 100.0%
IDEMA Reliability Symposium 12
Summary of clearanceSummary of clearance--based reliability modelbased reliability model
A clearance-based reliability model was developed and successfully tested.
The following parameters affecting clearance have been measured and incorporated in the model:
Initial clearance distribution Operating temperatureAltitudeInternal drive humidity (Developed theoretical model of humidity affect on clearance)
On the basis of the excellent correlation with experiment we conclude that. slider-disk clearance should be treated as a critical parameter by which to design a mechanically reliable magnetic hard disk drive.
Cautionary note: The impact of humidity and temperature on drive failure rates in the field cannot be attributed exclusively to the clearance effects discussed above.
IDEMA Reliability Symposium 13
“Fly-height On Demand” (FOD) conceptHead-disk spacing is controlled by resistive heating of the pole tip region of the head
Head-disk Spacing Control
Thermal Expansion
Element spacingNominal FH
FOD Off FOD On
4.5nm11nm
Disk
Thermal actuation is only performed during read and write operations
Since head – disk occur only during FOD actuation, the “effective” duty cycle, and AFR are reduced
Clearance measured by magnetic signal change via the Wallace spacing loss equation
FOD allows for a fourfold decrease in the std. deviation of the clearance.
Current FOD systems actively compensate for changes in drive temp, altitude, and humidity
Future designs promise to supply real time feedback on spacing fluctuations (analogous to position error signal)
4.5nm5.2 nm
FH
02468
10121416
0 20 40 60 80 100
120
140
160
180
200
220
240
FOD [DAC]FH
[nm
]
14IDEMA Reliability Symposium
Does FH control solve all our problems?Finer Details – Glide Avalanche
Physico-chemical properties of disk lubricant Long-term effects of temperatureOther adverse impacts of high humidity
The Future Challenge
The effect of intermolecular forces when macroscopic objects areseparated by nanoscopic distances – van der Waals interaction forces
FH Distribution
DISK
Distribution of DISK GLIDE AVALANCHE
Mean Clearance
4nm
5nm
IDEMA Reliability Symposium 15
Molecularly Molecularly --Thin PFPE Lubricant FilmsThin PFPE Lubricant FilmsEnergetics of molecularly thin PFPE films are dependent on the amount of material present
Surface-induced confinement results in properties that deviate substantially from bulk material. Anomalous materials properties – presence of disjoining pressure unique to molecularly-thin films Tthickness dependent film properties
• Lubricant displacement by flying head• diffusion coefficients (effective viscosity), • Adhesive forces• Friction forces
Oscillatory polar surface energy of PFPE films indicates layering phenomenon
Thermodynamic instabilities developReadily apparent in terraced spreading
Free
Ene
rgy
Film Thickness
1st Monolayer
2nd layer (unstable)
3rd layer(Stable)
(Stable)
Film
Thi
ckne
ss (A
)
0 10 20 30 4015
16
17
18
19
20
21
22
∆γd
Dis
pers
ive
Sur
face
Ene
rgy
(mJ/
m2 )
Lubricant Film Thickness (A)
Dispersive Surface Energy
IDEMA Reliability Symposium 16
Lubricant Stability Lubricant Stability –– AutophobicAutophobic dewettingdewettingOscillatory polar surface energy, χp results in a disjoining pressure ( Π = - dχ/dh) that changes sign as a function of film thickness Thermodynamic stability requires that Π >0When Π < 0, film is unstable and will spontaneously dewet.
Spontaneous dewetting
Π < 0 : Dewetted lubricant droplets
Disjoining Pressure Zdol (MW =4000)
dewetting
26.0 A
30.0 A
OSA ImagesΠ > 0No dewetting
Waltman, Khurshudov, TyndallTrib. Lett. (2002), 12, 163.
IDEMA Reliability Symposium 17
Impact of Impact of dewettingdewetting
9.6 A
9.9 A
10.3 A
Zdol 1350
Dewetting of Zdol (MW = 1350)Homogenous films observed when film
thickness, h, < 10 Lubricant dewetting occurs at h > 10
Clearance measurementsAcoustic emission results
0 5 10 15 20 25-5
-4
-3
-2
-1
0
1
Cle
aran
ce C
hang
e (n
m)
Lubricant Thickness (A)
Clearance severely reduced when lubricant droplets formed
Lubricant Lubricant dewettingdewetting produces produces ““soft particlessoft particles”” that can interact with the that can interact with the flying sliderflying slider
∆Z > 4nm
IDEMA Reliability Symposium 18
Lubricant Stability Lubricant Stability -- DewettingDewettingPFPE lubricant layering on surfaces is ubiquitous
Other surfaces
0
60
40
20
1.0 0.8 -0.20.6 0.4 0.2 0
Zdol 2800 on Cu
Thic
knes
s (A
)
Distance [mm]
Impact of PFPE Molecular Weight
Dew
ettin
gTh
ickn
ess (
A)Effect of carbon composition
0 1000 2000 3000 4000 500005
101520253035404550
OSATitrationDisjoining Press.
Mon
olay
er T
hick
ness
(A)
Zdol Molecular Weight (Mn)
gTerraced Spreading on Cu Fractal dewetting on Si
IDEMA Reliability Symposium 19
Moisture Moisture ––related failures modesrelated failures modesHigh moisture levels inside drive interior can adversely impact reliabilityFailure mechanism #1: Corrosion (well-known phenomenon)
Failure mechanism #2: Clearance loss due to supersaturation under flying head (included in failure model discussed above)Failure mechanism #3: Condensation of long-lived water microdroplets on both heads and disks
Droplets are abnormally long-livedDroplets reduce clearance / increase head-disk interactionDecreased reliability
45 min After Exposure
160 min After Exposure
6 hr After Exposure
21 hr After Exposure
45 hr After Exposure
4 days After Exposure
11 days After Exposure
51 days After Exposure
IDEMA Reliability Symposium 20
Schematics of Lubricant BondingSchematics of Lubricant Bonding
Carbon over coat
Free-Lube
Bonded-Lube
Z-DOL
-OH functionality = Lower affinity for disk Di-hydroxyl = Higher affinity for disk
Z-Tetraol
PFPE lubricants are available with a variety of backbone structures, molecular weights and functional end-groups.
Lubricants with functionalized end-groups will spontaneously bond to the disk.
Extent of lubricant bonding is dictated by the lube, the temperature and RH.
10000 20000 30000 40000 50000 600000.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
80% Humidity
40% Humidity
Zdol 4000
Bon
ded
Frac
tion
time (min)0 20000 40000 60000 8000
0.2
0.4
0.6
0.8
1.0Z-Tetraol
Bon
ded
Frac
tion
Time (min)
Z-Tetraol 80% RH
0 20000 40000 60000 800000.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
Bond
ed F
ract
ion
Time (min)
IDEMA Reliability Symposium 21
Lubricant bonding Lubricant bonding Exposure to elevated temperature induces a
(reversible) bonding of the lubricant to the disk.
Bonding adversely impacts lubricant mobility and contact forces 1 10 100 1000 10000
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
k(t) = kBt-0.5
∆ Bo
nded
Fra
ctio
n
Time (min)
T = 50C
20% Bonded Lube 50% Bonded Lube 90% Bonded Lube
Adhesion hysteresis increases with bonded fractionForce required to separate the head and disk surfaces
increases as the lube bonding ratio increases.Severity of contact (time in contact) increases as bonded
fraction increasesHDI failure rate will scale with the adhesion hysteresis
0 20 40 60 80 1000.0
0.2
0.4
0.6
0.8
1.0
Total Zdol Thickness = 10.5 +/- 0.5 A
Rel
ativ
e C
SS D
urab
ility
% Bonded Zdol
Adhesion hysteresis
IDEMA Reliability Symposium 22
Phase transitions in PFPE lubricantsPhase transitions in PFPE lubricants
0 20 40 60 80 100 120
78
80
82
84
86
88
90
"liquid-like""glass-like"
Phas
e sh
ift, d
egre
es
Temperature, °C
1 10 100 1000 10000
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
k(t) = kBt-0.5
(T = 90C)k(t) = kBt
-0.5
(T = 50C)∆ Bo
nded
Fra
ctio
n
Time (min)
PPFE bonding kinetics are non-classicalTime dependent rate coefficients1D diffusion – limited reactionFunctional form of rate coefficients can change
as a function of temperatureResults indicative of a phase transition
Shear modulated AFM measurements confirm a transition from solid-like to liquid-like behavior
Transition temperature is dependent on the lubricant molecular weight, PFPE structure and nature of carbon surface
Impact of this change in lubricant mobility on HDI reliability has not been addressed
23IDEMA Reliability Symposium
Future (current) Challenge Future (current) Challenge -- Intermolecular ForcesIntermolecular Forces
0 2 4 6 8 1 0 1 2
0
5
1 0
1 5
2 0
V D W A t t ra c t iv e F o rc e s
(A = 3 .6 x 1 0 -1 9 J )
8 A Z d o l 4 0 0 0 /C N x
∆ Fr
eque
ncy
d ( n m )
Quartz Resonator (f = 10 - 30 MHz)v = 0.5 - 1.5 m/sA = 20 nm
Si3N4R = 2.3 mm
PFPE + carbon
Z Resolution = 0.1 nm
d
Force Transducer (FN)
It is well known that attractive, intermolecular forces are generated when macroscopic bodies are separated by nanoscopic distances (d < 10 – 20nm). For van der Waals forces, the attractive force is dependent on three factors:
g = Geometric factor defining the ‘effective’ interaction areaA = defines interaction strength between materials on the two surfaces
We operate the head-disk interface in a regime where these forces are not insignificant, and will only get more prominent in future products
2dgAF ∝
Experimental
IDEMA Reliability Symposium 24
Impact of VDW forces at the headImpact of VDW forces at the head--disk interfacedisk interface
Slider
Carbon OvercoatPFPE
ΠFN d
5 10 15 20 25 30 35-20
-15
-10
-5
0
5
10
15
20
unperturbed
Zdol 4000/ CNxImpact of macroscopic body at distance d
d = 8nm
d = 6nm
d = 4nm
d = 7nm
d = 5 nm
Πef
f (at
m)
Lubricant Thickness ( )
The attractive VDW forces that develop at the HDI reduces the magnitude of the disjoining pressure resulting in a vertical expansion of the lubricant film
Slider
Carbon OvercoatPFPE
Lube Jum p
Slider
Carbon OvercoatPFPE
ApproachAt sufficiently close distances, the lube will “jump” from the disk to the head.
This phenomenon will limit the minimum separation distance, and thereby contribute to the glide avalanche.
25IDEMA Reliability Symposium
Effect of Molecular Weight on Glide AvalancheEffect of Molecular Weight on Glide AvalancheFr
ictio
n, g
Linear Velocity, m/sec
13.6 7.310.5
12 A Zdol 5391
12 A Zdol 3216
10 A Zdol 1350
h
Lower MW Zdol Higher MW Zdol
slider
disk
Clearance Measurements vs Lubricant Molecular WeightA) Reduced rotational velocity B) Reduced pressure
Both techniques show clearance decreases with increasing PFPE molecular weight
(Khurshudov and Waltman, Trib. Lett. (2001), 11, 143)
∆Z ~ 2nm
26IDEMA Reliability Symposium
Effect of film thickness on Slider disk interactions
Disjoining pressure characterizing film adhesion decreases rapidly with increasing film thickness
Thicker films should jump to contact sooner than thinner filmsExperimental results confirm this expectation
These results also indicate that Intermolecular forces are currently limiting glide avalanche.
Nor
mal
For
ce (a
rb.u
nits
)
0
10
20
Zdol 4000/CNx
7A
13A
35A
0 2 4 6 8 10 12
Physical Separation (nm)
contact
28A
27IDEMA Reliability Symposium
Intermolecular Forces: AdhesionIntermolecular Forces: AdhesionPrevious results suggest that decreased lubricant film thickness would be
beneficial from a clearance perspectiveThe Adhesive Force generated between the disk and slider (stickiness of interface), increases with decreasing film thickness, however
Increased adhesive (and frictional) force at low film thickness will increase the severity of any head-disk contact
0 5 10 15 20 25 30 350.0
0.2
0.4
0.6
0.8
1.0
1.2F ad
h / g
(arb
. uni
ts)
∆γd + ∆γp (mJ-m-2)
Thick lube (20A)
Thin lube (7.5A)
)(4 hRFadh γπ ∆=
28IDEMA Reliability Symposium
PrePre--contact Frictional Forcescontact Frictional ForcesPre – contact Frictional forces are generated at low clearances in films of
thicknesses used at the HDI
CNx/Zdol 4000
file:CNxqcm1
35A
11A
0 1 2 3 4 5 6 7 8 9 10
15A
0.0
0.5
1.0
1.5
2.0
Nor
mal
ized
Fric
tion
Forc
e (a
rb. u
nits
)
Separation Distance (nm)
28A
Physical contactFriction detected
d = 1.5nm
Friction starts at nominally 1.5nm prior to establishment of physical contact
Friction signal shows displays a maximum value prior to contact
Magnitude of friction maximum scales inversely with film thickness
The origin of these phenomenon are not well understood
This could also contribute to defining our minimum separation distance.
Friction maximum
29IDEMA Reliability Symposium
SummarySummaryHDD failure rates can be dominated by the HDIHDI failure rates depend strongly on clearance
Mean clearanceStd. deviation of clearanceSensitivity to environmental parameters
Interface materials (lube + carbon) dictate HDI performance and can impact clearance
Lube thickness, type and molecular weightCarbon type and thicknessTemperature and humidity
Physics underlying many of the interface materials properties is still not understood completely
Intermolecular forces (both adhesive and friction) will fundamentally limit the lowest head-disk clearance that can be obtained (2 – 3nm)