This page covers the degradation of ferroelectric random access memory: commercial PZT (lead zirconate titanate) FeRAM with one-transistor one-capacitor or two-transistor two-capacitor cells, and the emerging ferroelectric hafnium-zirconium oxide (HZO) capacitors. It covers retention loss by thermal depolarization, imprint, fatigue, the endurance budget of the destructive read, field-limited breakdown in HfO2 films, and environmental and radiation limits. The cell physics is on the Ferroelectric Random Access Memory (FeRAM) device page, the ionizing-dose models on the Radiation page and the moisture models on the Environmental and Chemical page; ferroelectric FETs and tunnel junctions are research devices and are not covered.

1. Why It Happens

Trigger: Storage and operation at high temperature, which depolarizes the film and builds imprint (TI; Cypress); switching cycles, which every read adds because the read is destructive (TI; Chae et al.); high cycling fields in thin HfO2 films (Lyu et al.); humidity, bias and ionizing dose (Sharma et al.).
Mechanism, depolarization and retention: The remanent polarization falls with temperature and vanishes near the Curie temperature, about 430 $^{\circ} C$ for PZT; below it the loss is reversible and is restored by the next write-back (TI).
Mechanism, imprint: A state held for long at high temperature becomes preferred, and the opposite state becomes harder to write and read; oxygen vacancies diffuse to the electrode interface and screen the polarization, which offsets the loop (TI; Miyazawa et al.). Imprint is permanent (TI).
Mechanism, fatigue: Repeated switching pins domains at defects; in Pt/PZT/Pt capacitors oxygen and Pb vacancies near the bottom electrode form fatigue centres (Chae et al.).
Mechanism, hydrogen: Hydrogen from processing or plastic packages reduces PZT at the Pt interface and degrades the hysteresis, so hydrogen barrier films are needed (Sharma et al.; patent 9842662).
What sets the rate: The temperature, with an apparent activation energy of about 1.4 $eV$ for retention (Cypress table, computed); the number of switchings, not their frequency (Chae et al.); the pulse asymmetry (Chae et al.); the field in HfO2 films (Lyu et al.).
Failure criterion: Data retention below the specified years at temperature (Cypress); a read margin of the ‘1’ state below the reference (patent 9842662); a switchable polarization too small to sense, or a hard breakdown (Lyu et al.).

2. Where to Look

Location: The bottom electrode interface of PZT capacitors, where vacancies concentrate (Chae et al.); the electrode interfaces where imprint charge collects (Miyazawa et al.); the hydrogen barrier around the capacitor (patent 9842662).
Signature: A hysteresis loop shifted along the voltage axis (imprint) (Lyu et al.); a smaller switched charge after many cycles (fatigue) (Chae et al.); read failures after HAST or HTOL that disappear on rewriting, which are retention failures, not hard failures (Sharma et al.).
Confirm with: Hysteresis and switched-charge measurements after cycling and baking (Chae et al.; Lyu et al.); a screening bake followed by a hot margin test with a relaxation pause (patent 9842662); retention bakes at several temperatures (Cypress; TI).
Look-alikes: Reversible depolarization, which disappears after a write-back, against permanent imprint (TI); a failure with a higher supply current after ionizing dose, which is radiation damage rather than wear-out (Sharma et al.).

3. Prevention

Design: Use a 2T-2C cell for automotive temperature ranges (Everspin); count every access, read or write, against the endurance, and spread accesses over rows (TI; Cypress).
Material and process: Use hydrogen barrier films around the capacitor and keep them uniform (patent 9842662; Sharma et al.); control oxygen vacancies at the electrodes (Miyazawa et al.; Chae et al.); for HfO2, keep the cycling field low, since breakdown cycles fall steeply with field (Lyu et al.).
Operation: Stay within -40 to +85 $^{\circ} C$ for the rated ten years (Cypress; TI); program after reflow where the data sheet requires it (TI); refresh fatigued PZT with a field above the coercive field where the design allows it (Chae et al.).
Screening and test: Qualify retention with a 1000 $h$ bake at 125 $^{\circ} C$ (TI); screen imprint-weak cells with a bake above 85 $^{\circ} C$ and a hot margin test (patent 9842662); test after cycling preconditioning (Sharma et al.).

Retention and Imprint

Name	Formula	Related Formulae	Variables/Constants
Retention against temperature from the data-sheet table	$t_{D R} (T) = t_{D R} (T_{0}) exp [\frac{E _{a}}{k _{B}} (\frac{1}{T} - \frac{1}{T _{0}})]$	$E_{a} = \frac{k _{B} l n ( t _{D R, 2} / t _{D R, 1} )}{1/ T _{2} - 1/ T _{1}}$	$t_{D R} : minim u m d a t a r e t e n t i o n$ $T an d T_{0} : ab so l u t e ambi e n t t e m p er a t u r es$ $E_{a} : a pp a r e n t a c t i v a t i o n e n er g y$ $k_{B} : B o l t z mann co n s t an t (8.617 \times 1 0^{- 5} eV / K)$ $t_{D R, 1} an d t_{D R, 2} : r e t e n t i o n a t T_{1} an d T_{2}; C y p r ess C Y 15 B 101 N an d F M 24 V 10 : 10 y e a r s a t 85^{\circ} C, 38 y e a r s a t 75^{\circ} C an d 151 y e a r s a t 65^{\circ} C, 1 \times 1 0^{14} cy c l es o v er t h e o p er a t in g t e m p er a t u r e, - 40 t o + 85^{\circ} C; T I : t e n y e a r s a t 85^{\circ} C; E v er s p in : F R A M r e t e n t i o n d e g r a d es ab o v e 85^{\circ} C an d a u t o m o t i v e g r a d e F R A M n ee d s a 2 T - 2 C ce l l t o co m p e n s a t e f or t h e w e ak er p o l a r i z a t i o n a t hi g h t e m p er a t u r e$ $(co m p u t e d : E_{a} = 1.43 eV f r o m 85 t o 75^{\circ} C, 1.40 eV f r o m 75 t o 65^{\circ} C an d 1.42 eV f r o m 85 t o 65^{\circ} C; e x t r a p o l a t e d w i t h 1.42 eV, 0.88 y e a r s a t 105^{\circ} C an d 0.10 y e a r s (ab o u t f i v e w ee k s) a t 125^{\circ} C, o u t s i d e t h e r a t e d r an g e (m y e x t r a p o l a t i o n))$
Equivalent time of a retention qualification bake	$t_{u se} = t_{bak e} exp [\frac{E _{a}}{k _{B}} (\frac{1}{T _{u se}} - \frac{1}{T _{bak e}})]$	$T_{C} \approx 43 0^{\circ} C (PZT, complete depolarization)$	$t_{u se} : e q u i v a l e n t t im e a t t h e u se t e m p er a t u r e T_{u se}$ $t_{bak e} : bak e t im e a t T_{bak e}$ $E_{a} : a c t i v a t i o n e n er g y$ $k_{B} : B o l t z mann co n s t an t (8.617 \times 1 0^{- 5} eV / K)$ $T_{C} : t r an s i t i o n (C u r i e) t e m p er a t u r e; T I : q u a l i f i c a t i o n i s a c u m u l a t i v e bak e o f 1000 h a t 125^{\circ} C, w i t h an im p r in t bak e a t 125^{\circ} C an d a d e p o l a r i z a t i o n bak e a t 85^{\circ} C, t h er ma l d e p o l a r i z a t i o n b e l o w T_{C} i s n o t p er man e n t an d i s r es t or e d b y t h e n e x t r e a d an d w r i t e - ba c k, i t i s in s i g ni f i c an t f r o m - 40 t o + 85^{\circ} C, an d t h e M S P 430 F R 57 xx m u s t b e p r o g r amm e d o n l y a f t er r e f l o w b ec a u se r e t e n t i o n i s n o t e n s u r e d ab o v e T_{s t g} = 95^{\circ} C$ $(co m p u t e d : w i t h E_{a} = 1.42 eV f r o m t h e C y p r ess t ab l e, 1000 h a t 125^{\circ} C e q u a l s 11.5 y e a r s a t 85^{\circ} C, co n s i s t e n t w i t h t h e t e n - y e a r s p ec i f i c a t i o n (m y c h ec k))$
Imprint offset from interface charge	$V_{o f f} = \frac{σ _{in t} δ}{ε _{0} ε _{i}}$	$E_{im p} = \frac{V _{c +} + V _{c -}}{2 t _{f}}$	$V_{o f f} : s hi f t o f t h e h y s t er es i s l oo p a l o n g t h e v o l t a g e a x i s$ $σ_{in t} : c ha r g e d e n s i t y a t t h e e l ec t r o d e in t er f a ce$ $δ : t hi c k n ess o f t h e in t er f a ce (d e a d) l a y er$ $ε_{0} : v a c uu m p er mi tt i v i t y$ $ε_{i} : i t s r e l a t i v e p er mi tt i v i t y$ $E_{im p} : im p r in t f i e l d$ $V_{c +} an d V_{c -} : p os i t i v e an d n e g a t i v e coer c i v e v o l t a g es (s i g n e d)$ $t_{f} : f i l m t hi c k n ess (in t er f a ce - l a y er scr ee nin g m o d e l, ba c k g r o u n d k n o w l e d g e); M i y a z a w a e t a l . : im p r in t in P Z T i s e x pl ain e d b y o x y g e n v a c an c i es t ha t d i f f u se t o t h e in t er f a ce an d scr ee n t h e p o l a r i z a t i o n, w i t h t h e d i f f u s i o n co n s t an t f r o m m o l ec u l a r d y nami cs; T I : im p r in t s t r e n g t h e n s t h e s t or e d s t a t e an d w e ak e n s t h e o pp os i t e o n e, i s p er man e n t an d occ u r s a t l o w er t e m p er a t u r e t han d e p o l a r i z a t i o n; p a t e n t 9842662 : im p r in t a pp e a r s a s a s ma l l er r e a d ma r g in o f t h e hi g h er - c a p a c i t an ce^{'} 1^{'} s t a t e; L y u e t a l . (H Z O o n S i) : l oo p s hi f t s o f - 0.08 V a t 4.6 nm an d ab o u t - 0.36 V a t 6.9 nm, an d t h e im p r in t f i e l d f a l l s f r o m ab o u t 300 t o 100 kV/cm w i t h cy c l in g$ $(co m p u t e d : - 0.36 V o v er 6.9 nm i s 0.52 MV/cm an d - 0.08 V o v er 4.6 nm 0.17 MV/cm a s a v er a g e f i e l d s)$
Screening of weak cells by a bake and a hot margin test	$V_{r e f, 0} = V (B F C = 0.1) from a fit to the lowest fail counts$	$Δ V_{ma r g in} = V_{r e f, 0} - V_{r e f, s p ec}$	$V_{r e f, 0} : r e f er e n ce v o l t a g e a t w hi c h n o bi t f ai l s$ $B F C : bi t - f ai l co u n t$ $Δ V_{ma r g in} : t es t ma r g in t o t h e s p ec i f i e d r e f er e n ce; p a t e n t 9842662 : o n e h o u r a t 85^{\circ} C d oes n o t d e p o l a r i z e s u s p ec t ce l l s e n o ug h, t h e scr ee n bak e i s ab o v e 85^{\circ} C f or a t l e a s t ab o u t 20 min u t es (e x am pl e o n e h o u r a t 155^{\circ} C f or P Z T), t h e seco n d t es t i s ab o v e ab o u t 60^{\circ} C (e x am pl e 85^{\circ} C) a f t er a r e l a x a t i o n p a u se o f a t l e a s t ab o u t 10 s, r e f er e n ce s t e p s a r e ab o u t 5 mV, t h e f i t u ses t h e f o u r l o w es t f ai l co u n t s, t h e ma r g in i s o f t h e or d er o f a f e w mV, an d n o n - u ni f or m h y d r o g e n ba r r i er f i l m s a r e a so u r ce o f v a r ia t i o n$ $(co m p u t e d : w i t h E_{a} = 1.42 eV, o n e h o u r a t 155^{\circ} C e q u a l s ab o u t 1800 h, or 76 d a y s, a t 85^{\circ} C (m y es t ima t e), 1800 t im es t h e 1 h a t 85^{\circ} C t ha t t h e p a t e n t f in d s in s u f f i c i e n t)$

Fatigue and Endurance

Name	Formula	Related Formulae	Variables/Constants
Endurance budget of a destructive-read memory	$t_{l i f e} = \frac{N _{e n d}}{f _{a cc}}$	$N_{a cc} = N_{r e a d} + N_{w r i t e} (every read is followed by a write-back)$	$t_{l i f e} : t im e t o r e a c h t h e r a t e d cy c l es$ $N_{e n d} : r a t e d e n d u r an ce$ $f_{a cc} : a ccess r a t e o f t h e m os t u se d l oc a t i o n$ $N_{a cc} : a ccesses t ha t co u n t t o w a r d e n d u r an ce$ $N_{r e a d} an d N_{w r i t e} : r e a d s an d w r i t es; T I (M S P 430) : 1 \times 1 0^{15} cy c l es p er ce l l, f l a s h t y p i c a l l y 1 \times 1 0^{5}, r e a d s an d w r i t es co u n t t h e s am e, a l oc a t i o n a ccesse d 0.88 \times 1 0^{6} t im es p er seco n d r e a c h es 1 \times 1 0^{15} in ab o u t 36 y e a r s, an d n o f ai l u r e ha d b ee n f o u n d in co n t in u o u s cy c l in g; C y p r ess : 1 \times 1 0^{14} r e a d / w r i t es o v er t h e o p er a t in g t e m p er a t u r e, co u n t e d p er r o w so t ha t a r o w a ccess co u n t s o n ce w ha t e v er t h e n u mb er o f w or d s u se d; E v er s p in : F R A M 1 \times 1 0^{14} cy c l es w i t h k n o w n w e a r - o u t m ec hani s m s, an d t h e d es t r u c t i v e r e a d n ee d s a r es t or e$ $(co m p u t e d : 36.0 y e a r s a t 1 \times 1 0^{15} an d 3.6 y e a r s a t 1 \times 1 0^{14} f or 0.88 \times 1 0^{6} a ccesses p er seco n d; 3.2 y e a r s a t 1 \times 1 0^{14} f or 1 \times 1 0^{6} a ccesses p er seco n d)$
Fatigue of the switchable polarization	$P_{s w} (N) \approx P_{s w, 0} (N < N_{o n se t}), P_{s w} ↓ with N (N > N_{o n se t})$	$A = \frac{V _{+}^{p} - V _{-}^{p}}{V _{+}^{p} + V _{-}^{p}}, E_{c} = \frac{V _{c}}{t _{f}}$	$P_{s w} : s w i t c hab l e p o l a r i z a t i o n$ $P_{s w, 0} : ini t ia l v a l u e$ $N : s w i t c hin g cy c l es$ $N_{o n se t} : cy c l es a t t h e o n se t o f f a t i g u e$ $A : p u l se a sy mm e t r y, V_{+}^{p} an d V_{-}^{p} : p os i t i v e an d n e g a t i v e p e ak v o l t a g es$ $E_{c} : coer c i v e f i e l d$ $V_{c} : coer c i v e v o l t a g e$ $t_{f} : f i l m t hi c k n ess; C ha e e t a l . (so l - g e l P b (Z r 0.52 T i 0.48) O 3, 300 nm, P t e l ec t r o d es, 32.7 μ C/c m^{2} ini t ia l) : f a t i g u e s t a r t s a f t er ab o u t 1 \times 1 0^{4} cy c l es f or 3 t o 7 V p e ak s, n e a r l y in d e p e n d e n t o f t h e v o l t a g e b u t f a s t er a t hi g h er v o l t a g e, an d d e p e n d s o n t h e n u mb er o f s w i t c hin g s r a t h er t han t h e f r e q u e n cy; - 10 V o n t h e t o p e l ec t r o d e r eco v er s ab o u t 70 p er ce n t o f t h e p o l a r i z a t i o n in 1 s an d n e a r l y a l l in 100 s, + 10 V g i v es l i ttl e r eco v er y, r eco v er e d f i l m s r e - f a t i g u e ab o u t t e n t im es f a s t er, an d f or A < - 0.3 t h er e i s l i ttl e f a t i g u e a f t er 4 \times 1 0^{9} cy c l es w hi l e t h e coer c i v e f i e l d r i ses f r o m 52 t o 70 kV/cm a t A = - 0.4; m ec hani s m p r o p ose d : o x y g e n an d P b v a c an c i es n e a r t h e b o tt o m e l ec t r o d e p in d o main s$ $(co m p u t e d : 1.5 V o v er 300 nm i s 50 kV/cm, c l ose t o t h e r e p or t e d 52 kV/cm; t h e \pm 5 V t es t p u l ses a r e 167 kV/cm, 3.2 t im es E_{c})$
Field-limited endurance of ferroelectric HfO2	$E_{a pp} = \frac{V}{t _{f}}, N_{B D} falls steeply with E_{a pp}$	$E_{c} \propto t_{f}^{- 2/3}$	$E_{a pp} : a ppl i e d f i e l d$ $V : cy c l in g v o l t a g e$ $t_{f} : f i l m t hi c k n ess$ $N_{B D} : cy c l es t o ha r d b r e ak d o w n$ $E_{c} : coer c i v e f i e l d (t hi c k n ess sc a l in g r e p or t e d b y L y u e t a l .); L y u e t a l . (e p i t a x ia l H f 0.5 Z r 0.5 O 2 o n S i, 4.6 t o 18.4 nm) : t h e 4.6 nm f i l m k ee p s 2 P_{r} ab o v e 6 μ C/c m^{2} a f t er 1 \times 1 0^{11} cy c l es a t 2.5 V w i t h o u t b r e ak d o w n, b r e ak s d o w n a f t er 1 \times 1 0^{10} cy c l es a t 3.0 V an d 1 \times 1 0^{9} a t 3.5 V, l e ak a g e r i ses a f t er ab o u t 1 \times 1 0^{5} cy c l es a t 3.5 V an d 1 \times 1 0^{7} a t 3 V, a 9.7 nm f i l m b r e ak s d o w n a f t er 5 \times 1 0^{8} cy c l es a t 4.5 V an d 1 \times 1 0^{5} a t 5 V, t h e 18.4 nm f i l m f a l l s b e l o w 2 μ C/c m^{2} a f t er 1 \times 1 0^{6} cy c l es a t 6 V, w ak e - u p i s ab se n t ab o v e 4.6 nm, an d t h e e x t r a p o l a t e d 2 P_{r} a f t er 10 y e a r s e x cee d s 12 μ C/c m^{2}$ $(co m p u t e d : 2.5, 3.0 an d 3.5 V o n 4.6 nm a r e 5.4, 6.5 an d 7.6 MV/cm, so 2.1 MV/cm m or e f i e l d cos t s ab o u t tw o d ec a d es o f cy c l es t o b r e ak d o w n; 2 P_{r} f a l l s f r o m ab o u t 14 t o 6.6 μ C/c m^{2}, t o 47 p er ce n t, a f t er 1 \times 1 0^{11} cy c l es a t 2.5 V)$

Environmental and Radiation Limits

Name	Formula	Related Formulae	Variables/Constants
Environmental and radiation limits of PZT FeRAM	$D_{f ai l} \approx 30 to 40 krad (Si), T_{f ai l} \approx 20 0^{\circ} C (500 h)$	$P_{r} (T) \propto (T_{C} - T)^{1/2} (T < T_{C})$	$D_{f ai l} : t o t a l i o ni z in g d ose a t w hi c h f ai l u r es a pp e a r$ $T_{f ai l} : s t or a g e t e m p er a t u r e a t w hi c h f ai l u r es a pp e a r e d$ $P_{r} : r e man e n t p o l a r i z a t i o n$ $T : t e m p er a t u r e$ $T_{C} : C u r i e t e m p er a t u r e (m e an - f i e l d f or m f or a seco n d - or d er t r an s i t i o n, ba c k g r o u n d k n o w l e d g e; P Z T i s c l ose t o f i r s t or d er, so i t i s a g u i d e o n l y); S ha r ma e t a l . (co mm er c ia l F R A M s, 10000 r e w r i t e cy c l es a s p r eco n d i t i o nin g) : n o r e t e n t i o n or im p r in t f ai l u r es in 1000 cy c l es o f - 65 t o 150^{\circ} C, tw o ha r d f ai l u r es a t 400 h o f H A S T (130^{\circ} C, 85 p er ce n t R H), o n e a t 500 h o f H T O L (125^{\circ} C, 5.5 V), man y r e a d f ai l u r es t ha t co u l d b e r e w r i tt e n (r e t e n t i o n f ai l u r es), n o f ai l u r es a f t er a g in g a t 150 an d 175^{\circ} C b u t f ai l u r es a f t er 500 h a t 200^{\circ} C, f u n c t i o na l t o 10 krad, hi g h er s u ppl y c u r r e n t s a f t er 30 krad an d se v er e f ai l u r es a t 40 krad a t 83.3 rad/s, l i ttl e r eco v er y a f t er o n e d a y o f ann e a l in g, an d l i t er a t u r e o n h e a v y - i o n l a t c h - u p; ba c k g r o u n d m ec hani s m s c i t e d t h er e : t h er ma l d e p o l a r i z a t i o n, h y d r o g e n f r o m pl a s t i c p a c k a g es r e d u c in g P Z T a t t h e P t in t er f a ce, an d o x y g e n v a c an c i es t ha t l oc k d o main s; T I : s t a t i c so f t er r or r a t e b e l o w 0.051 F I T / M b$ $(co m p u t e d : t h e 40 krad f ai l u r e l e v e l i s f o u r t im es t h e 10 krad l e v e l t ha t p a sse d; T_{C} = 430^{\circ} C (T I) i s 345 K ab o v e 85^{\circ} C)$

Important Questions

Ques: Is FeRAM endurance $1 \times 1 0^{14}$ or $1 \times 1 0^{15}$ cycles?

The sources give both. TI rates the MSP430 FRAM cells at $1 \times 1 0^{15}$ cycles and states that no failure had been found in continuous cycling, while Cypress rates its F-RAM chips at $1 \times 1 0^{14}$ read/writes over the operating temperature, and Everspin quotes $1 \times 1 0^{14}$ with known wear-out mechanisms. The values are specifications for different products and test conditions rather than measured limits, so neither is a physical constant (my reading). For a design that does not use one of the TI parts, $1 \times 1 0^{14}$ is the more likely and the safer value. The difference matters for the most used location: at $0.88 \times 1 0^{6}$ accesses per second, as in the TI example, $1 \times 1 0^{15}$ lasts 36 years and $1 \times 1 0^{14}$ only 3.6 years (computed), since every read is followed by a write-back. Reference: Texas Instruments, ‘MSP430 FRAM Quality and Reliability’, Application Report SLAA526A (March 2012, revised May 2014) (fetch-tool summary); Cypress Semiconductor (Infineon), ‘CY15B101N 1-Mbit (64K x 16) Automotive F-RAM’ data sheet, Rev. *F (December 2018), and ‘FM24V10 1-Mbit Serial (I2C) F-RAM’ data sheet, Rev. *J (December 2018) (fetch-tool summaries); C. Bohac, Everspin Technologies, ‘Comparing Technologies: MRAM vs. FRAM’, Application Note 02130 (March 2013) (fetch-tool summary).

Ques: What activation energy does the FeRAM retention table imply, and what does it predict at 125 degrees Celsius?

The Cypress data sheets give a minimum of 10 years at 85 $^{\circ} C$ , 38 years at 75 $^{\circ} C$ and 151 years at 65 $^{\circ} C$ . Fitting an Arrhenius law to these points gives 1.43 $eV$ between 85 and 75 $^{\circ} C$ and 1.40 $eV$ between 75 and 65 $^{\circ} C$ , so the table is consistent with a single activation energy of about 1.42 $eV$ (computed). Extrapolated, this gives 0.88 years at 105 $^{\circ} C$ and about five weeks at 125 $^{\circ} C$ (my extrapolation), which is why the parts are rated only to 85 $^{\circ} C$ and why Everspin states that FRAM retention degrades above 85 $^{\circ} C$ . The same energy turns the 1000 $h$ bake at 125 $^{\circ} C$ of the TI qualification into 11.5 years at 85 $^{\circ} C$ , consistent with the ten-year specification. Reference: Cypress Semiconductor (Infineon), ‘CY15B101N 1-Mbit (64K x 16) Automotive F-RAM’ data sheet, Rev. *F (December 2018), and ‘FM24V10 1-Mbit Serial (I2C) F-RAM’ data sheet, Rev. *J (December 2018) (fetch-tool summaries); Texas Instruments, ‘MSP430 FRAM Quality and Reliability’, Application Report SLAA526A (March 2012, revised May 2014) (fetch-tool summary); C. Bohac, Everspin Technologies, ‘Comparing Technologies: MRAM vs. FRAM’, Application Note 02130 (March 2013) (fetch-tool summary).

Ques: How is imprint different from thermal depolarization?

Depolarization is a loss of polarization at high temperature that is not permanent: below the transition temperature, about 430 $^{\circ} C$ for PZT, the cell regains its polarization at the next read and write-back, and from -40 to +85 $^{\circ} C$ it is insignificant (TI). Imprint is a permanent preference for the stored state that makes the opposite state harder to write and read; it occurs at lower temperatures than depolarization, for example in long bakes (TI). Miyazawa et al. explain it by oxygen vacancies that diffuse to the electrode interface and screen the polarization, which shifts the loop along the voltage axis. Patent 9842662 notes that it shows up as a smaller read margin of the ‘1’ state, and that one hour at 85 $^{\circ} C$ is too weak to reveal weak cells. Reference: Texas Instruments, ‘MSP430 FRAM Quality and Reliability’, Application Report SLAA526A (March 2012, revised May 2014) (fetch-tool summary); H. Miyazawa, Y. Hamada, T. Higuchi, E. Natori, T. Kijima, T. Shimoda, M. Yoshiya and T. Oguchi, ‘A new model of imprint mechanism in ferroelectric memory’, MRS Online Proceedings Library 902, 0902-T10-43 (2005) (abstract only); US Patent 9842662, ‘Screening for data retention loss in ferroelectric memories’ (assignee and year not in the text read) (fetch-tool summary).

Ques: Why does a PZT capacitor fatigue, and can it be recovered?

Chae et al. cycled 300 $nm$ sol-gel PZT capacitors with Pt electrodes and found that fatigue begins after about $1 \times 1 0^{4}$ cycles for peak voltages of 3 to 7 $V$ , progresses faster at higher voltage and depends on the number of switchings rather than the frequency. They attribute it to oxygen and Pb vacancies near the bottom electrode that pin domains. A dc voltage of -10 $V$ on the top electrode recovered about 70 per cent of the polarization in 1 $s$ and nearly all of it in 100 $s$ , while +10 $V$ did little, and recovery needs a field above the coercive field. Recovered films fatigue again about ten times faster, so the defects remain. Asymmetric pulses with $A < - 0.3$ showed little fatigue after $4 \times 1 0^{9}$ cycles. Reference: B. G. Chae, C. H. Park, Y. S. Yang and M. S. Jang, ‘Asymmetry in fatigue and recovery in ferroelectric Pb(Zr,Ti)O3 thin-film capacitors’, Applied Physics Letters 75, 2135 (1999) (arXiv:cond-mat/0210690, fetch-tool summary).

Ques: Do reads wear out FeRAM?

Yes, because the read is destructive. Reading a ferroelectric capacitor switches it if it holds the opposite state, so every read is followed by an automatic write-back, and TI states that read and write endurance are therefore the same and any access counts. TI also notes that the write-back is protected against power loss by isolating the FRAM supply and storing enough charge to finish it. Cypress counts endurance per row: one access to a row counts once whether one or all of its words are used. A cache that serves repeated reads removes them from the count, and in TI’s example a fully cached loop causes no FRAM accesses. Everspin adds that the read signal margin shrinks over repeated reads, while MRAM reads do not disturb the state. Reference: Texas Instruments, ‘MSP430 FRAM Quality and Reliability’, Application Report SLAA526A (March 2012, revised May 2014) (fetch-tool summary); Cypress Semiconductor (Infineon), ‘CY15B101N 1-Mbit (64K x 16) Automotive F-RAM’ data sheet, Rev. *F (December 2018), and ‘FM24V10 1-Mbit Serial (I2C) F-RAM’ data sheet, Rev. *J (December 2018) (fetch-tool summaries); C. Bohac, Everspin Technologies, ‘Comparing Technologies: MRAM vs. FRAM’, Application Note 02130 (March 2013) (fetch-tool summary).

Ques: What limits the endurance of ferroelectric HfO2 films?

Mainly the field and dielectric breakdown, rather than slow fatigue. Lyu et al. cycled epitaxial HZO films of 4.6 to 18.4 $nm$ on silicon. The 4.6 $nm$ film kept 2Pr above 6 $μ C/c m^{2}$ after $1 \times 1 0^{11}$ cycles at 2.5 $V$ with no breakdown, but broke down after $1 \times 1 0^{10}$ cycles at 3.0 $V$ and $1 \times 1 0^{9}$ at 3.5 $V$ . These are fields of 5.4, 6.5 and 7.6 $MV/cm$ , so about 2.1 $MV/cm$ more field cost about two decades of cycles (computed). Leakage rose after $1 \times 1 0^{5}$ to $1 \times 1 0^{7}$ cycles before breakdown. Thicker films fatigued more: the 18.4 $nm$ film fell below 2 $μ C/c m^{2}$ after $1 \times 1 0^{6}$ cycles. The thinnest film also showed no wake-up and an extrapolated 2Pr above 12 $μ C/c m^{2}$ after ten years, which is why sub-5 $nm$ films are attractive. Reference: J. Lyu, T. Song, I. Fina and F. Sanchez, ‘High polarization, endurance and retention in sub-5 nm Hf0.5Zr0.5O2 films’, Nanoscale 12, 11280-11287 (2020) (arXiv:2005.14477, fetch-tool summary).

Ques: How robust are commercial FeRAMs to humidity, heat and radiation?

Sharma et al. stressed commercial FRAMs after 10 000 rewrite cycles. Temperature cycling from -65 to 150 $^{\circ} C$ for 1000 cycles caused no retention or imprint failures, and aging at 150 and 175 $^{\circ} C$ none, but failures appeared after 500 $h$ at 200 $^{\circ} C$ . HAST at 130 $^{\circ} C$ and 85 per cent RH gave two hard failures at 400 $h$ , HTOL at 125 $^{\circ} C$ and 5.5 $V$ one at 500 $h$ , and many read failures that could be rewritten, which they class as retention failures. The parts stayed functional to 10 $krad$ , drew more supply current after 30 $krad$ and failed severely at 40 $krad$ , with little recovery after annealing. The literature they cite adds heavy-ion latch-up and hydrogen damage from plastic packages. Reference: A. K. Sharma, A. Teverovsky, T. W. Dowdy and B. Hamilton, ‘Evaluation of Data Retention and Imprint Characteristics of FRAMs Under Environmental Stresses for NASA Applications’, NASA GSFC and NAVSEA Crane (NTRS 20040053358; year not stated in the text read) (fetch-tool summary); Radiation page of this vault; Environmental and Chemical page of this vault.

Quartz 5

Explorer

Ferroelectric Random Access Memory (FeRAM, PZT and HfO2)

1. Why It Happens

2. Where to Look

3. Prevention

Retention and Imprint

Fatigue and Endurance

Environmental and Radiation Limits

Important Questions

Ques: Is FeRAM endurance $1 \times 1 0^{14}$ or $1 \times 1 0^{15}$ cycles?

Ques: What activation energy does the FeRAM retention table imply, and what does it predict at 125 degrees Celsius?

Ques: How is imprint different from thermal depolarization?

Ques: Why does a PZT capacitor fatigue, and can it be recovered?

Ques: Do reads wear out FeRAM?

Ques: What limits the endurance of ferroelectric HfO2 films?

Ques: How robust are commercial FeRAMs to humidity, heat and radiation?

Graph View

Table of Contents

Backlinks

Quartz 5

Explorer

Ferroelectric Random Access Memory (FeRAM, PZT and HfO2)

1. Why It Happens

2. Where to Look

3. Prevention

Retention and Imprint

Fatigue and Endurance

Environmental and Radiation Limits

Important Questions

Ques: Is FeRAM endurance 1×1014 or 1×1015 cycles?

Ques: What activation energy does the FeRAM retention table imply, and what does it predict at 125 degrees Celsius?

Ques: How is imprint different from thermal depolarization?

Ques: Why does a PZT capacitor fatigue, and can it be recovered?

Ques: Do reads wear out FeRAM?

Ques: What limits the endurance of ferroelectric HfO2 films?

Ques: How robust are commercial FeRAMs to humidity, heat and radiation?

Graph View

Table of Contents

Backlinks

Ques: Is FeRAM endurance $1 \times 1 0^{14}$ or $1 \times 1 0^{15}$ cycles?