Magnetic random access memory using a non-linear memory...

Details Magnetic random access memory using a non-linear memory... Magnetic random access memory using a non-linear memory...

2000-04-13

2003-02-04

Dinh, Son T. (Department: 2824)

Static information storage and retrieval

Systems using particular element

Magnetic thin film

C365S145000, C365S171000

Reexamination Certificate

active

06515897

ABSTRACT:

BACKGROUND OF THE INTENTION
1. Field of the Invention
The present invention relates generally to the field of nonvolatile memory devices for use in computers and other devices. More particularly, the present invention relates to nonvolatile memory arrays that use magnetic tunnel junction memory elements as individual memory cells.
2. Description of the Related Art
Certain types of magnetic memory cells that use the magnetic state of a ferromagnetic region for altering the electrical resistance of materials located near the ferromagnetic region are collectively known as magnetoresistive (MR) memory cells. An array of magnetic memory cells is often called a magnetic random access memory (MRAM).
In comparison to metallic MR memory cells, which are based on giant magnetoresistance (GMR) or anisotropic magnetoresistance (AMR) devices, MRAM memory cells are based on magnetic tunnel junction (MTJ) devices and rely on substantially different physical principles. For example, GMR devices include at least two ferromagnetic layers that are separated by a thin metallic layer. In contrast, an MTJ device has two ferromagnetic layers that are separated by a thin insulating tunnel barrier. The magnetoresistance of an MTJ device results from a spin-polarized tunneling of conduction electrons between the two ferromagnetic layers that depends on the relative orientation of the magnetic moments of the two ferromagnetic layers.
FIG. 1A
shows a portion of a conventional MRAM array that uses conventional magnetoresistive memory cells, such as disclosed by U.S. Pat. No. 5,640,343 to Gallagher et al. (the Gallagher ′343 patent). The MRAM array shown in
FIG. 1A
includes a set of electrically conductive traces
1
-
3
in a horizontal plane that function as parallel word lines and a set of electrically conductive traces
4
-
9
in another horizontal plane that function as parallel bit lines. The word lines are oriented in a different direction from the bit lines, preferably at a right angle, so that the two sets of lines intersect when viewed from above. The MRAM array of
FIG. 1A
is referred to as a cross point array because memory cells are placed at the intersection point of crossing lines. The CAM array of
FIG. 1A
is formed on a substrate (not shown), such as a silicon, on which there would be other circuitry (also not shown). For clarity, a layer of insulative material that is located between the bit lines and word lines within the MRAM other than the intersecting regions is not shown. While three word lines and six bit lines are illustrated in
FIG. 1A
, the total number of lines is typically much larger.
A conventional memory cell
10
is located at each crossing point of the word lines and bit lines within an intersection region that is vertically spaced between the respective sets of lines.
FIG. 1B
shows an enlarged view of a conventional magnetoresistive memory cell
10
. Memory cell
10
includes a vertical stack of a diode-like selection device
11
, e.g., a silicon junction diode, connected electrically in series with a magnetic tunnel junction (MTJ) device
12
. Memory cell
10
can be fabricated very densely because the cell has only two terminals and has a vertical current path through selection device
11
and MTJ
12
.
Selection device
11
is a silicon junction diode that is formed from an n-type silicon layer
13
and a p-type silicon layer
14
. The n-type silicon layer
13
is formed on and connected to word line
3
. The p-type silicon layer
14
is connected to the MTJ
12
via a tungsten stud
15
. MTJ
12
is formed from a series of layers of material that are stacked one on top of the other. MTJ
12
includes a template layer
16
, such as Pt, an initial ferromagnetic layer
17
, such as permalloy (Ni—Fe), an antiferromagnetic layer (AF)
18
, such as Mn—Fe, a fixed ferromagnetic layer (FMF)
19
, such as Co—Fe or permalloy, a thin tunneling barrier layer
20
of alumina (Al
2
O
3
), a soft ferromagnetic layer (FMS)
21
, such as a sandwich of thin Co—Fe with permalloy, and a contact layer
22
, such as Pt. Additional details regarding conventional memory cell
10
are provided in U.S. Pat. No. 5,640,343 to Gallagher et al., which is incorporated by reference herein.
Diode
11
is necessary for preventing currents from flowing through alternate current paths, referred to herein as sneak currents. For instance, if in
FIG. 1A
, word line
2
is grounded and a bias of V
A
is applied to bit line
7
, a signal current will flow through the selected cell (in this case the cell connecting word line
2
and bit line
7
, referred to herein as cell
2
,
7
). A very small current will flow through alternate paths, for example, via cells
2
,
6
,
3
,
6
and
3
,
7
because the diode in cell
3
,
6
is reverse biased. The reverse currents for all of the unselected diodes in the memory matrix contribute to the total sneak current. In order to maintain the signal current to be of the same order of magnitude as the total sneak current for an n×n matrix, the rectification ratio of the selected diode must be greater than n
2
. Such a requirement for a diode requires a device-quality semiconducting diode material, which cannot easily be grown on top of a metallic word line, and implies that the diode will have a high resistance.
Additionally, the resistance of the diode, and more particularly the differential resistance of the diode, should be less than the resistance of the MTJ device so that sensing circuitry can easily detect changes in resistance of the MTJ device that represent the different states of the MTJ device. Thus, a large diode resistance implies a correspondingly large MTJ resistance so that MTJ device resistance changes can be easily detected. A high overall diode and MTJ resistance results in slow performance caused by RC delays, in addition to limited power levels during a sensing operation. Thus, a drawback associated with the Gallagher ′343 memory cell is the high resistance of the diode between the two metal layers.
U.S. Pat. No. 5,734,605 to Zhu et al. discloses an alternative conventional MRAM cell that uses a transistor as a selection element for a memory cell. The Zhu et al. MRAM cell uses more space than a cross point cell because the cell has more than two terminals. Moreover, the cell must have a connection from the MTJ to the silicon surface where the transistor is located. The transistor also occupies more area than the tunnel junction occupies.
A very small, 2×2 MTJ element array is disclosed by Z. G. Wang et al. in Spin Tunneling Random Access Memory, IEEE Transactions on Magnetics, Vol. 32 No. 5, 1996, pp. 4022-4024. Although individual cells in the disclosed 2×2 array can be read without diodes, the non-linearity in these standard Al
2
O
3
MTJs is insufficient for permitting selection in arrays that are larger than 2×2 without significantly increasing the time to read a memory cell. Moreover, arrays having a 2×2-array size are too small to be of technological use because the support circuitry would form too large an area on a chip, thereby making the storage density much too small.
To overcome the very small overall MRAM array size of Wang et al., U.S. patent application Ser. No. (Atty Docket No. AM9-99-0112) to Monsma et al. discloses an MTJ cell that uses a tunneling device having a non-linear current dependence through the device as a function of bias voltage. According to Monsma et al., amorphous or polycrystalline barriers are used for forming a non-linear tunnel selection device, such as a metal-barrier-metal (MBM) device, that is fabricated in series with the MTJ to form an MBM-MTJ memory cell. The voltage drop across an unselected MBM is V
mbm,u
=0.5V
A
, with V
A
being the applied bias voltage used for selecting a cell. The voltage drop across a selected MBM is V
mbm
,
s
=V
A
−V
mtj
, with V
mtj
being the voltage drop across the selected MTJ (about 0.2 V). Consequently, the difference in voltage drop across a selected MBM and an unselected MBM is 0.5V
A
−V
mtj
, and not 0

Affiliated with

Also associated with

Rating

Magnetic random access memory using a non-linear memory... does not yet have a rating. At this time, there are no reviews or comments for this patent.

If you have personal experience with Magnetic random access memory using a non-linear memory..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Magnetic random access memory using a non-linear memory... will most certainly appreciate the feedback.

Rate now

Comments { 0 }

Profile ID: LFUS-PAI-O-3179600