<%BANNER%>

Transport properties at nano scales via first principles studies

University of Florida Institutional Repository
xml version 1.0 encoding UTF-8
REPORT xmlns http:www.fcla.edudlsmddaitss xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.fcla.edudlsmddaitssdaitssReport.xsd
INGEST IEID E20110115_AAAADZ INGEST_TIME 2011-01-15T22:37:34Z PACKAGE UFE0008621_00001
AGREEMENT_INFO ACCOUNT UF PROJECT UFDC
FILES
FILE SIZE 27029 DFID F20110115_AACSYH ORIGIN DEPOSITOR PATH zhang_c_Page_027thm.jpg GLOBAL false PRESERVATION BIT MESSAGE_DIGEST ALGORITHM MD5
9f72f8fe551428df244fffd41647baac
SHA-1
9e279a2720eaf0df2d5d609e97849f433ec89dc9
77293 F20110115_AACTCZ zhang_c_Page_102.QC.jpg
6898628288086f265f94bd978f7db6c0
2645141fab7fbeb9cafa8a8a3b5e37916d90ed0c
21510 F20110115_AACSXS zhang_c_Page_019thm.jpg
fd32d012a7839441743465a2f5822642
2af2687cbd3d3306deed33ca4975454fc7d05c35
62137 F20110115_AACSYI zhang_c_Page_027.QC.jpg
cce9e982f484a4280fe93a69ab87692c
ff31b30c0062466fd957d8fe37b91dbf0bd8df57
66921 F20110115_AACSXT zhang_c_Page_019.QC.jpg
4471b9acfad063b86c29708242b61131
23e724d023c1450c539142d28cb193bc2db69de7
1053954 F20110115_AACSBA zhang_c_Page_078.tif
0a05adfef7396a79f02b2cea0f52acf9
b360b6fa9fe95589f9b28a327171deabab66313a
20029 F20110115_AACSYJ zhang_c_Page_028thm.jpg
27b564714dfff67a5610e35fe7a5a5aa
ce9634607c0d7e66ec5dfe83aa6473a487d0cbd6
21844 F20110115_AACSXU zhang_c_Page_020thm.jpg
4e37a858725fa6e0fcb69c9b97ea136e
b32c83656fa70553562fcd9561f2a259b3176d25
25477 F20110115_AACSBB zhang_c_Page_059thm.jpg
a81fc7102d8680a04d63ca14ccdcf5b9
33cce38eed0c8318356299bd55927416a4927c3f
57329 F20110115_AACSYK zhang_c_Page_028.QC.jpg
a78ba3956433a99561025075b15f0b92
e1a86d7920840be439c2fcf271ab861c7fb5d808
65703 F20110115_AACSXV zhang_c_Page_020.QC.jpg
9f0d707b33c79eb50e8ba6d251efd0cf
e775f6f2a5048c8d9045b5cb89e39838a0c4889e
43372 F20110115_AACSBC zhang_c_Page_023.pro
3f5bae20e48eafcd112879b57c1d92fa
3c017aadf2ee66c52a721f5fa0a0f5fc0e0efd30
17825 F20110115_AACSYL zhang_c_Page_030thm.jpg
bf48de0cd1cb1af657915fec56bcdd61
f75a0c9fffa9b11f8d2447ad3ffb097a2a979ee9
23243 F20110115_AACSXW zhang_c_Page_021thm.jpg
31783fe9f9b7ec5f3c2cda5000eadaca
def6f4f7e59c7e1f968a6c2869fb00124796c6a9
65332 F20110115_AACSBD zhang_c_Page_048.QC.jpg
2e09b6da93c590bf5ccf0535ab6be998
4d8f73211f32dde1b354214700120a7529d6fc0f
52523 F20110115_AACSYM zhang_c_Page_030.QC.jpg
796adb2f29468966c37dfd5cb88a2769
d17d8de66217a792f788c3af964f290582222b60
73281 F20110115_AACSXX zhang_c_Page_021.QC.jpg
3c96559aef58300c8d7943acdc6928aa
cec36520240ea8f88ab5f70c4c7c49821655c4c5
16017 F20110115_AACSBE zhang_c_Page_091.pro
9d6c3afdeb6a075d17eb3613266eb977
2e29cfec35797852d1c8e0de578e9ed4b0df66e9
71574 F20110115_AACSZA zhang_c_Page_041.QC.jpg
7e1fa1cd432070410d4efc1315739147
17b2e026e3eac81110528909463af2baccbec286
19081 F20110115_AACSYN zhang_c_Page_031thm.jpg
ea2594152bd4fe80299d7fdb56f349d8
d8f431d36e7eb21b73f2a01a5804b478a351938d
20801 F20110115_AACSXY zhang_c_Page_022thm.jpg
2ed938ef54f3e29fba0b42eb8d8474d2
4f3f98a7f70008b6ce4de041d0a96a32528b44a9
175232 F20110115_AACSBF zhang_c_Page_054.jpg
37183649412e45a871df0ce1647e3a09
613e041209eaa7d648132ec6e9b3e85d740724de
25066 F20110115_AACSZB zhang_c_Page_042thm.jpg
27e476f3fc45ec3282cee48a952a0b9a
ec798b4b869f3cc78d3890475f805001cb5ccfcb
64309 F20110115_AACSYO zhang_c_Page_031.QC.jpg
4e203485528ce7f983b0480ef44fc5ab
2f2b8cc6c74c9790fa94411ac8d42d78df75b2ad
63040 F20110115_AACSXZ zhang_c_Page_022.QC.jpg
1248c3d0f3ff7cebdc2a60527e257ef1
f71c6078774a40c50aa3ac80864ee805c8dd5d42
32721 F20110115_AACSBG zhang_c_Page_101.pro
2c393a77b375cae466ed8d8d57ad2426
71feda090069104afbd63fb128ab3e8838d735e4
80971 F20110115_AACSZC zhang_c_Page_042.QC.jpg
c213a4040fee6eed842710b428750816
a53ca7a09265dadabfcc7c1878ce4a071049ebc7
25098 F20110115_AACSYP zhang_c_Page_032thm.jpg
0f07fac765a143a6244246e3beb46c86
692e0f802f80b633701e218e4d2d413876b0ecb5
F20110115_AACSBH zhang_c_Page_016.tif
2ce055ea7442916ade2f73d6657a0234
7768a07f1ac885d52dc73f59044f40d88030b1b6
25185 F20110115_AACSZD zhang_c_Page_043thm.jpg
4325f2a0a998078eca42c05d2a644c06
060846fcc41c59e6069a2e1523bd5a06b8160735
79848 F20110115_AACSYQ zhang_c_Page_032.QC.jpg
24d66436d3c5c7dbf466498efef156ef
fa67249acd07262c0ff143be45fd91fc6be3af4a
1836 F20110115_AACSBI zhang_c_Page_012.txt
2de7921bc5ea6b32930588e67ada0fbb
ac6bf46703b4c5655f816c20503680e8303b8c57
77016 F20110115_AACSZE zhang_c_Page_043.QC.jpg
49e36ec64facdb0fbf0b2e67360a3a5b
a2dcf44909a11cc90439922b4a5d250b692ce0db
60800 F20110115_AACSYR zhang_c_Page_033.QC.jpg
05022597a490e2c794dd0618f02e9b82
25e085ec4729e25649ac50111ead3ed94e12dda1
1768 F20110115_AACSBJ zhang_c_Page_075.txt
16f7664902bfdc0c0806532919f376da
f68fd513dd2168caccf121ca2009822f250a9057
F20110115_AACSAV zhang_c_Page_015.tif
8fdb3fc7392f2a7ec2ffa862444f2dca
39448abb377a34d47c7d986a4e5fab04cfd9dfc3
24378 F20110115_AACSZF zhang_c_Page_044thm.jpg
76ac06360527d3cafad28d773ee7256e
836b08b1991f38cb01f597b1996e31364e3b7059
157330 F20110115_AACSBK zhang_c_Page_066.jpg
b05e9737701c03354f1d353b30a695aa
dd57bf35681e038a13b636b9d7659192e3c9a72c
19010 F20110115_AACSAW zhang_c_Page_055thm.jpg
57823c0145690acea1aff5f04868458e
774668c0b6404b4315e15d94f318d46e3cabb287
78324 F20110115_AACSZG zhang_c_Page_044.QC.jpg
6408da623e4edddd3290bd17cc448f35
e8e7d7b63c40f65b9f665661982b391a80229329
47329 F20110115_AACSYS zhang_c_Page_035thm.jpg
51979bae39e698b09528e038eb29e34f
2f996efabacb45229d6a3b91dbd41863773ee4b2
68458 F20110115_AACSBL zhang_c_Page_075.QC.jpg
7d3cfd2ff6ba73ebdf706d869713df9b
c82d3fffc382a06355725fbfc4acd31021ea1dc3
F20110115_AACSAX zhang_c_Page_100.tif
3e8e436cbcfe8d0980ebe43a0d71362a
2954d19e04b33978565601c2f23c2db6347f6c3a
20289 F20110115_AACSZH zhang_c_Page_045thm.jpg
745b3dd645e29967106d0b15454921f9
e0ec0a27fe250da318a072539fa7e7e63260dd69
45128 F20110115_AACSYT zhang_c_Page_036thm.jpg
a81eaa546620416dd0a6967229d7bdea
2d0a54be8711a5b5ea241b6b08836f29c83ccd37
1602 F20110115_AACSCA zhang_c_Page_037.txt
88e2e92b463a46c2cb42497c27b4b6b5
49aaece22fb38b6c69cb3c3c5994088be33ec0bc
45610 F20110115_AACSBM zhang_c_Page_041.pro
90591529373ea0efe617d31f65003662
ecb8e1f0a98267d34e20353d21f426a0ea16ef2d
63799 F20110115_AACSAY zhang_c_Page_079.QC.jpg
8fb9ec7f1f8250ee9a849d3854da8925
cb4c97fd5b8485857457cfd73f28fe2736758b6d
61163 F20110115_AACSZI zhang_c_Page_045.QC.jpg
8788a915e7abaa6b9f09d427ed559582
f86ef9caa2fa1d75421e7370e19ac85092998503
87422 F20110115_AACSYU zhang_c_Page_036.QC.jpg
37dca9c75f93134acf3cddca91c919fa
9b9e38542b58efc007073437360b99d3cffab3d8
F20110115_AACSCB zhang_c_Page_101.tif
f2bf0ba176a4c653ec1c4e32e2a5408d
5efcd2a7d95386374fe7a8dfb83d7bdc77a9a645
888 F20110115_AACSBN zhang_c_Page_061.txt
224b387d862967ef6ee8575516b50d88
19107bdf0165807230b42b22a79e6cc1075daa42
50341 F20110115_AACSAZ zhang_c_Page_102.pro
73a2daf8ab447512d2d13bc8a569f770
c2e7fd0ca7ad7b7eee1f2a76895817e7640c2a3a
19569 F20110115_AACSZJ zhang_c_Page_046thm.jpg
419c296d345b9261cb1462c1667d7c37
64a93ab16dc32e8eb8f08bac4ed41088d3d832d2
45255 F20110115_AACSYV zhang_c_Page_037thm.jpg
b1857572d47ef3e70e388f9efaceddb1
3c184ae3b651a7cf1f492e5a0967607a1250fc26
F20110115_AACSCC zhang_c_Page_068.tif
623b46d94bb58fb94b1cfa1cc563bbc0
431c628fc2ca8c88ce5a13fb4e6a3436ec6c8d32
212044 F20110115_AACSBO zhang_c_Page_102.jpg
80c8fa95fe77b5ee8c5b94df2f830a0d
fda54d2462571f699ac2662726dad9ec7323ee7b
59420 F20110115_AACSZK zhang_c_Page_046.QC.jpg
35a80f3707ff945d0413aef2cfcd4332
6f3cda2953bc485904e845d1fa1fc35af415e807
44167 F20110115_AACSYW zhang_c_Page_038thm.jpg
0ca47f04af58eea4f875a0eb82265e80
ab54109485949d0e6e6938bb3d958a0e2b0ecc7d
F20110115_AACSCD zhang_c_Page_041.tif
c618e7148804b0cbd7dcc0f4e9873ee3
31c71767617ba4d5368643aa6385374e28dbcdbd
F20110115_AACSBP zhang_c_Page_095.tif
a8ee81bf6b36bcba9363a3bb0286d3f2
82b574f3ad10e26bbe5916301d9d47d5fe7ff32f
24283 F20110115_AACSZL zhang_c_Page_047thm.jpg
8091980905deb646906a8f50d6b73c69
c8518bf5056887dc4a4050a98668b0ddb701dd6b
78278 F20110115_AACSYX zhang_c_Page_038.QC.jpg
9ff74c9b965e2db266db7a42c5273b4f
3f84c16f2e62cb12f0dfbea8b2c4fcd0f34bccc4
F20110115_AACSCE zhang_c_Page_102.tif
ed8cb555e5c4b653728f78615802d31e
b001cf4a71c0c653ae41b876404f824aa92adab2
25280 F20110115_AACSBQ zhang_c_Page_096.pro
6bd0de204a489f51508927c1f5b4fd75
fa9357c83ef3d58468d2e5208818d812bba2a5fa
68324 F20110115_AACSZM zhang_c_Page_047.QC.jpg
cd02a6316a99d99e2cb45e687b32b8df
c88e65aedec259cede74fe3c086c8b17f0cced2e
59253 F20110115_AACSYY zhang_c_Page_039.QC.jpg
4ceb2682d7884de29d8dc0301266b61d
75800d4852164bb97e94d1c7b31acfdad4cf5418
78428 F20110115_AACSCF zhang_c_Page_028.jp2
99d004c21551a4ef91beb48270bf5585
9c60da7adde3b4f1c7cd1d5602c96e3f09354052
74671 F20110115_AACSBR zhang_c_Page_082.jp2
fc5382fcdad27933f368c529abd042b0
25f1e671357b8e9fc822787a25f59de413508e66
21858 F20110115_AACSZN zhang_c_Page_048thm.jpg
4864e1570471270f4d6a3f04045874df
0a820133a42aced4e4405006847bb4fff32b40f2
22529 F20110115_AACSYZ zhang_c_Page_041thm.jpg
b67d06da94e36fe8efec55f7a905f369
f9596b567fc2e790fb350c910368fb37baf87fa3
25271604 F20110115_AACSCG zhang_c_Page_039.tif
c48d4b770b869da2924074167b2fbc69
3247fe4f9e059387e4081a528c2c0dfab462bbb0
1659 F20110115_AACSBS zhang_c_Page_022.txt
7e4050156d2b4f49b3dad7dcf903bbdc
64f301d93a77f2b56ec8b6c169391e24238094e4
18894 F20110115_AACSZO zhang_c_Page_049thm.jpg
993163d105e08966a0a16c02b2c34140
e1650515012e0751f0543f2dbb209f65141c7300
92686 F20110115_AACSCH zhang_c_Page_013.QC.jpg
0a56defe616c74324924f9fe9102ee5f
eae8161cf69cb17a23900a34b891a6f5fca23014
17634 F20110115_AACSBT zhang_c_Page_052thm.jpg
c214e35719746a3c1adb6de73617c489
9568f853c801a164fdc4bd1e6eb5234622f8fe37
52389 F20110115_AACSZP zhang_c_Page_049.QC.jpg
65fe0df266e784e9c847a30d7e3d77fb
57a264c97ba394b75e08eb734b46fe9959571450
17546 F20110115_AACSCI zhang_c_Page_004thm.jpg
0f4bc5f07b43e257ec7e95d623b3aad7
ab9fb450c7c18037201fe5162dbd788462c7a437
536903 F20110115_AACSBU zhang_c_Page_039.jp2
445777c36d91d01a4ceef474b4264958
476eb65f46e0a5e667a35c1706f73f716758d9be
53702 F20110115_AACSZQ zhang_c_Page_050.QC.jpg
91353ae0caf9d87d0728f4776c8ae7fd
f5b6b47b5d522103bab9d5a3f6ed80e9b1cb2d26
1771 F20110115_AACSCJ zhang_c_Page_074.txt
7c8335a13d1dcbd736b914e2745d4264
2ed748891558f101673d26d70d4e1a22307f6381
1791 F20110115_AACSBV zhang_c_Page_080.txt
7919e828324fe5636f3c95287fff4de7
eca5c25ccd156bedb20daa7f95c9e0c4b41431a1
23382 F20110115_AACSZR zhang_c_Page_051thm.jpg
caacaaf8bb45ada9bfa0af3f30a8520c
a0c2a7374e2597dc8f53b8be64c9a42bb11ce61d
67352 F20110115_AACSCK zhang_c_Page_089.jp2
e068450f7d04787673089abe721fe201
8d9d0260910852e708091d55a24ae1a5b1c55090
33430 F20110115_AACSBW zhang_c_Page_017.pro
4d3787dcf736ce0c4bca5a195e3ebe02
8a2cdb6c9c7e3cb89331cab5f59e330bd120eda4
71342 F20110115_AACSZS zhang_c_Page_051.QC.jpg
b7e6969cfad0e3e720a1a8d72167e6c0
9fd25c041476c418b8c6c5a291b77f4c2283ce41
138435 F20110115_AACSCL zhang_c_Page_018.jpg
2979df3fef8e73cbfb0f8ccbe95ecb16
59927db455de154562afcb4ca61e5d9c3e2e4b2e
166035 F20110115_AACSBX zhang_c_Page_088.jpg
541b722fca09e070d04661f353d21c96
61b134f454e52eca002d8df0599d79621d23b36e
69355 F20110115_AACSCM zhang_c_Page_004.jp2
8612ea9752027e4f3f118771636bddb1
f4e1ca3366300acf20788d67b8fe55a101fd7160
56579 F20110115_AACSBY zhang_c_Page_008thm.jpg
9f796a226f5cda9183db283c82d62c52
6dc34ef79d544a2c3dcd21acf8dd38233091e194
56678 F20110115_AACSZT zhang_c_Page_052.QC.jpg
91742d1f20d65ea336dbfc283027fedc
fb48fe296489909f46504c772ed911a3a8623699
79910 F20110115_AACSDA zhang_c_Page_059.QC.jpg
d9fced56ce31f6283b37d2a7aa281326
209ec76725f8dd1008aa3fc36da0d083adad792c
21236 F20110115_AACSCN zhang_c_Page_092.pro
046993ccfe075d57d956a684a402f2e1
d827aa038b619ba2a2b7e4ba3e07afac282b8d43
79881 F20110115_AACSBZ zhang_c_Page_029.jp2
d364614c093b5df6521581c090630efa
3703b226c46b20fada929dc52b43e4bf428132f1
19988 F20110115_AACSZU zhang_c_Page_053thm.jpg
8f1a4cc1a753b797a779dcb8a2a95cf5
837f821d7c7920cc1a4f050619dd8164eb8a5790
20206 F20110115_AACSDB zhang_c_Page_033thm.jpg
e0f84c5d30a962505a9a062a62d2a47c
a9bcc1839225fd1320953619d23123493ad38092
F20110115_AACSCO zhang_c_Page_006.tif
e366c80020d74d1cbc981455fa303a9c
227001a85461299d77d722ad122b722adb822358
65917 F20110115_AACSZV zhang_c_Page_053.QC.jpg
f5b6680b5342e6464f56dab6a4861fb5
dffabcde634b8748e840a0e33fb83742e7de23bd
166296 F20110115_AACSDC zhang_c_Page_038.jpg
307d85735b91e5459706970f5af1b80d
201a5210d66d3f56d34fb26fc1a93fce71df5d88
34939 F20110115_AACSCP zhang_c_Page_072.pro
b53bc7782a9fe5b909856c795c949bab
9664881ae7dd388ec79d018ebfb06cc140c17003
21372 F20110115_AACSZW zhang_c_Page_054thm.jpg
e40b6eba26e69bc63b8ca02d64b6f61c
7b865396eade76eff442a89a8f3331e70dafae0b
5079 F20110115_AACSDD zhang_c_Page_098.pro
0ad758e8e9a7cce160500cbb7b01dbaf
dcbcc073105a4d3fb54dc942fc599b82a8a395fb
168796 F20110115_AACSCQ zhang_c_Page_020.jpg
777a7e2047b1d17069caeb4dcfac20b9
290af3690f44da0d528206e980530858eb71865a
64894 F20110115_AACSZX zhang_c_Page_054.QC.jpg
4057cd252388c479a3e0860100fb98a8
7fb1f1348c9882f96dfdd42bea5cc0f66e78f10e
1561 F20110115_AACSDE zhang_c_Page_090.txt
c792b6c911989d4b69a860e93e672579
784347d12593a4afdf3cb892d8024a88df2ab700
55619 F20110115_AACSCR zhang_c_Page_096.jp2
c3bfd7145ca45e75788a8988efebe907
3320250e43efda9ebb9ffda1d4959f32177408cb
58952 F20110115_AACSZY zhang_c_Page_055.QC.jpg
73712a3a808584af0374366bf9ca1321
45d3eba109bbe905cc0677f2bf0cbc7b702caac8
57689 F20110115_AACSDF zhang_c_Page_081.QC.jpg
f9048ee5c3a01da7356e9d0337e4bc8c
bc0b6b54c605275dc6bd679d6931fcc968174434
25574 F20110115_AACSCS zhang_c_Page_034thm.jpg
1cb12f5cb6b74964cefd1c1a90443c12
6abb11addf761664b66bc84bfe82a51127f9cc05
25706 F20110115_AACSZZ zhang_c_Page_056thm.jpg
1235aa6d224c877fbd0fc3886eb2b1de
5bc712ccd5816cd132cf9c956137629652a5ec94
112251 F20110115_AACSDG zhang_c_Page_096.jpg
405e9d12c30080c8865f89a149573491
98b8076ce43296ff85dba2ba79322bfbacbc01b8
F20110115_AACSCT zhang_c_Page_051.tif
a693314c945c9267da340957c68b3c45
1821a3b1f6ba74eb5fb838be7040f1a9a88e8d2e
2677 F20110115_AACSDH zhang_c_Page_003thm.jpg
c2b7a36d40fd664289072b9fa091d9aa
a4ab66a01ae368605050ddd166b49259f78d1aa2
97967 F20110115_AACSCU zhang_c_Page_103.jp2
f1b9c4225808a0397fa84f3059734be0
d3cea69a0688ca633bdf3cbe66f5a4e19a344517
19921 F20110115_AACSDI zhang_c_Page_029thm.jpg
e584490cddae7ecdc58585eea474eeb1
0dee15d52061372c685924dca42d08537ada447a
1826 F20110115_AACSCV zhang_c_Page_014.txt
cd35764b882ff2b47f2f66c6d4e59ec2
60e5c635cc6f5138bed4f432950db7beb3bde482
220005 F20110115_AACSDJ zhang_c_Page_034.jpg
464430ac4f5c079642feffd76964743c
3ad33345e28533dc8e2ece68d30b5b50a4723805
F20110115_AACSCW zhang_c_Page_037.tif
2b8174f092ba3a6ffb29f4abe67d5afd
2779faee486d9e9e184f3a13a11fcfac4a4bef16
15036 F20110115_AACSDK zhang_c_Page_073thm.jpg
84aa3371d88b592e39c151503da3b92f
9076205d7601270789edfbcd7838950b3394de2f
F20110115_AACSCX zhang_c_Page_054.tif
98bd1fa9b71fbfcfb325b1245c16f96c
193bb309f334f2717242db9ea922c48296cb9567
F20110115_AACSDL zhang_c_Page_022.tif
30a6ff81f7c0aaef376b0e89591362a9
55c7b93b530d4f32d5570c377be5f5479a7efad5
1619 F20110115_AACSCY zhang_c_Page_030.txt
4a128f5089abaac8914055c9ea19d086
3132407cb49d3ef7f5f957e7a4899bbcacdc8d9a
4650 F20110115_AACSEA zhang_c_Page_040.pro
e30ee28dde596cd6841a72263800aa64
5aeee9b247f7e4e3218b8629e9b4f3658ef60876
2070 F20110115_AACSDM zhang_c_Page_071.txt
1c71edf903e6f067d98ba1c20da9fd1f
742d28c7a4cc6f02a926d24e877f424c66a26ceb
30296 F20110115_AACSCZ zhang_c_Page_015.pro
4a05994c257cef194293c83967c18c21
4af71aff872aaa436ae4fabc1d6f4759327858b3
90870 F20110115_AACSEB zhang_c_Page_035.QC.jpg
6f495530549ed6acb7ec2452068ce756
e0fe419d89a95d3251571aa5d786b2cfa7e73f93
42305 F20110115_AACSDN zhang_c_Page_074.pro
9717f0e044a6903af6b81938fdfa50a2
ae55623f71d7c5399b24f3f436b1506873f0e425
79193 F20110115_AACSEC zhang_c_Page_063.QC.jpg
067779a4dee2a0732b8b228d283eb7d5
85371ab2dcebc3102c060904ff227622c3fae0b2
38475 F20110115_AACSDO zhang_c_Page_070.QC.jpg
558b7c17deb4b3a63fb24d978f291217
828d1a6b1e2f0be112a7e397a492d5783bae7ff7
82029 F20110115_AACSED zhang_c_Page_034.QC.jpg
1927b528e4967a9a79fd144ead9d8651
98e481de83d8aa519f8b4bc2a14c36e0d8409aca
89499 F20110115_AACSDP zhang_c_Page_037.QC.jpg
e69975d7b59645e1bc706799cd410369
e0da5ec200ca363f035ea24957240ebeb7c96db1
26619 F20110115_AACSEE zhang_c_Page_095.QC.jpg
d8b1905729769c0512538d418ff509ba
9407fd889d1ca67a4a0909edcf75a51bc9f59cd7
51528 F20110115_AACSDQ zhang_c_Page_042.pro
43eebf00eb94338ab8b3a63482bb1889
5b555f28e95f30e8c4954cd1e73b0228b16e3f09
1492 F20110115_AACSEF zhang_c_Page_038.txt
d71a2c76adf9c3b37060989628b34e1e
8af62ef1709972aa64ba8660273b6175deb722fe
12222 F20110115_AACSDR zhang_c_Page_040.QC.jpg
0ee2c88a9298ea5dedf7815fbecedeed
622ecc9097f278cea8f6cb5c5faabb7c71c99781
34803 F20110115_AACSEG zhang_c_Page_037.pro
a9a203489a49730ce44a4c96b86d1b0c
a9250bc3765de6607d389dbb106141b0366ba063
39023 F20110115_AACSDS zhang_c_Page_058.pro
00527c5ca3b9a87e9377466b5c257597
932d0436b99f21911bd4a66b029cda036d762f7e
44648 F20110115_AACSEH zhang_c_Page_075.pro
6819b5c3d26c71439e7af31157836743
c23e36d9ac9df6cf3d7614ba3e85a4a9ec8d8bc6
60694 F20110115_AACSDT zhang_c_Page_016.QC.jpg
2ccb7f1dd5ba4396df43372591a9eda0
12254e4473313bb3c9ba34b4c8faebbfeb52b308
82354 F20110115_AACSEI zhang_c_Page_086.jp2
f6ec169c54d47b5be657ebf099f3b4c3
cc58ebde7eaaf8978a949af1d611a22de54a509a
77787 F20110115_AACSDU zhang_c_Page_024.QC.jpg
ebc237ae3e7067cf17a9332f82e739c0
a6014054d116c3ef50d7d55655d9b822d3f080c3
37138 F20110115_AACSEJ zhang_c_Page_039thm.jpg
ca031f666590cc72eac96f4838b000f9
59adb1506c7d4095c8a78c61402511fba3e84845
2033 F20110115_AACSDV zhang_c_Page_102.txt
4bc168b3dfbf8dcb3038c181cd837d13
13826b3905bb117cc4ceeaa8a9876a9255afdc98
64090 F20110115_AACSEK zhang_c_Page_064.QC.jpg
8f132d6a399dc54e0c699dd7d473f8eb
7bf27b76c84649898ff55cbf3705feb32ed9fe61
29252 F20110115_AACSDW zhang_c_Page_098.jpg
dd828cf96b4290e32b657e9495ddbe02
2daa8f101bfbe97f0ebebc148a6545640fe8d8d2
1442 F20110115_AACSEL zhang_c_Page_018.txt
0b7e67fd7dbd80a0ab62c54cf79d1e3e
f3eab26ec7b2b25fdfad66ea668830df5a66be8c
19703 F20110115_AACSDX zhang_c_Page_081thm.jpg
fe04e6af14e1dea35b5c9ebc2f0667d0
09c34d7649a343520c473f20200a6172987209db
35973 F20110115_AACSFA zhang_c_Page_066.pro
90a00de2a55c00e516784c953a3ee4dd
7a32c232f106ac630e0a6649c5db5dcc3b089d42
1051970 F20110115_AACSEM zhang_c_Page_009.jp2
66c1f8ab0774c6c4517fbc9c03b00ff1
129a1f487868707ddc0537e772dd7d89c141f38c
4752 F20110115_AACSDY zhang_c_Page_040thm.jpg
3db37c619a3f9c1088cd59406067cce4
80d3ad5172c45e3480ba250ef1b13a6c72f704a7
82421 F20110115_AACSFB zhang_c_Page_099.jp2
1ec073c2e0ad8170d7610a86b569da4a
75afb46a3736c82850b91fd768bde67dda5cf8c2
44718 F20110115_AACSEN zhang_c_Page_100.QC.jpg
f6b7e200f6825fa78ea5900a07dc8750
f52b58c60f1648c6edda30d3cf0027a7248369f2
111460 F20110115_AACSDZ zhang_c_Page_034.jp2
4b0ebcd23822cef9f621243b41aa254f
657d195f1cb0990ded5b36023f99f3f9df0bb4eb
19678 F20110115_AACSFC zhang_c_Page_058thm.jpg
f306c0d064a115970a88e574c0f5b167
7fe81b0083849336d959a13643e89e26194233ff
F20110115_AACSEO zhang_c_Page_009.tif
e8afe33a60318fc5c960caaead5a135f
c6500cba6506edaf852acbd2749c37b7809becee
1666 F20110115_AACSFD zhang_c_Page_050.txt
68b72715efd537cbd0fd1ebb49d11077
587b4f049f47e308539e3c65878a4141065189bf
18178 F20110115_AACSEP zhang_c_Page_050thm.jpg
6845c5186726a7b52dfcc1870c6716d8
3f735d76f5f06e1be4b23f2f9df59fc552b2f9fe
1328 F20110115_AACSFE zhang_c_Page_101.txt
bb4330c5520f340aaaef3b28ad156c61
5d6637d29d36d5b108e36f1227932cd1a71bc03a
9250 F20110115_AACSEQ zhang_c_Page_095thm.jpg
da7d1aa58c158de9b6046f2f7ca7ec99
65f843f6d3e1616f84c510885b515b3acadf1fed
1089 F20110115_AACSFF zhang_c_Page_011.txt
c113066cf9e1142936f701b4201a3104
a93bbf123e3d8b25922087068bbc18b3e4d5793d
57145 F20110115_AACSER zhang_c_Page_029.QC.jpg
7397d150b26578aee6099f9d245149a0
7073eaf14f2e1de9564a91677eadbc79fa279de6
13255 F20110115_AACSES zhang_c_Page_069thm.jpg
de22b1234b56c43ef7fe636012f469ae
4aa5614728030a6c1ffcac4826684dc709f1c2fb
20068 F20110115_AACSFG zhang_c_Page_010thm.jpg
b1680c86aa7ec6fa3dcf37bf70e202d1
f0788f89378d8387b6805b8c0c32d49d01099d3f
29379 F20110115_AACSET zhang_c_Page_089.pro
bf6b77e79bd9001e2395dcd25f097a54
c0dddf0977c1e5059df3378f9f9f4d84b23c7c94
157384 F20110115_AACSFH UFE0008621_00001.xml FULL
604644b8688ddab21455f32011a6e3d3
2b7fbf887755204d20d1f663f8d99942487bb6e1
82681 F20110115_AACSEU zhang_c_Page_064.jp2
982a0f52c8c973e668ef1d10607073b1
58db287fd0e3ea49fa2737710e832ec7fea5422b
F20110115_AACSEV zhang_c_Page_012.tif
7e5450399e89071f55bfbe907948b69b
67e530b138107bbcc8a394ce70c23725a0966807
94632 F20110115_AACSEW zhang_c_Page_012.jp2
2d09ff578dcfff1d3e650f0174962591
5000e3302319b1ab4a89d64ce3156b952e530755
51653 F20110115_AACSFK zhang_c_Page_001.jpg
3535242c2cc854f3825941e2fc82dae0
4cb9f99a8becfec9dbdd4efa8a5742cb4fe17a9d
52712 F20110115_AACSEX zhang_c_Page_071.pro
a489238dd673c647a3e07d04aaa54102
ded7f93d5552d5b2e4494fbbf86582d98a04c353
14194 F20110115_AACSFL zhang_c_Page_002.jpg
9d7bcb5647d208c6c54606afd9b69446
ac37ddec2fb372d8d7a8e33d2810bba316b576b2
44029 F20110115_AACSEY zhang_c_Page_080.pro
f78391acc440e42ee7ed0bf5d9333659
8c5355a51f690df50f0ee5a17fb9ccf8e92d1ece
139910 F20110115_AACSGA zhang_c_Page_017.jpg
7ec94f68e028df5a8fe8bab892713cc1
5fac251daff07aa23a453aa13ba3af5285bf64d0
11852 F20110115_AACSFM zhang_c_Page_003.jpg
e4ac1b8bae2d7bfade38fde7bf876478
82c95d0134a7d8f1f895c0a57ef627f3878f404c
F20110115_AACSEZ zhang_c_Page_017.tif
acf5dfe16bfa91fbacf9e745152aa2db
8ae7badf3335f6482bb87267a4384512c3f3cfaf
173855 F20110115_AACSGB zhang_c_Page_019.jpg
df40ce0889bff73705a9b535981164a8
25f830d361e29f150be1a6cd8b2524b3f59ff8e5
139693 F20110115_AACSFN zhang_c_Page_004.jpg
88b5326af38c37365d4f1476f3564287
250f3516e032c21030ae1c051569f96b73bf3d9a
195371 F20110115_AACSGC zhang_c_Page_021.jpg
1907979ae935a32c55484be4c0cb632f
47452cf8edb139cfccba6b83934267c435759780
274029 F20110115_AACSFO zhang_c_Page_005.jpg
507090daab57ebc1e7a3934bcdc45bf1
12918c05f05819029befbb1c4c7e97d6b70d11e1
162573 F20110115_AACSGD zhang_c_Page_022.jpg
573d97f4ad524dee8b1d086b32f33f28
d74fa018994d50f73009b2c6a1f3200ed707c9c2
107451 F20110115_AACSFP zhang_c_Page_006.jpg
641e97ef910184b73f192024486aa31e
08605984a5a951b740a15a2f32fd6585992d23c2
185079 F20110115_AACSGE zhang_c_Page_023.jpg
06c58557aec97999a0128291fe0fba83
314d46550aace81ce32b7beb0a8c72a1496fc527
338455 F20110115_AACSFQ zhang_c_Page_007.jpg
ed23f7a65daa6d3867af5ba74c6c46b2
e98cd07f53470d402e01872e226c1cfacba1ced9
206292 F20110115_AACSGF zhang_c_Page_024.jpg
6045df50afcb4772fd9cbc16620c8b01
d85a73242d7f8976eb36f01246350ea00acc7062
374175 F20110115_AACSFR zhang_c_Page_008.jpg
a5411e9e794f1806cb1ba830014f4257
f504f3c1ae3d5199525c497bee6c59ed4fd8d412
196965 F20110115_AACSGG zhang_c_Page_025.jpg
21b993b5afba9daebf29d202549e04ff
b8128774e8fbfd301933d0bd548007995bee5e43
184042 F20110115_AACSFS zhang_c_Page_009.jpg
1bee5d3c52f4901382541cef1b3e0a90
c26d2c6126b37652a68a186942b5ad90d21c1284
190703 F20110115_AACSGH zhang_c_Page_026.jpg
5f5b815015674ba6203ab64a9570092f
9d5e2360a9c9c2f1863ec78c3e8e87eca3ac968b
180921 F20110115_AACSFT zhang_c_Page_010.jpg
9693d66cc1b32cf4211915739d5c2334
1adeebfa49e04e325376140a36f598daec529f5a
155196 F20110115_AACSGI zhang_c_Page_027.jpg
027e42da3d9909aeb3ea60db09aac587
ecd0c8cfe9be1cb82acb93c91544937943e5d5e6
121174 F20110115_AACSFU zhang_c_Page_011.jpg
28ebc19dd7691f708fa6a378a90a6f11
161d3dc0bdfb0ff9873398e7fcba95afe8ae2f7e
152222 F20110115_AACSGJ zhang_c_Page_028.jpg
04551954e560d618dd38d28f35c35953
7b5abc45d8db836e59a2ef3dc429da595af5bc67
182530 F20110115_AACSFV zhang_c_Page_012.jpg
1207114289be52b489d52aeb7f4b2207
cab391770104dabe4c3c0e6c3270fcf9dc810c19
152117 F20110115_AACSGK zhang_c_Page_029.jpg
0b280d6e5ad12f7af147ad03f18bec03
9e29bbc6c0241ca0b589b7a4c77e7b27d8362fcc
196871 F20110115_AACSFW zhang_c_Page_013.jpg
c5aa01b82c32798ec2602bba9f5c169b
0a9faf1547da7c43de702d8f3a5b269e5aad09a6
184596 F20110115_AACSHA zhang_c_Page_047.jpg
d74f3cf8582de201cbd5e88a05ad1498
56a661fc7f91b125ab3fa4f35584313d76014f8f
143576 F20110115_AACSGL zhang_c_Page_030.jpg
e113b572213bf6eeb1a812d98954d1c3
89fc94b2c1e632ffacc04ea915b340dcc681e129
192309 F20110115_AACSFX zhang_c_Page_014.jpg
63ea295f024f7fcdeb0f61aeeb9a3b53
ea66bb2b3a0dc225c9078b5cb09b5dfbb591ca2a
172979 F20110115_AACSGM zhang_c_Page_031.jpg
d004f01b073eaea3e92d09f96cb05d66
0d00ca972ccd7db3531dcb8f4b637a823144bc2c
138045 F20110115_AACSFY zhang_c_Page_015.jpg
574b24881cf491844988c491befde15e
75c643b5105aef529d80db00195829771214de8c
176616 F20110115_AACSHB zhang_c_Page_048.jpg
6c84a3efb54fefded7ec8bdeeaa7d4cd
c688b801cee796f0bbe691963a87d2f0bbb44133
209815 F20110115_AACSGN zhang_c_Page_032.jpg
d1d75e01aa27b2571a746b819ebf2be4
d6d0dc4d1f8154e5e217441cdfe31070dfa3b946
164689 F20110115_AACSFZ zhang_c_Page_016.jpg
ed96e73888a37da0eec36628702f1dbb
92153c5d922a12eff42257b0d5259c84c980f9ed
139307 F20110115_AACSHC zhang_c_Page_049.jpg
511fd878a700d9c3ac288ea56e5c521a
fbfcf94effb3c536b1f04f3e87082a451ad12e87
170362 F20110115_AACSGO zhang_c_Page_033.jpg
465010f12a7169ca8175d93826a85de4
2b385d46391cd62983353a55ca2f5b9496c6b3d0
135010 F20110115_AACSHD zhang_c_Page_050.jpg
9c324d216a5dfa308078f302c5755680
8b261494b619c34040a22154b06583416b7accb4
201234 F20110115_AACSGP zhang_c_Page_035.jpg
442e757dfa9ccf49f1f87b7eb2642eec
16ad659bb07f5d597d9532d7d4b6aa1599ae2df4
192034 F20110115_AACSHE zhang_c_Page_051.jpg
229c170608e706cdc54fee6a848dd772
dd34b0a9e9bca68cb4d3ad1201c1777d89811dd6
203623 F20110115_AACSGQ zhang_c_Page_036.jpg
c0ac71d7ffba3627d67528d6a50b02a9
51aad59654aae970fd6956c9d3758f35531329c2
135678 F20110115_AACSHF zhang_c_Page_052.jpg
c307618b1b0c4000d716680625ab34ec
78cc46aacf7fd8be0710cf276216c763a127c211
205970 F20110115_AACSGR zhang_c_Page_037.jpg
dd130b6a992c617911e5b696df62db35
4c4adca9e146e9d8b931e471b468751215a7e624
171788 F20110115_AACSHG zhang_c_Page_053.jpg
2bfe8da4c949efe9dd73570e6031f147
bf73d5eedca575778aa98f234244f60aa92fe8be
107103 F20110115_AACSGS zhang_c_Page_039.jpg
667125a0460d10cbdbf7adfff77f6654
b99d8bcb281823a48ee34ce745a533e24078986e
151456 F20110115_AACSHH zhang_c_Page_055.jpg
512485c2c367f14e86360cc93ac6159c
89e1554df13da7de442ca48f7f716a56965a610c
28034 F20110115_AACSGT zhang_c_Page_040.jpg
5697fc673a5a04be7680a8488faaaa63
2803e5d5459d2167a5ccba0c83a79922e70680c5
215300 F20110115_AACSHI zhang_c_Page_056.jpg
cead2298fbe1f2e6af655c33ec58bd00
9d897c5d9cb292d14ddc4b6ba9f6d5a4a05368c8
195728 F20110115_AACSGU zhang_c_Page_041.jpg
1693bfe217588e2f71516ba096a3330e
360a01120530ef23618c6784937e3c8f9e05ad3e
184944 F20110115_AACSHJ zhang_c_Page_057.jpg
0dc46abb81a4994a06153b72fe9c9ede
206b62907b482d7cdf76e0ffc20124d6dbf7789d
215613 F20110115_AACSGV zhang_c_Page_042.jpg
ffbadfed955223aa90392bcd9a293302
2c95560399de5d34d8a7b75fab60132528306381
171814 F20110115_AACSHK zhang_c_Page_058.jpg
e6cefdb17d02aa35ceec05d44eef33c1
752dfa1f8768d7939afd7de64824487865b99361
209524 F20110115_AACSGW zhang_c_Page_043.jpg
b49f1a437bf9f5e54cbe2dab6714670d
1d1280641c9f263d941895a12dac8f5b9606f854
215116 F20110115_AACSHL zhang_c_Page_059.jpg
fcfc50b33fea4174933b5da9b7a46a67
f069e69f753b4d64cac4b085a0ee4c20c046096f
210220 F20110115_AACSGX zhang_c_Page_044.jpg
27ef567f3e782f0bd0d122c6d72b5fef
13fe5156b39d2b11e98121015f2c98384458a813
187567 F20110115_AACSIA zhang_c_Page_075.jpg
c203ce1f1a911c963969352722fff9ad
e9c3b6aab10478cf7880012b8ea171c1539ec28b
192078 F20110115_AACSHM zhang_c_Page_060.jpg
70096e91e08792adedc4ad8345d6cbef
edc88f4b4db943030c9f6243626a2355ca1f6af4
164362 F20110115_AACSGY zhang_c_Page_045.jpg
ea8f12307afd44b780c701b3e7b0383b
cb274b8b7938c05844486656e50fba27f73bc3a6
212252 F20110115_AACSIB zhang_c_Page_076.jpg
de85f24fbd79158dc11aa227eb702aa8
7307e9719f3700aa487f70599e5b4af90934dac4
105999 F20110115_AACSHN zhang_c_Page_061.jpg
c1554d35c3a14e52449240b547ade3c3
c19f0876f2c7bd745242a09c8f427a90be17420e
157041 F20110115_AACSGZ zhang_c_Page_046.jpg
7cd4d54393c432523c1f4fbec263fa6d
5644cf858e5c00545ef274d459888bb953a77544
119591 F20110115_AACSHO zhang_c_Page_062.jpg
48c3c4ffdae270f3df8eb68214de18c4
d06eaae7a7711ab8e9cd43eb8c36be99532a8e03
200735 F20110115_AACSIC zhang_c_Page_077.jpg
29ae274398818646ed80f001b382cede
547d81c3de846200644c839611fa05ae727f42cd
205594 F20110115_AACSHP zhang_c_Page_063.jpg
db2cb3b593b9932e6058dc51f3372e73
9a50b3fb1c7380183787f9512e38de8fc096fec5
163030 F20110115_AACSID zhang_c_Page_078.jpg
da9ba13c9bf7b3d8f962ee44462d6557
e6a90f18277fbf491d91b45eed2049bbec1fad59
167035 F20110115_AACSHQ zhang_c_Page_064.jpg
2b79754f5b347d6fb45d7e30db94fc22
ce1101559ac7c99df2eb9a3b064c787ab076c379
175501 F20110115_AACSIE zhang_c_Page_079.jpg
cf85267ba909fd935f6e9b36a828dba2
b8992ea2b44b9efee2e672d8ab219357742a664c
210281 F20110115_AACSHR zhang_c_Page_065.jpg
3e6c0c2e1afd5abf557c35785a3aabc9
5da6bed7da49c3b3a9d9e93a5e29533294a2a745
186374 F20110115_AACSIF zhang_c_Page_080.jpg
4d30c4cee440c85b294d62f466052025
5940e3e070714a3a039b3a1de1ff8053f79477fb
165418 F20110115_AACSHS zhang_c_Page_067.jpg
d7a3a75a3d5cc6a9d3fd732cb7b9a65d
687ad342e071d9829d53d292f6b4ef24ec9e90a9
145691 F20110115_AACSIG zhang_c_Page_081.jpg
4dc45a4a454f7c691a587a6cb1bbd004
4fb9490b702082de8fd42aea8d9ab5d513aa0f15
208786 F20110115_AACSHT zhang_c_Page_068.jpg
9359a654bf030ea94736c4a5d6012ac3
f06f65bc911985cbe4c82d7636305e38fc895e24
151357 F20110115_AACSIH zhang_c_Page_082.jpg
132b328eb55ffdf55bc9af9fd91248f5
99e426d270294c27b41d3713fd2aeae7492d92e5
105757 F20110115_AACSHU zhang_c_Page_069.jpg
54c90a9a8c2be6dc16a4fe933a374bed
24778304b27a65cf82735f053df041cb84ea9051
165769 F20110115_AACSII zhang_c_Page_083.jpg
258776c5d5c459e2ec01c6d5d6b7577c
559c523955415e2fa9773962bb7a44c6e6834196
118211 F20110115_AACSHV zhang_c_Page_070.jpg
eeea07d6a4145bcf23d6caedc80ff25a
dc6cd08962a4126d8de6e5834846ea67dfbcb0fc
166422 F20110115_AACSIJ zhang_c_Page_084.jpg
737d5207d54d6ab94051ee59e274ed3e
4be2456ec2a8d2956ac51b55dd45cadcf8f54e54
216413 F20110115_AACSHW zhang_c_Page_071.jpg
9a5cb513dea593532685067baf5b1f2c
b1e7b21a09effb8ab6e554ccf3375020292fb0f6
106134 F20110115_AACSIK zhang_c_Page_085.jpg
b20f6f468a734c1f4643666364f5d4b6
7b984839ea365b0d4b484227df11bcca8dbe6626
166585 F20110115_AACSHX zhang_c_Page_072.jpg
943f53920d7e35144041f1a69b35b67c
28d692b7c5c4b042b8f3dc103b1876389e49ecb3
47885 F20110115_AACSJA zhang_c_Page_105.jpg
20f86f0356cbb061b92b084489cb21d2
ae88c4ca050ad1533fc074ea4a5d72ba8fb02e4c
160586 F20110115_AACSIL zhang_c_Page_086.jpg
9bf02d5de28df1a6f4812210b344b0ee
bbc92b470d8572df2de25eafcbd957bdb7844e7f
125174 F20110115_AACSHY zhang_c_Page_073.jpg
49195aa9834789d8658a5e6c9b6f0687
40fc913a7a8385b82cde15a8b8aab69510ffe2f2
23194 F20110115_AACSJB zhang_c_Page_001.jp2
792871a5f4115d6f91a82b34773e97c4
74d661742128bd15296fa8337a7fdab4ab746fdf
159923 F20110115_AACSIM zhang_c_Page_087.jpg
507570c730e64343e696d521924e9398
748f1448ddd5022deb707dd8ec8a0158db9652d9
180187 F20110115_AACSHZ zhang_c_Page_074.jpg
9d12c1e6f7359e1a6eeb3ee949816de2
e5b4c3b4d9d6b477776c190163c02101b80aefb5
5380 F20110115_AACSJC zhang_c_Page_002.jp2
7405bb459fe7302ef5ddd3c58b83ff28
49a8b7076c5aa186667ff904c563426298907f1c
132028 F20110115_AACSIN zhang_c_Page_089.jpg
ab138cf3c49104fcd2e4d9433045b1ff
94eaaf6f6e657adf0de0e4f2e9ebe26f44975ce7
146956 F20110115_AACSIO zhang_c_Page_090.jpg
41cdc1866dda0f437dc9c2ef35244041
2f49dbd2f8a15176265fc8b9e17f82b3a3b135a9
4130 F20110115_AACSJD zhang_c_Page_003.jp2
7f09617dff18eab59659cb48ba7d4000
70adc65127099ede0525ceeb7eaddfe69e8ce387
70928 F20110115_AACSIP zhang_c_Page_091.jpg
71a7dc8296a29fc0faeaed0ce800c7bd
31bbed05fc60ce9453068f7a2c615e34f4cea3b8
1051969 F20110115_AACSJE zhang_c_Page_005.jp2
25b15f3cecba7eb7a1a762f080a4c287
8a801eaac5fd8cb1c97fba17c6533078784b6bb3
97833 F20110115_AACSIQ zhang_c_Page_092.jpg
764af0dfeed6c07755ea88f55a105701
68d15813b655d0232b7c4c8c7d3cc600866dcdc0
601697 F20110115_AACSJF zhang_c_Page_006.jp2
5f0580b619171ece18b0c07c0adb612c
b9979ca3621f515ae553460166b2720091d8c00a
100141 F20110115_AACSIR zhang_c_Page_093.jpg
5a3cdcd5b92f4eee358196486a432a66
8dca940998192c5fb8a36d3ff22f885c817e4333
1051977 F20110115_AACSJG zhang_c_Page_007.jp2
3891d521a7f148e05729278249526a37
df973fcf9f74c88277751cebdf981ce4af546788
184146 F20110115_AACSIS zhang_c_Page_094.jpg
f37e27378c8207b49ed9186bd77172b8
ff499850760c1b1c5899b35491e29cc85f4d9096
1051985 F20110115_AACSJH zhang_c_Page_008.jp2
c9e9d84207e9343cdc588519f8a230c7
0177ebb6b7c77622bc36b25da094aee17b523706
64531 F20110115_AACSIT zhang_c_Page_095.jpg
62eb3d33c6c000216fd1d1ab29fa9f50
3e6f3ee6ef9504a9d3da5d9987f22445c028a47c
88000 F20110115_AACSJI zhang_c_Page_010.jp2
63d07e0fea6e751bf5b350506a9d469d
7d4f054ecd56cb58a23886aaaef1efdef9b99763
175927 F20110115_AACSIU zhang_c_Page_097.jpg
0048f1b99e0ed1b0d1dbca7772429ef4
1f01450bc40aa80e80427c9c52d5ee188e5f824e
62632 F20110115_AACSJJ zhang_c_Page_011.jp2
ebdbdfa819b0777ac41c63ea666b20b5
6104510f52e803811b620708d0dc7f09256e09b3
163814 F20110115_AACSIV zhang_c_Page_099.jpg
b1a90bc1ebe7686d087af3786ca6d0c4
4780a5aa96bdf4e845fd7217b2d9861a65d35f47
880176 F20110115_AACSJK zhang_c_Page_013.jp2
3484670ddf09bdd0e9e94e485acdba63
774f66faeabf86cd5498d2a3098b00c5c0d8ce0f
119197 F20110115_AACSIW zhang_c_Page_100.jpg
b6af92c3fd858bebf7a7dad25b99b7aa
d8508da76303882865d954670ba3d2c25a6e781b
87914 F20110115_AACSKA zhang_c_Page_031.jp2
5bc37b26377f3c19ec43405b073f4b3f
1b4af84087592413bdb05939cf876b58b5b0847f
98704 F20110115_AACSJL zhang_c_Page_014.jp2
8aa04f1382a64b5f315231ec900e1b32
46a8f32b823b8506ee86bd664dcb4a32257ae88b
150168 F20110115_AACSIX zhang_c_Page_101.jpg
acc4815fe8bc111fc11f4bcbad6e9968
ebed5ff78a67582e890f82a2b441143a185485df
107418 F20110115_AACSKB zhang_c_Page_032.jp2
b8f9e09dc4bbd3df8d1426811a86c567
590cc7d54c0e4622349910e1b85d9dd6d736290d
66813 F20110115_AACSJM zhang_c_Page_015.jp2
502549a52043eb36ff60839cef859a6c
50ad9a724d2d3bfe18360bdf70cff132fca70596
208376 F20110115_AACSIY zhang_c_Page_103.jpg
22766dc97fa95d16079f281c20b5a899
d46f9c27f2529ba561064fb878c0698ad1f8e240
87002 F20110115_AACSKC zhang_c_Page_033.jp2
c392f91b4a17eb3a3750ffa4d7804442
a0c2612541e2ae1d92f6d717b319ce4c428b37f9
82039 F20110115_AACSJN zhang_c_Page_016.jp2
61d3a90292c132dd833c362993361a91
d2f276d01eb5d0f664c005bf81bd0b13c4cbc7ec
215117 F20110115_AACSIZ zhang_c_Page_104.jpg
e070b9f2a839f39e0b72700f6a76dcad
4f8db8355ab13cdd4d618ce874c4902ec1009aa7
914891 F20110115_AACSKD zhang_c_Page_035.jp2
2536da9441894ff8b97b52414352e836
a8958e14621d3df44c3453ccd2ea510c17597870
72421 F20110115_AACSJO zhang_c_Page_017.jp2
23b4f3c9397bab26e6b49ac1755b529c
5fc9d2df92c737441ef6a5bc5427a6ec0cc052de
67762 F20110115_AACSJP zhang_c_Page_018.jp2
cf96c717fb82b5bbb68fdf997bc24658
b345a38ec318954b5be8a37b89e7907fa69b1c24
88282 F20110115_AACSJQ zhang_c_Page_019.jp2
06d66cfef133b5f96a3f7149aab004dd
f4a69a2490c63a2700a209c5f1acebca63b0da05
1051978 F20110115_AACSKE zhang_c_Page_036.jp2
0982468ce6f65cdd47eb797da4dae83e
f410425def64f8cc1b76466812332a0672fe210d
85893 F20110115_AACSJR zhang_c_Page_020.jp2
11ec118b0aa4ed8b9886688427a0d8c0
5a8246da782f370faf4bfb61820f9b39942343d2
1051912 F20110115_AACSKF zhang_c_Page_037.jp2
ff6be70ff776ca0fd2c10b0e2cf621ef
e9b3966947ea5d4534f06ff9410099d65937d058
99674 F20110115_AACSJS zhang_c_Page_021.jp2
6c71e4ac5aee75ba9bc5efa54645fa29
5b2d4d79532a97ac68b3ab76330fe6303b4cea74
797898 F20110115_AACSKG zhang_c_Page_038.jp2
824198f700fb1ecd33e4b216fee11926
7da86eab3a41d88a617985373575e33e0cd847eb
86262 F20110115_AACSJT zhang_c_Page_022.jp2
7cdcbeceb2b33989dcaa2a3f457b373a
0478851f17ac6ca5e1837d8e3b0e38a510253421
13152 F20110115_AACSKH zhang_c_Page_040.jp2
0598913445a8ad2711f065250ecb02e4
01e341c0de42c93e6a7a242d4783f202abbf7d47
93794 F20110115_AACSJU zhang_c_Page_023.jp2
9bfc52b9d57e298b6a7b1a13e104bd1f
b6b8bb9015a3f630a17f22635ec2dc8ea8ad1115
99813 F20110115_AACSKI zhang_c_Page_041.jp2
82f044dbde811e5e5dbcc081757bd392
450b9f5db6cf1e9687ed138eaf32319bfec4017a
106500 F20110115_AACSJV zhang_c_Page_024.jp2
f2ecf2871c7aedff7d671d4ec29027a7
f3d1e25d951993e40bf09c9ed43bcd054468d1e8
111687 F20110115_AACSKJ zhang_c_Page_042.jp2
7972484f8100941dab70853fb8cbc06c
45fbd0c408a5b64eb77705082e9b0eab0fa161ce
1042797 F20110115_AACSJW zhang_c_Page_025.jp2
17ab563d1c623973e42ec542c7787e93
57dcf9c3d6dcedcbf6eb38bd07710188771eb32c
108521 F20110115_AACSKK zhang_c_Page_043.jp2
c2cee9459b88bd8c863eafd0b44e569f
d39c8b64503235e4ac8b2ccc329fe513b5fc8092
853468 F20110115_AACSJX zhang_c_Page_026.jp2
cd5f8a7d609af83be6ab5f5d66973013
1f65f9993b63af77a9b388384d91cb1053215aa8
110580 F20110115_AACSLA zhang_c_Page_059.jp2
ffc3dd858517b6c75534fdf5b90a23b3
5f2213d515acffb039463f192286ff8f3c5d92ff
105724 F20110115_AACSKL zhang_c_Page_044.jp2
bf6ce21f35f7306fab7fa4fbdb8b7c60
98ff371e1964c0ba93cfece619de76cd0b2dc611
745170 F20110115_AACSJY zhang_c_Page_027.jp2
1d7f232eac34afae73855f15b9cd08d8
11ba7145f65954488b16e8334fa4cf0275ee629f
99926 F20110115_AACSLB zhang_c_Page_060.jp2
d176b52abddf35faf15787386fa8d1e5
1b84d4e00185c1d40815e1f5c8d50e47e6c2f11a
82524 F20110115_AACSKM zhang_c_Page_045.jp2
fdff1543df9fa24810e69a93e3b750a5
01751282cfea6f333fcc1cf12dada9d54c95c1a5
73284 F20110115_AACSJZ zhang_c_Page_030.jp2
122d2090844b268d84807e5ad482d539
ff34792ed01fad58e6b77602c95952e4829cad4b
93455 F20110115_AACSLC zhang_c_Page_061.jp2
328f12f2abe609f905b5a02c970abd0c
60e73648bf6964d8110e7bf5c5b3bc05b83d7127
81132 F20110115_AACSKN zhang_c_Page_046.jp2
5141a2e0b70e49f5b87e6dca3680060d
9e07a2a51a8232471866da63ed4ad2f799727464
55545 F20110115_AACSLD zhang_c_Page_062.jp2
57de306dd161c8de689b24d3f6f79c7a
95fd4a926dec645090ebf6120871437026e11afc
99756 F20110115_AACSKO zhang_c_Page_047.jp2
755837a10e2fec467e149613ee463be4
7aab7f9e30444d18d9168ca6a12b7c8d4cf266d4
103815 F20110115_AACSLE zhang_c_Page_063.jp2
8936307c62fc27f57ef8c92b4ffb580e
8957f02a59118082a4ab446b48c1bc7f34d6e447
90636 F20110115_AACSKP zhang_c_Page_048.jp2
5f5c162292fbe99754d64af762416cf7
388c24355da54926541e745b3d1364321df2c3e6
73606 F20110115_AACSKQ zhang_c_Page_049.jp2
78968e1be093e55700509d7daa5d1f95
4978345f67f1f2102f2a5e24bf46eee07eee21d3
110106 F20110115_AACSLF zhang_c_Page_065.jp2
311884dbd2f1b0e1e5582466e66136cb
5856e0841380729a8b82c956d495a62edc506cba
70016 F20110115_AACSKR zhang_c_Page_050.jp2
924022aa1c3d9571bb6e5b41d56f7a50
85147b834b7dc8040f3be6f7c7625007fd7e78e0
81705 F20110115_AACSLG zhang_c_Page_066.jp2
29cd1f4ba0b58d6aa1de74be2dd382a0
cad428cabed92dfef0255c96d45a442aa4a324fb
96984 F20110115_AACSKS zhang_c_Page_051.jp2
3cd94015c9b2244cc69fcbd189827e96
b42824e45f7b90a7f821c6ce1c13c5f27bd38d7e
80783 F20110115_AACSLH zhang_c_Page_067.jp2
c59a26644c7c8be20d42ad660bfe000f
5a8bb94dd15e3a5056c8e27be66a2a27fc53b7ce
72198 F20110115_AACSKT zhang_c_Page_052.jp2
1ce367601f66df68e1a716c71e762ac5
2142e899a18bc01b16765dcb9172afbcdf897ef8
105271 F20110115_AACSLI zhang_c_Page_068.jp2
ac22487e443e9fca7d982c7173bf07a8
35f1ed96951418ebb8c943dd9d506faf6c051f88
87870 F20110115_AACSKU zhang_c_Page_053.jp2
2e507ad1244a389c62f3d23ad4023208
57c7ba3b1ff72fc81643270683ed9551d55b0dbb
52893 F20110115_AACSLJ zhang_c_Page_069.jp2
d96e9d6f2abb025c98eaca20726694f9
1e475212fe5efc85d389c37447f4cbc9d4abd0a9
88862 F20110115_AACSKV zhang_c_Page_054.jp2
71990ac8ed8fb321cdaec91adb22f106
a766d96bce3157da2fb9b72d03494d6041380854
98882 F20110115_AACSLK zhang_c_Page_070.jp2
4f84f4ea1ba721b5d238bd81dd0c147d
9b3f584ee779ea215f1b53afc46f3933a58d3797
76509 F20110115_AACSKW zhang_c_Page_055.jp2
8a939ffc49d511b7238384e7f96539ad
a7d07c5cf25403ab0eb93b5e8b0a8363c5eab003
85997 F20110115_AACSMA zhang_c_Page_088.jp2
44051db2f1381eb43d9f85e472ace141
9202d6e2937a848667e0ad61365d6f68954b4c46
111010 F20110115_AACSLL zhang_c_Page_071.jp2
eac16bea7a805b077f34ac532e28873a
71248caad5aa7b1eb803b77da4eee231ac1be85e
110638 F20110115_AACSKX zhang_c_Page_056.jp2
d3a276ab4a07476501d8c646f3bf7bdf
47c5595e5f0e08cd8093e46cbbdd8b71fdd25af7
75214 F20110115_AACSMB zhang_c_Page_090.jp2
c458e6f94994e8353f57afa458b8c2c6
20aaef48081a733fb4527954c3f14869f6c25a72
81696 F20110115_AACSLM zhang_c_Page_072.jp2
7aac4787ddf30286fd25866f2508060b
f76f6f9a1d7554423346d278764f424800d09fe3
89322 F20110115_AACSKY zhang_c_Page_057.jp2
84508bc01d1e21f8c3093bbf9e18b796
d079a5bd171f2c670698cc67940a72adf3d425a1
33815 F20110115_AACSMC zhang_c_Page_091.jp2
e5551271200259db5dbe1652c6b8417b
9b836d7bc3405c613fdf5cf81adf101af549ec38
64189 F20110115_AACSLN zhang_c_Page_073.jp2
9cce4c8c6acd36eda4fbe6ec0eb3b0f3
f72f10e3661131b0a1105057e4193e391460ac98
83838 F20110115_AACSKZ zhang_c_Page_058.jp2
a8a3a5d530b475017813489709941fd1
b8d85fdc5e7b32315cbb2170eb8c313aca7f3395
50582 F20110115_AACSMD zhang_c_Page_092.jp2
a8c1ca0c36b0871b3a8963f72abadcf0
72348094a05886d6e4102425ac8bd525a48ef6d4
91666 F20110115_AACSLO zhang_c_Page_074.jp2
1261d274661764c05e39d4c4a399bf16
d8361472a205110fc9ff63da1d89658c03b46193
52783 F20110115_AACSME zhang_c_Page_093.jp2
4e7e9dd00d1a3cdc3c19499abd58108c
5384a56297c56a8fffa8433b86b443b74a9d3cf2
95940 F20110115_AACSLP zhang_c_Page_075.jp2
e69f8c1d747fd3f4ea7118d4dc22b799
4cabe8f42624fdea3ac27df04ebb037219121a2c
94003 F20110115_AACSMF zhang_c_Page_094.jp2
1c9ac1fc3a9ae5ace7192b70d565fbce
3bf68d8334aeb235ee57c3e696453b361caaf75c
106503 F20110115_AACSLQ zhang_c_Page_076.jp2
6b4b042044fbc07558b3212bc93b2d30
c23f16c56d00f57f98e2a3ca8ddf9466e5a5fa32
106492 F20110115_AACSLR zhang_c_Page_077.jp2
767adad32abc26c3e21b16c90619fc3b
ab86a757683dc330be7c8f9268214de38c04f16f
34523 F20110115_AACSMG zhang_c_Page_095.jp2
8c51450bd5ed425dd0cc7a222e002d85
eb026b3c254abdf17d95303621558dc1a410ed81
86722 F20110115_AACSLS zhang_c_Page_078.jp2
3ed0db2100930ffcb1da05f53ccaf5c6
670cff3d7f968db3545e28574a271c4800101acd
763033 F20110115_AACSMH zhang_c_Page_097.jp2
9aa1528bef91bc7a0f83e838bd13eb6c
d3b6051168b28bf69f740c3050ba0ca02e32853f
92857 F20110115_AACSLT zhang_c_Page_079.jp2
5663a5cbcf5b38e53fbe33592e1b332b
db8f0ecacf0b96327c46947d5d7a62c563057f33
14818 F20110115_AACSMI zhang_c_Page_098.jp2
9c1ae9462dc7441c8d0f19310ce8985f
8fc279dfcd28cd80ba950520e2201a239507826c
93837 F20110115_AACSLU zhang_c_Page_080.jp2
0cd3be8278c85c23a1ee88bd187c9b26
aa40fee91a4e544e91e55a83d34ceefb9735964c
63610 F20110115_AACSMJ zhang_c_Page_100.jp2
207a4dc65f376676b0c0389440023422
40f11c357149f02f8fadccab0d0e4561d26ae3ed
73445 F20110115_AACSLV zhang_c_Page_081.jp2
1eee020ec063fd9e12272063c63f238e
50af48cff79c283499ea849c1e095c7893b1cfb7
69365 F20110115_AACSMK zhang_c_Page_101.jp2
ef718500a7a178bfc71cfa2bdc347c01
f56ccea1d03981c30a421a6072549218ff8e6073
86785 F20110115_AACSLW zhang_c_Page_083.jp2
958f53e62bca03e381283cbbccf6205b
b35845e40bc49d1a633d01686e0f2f3b994726b8
105968 F20110115_AACSML zhang_c_Page_102.jp2
b202fa1944409491c7f4b14094a29790
2070514b9cb95ef53d300f2585ea115f9821b023
87410 F20110115_AACSLX zhang_c_Page_084.jp2
630cb711b26f5498fa982cd45ab0e232
9dcefac73eef497f6a7c4b7592888873bf2e810c
F20110115_AACSNA zhang_c_Page_019.tif
78c782092d022beb92b90dc1b7a4dbfd
1fdde26294f2db9c1344daef663058ace8434337
99359 F20110115_AACSMM zhang_c_Page_104.jp2
6fed2af418887a8fd3798ee3a7349985
cca87c1ed008f2b6d44ab7d61b060130adfaf881
59423 F20110115_AACSLY zhang_c_Page_085.jp2
23760956f059eed4b13a0b05a79f9893
7128c23908b3787a1329c68d05f180e502b21035
F20110115_AACSNB zhang_c_Page_020.tif
d80c0a69a58bffc4d0f10c6fb89adce2
7e526b876629081195219dda45ee072f5011837c
23118 F20110115_AACSMN zhang_c_Page_105.jp2
7b1d78f16080f3e9d187f2fdd09b6c71
e52468f5bb0add657647bec436467e4c5c81adec
81650 F20110115_AACSLZ zhang_c_Page_087.jp2
87f4d5f238ce4534909ac2906cfc28d4
4e7d4f3645eb674ba33f121feb1b18fc354cce72
F20110115_AACSNC zhang_c_Page_021.tif
5c030996d21f0895d02e35cfec72037a
076c42119042608f4dd14540dda0e6fdb66737f6
F20110115_AACSMO zhang_c_Page_001.tif
a3a405aca9b451d46edeaa72901d01e9
6405eda82ee35728b457e53655f0a8e155ee875d
F20110115_AACSND zhang_c_Page_023.tif
b8a45211a60f51ec4c65d6048e76650d
11d2e23eb2498f7d0cc8a259b155c766cb9a3b8f
F20110115_AACSMP zhang_c_Page_002.tif
d367c34fe5c48bd8196a796f2517335c
70a7f0b904f4e65cfa0b655608cd21777bb6a4de
F20110115_AACSNE zhang_c_Page_024.tif
b5b4aec68cd4e74e8d305b88edcdba5e
cd40019ec950c7e46775a550c6a6e7f8717a8139
F20110115_AACSMQ zhang_c_Page_003.tif
ae78d1ec536f5d42c572bef8a7471245
15edaffb8b4dcc2185dd7d7897bd18c5f2c6fd9a
F20110115_AACSNF zhang_c_Page_025.tif
4c655fd713111d3bfb01ab6060eff8e0
debb26999332abdc727a3ada157714ea2220a17d
F20110115_AACSMR zhang_c_Page_004.tif
0ca11c84c9f39ddf66280e8d4506da4a
98b403bbe2cb86352502d10757ee7d8bc5fbcb73
F20110115_AACSNG zhang_c_Page_026.tif
2de5eae3500aa51d75070012cf83107c
d8c02722d2fa4e5ef3afc5fa3af9b2f88a081304
F20110115_AACSMS zhang_c_Page_005.tif
9b1e16512c40304f20d75eea0e252995
323599b729778afb2fb617aa179586ed2024c80b
F20110115_AACSMT zhang_c_Page_007.tif
7810f3d43140a6211288a13dda82eb10
c1ae97675a140e820012ca84a6a47907d1ca9df6
8423998 F20110115_AACSNH zhang_c_Page_027.tif
eba60731788351590de8d252e7a39926
6e0120cdef12842bb02610b6a389921bd7faa1d2
F20110115_AACSMU zhang_c_Page_008.tif
b33f4665fbb93b6b23c3ebe952a0ad7c
5e6c194a29d1fe4efda0a6c1f51d1e27a144160b
F20110115_AACSNI zhang_c_Page_028.tif
e35d9254c33070b58b8a1f9ea22b629c
12f10e47e407b8a098c6e1864e930e7b9e0bf745
F20110115_AACSMV zhang_c_Page_010.tif
2cc2cfe5efc21ec5fa56c4c7fc654a08
d32ce8c879adbb475b009b3ebfdea2adf95a60ee
F20110115_AACSNJ zhang_c_Page_029.tif
e1295a26053e5416f28698ecc118dfca
2f5b6f092453edf302f1170cb8d9e8223f7858a9
F20110115_AACSMW zhang_c_Page_011.tif
af74ab22777afed3285a0f347e891b03
2ac355f36982bfcde5f842ea8d11e8babecaa3a3
F20110115_AACSNK zhang_c_Page_030.tif
624d2ad51dad308726afb08f97f090d8
0f3e35f4eb245ee4ca8caed966fda493fec7924f
F20110115_AACSMX zhang_c_Page_013.tif
b25f05d0f179e4def498684e38f5a1ca
f70de06a030358ca5410570ae613ddfea0204af7
F20110115_AACSOA zhang_c_Page_049.tif
fd66c6e03f487cbd6e470518843d3405
ff3e7904623b1de5a8e406abc2ec91f200f1d98a
F20110115_AACSNL zhang_c_Page_031.tif
e95558ad04e9c57f41578d1d556f1716
6e99dda9a4db3daf42957f400bff3272be7fb4a7
F20110115_AACSMY zhang_c_Page_014.tif
e6922497b249357140ec24d104b57dcd
86617879a948b219973c74a077bc24569e107723
F20110115_AACSOB zhang_c_Page_050.tif
0cf76b161473079374b1bd197d4b2154
eda8c646077b034557f6955d7b595293b03a9062
F20110115_AACSNM zhang_c_Page_032.tif
5deb9568f48f62ab925839496bae5e75
937a5499fd0da2b74de6ffae0af29e41e177949b
F20110115_AACSMZ zhang_c_Page_018.tif
b35d68f0455cf7a8687be6ca1754ff98
486c61c1fd06fb820f888be0082b5ab26f766913
F20110115_AACSOC zhang_c_Page_052.tif
e84ace5a451800ced60174473db0196a
5673d2553d5c22ad280d7373c63b6bbed9191720
F20110115_AACSNN zhang_c_Page_033.tif
047f6a91a3d97f57e9ef50dd72c2046f
a8f129bf40adb0e423466a47db6e8d415c543d31
F20110115_AACSOD zhang_c_Page_053.tif
927dba66b67cfb48c17e2a52f2501cb2
12f3ac783764c3e78054087d4514f60cce3ec4af
F20110115_AACSNO zhang_c_Page_034.tif
15b03bb9a31b8c5bcb374543493aad62
477b6f6d42461f222931dc1657b0c6ec9d06efef
F20110115_AACSOE zhang_c_Page_055.tif
42d2e8b492afc62dbaa93d62c8a42c32
6a4a2a0b23a498d5e9684e8e9acea07d534bc8bd
F20110115_AACSNP zhang_c_Page_035.tif
039d4bb54bc6ddd32a997b1345777d78
7fda64a339380779c588d19f8fc8c775c70e262f
F20110115_AACSOF zhang_c_Page_056.tif
a320199953c7a53b37f6b8c82018b910
8728289523e9473c34b9e319d2248758ab1331f3
F20110115_AACSNQ zhang_c_Page_036.tif
f4f6fa28dc7821a4153937b8a4620a3a
2179a45dcdbc90d031c28215afff613ced564bf0
F20110115_AACSOG zhang_c_Page_057.tif
a8caa70a5d8adba2858bc26f9b6288e8
1ba47aa2ff6b12c1685346d0c7ce9b3f9b6887ba
F20110115_AACSNR zhang_c_Page_038.tif
6a71e6071534bcf34f5762eefcf92361
af56e9998792be2e0e77622bf2bd577daa2035ec
F20110115_AACSOH zhang_c_Page_058.tif
5218fdcecb199304e29f0a54f9752948
4264e22e99afb938ba6ba12c6698238ed0d60112
F20110115_AACSNS zhang_c_Page_040.tif
1595c4a55863945176777b385866430a
b3fb5e3dcebeb70ec11ecc69141e44bf458567e4
F20110115_AACSNT zhang_c_Page_042.tif
9385a3873051df7f79c87a6407f11043
d07377691338d41daa1046ce16b190ef87798734
F20110115_AACSOI zhang_c_Page_059.tif
6c3e6b8b7c0f4bfd7847535b0d5f9e60
b0b47f462c2c77fb2ef866afb02a8c8ac09ae964
F20110115_AACSNU zhang_c_Page_043.tif
96f7d1365ef13991135fd13bb94038cb
f59f67f72a446ad5b13170aefac1308c54f4278f
F20110115_AACSOJ zhang_c_Page_060.tif
5e52353af182ad6e28da4f10fc07980e
f9be9c61abb0db2689ff1a23289c507cbc863e62
F20110115_AACSNV zhang_c_Page_044.tif
0957c292edac4fb4b4918e117e9a25a4
d8a309a503e61ffccc2d80d09584dbe5df797d3a
F20110115_AACSOK zhang_c_Page_061.tif
af67a4e15310c3ac7c07cec7213b1d2c
53e79b9c25d5533f6ed2e8cf447a1d21a7d6013c
F20110115_AACSNW zhang_c_Page_045.tif
21f2864e63e706fcb16921685e0da5f5
80f2e217353a2141354dd3f22275581113f77e12
F20110115_AACSPA zhang_c_Page_079.tif
ce59cbe84bc9688f2c8790d5b33a9307
f09b4f567f18b356b8ebb164740033f04bd0c15e
F20110115_AACSOL zhang_c_Page_062.tif
3db6f4e4366bf23534e123609c115197
2c9e37b1ef7d307ca6c82e63ad30fd4ff76e0cb0
F20110115_AACSNX zhang_c_Page_046.tif
4bcf90a6afac36582626b61b5c0d4e9c
5495eb247f03866c54cf221637457f7c7aa96905
F20110115_AACSOM zhang_c_Page_063.tif
5da393d3604fdbddf403d32833d387c8
2c9db9d37b7048067c91ad3bcc8e487915927a85
F20110115_AACSNY zhang_c_Page_047.tif
9d0eecbc3e91864fad105b88f7115050
dbcb6c2dfb1bc21e733a58cdce184f63f8aec7b6
F20110115_AACSPB zhang_c_Page_080.tif
70f30fa2c52238d15d14afdbeea9a974
4fe42502e3880b8617488392d3324e911003c974
F20110115_AACSON zhang_c_Page_064.tif
4018e1e73bcea4c7eb65480a20dbd0be
fe9c6bc9b1a60d15c740b9902bfea36191d83a94
F20110115_AACSNZ zhang_c_Page_048.tif
fc07371d42451c47acaa6e07f5a23f92
bbd913211537ab8d35f99fa4df80940197caca62
F20110115_AACSPC zhang_c_Page_081.tif
7fcad3a996f9ad506720f8ec7835de3c
fd02b1065b2c8c9cd26a95974518274d47779e2a
F20110115_AACSOO zhang_c_Page_065.tif
cd2e7baacacc93643a9b313b74137e2e
5d9b44f480d9042d57962093295ca90efc495552
F20110115_AACSPD zhang_c_Page_082.tif
3a4869c9064d15c66e639057c384ce66
8400c430096842877424536c828a3fd7ff23e66a
F20110115_AACSOP zhang_c_Page_066.tif
9d159afff0f6a83f8ad065b6f58ab6d4
db6e0447511404b733956d5057ec2da8111b61f1
F20110115_AACSPE zhang_c_Page_083.tif
5942206673358015a9ad7078bf94babd
e2d730b470a230f7b8435b2ef50c6e8dea267bb7
F20110115_AACSOQ zhang_c_Page_067.tif
33480b5769fb2d1749518394b7af0900
ef2a5e1f0ffe11b6ba8bcd2583d9b0cd99abb18e
F20110115_AACSPF zhang_c_Page_084.tif
76951e8435acc5196030912de1ff57ba
d5ba3cc1f4c8025719583c0e506d6ee91a5e96b4
F20110115_AACSOR zhang_c_Page_069.tif
1608cd2fe47c4ec8e5cb0239300eca04
82bbdd1ee3df9ccc313a3dfaabe2f74dc3c6ed60
F20110115_AACSPG zhang_c_Page_085.tif
6a78942e80084db7c81b7d22ddfa00a0
fa96724a3261faae78893a400bd945d07a8955d2
F20110115_AACSOS zhang_c_Page_070.tif
678631bf0d08edbb867d4b1511b1c4b5
9d9ff8d249f7dc1ec1648b9429aafe688539a478
F20110115_AACSPH zhang_c_Page_086.tif
f24099c75164d7a03478a50239035a71
1692c4c88c7bd7294d094c2a8f265fc4ecb8b272
F20110115_AACSOT zhang_c_Page_071.tif
c52f97983041a323aa3ef39c2d0d7947
1f21b82cc55b180048602fbbe3ebab17a6d6de1a
F20110115_AACSPI zhang_c_Page_087.tif
b4a50a80422c920e0c7e6f13f7175ab8
c296221735029b02744c5289251b8e325734c53b
F20110115_AACSOU zhang_c_Page_072.tif
b24e7df318680a8863efbf9414a5d034
b6806413bdc5f4855913a4c21a9c2d7130a43fee
F20110115_AACSOV zhang_c_Page_073.tif
1b806a49ebbf69d766b1d5f89e0dc9b2
5ccd4adf09a3511105273ab2440f4a16f946e169
F20110115_AACSPJ zhang_c_Page_088.tif
6228d5c7a47872f5b2787ee2d8b08721
56f008347972c29b2a56923b511e545333f31822
F20110115_AACSOW zhang_c_Page_074.tif
a3f0cbad78f2aa773098d29a307f2dc7
1aef293c82b3a5702377c4724309a8b18bc98ab9
F20110115_AACSPK zhang_c_Page_089.tif
a20ea2859da0dfa156791a6fb78977db
8b9093aaca21a16865507aa97b3abe730abfe906
F20110115_AACSOX zhang_c_Page_075.tif
c9bdb7dfe138388322790aec42038a4f
3d96786d7078920901d8cdf1a8ab5d3206428e20
31173 F20110115_AACSQA zhang_c_Page_004.pro
07538967ca47c5c51ceb7777dac80e11
9af4c18ab87f9ae792ad7421a5c0df32868eb353
F20110115_AACSPL zhang_c_Page_090.tif
a1e29a891d43adf87807c4f08ec1862d
348b7342c7bfa02c1213b1e742ee09fea21e09b2
F20110115_AACSOY zhang_c_Page_076.tif
7f4c577993d9cbee7ce5c495e95a62b8
105fc3706390902bb2655ef7aba885db5b3d3ada
63271 F20110115_AACSQB zhang_c_Page_005.pro
a8e53f1556d3f9502ffe33ab43138bb3
465e4e12777b1f7d30ee957d632f6cc41e0df36d
F20110115_AACSPM zhang_c_Page_091.tif
a18b2bfddb88bd81b217c19d7bf0fc42
44d87748a7e603754f290ebea8f145e96b30ba1d
F20110115_AACSOZ zhang_c_Page_077.tif
fbf98d79b2efd3fa21253960c5e61aea
fd3cba5964f579bce87b175c49a51e4cf69215b2
17892 F20110115_AACSQC zhang_c_Page_006.pro
b5f4f1757cb8daa2f41e6d180859035b
33eaf3ca281fe50059584306d1f674a5334f9ca3
F20110115_AACSPN zhang_c_Page_092.tif
c9f92570b52ad3082711df833d45673f
e0030c036ffbf81e6bb8b92d0c532699dcdc6861
69561 F20110115_AACSQD zhang_c_Page_007.pro
998be967487571a23ab91b348ae6ef49
81d870f24c6b4425ddeaa630edd48e79e2ec293c
F20110115_AACSPO zhang_c_Page_093.tif
72345fc59b426599389eb8b3b4b12ab2
2f8ed40956b123cfe293998b4b03c1832c11bcf6
78457 F20110115_AACSQE zhang_c_Page_008.pro
aa0adbcca78ed8d9e953847a03e36a8b
7eadff7a09c3f156e08c8bdd2c288b9ae5e7540f
F20110115_AACSPP zhang_c_Page_094.tif
93feaf7b56e62f1fc808e899aa8b216c
fd9589ce3c11e0eb32dd22782f2634e5659c0b2e
31726 F20110115_AACSQF zhang_c_Page_009.pro
86fdf8dfb9789ea8121a5e00b0316974
e9653672f1e938463b55b571c4a221b0176907db
F20110115_AACSPQ zhang_c_Page_096.tif
4ed06e5c8013de22cfe1a6f1b795da52
adc1d5a1f50578736dc467c032ba1b3db11a7383
40841 F20110115_AACSQG zhang_c_Page_010.pro
5ab2cda978d12b7498c481534f4bbc1a
9690817d91459e7b429c369c3a17aab1e5133213
F20110115_AACSPR zhang_c_Page_097.tif
c91219d40002dadfdf8da0989c16cb61
5d2a7f4d38cd3766baed4af31034f5ffba489ffe
27401 F20110115_AACSQH zhang_c_Page_011.pro
c5d48a608e6c504af89b97d4679ff388
142627356fe5a7efbdc77e4841d274d34d813e17
F20110115_AACSPS zhang_c_Page_098.tif
6605e132eee913863a03d253d9263245
01963fa961350e3cd23ac56ff24066432f51ed15
44632 F20110115_AACSQI zhang_c_Page_012.pro
55a06957a046ab05f4295efbc26eaab1
c00e3ed31bc22bc62cfa9f8fd55a1f8de573514c
F20110115_AACSPT zhang_c_Page_099.tif
3748984c6857cd6513342c1a48430bca
91cc02259b6c87054da21608659c333e7bdc2f63
38965 F20110115_AACSQJ zhang_c_Page_013.pro
0c1a3bde69110f2692f2cabc1d801133
c9498716fc93b7304e011fc112fdb7e0f1b78327
F20110115_AACSPU zhang_c_Page_103.tif
1ddf88fd3297e0033670aa284797332e
dea37e815f50856fa08488ce3ebab4680f0ce48b
F20110115_AACSPV zhang_c_Page_104.tif
940dc71b5677bfd55e4c12aecc5c6bab
a57f2cc8099108900494d47da4c9f93930ad1aa9
45755 F20110115_AACSQK zhang_c_Page_014.pro
64ab8fb21b67183307bb0e656b851e2d
cee6fc274f31c072be204dfc6b61d1a1a6b673e8
F20110115_AACSPW zhang_c_Page_105.tif
e5cdf0eade7f8f94e7307c8202b8d8fa
d3656c184dc2fde878748b024a6fdff30a42ca66
43917 F20110115_AACSRA zhang_c_Page_033.pro
5d2289a3c1c5e4a69c8bc2f713e4a61f
165278c4c52d268313028c62b83b8c23e771b698
37880 F20110115_AACSQL zhang_c_Page_016.pro
bc1412773de914ba6b7d4bb674cee2c3
0568e992607efafccafd7a110f0031af8cf357e1
7778 F20110115_AACSPX zhang_c_Page_001.pro
4e0fbb7897b8cab4479162d30d76d16c
eae71bd67bc4304d33c7f63faf27aeb23a59fb50
52818 F20110115_AACSRB zhang_c_Page_034.pro
f94e026e61638bd7dbe49e0744adf049
1dbf401e057492ff2550caeec1962c41f39ea098
31516 F20110115_AACSQM zhang_c_Page_018.pro
73c74ee91d31ed3590884a70d6c44da4
54c4c41664582efbcb0ef5e8ea62b0c4046ed26a
1043 F20110115_AACSPY zhang_c_Page_002.pro
bd61a99d843c7063b10578b56f6b596a
a64cbfa38a72b26d2ef427dd48ece8900b656832
39847 F20110115_AACSRC zhang_c_Page_035.pro
511777cd33cc097ace263083aee71fde
ad097563c10abbd33f70084e240bf7148da06d4a
40874 F20110115_AACSQN zhang_c_Page_019.pro
6286d601a5c3426cea21946195a9bc97
2556c881a49b86959645e161770bcfabca523db7
649 F20110115_AACSPZ zhang_c_Page_003.pro
e04a7135637fde5e9a033c8f08be5ef2
d4dfcbdf7f5cf041a30d64c2a6b19bae598ea52d
29571 F20110115_AACSRD zhang_c_Page_036.pro
7c55edf20cd1f12a406d5f66db75b352
893d6c0abcbf95c844057542edd11df5f9e7a7da
38448 F20110115_AACSQO zhang_c_Page_020.pro
00c06701641321f520fc5b26b5e4948c
2a7de631b610726249267d5fd7d4d46e34997406
31400 F20110115_AACSRE zhang_c_Page_038.pro
21f95de8f64bc41d63f7b4c5c68e7d47
b68cd90e8b7165e515975b767202dc2dd7ae857a
47241 F20110115_AACSQP zhang_c_Page_021.pro
b60194b94d78b6aa07058e8f71e54b3c
5ba6b6749f106217bacccd0fd56fa481e315542d
11026 F20110115_AACSRF zhang_c_Page_039.pro
a8d82c8a2b41899cc4ee63c16103e39e
76d51fe993b236fd8b7270493cca6b3dace80882
39472 F20110115_AACSQQ zhang_c_Page_022.pro
f091f551dac27f229b6c38d485a9ab2d
4978cdd8ca854b599e13dbd75bcf41c79ad0cc46
50191 F20110115_AACSRG zhang_c_Page_043.pro
5cff3b33c555862175e4fb04cf5a980b
372c8e79c3ce2f80e95be6e2cb5156603fd855ae
50027 F20110115_AACSQR zhang_c_Page_024.pro
4e58e7b94c82e418d78fab96c58dc665
201b4ee5e2c052e9d8598d6d0bda1be4127a43ef
49889 F20110115_AACSRH zhang_c_Page_044.pro
ae8a26003a69c8a314d99ca89ae895b2
810fc993c683219f685b4e78ff1bbfce88c8f033
30934 F20110115_AACSQS zhang_c_Page_025.pro
2b440689a658435371c77d569435e2b8
172e0d7f0ef3f5c64f770088c79e3948e00d624f
38470 F20110115_AACSRI zhang_c_Page_045.pro
dfaf420f9bf8b83a1e889d20114c262a
789aaa98ca09fa6d3f0b137d8a6bb8eb56a1a554
37091 F20110115_AACSQT zhang_c_Page_026.pro
e2163eea4f18756b37ea3686a7605e70
e48925ea861200250f955ed437f106ef8c041268
36909 F20110115_AACSRJ zhang_c_Page_046.pro
2e6d19f9229d176fd2a1f20bb43289e5
a15b73c5c520e1d7787048502d7c82eb2654787c
35143 F20110115_AACSQU zhang_c_Page_027.pro
dffd0e12b2dda051d65216bf4a3c8c10
fbd9ef1486e07b7630f0a9507e4615994453757a
46396 F20110115_AACSRK zhang_c_Page_047.pro
76ce6d3ddd250cf4368b85acf8351bee
780f1a6ab5dbd63001d22e0d0eb17be5b5dffc65
36392 F20110115_AACSQV zhang_c_Page_028.pro
69922da81ee1475a027c4b838e715b79
aac4ad3554b044381e86f2bd51f2af20b0588547
38224 F20110115_AACSQW zhang_c_Page_029.pro
877282b9f78136c829dac286b41cbf98
f94ad844aad968d5f14486c18dd927738ee5f8d2
42284 F20110115_AACSRL zhang_c_Page_048.pro
9b319884ae37f7e4e4a30b491eab80e1
11e21461ac37a65bef011839a75ae9079d3c6719
31092 F20110115_AACSQX zhang_c_Page_030.pro
3c64e9265230f4ff95d9c60889f42edb
5464485f1f8f9a1d4f05d89307f8e043c61754c5
38014 F20110115_AACSSA zhang_c_Page_064.pro
7810eddaf355a6fc0351a5649db36d7b
461bd57919c0e69e4b6c860e52b0c20aea53ebb7
33621 F20110115_AACSRM zhang_c_Page_049.pro
c42e5df70bf8030f8c1277e8f515ea2f
8ed542eb0030bbc5991e0b007e6015d59040d4f2
39337 F20110115_AACSQY zhang_c_Page_031.pro
41c8b4d132b1e7878d44d33b6334c37f
ae82ca9f25f135e0dadba76eeeba34f9d66e61df
52319 F20110115_AACSSB zhang_c_Page_065.pro
79b309700dc0f4413beba8d359bc4a11
9cbe3a09294fd8732afc634eeed12bf97f0a5c76
32215 F20110115_AACSRN zhang_c_Page_050.pro
f24d9287672f74219e103c1a4b6a0724
ddea50974e3201548e7c4b197a60113534ce3541
50163 F20110115_AACSQZ zhang_c_Page_032.pro
5f97e44463f710eb454c01109fe279d0
64fd7c15fd4707d325cbeab0254d1caae456cd3f
38179 F20110115_AACSSC zhang_c_Page_067.pro
610412ce47739098a254bbe67389bf45
3fbe51fb2727c09f99df7ea70360ad06abb504dc
46209 F20110115_AACSRO zhang_c_Page_051.pro
f9d632f4479b0945914091327d864eee
1f03c6ba2338bc1f137074b9b67fb06532c393eb
50143 F20110115_AACSSD zhang_c_Page_068.pro
13c788ade5f5807d7936f2e26a442cfa
ae66894b4b7c443d1c35b440d1a656ed463f9a75
32349 F20110115_AACSRP zhang_c_Page_052.pro
9643c1afb76e6157f9c3d9074cae8ff6
6c59e12cede36ef1f3733edfb680dfb432dce8af
9911 F20110115_AACSSE zhang_c_Page_069.pro
38cb6029aace6aa61bbb968bec996e0e
ae94cc927257e5eea92d48bfe01f2f461713a666
41070 F20110115_AACSRQ zhang_c_Page_053.pro
fe4ed7630bf9e91482ad01b1ac86c86a
c7e2afd577c2648b34366a5e37464d45921cf745
16406 F20110115_AACSSF zhang_c_Page_070.pro
ba8d4cdd85bd42739a69d8fde30f018c
df3468e49b2da7eacf136a8a0e110beaa9b4a035
41728 F20110115_AACSRR zhang_c_Page_054.pro
aa3295c7c025502b1e7a6e04f59e70fc
7900eaae00bde6701deb882c15140c518eff11c4
28910 F20110115_AACSSG zhang_c_Page_073.pro
6cead54caf2fb4aae9f26297cd0a5c9f
7853601dfa29905f093583da3e0f9c6d68102293
35788 F20110115_AACSRS zhang_c_Page_055.pro
b7b1dabc4fd9bd0bdd94d8278ca2e812
1be664f4d7031c2bf70c760db882d8e7b9b6c2ee
50477 F20110115_AACSSH zhang_c_Page_076.pro
83eb6cae1d2a0c3050c0e7106a8d496d
4422dabe685d6d688bffb650b173c86b227131e4
52809 F20110115_AACSRT zhang_c_Page_056.pro
b8d956497d0fc8604e7873a87129708b
eabbaab43959fd8095a278c42f1341e8950cedcd
50171 F20110115_AACSSI zhang_c_Page_077.pro
be09e56a9a6bc031ace460cf5cc9d931
b6cd9e0f5f2c3423f9b7c0d9e7cbd47fcebbc32d
41594 F20110115_AACSRU zhang_c_Page_057.pro
7d659b9f09dcbd5d6b72b898ec293200
0c7de29aa3545e07445c7cc28115a8341c1846dc
41600 F20110115_AACSSJ zhang_c_Page_078.pro
eff6bce741342516ddfdeb05c44883b0
ac41ecadeb348934d5112d2a11ea32948ff2e8ac
52390 F20110115_AACSRV zhang_c_Page_059.pro
75b451ebf0d7444f1c406e818ed62bb0
ffd9b0bcda966f0dc92cb724eee4a95099faadde
44084 F20110115_AACSSK zhang_c_Page_079.pro
602bbfbf5afeb1eef10a34aa7ac2e428
b1c51c0df3a2196321ddf08c0f141a0d3aa39e4c
47245 F20110115_AACSRW zhang_c_Page_060.pro
bbc54aabb496abe3150cd0b2e255df60
5256669d4912e2bcfbb370a8f590819bce0e7179
34503 F20110115_AACSSL zhang_c_Page_081.pro
e54080174e829c73eaf4e414a3a287c2
aa116bc3c32118f0bcf8c633851d7639c948675d
15963 F20110115_AACSRX zhang_c_Page_061.pro
ef00ff81dd9622b3e5b48ce6978794aa
d3c8d28fcedfa492147c3ba5d597b90ba9b00e14
47512 F20110115_AACSTA zhang_c_Page_103.pro
10bb21d9500130dcee2fa8b587fa2488
319ce5f1d470b9ae3abde7ebf80dc81ecdf143af
15318 F20110115_AACSRY zhang_c_Page_062.pro
071f342fb451eb0d107852351054c724
13f9c004c2aa71396adf279f59cffd9cd18cb017
47168 F20110115_AACSTB zhang_c_Page_104.pro
1415efe2601d0c3cc03390c234959dea
aa12f9c2cb213df91f27db125f03d7cb6a17120f
35931 F20110115_AACSSM zhang_c_Page_082.pro
1a04ecaf03868d20b96b71ea28e9c23b
9b5ea4c18cccc61ebd9be59c7a7ada2ae2efd49b
8743 F20110115_AACSTC zhang_c_Page_105.pro
79b029de07cab95577f699cd9dd196d0
a518b45a92e20e303200042c947f8e809f1abcf5
40449 F20110115_AACSSN zhang_c_Page_083.pro
80dbe652e06028ad3a49c84bc145f270
446d83324192c03701c9d4c9548f74a10380e44d
49178 F20110115_AACSRZ zhang_c_Page_063.pro
759f9b210561d1784625655bdcfc299f
0282dca17c5deb2e9541c38859e8bb0f16fa7d88
424 F20110115_AACSTD zhang_c_Page_001.txt
8c99a97255c351c284d150e26142229b
6ebb64b60ddbcbfd06ba9436e30fdcd465bb058b
41147 F20110115_AACSSO zhang_c_Page_084.pro
a179a28bc43d062181b218ad87bee23c
10c3df38e3667e987999d8ba971d7b9905dc557f
93 F20110115_AACSTE zhang_c_Page_002.txt
677111f3bf6950cb1d0f7287394d3711
b2235f589e8a0b73981a91c56ebfe4a84428cdc2
26410 F20110115_AACSSP zhang_c_Page_085.pro
28af1ba7583e07c434cac110a2f15dab
2b23f2be70ad2bdaa7dc4c875edf655452b692c0
83 F20110115_AACSTF zhang_c_Page_003.txt
cc69670c78eae3bee9fcaf02ca3c64d8
026ae5b48d0831dbb8ecb9fb3dc81697b02b700f
38301 F20110115_AACSSQ zhang_c_Page_086.pro
b6bf6232c6e5f58cb4e2858f92eb30f0
bc2d6dc67aa5fb55cd4f0136969c44f76abb11a8
1306 F20110115_AACSTG zhang_c_Page_004.txt
a413e8be25cc4ff894a872305b5f264c
7a1308ac107e5aa657dd535e98fba31ba7286eb8
38420 F20110115_AACSSR zhang_c_Page_087.pro
214169539c789e6153eb77eb540b3153
640fc8f6d48f6a22ffc2e343b24ca43143079d8a
2620 F20110115_AACSTH zhang_c_Page_005.txt
ab56786ef7f609bfc15936182f225175
200108ea1e86bfd3bf234dcef04ac9e8fd45b7f3
39858 F20110115_AACSSS zhang_c_Page_088.pro
d1151478264b892bbfb2a7d0786317ac
96a045f7e8b56862d9697a58acefd363602306c8
711 F20110115_AACSTI zhang_c_Page_006.txt
7b8306dc2413ef98b38ee76651e380c8
19ad34aa2d4d7b8c8b4e106fdfcff137eb92a535
33464 F20110115_AACSST zhang_c_Page_090.pro
2688d28d3ca795208ad284096a519741
7167a59a2d96c06b62b9be9d45275a4cd0ecdf94
2815 F20110115_AACSTJ zhang_c_Page_007.txt
7eea62b70e9694e0473f72d455a7c84f
b4186b7bd19f585b2104429025c0943ed76dd684
22992 F20110115_AACSSU zhang_c_Page_093.pro
c95fe7d0ff1e1b021efccd52eb0561e0
ec899461d06a71e0e2e44ead0434481769fc07da
3153 F20110115_AACSTK zhang_c_Page_008.txt
3933665a5df0e9f6f277597ed65c8041
e3415ea76bbe66e1369ad7e4f10a0172fac58dfc
42986 F20110115_AACSSV zhang_c_Page_094.pro
e0fbd1ca6dad06274ae7acec85ab9169
df781acd52fad8e1ff8042c083ae7882c06b1dfe
1276 F20110115_AACSTL zhang_c_Page_009.txt
18ba1fca5916f2e38613a24042b0b826
c3f902dadacd418e5179e2b6da307319c95c9c1a
14007 F20110115_AACSSW zhang_c_Page_095.pro
8f64797dc6b0dc6153751f834cdf6acf
c1bc47d52d0155d1ff72fd0877c0b64aa9d4cbb6
1761 F20110115_AACSTM zhang_c_Page_010.txt
e268b865896a3e982c86fe1dfb742329
06c11fbd1083efa955601ad28a16c09b1404189f
34152 F20110115_AACSSX zhang_c_Page_097.pro
4d84ffd90b7365fce78e8aa863413e37
aa2183cb9e95d01697ef7e0c5c08b242547084f2
1785 F20110115_AACSUA zhang_c_Page_029.txt
7d747d9a54eb5d600a1eef4d38cac233
a4b176f9dfe0c0b9324a4eb24f794b4798d8cc7d
38596 F20110115_AACSSY zhang_c_Page_099.pro
4903905d956c5c43011d9fcc5fbb09cc
77730de31675de29669f3f483fdfdc0b4c846791
1932 F20110115_AACSUB zhang_c_Page_031.txt
662a07f643f541acec25d9bf22c37cbc
eb618fd0666a4ecae35f0644bff07158cc25a6df
1599 F20110115_AACSTN zhang_c_Page_013.txt
7280bdd9216e691a7d0d54edf9f9a308
62522f8d2d71e532a47689500f25dfc24c72ac1e
30164 F20110115_AACSSZ zhang_c_Page_100.pro
8ce4379ff61ce96c9afe11b5ae87eb5c
835a6c6c6f6530d01abaca9d0acc25262a6717a9
1973 F20110115_AACSUC zhang_c_Page_032.txt
df7e2b7689646d83f58ef2f68ce953eb
1f2e61dddfdb13407d11ead3c3f7f14200323082
1538 F20110115_AACSTO zhang_c_Page_015.txt
40b113d7334a0bdfafff399bbd4028ea
7d819975c3f5c350b27d8b529743c80b1eb3c8c0
2255 F20110115_AACSUD zhang_c_Page_033.txt
0b6b5ec86e9a5e60210bb287f4e30c84
8281fc9b25e829fc91763afaaa43cab04cd97b06
1795 F20110115_AACSTP zhang_c_Page_016.txt
4cfc6cb689a8d222b35e5eddfb28e7e1
abdd4de6f2ac5a691ce9df2a0653eb6d251f5b5e
2108 F20110115_AACSUE zhang_c_Page_034.txt
e3b6a5cfc57bd75cd634399f7a4ad383
8a217fd19003ccec3d3a0ab5ed3f30220c69e084
F20110115_AACSTQ zhang_c_Page_017.txt
ee995d7dc8730f8607ee130594855c85
f721f25aee78aa3e290a80934eb1e59d543e1bdb
1714 F20110115_AACSUF zhang_c_Page_035.txt
d03d61872f11d972adc65fc5676ecb85
188238422c6424ed45883fa4117a64dd2b11dc5c
1778 F20110115_AACSTR zhang_c_Page_019.txt
d233c58606b2adb21215f2d2bfa5e725
a8b74829835c5db23deb887e35da3d67b9a85474
79309 F20110115_AACTAA zhang_c_Page_056.QC.jpg
beb4cd28820197821d2273c39112ce40
350b06af3a8a18a068951cbb50c4212cf7484617
1462 F20110115_AACSUG zhang_c_Page_036.txt
d984864e721453b086f0f1ef4321b245
53bdb781cfacebcbf2cd6e4c37773b2cf3997dfa
1733 F20110115_AACSTS zhang_c_Page_020.txt
142575482f892bbede72301d9e29bbc7
58484f7ee35dc5e67ae0f5238d7369d2a0cd2088
21299 F20110115_AACTAB zhang_c_Page_057thm.jpg
bbbba435aa63add07f831abe7ddd9195
7eb155c5aa5de8868299c070b8b07dcb4648076b
663 F20110115_AACSUH zhang_c_Page_039.txt
fc00efa0c3b94a0558f71128d8f0ce06
3f53aa8be63179654d92f11bd559d0e82c6b5de3
1975 F20110115_AACSTT zhang_c_Page_021.txt
581da266210b42281b5b02c74a2d5183
6679e8277e9ba0c4390d73175bd7eeaffc7ed37d
63128 F20110115_AACTAC zhang_c_Page_057.QC.jpg
e359239e7ac21dd5c5b2e6ac5ad4b9d6
f75e7acb900b3502647d3044b98ec2ec7f99aaaf
227 F20110115_AACSUI zhang_c_Page_040.txt
b37b5b0d505dd71671b840febcc11950
a15b3079f4b3e930698b4916ff9d125fdfd458d8
1806 F20110115_AACSTU zhang_c_Page_023.txt
fb6853e98e8950bd42365a6470f9ce38
8e5bd92980db4bf116e419974d4744173badd57f
61801 F20110115_AACTAD zhang_c_Page_058.QC.jpg
4e82c9a5e6cfd361babfed10be51982b
08f4f2d814d8e095979febf1fed36da388eca78c
1875 F20110115_AACSUJ zhang_c_Page_041.txt
be305b505b44d63ecedafa6ae3c7e864
03fccd4e01ec53a276ade973029dc986169b9aa1
F20110115_AACSTV zhang_c_Page_024.txt
6eee21c2069d6b6c7562317ec340265e
26e4ec9922ac14e1cbac0cac086e0fdd5486269c
23213 F20110115_AACTAE zhang_c_Page_060thm.jpg
8fc3e221074d698e2b0747b8166725ab
758c3c02f4cea5b746b3d68d8159b64334bef861
2021 F20110115_AACSUK zhang_c_Page_042.txt
2d06eb79c812f201c15f2ec5f3ff9e5a
a99e5001c0faa92d248d0edfd3de9d232dc64468
1313 F20110115_AACSTW zhang_c_Page_025.txt
8d6e6195b9204338473a15b312275180
cd6f79473cf95d6e7bab5b0ee235ef34461ed764
73379 F20110115_AACTAF zhang_c_Page_060.QC.jpg
d3f63f25f2856f09609feb1ecb9858f5
8c3c230df4dc8ea2f3a8a4c363dc77f15e3046d1
1982 F20110115_AACSUL zhang_c_Page_043.txt
b76148a822a10a9db71967483f6b9ce4
b7cc2071160b406a42aa91b224c58ce49d5ad732
1709 F20110115_AACSTX zhang_c_Page_026.txt
b2d667e0e2a29263629399d95000c34c
7b1210c0d8ce8fecb9c393ba2b10d39ba913ec70
11633 F20110115_AACTAG zhang_c_Page_061thm.jpg
f61afd3a440921ece210dce9146bdaca
2bdd4636d5f4fc1c9f19856853b9a7ab5786a6f7
2064 F20110115_AACSVA zhang_c_Page_059.txt
169978f347c40396902ddccc135829b0
7f1f5163be8df3cea4010e21b02f7a77c4f0bff8
1969 F20110115_AACSUM zhang_c_Page_044.txt
b9e0886592571652d4139be11fae3290
acc516839feceb19091cb3ab039c30cbd835eb93
1681 F20110115_AACSTY zhang_c_Page_027.txt
ffdac107273fcf72be05c7bf76e03881
604361fed9d8a6c9698c72d85cfc365337c67a8f
35463 F20110115_AACTAH zhang_c_Page_061.QC.jpg
8baa515d6992b10742ba77ef63321fe3
f77bdc22a266d8b9415c6e74082a0899dbe8c3e1
1957 F20110115_AACSVB zhang_c_Page_060.txt
7c5ff6f5f137267efd446c8d39167f8e
b83e54fd0e958e6b0c7f243562287c8ed31e2408
1777 F20110115_AACSUN zhang_c_Page_045.txt
9d1c38057ca9e2e60f34bd63d51880a0
932c252dce5f20443958f2541201eb5513d5ea25
1802 F20110115_AACSTZ zhang_c_Page_028.txt
494655b1518e23c232260f389956709f
904de8960c24d26d56afa6ba0db07520e75bc9ec
15585 F20110115_AACTAI zhang_c_Page_062thm.jpg
9b1ea327a9c09a5e44b00d16a4f4b3a5
9cc466bc5a85e9a6b79b944fa065fc90a3a231fd
744 F20110115_AACSVC zhang_c_Page_062.txt
0bfb3d8615fc85f7dba4b66ae92f7f78
09ee8d592bf53e548848e20c53d77fdca0e081af
42353 F20110115_AACTAJ zhang_c_Page_062.QC.jpg
3633955bf8a181323fe494b1a13bef03
f035f19133bd8822529581840035403ae45968bf
1939 F20110115_AACSVD zhang_c_Page_063.txt
a2d7e24a997f44eca5fab95f52a4fb42
9f1a6a55217fc75ad12ac0ab6538378e979447d0
1838 F20110115_AACSUO zhang_c_Page_046.txt
67c9980a8cc0a6f49f8ed327252916a5
afb78cc1a95fe7843f10feceed42adc596e58e63
23757 F20110115_AACTAK zhang_c_Page_063thm.jpg
fe3c62742675a14b643361f1c459bdfe
3e4a8af52d3bb5ce7b44f9ef39cc95fbb94a7a3d
1752 F20110115_AACSVE zhang_c_Page_064.txt
db7ae07fb75898b5d712980429434604
6f72f5824e0610d5bd5aa0e743118481c90f54eb
1915 F20110115_AACSUP zhang_c_Page_047.txt
43937f11558779d2709b1191ef486b9b
968a99e550371616c7de211714a949e8a127c9d3
19161 F20110115_AACTAL zhang_c_Page_064thm.jpg
7bcca03ad1e7157be4d74828ad4ebba2
804f21df23d348f8c49780c0ff44fa08411f52ff
2050 F20110115_AACSVF zhang_c_Page_065.txt
f86deda2ce2ababb659c1b62b9bd98cc
8e16941df54050e62f81c686b70e5abf8662f67f
1867 F20110115_AACSUQ zhang_c_Page_048.txt
9a2f67d0e2ece35cc2c2cee9bd60e9d1
7b7a1c0930fd6e977dc2cc97d91541531eac237e
25812 F20110115_AACTAM zhang_c_Page_065thm.jpg
9c05de47ea1af88e82ae6234173fd893
4fda8db8385ab907dea980f6b14870531204d1b9
1870 F20110115_AACSVG zhang_c_Page_066.txt
e1ad896e1293c22f34bd6e77235fe2e4
042a22b9ffc7dd35ce81d4032761901a8833f087
1630 F20110115_AACSUR zhang_c_Page_049.txt
6c00643fd082dd236fa0f77ab8b2ff9a
0bf2d70067e582d3a409e110c9cb8c661184363c
47461 F20110115_AACTBA zhang_c_Page_073.QC.jpg
2bcee9839604c373e3ac47c6eb7d99b5
1bc9994eb210695b977fffddbf87fc6ff4fe5847
77774 F20110115_AACTAN zhang_c_Page_065.QC.jpg
d3bec2c792b98fc8df23d7631ca2ead4
e1a3c574509a3161a89888294822664e61ace595
1775 F20110115_AACSVH zhang_c_Page_067.txt
21e8650bb8f75a7547faecfaf4fdc07c
6edb3461088615a51b3d18763d6dd3d54066ab38
1968 F20110115_AACSUS zhang_c_Page_051.txt
b039b07d0d68c87a895bdfde65dc30b3
c052ee2bf0cac2e44e9a7192e751c48ed248e2d1
21881 F20110115_AACTBB zhang_c_Page_074thm.jpg
d38441804feabbe108778a814b7f766b
aed157090bc8b6f3c5dd7e9fbb4ec8ef6fd48f18
19326 F20110115_AACTAO zhang_c_Page_066thm.jpg
d29097a8426d73ec7721c62b3e44d9fe
3bc3b8bcfb0a27342a01dd2f960c70f9ce92e8ef
1980 F20110115_AACSVI zhang_c_Page_068.txt
b5705cad52b26a780e7af86fcd7909e0
4717b869f5e7cfc3541202c724daefb1560ee1c9
1682 F20110115_AACSUT zhang_c_Page_052.txt
aefcac0d603fd4e6f0ef74f417b9c58d
79bb70f30f0a2861268801c70a727327d6cea3be
67483 F20110115_AACTBC zhang_c_Page_074.QC.jpg
a050c6aa0543011f4472cbf8848dc1a2
dc08f50c6a28698c0e6c9344ce9e1a9c5e9fe630
57564 F20110115_AACTAP zhang_c_Page_066.QC.jpg
e847eaa3296b4176acd26633e3fd54d9
57c82263f29e33bc2f02e9067daca10764af12a4
511 F20110115_AACSVJ zhang_c_Page_069.txt
4d306904ed103ca9ce3530dfbc4114c1
637639dbf6a652a38ef0c89b62dbc7f33d0b1a0f
1891 F20110115_AACSUU zhang_c_Page_053.txt
776de85bfa4391a3323c398a5662cb30
fa7a1f4e3d317d8bb17afb427e06d9066cbcdee2
22419 F20110115_AACTBD zhang_c_Page_075thm.jpg
573c1399754c752410ca3aced33f0f08
e7df7e85749ddd06bcbcffc27178fad24a0f4785
20301 F20110115_AACTAQ zhang_c_Page_067thm.jpg
9c9861a6309f76a635266e0fe27124e1
d0285566746e2dc97705072269d24d7fc0cf12d6
902 F20110115_AACSVK zhang_c_Page_070.txt
5c132c4d2443baa02549a611b27913f6
047b772123e362bbfcc37bd760ce2360d8ed4e51
2037 F20110115_AACSUV zhang_c_Page_054.txt
74c60c86708eab85dc665985edaa6f40
9e4aa243b9f74778fd84a69a4e07b46d905a78f0
24589 F20110115_AACTBE zhang_c_Page_076thm.jpg
68944b76e5cd32da8a61b01a60d246a3
846154207246a154334c43c5ac5072d38c22a359
60537 F20110115_AACTAR zhang_c_Page_067.QC.jpg
83f1a5b7a93694a0fa2efc1850316018
6e8b7b3b750de7f6759938aa1325535e3ea81c1c
1638 F20110115_AACSVL zhang_c_Page_072.txt
2f6ee4966c4381f4c596b4108fd34d55
14db64ae45666324165abbf5416ec4990939f049
1755 F20110115_AACSUW zhang_c_Page_055.txt
71797757c823915926ddb6c285e59a6f
534ea4a6532f0827574f406a09f78a15d57c8fe7
77857 F20110115_AACTBF zhang_c_Page_076.QC.jpg
d1403094cfd7a76eaa64622959824d70
bea0f3dea03f7077c437b0b679fc9fa0a4166fd2
1395 F20110115_AACSWA zhang_c_Page_091.txt
0ddac0e5ca1fc25453defb3f586f1dc9
46fcbe3bbb286cf285a99b92781e0a66fbd6ba7c
24390 F20110115_AACTAS zhang_c_Page_068thm.jpg
5d888c7753dbc219deee924548f6f83d
6665be187d7641dc62d96ba76462117455d0a0b4
1147 F20110115_AACSVM zhang_c_Page_073.txt
3b8bad137085994be31f7ea3dfa67493
14f4275b975eb6badcb491f6090ccc9921200b63
2068 F20110115_AACSUX zhang_c_Page_056.txt
c34190d2404af3dc9ea8ac77bbe72aca
77db7a4de5749884b51c140624ae374b7bb00a52
24850 F20110115_AACTBG zhang_c_Page_077thm.jpg
0c81d70a1cfe03e4e82aabc840a9c4b0
bde5eab8fa8772e41c1b9e449df3647572db16cd
1033 F20110115_AACSWB zhang_c_Page_092.txt
56180ddc4ffc2fbb7f4470ea5142ba7e
b030c831c19c44e76b5c0b029beefeec8d393f5e
75426 F20110115_AACTAT zhang_c_Page_068.QC.jpg
e5991bd4621b93ff5927f9aaa5f9bceb
5bac43b5fc79142eff0fb0f2b2d9a50b46b87f53
2012 F20110115_AACSVN zhang_c_Page_076.txt
03e63376102d1d226ba2975dbebc6da2
26ac494d304e01b621e42b7e301f0a8724f823b8
1879 F20110115_AACSUY zhang_c_Page_057.txt
ad2231759158fef4ef4389fe96f3836c
85d75bc27c74af950ed7d59db852344ca414642c
78507 F20110115_AACTBH zhang_c_Page_077.QC.jpg
dd288b6a205332718b19131cb7978ceb
27ff6fb413ab4a164613ebce71767e5d8c5f2b72
962 F20110115_AACSWC zhang_c_Page_093.txt
ce3c162b196644d662418faee4b8f2cc
ffd06a2037972d593e55c3c7c890081f694fc8a7
37426 F20110115_AACTAU zhang_c_Page_069.QC.jpg
fec66d7e660d395386566c4e7aeea4fb
a35e8d5758c37793f4c9424dff82fa7a5a5f16d9
1979 F20110115_AACSVO zhang_c_Page_077.txt
557a13e1f05003794ab4ebf3c13dab41
44cc778c91d16793fe161a6ffd33528288f88e06
1781 F20110115_AACSUZ zhang_c_Page_058.txt
2da05d4289e9446a15a9eb43908bf3a2
f2eeb9b46b2fc36755885cf422be2c161fe2e3a3
21749 F20110115_AACTBI zhang_c_Page_078thm.jpg
e05d81d84592d31b5acd87509bbc9e1b
5b082d920e1a96f0eb1e6c18b312efbfd2acc0bf
1769 F20110115_AACSWD zhang_c_Page_094.txt
705f0c08cfad56bf66a2bdd7d1123e57
616e2aca26bbcc1e987e24a40369c945952023f8
13512 F20110115_AACTAV zhang_c_Page_070thm.jpg
a3309ee149df18b04bd6ced102e3162a
60ef8cbea9c7d1d52e64cb56acbee095ebc52c83
63995 F20110115_AACTBJ zhang_c_Page_078.QC.jpg
ae9948fed3073b38974418a2e0d63a27
fc5c6687265b33505e75d5dffdc26a73d425a915
604 F20110115_AACSWE zhang_c_Page_095.txt
2b9316d5d1c51cda0804a344df8451c7
a431dd77ca581f53dd77f1f803dcaadfd7ae0ee1
25959 F20110115_AACTAW zhang_c_Page_071thm.jpg
838267ca82cadd39cfad8a6d9254460e
3f1863f210e832b3ffcfa252cac8d123649a6f19
1850 F20110115_AACSVP zhang_c_Page_078.txt
71bc086d754b378a31002b47ce30b9df
9fccd2e4c0c950e7bd00b68f92540a7f8df6c81c
22772 F20110115_AACTBK zhang_c_Page_079thm.jpg
cb20fbaa496df7216f69572573ac0141
9d710c0a9de7c62f62259f3656bcd73e2f1c9e7c
1359 F20110115_AACSWF zhang_c_Page_096.txt
2a9be71638ee55ab154ff49a894bc915
a4abe35626836d2cae465fb9b38507981a3419f2
80212 F20110115_AACTAX zhang_c_Page_071.QC.jpg
47c2c722b0c5280f110e3ef68e69ef5c
2c6ac43d45c3379da987a4732b30e38871b3515c
1794 F20110115_AACSVQ zhang_c_Page_079.txt
da94a0d8942bed421ee66ad4488692c7
bb26e2a65bbf199761b3083cacea8120a3907304
22816 F20110115_AACTBL zhang_c_Page_080thm.jpg
50c3ba6a685bebec40afa19ea918fbf1
1d7a97d32819bd383e442651ac2e368d58e526e9
1543 F20110115_AACSWG zhang_c_Page_097.txt
20f63f97c0c0c969604450302078b6c9
4df30fae7cfc88bcb0b183e499ace882a7f24351
20648 F20110115_AACTAY zhang_c_Page_072thm.jpg
a7a7cdd85f9c576f49be3604f4fd2579
abbeacd25294821d39c65d8a1edc25bec6401dec
1732 F20110115_AACSVR zhang_c_Page_081.txt
637c2f5f26d0f9de53603f262719be00
6e3c628f59b4828eef1271ffaa4b8d285203e1f5
60648 F20110115_AACTCA zhang_c_Page_088.QC.jpg
eefadd8020e8472eff9d3ae1a87c75fb
3b95f1dd3c28412fbd70a5a9bc976f8e2f50cb4e
64416 F20110115_AACTBM zhang_c_Page_080.QC.jpg
e60f36aecbd25e0dabe58034faf19c70
324d4e586f02d6d23b1f79efbd961e5d1c6ac185
311 F20110115_AACSWH zhang_c_Page_098.txt
0b9037b345885df949241ce20e2a7cc6
d66d8af0f0eafddf7ee49e42b9874976ea8211bd
1737 F20110115_AACSVS zhang_c_Page_082.txt
c261dd3bd6a28bd8fd965ed72f6df639
703a3b78a09d0bbb5902beb5881f19910a9a1a89
16326 F20110115_AACTCB zhang_c_Page_089thm.jpg
d82a5951636c3b3ed0d21b70336aeec4
2404146d52d6a9833a8e2ff0a4ce9a6be67fd9f1
20186 F20110115_AACTBN zhang_c_Page_082thm.jpg
600b544674acef7bf19703e4a82faaa2
8441e44d2ddfb837dd702c6232e1cb29552426f4
1849 F20110115_AACSWI zhang_c_Page_099.txt
4e0d9d64fb016e501b19a51a1fb2a46f
60d6c31b8ba7c5ddac2639d658994c03bbb4f788
63107 F20110115_AACTAZ zhang_c_Page_072.QC.jpg
d081185787d0502e125c2d251ddd8ddf
18867770a602eaf636134f8aa6a1fa5109d140c3
1912 F20110115_AACSVT zhang_c_Page_083.txt
b6bc91974620737b6c53867a3bbf14b9
90f667d5d822a08ee356ffec5092aaaea066a490
49859 F20110115_AACTCC zhang_c_Page_089.QC.jpg
d4ab8996aa497424eae136c438df562e
f247dc789a3906dd9a2f076dda5ee880456f0caf
61289 F20110115_AACTBO zhang_c_Page_082.QC.jpg
716919a3fcb45d1f7e96960921da8531
cd536c8bd03614384adaf15d5059515567cc2732
1542 F20110115_AACSWJ zhang_c_Page_100.txt
101a056ecc5af29b9ab9f2d3077f5f09
1ca4c2a06c7f837bf2462ae929db7fc2091936d7
1940 F20110115_AACSVU zhang_c_Page_084.txt
d08fe4fb061c40a9e6a2baf05320b2c8
522df6204abc7ad10eeef897de71f6cd61e55431
19225 F20110115_AACTCD zhang_c_Page_090thm.jpg
7c1826e6f086708078d8353c324e41bd
74a7b892e9384520b843605337b5dfb2ce596d6a
21639 F20110115_AACTBP zhang_c_Page_083thm.jpg
6352de5c3a7632af31eb4e0e823963e2
7ba5dd81ad5114a3029a78d323f28276696f0676
1917 F20110115_AACSWK zhang_c_Page_103.txt
d7803dc77dce6fd478d24e1ee14636fb
5b7dba6ad2998aaa61016cabe2dfa9759b14a0e9
1582 F20110115_AACSVV zhang_c_Page_085.txt
3d5a171d0a7075bc14f85f421b35bb6f
615a30bffbaa91b0a2c09da7749d491dc3748e0f
47173 F20110115_AACTCE zhang_c_Page_090.QC.jpg
e41d7046dcbb7683e12bc9081cc79136
fa41e943023670f4d61cc127552fe87caca328d8
62037 F20110115_AACTBQ zhang_c_Page_083.QC.jpg
db45b0d9b674bfc82b226ea699461b44
67ee873aafa0b0bf16ad02e5150a7c668bd7d1fd
F20110115_AACSWL zhang_c_Page_104.txt
761c0f87a61a45aaa377b9b9cb60d706
271fa7705a9513691ee0d2b25953bd5777ccdd89
1907 F20110115_AACSVW zhang_c_Page_086.txt
d325da7bde90787b66c1173b21058d80
2a1334780401c5641cf418d30c1521484b937028
9973 F20110115_AACTCF zhang_c_Page_091thm.jpg
f7d279080fd859d5ac44ef303053ae67
1f9d5c2442148a7cf487111d8afd041191532ac6
20838 F20110115_AACTBR zhang_c_Page_084thm.jpg
442122202763097dd774771bff56d1eb
015ecdafa1792241cf44b1b75744cdf483e8fad6
395 F20110115_AACSWM zhang_c_Page_105.txt
a8a15146633d0bd3814df42b0c5c1b67
c7678ddc110480202444c1db941f32bfcec0ecf2
F20110115_AACSVX zhang_c_Page_087.txt
7549176f1da157418d724530622e4826
1367c19c18fe9b3241a992e70ea4c2152fb3b3e9
26990 F20110115_AACTCG zhang_c_Page_091.QC.jpg
11a36c10791c2fd0f3e6ed56971f0948
5bad366a8c9c18f81ddb45200873827fb741c5c6
139036 F20110115_AACSXA zhang_c_Page_008.QC.jpg
155e8f156fad5af6a14c47b208fb9378
8bf51077a5ddb62b61bbeabd85d0a509dc10de8d
62480 F20110115_AACTBS zhang_c_Page_084.QC.jpg
5e34b2122597d265996e9403903dc560
d4f545e27a04e70b9c6c8f4cb5b9a5dff7aa09ff
6675 F20110115_AACSWN zhang_c_Page_001thm.jpg
7ff6436a9d8596124394429317f5fe2b
34b225b00cfc9cbe0e4a57e324ef47ae60822a51
1823 F20110115_AACSVY zhang_c_Page_088.txt
8c6a9339a9d7ccb33c6aea5716ad2441
c6930c2fcc19ff18c44bf86eaef7fbdf59c60a1f
13909 F20110115_AACTCH zhang_c_Page_092thm.jpg
638c0b25fc534e19ae5a5e91f351f95d
e50fc76cce6675cacc2257bbd566bce41565348e
40275 F20110115_AACSXB zhang_c_Page_009thm.jpg
f2bdc4b88a01ac1a3cb5ecdcccdbc4ed
faf7c1524feb967f344d1f0c66cb986324a49fe5
14744 F20110115_AACTBT zhang_c_Page_085thm.jpg
90036a8d80a5bdd2895a31960f1e9e43
5f8e0a1164c5a0c65061cda8a75a9243662d8e8d
1395584 F20110115_AACSWO zhang_c.pdf
8528fde5358baee991bbbc740c0e5979
e5094db6a98534f0d9a00e4fdcb9fabf9d5faf5d
1453 F20110115_AACSVZ zhang_c_Page_089.txt
028c6d44258261b4f66ddfe99df395d2
fe76b165b7b82f46266fe996a522da0303e75ca8
34006 F20110115_AACTCI zhang_c_Page_092.QC.jpg
3a5280201e735fc4b8f9c85be1a46bad
20cc6ea450dc7b8a6b3af2ab0a622062d43eb7ea
76667 F20110115_AACSXC zhang_c_Page_009.QC.jpg
febd5f8b255bcc770ba17ca1f91b3be8
27ee8e707f115528818fcadb07bb689caade66fc
43557 F20110115_AACTBU zhang_c_Page_085.QC.jpg
ced64763a9ca75c5d8b2e032f0125fa1
d5fe681ba9d3f92f26607ca337fc574e54be9b45
17624 F20110115_AACSWP zhang_c_Page_001.QC.jpg
b5da922d6224eeb46cf05a56f97f7ddf
6a97334d08fcc7d7b75705741d353e4b7fa8c127
13071 F20110115_AACTCJ zhang_c_Page_093thm.jpg
27e0fecad925b1906301b95d9fc90443
609c86183f3fb1759f4b79f42de28e6ce2492ca4
63168 F20110115_AACSXD zhang_c_Page_010.QC.jpg
1b7008a30aac5995710954ab469578da
6995036750d69d524be691dba4ad7878914c4273
20232 F20110115_AACTBV zhang_c_Page_086thm.jpg
e120884636589602408f2a689b64a30b
3be5d1bcbc9cec827cfebb9b5496a13203d77ccd
39718 F20110115_AACTCK zhang_c_Page_093.QC.jpg
2f351e9c807b6f4e988529b6f5e204fe
801d71e713d9bcaf404eb51149512c939e86ddf9
14530 F20110115_AACSXE zhang_c_Page_011thm.jpg
99a4f505be7635e07c6d4bcd00f65072
a88e89c0b07b7cb92926a99c27ceb7a354e883d0
58105 F20110115_AACTBW zhang_c_Page_086.QC.jpg
cc745152f06504a8a4c38e1c810f0846
9366cbcb5c06f0fcd98ce433b0c0a4024173beac
2951 F20110115_AACSWQ zhang_c_Page_002thm.jpg
ebc7a5a4513b343920e5553ff1adc998
66f4560bb5797cba1b2b3595f484b86ed868d429
22547 F20110115_AACTCL zhang_c_Page_094thm.jpg
670e5b158e8d521ea86886c73a8b9797
f5df4a6aa554f5561cd7e8ab812ec4b07dd24497
46122 F20110115_AACSXF zhang_c_Page_011.QC.jpg
684558dc9f913cf123ad326ea9d6d1ef
266421551e0cac5d41a5092c69d0c58861b81f77
22380 F20110115_AACTBX zhang_c_Page_087thm.jpg
3a6796305c3e62be9bfe7959a52700ce
c3809ea08037e95afb06d467b46db5b565dbd310
5496 F20110115_AACSWR zhang_c_Page_002.QC.jpg
f76c45468e58b9d79e4cc3ffd210849f
ef88eca36d834f8da448e11d05f88cd69750f8ce
24025 F20110115_AACTDA zhang_c_Page_103thm.jpg
12b703c950ad10fdcb2f2effc367d17a
e443f29c689f19b48818804da205f3eb8b93e9bd
69081 F20110115_AACTCM zhang_c_Page_094.QC.jpg
6e044602f8b78514c0cd3ed8beacb3de
030b49faf7ab9b7ffd8df54db715407fb7c1e936
22302 F20110115_AACSXG zhang_c_Page_012thm.jpg
e7efcc005ff476d04ce18a9416164ef3
21d73b20f39957fd980c6bf114ab2987b04b1199
59605 F20110115_AACTBY zhang_c_Page_087.QC.jpg
8864bac8f7de6a41762b61748356a222
f78a7cf770ca47b8d37a05c5089647b8b7ef3fb4
4461 F20110115_AACSWS zhang_c_Page_003.QC.jpg
b2f1ab5768e9a218238387b9e2907811
e9cbe005094e06496dd9beb1b6122f44cc846652
75241 F20110115_AACTDB zhang_c_Page_103.QC.jpg
890ef66d46b00be8c28eda7767c53663
3ed8b437dd3bc707b2f1d939f76838b4f05f5ab4
14721 F20110115_AACTCN zhang_c_Page_096thm.jpg
610bc4ebd4d11902ed36a27335051a40
a8eb02e8d3e38d52e827bfcd37a1c81e370093c7
68976 F20110115_AACSXH zhang_c_Page_012.QC.jpg
dcaa325efb3f59e0946eed988a1a7fcd
45b0971397610e8d9ca4d8c545ea8bf3a47ceb76
21293 F20110115_AACTBZ zhang_c_Page_088thm.jpg
63fbefe16caca476830651fdf0c5fe0d
69655e63ba9110fb4979b1be974b0095a8a699a2
50555 F20110115_AACSWT zhang_c_Page_004.QC.jpg
d8597ebe9b84343e64eca14912f6774a
33896b9c019ebadf511b5f6418a6cdfdd93005db
21564 F20110115_AACTDC zhang_c_Page_104thm.jpg
a9d46cc5bf1614de01a1dcb9f2e0242e
602614c1d5dc1de7de0bad46ca3ce829a34ff845
43681 F20110115_AACTCO zhang_c_Page_096.QC.jpg
91cd1098cded08ed3d2e4807f97d7eff
9d2673d10d6eeba7b7c1cc976c255268c8d5e967
46774 F20110115_AACSXI zhang_c_Page_013thm.jpg
0edec02fdbd4d0cd5335894b6d18270f
77e7c02f2e558362afb3c1ad6064ed554e54fe85
47723 F20110115_AACSWU zhang_c_Page_005thm.jpg
dde0674b7d9d0952ea0d9f3f72afe2de
30f63ed9d18830ddc68d4c40ea9938d15cd31583
74583 F20110115_AACTDD zhang_c_Page_104.QC.jpg
b588b8a5f67f9fb3682dd00c7b003d77
3c0910aeefd7da78a475c2046cd1e42a64ed64b4
45269 F20110115_AACTCP zhang_c_Page_097thm.jpg
8ec341860388e2687220449f2974158c
b77a2d8da4bb1bb72fba713d53e24736d9661413
22956 F20110115_AACSXJ zhang_c_Page_014thm.jpg
f7a0ff99dce2bde5371def551270250e
f93a613f5c0021cebd5d029bb7edc7566b098f8b
108533 F20110115_AACSWV zhang_c_Page_005.QC.jpg
73d771df59dab9a4ac36d47cdf1e43b7
37a84da1481899dc594a95fcf75e4dbbd8844490
7168 F20110115_AACTDE zhang_c_Page_105thm.jpg
034a3bee7c14cac1be4834d6f8687ea7
cbe9f900e76b4d55f8a14b03cf8aaf53ff80606b
84964 F20110115_AACTCQ zhang_c_Page_097.QC.jpg
d6e61c6b84e978ece3425be1b93dcbb3
ec37deb4e84c9adc01f49395b9bb3f994df234b6
69994 F20110115_AACSXK zhang_c_Page_014.QC.jpg
09d0bf1669fe9126418f6b737dc6b7a2
c07156d646d2732736e1388374353b1933b638d5
18433 F20110115_AACTDF zhang_c_Page_105.QC.jpg
f554907ff49fb8a669c304a69d235913
878331f498bf48808963332bb89f346a7c4e892f
5046 F20110115_AACTCR zhang_c_Page_098thm.jpg
07e7b15e97d00e81bf4af090d4a19c3d
a79a7a7f482385696578ce82fa45a2675ba9c625
16567 F20110115_AACSXL zhang_c_Page_015thm.jpg
e14e35dfb4e35bc1ddccfee012e855b8
e9a13831f92e11a5cf742c9195e8da2f261b74de
34133 F20110115_AACSWW zhang_c_Page_006thm.jpg
8458107925a2c881dbb0c069bdc3f7d5
c65f8ad66dc38b3e76fc3cb28072a4f04e6379a3
121577 F20110115_AACTDG UFE0008621_00001.mets
550670c27ffe57db761e5908d7361c94
3038d1cc43a50ae2c85789e60d986c84ce634545
22550 F20110115_AACSYA zhang_c_Page_023thm.jpg
1e9d7e388291b457daee8fe2ca9a28c8
3ac1a42037205b981f260cb138bc5aeb6f980ed2
10538 F20110115_AACTCS zhang_c_Page_098.QC.jpg
d674b8891ece8ef3c6d7b88c766da495
f6677fd8372a1e04594e4d62f8d67e34849aed60
54398 F20110115_AACSXM zhang_c_Page_015.QC.jpg
bd29138968fdc90fe2eac9df2bd596cf
800f05f12493ed3a2552dda1a7f6a0ec97507530
50153 F20110115_AACSWX zhang_c_Page_006.QC.jpg
395208c55ad943b7d7c98cdd953ef2c1
beb9e9b357eeddde9a0057eba196afa7d95273cc
68918 F20110115_AACSYB zhang_c_Page_023.QC.jpg
b4ba6e8a798b104d4e7eea3f6fb34cd0
d3cbcf703e4d922f6415dc67b7afec41ed76a20c
20256 F20110115_AACTCT zhang_c_Page_099thm.jpg
47a4a3b863d77e32bc3e206a998066e3
cf3f77104283d505169c5ae2475334cb3c9a8df4
21242 F20110115_AACSXN zhang_c_Page_016thm.jpg
0473eb74f2d171ccb942edca13564c87
fcd7d4315896b8ec314c0c0f78aad06818b3ab43
52777 F20110115_AACSWY zhang_c_Page_007thm.jpg
9af062e471f8e8332ed558a89b7a6b9b
5d6d0606cfbbbe7643f6498677d1e9d8bff1c778
25297 F20110115_AACSYC zhang_c_Page_024thm.jpg
747f28622bf38bd45316e076fcb21a16
aa0e32a651990478963c1c871e2039fc0cc7d20d
61949 F20110115_AACTCU zhang_c_Page_099.QC.jpg
c8bcd1c8a22d4a0b763e7787eed3a948
19fbfacacb5f0be430f8995a683e2d94fe8d73d4
19083 F20110115_AACSXO zhang_c_Page_017thm.jpg
6f5a7a4524a27c74b2ed971c5df4b840
a1b62ada634c293c04462e9742334f6dc9ae3806
122194 F20110115_AACSWZ zhang_c_Page_007.QC.jpg
56b0ee1b449e62267b42d0c274cb27f1
93fa0d6b7e053fbc3b14fba70674dbbb57575321
46726 F20110115_AACSYD zhang_c_Page_025thm.jpg
e6eb12b5bae7198f7d16a9bef432476f
b83cdf7e1683314fd6a8389fb5b41de16f1fb096
16547 F20110115_AACTCV zhang_c_Page_100thm.jpg
54a0f73700ec87787275a91a589c5ff0
b9ea0b87eaf2acb18722856e0d08ecc86cce277e
50179 F20110115_AACSXP zhang_c_Page_017.QC.jpg
0fad979e3bd1265a97584de0b627d549
d2348f1a42f084e489abbe91645cc913c082b83b
89849 F20110115_AACSYE zhang_c_Page_025.QC.jpg
a9c7ed7a8ee7a9f56b5a3be087ad393e
de989ef7b166f5fa3726bd3c9f1f77c7f0c138c2
17988 F20110115_AACTCW zhang_c_Page_101thm.jpg
8797884203a58df3a488140f2f8088e1
792f9ff9ba59e576fcde9001e7830bce5f7f9d68
17206 F20110115_AACSXQ zhang_c_Page_018thm.jpg
da81ea339585461c8df65ac2e6b9fd24
eef4518969969582b55cfaf165eea1cc958648ab
46683 F20110115_AACSYF zhang_c_Page_026thm.jpg
2f56eeafce849c90d695d5cc7e27f982
737a5aaef0a91aa69287a378044bc4677a1b30b4
53920 F20110115_AACTCX zhang_c_Page_101.QC.jpg
9721a965d92151fa49ab1139aa8f0a1d
aef9b290ceebdab3a19e5674daa96dbeb9cd9e22
86577 F20110115_AACSYG zhang_c_Page_026.QC.jpg
cf0e74afdb53d5305fc9a35a948661ba
65b48323d9b26ec835d72c6d5c91ece9300c7ac0
24125 F20110115_AACTCY zhang_c_Page_102thm.jpg
d56f25f6b3918d17b9af2217a10721b7
2feef2a9d6bb01d8f67d2da593b3f8abbf4c52a0
49613 F20110115_AACSXR zhang_c_Page_018.QC.jpg
bdf13f7eb300ef3b14ffbf1b6b1706bc
f5aa160b43d4f37b981635975d52f77bc5fe119b



PAGE 1

TRANSPORT PROPERTIES AT NANO SCA LES VIA FIRST PRINCIPLES STUDIES By CHUN ZHANG A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2004

PAGE 2

Copyright 2004 by Chun Zhang

PAGE 3

To my parents.

PAGE 4

ACKNOWLEDGMENTS First, I would like to thank my advisor, Dr. Hai-Ping Cheng. She brought me into the field of computational physics and exposed me to its frontier. She continuously supported me throughout the past four years and most importantly, taught me that independent thinking is true heart of the research. Research is lonely and boring, but thanks to her, it can also be interesting. I would also like to thank Professors Arthur Hebard, Selman Hershfield, Adrian Roitberg and Samuel B. Trickey for serving on my supervisory committee. Dr. Xiaoguang Zhang, a senior scientist at Oak Ridge National Lab, is another important person to my Ph. D study. Not only his knowledge, but also his attitude toward the research will have a long-term impact to my career. My life at the University of Florida would have been black and white without my friends. They are Dr. Wuming Zhu, Dr. Maohua Du, Dr. Linlin Wang, Dr. Ping Jiang and Dr. Yao He. I would also like to extend special thanks to the following professors at Fudan University, Dr. Yuanxun Qiu, Dr. Jingguang Che and Dr. Jian Zi. Becoming their friend is one of the most important reasons that I want to be a physicist. iv

PAGE 5

TABLE OF CONTENTS page ACKNOWLEDGMENTS .................................................................................................iv LIST OF FIGURES ..........................................................................................................vii ABSTRACT .........................................................................................................................x CHAPTER 1 OVERVIEW.................................................................................................................1 2 FIRST PRINCIPLES APPROACH COMBINING NGFT AND DFT........................4 2.1 Model Hamiltonian and Landauer-type Formula...................................................4 2.2 The Combination of DFT and NGFT for An Open System...................................8 2.2.1 Self-Consistent Procedure for Ground State of Closed System...................8 2.2.2 Self-consistent Procedure for Open System...............................................10 3 COHERENT ELECTRON TRANSPORT THROUGH AN AZOBENZENE MOLECULE: A LIGHT-DRIVEN MOLECULAR SWITCH..................................12 3.1 Introduction...........................................................................................................12 3.2 Tight-binding Model and Self-energies................................................................14 3.3 Conductance at Zero Bias.....................................................................................18 3.4 The Switch under A Finite Bias............................................................................23 4 QUANTUM TUNNELING THROUGH MAGNETIC JUNCTIONS......................30 4.1 Introduction...........................................................................................................30 4.2 Non-equilibrium Greens Function Formalism in Real Space.............................34 4.3 Self-consistency and Tunneling Current for A TMR Junction.............................37 4.4 Application to the Fe/FeO/MgO/Fe Tunneling Junction......................................43 4.5 Conclusion............................................................................................................61 5 PROBLEMS IN THE CURRENT NON-EQUILIBRIUM APPROACH FOR OPEN SYSTEMS BASED ON DFT.....................................................................................63 5.1 Introduction...........................................................................................................63 5.2 Hohenberg-Kohn Theorem for Open System under Finite Bias..........................65 v

PAGE 6

5.3 Exchange Energy for Uniform Electron Gas........................................................75 5.4 Conclusion............................................................................................................82 6 SUMMARY AND CONCLUSIONS.........................................................................83 APPENDIX A EXPANSION OF LESSER GREENS FUNCTION.................................................85 B DETAILS OF NON-EQUILIBRIUM SELF-CONSISTENT PROCEDURE...........88 LIST OF REFERENCES...................................................................................................90 BIOGRAPHICAL SKETCH.............................................................................................94 vi

PAGE 7

LIST OF FIGURES Figure page 1-1. Partition of the open system connected to two metal electrodes..................................2 3-1. Physical models of an azobenzene switch in contact with two gold leads: trans isomer (upper) and cis isomer (lower).....................................................................14 3-2. Tight-binding model of an open system connected to two semi-infinite leads. Indices denote nodes of device region and are the nearest node of left and right lead. Only the nearest-neighbor interaction is considered.......................................15 ,ij ,ab 3-3. A perfect lead that consists of two semi-infinite leads and a hopping term. Indices m and i belong to the left semi-infinite lead and i+1, j belong to the right semi-infinite lead...........................................................................................................................16 3-4. Density of states (DOS) projected onto the switch region (upper panel) and transmission as a function of energy (lower panel). The solid lines represent the trans configuration and dotted lines the cis. Intense peaks in the DOS are truncated to improve readability..............................................................................................19 3-5. Local DOS of the trans and cis configurations decomposed into contributions on the Au layers (lower panel), S atoms (middle panel), and the azobenzene molecule (upper panel). The solid lines represent the trans configuration and dotted lines cis.20 3-6. Local DOS at the Fermi energy projected on seven sites in the switch. From 1-7, or from left to right, Au lead, S atom, benzene ring, double-bonded N 2 another benzene ring, S atom and Au lead............................................................................22 3-7. Non-equilibrium self-consistent procedure. The non-equilibrium Greens function is calculated according to equation (2-11)...................................................................24 3-8. The difference of electronstatic potential between bias one volt and bias zero for cis configuration. The x-z plane is parallel to the N2 bond, cutting through the midpoint of Au unit cell...........................................................................................25 3-9. The transmission probability as a function of single-particle energy and the external bias voltage for cis configuration.............................................................................26 3-10. Current as a function of bias voltage for both trans and cis configurations. The red line represents the current of trans and the blue line represents the current of cis...27 vii

PAGE 8

3-11. The differential conductance as a function of bias voltage for both trans and cis configurations...........................................................................................................28 3-12. The transmission as a function of energy at 0.64, 0.72 and 0.8 Volts for cis configuration............................................................................................................28 4-1. Shifts in the density of states in each atomic layer at a bias voltage of 0.54 V..........44 4-2. Derivative of the capacitive charge with respect to the bias voltage. The charge on the right electrode includes that on the oxygen atom in the interface MgO layer. Net charge' includes that on both the electrode and the dielectric charges on the MgO layers on the same side of the midpoint of the barrier....................................46 4-3. Derivative of the moment with respect to the bias voltage. The labels left and right indicate that the layers to the left (right) of the midpoint of the barrier are summed. The contribution to the sum from the nonbulk layers of the iron electrodes is taken to be the difference between the moment of that layer and the bulk moment in the electrode on the same side..............................................................................47 4-4. Zero bias transmission probability as a function of k // (a) Majority spin channel for parallel alignment of the moments; (b) The dominant spin channel for antiparallel alignment of the moment..........................................................................................50 4-5. Current density as a function of bias voltage for majority and minority spin channels for the case of parallel alignment of the moments between electrodes....................51 4-6. Current density as a function of bias voltage for majority and minority spin channels for the case of antiparallel alignment of the moments between electrodes..............51 4-7. Transmission probability at energy RE as a function of k // along the line of for the majority spin channel for parallel alignment of the moments at different bias voltages..............................................................................................53 xkk y 4-8. Transmission probability at energy RE as a function of k // along the line of for the minority spin channel for antiparallel alignment of the moments at different bias voltages..............................................................................................55 xkk yyy 4-9. Integrated transmission probability as a function of k // along the line of for the majority spin channel for parallel alignment of the moments at different bias voltages.....................................................................................................................56 xkk 4-10. Integrated transmission probability as a function of k // along the line of xkk for the majority spin channel for parallel alignment of the moments at different bias voltages.....................................................................................................................58 viii

PAGE 9

4-11. Integrated transmission probability as a function of k // for the majority spin channel for parallel alignment of the moments at bias 0.378 V............................................59 4-12. Integrated transmission probability as a function of k // for the minority spin channel for antiparallel alignment of the moments at bias 0.378 V......................................59 4-13. Magnetoresistance ratio as a function of bias voltage. Point A denotes the MR ratio at zero bias by removing all contribution from interface resonance states..............61 5-1. Exchange energy at different bias voltages for different (Electrons/m 3 )..............78 5-2. Split of the exchange energy: (a) the exchange between non-equilibrium electrons and the exchange between non-equilibrium and equilibrium electrons and (b) the exchange between equilibrium electrons.................................................................80 5-3. Exchange energy as the functional of /n for different electron densities............81 A-1. Expansion of lesser Greens function. The time of the perturbation can be either on + branch or on branch............................................................................................86 ix

PAGE 10

Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy TRANSPORT PROPERTIES AT NANO SCALES VIA FIRST PRINCIPLES STUDIES By Chun Zhang December 2004 Chair: Hai-Ping Cheng Major Department: Physics There are two main difficulties for the first principles study of transport properties at the nano scale. The first is that many-body interactions need to be taken into account for the infinite system without periodic boundary conditions. The other is that the system is usually in a non-equilibrium state. Both of these two difficulties are beyond the ability of conventional first principles methods to reconcile. Recently, a new first principles approach which combines the Non-equilibrium Greens Functions Technique (NGFT) and the Density Functional Theory (DFT) was proposed. DFT has been proved to be successful in molecular and solid state physics. Currently used DFT approximations can take into account most many-body effects and NGFT naturally includes the non-equilibrium effects. The new approach uses NGFT to treat the non-periodic boundary conditions and DFT to treat many body interactions. This approach has been successfully used in molecular electronics. The thesis is organized in the following way. First: we introduce the main ideas of combining NGFT and DFT, and then apply this method to a light-driven molecular switch. The switch, made of a single molecule, is one of the most x

PAGE 11

important elements of nano-electronics. However, most proposed molecular switches are driven either by an external bias voltage or by STM manipulation, neither of which is ideal for nano-scale circuits. The switch we designed has a high on-off conductance ratio and more importantly, can be driven by photons. In following chapters, we generalize the method to the spin-dependent case and apply it to a magnetic layered structure. We implemented the method within the framework of the Layer Korringa-Kohn-Rostoker (LKKR) approach, which is particularly well-adapted to the layered structure and found a bias-enhanced tunneling magneto-resistance (TMR) for the Fe/FeO/MgO/Fe junction. Our results are important not only for application, but also for understanding of the voltage-dependence of TMR for layered structures. The experimental studies show that the bias voltage usually kills the TMR of amorphous magnetic tunneling junctions. Our study shows that for an impurity-free layered structure, a different behavior of TMR may occur. xi

PAGE 12

CHAPTER 1 OVERVIEW As the trend of miniaturization continues, we eventually enter the regime of nano-electronics, the technology of building electronic devices on the basis of a single molecule. Experiments have shown the great potential of modern technologies, such as STM and AFM, to make molecular transistors and switches etc. [1-5]. For theoretical studies, the circuit in which the molecule is embedded provides great difficulties. First, the molecule is connected to the circuit through two open boundaries. It is crucial to describe the open boundaries in a way that the interaction between the molecule and the complicated circuit is taken into account. Second, the circuit is almost always far away from the equilibrium state and in most cases, varying with time. The time-dependent process is not the focus of this thesis. We assume that the whole circuit is in a steady state. Our goal is to find the steady-state solution for an open system far away from equilibrium. An open system connected to two metal electrodes usually has geometry such as shown in Fig. 1. The whole system can be divided into three parts, the left lead, the right lead and the device region. The device region includes the molecule and enough portions of the two metal electrodes so that the electrons in the two leads are not affected by interactions in the device region because of screening. A finite bias is applied to the system. The goal of the research is to relate the steady-state current flowing through the device region to the external bias voltage. Traditionally, the electron transport phenomena through such an open system are studied by the Boltzman equation in 1

PAGE 13

2 conjunction with the effective-mass assumption. The Boltzman equation is based on a probabilistic description for the evolution of a single-particle distribution which is a function of classical variables ),,(tprf r and p However, when the device region is at nano-scale, the phase coherence of charge carriers is usually maintained in the whole device region so that quantum effects dominate. For this kind of coherent quantum transport phenomena, the first principles approach which combines the non-equilibrium Greens functions technique (NGFT) and Density Functional Theory (DFT) has proven to be successful [6-22]. The method can be summarized in two steps. First, a non-equilibrium self-consistent procedure based on DFT and NGFT is used to determine the single-particle effective potential. Second, a Landauer-type formula is used to calculate the current, which relates the current to the integral of the single-particle transmission probability over energies. Current I Left lead Right lead Device Region Figure 1-1. Partition of the open system connected to two metal electrodes. The approach combining NGFT and DFT is based on the single-particle picture. The tunneling process is explained within this framework as follows. The Bloch electrons in two leads tunnel through the molecular orbital in the device region. The success of this picture can be best seen in the work recently published by a Denmark group [21], where

PAGE 14

3 the current flowing through a broken Au nanowire is calculated. At small bias voltages, the theory agrees well with the experiment. The approach can also be equivalently built on a scattering wave formulation [8-11]. The scattering wave approach has so far been implemented only with jellium electrodes, due to the difficulties associated with computing and storing Bloch wave functions of realistic leads. In Chapter 2, we introduce the details of the two-step approach. In Chapter 3, we apply the method to the Au/Azobenzene/Au junction. Our calculation shows that the azobenzene molecule is a good candidate for the light-driven molecular switch. In Chapter 4, we generalize the method to spin-dependent tunneling junctions with application to a layered magnetic Fe/FeO/MgO/Fe junction. The calculation suggests a bias-enhanced tunneling magneto-resistance (TMR), which is important for both application and the understanding of TMR through a layered structure. In Chapter 5, we discuss the problems that exist in the current two-step approach. For the first step, in which the effective single-particle potential is determined, DFT is the biggest problem. The Hohenberg-Kohn theorem, the basis of DFT, is correct only for the ground state [23]. The use of DFT for the open system far from equilibrium state is problematic. We discuss this problem in detail and present a possible way to introduce non-equilibrium effects into DFT. For the second step, in which a Landauer-type formula is used to calculate the current, the problem is that the Kohn-Sham orbitals are used to calculate the current. Such orbitals are not genuine physical quantities, which causes trouble in choosing basis sets and functionals.

PAGE 15

CHAPTER 2 FIRST PRINCIPLES APPROACH COMBINING NGFT AND DFT 2.1 Model Hamiltonian and Landauer-type Formula As we mentioned in the previous chapter, the system of interest can be divided into three parts, the left lead, the device region and the right lead. To relate the current flowing through the device region to the external bias is the goal of our research. The whole system is infinitely large, but without periodic boundary conditions, which provides a big challenge to the theoretical study. However, if one assumes that the two leads are non-interacting, an exact formula which relates the current to local properties of the device region can be derived [24]. The derivation is outlined in following paragraphs. The starting point is the Hamiltonian CouCenRLHHHHH (2-1) where LH RH and CenH denote the Hamiltonian for the left lead, the right lead, and the device region respectively. The last term CouH denotes the coupling between leads and the device region. Since the two leads are non-interacting, we have nRLknknkCounnnCenRLkkkkRLchdcVHddHHccH,,,int,,,,,,.].[}){},({ (2-2) 4

PAGE 16

5 where indices ,k denote the channel of left and right leads (Bloch wave vectors and spin) and denotes the single-particle basis of the device region. The key issue is how to treat the interaction of the device region. n The expression for the current can be derived by taking the time derivative of the total number of electrons in the left lead LLLNHieNeJ, (2-3) where and is the current from the left lead into the device region. Inserting the Hamiltonian into the expression for the current, we have LkkkLccN, LJ nLknknkknnknLkknnknknkLGVGViecdVdcVieJ,,*,,,,*,, (2-4) where the current is expressed in terms of two non-equilibrium lesser Greens functions. Now, the problem becomes how to calculate the lesser Greens functions. This can be achieved by a perturbation procedure. At t the two leads and the device region are well separated, which means that there is no the coupling term in the Hamiltonian. The three parts maintains their own Fermi energies. Without loss of generality, we assume that the Fermi energy of the right lead R is higher than the Fermi energy of left lead L As we gradually turn on the coupling from to t 0 t the system will achieve a steady state at 0 t with a current from left to right. Taking the coupling term in the Hamiltonian to be the perturbation, the lesser Greens functions can be expanded as J

PAGE 17

6 mamnkkmnrkkmknkmakknmkkrnmmkknGgGgVGgGgGVG][][,,,,,,*,, (2-5) where runs over all basis states in the device region ,denotes the Greens function for non-interacting leads and the upper indices m g a r denote retarded and advanced Greens functions respectively. The details of the expansion can be found in Appendix A. Introducing the equation (2-5) into equation (2-4) and doing a Fourier transform gives ]})()()()([Re{22,,,*,, nmakkrnmkkmnLkmknkLGgGgVVdeJ (2-6) where the summation runs over all channels of the left lead. Since the lead is non-interacting, the Greens functions can be calculated as agg, igifgkakkkkkk1)()()(2)(,, (2-7) where )( kf is the Fermi-Dirac distribution function. Insertion of the equation (2-7) into equation (2-6) gives the current as ]})([{2GGGfTrdieJarLLL (2-8) where the line-width function L is defined as LkmknkmnkLVV)()()(2)]([* (2-9) in which the term is the density of states in the lead. A similar formula can be derived for, the current from the right lead into the device region, so the total current is given by RJ )(21RLJJJ which is

PAGE 18

7 ]}][[]{[22arRRLLRLGGffGTrdieJ (2-10) where the current is related to the local properties of the device region. In equation (2-10), the Greens functions in the device region are solvable for only a few models. The simplest one among them is the freelectron model, which means that the device region is also non-interacting. For the free-electron model, we have [15, 20, 25] ][RRLLarffiGGG (2-11) and then the current can be written as }{][LrRaRLGGTrffdeJ (2-12) which is a Landauer-type formula. Since the interaction in both leads and the device region usually plays an important role in the tunneling process at nano scales, equation (2-12), which is based on the free-electron model, seems to be uninteresting, unless it is combined with a mean field theory. DFT is one kind of mean field theory. It is based on the single-particle picture. Within present-day approximations, it is able to take into account most many-body effects. Moreover, due to the success of DFT in the electronic structure calculation of molecules, it is natural to combine DFT and the Landauer-type formula to study molecular electronics.

PAGE 19

8 2.2 The Combination of DFT and NGFT for An Open System 2.2.1 Self-Consistent Procedure for Ground State of Closed System DFT has been known to be powerful in the electronic structure calculation for both molecular and solid state systems [26-30]. The main ideas of DFT are outlined in the following paragraphs. The basis of DFT is the Hohenberg-Kohn theorem [23]. This theorem shows that the ground state of a closed system is uniquely determined by the electron density of the system, which means that average value of any physical quantity for the ground state can be expressed as a functional of the electron density. As an example, the total energy can be written as rdrVrFEext )()(][][ (2-13) where the term represents the entire external potential (including the ion-electron interaction) and the first term extV ][ F is a universal functional related to the electron-electron interaction and the kinetic energy of electrons. The theorem also shows that the true ground-state electron density minimizes the energy functional ][ E The only thing used in the proof of Hohenberg-Kohn Theorem is the minimum energy principle of ground state. On the basis of the Hohenberg-Kohn theorem, a self-consistent procedure based on the single-particle picture can be constructed. Assume that there is a non-interacting system which yields the same electron density as the interacting system of interest, then the electron density can be written in terms of the single-particle functions iirr2)()( (2-14)

PAGE 20

9 which are usually called Kohn-Sham orbitals. The kinetic energy of the non-interacting system can be calculated in terms of the Kohn-Sham orbitals (Hartree atomic units) iiisT2][2 (2-15) Then the exchange-correlation energy can be defined as ][][][][ HsxcETFE (2-16) where the term ][ HE is the classical Coulomb energy. The exact form of ][ xcE is unknown. However many kinds of exchange-correlation functionals, most based on the local density approximation, have been proposed and successfully used. By minimizing the energy functional, we get the Kohn-Sham equations )()()]([)()]([212rrrVrVrViiixcextH (2-15) where the first term is the well-known classical Coulomb potential and the term is the exchange-correlation potential defined by HV xcV ][xcE The effective Kohn-Sham potential is defined as )]([)()]([)]([rVrVrVrVxcextHeff (2-16) Now, a self-consistent procedure (Kohn-Sham procedure) can be completed as follows. Given the electron density the effective Kohn-Sham potential ][ effV can be calculated according to equation (2-16). Knowing ][ effV we can get the Kohn-Sham orbitals i by solving equation (2-15) and then the new electron density can be calculated according to equation (2-14). Repeat the procedure until the self-consistency is achieved. The Hohenberg-Kohn theorem guarantees that the Kohn-Sham procedure yields the true

PAGE 21

10 ground state electron density (the V-representability problem for the electron density is not discussed in this thesis). Thus an interacting system can be solved by a self-consistent procedure based on the single-particle picture. 2.2.2 Self-consistent Procedure for Open System For an open system with the geometry shown in Fig. 1-1, there are two kinds of eigenstates, scattering states which are extended in all space and bound states which are localized in the device region. If the bias voltage is zero, DFT is still applicable due to the fact that the system is indeed in a ground state (the temperature is set to zero). In practice, since we cannot treat infinitely large system, certain approximations have to be made. The major approximation is that, due to screening effects, the electron density and effective Kohn-Sham potential in the two leads are not affected by the device region and thus can be set to the bulk values. The bulk is also infinitely large, but with the help of the Bloch theorem, we only need to do the DFT calculation in one unit cell. After fixing the effective Kohn-Sham potential of the two leads, the remaining problem is to treat the device region, which is finite, with two fixed boundaries. If the bias voltage is not zero, the use of DFT becomes problematic. In Chapter 5, we will discuss this problem in detail, but here, we just assume that the same approach can be generalized to the finite bias case. The goal is to calculate the electron density self-consistently in the device region with two fixed, open boundaries. The Greens functions technique described below is an efficient way to do this. The electron density in the device region is directly related to the lesser Greens function by dGi)(21 (2-17)

PAGE 22

11 For a single-particle Hamiltonian, the lesser Greens function can be calculated by the equation (2-11), where the retarded and advanced Greens functions are related to the effective potential of the device region and two open boundaries as )()()(1arRarLCenarHEG (2-18) In equation (2-18), the term CenH denotes the effective single-particle Hamiltonian and the two terms are self-energies for the left and right lead respectively. Two self-energy terms represents effects of open boundaries. We will see an example of calculating these two terms in Chapter 3. RL, As a conclusion, the self-consistent procedure for an open system can be summarized as follows. First, the two leads can be treated by a separate bulk calculation which provides two open boundaries for the device region. This only needs to be done once. Second, given the electron density in the device region, the effective Kohn-Sham potential can be calculated. Knowing the effective potential, the lesser Greens function can be computed and then a new electron density is given by equation (2-17). After self-consistency is achieved, the Landauer-type formula (2-12) can be used to calculate the current in the system. The non-equilibrium self-consistent procedure mentioned above is implemented in a Greens functions representation which employs a multi-band tight-binding model in Chapter 3 and is implemented in a wave function representation which includes spin in Chapter 4.

PAGE 23

CHAPTER 3 COHERENT ELECTRON TRANSPORT THROUGH AN AZOBENZENE MOLECULE: A LIGHT-DRIVEN MOLECULAR SWITCH 3.1 Introduction Electronic devices soon will be constructed of molecules or have features of molecular size. The results of recent experimental [1-5] and theoretical studies [6-20] suggest a brilliant future for molecular electronics. Among various efforts and activities, several schemes have been proposed to design and construct molecular switches that have two or more distinct states with vastly different conductance. Switching between on and off states can be performed by applying an external bias or by using a scanning tunneling microscope tip to manipulate the system [31-35]. These methods are not ideal. The former approach may interfere greatly with the function of a nanosize circuit, while the latter one imposes severe limitations in device applications. In addition, these methods are relatively slow. Alternative methods, such as those based on fast, light-driven processes, are therefore highly desirable. The photosensitive azobenzene molecule discussed here is one such candidate for an opto-electronic device. The azobenzene molecule has two stable conformations [36-38], trans and cis. The energy of the trans structure is 0.6 eV lower than the energy of the cis structure [39], and they are separated by a barrier of approximately 1.6 eV (measured from trans to cis) [40]. The trans structure is 1.98 longer than the cis. The molecule can switch reversibly from one structure to another under photoexcitation, with the structural change occurring on an electronically excited state [38, 41]. 12

PAGE 24

13 Previous work has shown that a nanoscale opto-mechanical device based on the conformation change of the azobenzene polymer can be designed. Experimentally [41], it is known that light with a wavelength of 365 nm transforms azobenzene units from the trans to the cis structure. As a result, the polymer contracts, thus converting the energy of electromagnetic radiation to mechanical work. A second pulse of light with a wavelength of 420 nm reverses the cis to the trans configuration. In this chapter, we report investigations of transport in the molecular wire, Au-S-azobenzene-S-Au. It consists of a single azobenzene molecule, which has been functionalized by replacing a hydrogen atom with a CH 2 S group to provide a good contact with gold, and two semi-infinite gold leads. Note that the name of the modified molecule is technically bis-(4-methanethiol-pheyl) diazene or BMTPD. Since one can add more azobenzene units to elongate the polymer, we simply follow Refs. [38, 39] and refer to the molecular system as azobenzene. As shown in Fig. 3-1 and consistent with previous chapters, the entire system is divided into three regions: left lead, switch, and right lead. The switch includes two layers of gold from each lead. The structures of the leads are chosen to be the same as in Ref. [42], which corresponds to a gold (111) crystal lattice. The electronic structure of the switch region is determined by treating it as a cluster. Full quantum calculations, based on density functional theory (DFT), are used to optimize the structure of a cluster consisting of the molecule, the sulfur atoms, and two rigid layers of gold. The optimization is performed with two different choices of the interlead distance, d. For the trans configuration, the calculated optimal d is 18.16 and, for the cis configuration, 16.23 In subsequent sections, we study the transport

PAGE 25

14 properties for both equilibrium and non-equilibrium case. In both cases, the localized basis set and the tight-binding model is employed. Figure 3-1. Physical models of an azobenzene switch in contact with two gold leads: trans isomer (upper) and cis isomer (lower). 3.2 Tight-binding Model and Self-energies As mentioned in Chapter 2, the key issue is to evaluate the self energies of two semi-inifinite leads, and. Once these two terms are computed, we can construct the retarded and advanced Greens functions and of the device region in the presence of two leads. Then the lesser Greens function can be calculated according to equation (2-11). L R rG aG The self-energy term represents the effects of the semi-infinite lead on the device region. Usually, a truncation of the Hamiltonian in real space is needed since we can not treat an infinitely large system. So, the tight-binding model is a natural choice. Fig. 3-2 sketches the model, where indices j i, denote nodes of the device region and are the nearest node of the left and right lead respectively. Note that hopping terms are operators [42]. ba, ',hh

PAGE 26

15 i j abh'h Figure 3-2. Tight-binding model of an open system connected to two semi-infinite leads. Indices j i, denote nodes of device region and are the nearest node of left and right lead. Only the nearest-neighbor interaction is considered. ba, Within the framework of tight binding, the Greens function can be written as ijG jjbbCenijijaaCeniiCenijijGhghGhGghGGG'' (3-1) where the two hopping terms are taken as perturbations and the Dyson equation is used. The term represents the Greens function of the device region without two leads and the term denotes the unperturbed Greens function of the leads. Then, self-energies can be defined as CenG g ''hghhghbbRaaL (3-2) where, are self-energies of the left and right lead respectively. Two hopping terms are projections of the effective Kohn-Sham Hamiltonian to a localized basis of each node (hopping terms are matrices also). The only problem remaining is two so-called surface Greens functions and. RL, aag bbg Since a perfect lead with periodic boundary conditions consists of two semi-infinite leads and a hoping term within the framework of tight-binding as shown in Fig. 3-3, it is possible to construct the Greens functions of semi-infinite leads on the basis of Greens functions of a perfect lead by a perturbation procedure. The whole procedure is described in the following.

PAGE 27

16 h m i i+1 j Figure 3-3. A perfect lead that consists of two semi-infinite leads and a hopping term. Indices m and i belong to the left semi-infinite lead and i+1, j belong to the right semi-infinite lead. We take the hopping term h to be the perturbation, then the node i is the surface node of the left lead and the node i+1 is the surface node of the right lead. G is used to denote the Greens function of the perfect lead and G 0 the Greens function of two semi-infinite leads. The Dysons equation reads qjpqipijijGVGjVGGiGG000 (3-3) where q p run over the whole system. Since V only contains two terms V i,i+1 and V i+1,i P equals to i or i+1. Because G 0 i,i+1 =0 (without perturbation, an electron can not propagate from i to i+1), we have jiiijiiiiiijhGgGVGG,1,11,0 (3-4) Similarly, (3-5) miiimiiiiimiGhgGVGG,1,1,,101,1,1 In order to solve g from equation (3-4) and (3-5), we have to know relations between G ij and G i+1,j This can be found by solving the Greens functions of the perfect lead. For a perfect lead, the tight-binding Hamiltonian is mmmmmmmmaahahaaahH110 (3-6) where are matrices, being the number of orbitals within one unit cell of the lead. The eigenvalue problem for Bloch states takes the form hh,0 NN N

PAGE 28

17 0),()()(*0kuehehhikdikd (3-7) where is the Bloch vector and is the lattice constant. The energy term k d contains an infinitesimal imaginary part, with iE 0 Equation (3-7) can be written as 02iiiuhhM (3-8) where, and It is to be noted that not all eigenvalues 0hM hh i of equation (3-8) for a given energy represent the Bloch state. Because equation (3-8) is quadratic in and has a dimension of N, the number of eigenvalues for a given energy is 2N. It is shown in reference [22] that, when the energy is not real, half of those eigenvalues are outside the unitary circle in the complex plane (we denote them as i ) and another half are inside the unitary circle (we denote them as i ). Using the superscripts to denote eigenvalues and eigenvectors for and, respectively, we can introduce two matrices, NN and 1*1)()(UUUU (3-9) where is a matrix, whose th column is the eigenvector and is a diagonal matrix, whose th diagonal element is the eigenvalue The matrices, U i iu i i and satisfy the matrix equation 0)()(0)()(22hhMhhM (3-10) The retarded Greens function is defined by the matrix equation. We have IGH)( .0)()(,1,1jijiijijGhGhGMjGHi (for j i ). (3-11)

PAGE 29

18 For a perfect lead, the solution of this equation is done via recursion relations of Greens functions, and. This gives the equation jiijAGG,1 jijiGAG,12,1 0. (3-12) )(2AhhAM Comparing with equation (3-10), we see that, for j i Introducing into equation (3-5), we have A jiijAGG,1 1)(hgii (3-13) Similarly, for j i we have (3-14) 11,1)(hgii 3.3 Conductance at Zero Bias In this section, we calculate the conductance of the two optimized structures. In this study, we focus on the linear-response regime. Within this limit, the entire system is in equilibrium and there is a common Fermi energy across the whole system. Because of charge screening by electrons in the metal, the two leads, as well as the switch, are charge neutral. The conductance is calculated by the Landauer formula [44], )(22FEThe (3-9) where is the transmission function at the Fermi energy. The effective potentials are calculated by the computational chemistry code NWCHEM [28]. In the calculation, we use Crenbs effective core potential (ECP) Gaussian basis [28] for Au atoms, STO-3G Gaussian basis [28] for all other atoms (C, S, N, H). The pbe0 exchange-correlation functionals [26] are used. For the wire system depicted in Fig. 3-1, the system is, in principle, infinite, and periodic boundary conditions are not appropriate. However, due to )(FET

PAGE 30

19 screening, the Kohn-Sham potentials in the two semi-infinite leads are not affected by the switch. As a result, the potentials can be computed in separate calculations of a one-dimensional gold wire with periodic boundary conditions [20, 22, 42]. -0.24-0.22-0.20-0.180.00.40.8 CisCisTransTransmission Density of states (in device area)Energy (Hartree)EfTrans0200400 aa'bb' Figure 3-4. Density of states (DOS) projected onto the switch region (upper panel) and transmission as a function of energy (lower panel). The solid lines represent the trans configuration and dotted lines the cis. Intense peaks in the DOS are truncated to improve readability. With Greens function techniques, the effect of the two leads on the switch can be absorbed into self-energy terms, which are calculated using the same method as mentioned in previous section. The transmission function T is evaluated by the Caroli formula [25] )(E F ][aRrLGGTrT (3-10) which is the kernel of the integral in equation (2-12). The density of states (DOS) in the switch region can be calculated from the trace of the retarded Greens function, which

PAGE 31

20 includes both scattering states and localized states. The DOS can then be projected onto any desired subsystem. Note that the subsystem can be chosen as a single atom or a group of atoms. The peaks in the DOS as a function of energy are often used as indications of the molecular orbital and, with the combination of the transmission as a function of energy, useful information can be extracted as to how any given molecular orbital contributes to the conductance. -0.24-0.22-0.20-0.180200 Local density of statesEnergy (Hartree)Au layersaa'b'0200 Trans Sulpher atoms a0200 CisCisTransCis EfTransAzobenzene moleculebb' Figure 3-5. Local DOS of the trans and cis configurations decomposed into contributions on the Au layers (lower panel), S atoms (middle panel), and the azobenzene molecule (upper panel). The solid lines represent the trans configuration and dotted lines cis. The calculated transmission as a function of energy and the DOS for the trans and the cis structures, with optimal interlead distance d, are shown in Fig. 3-4. Unlike the DOS of an isolated molecule, Fig. 3-4 displays a series of peaks superimposed on a broad background. These features originate from a combination of localized states from the molecule and a nearly continuous energy band of the two leads. At some energies, the DOS shows peaks, but the total transmission is very small. Notice, for example, peaks a

PAGE 32

21 and b, for the trans isomer, and peaks and for the cis isomer. These peaks are indications of localized states, which can be seen more clearly in the local density of states. 'a b As depicted in Fig. 3-5, the DOS of the switch region is decomposed into three subsystems: the two Au layers, the two S atoms, and the azobenzene molecule. For the trans isomer, peak is mainly localized on the contacts, namely the Au layers and S atoms, while peak b is localized mainly on the molecule. For the cis isomer, peak is localized mainly in the Au layers, while peak b has high density on the molecule. The DOS on the S atoms is very small in the cis isomer as compared to the same peak in the trans isomer. These comparisons indicate that the trans and cis configurations have quite different contacts with the leads. Furthermore, Fig. 3-5 indicates clearly that the broad background in the DOS derives primarily from the Au layers. a 'a Figure 3-5 shows that the Fermi energy of the system lies between the highest occupied molecule orbital (HOMO) and the lowest unoccupied molecule orbital (LUMO) of the azobenzene molecule, slightly closer in energy to the HOMO. Compared to the isolated molecule, the peaks in the local DOS projected on the molecule indicate that the molecular orbitals are perturbed by the presence of the leads and contribute to scattering states near the Fermi energy. Since we focus here on the zero-bias case, the conductance is determined completely by transmission at the Fermi energy, which is dominated by the tail of the electron distribution tunneling through the perturbed HOMO. Note that the local DOS of both the Au layers and the S atoms peak near the trans HOMO energy, but that the magnitudes of these peaks are minimal for the transmission function in Fig. 3-4. For the trans isomer, tunneling through the perturbed HOMO leads to a very broad and

PAGE 33

22 intense peak in the transmission function, but a sharp and smaller peak for the cis isomer (Fig.3-4, lower panel). As a result, tunneling at the Fermi energy is more rapid through the trans structure than through the cis structure, although the DOS of the cis isomer is higher than that of the trans isomer. The calculated zero-bias conductance for the trans structure is which indicates a moderately good conductor. In contrast, the conductance for the cis structure is nearly 2 orders of magnitude lower. 141021.0 1234567051015202530 Local density of statesAtom groups1: Left Au layer2: Left S atom3: Left benzene ring4: N=N5: Right benzene ring6: Right S atom7: Right Au layerCisTrans Figure 3-6. Local DOS at the Fermi energy projected on seven sites in the switch. From 1-7, or from left to right, Au lead, S atom, benzene ring, double-bonded N, another benzene ring, S atom and Au lead. 2 To understand the tunneling process at the Fermi energy, Fig. 3-6 depicts the local DOS at the Fermi energy, Fig. 3-6 depicts the local DOS at the Fermi energy for both the trans and the cis isomers. In this figure, we decompose the DOS into seven components: the two left Au layers, the left S atom, the left benzene ring with a CH 2 group, the double-bonded N 2 the right benzene ring with a CH 2 group, the right S atom, and the two right Au layers. As can be seen in the figure, the local DOS for the two molecular conformations differs significantly. For the cis isomer, the local DOS is much higher on

PAGE 34

23 the Au layers and the two benzene rings than on the S atoms and the N double bond. The nearly zero DOS on the right S atom and the oscillation of the DOS along different sites suggest that the scattering states are localized primarily on the benzene rings and the gold leads. Reflection of the scattering wave occurs at sites of low DOS (on S and N). Both localization and reflection contribute significantly to the resistance of the device. Compared to the cis isomer, the local DOS for the trans isomer displays a smoothly varying curve, suggesting that localization and reflection at the N double bond and the two S atoms contribute much less to resistance than in the cis configuration. In the azobenzene/gold system, the extended scattering states allow facile electron tunneling through the molecule. This result can also be interpreted from the viewpoint of molecular structure. In the cis configuration, due to the different orientations of the two benzene rings, orbitals on different rings have different configurations. Consequently, when electrons tunnel through the azo unit of the cis configuration, they are scattered more than they are in the trans configuration. A more illuminating way to understand the tunneling process is to decompose the conductance into conduction channels of two leads, which will be done in future work. The conductance of the cis isomer is significantly smaller than that of the trans. Since the azobenzene molecule can easily change between trans and cis configurations via excitation with light, we suggest that this system is a viable candidate as a light-driven molecular switch. The switch will work at room temperature due to the high thermal stability of both the trans and the cis configurations of the azobenzene molecule. 3.4 The Switch under a Finite Bias When a finite bias is applied, the whole system is out of equilibrium. In this section, we apply the non-equilibrium self-consistent procedure discussed in Chapter 2 to

PAGE 35

24 the Au/S/azobenzene/S/Au junction and relate the current through the switch to the external bias voltage. In the study, we assume that the geometry of the device region is the same as that of zero-bias case. We neglect the effects of the so-called current-induced force by making this assumption. dGi)(21 ][ VVeff Figure 3-7. Non-equilibrium self-consistent procedure. The non-equilibrium Greens function is calculated according to equation (2-11). The self-consistent procedure is depicted in Fig. 3-7. Since we are working within the framework of DFT, the most important physical quantity is the electron density. Given the electron density the effective potential can be computed as a functional of the density. Once we know the effective potential, the non-equilibrium Greens function can be calculated according to equation (2-11), where the retarded and advanced Greens functions are calculated in the same way as in the previous section. The details of the self-consistent procedure can be found in Appendix B. effV G One of the major accomplishments of the non-equilibrium first principles calculation is that the electrostatic potential due to the external bias can be computed self-consistently. Fig. 3-8 shows the difference of the electrostatic potential between bias one volt and zero bias for the cis configuration. In the figure, the y axis is along the current direction and the x-z plane is parallel to N double bond of cis configuration, cutting

PAGE 36

25 through the midpoint of the Au unit cell. The drop of the electrostatic potential across the switch area can be clearly seen. Figure 3-8. The difference of electronstatic potential between bias one volt and bias zero for cis configuration. The x-z plane is parallel to the N2 bond, cutting through the midpoint of Au unit cell. The current through the system is calculated by equation (2-12), where the current is expressed as the integral of the single-particle transmission probability over energy. The effects of external bias voltage on the current through the switch enter in two ways. First, as the bias voltage increases, the upper and lower limit of the integral changes so that more channels enter the integral. Second, the bias voltage changes the electrostatic potential, thus changes the transmission probability in turn. In Figure 3-9, we show the transmission probability as a function of both energy and bias voltage, where the transmission is expressed as ),(),(),(),(),(VEGVEVEGVETrVETaRrL (3-13)

PAGE 37

26 Figure 3-9. The transmission probability as a function of single-particle energy and the external bias voltage for cis configuration. There are two effects of the external bias on the transmission probability through a particular molecular orbital. First, as the external bias changes, the effective Kohn-Sham potential in the device region changes due to the change of electrostatic potential, thus the molecular orbital changes. On the other hand, the external bias shifts the Fermi energies of the two leads in opposite directions so that channels in both electrodes corresponding to the same molecular orbital change. Both of these two effects are taken into account in the first principles study. I-V curves through the trans and cis configuration are shown in Fig. 3-10. The differential conductance, defined as the derivative of the current with respect to the bias voltage, is shown in Fig. 3-11. One of the important results is that at zero bias, the conductance of trans configuration is two orders higher than the cis, which agrees with the equilibrium calculation. The current through the trans configuration is much higher

PAGE 38

27 than the cis for all bias voltages, which shows the high stability of the switch under a finite bias. Figure 3-10. Current as a function of bias voltage for both trans and cis configurations. The red line represents the current of trans and the blue line represents the current of cis. An interesting feature is shown in Fig. 3-10. For the cis configuration, when the bias increases from 0.7 Volts to 0.8 Volts, the current drops, which gives a negative differential conductance as shown in Fig. 3-11. This can be explained by consideration of Fig. 3-12, where the transmission as a function of energy at 0.64, 0.72 and 0.8 Volts are shown. As we mentioned before, as the bias voltage increases, more channels enter the integral of equation (3-1), which usually increases the current. However, this is not always the case. If the transmission through some channels which have major contribution to the current is greatly suppressed, it may not be compensated by the increase of channels. Fig. 3-10 clearly shows such an example, where as the bias voltage

PAGE 39

28 Figure 3-11. The differential conductance as a function of bias voltage for both trans and cis configurations. Figure 3-12. The transmission as a function of energy at 0.64, 0.72 and 0.8 Volts for cis configuration.

PAGE 40

29 increases from 0.72 Volts to 0.8 Volts, the transmission through the HOMO state is suppressed thereby leading to the negative differential conductance in Fig. 3-11.

PAGE 41

CHAPTER 4 QUANTUM TUNNELING THROUGH MAGNETIC JUNCTIONS 4.1 Introduction Quantum transport in open systems far from equilibrium is an important and challenging problem [1-5] as mentioned in previous chapters. This problem has attracted considerable attention in the context of semiconductor based devices for which the relevant electronic structures can be modeled using the effective mass approximation. More recently, considerable progress has been made with respect to application of quantum transport theory to molecular electronics. In particular, the effect of finite bias voltage has been studied with a number of first principles approaches [6-20] with various degrees of success. The preceding chapter is an example. The problem of spin-dependent tunneling in magnetic heterostructures, however, presents a significantly greater challenge than the spin-independent case to the theory of non-equilibrium transport because one must simultaneously deal with the complexities of non-equilibrium transport and complex electronic structures along with magnetism. Spin-dependent tunneling junctions are considered to be the most likely candidate for next-generation devices for spintronics. Compared to the trilayer or multilayer giant magnetoresistance (GMR) systems, magnetic tunneling junctions tend to have a relatively higher magnetoresistance ratio [45-48], which is an important factor in device designs. The magnetoresistance ratio is defined as (G p -G a )/G a where G p is the conductance for the case that the alignments of net magnetic moments of two leads are parallel and G a is the conductance for the case that the alignments of net magnetic moments of two leads are anti-parallel. One of the 30

PAGE 42

31 significant differences between a GMR system and a spin-dependent tunneling junction is that the former almost always operates within the linear response regime, while the latter usually has at least a bias voltage of a fraction of one Volt across the distance of a few nanometers, placing it far away from the equilibrium state. It has been observed experimentally [49] that the bias voltage across a tunneling junction can greatly reduce the tunneling magnetoresistance (TMR). Understanding the dependence of the TMR on the voltage bias has become one of the critical issues in the development of spin-dependent tunneling devices. There have been several first-principles studies [50-54] of tunneling conductance of trilayer junctions of the form FM/barrier/FM, where FM is a ferromagnetic material and the barrier layer is either a semiconductor or an insulator. These studies were limited to the linear-response regime. Researchers generally found exceptionally large TMR ratios compared to experiments, and strong dependence of TMR on the interface resonance states in the minority spin channel [51-52]. It is not clear whether the interface resonance states remain relevant under large voltage biases, which will significantly modify the interface electronic structure. Among the finite bias studies, Zhang and White [49] constructed a phenomenological model to explain the voltage-dependence of TMR in terms of the impurity scattering at the interface. They suggested that with a high quality barrier, a different voltage-dependence of TMR may occur. Some groups [55-57] have proposed different models suggesting that even without impurity scattering, the TMR also decrease with the bias. One work attempted to explain the TMR reduction in terms of magnon excitations at the interfaces [58]. However, these models are often based on overly simplified scattering potentials, e. g., simple step barriers, and other assumptions

PAGE 43

32 such as neglecting the k // (Bloch wave vector in two dimensional plan perpendicular to the current) dependence of the transmission and reflection probabilities. Such simplifications have been shown to lead to qualitatively incorrect conclusions in linear-response models [59, 60], including incorrect dependence of the conductance and magnetoresistance on the barrier thickness and height. Today, first-principles calculation [6-20] of electronic structure and tunneling conductance under a finite bias voltage has been successfully applied to molecular electronics. As mentioned in Chapter 1, the first-principles approach can be built on either non-equilibrium Greens functions formulation or scattering wave formulation and the scattering wave has so far been implemented only with jellium electrodes, due to the difficulties associated with computing and storing Bloch wave functions of realistic leads. The layer Korringa-Kohn-Rostoker (KKR) approach [61], which was used successfully in calculating linear-response transport properties of spin-dependent tunneling junctions [51, 52], uses an extended basis set, spherical waves inside atomic spheres and plane waves in the interstitial region. The scattering wave solution for incident Bloch wave functions is provided in Ref. 51. This allowed straightforward calculation of the two-terminal conductance within the linear-response regime. By extending this approach to treat finite bias, we can overcome the shortcoming of the scattering wave method and treat realistic leads. The test system for our calculation is a Fe/FeO/MgO/Fe(100) tunneling junction. This system has good lattice match between the Fe substrate and the MgO lattice. It has been successfully grown experimentally with demonstrated epitaxial growth [62-65]. The inclusion of a single atomic layer of FeO on one of the interfaces is important. An earlier

PAGE 44

33 linear-response calculation [52] for the Fe/MgO/Fe(100) tunneling junction showed a TMR ratio that is two orders of magnitude larger than experimental measurements [65]. In a subsequent calculation [66] this disagreement between theory and experiment was shown to be mostly due to the presence of the FeO layer on the interface which was observed experimentally [63, 64]. A finite voltage bias will change the interface electronic structure, which in turn will impact the TMR ratio. The Fe/FeO/MgO/Fe(100) tunneling junction would be a good system to examine whether the change in electronic structure due to the bias voltage alone can explain the change in TMR. Three effects of finite bias need to be considered in a first-principles theory. The first effect arises from the energy dependence of the number of channels in the wire and, in general, of the transmission probability of the conductor or the tunneling barrier. This can be accounted for by replacing the linear response formula with an integral over the energy. The second is the change in the electronic structure of the sample due to the Coulomb field of the accumulated charge induced by the flow of the current. This is usually small for good conductors, but can be significant in a tunneling device. The third one is the effect of the current on the exchange-correlation potential in a density functional theory approach. Most forms of this potential are developed for equilibrium electron systems, and those incorporating a current are usually rather complex and difficult to implement numerically. In this work we only consider the first two effects of a finite bias, and neglect the effect of a current on the exchange-correlation potential. In the following section, we will show that our approach is equivalent to the Keldysh Greens function approach within the single-particle picture, but without the

PAGE 45

34 need for the truncation of the matrix elements in the Hamiltonian and the overlap matrix as required by other first principles methods using the Keldysh Greens function. 4.2 Non-equilibrium Greens Function Formalism in Real Space The Keldysh formalism incorporates the open boundary conditions of a non-equilibrium system into the components of the Greens functions. Because not all of the Keldysh Greens functions are independent of each other, we can use the usual advanced and retarded Greens functions along with one additional component of the Keldysh Greens functions. Common practice is to use the lesser Greens function which is defined as G ),(),(),;,(''''trtritrtrG (4-1) where the angle brackets indicate the ensemble average and are field creation and destruction operators of electrons. There is a simple relation between the lesser Greens function and the electron density ),;,(),(trtriGtr Since we are only interested in the steady state solution, the time dependence is only through. So, the lesser Greens function can be Fourier transformed into 'ttt tdetrrGErrGdEeErrGtrrGtiEtiE);',(21);',();',();',( (4-2) Note that writing the Greens function with a single energy argument implies that the scattering in the tunneling region is elastic. The charge density is also independent of time, dEErrGir),,(21)( (4-3)

PAGE 46

35 The problem remains to evaluate );',(ErrG For non-interacting Hamiltonians, the approach in the literature [20, 22] requires the decoupling of the two electrodes. This effectively truncates the range the Hamiltonian at a finite distance, and may not be appropriate for some systems. Here, we take a different approach using the scattering wave representation. Our approach will not be easily generalized to interacting electron systems, but is valid within the typical first principles methodology based on the density functional theory. We rewrite the Fourier transformed Greens function in the form tdetrttriErrGtiE),(),(2),',(' (4-4) In the framework of mean-field theory, these two operators can be expanded as single-electron operators nnnnnnCtrtrCtrtr),(),(),(),(* (4-5) where n is the stationary state single particle wave function and two terms are creation and destruction operators for this stationary state. nnCC, Substituting equation (4-5) into equation (4-4), we have nnnnnnnnnnnrrfEEirrCCEEiErrG)()()()()()(),,('*'*' (4-6) where the product is the number of particles that occupy the state nnCC n and its ensemble average is the distribution function Equation (4-6) is a general expression for the lesser Greens function in terms of the stationary wave functions and the nf

PAGE 47

36 corresponding distribution function. Now, we need to adapt this expression for the tunneling junction system. An open system connected to two electrodes has a general geometry as depicted in Fig 1-1. As before the whole system is divided into three regions, the left lead, the device region and right lead. The wave functions and the distribution function must be considered separately within each region and continuous at boundaries. There are two types of stationary states: scattering states and bound states. The basic assumption of our approach is that the left lead maintains an equilibrium distribution function )(LEf which is diagonalized by scattering function incident from the left lead, and likewise for the right lead distribution function )(REf and the scattering wave functions from the right lead. Both and are Fermi-Dirac distribution functions defined with their respective chemical potentials f f L and R Next, using the subscript l to denote all degenerate indices of and correspondingly for, equation (4-6) can be written as (neglecting the bound states for the moment) m lmmmmlllrrEEifrrEEifErrG)()'()()()()(),',(*'* (4-7) We compare this result with that obtained in Chapter 2. Rewrite equation (2-11) as aRraLrGGfGGfG (4-8) Both equations are valid within the single particle picture. However, equation (4-7) is more general than the equation (4-8), as it does not require the complete decoupling of the two electrodes, thus no truncation in the range of the matrix elements of the Hamiltonian is needed. In particular, equation (4-7) applies to the entire space whereas equation (4-8) can be applied only within the device region.

PAGE 48

37 Substituting equation (4-7) into equation (4-3), we obtain the scattering part of the electron density mmRmllLlscatEfEfr22)()()( (4-9) Including the contribution from the bound states, the lesser Greens function contains an additional term where and bbbbbrrfEEi)'()()(* bf b are the distribution function and the single-particle functions of bound states. Accordingly, the electron density also has an extra contribution from bbbbbEf2)( (4-10) The determination of is not unique and is still being investigated by researchers. bf 4.3 Self-consistency and Tunneling Current for A TMR Junction In this section we outline our approach for non-equilibrium self-consistent calculations of electronic structure of a TMR junction. Although the approach applies generally to any lead-device-lead ensemble, the particular Fe/FeO/MgO/Fe tunneling junction in our calculation has a layered structure with a two-dimensional periodicity along each layer. This approach allows a two-dimensional Fourier transform that yields a quasi one-dimensional problem for each reciprocal vector k // The left and right leads are semi-infinite in the direction perpendicular to the layers. They maintain different chemical potentials L and R Without loss of generality, we may assume RL The device region includes all of the tunneling junction and a sufficient number of layers of materials from the left and right leads, so that the potential and the charge density in the lead regions can be taken to be the asymptotic values.

PAGE 49

38 Within each electrode, the Bloch eigenstates can be divided into right-going states and left-going states. The scattering states when there is a device region can be categorized into those incident from the left lead traveling to the right, and those incident from the right lead traveling to the left, Their wave functions can be expressed in terms of the Bloch eigenstates within each electrode, and a scattering wave function in the scattering region, in the form ,,,'''mmmlsllllltt device region (4-11) Left lea d Right lead and ,,,'''''lllmsmmmmmtt device region (4-12) Right lead Left lea d where are the transmission coefficients and are the reflection coefficients. The superscripts indicate the direction of travel along the z direction, and the indices label Bloch states in the left and right electrodes, respectively, and span over all possible values of k ',tt ',rr ml, // and the spin. If one takes the jellium approximation in both leads, the Bloch wave functions become plane waves and become the scattering states used in Ref. 10. With the help of the layer KKR approach, we have gone beyond the jellium approximation and are able to deal with the realistic leads.

PAGE 50

39 At zero temperature, both distribution functions become the step functions )(,RLE Then from equations (4-9) and (4-10) we obtain )()()()(rrdErdErbEErl (4-13) where mmmElllEEErEEr22)()(,)()( (4-14) These equations hold for all regions including the device region and both leads. We can rewrite equation (4-13) as LRrdErrEReq)()()( (4-15) where the equilibrium part of the electron density corresponding to R is defined as RrrrdErbEEReq)()]()([)( (4-16) Note that the superscript R indicates the chemical potential R used in the calculation of yet the density is defined for all three regions. Req Req Now, let us consider the electron density in the two leads separately. In the right lead, the first term in equation (4-15) should give the asymptotic value of the electron density far away from the device region at zero bias. The second term, the non-equilibrium part, is due to the electrons transmitted from left. Neglecting the interference terms between different Bloch states, which should be averaged out in an ensemble average, we have within the right lead E

PAGE 51

40 lmmmlmErtEEr2)()()( (4-17) The electron density in the left lead contains more terms than the right lead. Again, neglecting the interference terms, we can write '2'''2)()()()()(lllllllllErREErEEr (4-18) The normalization of the Bloch wave functions is such that the integral over the volume of a unit cell llllllEEEErd223)()( equals to the density of states at energy E of the left lead, and likewise for the right lead. The self-consistent procedure for the non-equilibrium calculation is very similar to the usual equilibrium calculation for an interface system, except in this case, the charge densities and the electrostatic potentials for all three regions need to be iterated simultaneously, unlike the equilibrium case in which the semi-infinite (lead) parts are calculated using the bulk self-consistent calculations before the interface region is calculated. An additional consideration is given to the relative shifts in the electrostatic potential between the three regions. Although the leads should screen any net charge quickly so that the overall system should be charge neutral, the effect of dipole layers which cause constant shifts in the potential cannot be screened. The size of the dipole moment on each layer also depends sensitively on the convergence of the calculation. To overcome this problem, separate constant shifts in the potential are added to the Kohn-Sham potentials for the three regions. These shifts are adjusted at each LSDA iteration until the net charge within each region reaches zero respectively. In general the shifts in the potential differ from the bias voltage, as has been shown in earlier studies [21].

PAGE 52

41 In order to see how the spin-dependent calculation is done, we write the electron density in equation (4-14) in terms of spin explicitly ////////////2,,,2,,,)()()()(kkkEkkkEEErEEr (4-19) where the term is the spin index. The subscripts now comprise both k ml, // and Remember that sum over ,E ll, and sum over ,E mm, Then equation (4-15) can be written as LRrdErEReq)()(,, (4-20) where the equilibrium part is )()()()(,,,rrrdErbEEReqR (4-21) The last term means only those bound states which have spin are counted. The equation (4-21) can be used to calculate the electron density for each spin over all space. After obtaining the electron density for each spin, we use the common local spin-density functional to calculate the effective potential. Similarly, the current density for each spin is given by a Landauer-type formula lmlmlmtvvdEeJ,2, (4-22) where are the group velocities of the Bloch states, and only the Bloch states for the lmvv, spin are summed. The velocity factors are necessary because the transmission coefficients are defined based on the continuity of the wave function, not the flux.

PAGE 53

42 The equilibrium term in equation (4-20) can be evaluated using the standard contour integration technique for Greens function methods. The non-equilibrium term on the other hand, is more difficult to evaluate numerically. The integration over the energy for the non-equilibrium term has to be computed along the real axis, which is numerically noisy due to the singularities and resonances arising from the three-dimensional Bloch states in the leads. This problem is not unique to the scattering wave function approach employed here. Even if one were to compute the lesser Greens function directly, the integration still would need to be done along the real axis because the lesser Greens function is not analytic in both upper and lower complex energy planes. Thus a complex contour cannot be used in either approach to reduce the numerical noise of the non-equilibrium term. The numerics of the real axis integration can be improved greatly using the technique by Brandbyge et al [33]. We note that equation (4-20) can be equivalently written as (4-23) LRrdErELeq)()(',, where )()()()(,,,rrrdErbEELeqL (4-24) In principle, the electron density evaluated from either equation (4-20) or equation (4-24) should be identical. But because these equations use two different energy contours, the numerical errors are different. It is shown that the numerical error is minimized by combining the two electron densities in the following manner [67]: ')1(ccnew (4-25) where

PAGE 54

43 2,2,2,RneqLneqLneqc (4-26) with LRLRERneqELneqdEdE,,,, (4-27) 4.4 Application to the Fe/FeO/MgO/Fe Tunneling Junction The Fe/FeO/MgO/Fe junction can be grown by depositing MgO epitaxially onto Fe whiskers and then depositing another Fe electrode epitaxially onto the MgO layer. Experimentally it was shown that a single atomic layer of FeO forms when MgO is deposited onto Fe [62-64]. For the MgO/Fe interface, we use the same structure as in Ref. 52. The lattice constant of the Fe layer is fixed at 2.866 In order to match [100] layers of two materials, the MgO lattice constant is taken to be a factor of 2 larger than that of Fe. The distance between Fe and O layers is taken to be 2.16 while the distance between Fe and O atoms in the FeO layer is taken to be 2.1 Because we are unable to incorporate the considerations of chemical or structural disorder into the Keldysh Greens function formalism, a necessary step to account for the under occupancy of the oxygen site in the FeO layer, we assumed 100% oxygen in our calculation. Numerical convergence in terms of the number of energy points between two Fermi energies and the number of k // points in the two-dimensional reciprocal space needs to be considered carefully because the integration is along the real energy axis, and can be prone to be numerical errors, as just discussed. For the self-consistent part of the calculation, we use one energy point for biases below 0.01 Hartree and two energy points

PAGE 55

44 for biases between 0.01 and 0.018 Hartree, the largest bias in our calculation. We find increasing the number of energy points further has little effect on the self-consistent solution. For tunneling conductance calculation, we begin with three energy points in a uniform mesh for a bias of 0.001 Hartree. Then for each increment of 0.001 Hartree in the bias voltage, two additional energy points are added to the mesh. Our calculation at the bias of 0.001 Hartree showed that, when the number of k // points is increased from 2800 to 8262, the change in the conductance of parallel alignment of the magnetic moments in the electrodes is less than 1% and the change for antiparallel alignment is less than 3%. Therefore, 8262 k // points are used in the integration for all of the conductance calculations. In the following discussion, the majority and minority spin channel always refer to the left Fe lead. -0.3 -0.2 -0.1 0 0.1 0.2 0.3 Shift in DOS (eV) Fe O MgO Fe Mg FeO Majority Minority Figure 4-1. Shifts in the density of states in each atomic layer at a bias voltage of 0.54 V. It is interesting to examine the change in the electronic structure due to the applied voltage bias. The main effect of the voltage bias here is the accumulation of the charge on

PAGE 56

45 the interface layers next to the barrier. We find that the change in the net charge on the interface layers is small, and this change does not cause significant changes in the profile of the electronic density of states at each layer. We see the nearly constant shifts in the energy in the barrier that should be consistent with the shift in the electrostatic potential at each layer. The shifts in the DOS between E=-0.5 and E=0.5 Hartree for each layer for the bias voltage of 0.54 V is shown in Fig. 4-1 for parallel alignment of the moments. The DOS for the iron layers in the leads is shifted by roughly 2/1 V on either side of the barrier, and the shifts for the barrier layers correlate monotonically with the layer positions. The shifts for both spin channels are about the same for each layer. Two interesting features occur near the interfaces. On both interfaces, the first insulator layers (FeO or MgO) have the same shifts as the electrodes. This is somewhat surprising, noting that such alignment does not conform to the extrapolation of the shifts of other MgO layer. The nearly exact alignment of the DOS of the interface insulator layers with the electrodes does not agree with the model given by Ref. 34 in which an oscillatory behavior is predicted for the electrostatic potential near the interfaces. On the other hand, we find that the shifts in the DOS scale essentially linearly with the bias voltage, as one would expect. For negative biases, the shifts are the same in magnitude with the opposite signs. From the linear scaling of the DOS shifts with the bias voltage one would expect that the capacitive charge should also scale linearly with the voltage. Surprisingly, this is not the case. In Fig. 4-2 we show the capacitance as a function of the voltage. There are three different charges plotted in this figure, none of which strictly conforms to the usual definition of capacitive charge, due to the difficulty of separating the capacitive dVdQ/

PAGE 57

46 0 0.002 0.004 0.006 0.008 0.01 -0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5 0.6 dQ/dV (electrons/volt) Bias Voltage (volts) Net charge Left electrode Right electrode Figure 4-2. Derivative of the capacitive charge with respect to the bias voltage. The charge on the right electrode includes that on the oxygen atom in the interface MgO layer. Net charge' includes that on both the electrode and the dielectric charges on the MgO layers on the same side of the midpoint of the barrier. charge in the electrodes and the dielectric charge in MgO. The capacitive charge must be antisymmetric, i.e., with on opposite sides of the barrier. This antisymmetry is not possible with any partition of the atoms in our calculation, unless we include the charges on all iron and MgO layers on the same side of the midpoint of the barrier. The curve labeled net charge is this sum per two-dimensional (2D) unit cell. This sum is the same (with opposite signs) for both sides of the barrier. However, the sum includes both the capacitive and the dielectric charges. To separate out the capacitive charge at least approximately, for the left electrode we sum the charge for all iron layers but excluded the FeO layer. For the right electrode we sum the charge for all iron layers plus the charge of the oxygen atom on the interfacial MgO layer. The reason for including the oxygen atom is that it represents more than a quarter of the net change in the charge on Q

PAGE 58

47 the right side, consistent with our observation that this oxygen atom seems to be short circuited with the iron electrode. The two curves in Fig. 4-2 labeled left electrode and right electrode represent this approximation of capacitance. For comparison a model of capacitance per unit cell for two plates separated by a vacuum gap of 21 (which is the distance between two iron electrodes), is about 0.0044 electrons/V. Thus our self-consistent calculation yields an effective dielectric constant for the MgO film (coated with an atomic layer of FeO on one side) of around 2. Since in our calculation no lattice relaxation is included, this number should be compared to the optical dielectric constant measured experimently. The value derived from the index of refraction measurement [35] (n=1.7) for bulk MgO is 9.22n 0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04 -0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5 dm/dV ( /volt) B Bias Voltage (volts) Left Right Bulk electrode (left) Figure 4-3. Derivative of the moment with respect to the bias voltage. The labels left and right indicate that the layers to the left (right) of the midpoint of the barrier are summed. The contribution to the sum from the nonbulk layers of the iron electrodes is taken to be the difference between the moment of that layer and the bulk moment in the electrode on the same side.

PAGE 59

48 Because of the neutrality of the total system, the total net charge integrated over the 2D unit cells of all the layers on one side of the midpoint of the barrier should be the same with the opposite signs from both sides. This charge is shown as net charge in Fig. 4-2. Evidently, it is not symmetric with respect to the bias voltage. For positive biases, it drops quickly as the voltage increases. At about 0.5 V, it becomes negative. This means that at this voltage, any increase in the charge in the electrode due to the increasing bias is completely matched and canceled by the opposite dielectric charge from the MgO layer. An unexpected result is the change in the magnetic moment at the interfaces for parallel alignment of the moments. For a symmetric junction such as Fe/MgO/Fe, if the net moment increases with bias on one interface, it must decrease on the other interface due to symmetry. For the asymmetric junction of Fe/FeO/MgO/Fe, we find that the moments on both interfaces increase with positive bias and decrease with negative bias. Thus the derivative of the net moment with respect to the voltage is always positive, as shown in Fig. 4-3. Unlike the charge, there is no indication of saturation of the moment changes at the interface as a function of the bias. The asymmetric changes in the moments mean that the majority spin electrons react differently to the bias voltage than the minority spin electrons. Since the rate of the change in the moment is almost an order of magnitude larger than the rate of change in the charge, charge accumulation in the two spin channels must be opposite in sign. The curve labeled bulk electrode in Fig. 4-3 is the change in the moment in the bulk for the left electrode. The changes in the moment in the bulk regions on the two electrodes are the same in magnitude but opposite in sign. The change in the interface magnetic moment in response to the bias voltage is detectable experimentally if one can

PAGE 60

49 keep the right electrode sufficiently thin. The right electrode (which does not have an FeO layer at the interface) is the last deposited iron layer and its moment changes in the opposite direction as the interfacial moment as a function of the bias voltage. At zero bias, the conductance is given by lmmllmtvveG22 (4-28) It is important that our calculation from the finite bias approach extrapolates to the linear response results obtained in earlier works [51, 52, 66]. The k // -resolved conductance at zero bias, shown in Fig. 4-4, reproduces all of the features seen in earlier works. For the parallel case, the majority spin channel is dominated by the contribution from a region around k // =0. The conductance due to the minority spin channel is about two orders of magnitude smaller and is not shown in the figure. For the antiparallel case, our calculations show that the predominant contribution to the minority spin current comes from tunneling through the interface resonant states which correspond to sharp peaks of the transmission. The calculation gives a negative TMR of -28% at zero bias. We note that an earlier work obtained a small positive TMR ratio for the same junction [66], but our calculation yields a small negative TMR ratio. This difference appears to be due to the extreme sensitivity of the minority spin channel conductance to the convergence of the integration over k // The earlier calculation used 2800 k // points while we used significantly more k // points (8262). Nonetheless, the general trend studied, that of TMR as a function of oxygen concentration in Ref. 31, and the TMR as a function of bias voltage studied here, should remain valid as long as the number of k // points used is consistent within each study.

PAGE 61

50 -0.2 -0.15 -0.1 -0.05 0 0.05 0.1 0.15 0.2 Kx -0.2 -0.15 -0.1 -0.05 0 0.05 0.1 0.15 0.2 Ky 0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 Transmission (10 ) -5 (a) -0.2 -0.15 -0.1 -0.05 0 0.05 0.1 0.15 0.2 Kx -0.2 -0.15 -0.1 -0.05 0 0.05 0.1 0.15 0.2 Ky 0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 Transmission (10 ) -4 (b) Figure 4-4. Zero bias transmission probability as a function of k // (a) Majority spin channel for parallel alignment of the moments; (b) The dominant spin channel for antiparallel alignment of the moment.

PAGE 62

51 -40 -30 -20 -10 0 10 20 30 -0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5 Current density (10 A/m ) 4 2 Bias Voltage (Volts) Majority Minority Majority:Non-SCF Figure 4-5. Current density as a function of bias voltage for majority and minority spin channels for the case of parallel alignment of the moments between electrodes. -10 -8 -6 -4 -2 0 2 4 6 -0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5 Current density (10 A/m ) 4 2 Bias Voltage (Volts) Current:SCF Current:Non-SCF Figure 4-6. Current density as a function of bias voltage for majority and minority spin channels for the case of antiparallel alignment of the moments between electrodes.

PAGE 63

52 The tunneling current is calculated through equation (4-22). The effect of the bias voltage, through the change of the electro-chemical potentials in the two electrodes, enters this equation in two ways. First, the bias voltage can change the transmission probability for each conduction channel, i.e., for each k // point at each energy. Second, when the bias increases, the number of conduction channels in the leads that contribute to tunneling also increases. An interesting question is how much of these effects is captured by rigidly shifting the electrostatic potential for each atomic layer and neglecting the effects of charge rearrangement, as was commonly done in many early non-self-consistent calculations, and, conversely, how important is charge self-consistency in accurate determination of the TMR. For this purpose, we compare the fully self-consistent calculations of the tunneling current and the TMR with a calculation constructed from a self-consistent zero bias potential, in which the electrostatic potentials in the left and right electrodes are rigidly shifted by 2/1 V where V is the bias voltage, and the potentials of the barrier layers are shifted linearly with the bias according to their positions. Because of the asymmetry of the tunneling junction, the positive and negative biases give different tunneling currents. Figure 4-5 shows the I-V curves for the majority and minority spin channels for parallel alignment of the moments. As expected, we see a large linear regime in this figure. For low biases, the slope of the curve gives the linear response conductance a value of. This number agrees very well with )/(1015.127m the result obtained from equation (4-28) which is. Note also that the non-self consistent calculation yields the same results within the linear region (for bias )/(1017.127m

PAGE 64

53 voltage up to about 0.1 V) but deviates significantly at higher biases, reaching about twice the self-consistent results at 0.5 V. The current of minority spin channels actually 0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 -0.2 -0.15 -0.1 -0.05 0 0.05 0.1 0.15 0.2 Transmission (10 ) -5 Kx (Ky) .0_Volts .054_Volts .108_Volts Figure 4-7. Transmission probability at energy RE as a function of k // along the line of for the majority spin channel for parallel alignment of the moments at different bias voltages. yxkk shows an oscillatory behavior. Since it is negligibly small compared to the majority spin channels, we cannot see this in Fig. 4-5. Figure 4-6 shows the total tunneling current for antiparallel alignment of the moments. The curve labeled SCF is in fact calculated by simply flipping the moment alignments of half the layers of the corresponding parallel SCF calculations, thus this is not a truly SCF calculation for the antiparallel alignment. However, one fully SCF calculation for the antiparallel alignment was performed at V=0.5 V and the difference is negligible both in terms of charge transfer and tunneling current. The curve labeled non-SCF is calculated with the rigid shift of a self-consistent zero-bias electrostatic potential. For most of the bias voltages the tunneling currents agree between the SCF and non

PAGE 65

54 SCF calculations, except for large negative biases. The tunneling current exhibits an oscillatory behavior as a function of positive biases, probably due to the matching and mismatching of interface resonance states on both sides of the barrier as the bias voltage moves these states in energy and k // We will see this later from the analysis of integrated transmission. Let us consider a positive bias as an example to study the effects of bias on the tunneling process. By positive bias we mean that the electron flux is from the left (FeO side) to the right, or that the current is from the right to the left. We calculated the transmission probability for the spin channel with the largest contribution to the current at the energy RE and along a line of yxkk in the 2D reciprocal space, under different positive bias voltages. For the case of parallel alignment of the moments in both leads, Fig. 4-7 shows the results of the majority spin channel at bias voltages of 0.0, 0.054, and 0.108 V. All curves exhibit very similar structures. It is clear that the bias does not affect the transmission very much. This suggests that the linear regime of I-V curve for the majority spin channel may extend up to 0.108 V. For the antiparallel case, Fig 4-8 shows the results for the same range of bias voltages. Although all curves has the similar overall structure (the transmission is low at k // =0 and have one peak on each side of k // =0), the position, height and width of these peaks change dramatically as functions of bias. From 0.0 to 0.082 V, the peaks are enhanced, and from 0.082 to 0.108 V, the peaks are reduced. As discussed in an earlier linear response study [20], the position and the strength of these peaks are determined by several factors. Among them are the amplitude of the wave function at the interfaces on both sides of the barrier, and the decay rate across the barrier as determined by the complex band structure in the barrier. When

PAGE 66

55 interface resonance states exist, the amplitudes of the wave function can be enhanced by several orders of magnitude at the interface through coupling to those interface resonance states. This coupling gives rise to the sharp peaks in the transmission probability at certain k // points. Applying a bias voltage or changing it will change the energy and k // positions of these interface resonance states, in particular the relative positions of interface resonance states on both sides of the barrier can be moved. This in turn can cause large changes in the transmission probability at corresponding k // values. 0 0.5 1.0 1.5 2.0 2.5 3.0 -0.2 -0.15 -0.1 -0.05 0 0.05 0.1 0.15 0.2 Transmission (10 ) -5 Kx (Ky) .0_Volts .054_Volts .082_Volts .108_Volts Figure 4-8. Transmission probability at energy RE as a function of k // along the line of for the minority spin channel for antiparallel alignment of the moments at different bias voltages. yxkk As mentioned at the beginning of this section, when the bias voltage increase, the number of conduction channels in the leads that contribute to tunneling also increase. In order to examine this effect, we define the integrated transmission as LRdEktkTRL)(1)(//// (4-29)

PAGE 67

56 In Fig. 4-9 and 4-10, we show this integrated transmission probability as a function of k // along a line of k x =k y in the 2D reciprocal space. The patterns shown in these two figures closely resemble those in Fig 4-7 and 4-8. However, the integration over the energy has smeared out to some degree the effects due to the interface resonance states, so that the change in transmission probability as function of bias is less dramatic. Based on these figures, we can expect that the majority spin tunneling current for the parallel case should scale approximately linearly with the bias, and the minority spin tunneling current and the antiparallel tunneling current should increase more slowly than linear with the bias. This behavior should lead to an increase of the TMR ratio as a function of the bias voltage. 0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 -0.2 -0.15 -0.1 -0.05 0 0.05 0.1 0.15 0.2 Integrated transmission (10 ) -5 Kx (Ky) .0_Volts .054_Volts .108_Volts Figure 4-9. Integrated transmission probability as a function of k // along the line of for the majority spin channel for parallel alignment of the moments at different bias voltages. yxkk Although the negative bias affects the system differently, the general feature is the same: the transmission probability of dominant channel of antiparalle case is much more

PAGE 68

57 sensitive to bias than that of parallel case. At small bias, the antiparallel tunneling current also increases more slowly than the parallel tunneling current as a function of bias voltage. Figures (4-12) and (4-13) show the integrated transmission as a function of k // for all k // points at a higher bias of 0.378 V. Compared to Fig 4-4, the broad peak in the parallel case becomes broader, a result of the integration over the contribution from all channels at different energies. For the antiparallel case, the structure of the peaks due to the interface resonance states becomes more complicated than at zero bias. The TMR at finite bias is defined as papJJJ/)( where are total currents for the parallel and antiparallel case respectively and its linear-response limit is (G apJJ, p -G a )/G a as given in the introduction. The voltage dependence of the TMR is shown in Fig. 4-13. Previous calculations that used simplified models [49, 50, 58] yielded TMR ratios that are large and positive at zero bias, then decrease rapidly with bias. Our calculations yield a very low TMR ratio at low biases, even negative at 0 and 027.0 V. Then, for positive biases, the TMR ratio increases rapidly as the bias voltage is changed from 0.054 to 0.108 V. This increase is caused by the decrease of the antiparallel current. When the bias voltage is larger than 0.108 V, the TMR ratio starts to oscillate with the bias, but overall still rises with the bias, albeit more slowly. For negative biases, the TMR increases from -0.027 to -0.432 V and decreases at -0.486 V, which is caused by slow increase of parallel tunneling current. The comparison between the SCF calculation and the non-SCF calculation for TMR is similar to that of the tunneling current for parallel spin alignment. This indicates that the charge self-consistency is more important for the majority spin channel than the

PAGE 69

58 0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 -0.2 -0.15 -0.1 -0.05 0 0.05 0.1 0.15 0.2 Integrated transmission (10 ) -5 Kx (Ky) .0_Volts .027_Volts .054_Volts .082_Volts .108_Volts (a) 0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 -0.2 -0.15 -0.1 -0.05 0 0.05 0.1 0.15 0.2 Integrated transmission (10 ) -5 Kx (Ky) .108_Volts .136_Volts .163_Volts (b) Figure 4-10. Integrated transmission probability as a function of k // along the line of for the majority spin channel for parallel alignment of the moments at different bias voltages. yxkk

PAGE 70

59 -0.2 -0.15 -0.1 -0.05 0 0.05 0.1 0.15 0.2 Kx -0.2 -0.15 -0.1 -0.05 0 0.05 0.1 0.15 0.2 Ky 0 0.5 1.0 1.5 2.0 2.5 3.0 Transmission (10 ) -5 Figure 4-11. Integrated transmission probability as a function of k // for the majority spin channel for parallel alignment of the moments at bias 0.378 V. -0.2 -0.15 -0.1 -0.05 0 0.05 0.1 0.15 0.2 Kx -0.2 -0.15 -0.1 -0.05 0 0.05 0.1 0.15 0.2 Ky 0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 Transmission (10 ) -5 Figure 4-12. Integrated transmission probability as a function of k // for the minority spin channel for antiparallel alignment of the moments at bias 0.378 V.

PAGE 71

60 minority spin channel, presumably due to the sensitivity of the majority spin tunneling current to the chemical bond configuration at the interfaces [66]. The tunneling through the interface resonance states leads to the sharpest peaks of the transmission. Due to the presence of roughness and impurities at the interfaces, the interface resonance states may not be present or sharply located in experimental measurements. In Fig. 4-13 we also indicated the TMR ratio at zero bias if we removed all of the contributions from the interface resonance states. Although the TMR ratio increases significantly, this increase does not change the overall trend of increasing TMR ratio with bias. Thus we conclude that at least for Fe(100) electrodes, where a single atomic layer of FeO is formed at one of the interfaces, the TMT ratio is close to zero at small biases, and should increase with the bias voltage. We believe that this conclusion should also apply to barrier materials other than MgO, especially when the barrier layer is asymmetric as a result of the order of deposition, so that the TMR ratio is low at small biases. This conclusion supports the suggestion by Zhang and White [9], that for high quality barrier, the voltage-dependence of TMR may be different than the observation in earlier experiments that the TMR ratio decreases with bias. For the system we studied, this unusual behavior is based on following facts. First, for the parallel case, the transmission of dominant conduction channels (peak for majority spin channel) is not sensitive to the bias. This gives a large linear regime as a function of bias. For the antiparallel case, the dominant conduction channels are sensitive to the interface electronic structure, in particular, interface resonance states, which in turn are sensitive to the bias voltage. Often the maximum of transmission occurs when a tunneling state couples efficiently to both leads. This may be the case when interface resonance

PAGE 72

61 -0.5 0 0.5 1 1.5 2 2.5 3 3.5 4 -0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5 TMR Bias Voltage (Volts) TMR:SCF TMR:Non-SCF A Figure 4-13. Magnetoresistance ratio as a function of bias voltage. Point A denotes the MR ratio at zero bias by removing all contribution from interface resonance states. states are present on both interfaces. When the bias increases, it is likely that the once efficient coupling to both leads is greatly reduced due to the change in the interface resonance states. This would lead to a great reduction in the tunneling current for the minority spin channel and for the antiparallel case. Thus, it is possible for TMR to increase with the bias. Although one may expect that for amorphous barriers the effects of interface resonance states are much reduced, the impurity states at the interface may play the same role as the interface resonance states in the system we discussed in this work, so that the argument presented here may still be qualitatively true. Thus, it may be possible to observe the similar behavior of TMR even for amorphous barriers. 4.5 Conclusion In conclusion, we have completed the first spin-polarized calculation using the non-equilibrium first-principle method for spin-dependent tunneling junction under a

PAGE 73

62 finite bias. Application of this method to the layer KKR treatment of the Fe/FeO/MgO/Fe tunneling junction leads to surprising results that are not anticipated by simple models but are reasonable in view of the change in the electronic structure at the interface in response to the bias voltage. Our calculations indicate a negative TMR ratio at very low biases and a rapid increase of the TMR ratio at finite biases. We believe that this behavior is not limited to the Fe/FeO/MgO/Fe tunneling junctions and may be observed experimentally in other systems. This result has import practical implications since increasing the TMR ration at finite bias voltages is one of the important goals of spintronics research. Our calculations also show significant deviations from non-SCF calculations using rigidly shifted electrostatic potentials at bias voltages about 0.1 V, as well as the nonlinear capacitive charging effect and the asymmetric magnetic moment change as a function of bias voltage. All these results highlight the important role of a fully self-consistent non-equilibrium calculation.

PAGE 74

CHAPTER 5 PROBLEMS IN THE CURRENT NON-EQUILIBRIUM APPROACH FOR OPEN SYSTEMS BASED ON DFT 5.1 Introduction As discussed in previous chapters, for the open system under a finite bias, the non-equilibrium first-principles approach has been proved to be successful. The approach is based on density functional theory, and the typical procedure can be summarized in two steps as we mentioned in Chapter 2: the first is the self-consistent procedure based on DFT to determine the effective single-electron potential. Here, of course, the density functional used is valid for ground states only, but is assumed to be valid for non-equilibrium systems because no alternative approaches exist. The second step is to use the Landauer conductance formula to calculate the current. The Landauer formula relates the current to the integral of the single-particle transmission function over energy. Such approaches have been used in many systems with varying degrees of success: the shape of the I-V curve can be compared to the experiment and the structure of I-V curve can be explained by molecular orbitals. All these achievements are important for the understanding and future design of molecular electronics. However, large discrepancies are often seen between theoretical calculations and actual measurements of I-V curves. Among possible sources of error in the theory, the use of the particular type of the density functional in a calculation is often a cause of concern. Much progress has been made in finding an energy functional form that works well for both the metallic leads and the 63

PAGE 75

64 molecular junction. Unfortunately, the problem is more fundamental than just the functional. The basis of DFT is the first Hohenberg-Kohn theorem [23], which states that for the ground state, the external potential is determined, within an additive constant by the electron density )(rVext 0V )(r Since all ground state properties are determined by the Hamiltonian, which is uniquely defined once the external potential is determined, the one-to-one correspondence between and )(rVext )(r guarantees that all ground state properties are determined by the electron density )(r In other words, )(r is the basic variable of the system at ground state and all physical quantities can be expressed as functionals of )(r The constant potential is trivial for ground state, since if changes, the Fermi energy and the electronic structure of the system simply shifts accordingly. The theorem can be easily proved through the fact that the ground state has the lowest energy. However, the same proof does not apply to an open system under a bias voltage, since the system is in a non-equilibrium state, and there is no minimum energy principle for such a state. Before we use the DFT in such a system, we have to question whether the electron density 0V 0V )(r remains a fundamental variable that uniquely describes the system. In the following section, we show that the electron density alone is not sufficient to describe a non-equilibrium open system. Then we propose a modification of the Hohenberg-Kohn Theorem for an open system at finite bias. We follow that with a numerical study that gives a functional form of the exchange energy for non-equilibrium open systems.

PAGE 76

65 5.2 Hohenberg-Kohn Theorem for Open System under Finite Bias To make our discussion more concrete, we choose a model open system with the general structure shown in Fig. 1-1. The system is divided into three regions: the left lead, the scattering region (the device region), and the right lead. The scattering region consists of the left contact, the barrier and the right contact. The two contacts are parts of leads that connect to the middle barrier. The leads are connected with two reservoirs at infinity (not shown in Fig. 1-1) and the chemical potentials of the two reservoirs are constants. A discussion of the contact between leads and reservoirs can be found in reference 68. The two leads are often to be assumed to be internally at thermodynamic equilibrium [20], but not in equilibrium with the contacts or the scattering region through the contacts. The electronic structure of the two leads far away from the scattering region are assumed to be identical to their bulk values [20], and two separate calculations based on first-principles method are often performed to obtain properties of the leads. Furthermore, the temperature of both leads is assumed to be zero. The bias voltage is defined as the difference between Fermi energies L and R maintained by the two leads connected to two reservoirs, respectively. The purpose of the study is to calculate the current through the system as a function of the bias voltage. Let us now examine the question whether the electron density )(r can uniquely determine the transport properties of the system, specifically, whether )(r alone is sufficient in determining the current through the system. Because the current is a function of the bias voltage, if )(r is the basic variable of the system in the same way it is for a ground state, knowing )(r should allow one to determine the voltage bias across the scattering region.

PAGE 77

66 In both leads, the potential and the electron density can be approximately taken as the ground state bulk value, due to the screening by the two contacts. This is the approximation generally used in the previous studies. Suppose the first Hohenberg-Kohn theorem for the ground state can apply at least in the two leads. Then, according to the theorem, the Fermi energy L is decided by the electron density in the left lead )(rL to an constant and the Fermi energy LV R is decided by the electron density in the right lead )(rR to another constant Since both leads are connected to a battery, these two constants, and should not be independent. However, according to the ground state Hohenberg-Kohn theorem, if we shift Fermi energies of left and right leads, thus changing the voltage bias across the scattering region, the electron density in the two leads should remain unchanged. In another words, these two constants, and are independent if only given electron density. RV LV RV LV RV The result is that the voltage bias RL is completely arbitrary, if one is only given the electron density in the two leads. Since the bias voltage should not depend on the scattering region, we come to the conclusion that for the open system under a finite bias, the electron density )(r alone is not sufficient to describe all properties of the system, thus conventional DFT is not applicable. The Hohenberg-Kohn theorem must be modified in order to be applied to such systems. 5.2.1 Voltage-dependence of the Density Functional In order to see how the Hohenberg-Kohn theorem should be changed, we start from a rigorous approach to this problem based on the adiabatic time-dependent procedure. We take the same time-dependent process as that in reference 24 and reference 25: at time the unperturbed system consists of three uncoupled regions, the left lead, the t

PAGE 78

67 scattering region, and the right lead. Each one maintains its own equilibrium state (ground state for zero temperature). The Fermi energies for the left and right leads are L and R respectively. The Fermi energy of the scattering region does not need to be the same as left or right leads, but for simplicity, we assume that it is in the same equilibrium state as the right lead. The bias voltage is defined as RL The couplings between leads and the scattering region turn on slowly from t to 0 t such that at any time, the system is in the stationary state. The goal is to relate the current at with the couplings completely turned on to the unperturbed Fermi energies 0t L and R Given the initial state, the properties of the time-dependent system can be obtained by solving the time-dependent Schrdinger equation )()()(ttHtti (5-1) The Hamiltonian can be written as ictijiijeeiiextiierVtVrVrVHtVHtH)1)((')()()(21)()(200 (5-2) where, in the Hamiltonian, the first term 0H consists of the kinetic energy, the nuclear-electron interaction and the electron-electron interaction The second term is an extra time-dependent external potential, where can be chosen to be two barriers in the two contacts and zero in all other regions. The two barriers should be high enough such that at the transmission probability of electrons tunneling through these two barriers is negligible; thus, the two leads and the middle barrier are well separated. The extV eeV 'V t

PAGE 79

68 constant c should be positive and small enough such that the couplings between leads and the middle barrier turn on adiabatically. For such a time-dependent system, Runge and Gross [69] showed that the wave function at time, t )(t can be determined by the time-dependent electron density )(t and the initial state, up to a time-dependent phase factor, )(),()()( tetti (5-3) where, (, is a functional of the electron density and the initial state. Particularly, if the initial state is the stationary ground state at t then according to the first Hohenberg-Kohn theorem for ground state, it is completely determined by the initial electron density; hence )( can be eliminated in the equation (5-3), so that )(t is determined up to a phase factor by only. Neglecting the phase factor, we come to the time-dependent DFT in this case, which claims that any time-dependent physical quantity can be expressed as the functional of time-dependent electron density. However, for our case, at t the system is not in the ground state. The initial state )( should be constructed to represent three local equilibrium states for three uncoupled regions with two different Fermi energies L and R which actually is an excited state at zero temperature characterized by L and R Apparently, the initial state is determined by the Hamiltonian and these two Fermi energies. Since a trivial constant added to the Hamiltonian will not affect the initial state, but will shift L and R by the same amount, we know that the initial state is determined by the Hamiltonian and the voltage bias RLbV

PAGE 80

69 Applying the first Hohenberg-Kohn theorem for the ground state to three regions which are in local equilibrium states, the Hamiltonian can be determined by the initial electron density to a constant. As we discussed before, the bias voltage is independent of the electron density, so the initial state is a functional of electron density and the voltage bias. bV At, the Equation (5-3) can be written as, 0t btiVtet),()()( (5-4) which means that at a given bV )(t is determined by to a time-dependent phase factor. bV, can be equivalently written as bVb| which means that the density functional form depends on the bias voltage We come to a conclusion that all steady-state properties at can be expressed as functionals of bV 0t at a given bias voltage. The most important of these functionals is the exchange-correlation energy functional bV xcE The voltage-dependence of those density functionals actually comes from the voltage-dependence of the initial state. The first Hohenberg-Kohn theorem for the equilibrium state is just a special case for 0 bV The voltage-dependence of the density functionals introduce substantial difficulty to the implementation of the theory, since we need different functional forms for different bias voltages. It is even worse than that: since the voltage-dependence of the initial state varies with the system, the voltage-dependence of the density functionals varies with the system, which means we don't have universal density functional for all systems even when they are under the same bias voltage. This disadvantage is caused by the fact that

PAGE 81

70 the basic variable of the functional, the function )(r can not determine the bias voltage and is not sufficient to describe the system. In order to get rid of this disadvantage, we need to introduce other functions as variables of the functional, which together with bV can completely determine the system. 5.2.2 The Non-equilibrium Electron Density The key to eliminate the voltage-dependence of the functional in the equation (5-4) is to introduce other functions into the functional which can determine the bias voltage and thus determine the initial state bV )( together with the electron density For the non-equilibrium system, the electron density can be expressed as [20, 22], ),(21)(rGdir (5-5) where, is the non-equilibrium lesser Green's function defined for the stationary state. For the system under study, which is characterized by two Fermi energies ),(rG L and R (suppose RL ), the electron density can be divided into the equilibrium part eq and the non-equilibrium part n )()()(rrrneq (5-6) where, RRrGdirrGdirneq),(21)(),(21)( (5-7) The term eq is called the equilibrium density in the sense that it can be equivalently written as an equilibrium integral [20],

PAGE 82

71 RrGdrReq),(Im1)( (5-8) where, is the usual retarded Green's function. RG We want to show that if given both eq and n at the initial time, the bias voltage is determined and then the initial state is determined. When t the left lead is well separated with respect to other regions. The left lead is in a local equilibrium state with the Fermi energy L while the scattering region and the right lead are in a local equilibrium state with the Fermi energy R In this case, n in the right lead is zero and in the left lead it also can be written as an equilibrium integral, LRrGdrRn),(Im1)( (5-9) Since the left lead is in local equilibrium, the first Hohenberg-Kohn theorem for equilibrium state applies. We know that in the left lead, the Hamiltonian is determined by )()()(rrrneq to an arbitrary constant. If a constant is added to the Hamiltonian, changes to. Then, with 0V )(RG )(0VGR eq given, from equation (5-8), we see that R changes to 0VR Introduce the new R and into the equation (5-9), we see that with the given )(RG n L becomes 0VL Hence the bias voltage RLbV is completely determined without any arbitrary constant. Then, the Equation (5-4) can be rewritten as, )0(),0()0()0(ttetneqti (5-10)

PAGE 83

72 where, the voltage-dependence of the functional is eliminated. Based on the functional neq , all steady state properties can be expressed as functionals of eq and n 5.2.3 The Stationary Principle and the Single Particle Picture For the time-dependent system, we can define the action A as, 0)()()(dtttHtitA (5-11) where )(t is the many body wave function. If )(t is normalizable, then given the initial state, )( the stationary requirement of the action, 0)(tA (5-12) is completely equivalent to equation (5-3), where )(t is an arbitrary change of the wave function under the conditions that 0)0( and the normalization is unchanged. There is no difficulty to normalize the wave function in the closed system, but it seems to be impossible to normalize the wave function for the open system which has the geometry shown in Fig (1-1). Thus, the Equation (5-12) can not be directly applied. However, for the case of time-dependent DFT, which means that the initial state is a stationary ground state, the action is the functional of the electron density AA As long as the wave function )(t yields the correct electron density, we still can use the Equation (5-12) to calculate the action and the stationary requirement becomes, 0)(tA (5-13) with the conditions that 0)0( and doesn't change the neutrality condition of two leads and the scattering region. This procedure can be generalized to the non-equilibrium

PAGE 84

73 initial state case. According to the previous subsection, for the open system under a finite bias, the action can be written as neqA and the stationary principle becomes, 0)()(tAtAneq (5-14) where the arbitrary change of electron densities should not change the initial conditions nor the neutrality condition of the system. For the adiabatic time-dependent process, the action reduces to 0)(dttEA (5-15) Then, the Equation (5-14) becomes, 0)()(tEtEneq (5-16) where, E is the total energy of the system at time t. Both the ground state and the time-dependent DFT work in the framework of a single-particle picture. We follow the same procedure. Consider an independent particle system, which has a same geometry as shown in Fig. 1-1 and whose single particle orbitals can be written as ,m where denotes all good quantum numbers other than the energy m At the initial time, since the two leads are well-separated with respect to the scattering region, ,m consists of three kinds of states, ,l ,r and ,b which are localized in the left lead, the right lead, and the scattering region respectively. Those three kinds of states obey the equilibrium distribution functions of three regions respectively. When the temperature is zero, those distribution functions are step functions. The adiabatic time-dependent procedure described in previous sections will not change the occupation number of each state when it evolves with the time. Further

PAGE 85

74 assume that at the initial time, ,l ,r and ,b are perfectly normalized such that the neutrality conditions are satisfied in all three regions. Since the system is non-interacting, both Green's functions and are diagonalized by )(G )(rG ,m It is straightforward that equation (5-7) becomes, RLRmmmnmmeqf,2,,2,)( (5-17) where is the occupation number. Since the occupation number of each state does not change with time and mf RL n is, RLlln,2, (5-18) Assume that eq and n yielded by equaton (5-17) and (5-18) are the same as those of the true interacting system under study. Then, similar to the equilibrium state DFT, we write the total energy as VETE where T is the single-particle kinetic energy, RLRlllnmmmeqneqTTTTT,2,22121 (5-19) The potential energy can be written as, VE )()(121)()(212121rrrrrdrdEEErrVrdECoulxcCoulextV (5-20)

PAGE 86

75 where is the classical electron-electron repulsion energy and is the exchange-correlation energy. is the functional of CoulE xcE VE eq and n We can define two effective potentials based on as, VE nVneffeqVeqeffEVEV (5-21) The Equation (5-16) can be written in terms of those orbitals as, 0)21(0)21(,2,2RLRmneffmeqeffVV (5-22) There are two boundary conditions for the system under study. The open boundary condition incorporated with equation (5-22) yields the scattering states and the bound state boundary condition yields bound states. Unlike equilibrium DFT, we have two different effective potentials instead of one. The non-equilibrium electrons see a different mean field potential than the equilibrium electrons do. The self-consistent procedure makes two electron densities eq and n to converge respectively, and the current is carried by the non-equilibrium electrons. 5.3 Exchange energy for uniform electron gas The exchange-correlation energy can be written as the sum of the exchange part and the correlation part, cxxcEEE Both and are functionals of xE cE eq and n They are very complicated for a real, inhomogeneous system. In this section, as a case study, we calculate the exchange energy for the uniform electron gas under a finite bias, which represents a perfect conductor connected to two reservoirs infinitely far away.

PAGE 87

76 This study serves as a test for the theory described in the previous section. For the uniform system, if we take periodic boundary condition as )()(xlx then the single-particle orbitals can be written as, rkizkykxkizyxeVe l kkkzyx 2/1)(2/311),,( (5-23) where zyxzyxnlk,,,,2 with are integers. Set the axis to go from left to right, which is the direction of the current. There are two kinds of states for this system: the right-going states () and the left-going states ( zyxn,, z 0zk 0 zk ). From equations (5-17) and (5-18), we see that the occupation of those states is: below, all states are occupied and between and, only right-going states are occupied. Define two quantities and by ek ek tk ek tk LtRemkmk2,22222 Two electron densities eq (which comes from all states below ) and ek n (which comes from all right-going states above ), can be related to and After some algebra, we have ek ek tk 233 eeqk and )(61332etnkk We can see more clearly from this case that the total electron density neq alone is not enough to describe the system: two systems with the same have different electronic structures if they have different set of eq and n The exchange energy can be evaluated by the equation [30] xE 2122112),(141rdrdrrrEx (5-24) where ),(21rr is the density matrix and its diagonal part is the electron density. It can be

PAGE 88

77 calculated as krkieVrr 122),(21 where the summation is over all occupied For a large number of occupied states, the summation can be replaced by an integral, k 1212202/023200023214141),(rkikkrkikeddSindkkeddSindkkrrtee (5-25) where the first integral is the contribution of all states below which is an equilibrium state integral and can be evaluated analytically. We denote this term as ek eq The second integral is the contribution of occupied states between and, which can be denoted as ek tk n For the non-equilibrium integral, since only states with positive are occupied, the variable zk of the second integral goes from 0 to 2/ We introduce two variables 2/)(21rrR and Then the density matrix is only a function of and the exchange energy can be written as, 12rs s 2)(141ssKKRdEx (5-26) where K is a constant since the system is uniform. For a given electron density when the bias voltage increases, n increases and eq decreases until at a certain bias voltage maxV eq becomes zero and n equals to defines the biggest difference between maxV L and R we can have for a uniform system with a given at zero temperature. Given the electron density and the bias voltage RL we can numerically evaluate and K Fig. 5-1 shows the ratio as a function of total electron density and bias voltage. denotes the term calculated at zero bias. From the 0/KK 0K

PAGE 89

78 figure, we clearly see the voltage-dependence of the exchange energy for a given electron density This supports the analysis of the previous section: the electron density alone is not enough to describe the system. Also, the -dependence of the exchange energy at a given bias voltage, which will determine the form of the density functional, varies with the bias voltage. It means that we will have difficulty in finding a universal functional for all bias voltages, which agrees with the previous section too. =7.30 =33.77=17.29=2.16 Figure 5-1. Exchange energy at different bias voltages for different (Electrons/m 3 ) We notice that although the voltage-dependence of the exchange energy varies with the electron density, the behavior is similar for all electron densities: as the voltage increases, the exchange energy increases until it reaches a maximum and then decreases. In order to understand this behavior better, we rewrite the Eq. (5-26) as,

PAGE 90

79 sdsKsdsKsdsKKKKKneqneqneqnneqeqneqneq)(141141141**22 (5-27) The first term is the contribution of the exchange of equilibrium electrons. The second term comes from the non-equilibrium electrons and the third one denotes the contribution of exchanging between equilibrium and non-equilibrium electrons. The voltage-dependence of these three terms for eqK nK 29.7 Electrons/m 3 is shown in Fig. 5-2. At bias zero, since n is zero, there are no non-equilibrium electrons. Hence and are zero. As the bias voltage increases, nK neqK eq decreases, which leads to the decrease of, and eqK n increases, which leads to the increase of both and. The term reaches its maximum at a certain bias voltage and then decreases. At the maximum bias voltage, nK neqK neqK n and there are no equilibrium electrons, then both and are zero at this bias voltage. eqK neqK According to the previous section, the exchange energy can be expressed as a functional of and n We write K as, nfKK ,10 (5-28) where, 0K is the one for the equilibrium system with the electron density ; this term is well known. We plot the numerical results of as the function of 0/KK /n for different electron densities in Fig. 5-3.

PAGE 91

80 =7.30 (a) =7.30 (b) Figure 5-2. Split of the exchange energy: (a) the exchange between non-equilibrium electrons and the exchange between non-equilibrium and equilibrium electrons and (b) the exchange between equilibrium electrons.

PAGE 92

81 The numerical results show that we have a universal functional nf for all electron densities. These data can be fit to the function, nnbaf2 (5-29) where and 0497.0a 0436.0b =7.30 =2.16 =17.29 =33.77 / Figure 5-3. Exchange energy as the functional of /n for different electron densities. Then the exchange energy can be written as, nnxxbardCE223/41 (5-30) where, 7386.0xC Equation (5-30) gives the exchange energy functional of the uniform system for all bias voltages. It clearly shows the dependence on the non-equilibrium electron density as well as the dependence on the total electron density, which support the analysis of the

PAGE 93

82 previous section. Whether or not this functional can be generalized to the inhomogeneous system has to be resolved in a future study. 5.4 Conclusion In conclusion, we have shown that the transport properties of an open system under a finite bias can not be determined by the total electron density alone. We introduce two electron densities eq and n to describe such a system and show that the non-equilibrium electrons experience different effective potential from the equilibrium electrons. As a test, we study the exchange energy of the uniform electron gas under a finite bias. The numerical results clearly show the dependence to the non-equilibrium electron density, which support the theory. We fit the numerical data to the exchange energy functional. The applicability of the functional to the inhomogeneous systems needs to be addressed in future studies.

PAGE 94

CHAPTER 6 SUMMARY AND CONCLUSIONS We introduce the main ideas of first principles method for open system under a finite bias. The method has been successfully applied to study the transport properties at nano scale, for example, molecular electronics. The achievements of the method can be summarized as follows. The shape of calculated I-V curve can be compared to the experiments and peaks in the curve can be explained by tunneling through molecular orbitals. We apply the state-of-art approach which combines the non-equilibrium Greens functions technique and DFT to Au/S/Azobenzene/S/Au molecular junction. The results show that the azobenzene molecule is a good candidate of light-driven molecular switch, which is important for application of nanoelectronics. The I-V curve of cis molecular shows a negative differential conductance at a certain bias, which is due to the suppression of the tunneling through HOMO due to bias voltage. We also generate the method to include spin and implement the method in a wave function representation. The method is used to study the TMR through a layered Fe/FeO/MgO/Fe magnetic junction. The results suggest a bias-enhanced TMR which is rarely seen in previous experiments. The enhancement of TMR due to bias voltage is explained by the effects of tunneling through the surface resonance states. Our analysis suggests that this phenomenon should be able to seen in other systems. In the final part of the thesis, we discussed the problem in the current approach and suggest that in order to correctly describe the non-equilibrium effects introduced by the 83

PAGE 95

84 external bias, an additional parameter other than the total electron density should be introduced into the equilibrium DFT. For a case study, we calculate the exchange energy for an uniform electron gas and clearly see the voltage-dependence of exchange energy. By introducing non-equilibrium electron density, we show that the voltage-dependence of the functional can eliminated. This work is supported by the Alumni Fellowship of University of Florida and the US Department of Energy under contract DE-FG02-02ER45995.

PAGE 96

APPENDIX A EXPANSION OF LESSER GREENS FUNCTION The non-equilibrium Greens functions are defined as )'1()1()'1,1(CTiG (A-1) where, the shorthand notation ),(111tx is used. The time-order operator is defined on two time branches, + branch, which runs from CT t to t and branch, which runs in the opposite direction. With two time labels, and which can be located on either of two time branches, we have four kinds of non-equilibrium Greens functions: 1t '1t )'1,1()'1,1()'1,1()'1,1()'1,1(~ttGGGGG '11'11'11'11,,,,tttttttt (A-2) These four Greens functions can be denoted as the time-ordered Greens function the greater Greens function the lesser Greens function and the antitime-ordered Greens function tG G G tG ~ For the Hamiltonian which is given by equation (2-1), a particular lesser Greens function defined as nkG, )()(,tcditGknnk (A-3) 85

PAGE 97

86 where, the indices ,k denote channels of the left lead and denotes the basis states of the device region. The angle brackets indicate the ensemble average. The time of the operator lies on the branch and the time of the operator lies on the + branch. n nd kC Figure A-1. Expansion of lesser Greens function The time of the perturbation can be either on + branch or on branch. nkG, The coupling term CouH is taken to be the perturbation V Then, the Dyson equation can be used to expand the Greens functions, where the term is the Greens function without the perturbation. Inserting the basis states into the Dyson equation, we have, VGGGG00 0G mmnmkkknkGVgnVGGknGknGkG,,00, (A-4) where, the term nGk0 is zero because electrons can not propagate from the left lead to the device region without the coupling term and the term is the bare Greens function of the left lead. The equation (A-4) can be used for both equilibrium and non-equilibrium Greens functions. For non-equilibrium case, since the time of the perturbation can be either on + branch or branch, the equation (A-4) gives two terms. kkg, + nt ktnt mmnmktkkGVg,,mtmnmkkkGVg ~ ,,ktVtVt +

PAGE 98

87 As an example, the expansion of the lesser Greens function contains two terms as shown in Fig. A-1. Thus, can be written as nkG, nkG, mtmnkkmntkkmknkGgGgVG][ ~ ,,,, (A-5)

PAGE 99

APPENDIX B DETAILS OF NON-EQUILIBRIUM SELF-CONSISTENT PROCEDURE As mentioned in Chapter 3, the device region includes one unit cell of the left lead (denoted as L), the molecule (denoted as M), and one unit cell of the right lead (denoted as R). The first step of the calculation is to compute self-energies of two leads according to equation (3-2). Since the tight-binding model is used to describe two leads, only the part L (R) of the device region is affected by the left (right) semi-infinite lead. Thus the dimension of the self-energy of the left (right) lead L ( R ) is the number of basis states of one unit cell of left (right) lead. In our calculation, the Gaussian basis set is used. Due to screening, self-energies of two leads are not affected by the device region, so they dont change in the self-consistent procedure. The self-consistent procedure starts from an initial electron density )(0r in the device region. Given the electron density, the Hamiltonian can be calculated as )(][')'(')(21002rVVrrrrdrVHbiasxcion (B-1) where, the last term )(rVbias is the linear voltage drop, ydrVLRbias)( The current is flowing in y direction and d measures the distance between two leads. The retarded and advanced Greens function can be achieved by RLarHSG1, (B-2) where, S is the overlap matrix of Gaussian basis states and H is the Hamiltonian matrix. 88

PAGE 100

89 In order to see it more clearly, equation (B-2) can be expanded as 1RRRRRMRMRLRLRRMRMMMMMLMLMRLRLMLMLLLLLLrHSHSHSHSHSHSHSHSHSG (B-3) where, indices L, M, and R denote three parts of the device region respectively. Then, the lesser Greens function can be computed according to equation (2-11) and the density matrix is determined by dGiijij)(21 (B-4) where, indices j i, denote Gaussian basis states. The electron density can be directly related to the density matrix as jijijijirrrjjiirrrr,*,)()()( (B-5) where, )(ri is the Gaussian basis in real space. The electric static potential is defined as )(')'(')()(rVrrrrdrVrVbiasiones (B-6) An interesting quantity is the difference between the electric static potential under a finite bias and the electric static potential of zero bias, ,0esV ')'(')(0rrrrdrVVVVbiaseseses (B-7) which shows the effects of the voltage bias on the electric static potential.

PAGE 101

LIST OF REFERENCES [1] C. Joachim, J. Gimzewski, R. Schlittler, C. Chavy, Phy. Rev. Lett 74 2102 (1995). [2] M. A. Reed, C. Zhou, C. J. Muller, T. P. Burgin, and J. M. Tour, Science 278 252 (1997). [3] D. J. Wold, R. Haag, M. A. Rampi, and C. D. Frisbie, J. Phys. Chem. B 106 2813 (2002). [4] M. A. Reed, Proc. IEEE 97, 652 (1999). [5] C. Dekker, Phys. Today 22 (1999). [6] A. Aviram, and M. A. Ratner, Chem. Phys. Lett. 29 277 (1974). [7] N. D. Lang, Phys. Rev. B 52, 5335 (1995). [8] N. D. Lang, Phys. Rev. B 55, 4113 (1997). [9] N. D. Lang and Ph. Avouris, Phys. Rev. Lett. 81, 3515 (1998). [10] M. Di Ventra, S. T. Pantelides, and N. D. Lang, Phys. Rev. Lett. 84, 979 (2000). [11] N. D. Lang and Ph. Avouris, Phys. Rev. Lett. 84, 358 (2000). [12] C. C. Wan, J.-L. Mozos, G. Taraschi, J. Wang, and H. Guo, Appl. Phys. Lett. 71, 419 (1997). [13] J. Wang, H. Guo, J.-L. Mozos, C. C. Wan, G. Tarachi, and Q. Zheng, Phys. Rev. Lett. 80, 4277 (1998). [14] G. Tarachi, J.-L. Mozos, C. C. Wan, H. Guo, and J. Wang, Phys. Rev. B 58, 13138 (1998). [15] Yongqiang Xue, Supriyo Datta and Mark A. Ratner, Chem. Phys. 281, 151 (2000). [16] H. J. Choi, J. Ihm, Y. Y. Yoon, and S. G. Louie, Phys. Rev. B 60, 14009 (1999). 90

PAGE 102

91 [17] U. Landman, R. N. Barnett, A. G. Scherbakoy, and P. Avouris, Phys. Rev. Lett. 85, 1958 (2000). [18] B. Larade, J. Taylor, H. Mehrez, and H. Guo, Phys. Rev. B 64, 075420 (2001). [19] H. J. Choi, J. Ihm, S. G. Louie, and M. L. Cohen, Phys. Rev. Lett. 84, 2917 (2000). [20] J. Taylor, H. Guo, and J. Wang, Phys. Rev. B 63, 121104 (2001). [21] H. Mehrez, Alex. Wlasenko, Brian Larade, Jeremy Taylor, Peter Grutter, and Hong Guo, Phys. Rev. B 65, 195419 (2002). [22] Y. Xue, S. Datta, S. Hong, R. Reifenberger, J. I. Henderson, and C. P. Kubiak, Phys. Rev. B 59, 7852 (1999). [23] P. Hohenberg, W. Kohn, Phys. Rev. 136, B864 (1964). [24] Yigal Meir and Ned S. Wingreen, Phys. Rev. Lett. 68, 2512 (1992). [25] C. Caroli, R. Combescot, Pl. Nozieres, and D. Saint-James, J. Phys. C 4, 916 (1971). [26] C. Adamo and V. Barone, J. Chem. Phys. 110, 6158 (1998). [27] A. Szabo and N. S. Ostlund, Modern Quantum Chemistry: Introduction to Advanced Electronic Structure Theory (Dover Publications, New York, 1996). [28] D. E. Bernholdt, E. Apra, H. A. Fruchtl, M. F. Guest, R. J. Harrison, R. A. Kendall, R. A. Kutteh, X. Long, J. B. Nicholas, J. A. Nichols, H. L. Taylor, A. T. Wong, G. I. Fann, R. J. Littlefield, and J. Nieplocha, Int. J. Quantum Chem., Quantum Chem. Symp. 29, 475-483 (1995). [29] R. M. Dreizler and E. K. U. Gross, Density Functional Theory (Springer-Verlag, Berlin, Germany, 1990). [30] R. G. Parr and W. T. Yang, Density-Functional Theory of Atoms and Molecules (Oxford University Press, Oxford, UK, 1989). [31] Francesca Moresco, Gerhard Meyer, and Karl-Heinz Rieder, Phys. Rev. Lett.. 86, 672 (2000). [32] J. M. Seminario, P. A. Derosa, and J. L. Bastos, J. Am. Chem. Soc. 124, 10266 (2001). [33] R. Gutierrez, G. Fagas, G. Cuniberti, F. Grossmann, R. Schmidt, K. Richter, Phys. Rev. B 65, 113410 (2002). [34] P. Orellana and F. Claro, Phys. Rev. Lett. 90, 178302 (2003).

PAGE 103

92 [35] E. G. Emberly and G. Kirczenow, Physl. Rev. Lett. 91, 188301 (2003). [36] S. Monti, G. Orlandi, and P. Palmieri, Chem. Phys. 71,87 (1982). [37] H. Rau, J. Photochem. 26, 221 (1984). [38] T. Nagele, R. Hoche, W. Zinth, and H. Wachtveitl, Chem. Phys. Lett. 272, 489 (1997). [39] F. W. Schulze, H. J. Petrick, H. K. Cammenga, and H. Klinge, Z. Phys. Chem. Neue Fol. 107, 1 (1977). [40] E. R. Talaty and J. C. Fargo, Chem. Commun. 2, 65 (1967). [41] T. Hugel, N. Holland, A. Cattani, L. Moroder, M. Seitz, H. Gaub, Science 296, 1103 (2002). [42] P. S. Krstic, X.-G. Zhang, and W. H. Buttler, Phys. Rev. B 66, 205319 (2002). [43] S. Sanvito, C. J. Lambert, J. H. Jefferson, A. M. Bratkovsky, Phys. Rev. B 59, 11936 (1999). [44] R. Landauer, IBM J. Res. Dev. 1, 233 (1957). [45] S. A. Rishton, Y. Lu, R. A. Altman, A. C. Marley, X. P. Bian, C. Jahnes, R. Viswannathan, G. Xiao, W. J. Gallagher, and S. S. P. Parkin, Microelectron. Eng. 35, 249 (1997). [46] S. S. Parkin, K. P. Roche, M. G. Samant, P. M. Rise, R. B. Beyers, R. E. Scheuerlein, E. J. O'Sullivan, S. L. Brown, J. Bucchigano, D. W. Abraham, Y. Lu, M. Rooks, Pl L. Trouilloud, R. A. Wanner, and W. J. Gallagher, J. Appl. Phys. 85, 5828 (1999). [47] J. S. Moodera, E. F. Gallagher, K. Robinson, and J. Nowak, Appl. Phys. Lett. 70, 3050 (1997). [48] C. T. Tanaka, J. Nowak, and J. S. Moodera, J. Appl. Phys. 86, 6239 (1999). [49] J. Zhang and R. M. White, J. Appl. Phys. 83, 6512 (1998). [50] P. Zahn and I. Mertig, Phys. Rev. Lett. 75, 2996 (1995). [51] J. M. MacLaren, X.-G. Zhang, W. H. Butler, and Xindong Wang, Phys. Rev. B 59, 5470 (1999). [52] W. H. Butler, X.-G. Zhang, T. C. Schulthess, and J. M. MacLaren, Phys. Rev. B 63, 054416 (2001). [53] J. Mathon and A. Umerski, Phys. Rev. B 63, 220403(R) (2001).

PAGE 104

93 [54] O. Wunnicke, N. Papanikolaou, R. Zeller, P. H. Dederichs, V. Drchal, and J. Kudmovski, Phys. Rev. B 65, 064425 (2002). [55] Xindong Wang, J. Appl. Phys. 83, 6518 (1998). [56] A. H. Davis and J. M. MacLaren, J. Appl. Phys. 87, 5224 (2000). [57] T. Nagahama, S. Yuasa, Y. Suzuki, and E. Tamura, J. Appl. Phys. 91, 7035 (2002). [58] S. Zhang, P. M. Levy, A. C. Marley, and S. S. P. Parkin, Phys. Rev. Lett. 79, 3744 (1997). [59] J. M. MacLaren, X.-G. Zhang, and W. H. Butler, Phys. Rev. B 56, 11827 (1997). [60] W. H. Butler, X.-G. Zhang, T. C. Schulthess, and J. M. MacLaren, Phys. Rev. B 63, 092402 (2001). [61] J. M. MacLaren, S. Crampin, D. D. Vvednsky, R. C. Albers, and J. B. Pendry, Comput. Phys. Commun. 60, 365 (1990). [62] W. Wulfhekel, M. Klaua, D. Ullmann, F. Zavaliche, J. Kirschner, R. Urban, R. T. Monchesky, and B. Heinrich, Appl. Phys. Lett. 78, 509 (2001). [63] H. L. Meyerheim, R. Popescu, J. Kirschner, N. Jedrecy, M. Sauvage-Simkin, B. Heinrich, and R. Pinchaux, Phys. Rev. Lett. 87, 076102 (2001). [64] H. L. Meyerheim, R. Popescu, N. Jedrecy, M. Sauvage-Simkin, R. Pinchaux, B. Heinrich, and J. Kirschner, Phys. Rev. B 65, 144433 (2002). [65] M. Bowen, V. Cros, F. Petroff, A. Fert, C. M. Boubeta, J. L. Costa-Kramer, J. V. Anguita, A. Cebollada, F. Briones, J. M. DeTeresa, L. Morellon, M. R. Ibarra, F. Guell, F. Peiro, and A. Cornet, Appl. Phys. Lett. 79, 1655 (2001). [66] X.-G. Zhang, W. H. Butler, and A. Bandyopadhyay, Phys. Rev. B 68, 092402 (2003). [67] M. Brandbyge, J. Mozos, P. Ordejon, J. Taylor, K. Stokbro, Phys. Rev. B 65, 165401 (2002). [68] T. Ando, Y.Arakawa, K. Furuya, S. Komiyama and H. Nakashima, Mesoscopic Physics and Electronics (Springer-Verlag, Berlin, 1998). [69] E. Runge and E. K. U. Gross, Phys. Rev. Lett. 52, 997 (1984).

PAGE 105

BIOGRAPHICAL SKETCH Chun Zhang was born on August 3, 1973, in Yongshun County, Hunan Province, China. He got his B. S degree in Nuclear Science in Fudan University, Shanghai, China in 1996 and M. S degree in physics in the same University in 2000. He entered the physics department of University of Florida in August, 2000. 94


Permanent Link: http://ufdc.ufl.edu/UFE0008621/00001

Material Information

Title: Transport properties at nano scales via first principles studies
Physical Description: Mixed Material
Language: English
Creator: Zhang, Chun ( Dissertant )
Cheng, Hai Ping ( Thesis advisor )
Hebard, Arthur F. ( Reviewer )
Herschfield, Selman ( Reviewer )
Roitberg, Adrian ( Reviewer )
Trickney, Samuel B. ( Reviewer )
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2004
Copyright Date: 2004

Subjects

Subjects / Keywords: Physics thesis, Ph.D
Dissertations, Academic -- UF -- Physics

Notes

Abstract: There are two main difficulties for the first principles study of transport properties at the nano scale. The first is that many-body interactions need to be taken into account for the infinite system without periodic boundary conditions. The other is that the system is usually in a non-equilibrium state. Both of these two difficulties are beyond the ability of conventional first principles methods to reconcile. Recently, a new first principles approach which combines the Non-equilibrium Green's Functions Technique (NGFT) and the Density Functional Theory (DFT) was proposed. DFT has been proved to be successful in molecular and solid state physics. Currently used DFT approximations can take into account 'most' many-body effects and NGFT naturally includes the non-equilibrium effects. The new approach uses NGFT to treat the non-periodic boundary conditions and DFT to treat many body interactions. This approach has been successfully used in molecular electronics. The thesis is organized in the following way. First: we introduce the main ideas of combining NGFT and DFT, and then apply this method to a light-driven molecular switch. The switch, made of a single molecule, is one of the most important elements of nano-electronics. However, most proposed molecular switches are driven either by an external bias voltage or by STM manipulation, neither of which is ideal for nano-scale circuits. The switch we designed has a high on-off conductance ratio and more importantly, can be driven by photons. In following chapters, we generalize the method to the spin-dependent case and apply it to a magnetic layered structure. We implemented the method within the framework of the Layer Korringa-Kohn-Rostoker (LKKR) approach, which is particularly well-adapted to the layered structure and found a bias-enhanced tunneling magneto-resistance (TMR) for the Fe/FeO/MgO/Fe junction. Our results are important not only for application, but also for understanding of the voltage-dependence of TMR for layered structures. The experimental studies show that the bias voltage usually kills the TMR of amorphous magnetic tunneling junctions. Our study shows that for an impurity-free layered structure, a different behavior of TMR may occur.
Subject: density, light, magnetoresistance, molecular, nanoelectronics, nonequilibrium, quantum, spin, transport
General Note: Title from title page of source document.
General Note: Document formatted into pages; contains 105 pages.
General Note: Includes vita.
Thesis: Thesis (Ph.D.)--University of Florida, 2004.
Bibliography: Includes bibliographical references.
General Note: Text (Electronic thesis) in PDF format.

Record Information

Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
System ID: UFE0008621:00001

Permanent Link: http://ufdc.ufl.edu/UFE0008621/00001

Material Information

Title: Transport properties at nano scales via first principles studies
Physical Description: Mixed Material
Language: English
Creator: Zhang, Chun ( Dissertant )
Cheng, Hai Ping ( Thesis advisor )
Hebard, Arthur F. ( Reviewer )
Herschfield, Selman ( Reviewer )
Roitberg, Adrian ( Reviewer )
Trickney, Samuel B. ( Reviewer )
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2004
Copyright Date: 2004

Subjects

Subjects / Keywords: Physics thesis, Ph.D
Dissertations, Academic -- UF -- Physics

Notes

Abstract: There are two main difficulties for the first principles study of transport properties at the nano scale. The first is that many-body interactions need to be taken into account for the infinite system without periodic boundary conditions. The other is that the system is usually in a non-equilibrium state. Both of these two difficulties are beyond the ability of conventional first principles methods to reconcile. Recently, a new first principles approach which combines the Non-equilibrium Green's Functions Technique (NGFT) and the Density Functional Theory (DFT) was proposed. DFT has been proved to be successful in molecular and solid state physics. Currently used DFT approximations can take into account 'most' many-body effects and NGFT naturally includes the non-equilibrium effects. The new approach uses NGFT to treat the non-periodic boundary conditions and DFT to treat many body interactions. This approach has been successfully used in molecular electronics. The thesis is organized in the following way. First: we introduce the main ideas of combining NGFT and DFT, and then apply this method to a light-driven molecular switch. The switch, made of a single molecule, is one of the most important elements of nano-electronics. However, most proposed molecular switches are driven either by an external bias voltage or by STM manipulation, neither of which is ideal for nano-scale circuits. The switch we designed has a high on-off conductance ratio and more importantly, can be driven by photons. In following chapters, we generalize the method to the spin-dependent case and apply it to a magnetic layered structure. We implemented the method within the framework of the Layer Korringa-Kohn-Rostoker (LKKR) approach, which is particularly well-adapted to the layered structure and found a bias-enhanced tunneling magneto-resistance (TMR) for the Fe/FeO/MgO/Fe junction. Our results are important not only for application, but also for understanding of the voltage-dependence of TMR for layered structures. The experimental studies show that the bias voltage usually kills the TMR of amorphous magnetic tunneling junctions. Our study shows that for an impurity-free layered structure, a different behavior of TMR may occur.
Subject: density, light, magnetoresistance, molecular, nanoelectronics, nonequilibrium, quantum, spin, transport
General Note: Title from title page of source document.
General Note: Document formatted into pages; contains 105 pages.
General Note: Includes vita.
Thesis: Thesis (Ph.D.)--University of Florida, 2004.
Bibliography: Includes bibliographical references.
General Note: Text (Electronic thesis) in PDF format.

Record Information

Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
System ID: UFE0008621:00001


This item has the following downloads:


Full Text












TRANSPORT PROPERTIES AT NANO SCALES VIA FIRST PRINCIPLES STUDIES


By

CHUN ZHANG


















A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA


2004



























Copyright 2004

by

Chun Zhang



































To my parents.















ACKNOWLEDGMENTS

First, I would like to thank my advisor, Dr. Hai-Ping Cheng. She brought me into

the field of computational physics and exposed me to its frontier. She continuously

supported me throughout the past four years and most importantly, taught me that

independent thinking is true heart of the research. Research is lonely and boring, but

thanks to her, it can also be interesting.

I would also like to thank Professors Arthur Hebard, Selman Hershfield, Adrian

Roitberg and Samuel B. Trickey for serving on my supervisory committee.

Dr. Xiaoguang Zhang, a senior scientist at Oak Ridge National Lab, is another

important person to my Ph. D study. Not only his knowledge, but also his attitude toward

the research will have a long-term impact to my career.

My life at the University of Florida would have been black and white without my

friends. They are Dr. Wuming Zhu, Dr. Maohua Du, Dr. Linlin Wang, Dr. Ping Jiang and

Dr. Yao He.

I would also like to extend special thanks to the following professors at Fudan

University, Dr. Yuanxun Qiu, Dr. Jingguang Che and Dr. Jian Zi. Becoming their friend

is one of the most important reasons that I want to be a physicist.
















TABLE OF CONTENTS

page

A C K N O W L E D G M E N T S ................................................................................................. iv

LIST OF FIGURES ......... ..................................... ....... vii

A B ST R A C T .......... ..... ...................................................................................... x

CHAPTER

1 O V E R V IEW ................................. .. .................................................... 1

2 FIRST PRINCIPLES APPROACH COMBINING NGFT AND DFT.....................4

2.1 M odel Hamiltonian and Landauer-type Formula ..................................................4
2.2 The Combination of DFT and NGFT for An Open System ...................................8
2.2.1 Self-Consistent Procedure for Ground State of Closed System ...................8
2.2.2 Self-consistent Procedure for Open System ........ ........... .................. 10

3 COHERENT ELECTRON TRANSPORT THROUGH AN AZOBENZENE
MOLECULE: A LIGHT-DRIVEN MOLECULAR SWITCH............... ...............12

3.1 Introduction ................................................. ........ .. ....................... 12
3.2 Tight-binding M odel and Self-energies...................... ..... ................. 14
3.3 C onductance at Z ero B ias......................................................................... ... ... 18
3.4 The Sw itch under A Finite Bias....................................... ......................... 23

4 QUANTUM TUNNELING THROUGH MAGNETIC JUNCTIONS ......................30

4 .1 Introdu action ................ .............. ....... ...... ..................................... 30
4.2 Non-equilibrium Green's Function Formalism in Real Space ...........................34
4.3 Self-consistency and Tunneling Current for A TMR Junction...........................37
4.4 Application to the Fe/FeO/MgO/Fe Tunneling Junction.............................. 43
4.5 Conclusion ..................... .................... ...........................61

5 PROBLEMS IN THE CURRENT NON-EQUILIBRIUM APPROACH FOR OPEN
SYSTEMS BASED ON DFT .............. .............. .... ............... 63

5.1 Introduction .................. ...... ............ ....................... ................ .....63
5.2 Hohenberg-Kohn Theorem for Open System under Finite Bias ........................65









5.3 Exchange Energy for Uniform Electron Gas............. .. ........... ............... 75
5.4 Conclusion ..................................... ................................ .......... 82

6 SUMM ARY AND CONCLUSIONS..................................................................... 83

APPENDIX

A EXPANSION OF LESSER GREEN"S FUNCTION ....................... ................85

B DETAILS OF NON-EQUILIBRIUM SELF-CONSISTENT PROCEDURE ...........88

L IST O F R E F E R E N C E S ......... .. ............... ................. ................................................90

BIOGRAPH ICAL SKETCH ...................................................... 94















LIST OF FIGURES


Figure p

1-1. Partition of the open system connected to two metal electrodes ............................2

3-1. Physical models of an azobenzene switch in contact with two gold leads: trans
isomer (upper) and cis isomer (lower). ........................................ ............... 14

3-2. Tight-binding model of an open system connected to two semi-infinite leads. Indices
i, j denote nodes of device region and a, b are the nearest node of left and right
lead. Only the nearest-neighbor interaction is considered. .....................................15

3-3. A perfect lead that consists of two semi-infinite leads and a hopping term. Indices m
and i belong to the left semi-infinite lead and i+1, j belong to the right semi-infinite
lead ................................................................................. 16

3-4. Density of states (DOS) projected onto the switch region (upper panel) and
transmission as a function of energy (lower panel). The solid lines represent the
trans configuration and dotted lines the cis. Intense peaks in the DOS are truncated
to im prove readability. ......................................... ......................... 19

3-5. Local DOS of the trans and cis configurations decomposed into contributions on the
Au layers (lower panel), S atoms (middle panel), and the azobenzene molecule
(upper panel). The solid lines represent the trans configuration and dotted lines cis.20

3-6. Local DOS at the Fermi energy projected on seven sites in the switch. From 1-7, or
from left to right, Au lead, S atom, benzene ring, double-bonded N2, another
benzene ring, S atom and Au lead ............................... .....................22

3-7. Non-equilibrium self-consistent procedure. The non-equilibrium Green's function is
calculated according to equation (2-11). ....................................... ............... 24

3-8. The difference of electronstatic potential between bias one volt and bias zero for cis
configuration. The x-z plane is parallel to the N2 bond, cutting through the
m idpoint of A u unit cell. ................................................................. .....................25

3-9. The transmission probability as a function of single-particle energy and the external
bias voltage for cis configuration. ........................................ ......................... 26

3-10. Current as a function of bias voltage for both trans and cis configurations. The red
line represents the current of trans and the blue line represents the current of cis...27









3-11. The differential conductance as a function of bias voltage for both trans and cis
configurations............................................................................................. 28

3-12. The transmission as a function of energy at 0.64, 0.72 and 0.8 Volts for cis
configuration ...................................................... ................. 2 8

4-1. Shifts in the density of states in each atomic layer at a bias voltage of 0.54 V..........44

4-2. Derivative of the capacitive charge with respect to the bias voltage. The charge on
the right electrode includes that on the oxygen atom in the interface MgO layer.
'Net charge' includes that on both the electrode and the dielectric charges on the
MgO layers on the same side of the midpoint of the barrier...............................46

4-3. Derivative of the moment with respect to the bias voltage. The labels 'left' and
'right' indicate that the layers to the left (right) of the midpoint of the barrier are
summed. The contribution to the sum from the nonbulk layers of the iron electrodes
is taken to be the difference between the moment of that layer and the bulk moment
in the electrode on the sam e side ....................................... ................. ......... 47

4-4. Zero bias transmission probability as a function of k/, (a) Majority spin channel for
parallel alignment of the moments; (b) The dominant spin channel for antiparallel
alignment of the moment......................................... ...... ............... 50

4-5. Current density as a function of bias voltage for majority and minority spin channels
for the case of parallel alignment of the moments between electrodes ..................51

4-6. Current density as a function of bias voltage for majority and minority spin channels
for the case of antiparallel alignment of the moments between electrodes.............51

4-7. Transmission probability at energy E = /, as a function of k// along the line of
k = ky for the majority spin channel for parallel alignment of the moments at
different bias voltages. ..................... ................ .............................53

4-8. Transmission probability at energy E = /, as a function of k// along the line of
k = ky for the minority spin channel for antiparallel alignment of the moments at
different bias voltages. ..................... ................ .............................55

4-9. Integrated transmission probability as a function of k// along the line of kx = k, for
the majority spin channel for parallel alignment of the moments at different bias
v o lta g e s ....................................................................... 5 6

4-10. Integrated transmission probability as a function of k// along the line of k = ky for
the majority spin channel for parallel alignment of the moments at different bias
v o lta g e s ....................................................................... 5 8









4-11. Integrated transmission probability as a function of k// for the majority spin channel
for parallel alignment of the moments at bias 0.378 V. ................. ...............59

4-12. Integrated transmission probability as a function of k// for the minority spin channel
for antiparallel alignment of the moments at bias 0.378 V. ................................59

4-13. Magnetoresistance ratio as a function of bias voltage. Point A denotes the MR ratio
at zero bias by removing all contribution from interface resonance states .............61

5-1. Exchange energy at different bias voltages for different p (Electrons/m3) .............78

5-2. Split of the exchange energy: (a) the exchange between non-equilibrium electrons
and the exchange between non-equilibrium and equilibrium electrons and (b) the
exchange between equilibrium electrons. ..................................... ............... 80

5-3. Exchange energy as the functional of p/ p for different electron densities ..........81

A-1. Expansion of lesser Green's function. The time of the perturbation can be either on
+ branch or on branch...................... .. .. ......... .. ........................ ............... 86















Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy

TRANSPORT PROPERTIES AT NANO SCALES VIA FIRST PRINCIPLES STUDIES

By

Chun Zhang

December 2004

Chair: Hai-Ping Cheng
Major Department: Physics

There are two main difficulties for the first principles study of transport properties

at the nano scale. The first is that many-body interactions need to be taken into account

for the infinite system without periodic boundary conditions. The other is that the system

is usually in a non-equilibrium state. Both of these two difficulties are beyond the ability

of conventional first principles methods to reconcile. Recently, a new first principles

approach which combines the Non-equilibrium Green's Functions Technique (NGFT)

and the Density Functional Theory (DFT) was proposed. DFT has been proved to be

successful in molecular and solid state physics. Currently used DFT approximations can

take into account 'most' many-body effects and NGFT naturally includes the non-

equilibrium effects. The new approach uses NGFT to treat the non-periodic boundary

conditions and DFT to treat many body interactions. This approach has been successfully

used in molecular electronics. The thesis is organized in the following way. First: we

introduce the main ideas of combining NGFT and DFT, and then apply this method to a

light-driven molecular switch. The switch, made of a single molecule, is one of the most









important elements of nano-electronics. However, most proposed molecular switches are

driven either by an external bias voltage or by STM manipulation, neither of which is

ideal for nano-scale circuits. The switch we designed has a high on-off conductance ratio

and more importantly, can be driven by photons. In following chapters, we generalize the

method to the spin-dependent case and apply it to a magnetic layered structure. We

implemented the method within the framework of the Layer Korringa-Kohn-Rostoker

(LKKR) approach, which is particularly well-adapted to the layered structure and found a

bias-enhanced tunneling magneto-resistance (TMR) for the Fe/FeO/MgO/Fe junction.

Our results are important not only for application, but also for understanding of the

voltage-dependence of TMR for layered structures. The experimental studies show that

the bias voltage usually kills the TMR of amorphous magnetic tunneling junctions. Our

study shows that for an impurity-free layered structure, a different behavior of TMR may

occur.














CHAPTER 1
OVERVIEW

As the trend of miniaturization continues, we eventually enter the regime of nano-

electronics, the technology of building electronic devices on the basis of a single

molecule. Experiments have shown the great potential of modem technologies, such as

STM and AFM, to make molecular transistors and switches etc. [1-5]. For theoretical

studies, the circuit in which the molecule is embedded provides great difficulties. First,

the molecule is connected to the circuit through two open boundaries. It is crucial to

describe the open boundaries in a way that the interaction between the molecule and the

complicated circuit is taken into account. Second, the circuit is almost always far away

from the equilibrium state and in most cases, varying with time. The time-dependent

process is not the focus of this thesis. We assume that the whole circuit is in a steady

state. Our goal is to find the steady-state solution for an open system far away from

equilibrium.

An open system connected to two metal electrodes usually has geometry such as

shown in Fig. 1. The whole system can be divided into three parts, the left lead, the right

lead and the device region. The device region includes the molecule and enough portions

of the two metal electrodes so that the electrons in the two leads are not affected by

interactions in the device region because of screening. A finite bias is applied to the

system. The goal of the research is to relate the steady-state current flowing through the

device region to the external bias voltage. Traditionally, the electron transport

phenomena through such an open system are studied by the Boltzman equation in









conjunction with the effective-mass assumption. The Boltzman equation is based on a

probabilistic description for the evolution of a single-particle distribution f(ri, p, t)

which is a function of classical variables r and p However, when the device region is at

nano-scale, the phase coherence of charge carriers is usually maintained in the whole

device region so that quantum effects dominate. For this kind of coherent quantum

transport phenomena, the first principles approach which combines the non-equilibrium

Green's functions technique (NGFT) and Density Functional Theory (DFT) has proven to

be successful [6-22]. The method can be summarized in two steps. First, a non-

equilibrium self-consistent procedure based on DFT and NGFT is used to determine the

single-particle effective potential. Second, a Landauer-type formula is used to calculate

the current, which relates the current to the integral of the single-particle transmission

probability over energies.


Current I




Left lead Device Region Right lead



Figure 1-1. Partition of the open system connected to two metal electrodes.

The approach combining NGFT and DFT is based on the single-particle picture.

The tunneling process is explained within this framework as follows. The Bloch electrons

in two leads tunnel through the molecular orbital in the device region. The success of this

picture can be best seen in the work recently published by a Denmark group [21], where









the current flowing through a broken Au nanowire is calculated. At small bias voltages,

the theory agrees well with the experiment.

The approach can also be equivalently built on a scattering wave formulation [8-

11]. The scattering wave approach has so far been implemented only with jellium

electrodes, due to the difficulties associated with computing and storing Bloch wave

functions of realistic leads.

In Chapter 2, we introduce the details of the two-step approach.

In Chapter 3, we apply the method to the Au/Azobenzene/Au junction. Our

calculation shows that the azobenzene molecule is a good candidate for the light-driven

molecular switch.

In Chapter 4, we generalize the method to spin-dependent tunneling junctions with

application to a layered magnetic Fe/FeO/MgO/Fe junction. The calculation suggests a

bias-enhanced tunneling magneto-resistance (TMR), which is important for both

application and the understanding of TMR through a layered structure.

In Chapter 5, we discuss the problems that exist in the current two-step approach.

For the first step, in which the effective single-particle potential is determined, DFT is the

biggest problem. The Hohenberg-Kohn theorem, the basis of DFT, is correct only for the

ground state [23]. The use of DFT for the open system far from equilibrium state is

problematic. We discuss this problem in detail and present a possible way to introduce

non-equilibrium effects into DFT. For the second step, in which a Landauer-type formula

is used to calculate the current, the problem is that the Kohn-Sham orbitals are used to

calculate the current. Such orbitals are not genuine physical quantities, which causes

trouble in choosing basis sets and functionals.















CHAPTER 2
FIRST PRINCIPLES APPROACH COMBINING NGFT AND DFT

2.1 Model Hamiltonian and Landauer-type Formula

As we mentioned in the previous chapter, the system of interest can be divided into

three parts, the left lead, the device region and the right lead. To relate the current

flowing through the device region to the external bias is the goal of our research. The

whole system is infinitely large, but without periodic boundary conditions, which

provides a big challenge to the theoretical study. However, if one assumes that the two

leads are non-interacting, an exact formula which relates the current to local properties of

the device region can be derived [24]. The derivation is outlined in following paragraphs.

The starting point is the Hamiltonian

H = H + R + Hce, + HoC (2-1)

where HL HR and HCe,, denote the Hamiltonian for the left lead, the right lead, and the

device region respectively. The last term HCo, denotes the coupling between leads and

the device region. Since the two leads are non-interacting, we have

HLR = kak,a
k,aeL,R

en = nt({d, {d) (2-2)

Hfou, = IZ[Vk,cT ldn + h.c.
k,aEL,R
n










where indices k, a denote the channel of left and right leads (Bloch wave vectors and

spin) and n denotes the single-particle basis of the device region. The key issue is how to

treat the interaction of the device region.

The expression for the current can be derived by taking the time derivative of the

total number of electrons in the left lead


J = -e(N -)ie([H, NL (2-3)


where NL = 'c-ck, and JL is the current from the left lead into the device region.
k,cxL

Inserting the Hamiltonian into the expression for the current, we have


JL =ie Vkn (ca d,) Vk~ (dc
k,aGL
(2-4)
e knka Vka,nG
k,aGL
n

where the current is expressed in terms of two non-equilibrium 'lesser' Green's

functions.

Now, the problem becomes how to calculate the 'lesser' Green's functions. This

can be achieved by a perturbation procedure. At t = -o, the two leads and the device

region are well separated, which means that there is no the coupling term in the

Hamiltonian. The three parts maintains their own Fermi energies. Without loss of

generality, we assume that the Fermi energy of the right lead /Ru is higher than the Fermi

energy of left lead/L,. As we gradually turn on the coupling from t = -c tot = 0, the

system will achieve a steady state at t = 0 with a current J from left to right. Taking the

coupling term in the Hamiltonian to be the perturbation, the 'lesser' Green's functions

can be expanded as










Gn,ka ka,m[G nmka,ka nmk+G ,ka
n (2-5)
ka,,n = Vka,m ka,kaG n k+ gakaM n
m

where m runs over all basis states in the device region g denotes the Green's function

for non-interacting leads and the upper indices r, a denote retarded and advanced Green's

functions respectively. The details of the expansion can be found in Appendix A.

Introducing the equation (2-5) into equation (2-4) and doing a Fourier transform gives


J, = 2e d Re{ VknVk m k ()Gm (0) + gK~ ,ka ()Gn( (o)]} (2-6)
f2;r kaeL
n,m

where the summation runs over all channels of the left lead. Since the lead is non-

interacting, the Green's functions g g" can be calculated as


gka,ka (0) = 27if(Eka )5(O) ka)
1 (2-7)
gka-,ka ka i


where f(ska) is the Fermi-Dirac distribution function. Insertion of the equation (2-7)

into equation (2-6) gives the current as


J = ie Tr{L [fL (Gr G) + G]} (2-8)
f 2;r

where the line-width function FL is defined as

[F (k )]mn = 2r ( pa (c )V. (,k )Vl (k) (2-9)
acL

in which the term pa is the density of states in the lead. A similar formula can be derived

for J, the current from the right lead into the device region, so the total current is given

1
byJ (JL JR),whichis
2










= ie dTr[FL -FR ]G +[fLFL fRFR][Gr -G]} (2-10)
2 f 2;r

where the current is related to the local properties of the device region.

In equation (2-10), the Green's functions in the device region are solvable for only a

few models. The simplest one among them is the free- lectron model, which means that

the device region is also non-interacting. For the free-electron model, we have [15, 20,

25]

G' = Gr''G
Y< = i[F/f, +FRR (2-11)

and then the current can be written as

J= elde[f fR ]Tr{GTRGTr } (2-12)

which is a Landauer-type formula.

Since the interaction in both leads and the device region usually plays an important

role in the tunneling process at nano scales, equation (2-12), which is based on the free-

electron model, seems to be uninteresting, unless it is combined with a mean field theory.

DFT is one kind of mean field theory. It is based on the single-particle picture. Within

present-day approximations, it is able to take into account 'most' many-body effects.

Moreover, due to the success of DFT in the electronic structure calculation of molecules,

it is natural to combine DFT and the Landauer-type formula to study molecular

electronics.









2.2 The Combination of DFT and NGFT for An Open System

2.2.1 Self-Consistent Procedure for Ground State of Closed System

DFT has been known to be powerful in the electronic structure calculation for both

molecular and solid state systems [26-30]. The main ideas of DFT are outlined in the

following paragraphs.

The basis of DFT is the Hohenberg-Kohn theorem [23]. This theorem shows that

the ground state of a closed system is uniquely determined by the electron density of the

system, which means that average value of any physical quantity for the ground state can

be expressed as a functional of the electron density. As an example, the total energy can

be written as

E[p] = F[p] + p()Vext (i)d (2-13)

where the term Vex, represents the entire external potential (including the ion-electron

interaction) and the first term F[p] is a universal functional related to the electron-

electron interaction and the kinetic energy of electrons. The theorem also shows that the

true ground-state electron density minimizes the energy functional E[p]. The only thing

used in the proof of Hohenberg-Kohn Theorem is the minimum energy principle of

ground state.

On the basis of the Hohenberg-Kohn theorem, a self-consistent procedure based on

the single-particle picture can be constructed. Assume that there is a non-interacting

system which yields the same electron density as the interacting system of interest, then

the electron density can be written in terms of the single-particle functions

p(/) = ~ (F) 2 (2-14)
1









which are usually called Kohn-Sham orbitals. The kinetic energy of the non-interacting

system can be calculated in terms of the Kohn-Sham orbitals (Hartree atomic units)

V2
T/[pT]= -- ). (2-15)

Then the exchange-correlation energy can be defined as

Exc [p] = F[p] T,[p] EH [p] (2-16)

where the term EH [p] is the classical Coulomb energy. The exact form of Ex[p] is

unknown. However many kinds of exchange-correlation functionals, most based on the

local density approximation, have been proposed and successfully used. By minimizing

the energy functional, we get the Kohn-Sham equations

V2 + VH [p](r) + Ve, () + V [P](ir), (i)= g () (2-15)


where the first term VH is the well-known classical Coulomb potential and the term Vxe


is the exchange-correlation potential defined by [-- The effective Kohn-Sham
9p

potential is defined as

VeY [p]() = VH [p](r) + Ve, (r) + VX [p]() (2-16)

Now, a self-consistent procedure (Kohn-Sham procedure) can be completed as follows.

Given the electron density p, the effective Kohn-Sham potential Vff [p] can be calculated

according to equation (2-16). Knowing VJ. [p], we can get the Kohn-Sham orbitals V, by

solving equation (2-15) and then the new electron density can be calculated according to

equation (2-14). Repeat the procedure until the self-consistency is achieved. The

Hohenberg-Kohn theorem guarantees that the Kohn-Sham procedure yields the true









ground state electron density (the V -representability problem for the electron density is

not discussed in this thesis). Thus an interacting system can be solved by a self-consistent

procedure based on the single-particle picture.

2.2.2 Self-consistent Procedure for Open System

For an open system with the geometry shown in Fig. 1-1, there are two kinds of

eigenstates, scattering states which are extended in all space and bound states which are

localized in the device region. If the bias voltage is zero, DFT is still applicable due to the

fact that the system is indeed in a ground state (the temperature is set to zero). In practice,

since we cannot treat infinitely large system, certain approximations have to be made.

The major approximation is that, due to screening effects, the electron density and

effective Kohn-Sham potential in the two leads are not affected by the device region and

thus can be set to the bulk values. The bulk is also infinitely large, but with the help of

the Bloch theorem, we only need to do the DFT calculation in one unit cell. After fixing

the effective Kohn-Sham potential of the two leads, the remaining problem is to treat the

device region, which is finite, with two fixed boundaries. If the bias voltage is not zero,

the use of DFT becomes problematic. In Chapter 5, we will discuss this problem in detail,

but here, we just assume that the same approach can be generalized to the finite bias case.

The goal is to calculate the electron density self-consistently in the device region

with two fixed, open boundaries. The Green's functions technique described below is an

efficient way to do this.

The electron density in the device region is directly related to the 'lesser' Green's

function by


p = G< (c)d (2-17)
2n i









For a single-particle Hamiltonian, the 'lesser' Green's function can be calculated by the

equation (2-11), where the retarded and advanced Green's functions are related to the

effective potential of the device region and two open boundaries as


Gr(a) = 1 (2-18)
E C y r(a) r

In equation (2-18), the term HC,, denotes the effective single-particle Hamiltonian and

the two terms EL ,R are self-energies for the left and right lead respectively. Two self-

energy terms represents effects of open boundaries. We will see an example of

calculating these two terms in Chapter 3.

As a conclusion, the self-consistent procedure for an open system can be

summarized as follows. First, the two leads can be treated by a separate bulk calculation

which provides two open boundaries for the device region. This only needs to be done

once. Second, given the electron density p in the device region, the effective Kohn-

Sham potential can be calculated. Knowing the effective potential, the 'lesser' Green's

function can be computed and then a new electron density is given by equation (2-17).

After self-consistency is achieved, the Landauer-type formula (2-12) can be used to

calculate the current in the system.

The non-equilibrium self-consistent procedure mentioned above is implemented in

a Green's functions representation which employs a multi-band tight-binding model in

Chapter 3 and is implemented in a wave function representation which includes spin in

Chapter 4.














CHAPTER 3
COHERENT ELECTRON TRANSPORT THROUGH AN AZOBENZENE
MOLECULE: A LIGHT-DRIVEN MOLECULAR SWITCH

3.1 Introduction

Electronic devices soon will be constructed of molecules or have features of

molecular size. The results of recent experimental [1-5] and theoretical studies [6-20]

suggest a brilliant future for molecular electronics. Among various efforts and activities,

several schemes have been proposed to design and construct molecular switches that have

two or more distinct states with vastly different conductance. Switching between on and

off states can be performed by applying an external bias or by using a scanning tunneling

microscope tip to manipulate the system [31-35]. These methods are not ideal. The

former approach may interfere greatly with the function of a nanosize circuit, while the

latter one imposes severe limitations in device applications. In addition, these methods

are relatively slow. Alternative methods, such as those based on fast, light-driven

processes, are therefore highly desirable. The photosensitive azobenzene molecule

discussed here is one such candidate for an opto-electronic device.

The azobenzene molecule has two stable conformations [36-38], trans and cis. The

energy of the trans structure is 0.6 eV lower than the energy of the cis structure [39], and

they are separated by a barrier of approximately 1.6 eV (measured from trans to cis) [40].

The trans structure is 1.98 A longer than the cis. The molecule can switch reversibly from

one structure to another under photoexcitation, with the structural change occurring on an

electronically excited state [38, 41].









Previous work has shown that a nanoscale opto-mechanical device based on the

conformation change of the azobenzene polymer can be designed. Experimentally [41], it

is known that light with a wavelength of 365 nm transforms azobenzene units from the

trans to the cis structure. As a result, the polymer contracts, thus converting the energy of

electromagnetic radiation to mechanical work. A second pulse of light with a wavelength

of 420 nm reverses the cis to the trans configuration.

In this chapter, we report investigations of transport in the molecular wire, Au-S-

azobenzene-S-Au. It consists of a single azobenzene molecule, which has been

functionalized by replacing a hydrogen atom with a CH2S group to provide a good

contact with gold, and two semi-infinite gold leads. Note that the name of the modified

molecule is technically bis-(4-methanethiol-pheyl) diazene or BMTPD. Since one can

add more azobenzene units to elongate the polymer, we simply follow Refs. [38, 39] and

refer to the molecular system as 'azobenzene'. As shown in Fig. 3-1 and consistent with

previous chapters, the entire system is divided into three regions: left lead, switch, and

right lead. The switch includes two layers of gold from each lead. The structures of the

leads are chosen to be the same as in Ref. [42], which corresponds to a gold (111) crystal

lattice.

The electronic structure of the switch region is determined by treating it as a

cluster. Full quantum calculations, based on density functional theory (DFT), are used to

optimize the structure of a cluster consisting of the molecule, the sulfur atoms, and two

rigid layers of gold. The optimization is performed with two different choices of the

interlead distance, d. For the trans configuration, the calculated optimal d is 18.16 A and,

for the cis configuration, 16.23 A. In subsequent sections, we study the transport









properties for both equilibrium and non-equilibrium case. In both cases, the localized

basis set and the tight-binding model is employed.








k=365 nm X=420 nm







Figure 3-1. Physical models of an azobenzene switch in contact with two gold leads:
trans isomer (upper) and cis isomer (lower).

3.2 Tight-binding Model and Self-energies

As mentioned in Chapter 2, the key issue is to evaluate the self energies of two

semi-inifinite leads, 2L and XR Once these two terms are computed, we can construct the

retarded and advanced Green's functions G' and G" of the device region in the presence

of two leads. Then the lesser Green's function can be calculated according to equation (2-

11).

The self-energy term represents the effects of the semi-infinite lead on the device

region. Usually, a truncation of the Hamiltonian in real space is needed since we can not

treat an infinitely large system. So, the tight-binding model is a natural choice. Fig. 3-2

sketches the model, where indices i, j denote nodes of the device region and a, b are the

nearest node of the left and right lead respectively. Note that hopping terms h, h' are

operators [42].













h:
a i J b

Figure 3-2. Tight-binding model of an open system connected to two semi-infinite leads.
Indices i, j denote nodes of device region and a, b are the nearest node of left
and right lead. Only the nearest-neighbor interaction is considered.

Within the framework of tight binding, the Green's function G, can be written as

G, = G ,c" + GC"h g hG, + GCh'gbh 'G, (3-1)

where the two hopping terms are taken as perturbations and the Dyson equation is used.

The term Gc"e represents the Green's function of the device region without two leads

and the term g denotes the unperturbed Green's function of the leads. Then, self-

energies can be defined as

S= h gh (3-2)
R = h gbbh

where, EL, 2R are self-energies of the left and right lead respectively. Two hopping terms

are projections of the effective Kohn-Sham Hamiltonian to a localized basis of each node

(hopping terms are matrices also). The only problem remaining is two so-called surface

Green's functions ga and gbb

Since a perfect lead with periodic boundary conditions consists of two semi-infinite

leads and a hoping term within the framework of tight-binding as shown in Fig. 3-3, it is

possible to construct the Green's functions of semi-infinite leads on the basis of Green's

functions of a perfect lead by a perturbation procedure. The whole procedure is

described in the following.









h



m i i+1 j


Figure 3-3. A perfect lead that consists of two semi-infinite leads and a hopping term.
Indices m and i belong to the left semi-infinite lead and i+1, j belong to the
right semi-infinite lead.

We take the hopping term h to be the perturbation, then the node i is the surface

node of the left lead and the node i+1 is the surface node of the right lead. G is used to

denote the Green's function of the perfect lead and Go, the Green's function of two semi-

infinite leads. The Dyson's equation reads

G = G, +(i GVGj) = GOVGj (3-3)

where p, q run over the whole system. Since V only contains two terms Vii+ and Vi+,i .

P equals to i or i+1. Because Gii+l=0 (without perturbation, an electron can not

propagate from i to i+1), we have

G, = GyV,,+G,G+, = glhG,+, (3-4)

Similarly,

G,,, = G y G, = g+ ,zh G, ,. (3-5)

In order to solve g from equation (3-4) and (3-5), we have to know relations

between Gi and Gi+,j. This can be found by solving the Green's functions of the perfect

lead. For a perfect lead, the tight-binding Hamiltonian is

H = hoa,, + C haa,, + h a+, am, (3-6)
m m

where h,,h are N x N matrices, N being the number of orbitals within one unit cell of

the lead. The eigenvalue problem for Bloch states takes the form









[ h, -(h- E)ekd -(h E*)ekd (k, e) = 0, (3-7)

where k is the Bloch vector and d is the lattice constant. The energy term e contains an

infinitesimal imaginary part, E' = E+ ir with r>> 0. Equation (3-7) can be written as

Az -h h- z 2=, =0, (3-8)

where, M+ = E+ ho and h, = h E It is to be noted that not all eigenvalues A, of

equation (3-8) for a given energy represent the Bloch state. Because equation (3-8) is

quadratic in A and has a dimension of N, the number of eigenvalues for a given energy is

2N. It is shown in reference [22] that, when the energy is not real, half of those

eigenvalues are outside the unitary circle in the complex plane (we denote them as A> )

and another half are inside the unitary circle (we denote them as A< ). Using the

superscripts + to denote eigenvalues and eigenvectors for e+ and E respectively, we

can introduce two Nx N matrices, a and f,

a = UA(U)
U (3-9)
8+ = U+A* (U) ),

where U+ is a matrix, whose i th column is the eigenvector u,, and A is a diagonal

matrix, whose i th diagonal element is the eigenvalue A2 The matrices, a and f,

satisfy the matrix equation

Ma h (h)+(a' )2 =
(3-10)
MA/1 + (h) -h (3+ )2 0

The retarded Green's function is defined by the matrix equation (E -H)G+ = I. We

have

(i( -H)G j)=8 M+G, -h G,, -(h )+ G+, = 0. (for i < j). (3-11)









For a perfect lead, the solution of this equation is done via recursion relations of

Green's functions, G- = AG+, andGi, = AG G,. This gives the equation

MA -h -(h ) A2 = 0. (3-12)

Comparing with equation (3-10), we see that, for i < j, A = a Introducing

G+ = AG+, into equation (3-5), we have

g, = a (h) (3-13)

Similarly, fori > j, we have

g ,+, = p+ (h+ ) (3-14)

3.3 Conductance at Zero Bias

In this section, we calculate the conductance of the two optimized structures. In this

study, we focus on the linear-response regime. Within this limit, the entire system is in

equilibrium and there is a common Fermi energy across the whole system. Because of

charge screening by electrons in the metal, the two leads, as well as the switch, are charge

neutral. The conductance is calculated by the Landauer formula [44],


2e2
ar= T(E ) (3-9)
h


where T(EF) is the transmission function at the Fermi energy. The effective potentials

are calculated by the computational chemistry code NWCHEM [28]. In the calculation,

we use Crenbs effective core potential (ECP) Gaussian basis [28] for Au atoms, STO-3G

Gaussian basis [28] for all other atoms (C, S, N, H). The pbe0 exchange-correlation

functionals [26] are used. For the wire system depicted in Fig. 3-1, the system is, in

principle, infinite, and periodic boundary conditions are not appropriate. However, due to










screening, the Kohn-Sham potentials in the two semi-infinite leads are not affected by the

switch. As a result, the potentials can be computed in separate calculations of a one-

dimensional gold wire with periodic boundary conditions [20, 22, 42].




400-
U a Trans b b

> 200 a




SE,
08
S Cis
E \ Trans
S04-


00,
-024 -022 -020 -018
Energy (Hartree)

Figure 3-4. Density of states (DOS) projected onto the switch region (upper panel) and
transmission as a function of energy (lower panel). The solid lines represent
the trans configuration and dotted lines the cis. Intense peaks in the DOS are
truncated to improve readability.

With Green's function techniques, the effect of the two leads on the switch can be

absorbed into self-energy terms, which are calculated using the same method as

mentioned in previous section. The transmission function T(EF) is evaluated by the


Caroli formula [25]



T= Tr[FLGrFRGa] (3-10)



which is the kernel of the integral in equation (2-12). The density of states (DOS) in the

switch region can be calculated from the trace of the retarded Green's function, which










includes both scattering states and localized states. The DOS can then be projected onto

any desired subsystem. Note that the subsystem can be chosen as a single atom or a group

of atoms. The peaks in the DOS as a function of energy are often used as indications of

the molecular orbital and, with the combination of the transmission as a function of

energy, useful information can be extracted as to how any given molecular orbital

contributes to the conductance.





Azobenzene molecule Trans b b'
200 Ef Cis


S 0-
o Trans
> 200- Sulpher atoms
a
SCis

a a' Trans b'
200- Au layers
Cis

-024 -022 -020 -018
Energy (Hartree)

Figure 3-5. Local DOS of the trans and cis configurations decomposed into contributions
on the Au layers (lower panel), S atoms (middle panel), and the azobenzene
molecule (upper panel). The solid lines represent the trans configuration and
dotted lines cis.

The calculated transmission as a function of energy and the DOS for the trans and

the cis structures, with optimal interlead distance d, are shown in Fig. 3-4. Unlike the

DOS of an isolated molecule, Fig. 3-4 displays a series of peaks superimposed on a broad

background. These features originate from a combination of localized states from the

molecule and a nearly continuous energy band of the two leads. At some energies, the

DOS shows peaks, but the total transmission is very small. Notice, for example, peaks a









and b, for the trans isomer, and peaks a' and b' for the cis isomer. These peaks are

indications of localized states, which can be seen more clearly in the local density of

states.

As depicted in Fig. 3-5, the DOS of the switch region is decomposed into three

subsystems: the two Au layers, the two S atoms, and the azobenzene molecule. For the

trans isomer, peak a is mainly localized on the contacts, namely the Au layers and S

atoms, while peak b is localized mainly on the molecule. For the cis isomer, peak a' is

localized mainly in the Au layers, while peak b' has high density on the molecule. The

DOS on the S atoms is very small in the cis isomer as compared to the same peak in the

trans isomer. These comparisons indicate that the trans and cis configurations have quite

different contacts with the leads. Furthermore, Fig. 3-5 indicates clearly that the broad

background in the DOS derives primarily from the Au layers.

Figure 3-5 shows that the Fermi energy of the system lies between the highest

occupied molecule orbital (HOMO) and the lowest unoccupied molecule orbital (LUMO)

of the azobenzene molecule, slightly closer in energy to the HOMO. Compared to the

isolated molecule, the peaks in the local DOS projected on the molecule indicate that the

molecular orbitals are perturbed by the presence of the leads and contribute to scattering

states near the Fermi energy. Since we focus here on the zero-bias case, the conductance

is determined completely by transmission at the Fermi energy, which is dominated by the

tail of the electron distribution tunneling through the perturbed HOMO. Note that the

local DOS of both the Au layers and the S atoms peak near the trans HOMO energy, but

that the magnitudes of these peaks are minimal for the transmission function in Fig. 3-4.

For the trans isomer, tunneling through the perturbed HOMO leads to a very broad and











intense peak in the transmission function, but a sharp and smaller peak for the cis isomer

(Fig.3-4, lower panel). As a result, tunneling at the Fermi energy is more rapid through

the trans structure than through the cis structure, although the DOS of the cis isomer is

higher than that of the trans isomer. The calculated zero-bias conductance for the trans


structure is 0.21 x 104 4-1 which indicates a moderately good conductor. In contrast, the

conductance for the cis structure is nearly 2 orders of magnitude lower.





30
1: Left Au layer
2: Left S atom
25- 3: Left benzene ring
S4: N=N
20 Cis 5: Right benzene ring
S6: Right S atom
>, 7: Right Au layer
S 15-
10
) \ 1


5- -- -
0



Trans U \
0- 0

1 2 3 4 5 6 7
Atom groups


Figure 3-6. Local DOS at the Fermi energy projected on seven sites in the switch. From
1-7, or from left to right, Au lead, S atom, benzene ring, double-bonded N2,
another benzene ring, S atom and Au lead.

To understand the tunneling process at the Fermi energy, Fig. 3-6 depicts the local

DOS at the Fermi energy, Fig. 3-6 depicts the local DOS at the Fermi energy for both the

trans and the cis isomers. In this figure, we decompose the DOS into seven components:

the two left Au layers, the left S atom, the left benzene ring with a CH2 group, the

double-bonded N2, the right benzene ring with a CH2 group, the right S atom, and the two

right Au layers. As can be seen in the figure, the local DOS for the two molecular

conformations differs significantly. For the cis isomer, the local DOS is much higher on









the Au layers and the two benzene rings than on the S atoms and the N double bond. The

nearly zero DOS on the right S atom and the oscillation of the DOS along different sites

suggest that the scattering states are localized primarily on the benzene rings and the gold

leads. Reflection of the scattering wave occurs at sites of low DOS (on S and N). Both

localization and reflection contribute significantly to the resistance of the device.

Compared to the cis isomer, the local DOS for the trans isomer displays a smoothly

varying curve, suggesting that localization and reflection at the N double bond and the

two S atoms contribute much less to resistance than in the cis configuration.

In the azobenzene/gold system, the extended scattering states allow facile electron

tunneling through the molecule. This result can also be interpreted from the viewpoint of

molecular structure. In the cis configuration, due to the different orientations of the two

benzene rings, orbitals on different rings have different configurations. Consequently,

when electrons tunnel through the azo unit of the cis configuration, they are scattered

more than they are in the trans configuration. A more illuminating way to understand the

tunneling process is to decompose the conductance into conduction channels of two

leads, which will be done in future work.

The conductance of the cis isomer is significantly smaller than that of the trans.

Since the azobenzene molecule can easily change between trans and cis configurations

via excitation with light, we suggest that this system is a viable candidate as a light-

driven molecular switch. The switch will work at room temperature due to the high

thermal stability of both the trans and the cis configurations of the azobenzene molecule.

3.4 The Switch under a Finite Bias

When a finite bias is applied, the whole system is out of equilibrium. In this

section, we apply the non-equilibrium self-consistent procedure discussed in Chapter 2 to









the Au/S/azobenzene/S/Au junction and relate the current through the switch to the

external bias voltage. In the study, we assume that the geometry of the device region is

the same as that of zero-bias case. We neglect the effects of the so-called current-induced

force by making this assumption.

1 G
2; i





Vff = V[p]

Figure 3-7. Non-equilibrium self-consistent procedure. The non-equilibrium Green's
function is calculated according to equation (2-11).

The self-consistent procedure is depicted in Fig. 3-7. Since we are working within

the framework of DFT, the most important physical quantity is the electron density.

Given the electron density p, the effective potential Ve, can be computed as a functional

of the density. Once we know the effective potential, the non-equilibrium Green's

function G' can be calculated according to equation (2-11), where the retarded and

advanced Green's functions are calculated in the same way as in the previous section.

The details of the self-consistent procedure can be found in Appendix B.

One of the major accomplishments of the non-equilibrium first principles

calculation is that the electrostatic potential due to the external bias can be computed self-

consistently. Fig. 3-8 shows the difference of the electrostatic potential between bias one

volt and zero bias for the cis configuration. In the figure, the y axis is along the current

direction and the x-z plane is parallel to N double bond of cis configuration, cutting










through the midpoint of the Au unit cell. The drop of the electrostatic potential across the

switch area can be clearly seen.

Cis under 1 Volts





0 0.5 .
'0.0







0-1



S(Angstros 10


Figure 3-8. The difference of electronstatic potential between bias one volt and bias zero
for cis configuration. The x-z plane is parallel to the N2 bond, cutting through
the midpoint of Au unit cell.

The current through the system is calculated by equation (2-12), where the current

is expressed as the integral of the single-particle transmission probability over energy.

The effects of external bias voltage on the current through the switch enter in two ways.

First, as the bias voltage increases, the upper and lower limit of the integral changes so

that more channels enter the integral. Second, the bias voltage changes the electrostatic

potential, thus changes the transmission probability in turn. In Figure 3-9, we show the

transmission probability as a function of both energy and bias voltage, where the

transmission is expressed as


T(E, V) = Tr(FL (E, V)G' (E, V)F (E, V)G' (E, V)).


(3-13)









Transmission as a function of energy and voltage for cis

1.0


HOMO
c / LUMO







0.0Ie 7 0\ \


nergyV) -4 00 ,9


Figure 3-9. The transmission probability as a function of single-particle energy and the
external bias voltage for cis configuration.

There are two effects of the external bias on the transmission probability through a

particular molecular orbital. First, as the external bias changes, the effective Kohn-Sham

potential in the device region changes due to the change of electrostatic potential, thus the

'molecular orbital' changes. On the other hand, the external bias shifts the Fermi energies

of the two leads in opposite directions so that channels in both electrodes corresponding

to the same molecular orbital change. Both of these two effects are taken into account in

the first principles study.

I-V curves through the trans and cis configuration are shown in Fig. 3-10. The

differential conductance, defined as the derivative of the current with respect to the bias

voltage, is shown in Fig. 3-11. One of the important results is that at zero bias, the

conductance of trans configuration is two orders higher than the cis, which agrees with

the equilibrium calculation. The current through the trans configuration is much higher










than the cis for all bias voltages, which shows the high stability of the switch under a

finite bias.



25


20-


a15


S0- Trans

SCis
5-



0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Voltage (Volts)


Figure 3-10. Current as a function of bias voltage for both trans and cis configurations.
The red line represents the current of trans and the blue line represents the
current of cis.

An interesting feature is shown in Fig. 3-10. For the cis configuration, when the

bias increases from 0.7 Volts to 0.8 Volts, the current drops, which gives a negative

differential conductance as shown in Fig. 3-11. This can be explained by consideration of

Fig. 3-12, where the transmission as a function of energy at 0.64, 0.72 and 0.8 Volts are

shown.

As we mentioned before, as the bias voltage increases, more channels enter the

integral of equation (3-1), which usually increases the current. However, this is not

always the case. If the transmission through some channels which have major

contribution to the current is greatly suppressed, it may not be compensated by the

increase of channels. Fig. 3-10 clearly shows such an example, where as the bias voltage
































0.0 0.2 0.4 0.6 0.8 1.0
Volts


Figure 3-11. The differential conductance as a function of bias voltage for both trans and
cis configurations.


1.0


0.8-


o 0.6

E
S0,4
I--

0.2


0.0
-6.


5


-5.5 -5.0
Energy (eV)


-4.5 -4.0


Figure 3-12. The transmission as a function of energy at 0.64, 0.72 and 0.8 Volts for cis
configuration.


., 2 Volts

\; J Q0, 2 Volts


0 Volts



Ef






29


increases from 0.72 Volts to 0.8 Volts, the transmission through the HOMO state is

suppressed thereby leading to the negative differential conductance in Fig. 3-11.














CHAPTER 4
QUANTUM TUNNELING THROUGH MAGNETIC JUNCTIONS

4.1 Introduction

Quantum transport in open systems far from equilibrium is an important and

challenging problem [1-5] as mentioned in previous chapters. This problem has attracted

considerable attention in the context of semiconductor based devices for which the

relevant electronic structures can be modeled using the effective mass approximation.

More recently, considerable progress has been made with respect to application of

quantum transport theory to molecular electronics. In particular, the effect of finite bias

voltage has been studied with a number of first principles approaches [6-20] with various

degrees of success. The preceding chapter is an example. The problem of spin-dependent

tunneling in magnetic heterostructures, however, presents a significantly greater

challenge than the spin-independent case to the theory of non-equilibrium transport

because one must simultaneously deal with the complexities of non-equilibrium transport

and complex electronic structures along with magnetism. Spin-dependent tunneling

junctions are considered to be the most likely candidate for next-generation devices for

spintronics. Compared to the trilayer or multilayer giant magnetoresistance (GMR)

systems, magnetic tunneling junctions tend to have a relatively higher magnetoresistance

ratio [45-48], which is an important factor in device designs. The magnetoresistance ratio

is defined as (Gp-Ga)/Ga, where Gp is the conductance for the case that the alignments of

net magnetic moments of two leads are parallel and Ga is the conductance for the case

that the alignments of net magnetic moments of two leads are anti-parallel. One of the









significant differences between a GMR system and a spin-dependent tunneling junction is

that the former almost always operates within the linear response regime, while the latter

usually has at least a bias voltage of a fraction of one Volt across the distance of a few

nanometers, placing it far away from the equilibrium state. It has been observed

experimentally [49] that the bias voltage across a tunneling junction can greatly reduce

the tunneling magnetoresistance (TMR). Understanding the dependence of the TMR on

the voltage bias has become one of the critical issues in the development of spin-

dependent tunneling devices.

There have been several first-principles studies [50-54] of tunneling conductance of

trilayer junctions of the form FM/barrier/FM, where FM is a ferromagnetic material and

the barrier layer is either a semiconductor or an insulator. These studies were limited to

the linear-response regime. Researchers generally found exceptionally large TMR ratios

compared to experiments, and strong dependence of TMR on the interface resonance

states in the minority spin channel [51-52]. It is not clear whether the interface resonance

states remain relevant under large voltage biases, which will significantly modify the

interface electronic structure. Among the finite bias studies, Zhang and White [49]

constructed a phenomenological model to explain the voltage-dependence of TMR in

terms of the impurity scattering at the interface. They suggested that with a high quality

barrier, a different voltage-dependence of TMR may occur. Some groups [55-57] have

proposed different models suggesting that even without impurity scattering, the TMR

also decrease with the bias. One work attempted to explain the TMR reduction in terms

of magnon excitations at the interfaces [58]. However, these models are often based on

overly simplified scattering potentials, e. g., simple step barriers, and other assumptions









such as neglecting the k// (Bloch wave vector in two dimensional plan perpendicular to

the current) dependence of the transmission and reflection probabilities. Such

simplifications have been shown to lead to qualitatively incorrect conclusions in linear-

response models [59, 60], including incorrect dependence of the conductance and

magnetoresistance on the barrier thickness and height.

Today, first-principles calculation [6-20] of electronic structure and tunneling

conductance under a finite bias voltage has been successfully applied to molecular

electronics. As mentioned in Chapter 1, the first-principles approach can be built on

either non-equilibrium Green's functions formulation or scattering wave formulation and

the scattering wave has so far been implemented only with jellium electrodes, due to the

difficulties associated with computing and storing Bloch wave functions of realistic leads.

The layer Korringa-Kohn-Rostoker (KKR) approach [61], which was used

successfully in calculating linear-response transport properties of spin-dependent

tunneling junctions [51, 52], uses an extended basis set, spherical waves inside atomic

spheres and plane waves in the interstitial region. The scattering wave solution for

incident Bloch wave functions is provided in Ref. 51. This allowed straightforward

calculation of the two-terminal conductance within the linear-response regime. By

extending this approach to treat finite bias, we can overcome the shortcoming of the

scattering wave method and treat realistic leads.

The test system for our calculation is a Fe/FeO/MgO/Fe(100) tunneling junction.

This system has good lattice match between the Fe substrate and the MgO lattice. It has

been successfully grown experimentally with demonstrated epitaxial growth [62-65]. The

inclusion of a single atomic layer of FeO on one of the interfaces is important. An earlier









linear-response calculation [52] for the Fe/MgO/Fe(100) tunneling junction showed a

TMR ratio that is two orders of magnitude larger than experimental measurements [65].

In a subsequent calculation [66] this disagreement between theory and experiment was

shown to be mostly due to the presence of the FeO layer on the interface which was

observed experimentally [63, 64]. A finite voltage bias will change the interface

electronic structure, which in turn will impact the TMR ratio. The Fe/FeO/MgO/Fe(100)

tunneling junction would be a good system to examine whether the change in electronic

structure due to the bias voltage alone can explain the change in TMR.

Three effects of finite bias need to be considered in a first-principles theory. The

first effect arises from the energy dependence of the number of channels in the wire and,

in general, of the transmission probability of the conductor or the tunneling barrier. This

can be accounted for by replacing the linear response formula with an integral over the

energy. The second is the change in the electronic structure of the sample due to the

Coulomb field of the accumulated charge induced by the flow of the current. This is

usually small for good conductors, but can be significant in a tunneling device. The third

one is the effect of the current on the exchange-correlation potential in a density

functional theory approach. Most forms of this potential are developed for equilibrium

electron systems, and those incorporating a current are usually rather complex and

difficult to implement numerically. In this work we only consider the first two effects of a

finite bias, and neglect the effect of a current on the exchange-correlation potential.

In the following section, we will show that our approach is equivalent to the

Keldysh Green's function approach within the single-particle picture, but without the









need for the truncation of the matrix elements in the Hamiltonian and the overlap matrix

as required by other first principles methods using the Keldysh Green's function.

4.2 Non-equilibrium Green's Function Formalism in Real Space

The Keldysh formalism incorporates the open boundary conditions of a non-

equilibrium system into the components of the Green's functions. Because not all of the

Keldysh Green's functions are independent of each other, we can use the usual advanced

and retarded Green's functions along with one additional component of the Keldysh

Green's functions. Common practice is to use the 'lesser' Green's function G' which is

defined as

G' (r, t; r t) = i ( (r',t ) (r,t)) (4-1)

where the angle brackets indicate the ensemble average and V'+, are field creation and

destruction operators of electrons. There is a simple relation between the lesser Green's

function and the electron density p(f, t) = -iG (F, t; F, t).

Since we are only interested in the steady state solution, the time dependence is

only through At = t t'. So, the lesser Green's function can be Fourier transformed into

G' (F, F'; At) = G' (r, '; E)e'EAtdE
1 e (4-2)
G'(,';E) -G'(F, ';t)e-EtdAt

Note that writing the Green's function with a single energy argument implies that

the scattering in the tunneling region is elastic. The charge density is also independent of

time,


() = 2fG' (,,E)dE. (4-3)
Z.7U> '









The problem remains to evaluate G' (r,'; E). For non-interacting Hamiltonians,

the approach in the literature [20, 22] requires the decoupling of the two electrodes. This

effectively truncates the range the Hamiltonian at a finite distance, and may not be

appropriate for some systems. Here, we take a different approach using the scattering

wave representation. Our approach will not be easily generalized to interacting electron

systems, but is valid within the typical first principles methodology based on the density

functional theory. We rewrite the Fourier transformed Green's function in the form


G<(i,',E)= /' (r',t + At)V/(r,t))e-'EAtdAt. (4-4)


In the framework of mean-field theory, these two operators can be expanded as single-

electron operators

+ (ir, t)= Y0(r, t)C +
n (4-5)
y(r, = On(r, t)C,
n

where o, is the stationary state single particle wave function and two terms C,, C, are

creation and destruction operators for this stationary state.

Substituting equation (4-5) into equation (4-4), we have

G< (, ',E)= i 9(E- E) J Cn )CC n (r')o (r)
(4-6)
= ij 3(E- En )fno (r')n (r)
n

where the product C C, is the number of particles that occupy the state o and its

ensemble average is the distribution function f,. Equation (4-6) is a general expression

for the lesser Green's function in terms of the stationary wave functions and the









corresponding distribution function. Now, we need to adapt this expression for the

tunneling junction system.

An open system connected to two electrodes has a general geometry as depicted in

Fig 1-1. As before the whole system is divided into three regions, the left lead, the device

region and right lead. The wave functions and the distribution function must be

considered separately within each region and continuous at boundaries. There are two

types of stationary states: scattering states and bound states. The basic assumption of our

approach is that the left lead maintains an equilibrium distribution function f (E /L )

which is diagonalized by scattering function '+ incident from the left lead, and likewise

for the right lead distribution function f (E / ) and the scattering wave functions

from the right lead 0 Both f, and f are Fermi-Dirac distribution functions defined

with their respective chemical potentials /L and /R Next, using the subscript / to

denote all degenerate indices of '+ and m correspondingly foro equation (4-6) can be

written as (neglecting the bound states for the moment)

G' (r,r',E) = if, 18(E- E,) 0(r' ), (r)+if 1,8(E -Em)m(r')^m(r). (4-7)
1 m

We compare this result with that obtained in Chapter 2. Rewrite equation (2-11) as

G' = fG'FLG" +fGrFRGa. (4-8)

Both equations are valid within the single particle picture. However, equation (4-7) is

more general than the equation (4-8), as it does not require the complete decoupling of

the two electrodes, thus no truncation in the range of the matrix elements of the

Hamiltonian is needed. In particular, equation (4-7) applies to the entire space whereas

equation (4-8) can be applied only within the device region.









Substituting equation (4-7) into equation (4-3), we obtain the scattering part of the

electron density

p, (r) = f (E k) 2 + f (E R) 2. (4-9)
/ m

Including the contribution from the bound states, the lesser Green's function contains an

additional term i- 3(E- Eb )fb* (r) (r') where fb and b are the distribution
b

function and the single-particle functions of bound states. Accordingly, the electron

density also has an extra contribution from


p Zf (E)b 2 (4-10)
b

The determination of fb is not unique and is still being investigated by researchers.

4.3 Self-consistency and Tunneling Current for A TMR Junction

In this section we outline our approach for non-equilibrium self-consistent

calculations of electronic structure of a TMR junction.

Although the approach applies generally to any lead-device-lead ensemble, the

particular Fe/FeO/MgO/Fe tunneling junction in our calculation has a layered structure

with a two-dimensional periodicity along each layer. This approach allows a two-

dimensional Fourier transform that yields a quasi one-dimensional problem for each

reciprocal vector k//. The left and right leads are semi-infinite in the direction

perpendicular to the layers. They maintain different chemical potentials P, and /R .

Without loss of generality, we may assume/ u > /R The device region includes all of the

tunneling junction and a sufficient number of layers of materials from the left and right

leads, so that the potential and the charge density in the lead regions can be taken to be

the asymptotic values.









Within each electrode, the Bloch eigenstates can be divided into right-going states

(+ and left-going states ( The scattering states when there is a device region can be

categorized into those incident from the left lead traveling to the right, !+, and those

incident from the right lead traveling to the left, 0! Their wave functions can be

expressed in terms of the Bloch eigenstates within each electrode, and a scattering wave

function in the scattering region, in the form

p + tl ,l, Left lead

= (+, device region (4-11)

to (0' Right lead
m

and

( +-t ,t' ,,,(+', Right lead

S= ( device region (4-12)
t'lm( Left lead


where t, t' are the transmission coefficients and r, r' are the reflection coefficients. The

superscripts + indicate the direction of travel along the z direction, and the indices 1, m

label Bloch states in the left and right electrodes, respectively, and span over all possible

values of k// and the spin.

If one takes the jellium approximation in both leads, the Bloch wave functions

become plane waves and 0+ become the scattering states used in Ref. 10. With the help

of the layer KKR approach, we have gone beyond the jellium approximation and are able

to deal with the realistic leads.









At zero temperature, both distribution functions become the step functions

(E /L, ). Then from equations (4-9) and (4-10) we obtain


p(r) = dEp (r) + dEp (r) +pb(r) (4-13)


where


p =(r) YZ (E- E,) ,
2 (4-14)
p (r) = Z (E Em) (
m

These equations hold for all regions including the device region and both leads. We can

rewrite equation (4-13) as


p(r) = p (r) + dEp (r) (4-15)
PlR

where the equilibrium part of the electron density corresponding to '/R is defined as

PlR
p (r)= dE[p+(r) +p(r) + pb(r) (4-16)


Note that the superscript R indicates the chemical potential /RU used in the calculation of


Pe, yet the density pe is defined for all three regions.

Now, let us consider the electron density in the two leads separately. In the right

lead, the first term in equation (4-15) should give the asymptotic value of the electron

density far away from the device region at zero bias. The second term, the non-

equilibrium part, is due to the electrons transmitted from left. Neglecting the interference

terms between different Bloch states, which should be averaged out in an ensemble

average, we have PE within the right lead










p(r)= 3(E Em)tm (r) (4-17)
Im

The electron density in the left lead contains more terms than the right lead. Again,

neglecting the interference terms, we can write


p (r)= c(E- E,) p (r)2 + 3(E- E,)R, R1,, (r) (4-18)
1 11'

The normalization of the Bloch wave functions is such that the integral over the volume


of a unit cell d3r L (E E,) 9, + 5(E E,) p equals to the density of states


at energy E of the left lead, and likewise for the right lead.

The self-consistent procedure for the non-equilibrium calculation is very similar to

the usual equilibrium calculation for an interface system, except in this case, the charge

densities and the electrostatic potentials for all three regions need to be iterated

simultaneously, unlike the equilibrium case in which the semi-infinite (lead) parts are

calculated using the bulk self-consistent calculations before the interface region is

calculated. An additional consideration is given to the relative shifts in the electrostatic

potential between the three regions. Although the leads should screen any net charge

quickly so that the overall system should be charge neutral, the effect of dipole layers

which cause constant shifts in the potential cannot be screened. The size of the dipole

moment on each layer also depends sensitively on the convergence of the calculation. To

overcome this problem, separate constant shifts in the potential are added to the Kohn-

Sham potentials for the three regions. These shifts are adjusted at each LSDA iteration

until the net charge within each region reaches zero respectively. In general the shifts in

the potential differ from the bias voltage, as has been shown in earlier studies [21].









In order to see how the spin-dependent calculation is done, we write the electron

density in equation (4-14) in terms of spin explicitly


PE,(r)= Z8(E- Ek)
2 (4-19)
P.,(r) =3(E -Ek, O)
k,

where the term a is the spin index. The subscripts l,m now comprise both k// and a .

Remember that PE+ sum over ,1 e a and p, sum over p,,m e u. Then equation (4-

15) can be written as


p,(r)= p", + fdEpE (r) (4-20)
PlR

where the equilibrium part is


pq, (r) = JdE[p (r) + PE (r)]+ p, (r). (4-21)


The last term means only those bound states which have spin c are counted.

The equation (4-21) can be used to calculate the electron density for each spin over

all space. After obtaining the electron density for each spin, we use the common local

spin-density functional to calculate the effective potential.

Similarly, the current density for each spin is given by a Landauer-type formula


J = eJdE e ti m (4-22)
m,lJc Vl

where v,, v, are the group velocities of the Bloch states, and only the Bloch states for the

a spin are summed. The velocity factors are necessary because the transmission

coefficients are defined based on the continuity of the wave function, not the flux.









The equilibrium term in equation (4-20) can be evaluated using the standard

contour integration technique for Green's function methods. The non-equilibrium term on

the other hand, is more difficult to evaluate numerically. The integration over the energy

for the non-equilibrium term has to be computed along the real axis, which is numerically

noisy due to the singularities and resonances arising from the three-dimensional Bloch

states in the leads. This problem is not unique to the scattering wave function approach

employed here. Even if one were to compute the lesser Green's function directly, the

integration still would need to be done along the real axis because the lesser Green's

function is not analytic in both upper and lower complex energy planes. Thus a complex

contour cannot be used in either approach to reduce the numerical noise of the non-

equilibrium term. The numerics of the real axis integration can be improved greatly using

the technique by Brandbyge et al [33]. We note that equation (4-20) can be equivalently

written as

I/'
', (r)= p dEp, (r) (4-23)
PlR

where

/PL
P~(r)= dE [p,(r)+p p(r)]+,p (r). (4-24)


In principle, the electron density evaluated from either equation (4-20) or equation

(4-24) should be identical. But because these equations use two different energy contours,

the numerical errors are different. It is shown that the numerical error is minimized by

combining the two electron densities in the following manner [67]:

p'W = CP + (1- c) p' (4-25)


where









L 2
c neq,T (4-26)
SL R
[neq, +[pneq,a]

with

,L = f dEp-
Pneq,T d ,(-
PR (4-27)

= f dEpE,
PR

4.4 Application to the Fe/FeO/MgO/Fe Tunneling Junction

The Fe/FeO/MgO/Fe junction can be grown by depositing MgO epitaxially onto Fe

whiskers and then depositing another Fe electrode epitaxially onto the MgO layer.

Experimentally it was shown that a single atomic layer of FeO forms when MgO is

deposited onto Fe [62-64]. For the MgO/Fe interface, we use the same structure as in Ref

52. The lattice constant of the Fe layer is fixed at 2.866 A. In order to match [100] layers

of two materials, the MgO lattice constant is taken to be a factor of /2 larger than that of

Fe. The distance between Fe and O layers is taken to be 2.16 A, while the distance

between Fe and O atoms in the FeO layer is taken to be 2.1 A. Because we are unable to

incorporate the considerations of chemical or structural disorder into the Keldysh Green's

function formalism, a necessary step to account for the under occupancy of the oxygen

site in the FeO layer, we assumed 100% oxygen in our calculation.

Numerical convergence in terms of the number of energy points between two Fermi

energies and the number of k// points in the two-dimensional reciprocal space needs to be

considered carefully because the integration is along the real energy axis, and can be

prone to be numerical errors, as just discussed. For the self-consistent part of the

calculation, we use one energy point for biases below 0.01 Hartree and two energy points









for biases between 0.01 and 0.018 Hartree, the largest bias in our calculation. We find

increasing the number of energy points further has little effect on the self-consistent

solution. For tunneling conductance calculation, we begin with three energy points in a

uniform mesh for a bias of 0.001 Hartree. Then for each increment of 0.001 Hartree in

the bias voltage, two additional energy points are added to the mesh. Our calculation at

the bias of 0.001 Hartree showed that, when the number of k// points is increased from

2800 to 8262, the change in the conductance of parallel alignment of the magnetic

moments in the electrodes is less than 1% and the change for antiparallel alignment is

less than 3%. Therefore, 8262 k// points are used in the integration for all of the

conductance calculations. In the following discussion, the majority and minority spin

channel always refer to the left Fe lead.


0.3
Majority +
Minority
0.2 0
QMg

> 0.1
(1D
a,
0
S 0 Fe FeO Fe
c

-0.1 -MgO +


-0.2


-0.3


Figure 4-1. Shifts in the density of states in each atomic layer at a bias voltage of 0.54 V.

It is interesting to examine the change in the electronic structure due to the applied

voltage bias. The main effect of the voltage bias here is the accumulation of the charge on









the interface layers next to the barrier. We find that the change in the net charge on the

interface layers is small, and this change does not cause significant changes in the profile

of the electronic density of states at each layer. We see the nearly constant shifts in the

energy in the barrier that should be consistent with the shift in the electrostatic potential

at each layer. The shifts in the DOS between E=-0.5 and E=0.5 Hartree for each layer for

the bias voltage of 0.54 V is shown in Fig. 4-1 for parallel alignment of the moments. The

DOS for the iron layers in the leads is shifted by roughly +1 / 2 V on either side of the

barrier, and the shifts for the barrier layers correlate monotonically with the layer

positions. The shifts for both spin channels are about the same for each layer. Two

interesting features occur near the interfaces. On both interfaces, the first insulator layers

(FeO or MgO) have the same shifts as the electrodes. This is somewhat surprising, noting

that such alignment does not conform to the extrapolation of the shifts of other MgO

layer. The nearly exact alignment of the DOS of the interface insulator layers with the

electrodes does not agree with the model given by Ref 34 in which an oscillatory

behavior is predicted for the electrostatic potential near the interfaces. On the other hand,

we find that the shifts in the DOS scale essentially linearly with the bias voltage, as one

would expect. For negative biases, the shifts are the same in magnitude with the opposite

signs.

From the linear scaling of the DOS shifts with the bias voltage one would expect

that the capacitive charge should also scale linearly with the voltage. Surprisingly, this is

not the case. In Fig. 4-2 we show the capacitance dQ / dV as a function of the voltage.

There are three different charges plotted in this figure, none of which strictly conforms to

the usual definition of capacitive charge, due to the difficulty of separating the capacitive











0.01 Left electrode

0 0.008 --
O

o 0.006 Right electrode ,
>---x- x-,"
S0.004 -
S,+ Net charge
C 0.002

0-

-0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5 0.6
Bias Voltage (volts)

Figure 4-2. Derivative of the capacitive charge with respect to the bias voltage. The
charge on the right electrode includes that on the oxygen atom in the interface
MgO layer. 'Net charge' includes that on both the electrode and the dielectric
charges on the MgO layers on the same side of the midpoint of the barrier.

charge in the electrodes and the dielectric charge in MgO. The capacitive charge must be

antisymmetric, i.e., with Q on opposite sides of the barrier. This antisymmetry is not

possible with any partition of the atoms in our calculation, unless we include the charges

on all iron and MgO layers on the same side of the midpoint of the barrier. The curve

labeled 'net charge' is this sum per two-dimensional (2D) unit cell. This sum is the same

(with opposite signs) for both sides of the barrier. However, the sum includes both the

capacitive and the dielectric charges. To separate out the capacitive charge at least

approximately, for the left electrode we sum the charge for all iron layers but excluded

the FeO layer. For the right electrode we sum the charge for all iron layers plus the

charge of the oxygen atom on the interfacial MgO layer. The reason for including the

oxygen atom is that it represents more than a quarter of the net change in the charge on









the right side, consistent with our observation that this oxygen atom seems to be 'short

circuited' with the iron electrode. The two curves in Fig. 4-2 labeled 'left electrode' and

'right electrode' represent this approximation of capacitance. For comparison a model of

capacitance per unit cell for two plates separated by a vacuum gap of 21 A (which is the

distance between two iron electrodes), is about 0.0044 electrons/V. Thus our self-

consistent calculation yields an effective dielectric constant for the MgO film (coated

with an atomic layer of FeO on one side) of around 2. Since in our calculation no lattice

relaxation is included, this number should be compared to the optical dielectric constant

measured experimently. The value derived from the index of refraction measurement [35]

(n=1.7) for bulk MgO is s = n2 =2.9.




0.04 .

0.035 Right

0.03 -

o 0.025

0.02 Left

E 0.015 -

0.01
Bulk electrode (left)
0.005 --- -


-0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5
Bias Voltage (volts)

Figure 4-3. Derivative of the moment with respect to the bias voltage. The labels 'left'
and 'right' indicate that the layers to the left (right) of the midpoint of the
barrier are summed. The contribution to the sum from the nonbulk layers of
the iron electrodes is taken to be the difference between the moment of that
layer and the bulk moment in the electrode on the same side.









Because of the neutrality of the total system, the total net charge integrated over the

2D unit cells of all the layers on one side of the midpoint of the barrier should be the

same with the opposite signs from both sides. This charge is shown as 'net charge' in Fig.

4-2. Evidently, it is not symmetric with respect to the bias voltage. For positive biases, it

drops quickly as the voltage increases. At about 0.5 V, it becomes negative. This means

that at this voltage, any increase in the charge in the electrode due to the increasing bias is

completely matched and canceled by the opposite dielectric charge from the MgO layer.

An unexpected result is the change in the magnetic moment at the interfaces for

parallel alignment of the moments. For a symmetric junction such as Fe/MgO/Fe, if the

net moment increases with bias on one interface, it must decrease on the other interface

due to symmetry. For the asymmetric junction ofFe/FeO/MgO/Fe, we find that the

moments on both interfaces increase with positive bias and decrease with negative bias.

Thus the derivative of the net moment with respect to the voltage is always positive, as

shown in Fig. 4-3. Unlike the charge, there is no indication of saturation of the moment

changes at the interface as a function of the bias. The asymmetric changes in the

moments mean that the majority spin electrons react differently to the bias voltage than

the minority spin electrons. Since the rate of the change in the moment is almost an order

of magnitude larger than the rate of change in the charge, charge accumulation in the two

spin channels must be opposite in sign.

The curve labeled 'bulk electrode' in Fig. 4-3 is the change in the moment in the

bulk for the left electrode. The changes in the moment in the bulk regions on the two

electrodes are the same in magnitude but opposite in sign. The change in the interface

magnetic moment in response to the bias voltage is detectable experimentally if one can









keep the right electrode sufficiently thin. The right electrode (which does not have an

FeO layer at the interface) is the last deposited iron layer and its moment changes in the

opposite direction as the interfacial moment as a function of the bias voltage.

At zero bias, the conductance is given by


G, =e2 2V m 2 (4-28)
Imeo V

It is important that our calculation from the finite bias approach extrapolates to the linear

response results obtained in earlier works [51, 52, 66]. The k//-resolved conductance at

zero bias, shown in Fig. 4-4, reproduces all of the features seen in earlier works. For the

parallel case, the majority spin channel is dominated by the contribution from a region

around k//=0. The conductance due to the minority spin channel is about two orders of

magnitude smaller and is not shown in the figure. For the antiparallel case, our

calculations show that the predominant contribution to the minority spin current comes

from tunneling through the interface resonant states which correspond to sharp peaks of

the transmission.

The calculation gives a negative TMR of -28% at zero bias. We note that an earlier

work obtained a small positive TMR ratio for the same junction [66], but our calculation

yields a small negative TMR ratio. This difference appears to be due to the extreme

sensitivity of the minority spin channel conductance to the convergence of the integration

over k//. The earlier calculation used 2800 k/ points while we used significantly more k/

points (8262). Nonetheless, the general trend studied, that of TMR as a function of

oxygen concentration in Ref. 31, and the TMR as a function of bias voltage studied here,

should remain valid as long as the number of k// points used is consistent within each

study.

















Transmission (10-5)


.0 0.2

.5 0.15

S0.05


00'


Transmission (10-4)


4.5 -
4.0 I
3.5 I
3.0 I
2.5 -
2.0 J
1.5 -
1.0 -


'II


I I.


0.5 "


.. "--'-^.
.: .. ..Z -


0.2

0 15

0 05 Ky


Figure 4-4. Zero bias transmission probability as a function of k//, (a) Majority spin
channel for parallel alignment of the moments; (b) The dominant spin channel
for antiparallel alignment of the moment.




























-40 L
-0.5


-0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5
Bias Voltage (Volts)


Figure 4-5. Current density as a function of bias voltage for majority and minority spin
channels for the case of parallel alignment of the moments between
electrodes.


6

4 .'Cu

2

0

-2

-4

-6

-8
-.--
-10
-0.5 -0.4


-0.3 -0.2 -0.1 0 0.1 0.2
Bias Voltage (Volts)


0.3 0.4 0.5


Figure 4-6. Current density as a function of bias voltage for majority and minority spin
channels for the case of antiparallel alignment of the moments between
electrodes.









The tunneling current is calculated through equation (4-22). The effect of the bias

voltage, through the change of the electro-chemical potentials in the two electrodes,

enters this equation in two ways. First, the bias voltage can change the transmission

probability for each conduction channel, i.e., for each k// point at each energy. Second,

when the bias increases, the number of conduction channels in the leads that contribute to

tunneling also increases. An interesting question is how much of these effects is captured

by rigidly shifting the electrostatic potential for each atomic layer and neglecting the

effects of charge rearrangement, as was commonly done in many early non-self-

consistent calculations, and, conversely, how important is charge self-consistency in

accurate determination of the TMR. For this purpose, we compare the fully self-

consistent calculations of the tunneling current and the TMR with a calculation

constructed from a self-consistent zero bias potential, in which the electrostatic potentials

in the left and right electrodes are rigidly shifted by + 1/ 2 V where V is the bias voltage,

and the potentials of the barrier layers are shifted linearly with the bias according to their

positions.

Because of the asymmetry of the tunneling junction, the positive and negative

biases give different tunneling currents. Figure 4-5 shows the I-V curves for the majority

and minority spin channels for parallel alignment of the moments. As expected, we see a

large linear regime in this figure. For low biases, the slope of the curve gives the linear

response conductance a value ofl. 15 x 107 /(m 2). This number agrees very well with

the result obtained from equation (4-28) which is1.17 x 107 /(Qm2) Note also that the

non-self -consistent calculation yields the same results within the linear region (for bias










voltage up to about 0.1 V) but deviates significantly at higher biases, reaching about

twice the self-consistent results at 0.5 V. The current of minority spin channels actually


3 .5 .
'0.0_Volts'
'0.054 Volts'--------
'0.108 Volts' ..........

S2.5
0

c 2.0
0

E 1.5 -

-1.0

0.5


-0.2 -0.15 -0.1 -0.05 0 0.05 0.1 0.15 0.2
Kx (Ky)

Figure 4-7. Transmission probability at energy E = /R as a function of k/ along the line
of kx = ky for the majority spin channel for parallel alignment of the moments
at different bias voltages.

shows an oscillatory behavior. Since it is negligibly small compared to the majority spin

channels, we cannot see this in Fig. 4-5.

Figure 4-6 shows the total tunneling current for antiparallel alignment of the

moments. The curve labeled 'SCF' is in fact calculated by simply flipping the moment

alignments of half the layers of the corresponding parallel SCF calculations, thus this is

not a truly SCF calculation for the antiparallel alignment. However, one fully SCF

calculation for the antiparallel alignment was performed at V=0.5 V and the difference is

negligible both in terms of charge transfer and tunneling current. The curve labeled 'non-

SCF' is calculated with the rigid shift of a self-consistent zero-bias electrostatic potential.

For most of the bias voltages the tunneling currents agree between the 'SCF' and 'non-









SCF' calculations, except for large negative biases. The tunneling current exhibits an

oscillatory behavior as a function of positive biases, probably due to the matching and

mismatching of interface resonance states on both sides of the barrier as the bias voltage

moves these states in energy and k/. We will see this later from the analysis of integrated

transmission.

Let us consider a positive bias as an example to study the effects of bias on the

tunneling process. By positive bias we mean that the electron flux is from the left (FeO

side) to the right, or that the current is from the right to the left. We calculated the

transmission probability for the spin channel with the largest contribution to the current at

the energy E = /u and along a line of kx = k, in the 2D reciprocal space, under different

positive bias voltages. For the case of parallel alignment of the moments in both leads,

Fig. 4-7 shows the results of the majority spin channel at bias voltages of 0.0, 0.054, and

0.108 V. All curves exhibit very similar structures. It is clear that the bias does not affect

the transmission very much. This suggests that the linear regime of I-V curve for the

majority spin channel may extend up to 0.108 V. For the antiparallel case, Fig 4-8 shows

the results for the same range of bias voltages. Although all curves has the similar overall

structure (the transmission is low at k/=0 and have one peak on each side of k/=0), the

position, height and width of these peaks change dramatically as functions of bias. From

0.0 to 0.082 V, the peaks are enhanced, and from 0.082 to 0.108 V, the peaks are

reduced. As discussed in an earlier linear response study [20], the position and the

strength of these peaks are determined by several factors. Among them are the amplitude

of the wave function at the interfaces on both sides of the barrier, and the decay rate

across the barrier as determined by the complex band structure in the barrier. When









interface resonance states exist, the amplitudes of the wave function can be enhanced by

several orders of magnitude at the interface through coupling to those interface resonance

states. This coupling gives rise to the sharp peaks in the transmission probability at

certain k// points. Applying a bias voltage or changing it will change the energy and k/

positions of these interface resonance states, in particular the relative positions of

interface resonance states on both sides of the barrier can be moved. This in turn can

cause large changes in the transmission probability at corresponding k// values.


3.0 ,
'0.0 Volts'
'0.054 Volts' -
2.5 -'0.082 Volts'
'0.108_Volts' i
o 2.0 \


1.5 -
E i
c 1.0


C0
0.5 ;



-0.2 -0.15 -0.1 -0.05 0 0.05 0.1 0.15 0.2
Kx (Ky)

Figure 4-8. Transmission probability at energy E = /R as a function of k/ along the line
of kx = ky for the minority spin channel for antiparallel alignment of the
moments at different bias voltages.

As mentioned at the beginning of this section, when the bias voltage increase, the

number of conduction channels in the leads that contribute to tunneling also increase. In

order to examine this effect, we define the integrated transmission as

1 "
T(k,) t(k, )dE. (4-29)
S R /'R










In Fig. 4-9 and 4-10, we show this integrated transmission probability as a function

of k// along a line of kx=ky in the 2D reciprocal space. The patterns shown in these two

figures closely resemble those in Fig 4-7 and 4-8. However, the integration over the

energy has smeared out to some degree the effects due to the interface resonance states,

so that the change in transmission probability as function of bias is less dramatic. Based

on these figures, we can expect that the majority spin tunneling current for the parallel

case should scale approximately linearly with the bias, and the minority spin tunneling

current and the antiparallel tunneling current should increase more slowly than linear

with the bias. This behavior should lead to an increase of the TMR ratio as a function of

the bias voltage.


3 .5 .
'0.0_Volts'
'0.054 Volts' -----
U 3.0 -
3.0 '0.108 Volts' ..........

c 2.5
C,/)
.(n 2.0
C
S 1.5
-0
S 1.0 -
0)
0.5 -


-0.2 -0.15 -0.1 -0.05 0 0.05 0.1 0.15 0.2
Kx (Ky)

Figure 4-9. Integrated transmission probability as a function of k// along the line of
kx = k, for the majority spin channel for parallel alignment of the moments at
different bias voltages.

Although the negative bias affects the system differently, the general feature is the

same: the transmission probability of dominant channel of antiparalle case is much more









sensitive to bias than that of parallel case. At small bias, the antiparallel tunneling current

also increases more slowly than the parallel tunneling current as a function of bias

voltage.

Figures (4-12) and (4-13) show the integrated transmission as a function of k// for

all k// points at a higher bias of 0.378 V. Compared to Fig 4-4, the broad peak in the

parallel case becomes broader, a result of the integration over the contribution from all

channels at different energies. For the antiparallel case, the structure of the peaks due to

the interface resonance states becomes more complicated than at zero bias.

The TMR at finite bias is defined as (J J) / Jp, where Jp, J, are total currents

for the parallel and antiparallel case respectively and its linear-response limit is (Gp-

Ga)/Ga as given in the introduction. The voltage dependence of the TMR is shown in Fig.

4-13. Previous calculations that used simplified models [49, 50, 58] yielded TMR ratios

that are large and positive at zero bias, then decrease rapidly with bias. Our calculations

yield a very low TMR ratio at low biases, even negative at 0 and + 0.027 V. Then, for

positive biases, the TMR ratio increases rapidly as the bias voltage is changed from 0.054

to 0.108 V. This increase is caused by the decrease of the antiparallel current. When the

bias voltage is larger than 0.108 V, the TMR ratio starts to oscillate with the bias, but

overall still rises with the bias, albeit more slowly. For negative biases, the TMR

increases from -0.027 to -0.432 V and decreases at -0.486 V, which is caused by slow

increase of parallel tunneling current.

The comparison between the SCF calculation and the non-SCF calculation for TMR

is similar to that of the tunneling current for parallel spin alignment. This indicates that

the charge self-consistency is more important for the majority spin channel than the

























0 1 1---I 1-,
-0.2 -0.15 -0.1 -0.05 0 0.05
Kx (Ky)

(a)


0 -
-0.2 -0.15 -0.1 -0.05


0.1 0.15 0.2


0 0.05 0.1 0.15 0.2


Kx (Ky)

(b)


Figure 4-10. Integrated transmission probability as a function of k// along the line of
kx = ky for the majority spin channel for parallel alignment of the moments at
different bias voltages.
















Transmission (10-5)


0.5




Kx 0.1 0.


.0I


S0.2
0.15
~ 0.1
' 005
Ky


Figure 4-11. Integrated transmission probability as a function of k// for the majority spin
channel for parallel alignment of the moments at bias 0.378 V.






Transmission (10-5)


1.4 -


1.2 -
1.0 I
0.8 -
I.1
0.6 .
0.4 0.

0.2 15
00.
i *C ..i :- -z .o- Ky

Kx Ci^i i I 0 2 1


Figure 4-12. Integrated transmission probability as a function of k// for the minority spin
channel for antiparallel alignment of the moments at bias 0.378 V.









minority spin channel, presumably due to the sensitivity of the majority spin tunneling

current to the chemical bond configuration at the interfaces [66].

The tunneling through the interface resonance states leads to the sharpest peaks of

the transmission. Due to the presence of roughness and impurities at the interfaces, the

interface resonance states may not be present or sharply located in experimental

measurements. In Fig. 4-13 we also indicated the TMR ratio at zero bias if we removed

all of the contributions from the interface resonance states. Although the TMR ratio

increases significantly, this increase does not change the overall trend of increasing TMR

ratio with bias. Thus we conclude that at least for Fe(100) electrodes, where a single

atomic layer of FeO is formed at one of the interfaces, the TMT ratio is close to zero at

small biases, and should increase with the bias voltage. We believe that this conclusion

should also apply to barrier materials other than MgO, especially when the barrier layer is

asymmetric as a result of the order of deposition, so that the TMR ratio is low at small

biases. This conclusion supports the suggestion by Zhang and White [9], that for high

quality barrier, the voltage-dependence of TMR may be different than the observation in

earlier experiments that the TMR ratio decreases with bias.

For the system we studied, this unusual behavior is based on following facts. First,

for the parallel case, the transmission of dominant conduction channels (peak for majority

spin channel) is not sensitive to the bias. This gives a large linear regime as a function of

bias. For the antiparallel case, the dominant conduction channels are sensitive to the

interface electronic structure, in particular, interface resonance states, which in turn are

sensitive to the bias voltage. Often the maximum of transmission occurs when a tunneling

state couples efficiently to both leads. This may be the case when interface resonance











'TMR:SCF' -B- i
3.5 'TMR:Non-SCF' ----------

3

2.5


1.5 \

1 .

0.5

0

-0.5
-0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5
Bias Voltage (Volts)

Figure 4-13. Magnetoresistance ratio as a function of bias voltage. Point A denotes the
MR ratio at zero bias by removing all contribution from interface resonance
states.

states are present on both interfaces. When the bias increases, it is likely that the once

efficient coupling to both leads is greatly reduced due to the change in the interface

resonance states. This would lead to a great reduction in the tunneling current for the

minority spin channel and for the antiparallel case. Thus, it is possible for TMR to

increase with the bias. Although one may expect that for amorphous barriers the effects

of interface resonance states are much reduced, the impurity states at the interface may

play the same role as the interface resonance states in the system we discussed in this

work, so that the argument presented here may still be qualitatively true. Thus, it may be

possible to observe the similar behavior of TMR even for amorphous barriers.

4.5 Conclusion

In conclusion, we have completed the first spin-polarized calculation using the

non-equilibrium first-principle method for spin-dependent tunneling junction under a









finite bias. Application of this method to the layer KKR treatment of the Fe/FeO/MgO/Fe

tunneling junction leads to surprising results that are not anticipated by simple models but

are reasonable in view of the change in the electronic structure at the interface in

response to the bias voltage. Our calculations indicate a negative TMR ratio at very low

biases and a rapid increase of the TMR ratio at finite biases. We believe that this behavior

is not limited to the Fe/FeO/MgO/Fe tunneling junctions and may be observed

experimentally in other systems. This result has import practical implications since

increasing the TMR ration at finite bias voltages is one of the important goals of

spintronics research. Our calculations also show significant deviations from non-SCF

calculations using rigidly shifted electrostatic potentials at bias voltages about 0.1 V, as

well as the nonlinear capacitive charging effect and the asymmetric magnetic moment

change as a function of bias voltage. All these results highlight the important role of a

fully self-consistent non-equilibrium calculation.














CHAPTER 5
PROBLEMS IN THE CURRENT NON-EQUILIBRIUM APPROACH FOR OPEN
SYSTEMS BASED ON DFT

5.1 Introduction

As discussed in previous chapters, for the open system under a finite bias, the non-

equilibrium first-principles approach has been proved to be successful. The approach is

based on density functional theory, and the typical procedure can be summarized in two

steps as we mentioned in Chapter 2: the first is the self-consistent procedure based on

DFT to determine the effective single-electron potential. Here, of course, the density

functional used is valid for ground states only, but is assumed to be valid for non-

equilibrium systems because no alternative approaches exist. The second step is to use

the Landauer conductance formula to calculate the current. The Landauer formula relates

the current to the integral of the single-particle transmission function over energy. Such

approaches have been used in many systems with varying degrees of success: the shape

of the I-V curve can be compared to the experiment and the structure of I-V curve can be

explained by molecular orbitals. All these achievements are important for the

understanding and future design of molecular electronics. However, large discrepancies

are often seen between theoretical calculations and actual measurements of I-V curves.

Among possible sources of error in the theory, the use of the particular type of the density

functional in a calculation is often a cause of concern. Much progress has been made in

finding an energy functional form that works well for both the metallic leads and the









molecular junction. Unfortunately, the problem is more fundamental than just the

functional.

The basis of DFT is the first Hohenberg-Kohn theorem [23], which states that for

the ground state, the external potential Vex,(r) is determined, within an additive constant

Vo, by the electron density p(r). Since all ground state properties are determined by the

Hamiltonian, which is uniquely defined once the external potential is determined, the

one-to-one correspondence between Ve, (r) and p(r) guarantees that all ground state

properties are determined by the electron density p(r) In other words, p(r) is the basic

variable of the system at ground state and all physical quantities can be expressed as

functionals of p(r) The constant potential Vy is trivial for ground state, since if Vy

changes, the Fermi energy and the electronic structure of the system simply shifts

accordingly. The theorem can be easily proved through the fact that the ground state has

the lowest energy. However, the same proof does not apply to an open system under a

bias voltage, since the system is in a non-equilibrium state, and there is no minimum

energy principle for such a state. Before we use the DFT in such a system, we have to

question whether the electron density p(r) remains a fundamental variable that uniquely

describes the system.

In the following section, we show that the electron density alone is not sufficient to

describe a non-equilibrium open system. Then we propose a modification of the

Hohenberg-Kohn Theorem for an open system at finite bias. We follow that with a

numerical study that gives a functional form of the exchange energy for non-equilibrium

open systems.









5.2 Hohenberg-Kohn Theorem for Open System under Finite Bias

To make our discussion more concrete, we choose a model open system with the

general structure shown in Fig. 1-1. The system is divided into three regions: the left lead,

the scattering region (the device region), and the right lead. The scattering region consists

of the left contact, the barrier and the right contact. The two contacts are parts of leads

that connect to the middle barrier. The leads are connected with two reservoirs at infinity

(not shown in Fig. 1-1) and the chemical potentials of the two reservoirs are constants. A

discussion of the contact between leads and reservoirs can be found in reference 68. The

two leads are often to be assumed to be internally at thermodynamic equilibrium [20], but

not in equilibrium with the contacts or the scattering region through the contacts.

The electronic structure of the two leads far away from the scattering region are

assumed to be identical to their bulk values [20], and two separate calculations based on

first-principles method are often performed to obtain properties of the leads. Furthermore,

the temperature of both leads is assumed to be zero.

The bias voltage is defined as the difference between Fermi energies '/, and '/R

maintained by the two leads connected to two reservoirs, respectively. The purpose of the

study is to calculate the current through the system as a function of the bias voltage.

Let us now examine the question whether the electron density p(r) can uniquely

determine the transport properties of the system, specifically, whether p(r) alone is

sufficient in determining the current through the system. Because the current is a function

of the bias voltage, if p(r) is the basic variable of the system in the same way it is for a

ground state, knowing p(r) should allow one to determine the voltage bias across the


scattering region.









In both leads, the potential and the electron density can be approximately taken as

the ground state bulk value, due to the screening by the two contacts. This is the

approximation generally used in the previous studies. Suppose the first Hohenberg-Kohn

theorem for the ground state can apply at least in the two leads. Then, according to the

theorem, the Fermi energy /Lu is decided by the electron density in the left lead PL (r) to

an constant VL and the Fermi energy /Ru is decided by the electron density in the right

lead PR (r) to another constant VR. Since both leads are connected to a battery, these two

constants, VL and VR, should not be independent. However, according to the ground

state Hohenberg-Kohn theorem, if we shift Fermi energies of left and right leads, thus

changing the voltage bias across the scattering region, the electron density in the two

leads should remain unchanged. In another words, these two constants, VL and VR, are

independent if only given electron density.

The result is that the voltage bias /L /i is completely arbitrary, if one is only

given the electron density in the two leads. Since the bias voltage should not depend on

the scattering region, we come to the conclusion that for the open system under a finite

bias, the electron density p(r) alone is not sufficient to describe all properties of the

system, thus conventional DFT is not applicable. The Hohenberg-Kohn theorem must be

modified in order to be applied to such systems.

5.2.1 Voltage-dependence of the Density Functional

In order to see how the Hohenberg-Kohn theorem should be changed, we start from

a rigorous approach to this problem based on the adiabatic time-dependent procedure. We

take the same time-dependent process as that in reference 24 and reference 25: at time

t = -oo, the unperturbed system consists of three uncoupled regions, the left lead, the









scattering region, and the right lead. Each one maintains its own equilibrium state

(ground state for zero temperature). The Fermi energies for the left and right leads are /u

and /Ru respectively. The Fermi energy of the scattering region does not need to be the

same as left or right leads, but for simplicity, we assume that it is in the same equilibrium

state as the right lead. The bias voltage is defined as/u Ru The couplings between

leads and the scattering region turn on slowly from t = -co tot = 0, such that at any time,

the system is in the stationary state. The goal is to relate the current at t = 0 with the

couplings completely turned on to the unperturbed Fermi energies '/L and /R Given the

initial state, the properties of the time-dependent system can be obtained by solving the

time-dependent Schrodinger equation

.0
S (t) = H(t)(t). (5-1)

The Hamiltonian can be written as

H(t) = H +V(t)

IH,= 1V,2+i + Vext(r,)+Z Vee(rj) (5-2)
2 1 I V(t)= V(t)( ect)


where, in the Hamiltonian, the first term H, consists of the kinetic energy, the nuclear-

electron interaction Vex and the electron-electron interaction V,. The second term is an

extra time-dependent external potential, where V can be chosen to be two barriers in the

two contacts and zero in all other regions. The two barriers should be high enough such

that at t = -oc, the transmission probability of electrons tunneling through these two

barriers is negligible; thus, the two leads and the middle barrier are well separated. The









constant c should be positive and small enough such that the couplings between leads

and the middle barrier turn on adiabatically.

For such a time-dependent system, Runge and Gross [69] showed that the wave

function at time t, ry(t), can be determined by the time-dependent electron density p(t)

and the initial state, up to a time-dependent phase factor,

,(t) = e 'O't)A/P(O, /(-o0)] (5-3)

where, [4p, r/(-co] is a functional of the electron density and the initial state.

Particularly, if the initial state is the stationary ground state at t = -0o, then according to

the first Hohenberg-Kohn theorem for ground state, it is completely determined by the

initial electron density; hence ((-oc) can be eliminated in the equation (5-3), so that

y/(t) is determined up to a phase factor by p only.

Neglecting the phase factor, we come to the time-dependent DFT in this case,

which claims that any time-dependent physical quantity can be expressed as the

functional of time-dependent electron density.

However, for our case, at t = -oc, the system is not in the ground state. The initial

state V,(-oc) should be constructed to represent three local equilibrium states for three

uncoupled regions with two different Fermi energies /L and /R which actually is an

excited state at zero temperature characterized by /L and /R Apparently, the initial

state is determined by the Hamiltonian and these two Fermi energies. Since a trivial

constant added to the Hamiltonian will not affect the initial state, but will shift /'L and

PR by the same amount, we know that the initial state is determined by the Hamiltonian

and the voltage biasVb = /u R .









Applying the first Hohenberg-Kohn theorem for the ground state to three regions

which are in local equilibrium states, the Hamiltonian can be determined by the initial

electron density to a constant. As we discussed before, the bias voltage is independent of

the electron density, so the initial state is a functional of electron density p and the

voltage bias Vb.

Att = 0, the Equation (5-3) can be written as,

f(t) = e p(t)/[P(t),1 ] (5-4)

which means that at a given Vb, ry(t) is determined by p to a time-dependent phase

factor. y/lp, b ] can be equivalently written as i/[Pb ] \ which means that the density

functional form io] depends on the bias voltage Vb. We come to a conclusion that all

steady-state properties at t = 0 can be expressed as functionals of p at a given bias

voltage V The most important of these functionals is the exchange-correlation energy

functional Ex, [].

The voltage-dependence of those density functionals actually comes from the

voltage-dependence of the initial state. The first Hohenberg-Kohn theorem for the

equilibrium state is just a special case forVb = 0.

The voltage-dependence of the density functionals introduce substantial difficulty

to the implementation of the theory, since we need different functional forms for different

bias voltages. It is even worse than that: since the voltage-dependence of the initial state

varies with the system, the voltage-dependence of the density functionals varies with the

system, which means we don't have universal density functional for all systems even

when they are under the same bias voltage. This disadvantage is caused by the fact that









the basic variable of the functional, the function p(r), can not determine the bias voltage

Vb, and is not sufficient to describe the system. In order to get rid of this disadvantage,

we need to introduce other functions as variables of the functional, which together with

p can completely determine the system.

5.2.2 The Non-equilibrium Electron Density

The key to eliminate the voltage-dependence of the functional in the equation (5-4)

is to introduce other functions into the functional which can determine the bias voltage

Vb and thus determine the initial state y,(-oo) together with the electron density .

For the non-equilibrium system, the electron density can be expressed as [20, 22],


p(r)= I'd (r, E) (5-5)
21i

where, G'(r, ) is the non-equilibrium lesser Green's function defined for the stationary

state. For the system under study, which is characterized by two Fermi energies /, and

/Ru (suppose /u > /,R), the electron density can be divided into the equilibrium part peq

and the non-equilibrium part p,,

p(r)) (r)+ pn(r) (5-6)

where,

1 ^R
2ii
P,, (r)=- 2IP dsG(r,E)
7 (5-7)

p (r) = I dsG (r, E)


The term Pe, is called the equilibrium density in the sense that it can be equivalently

written as an equilibrium integral [20],









PLR
pe (r)= Im dsG (r, E) (5-8)
-co

where, GR is the usual retarded Green's function.

We want to show that if given both Peg and p, at the initial time, the bias voltage

is determined and then the initial state is determined.

When t = -oo, the left lead is well separated with respect to other regions. The left

lead is in a local equilibrium state with the Fermi energy PuL while the scattering region

and the right lead are in a local equilibrium state with the Fermi energy pR In this case,

pn in the right lead is zero and in the left lead it also can be written as an equilibrium

integral,


p,(r) = -Im d~GR(r, ). (5-9)
i"
PLR

Since the left lead is in local equilibrium, the first Hohenberg-Kohn theorem for

equilibrium state applies.

We know that in the left lead, the Hamiltonian is determined by

p(r) = pe (r) + p, (r) to an arbitrary constant. If a constant V, is added to the

Hamiltonian, GR (E) changes to GR (E + V0) Then, with p,, given, from equation (5-8),


we see that PRt changes to PR + Vy. Introduce the new pR and GR (E) into the equation

(5-9), we see that with the given p,, PL becomes /L + V0. Hence the bias voltage

Vb = UL /UR is completely determined without any arbitrary constant. Then, the

Equation (5-4) can be rewritten as,

Y(t =0) = e '(t =o[pq (t = 0), p (t = 0)] (5-10)









where, the voltage-dependence of the functional is eliminated. Based on the functional

I[Pe, ,Pn j, all steady state properties can be expressed as functionals of Pe, and p,.

5.2.3 The Stationary Principle and the Single Particle Picture

For the time-dependent system, we can define the action A as,


A = ~(t)i- IH(t) V(t))dt (5-11)


where y,(t) is the many body wave function. If y(t) is normalizable, then given the

initial state, y(-oo), the stationary requirement of the action,

6A
=0 (5-12)
9(0

is completely equivalent to equation (5-3), where 38(t) is an arbitrary change of the

wave function under the conditions that 38/(0) = 0 and the normalization is unchanged.

There is no difficulty to normalize the wave function in the closed system, but it seems to

be impossible to normalize the wave function for the open system which has the

geometry shown in Fig (1-1). Thus, the Equation (5-12) can not be directly applied.

However, for the case of time-dependent DFT, which means that the initial state is a

stationary ground state, the action is the functional of the electron density A = A[p]. As

long as the wave function y(t) yields the correct electron density, we still can use the

Equation (5-12) to calculate the action and the stationary requirement becomes,

9A
= 0 (5-13)
8p(t)

with the conditions that 8p(0) = 0 and 8p doesn't change the neutrality condition of two

leads and the scattering region. This procedure can be generalized to the non-equilibrium









initial state case. According to the previous subsection, for the open system under a finite

bias, the action can be written as A[peq, np, and the stationary principle becomes,

6A 6A
=0 (5-14)
pe, (t) 8pn (0

where the arbitrary change of electron densities should not change the initial conditions

nor the neutrality condition of the system.

For the adiabatic time-dependent process, the action reduces to

0
A= E(t)dt. (5-15)


Then, the Equation (5-14) becomes,

= 5E0 (5-16)
3p,q (t) gp, (t)

where, E is the total energy of the system at time t.

Both the ground state and the time-dependent DFT work in the framework of a

single-particle picture. We follow the same procedure. Consider an independent particle

system, which has a same geometry as shown in Fig. 1-1 and whose single particle

orbitals can be written as #,f, where m denotes all good quantum numbers other than

the energy At the initial time, since the two leads are well-separated with respect to

the scattering region, fm, consists of three kinds of states, 0,,, and ;b, which are

localized in the left lead, the right lead, and the scattering region respectively. Those

three kinds of states obey the equilibrium distribution functions of three regions

respectively. When the temperature is zero, those distribution functions are step

functions. The adiabatic time-dependent procedure described in previous sections will not

change the occupation number of each state when it evolves with the time. Further









assume that at the initial time, 0, ,,0, and b ,, are perfectly normalized such that the

neutrality conditions are satisfied in all three regions.

Since the system is non-interacting, both Green's functions G (E) and G' () are

diagonalized by m,~ It is straightforward that equation (5-7) becomes,


Peq= Z ,e 2
m2 (5-17)
P, = fm( )W,2R
m,/L >E>/PR

where f, is the occupation number. Since the occupation number of each state does not

change with time and/ u > R,, p,, is,


P, = 2 (5-18)


Assume that Peq and p,, yielded by equaton (5-17) and (5-18) are the same as

those of the true interacting system under study. Then, similar to the equilibrium state

DFT, we write the total energy asE = T + E where T is the single-particle kinetic

energy,

T = T +T

Teq Z (O -V2 m). (5-19)
m,E<,uR
T,= Z (, -Iv20)
l,/L >E>/R 2

The potential energy E, can be written as,


E= rV, (r)p(E) +Ec, +E
E 1 J -1 (5-20)
2-: j- 2









where EcoU, is the classical electron-electron repulsion energy and Exc is the exchange-

correlation energy. E, is the functional of peq and p, We can define two effective

potentials based on E, as,


Jeq 8EV
Vef -
-Peq .(5-21)

Ve" = E
ef Pn

The Equation (5-16) can be written in terms of those orbitals as,


(E V2 -), R = 0
2 (5-22)
1
(E + V2 2 V,,) m,/i.>>>R = 0

There are two boundary conditions for the system under study. The open boundary

condition incorporated with equation (5-22) yields the scattering states and the bound

state boundary condition yields bound states.

Unlike equilibrium DFT, we have two different effective potentials instead of one.

The non-equilibrium electrons see a different mean field potential than the equilibrium

electrons do. The self-consistent procedure makes two electron densities Peq and p, to

converge respectively, and the current is carried by the non-equilibrium electrons.

5.3 Exchange energy for uniform electron gas

The exchange-correlation energy can be written as the sum of the exchange part

and the correlation part, Ex = Ex + Ec. Both Ex and Ec are functionals of Peq and p,.

They are very complicated for a real, inhomogeneous system. In this section, as a case

study, we calculate the exchange energy for the uniform electron gas under a finite bias,

which represents a perfect conductor connected to two reservoirs infinitely far away.









This study serves as a test for the theory described in the previous section.

For the uniform system, if we take periodic boundary condition as q(x + 1) = O(x),

then the single-particle orbitals can be written as,


(k, k, k) = e (kx+kk ) (5-23)
2V2
/3/2 W12


where kxy = -nx,y, with nx,y, are integers. Set the z axis to go from left to right,


which is the direction of the current. There are two kinds of states for this system: the

right-going states (k > 0) and the left-going states (k < 0). From equations (5-17) and

(5-18), we see that the occupation of those states is: belowke, all states are occupied and

between k, and k,, only right-going states are occupied. Define two quantities k, and k,

h2k2 h2k2
by = /R, -= L. Two electron densities P,, (which comes from all states below
2m 2m

ke) and p, (which comes from all right-going states above ke), can be related to k, and

k3 1
k,. After some algebra, we have peq = and p,,6 (k -k). We can see more
37T 6;T

clearly from this case that the total electron density p = Pe + p, alone is not enough to

describe the system: two systems with the same p have different electronic structures if

they have different set of Peg and p,.

The exchange energy Ex can be evaluated by the equation [30]


E = 1 ri(\,r) cI2clrl (5-24)
4 r ll

where (ri,2 ,) is the density matrix and its diagonal part is the electron density. It can be









calculated as p(r, r) = 2 Yer'12 where the summation is over all occupied k For a


large number of occupied states, the summation can be replaced by an integral,

S ke 2T I kt ,T/2 2,
p(, i) = k2dk SinOdO fdke'12 + k 2dk f SinOdO [de 12 (5-25)
0 0 0 ke 0 0

where the first integral is the contribution of all states below ke, which is an equilibrium

state integral and can be evaluated analytically. We denote this term as Pe. The second

integral is the contribution of occupied states between k, and k,, which can be denoted as

/,. For the non-equilibrium integral, since only states with positive kZ are occupied, the

variable 0 of the second integral goes from 0 tor /2 We introduce two variables

R = (tr + 2)/2 and S = '2. Then the density matrix is only a function of s and the

exchange energy can be written as,

E = -dRK
1 1 (5-26)
K= p(s) 2
4 s

where K is a constant since the system is uniform. For a given electron density p, when

the bias voltage increases, p,, increases and Peq decreases until at a certain bias voltage

Vmax, Pe becomes zero and p,, equals to p. Vmax defines the biggest difference

between /, and /R we can have for a uniform system with a given p at zero

temperature. Given the electron density p and the bias voltage L /R we can

numerically evaluate p andK. Fig. 5-1 shows the ratio K/Ko as a function of total

electron density and bias voltage. Ko denotes the term calculated at zero bias. From the










figure, we clearly see the voltage-dependence of the exchange energy for a given electron

density p. This supports the analysis of the previous section: the electron density alone


is not enough to describe the system. Also, the p -dependence of the exchange energy at


a given bias voltage, which will determine the form of the density functional, varies with

the bias voltage. It means that we will have difficulty in finding a universal functional for

all bias voltages, which agrees with the previous section too.



1.012


1.008 / /
o \ \ p= 3377
1.004 \ p=17.29


1.000- p=7.30


0.996- =2.16


0.992
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
Bias Voltage (Volts)



Figure 5-1. Exchange energy at different bias voltages for different p (Electrons/m3)

We notice that although the voltage-dependence of the exchange energy varies with

the electron density, the behavior is similar for all electron densities: as the voltage

increases, the exchange energy increases until it reaches a maximum and then decreases.

In order to understand this behavior better, we rewrite the Eq. (5-26) as,









K =Ke +K +K

Keq f Ieq

1 :1 2 (5-27)
K = 4 s 2


4 s

The first term Keq is the contribution of the exchange of equilibrium electrons. The

second term Kn comes from the non-equilibrium electrons and the third one denotes the

contribution of exchanging between equilibrium and non-equilibrium electrons. The

voltage-dependence of these three terms for p = 7.29 Electrons/m3 is shown in Fig. 5-2.

At bias zero, since p, is zero, there are no non-equilibrium electrons. Hence K,

and Keqn are zero. As the bias voltage increases, peq decreases, which leads to the

decrease ofK, and p, increases, which leads to the increase of both K, and Keqn The

term Keq n reaches its maximum at a certain bias voltage and then decreases. At the

maximum bias voltage, p, = p and there are no equilibrium electrons, then both Keq

and Keqn are zero at this bias voltage.

According to the previous section, the exchange energy can be expressed as a

functional of p and p,. We write K as,

K = Ko [p1]+ f[ p, p (5-28)

where, Ko [o] is the one for the equilibrium system with the electron density p; this term

is well known. We plot the numerical results of K / K as the function of p, / p for

different electron densities in Fig. 5-3.








80





12

i p12 =7.30 0
10- E0 o
x

8-
E"
0 0


,, \ /


2 o6 OO-
Lo
2- O m m
O-)

















3.0 =7.30 m m m m m






1.5\ -
1.0 0
(b)






3.0 p=7.30

2.5 -
2.00











electrons and the exchange between non-equilibrium and equilibrium


Sa a
0.0 0.2 0.4 0.6 0.8 1.0 1.2
Bias Voltage (Volts)




(a)
3 .5 ----------------------
















2.0Figure 5-2. Split of the exchange energy: (a) the exchange between non-equilibrium






-0.5electrons and (b) the exchange between equilibrium electrons.





electrons and (b) the exchange between equilibrium electrons.











The numerical results show that we have a universal functional f[ for all
I P


electron densities. These data can be fit to the function,



f PJ +b


(5-29)


where a = -0.0497 and b = 0.0436.


1.012-


1.008-


1.004-


1.000-


0.996-


0.992-


0.2 0.4 0.6 0.8


p' p



Figure 5-3. Exchange energy as the functional of p, / p for different electron densities.

Then the exchange energy can be written as,


E=-C 4/3 + P


(5-30)


where, C, = 0.7386.


Equation (5-30) gives the exchange energy functional of the uniform system for all

bias voltages. It clearly shows the dependence on the non-equilibrium electron density as

well as the dependence on the total electron density, which support the analysis of the


-m-p=7.30
0 p=2.16
+ p=17.29
x p=33.77









previous section. Whether or not this functional can be generalized to the inhomogeneous

system has to be resolved in a future study.



5.4 Conclusion

In conclusion, we have shown that the transport properties of an open system under

a finite bias can not be determined by the total electron density alone. We introduce two

electron densities pe and p, to describe such a system and show that the non-

equilibrium electrons experience different effective potential from the equilibrium

electrons. As a test, we study the exchange energy of the uniform electron gas under a

finite bias. The numerical results clearly show the dependence to the non-equilibrium

electron density, which support the theory. We fit the numerical data to the exchange

energy functional. The applicability of the functional to the inhomogeneous systems

needs to be addressed in future studies.














CHAPTER 6
SUMMARY AND CONCLUSIONS

We introduce the main ideas of first principles method for open system under a

finite bias. The method has been successfully applied to study the transport properties at

nano scale, for example, molecular electronics. The achievements of the method can be

summarized as follows. The shape of calculated I-V curve can be compared to the

experiments and peaks in the curve can be explained by tunneling through molecular

orbitals.

We apply the state-of-art approach which combines the non-equilibrium Green's

functions technique and DFT to Au/S/Azobenzene/S/Au molecular junction. The results

show that the azobenzene molecule is a good candidate of light-driven molecular switch,

which is important for application of nanoelectronics. The I-V curve of cis molecular

shows a negative differential conductance at a certain bias, which is due to the

suppression of the tunneling through HOMO due to bias voltage.

We also generate the method to include spin and implement the method in a wave

function representation. The method is used to study the TMR through a layered

Fe/FeO/MgO/Fe magnetic junction. The results suggest a bias-enhanced TMR which is

rarely seen in previous experiments. The enhancement of TMR due to bias voltage is

explained by the effects of tunneling through the surface resonance states. Our analysis

suggests that this phenomenon should be able to seen in other systems.

In the final part of the thesis, we discussed the problem in the current approach and

suggest that in order to correctly describe the non-equilibrium effects introduced by the






84


external bias, an additional parameter other than the total electron density should be

introduced into the equilibrium DFT. For a case study, we calculate the exchange energy

for an uniform electron gas and clearly see the voltage-dependence of exchange energy.

By introducing 'non-equilibrium electron density', we show that the voltage-dependence

of the functional can eliminated.

This work is supported by the Alumni Fellowship of University of Florida and the

US Department of Energy under contract DE-FG02-02ER45995.














APPENDIX A
EXPANSION OF LESSER GREEN'S FUNCTION

The non-equilibrium Green's functions are defined as

G(1,1') = -i( )(1 (A-l)

where, the shorthand notation 1 (x, t,) is used. The time-order operator Tc is defined

on two time branches, + branch, which runs from t = -co to t = +oc, and branch,

which runs in the opposite direction. With two time labels, tj and t,, which can be

located on either of two time branches, we have four kinds of non-equilibrium Green's

functions:

G(1,) t, t, e+
G'(1,1') tl e-,t, e+
G(J1,') = (A-2)
G'(1,1') tj +, t,e-
G' (1,1') ti, ti, -

These four Green's functions can be denoted as the time-ordered Green's function Gt,

the greater Green's function G', the lesser Green's function G', and the antitime-

ordered Green's function G'.

For the Hamiltonian which is given by equation (2-1), a particular lesser Green's

function Gk defined as


G (, (A-3)






86


where, the indices k, a denote channels of the left lead and n denotes the basis states of

the device region. The angle brackets indicate the ensemble average. The time of the

operator d+ lies on the branch and the time of the operator Cka lies on the + branch.


tka tV
+ tk. v 9kta,ka ka,m mGn
tn 1


tn

tkc

tv
'LW


g ka kc kcxm mn
m


Figure A-1. Expansion of lesser Green's function Gkx,,. The time of the perturbation can
be either on + branch or on branch.

The coupling term Hoo, is taken to be the perturbation V Then, the Dyson

equation G = G, + GVG can be used to expand the Green's functions, where the term

Go is the Green's function without the perturbation. Inserting the basis states into the

Dyson equation, we have,

Gk,,, = (ka G n)= (ka G n) + (ka GVG n) = gkz, mG,,,, (A-4)


where, the term (ka GG, n) is zero because electrons can not propagate from the left lead

to the device region without the coupling term and the term gk,,k is the bare Green's

function of the left lead. The equation (A-4) can be used for both equilibrium and non-

equilibrium Green's functions. For non-equilibrium case, since the time of the

perturbation can be either on + branch or branch, the equation (A-4) gives two terms.







87


As an example, the expansion of the lesser Green's function Gk contains two terms as


shown in Fig. A-1. Thus, Gk can be written as


Gkn = 7Vkam kakaGn gakaGmn. (A-5)
m














APPENDIX B
DETAILS OF NON-EQUILIBRIUM SELF-CONSISTENT PROCEDURE

As mentioned in Chapter 3, the device region includes one unit cell of the left lead

(denoted as L), the molecule (denoted as M), and one unit cell of the right lead (denoted

as R). The first step of the calculation is to compute self-energies of two leads according

to equation (3-2). Since the tight-binding model is used to describe two leads, only the

part L (R) of the device region is affected by the left (right) semi-infinite lead. Thus the

dimension of the self-energy of the left (right) lead YL (R ) is the number of basis states

of one unit cell of left (right) lead. In our calculation, the Gaussian basis set is used. Due

to screening, self-energies of two leads are not affected by the device region, so they

don't change in the self-consistent procedure.

The self-consistent procedure starts from an initial electron density p, (r) in the

device region. Given the electron density, the Hamiltonian can be calculated as

H 1v (')
H 2 2 +'+ V (r)+ + Vbl]+ (r), (B-l)

2- P-r

where, the last term Vb,, (r) is the linear voltage drop, Vb,K, (r) = -R y The current
d

is flowing in y direction and d measures the distance between two leads. The retarded and

advanced Green's function can be achieved by

Gr', = (B-2)
SS t t

where, S is the overlap matrix of Gaussian basis states and H is the Hamiltonian matrix.









In order to see it more clearly, equation (B-2) can be expanded as

K+SLL HLL HL +SML HML E+SRL HRL
G' = E '+S HLM ES -Ha ~+ Sm H -H (B-3)
E + -LR -HLR +S H E+S H~- RR R

where, indices L, M, and R denote three parts of the device region respectively.

Then, the lesser Green's function can be computed according to equation (2-11)

and the density matrix is determined by


jP, = IG; (8)dE, (B-4)

where, indices i, j denote Gaussian basis states.

The electron density can be directly related to the density matrix as

p()p) =( = )= (r i)(i p j)(j 6) = 01, ()p,,b0(), (B-5)
1,j 1,j

where, 0, (i) is the Gaussian basis in real space.

The electric static potential is defined as

S() () + Vb, () (B-6)


An interesting quantity is the difference between the electric static potential under a

finite bias and the electric static potential of zero bias, Ve ,


AV,, = V ,, V, = Vb, (i) + tep(c') (B-7)
which shows the effects of the voltage bias on the electric static potential.

which shows the effects of the voltage bias on the electric static potential.