1
00:00:01,450 --> 00:00:03,450
大家好我是 Shelly Kelly

2
00:00:03,450 --> 00:00:08,725
这节来展示如何使用 Artemis 

3
00:00:08,725 --> 00:00:14,350
分辨铁样品的第一壳层是氧还是硫

4
00:00:15,091 --> 00:00:17,550
这里是 Artemis

5
00:00:18,175 --> 00:00:21,625
主控制窗口

6
00:00:21,950 --> 00:00:25,625
到文件菜单

7
00:00:25,625 --> 00:00:28,950
打开项目文件

8
00:00:31,393 --> 00:00:34,375
到桌面

9
00:00:34,785 --> 00:00:40,525
打开之前的 Athena 文件

10
00:00:42,850 --> 00:00:48,300
可以看到之前的三组数据

11
00:00:48,300 --> 00:00:54,375
点击任一数据就会展示其 R 空间谱图

12
00:00:54,375 --> 00:00:59,700
利用画图窗口可以展示不同空间的数据

13
00:00:59,700 --> 00:01:01,700
选择中间的数据

14
00:01:01,700 --> 00:01:04,750
k 范围从 3 到 12

15
00:01:04,750 --> 00:01:07,825
加载选中的数据

16
00:01:08,300 --> 00:01:10,900
数据窗口中的对应信息出现

17
00:01:10,900 --> 00:01:16,500
之前在 Athena 里面的参数信息也转移到这个窗口中

18
00:01:16,500 --> 00:01:19,725
k 从 3 到 12

19
00:01:20,275 --> 00:01:26,850
拟合范围也就是反傅里叶变化的范围

20
00:01:26,850 --> 00:01:28,850
从 1 到 3

21
00:01:28,850 --> 00:01:34,875
这代表着我们会对 k 的一次方到三次方数据拟合

22
00:01:35,524 --> 00:01:40,525
下面勾选的选项是告诉 Artemis 

23
00:01:40,525 --> 00:01:44,625
实验数据会被包含并且拟合后会展示

24
00:01:46,575 --> 00:01:49,852
除了数据以外

25
00:01:49,852 --> 00:01:53,693
还要加载 Feff 模拟

26
00:01:54,325 --> 00:01:58,250
到 Feff 那里点击 Add

27
00:02:00,400 --> 00:02:04,200
到课程教材文件

28
00:02:12,109 --> 00:02:17,725
打开硫化铁的 inp 文件

29
00:02:21,675 --> 00:02:25,875
这个文件很好用

30
00:02:25,875 --> 00:02:30,875
因为含氧和含硫的路径都有

31
00:02:31,675 --> 00:02:35,750
这是晶体结构数据

32
00:02:35,750 --> 00:02:39,375
样品是硫化铁

33
00:02:40,125 --> 00:02:43,425
空间群在这里

34
00:02:43,425 --> 00:02:45,600
晶体常数在这里

35
00:02:45,600 --> 00:02:50,400
要计算的簇尺寸达到了 9

36
00:02:50,400 --> 00:02:53,650
有点大了，改成 6

37
00:02:54,064 --> 00:02:58,400
最长的路径只计算到 5 Å

38
00:02:59,250 --> 00:03:03,500
下面是晶胞中原子坐标

39
00:03:03,500 --> 00:03:07,100
勾选上的铁是中心原子

40
00:03:07,100 --> 00:03:09,750
这是指定的 K 边

41
00:03:09,750 --> 00:03:11,750
点击 Run Atoms

42
00:03:12,675 --> 00:03:17,875
页面会跳转到 Feff 界面

43
00:03:18,125 --> 00:03:21,100
你可以看到

44
00:03:21,100 --> 00:03:25,525
一些文件中结构参数信息

45
00:03:25,525 --> 00:03:27,525
这里标题

46
00:03:28,225 --> 00:03:33,175
中心原子被定在 K 边

47
00:03:33,175 --> 00:03:35,175
用这里的 1 代表

48
00:03:35,175 --> 00:03:40,125
下面代表着不同的 potential (势)

49
00:03:40,125 --> 00:03:41,850
会被用作计算

50
00:03:41,850 --> 00:03:44,347
第一个是中心铁原子

51
00:03:44,347 --> 00:03:46,950
第二个是邻近铁原子

52
00:03:46,950 --> 00:03:50,025
硫原子，氧原子

53
00:03:50,025 --> 00:03:52,275
往下翻

54
00:03:53,225 --> 00:03:57,350
可以看到第一壳层是氧原子

55
00:03:57,350 --> 00:04:04,825
共计有 6 个氧原子

56
00:04:04,825 --> 00:04:07,850
要检查这个列表中的参数是否合理

57
00:04:07,850 --> 00:04:13,075
铁氧之间距离大约 2.0 Å 是合理距离

58
00:04:13,075 --> 00:04:16,575
以及后面的硫原子

59
00:04:17,325 --> 00:04:20,950
确保一切正常后

60
00:04:20,950 --> 00:04:24,575
点击 Run Feff

61
00:04:27,025 --> 00:04:31,175
跳转到下一个页面 Paths

62
00:04:31,175 --> 00:04:36,225
这些路径接下来会被用作拟合

63
00:04:36,950 --> 00:04:40,000
现在打开数据窗口

64
00:04:40,000 --> 00:04:45,275
直接把铁氧路径拖到 Path list

65
00:04:45,725 --> 00:04:50,050
这样就可以用这条路径来拟合了

66
00:04:51,200 --> 00:04:53,425
现在可以把 inp 文件最小化了

67
00:04:53,425 --> 00:04:55,975
如果还想打开

68
00:04:56,300 --> 00:04:58,300
在 Feff calculations 下方

69
00:04:58,300 --> 00:04:59,900
鼠标左键点击对应文件即可

70
00:04:59,900 --> 00:05:01,989
再点击就关掉了

71
00:05:02,375 --> 00:05:04,375
数据窗口同理可证

72
00:05:04,625 --> 00:05:06,625
点击即可展示和最小化

73
00:05:07,325 --> 00:05:10,050
有了路径之后

74
00:05:10,050 --> 00:05:13,050
需要设置一些参数

75
00:05:13,050 --> 00:05:16,364
来描述具体的信息

76
00:05:16,625 --> 00:05:24,925
我们先加上……

77
00:05:25,084 --> 00:05:27,625
现在配位数是 6

78
00:05:28,250 --> 00:05:30,250
S02 被设为 1

79
00:05:30,250 --> 00:05:33,276
第一次尝试可以

80
00:05:33,925 --> 00:05:37,450
我们可以加入铁氧路径的 ΔE 

81
00:05:37,450 --> 00:05:43,150
和用于调整参考距离的 ΔR

82
00:05:43,650 --> 00:05:49,650
我们还可以设置 σ2 为绝对值 

83
00:05:50,475 --> 00:05:59,775
利用 abs(变量) 可以保证最后的数值不会为负数

84
00:05:59,775 --> 00:06:03,275
鼠标右键点击参数可以转移到猜测窗口

85
00:06:04,025 --> 00:06:08,350
这里是猜测窗口

86
00:06:08,350 --> 00:06:12,875
我们猜测 ΔE = 0

87
00:06:13,575 --> 00:06:17,650
同理设置 ΔR

88
00:06:17,650 --> 00:06:19,800
和 σ2

89
00:06:22,209 --> 00:06:25,700
重要的是这些参数需要跟参考路径足够接近

90
00:06:26,000 --> 00:06:31,200
σ2 数值大多在 0.00数字

91
00:06:31,200 --> 00:06:33,450
一开始猜个 0.003

92
00:06:34,000 --> 00:06:37,125
我个人喜欢在第一次拟合的时候

93
00:06:37,125 --> 00:06:39,375
把所有变量设为常量

94
00:06:39,375 --> 00:06:42,400
直接把路径加和在一起

95
00:06:42,400 --> 00:06:44,400
跟实验数据进行比较

96
00:06:45,153 --> 00:06:50,650
那现在就开始

97
00:06:50,650 --> 00:06:52,650
点击 Fit 

98
00:06:56,050 --> 00:06:58,725
可以看到

99
00:06:59,475 --> 00:07:05,775
傅里叶变换后的强度在上，实部图像在下

100
00:07:05,775 --> 00:07:09,825
红线是 Feff 模拟

101
00:07:10,325 --> 00:07:12,575
绿线是拟合范围

102
00:07:12,950 --> 00:07:17,050
可以通过画图窗口来改变图像展示方式

103
00:07:17,700 --> 00:07:19,725
如果你想回看这组图

104
00:07:19,725 --> 00:07:24,675
回到数据窗口点击 Rmr 就好

105
00:07:24,675 --> 00:07:27,975
代表着强度和实部

106
00:07:28,206 --> 00:07:31,000
如果想看其他的

107
00:07:31,000 --> 00:07:33,600
比如只想看强度

108
00:07:33,600 --> 00:07:35,400
像这样

109
00:07:35,400 --> 00:07:39,600
设定为强度，对应数据在列表中

110
00:07:40,208 --> 00:07:42,425
如果是实部图像

111
00:07:43,150 --> 00:07:45,150
也就只有实部谱图

112
00:07:45,150 --> 00:07:48,150
也可以把拟合窗口关掉

113
00:07:48,450 --> 00:07:51,125
如果你想的话，取消勾选这个就好

114
00:07:53,200 --> 00:08:00,175
看起来路径不匹配

115
00:08:00,175 --> 00:08:04,100
是 Feff 的路径不太对

116
00:08:04,600 --> 00:08:08,975
第一个发现是铁氧路径强度太强

117
00:08:09,300 --> 00:08:11,300
其次是距离太短

118
00:08:11,300 --> 00:08:13,325
所以距离需要改动

119
00:08:13,325 --> 00:08:15,750
强度也要减小

120
00:08:16,378 --> 00:08:19,575
要这么做的话

121
00:08:19,575 --> 00:08:25,450
先减少配位数

122
00:08:25,450 --> 00:08:27,450
现在是 6

123
00:08:28,150 --> 00:08:30,900
现在改成 1

124
00:08:31,275 --> 00:08:36,025
在 N 这一行必须是数字

125
00:08:36,550 --> 00:08:38,800
如果要让它设为变量

126
00:08:38,800 --> 00:08:40,800
得把参数加在下一行

127
00:08:40,800 --> 00:08:43,925
N 和 S02 都是强度相关参数

128
00:08:44,425 --> 00:08:49,250
把这个改成 amp * NO

129
00:08:49,250 --> 00:08:54,775
也就是 S02 乘以

130
00:08:55,425 --> 00:08:59,150
对 amp 和 NO 选 set

131
00:09:02,275 --> 00:09:06,975
我们把配位数从 1 设为 3

132
00:09:06,975 --> 00:09:11,215
amp 保持为 1

133
00:09:11,215 --> 00:09:13,975
S02 通常在 0.7 - 1

134
00:09:13,975 --> 00:09:15,975
先就这样

135
00:09:17,725 --> 00:09:23,975
还要调整下距离，加个 0.2

136
00:09:23,975 --> 00:09:27,500
相比于之前长了 0.2 Å

137
00:09:27,825 --> 00:09:30,850
在跑次拟合

138
00:09:32,050 --> 00:09:35,000
峰强度看起来好不少了

139
00:09:35,775 --> 00:09:40,800
高度接近

140
00:09:41,250 --> 00:09:45,300
但看起来不是很好

141
00:09:47,650 --> 00:09:49,650
距离更接近了

142
00:09:49,650 --> 00:09:57,150
但拟合谱图的形状跟实验数据差距很大

143
00:09:57,743 --> 00:10:02,375
但更接近些了，所以试一把

144
00:10:02,900 --> 00:10:07,800
把参数变为变量

145
00:10:07,800 --> 00:10:09,800
把 amp 设为

146
00:10:12,175 --> 00:10:18,175
直接把 amp 定成 0.8

147
00:10:18,175 --> 00:10:23,500
我是从铁片拟合中知道这个数值的

148
00:10:23,500 --> 00:10:26,850
从固定铁片的已知配位数得到的

149
00:10:26,850 --> 00:10:30,400
这个数值是可以从其他拟合中借鉴的

150
00:10:30,800 --> 00:10:33,499
设成 0.8

151
00:10:33,499 --> 00:10:36,379
剩下的是变量

152
00:10:38,043 --> 00:10:40,043
点击 Fit

153
00:10:40,350 --> 00:10:46,375
看起来形状吻合了

154
00:10:46,375 --> 00:10:48,375
看下各自的最优解是多少

155
00:10:49,567 --> 00:10:52,675
这有问题

156
00:10:52,675 --> 00:10:54,675
ΔE 是 18

157
00:10:55,500 --> 00:10:58,825
有问题

158
00:10:58,825 --> 00:11:03,500
ΔR 一会儿来细讲

159
00:11:04,096 --> 00:11:06,096
ΔR 是 0.28

160
00:11:06,096 --> 00:11:08,450
跟之前的猜测接近

161
00:11:09,100 --> 00:11:11,300
σ2 是 0.005

162
00:11:11,300 --> 00:11:12,975
不差

163
00:11:12,975 --> 00:11:15,825
氧配位数是 4

164
00:11:15,825 --> 00:11:17,825
也不差

165
00:11:18,100 --> 00:11:20,600
现在来看下 ΔE

166
00:11:20,600 --> 00:11:22,325
到底发生了什么

167
00:11:22,700 --> 00:11:26,250
如果到 k 空间去

168
00:11:26,925 --> 00:11:32,175
你可以看到拟合的数据

169
00:11:32,175 --> 00:11:35,850
从 2 Å-1 开始一下子跳起来

170
00:11:36,675 --> 00:11:43,925
是为了去匹配窗函数中的数据

171
00:11:44,307 --> 00:11:48,475
如果我们把 dele 设为初始值呢

172
00:11:49,400 --> 00:11:52,150
点击 use bset ift

173
00:11:53,050 --> 00:11:56,175
使用最优解

174
00:11:56,175 --> 00:11:58,175
只把 delr 改回 0 

175
00:11:58,175 --> 00:12:00,601
并且定为常量

176
00:12:02,850 --> 00:12:05,750
把 guess 改成 set

177
00:12:11,025 --> 00:12:13,825
再跑次拟合

178
00:12:13,825 --> 00:12:17,475
把整个数据往左移

179
00:12:17,500 --> 00:12:19,975
点击 Fit

180
00:12:21,325 --> 00:12:23,325
保持其他不变

181
00:12:23,350 --> 00:12:29,300
可以看到峰形状不吻合了

182
00:12:31,355 --> 00:12:35,125
到 k 空间

183
00:12:35,707 --> 00:12:38,300
把图移回 k = 0 

184
00:12:38,300 --> 00:12:42,825
信号现在完全不同步了

185
00:12:42,825 --> 00:12:46,432
拟合与实验图像不再接近

186
00:12:49,625 --> 00:12:55,025
有可能是我们在 Athena 中的 E0 有问题

187
00:12:55,025 --> 00:12:58,150
需要加上 18 eV 

188
00:12:58,500 --> 00:13:00,950
而不是我们之前选择的 E0

189
00:13:01,000 --> 00:13:04,325
回到 Athena

190
00:13:05,850 --> 00:13:08,375
看下这个猜测是否合理

191
00:13:09,047 --> 00:13:12,325
这是 Athena 文件

192
00:13:12,375 --> 00:13:15,150
关掉这些

193
00:13:17,725 --> 00:13:21,300
把剩下的参数都跟第二组数据改成一样的

194
00:13:22,807 --> 00:13:25,900
到 E 空间

195
00:13:26,675 --> 00:13:32,275
现在的 E0 在吸收边靠上的地方

196
00:13:32,750 --> 00:13:36,675
这里做个标记

197
00:13:37,325 --> 00:13:39,325
选 indicators

198
00:13:39,425 --> 00:13:44,225
用这个圆圆的按钮

199
00:13:45,325 --> 00:13:48,075
双击指定的地方

200
00:13:48,100 --> 00:13:49,650
点击 E 键

201
00:13:49,750 --> 00:13:53,200
棕线代表 E0 所在位置

202
00:13:54,775 --> 00:14:00,025
我们还可以再放个标记

203
00:14:00,050 --> 00:14:02,050
在加上 18 eV 之后的 E0

204
00:14:04,425 --> 00:14:07,575
也在 E 空间中

205
00:14:12,475 --> 00:14:14,475
那也就在这儿了

206
00:14:14,650 --> 00:14:17,325
我们知道

207
00:14:19,500 --> 00:14:21,500
这大约是 k = 2 的地方

208
00:14:21,500 --> 00:14:25,500
我们知道这里 k = 0 不合理

209
00:14:26,028 --> 00:14:30,175
E0 不能在吸收边这么以外的地方

210
00:14:30,892 --> 00:14:35,425
另一个解决方法是用铁硫路径

211
00:14:40,082 --> 00:14:44,900
打开 Artemis 

212
00:14:57,125 --> 00:14:59,600
这个先等下

213
00:15:02,300 --> 00:15:09,175
我们为铁硫路径再加个 Feff 模拟

214
00:15:11,525 --> 00:15:17,275
我们去到桌面

215
00:15:17,550 --> 00:15:20,850
选择硫化铁文件

216
00:15:21,400 --> 00:15:25,650
打开是陨硫铁样品，硫化铁

217
00:15:25,750 --> 00:15:29,500
确保所有参数合理

218
00:15:29,525 --> 00:15:31,475
簇大小改为 6

219
00:15:31,500 --> 00:15:33,200
最长路径不超过 5 Å

220
00:15:33,200 --> 00:15:34,575
K 边

221
00:15:34,600 --> 00:15:36,600
勾选中心原子 K

222
00:15:36,625 --> 00:15:38,625
点击 Run Atoms

223
00:15:38,650 --> 00:15:43,450
往下翻确认距离

224
00:15:43,881 --> 00:15:45,881
看起来可以

225
00:15:45,975 --> 00:15:51,425
铁硫距离大约 2.3 Å

226
00:15:51,425 --> 00:15:53,050
要比铁氧长点

227
00:15:53,050 --> 00:15:55,050
符合预计

228
00:15:55,100 --> 00:15:57,375
点击 Run Feff

229
00:15:59,700 --> 00:16:04,425
路径窗口出现

230
00:16:04,625 --> 00:16:10,875
第一条铁硫路径准备就绪

231
00:16:11,175 --> 00:16:15,350
我不知道之前有没有具体介绍过这一页

232
00:16:15,375 --> 00:16:16,500
Degen 代表简并度

233
00:16:16,500 --> 00:16:18,750
指一样的路径有多少个

234
00:16:19,363 --> 00:16:23,975
第一壳层参考路径距离 2.3 Å 

235
00:16:23,975 --> 00:16:25,125
也是成键距离

236
00:16:25,375 --> 00:16:31,325
散射路径类型里面用原子来代表

237
00:16:31,375 --> 00:16:36,925
从中心铁原子到硫原子再回到铁

238
00:16:37,075 --> 00:16:40,800
所有单散射路径是绿色

239
00:16:41,275 --> 00:16:46,775
排名 (Rank) 说明每条路径相对第一条路径的重要性

240
00:16:46,775 --> 00:16:50,200
说明第一条简并度为 2 的路径

241
00:16:50,200 --> 00:16:55,550
强度是其他这几条简并度为 1 的路径两倍

242
00:16:56,850 --> 00:17:01,325
说明各自路径的重要程度

243
00:17:01,350 --> 00:17:05,375
加载第一条路径

244
00:17:07,100 --> 00:17:09,100
关闭文件

245
00:17:09,100 --> 00:17:18,650
现在 Feff calculations 有两个不同文件

246
00:17:18,900 --> 00:17:23,950
这个路径在列表中后

247
00:17:24,130 --> 00:17:25,700
我们需要设置参数

248
00:17:25,700 --> 00:17:31,100
回到我们之前拟合的设置

249
00:17:31,428 --> 00:17:36,575
利用之前设置好的来设置铁硫路径

250
00:17:36,575 --> 00:17:39,675
只需要鼠标右键点击

251
00:17:39,675 --> 00:17:43,750
把参数扩展到勾选上的路径

252
00:17:44,575 --> 00:17:51,025
这样两条路径的参数会是一样的

253
00:17:51,900 --> 00:17:55,600
参数必须手动设置

254
00:17:55,600 --> 00:18:00,050
或是在 Data 选项下方设置所有简并度为 1

255
00:18:00,183 --> 00:18:03,275
或是设为 Feff 默认值

256
00:18:04,575 --> 00:18:08,500
由于氧硫差异

257
00:18:08,500 --> 00:18:10,895
我们需要改变参数

258
00:18:11,775 --> 00:18:16,850
鼠标右键参数选择 guess 

259
00:18:18,219 --> 00:18:22,375
然后要踢掉铁氧路径

260
00:18:22,375 --> 00:18:25,275
取消勾选铁氧路径选项

261
00:18:26,150 --> 00:18:30,475
回到猜测窗口

262
00:18:30,475 --> 00:18:31,875
左上方 GDS 点击就有

263
00:18:32,835 --> 00:18:39,425
把这些参数设为 skip

264
00:18:39,425 --> 00:18:41,425
拟合时软件不会考虑它们了

265
00:18:42,075 --> 00:18:43,608
选 skip

266
00:18:43,608 --> 00:18:47,025
再设置这些参数

267
00:18:47,575 --> 00:18:51,275
设配位数 4

268
00:18:51,275 --> 00:18:54,575
铁硫 ΔR 为 0 

269
00:18:54,900 --> 00:18:57,575
σ2 为 0.003

270
00:18:57,575 --> 00:19:02,025
跑次拟合看看差距

271
00:19:02,975 --> 00:19:09,350
红线是 Feff 路径加和结果

272
00:19:09,350 --> 00:19:11,350
蓝线是实验数据

273
00:19:11,350 --> 00:19:15,650
可以看到不仅距离吻合

274
00:19:15,650 --> 00:19:22,925
两者的形状也接近

275
00:19:26,700 --> 00:19:31,300
那现在正式拟合

276
00:19:31,300 --> 00:19:34,025
拟合结果应该不会差太远

277
00:19:34,412 --> 00:19:37,000
把参数设为 guess

278
00:19:37,000 --> 00:19:39,532
ΔE 也是变量

279
00:19:40,616 --> 00:19:43,150
试下

280
00:19:50,250 --> 00:19:57,100
现在 ΔE 是 -0.8 eV

281
00:19:57,100 --> 00:19:59,600
很好

282
00:20:00,059 --> 00:20:02,059
配位数是 6

283
00:20:02,880 --> 00:20:05,900
ΔR 是 0.06 

284
00:20:05,900 --> 00:20:10,775
σ2 是 0.015

285
00:20:10,775 --> 00:20:18,875
这个值比较大但可以接受

286
00:20:18,875 --> 00:20:22,475
使用最优解

287
00:20:22,475 --> 00:20:26,500
到 k 空间

288
00:20:28,690 --> 00:20:31,200
现在是 k 空间

289
00:20:31,200 --> 00:20:39,400
可以看到 k = 0 地方更吻合了

290
00:20:39,400 --> 00:20:45,600
中间一段的数据也吻合

291
00:20:45,925 --> 00:20:48,500
可以做反傅里叶变换

292
00:20:48,500 --> 00:20:54,500
仅针对拟合范围内的数据

293
00:20:54,500 --> 00:20:58,650
整体来说结果接近

294
00:20:58,650 --> 00:21:02,800
但是丢失了在 k = 8 Å-1 左右的细节

295
00:21:02,800 --> 00:21:05,100
这是为什么呢

296
00:21:05,500 --> 00:21:10,175
是来自于这一段的信号没有拟合

297
00:21:10,175 --> 00:21:14,600
但扔在拟合的范围内

298
00:21:15,150 --> 00:21:20,596
如果把拟合范围降到 2.3 Å

299
00:21:21,100 --> 00:21:27,325
这一段丢失的细节会消失

300
00:21:27,325 --> 00:21:31,100
也会降低不确定性

301
00:21:31,850 --> 00:21:33,850
试下

302
00:21:34,675 --> 00:21:36,675
再拟合次

303
00:21:37,150 --> 00:21:39,750
现在窗口只到了 2.3 

304
00:21:40,075 --> 00:21:45,675
不确定性下降

305
00:21:45,675 --> 00:21:50,251
比较不同次的拟合到历史窗口

306
00:21:50,251 --> 00:21:53,475
看第 7 次拟合

307
00:21:53,475 --> 00:21:55,350
也是最后一个

308
00:21:55,350 --> 00:21:59,375
不确定值在这儿

309
00:21:59,375 --> 00:22:01,900
ΔE 不确定值是 1.1 

310
00:22:02,475 --> 00:22:07,869
配位数是 5.5 ± 0.6

311
00:22:08,575 --> 00:22:11,250
如果到第 6 次拟合

312
00:22:11,900 --> 00:22:18,525
配位数的不确定值是 0.9

313
00:22:18,525 --> 00:22:20,525
然后减少了 0.5

314
00:22:20,525 --> 00:22:26,100
reduced chi-square 降低到原来的二分之一

315
00:22:26,100 --> 00:22:27,625
一个很明显的区别

316
00:22:29,100 --> 00:22:30,900
这是 539

317
00:22:30,900 --> 00:22:35,050
因为拟合范围选择不恰当

318
00:22:35,050 --> 00:22:36,850
两倍的差距很关键

319
00:22:36,850 --> 00:22:42,425
所以现在 reduced chi-square 是 243

320
00:22:42,425 --> 00:22:44,425
是个更好的拟合

321
00:22:44,425 --> 00:22:49,600
σ2 依然很大

322
00:22:49,600 --> 00:22:55,600
现在是 0.013 一个比较大的值

323
00:22:55,950 --> 00:23:00,215
如果回到硫化铁的 Feff 文件

324
00:23:00,215 --> 00:23:05,900
可以发现铁硫距离有好几个不同的

325
00:23:05,900 --> 00:23:07,900
而不像铁氧路径

326
00:23:08,525 --> 00:23:18,025
有 2.3, 2.4, 2.5, 2.6 和 2.7 Å

327
00:23:18,025 --> 00:23:25,725
可以试着加一个距离更长的铁硫路径

328
00:23:26,350 --> 00:23:28,350
试下

329
00:23:29,200 --> 00:23:31,325
加入这条路径

330
00:23:34,125 --> 00:23:39,700
勾选，并把参数设为一样

331
00:23:40,350 --> 00:23:44,850
像这样设置勾选路径

332
00:23:47,675 --> 00:23:50,900
这个改成铁硫 1

333
00:23:51,500 --> 00:23:54,800
把 ΔR 也改成对应的

334
00:23:54,800 --> 00:23:56,525
并且都设为变量

335
00:23:56,525 --> 00:24:00,350
两条路径的 σ2 设为一样

336
00:24:00,350 --> 00:24:04,075
铁硫 2 对应设置

337
00:24:06,425 --> 00:24:09,275
回到猜测窗口

338
00:24:10,175 --> 00:24:14,075
踢掉不需要的参数

339
00:24:14,750 --> 00:24:19,900
把两条铁硫路径配位数各自设为 3

340
00:24:20,225 --> 00:24:23,750
使用最优解

341
00:24:23,848 --> 00:24:28,375
并且全部定为常数看看

342
00:24:28,375 --> 00:24:30,375
改成 set

343
00:24:33,473 --> 00:24:35,473
点击 Fit

344
00:24:36,344 --> 00:24:41,225
σ2 现在有点太大了

345
00:24:41,825 --> 00:24:43,825
先这样吧

346
00:24:43,825 --> 00:24:50,859
画个图看下

347
00:24:50,859 --> 00:24:53,125
这儿的

348
00:24:53,787 --> 00:25:01,375
现在要拟合中看各自路径的贡献

349
00:25:01,624 --> 00:25:04,775
回到数据面板

350
00:25:05,750 --> 00:25:11,250
点击右上方蓝色按键

351
00:25:11,250 --> 00:25:13,250
把路径加到画图列表

352
00:25:13,575 --> 00:25:16,050
第二条也是

353
00:25:16,050 --> 00:25:20,125
如果想让它每次拟合后自动加载路径

354
00:25:20,400 --> 00:25:23,850
勾选 Plot after fit

355
00:25:24,513 --> 00:25:29,625
图以堆叠形式呈现

356
00:25:29,975 --> 00:25:34,450
我们可以让它们之间错开 0.2

357
00:25:36,200 --> 00:25:40,400
现在可以看到

358
00:25:41,125 --> 00:25:43,125
关掉窗口

359
00:25:43,125 --> 00:25:54,400
我们应该把拟合范围扩大到 2.8 Å 

360
00:25:56,271 --> 00:26:02,575
减小 σ2 数值

361
00:26:03,247 --> 00:26:05,525
回到猜测窗口

362
00:26:05,525 --> 00:26:07,400
减小一半

363
00:26:07,400 --> 00:26:09,400
鉴于有两条路径

364
00:26:10,050 --> 00:26:12,675
跑次拟合

365
00:26:15,477 --> 00:26:18,550
强度相比实验数据高了一些

366
00:26:18,550 --> 00:26:25,250
拟合后到 R 空间

367
00:26:25,975 --> 00:26:27,975
现在看起来像这样

368
00:26:30,275 --> 00:26:32,275
再试下

369
00:26:33,725 --> 00:26:35,725
改为 guess

370
00:26:37,125 --> 00:26:39,125
ΔE

371
00:26:43,075 --> 00:26:45,075
点击 Fit

372
00:26:48,525 --> 00:26:53,050
现在可以看到

373
00:26:53,050 --> 00:27:00,400
我们可以做到跟实验数据的特征信号匹配上了

374
00:27:00,400 --> 00:27:06,750
从 2.5 Å 开始拟合有些差距了

375
00:27:06,750 --> 00:27:13,025
把 Rmax 从 2.8 改成 2.5 

376
00:27:13,025 --> 00:27:15,150
之所以这么做

377
00:27:15,150 --> 00:27:18,825
是因为我觉得没把握去拟合这里

378
00:27:19,175 --> 00:27:24,850
不确定值大小是受这些不吻合的地方影响

379
00:27:25,550 --> 00:27:32,725
所以 R 范围要尽量选到真正被拟合的区间

380
00:27:33,344 --> 00:27:35,850
检查各自最优解

381
00:27:36,300 --> 00:27:41,550
ΔE 数值在正常范围

382
00:27:41,640 --> 00:27:44,275
σ2现在是 0.008

383
00:27:44,275 --> 00:27:51,525
通过使用两条路径来移除了一些混乱度

384
00:27:51,950 --> 00:27:56,300
配位数是 4.2 和 2.1

385
00:27:56,300 --> 00:27:58,600
总配位数是 6

386
00:27:58,600 --> 00:28:01,250
分成 4 和 2 也是合理的

387
00:28:01,700 --> 00:28:05,425
ΔR 数值在也正常范围

388
00:28:05,425 --> 00:28:07,425
0.04 和 0.1

389
00:28:07,425 --> 00:28:10,450
使用最优解

390
00:28:13,225 --> 00:28:15,225
我不是想这么做的

391
00:28:15,600 --> 00:28:18,075
点击 Use best fit

392
00:28:18,075 --> 00:28:20,075
把它们当做初始猜测数值

393
00:28:20,375 --> 00:28:26,175
去到历史窗口

394
00:28:26,175 --> 00:28:27,625
鉴于之前跑的拟合

395
00:28:27,625 --> 00:28:29,625
看下历史记录

396
00:28:30,642 --> 00:28:35,100
现在 reduced chi-square 是 309

397
00:28:35,634 --> 00:28:40,375
之前是 424

398
00:28:40,375 --> 00:28:46,625
这个改善是通过选择恰当拟合范围带来的

399
00:28:47,666 --> 00:28:49,666
 看下不确定值

400
00:28:50,100 --> 00:28:53,175
4.2 ± 0.8

401
00:28:53,559 --> 00:28:55,559
没有变太多

402
00:28:55,559 --> 00:28:59,250
4.0 ± 0.9

403
00:28:59,480 --> 00:29:04,250
你也许会说这些不确定值仍然有点大

404
00:29:04,250 --> 00:29:08,750
我很确定这是硫化铁样品

405
00:29:08,750 --> 00:29:13,400
不仅如此，配位数也应该是 6

406
00:29:13,400 --> 00:29:16,800
铁的配位数通常为 6

407
00:29:16,800 --> 00:29:20,500
虽然我不是一定想证明配位数是 6

408
00:29:20,500 --> 00:29:24,775
但想证明是 4 + 2 组合

409
00:29:25,200 --> 00:29:29,350
你可以通过数学表达式来实现

410
00:29:30,325 --> 00:29:33,938
比如

411
00:29:33,938 --> 00:29:46,000
第二壳层设为 6 减去第一壳层配位数的绝对值

412
00:29:46,825 --> 00:29:48,825
并设为define

413
00:29:49,200 --> 00:29:51,600
用这个数学表达式来“定义”

414
00:29:52,050 --> 00:29:57,025
与设为 amp = 0.8 不一样的是

415
00:29:57,025 --> 00:30:00,800
set 的数值只会在拟合一开始被考虑

416
00:30:00,800 --> 00:30:05,075
而不会随着参数改变而再次考虑

417
00:30:05,075 --> 00:30:07,075
如果定义一个变量

418
00:30:07,075 --> 00:30:10,050
像是第二条铁硫路径

419
00:30:10,050 --> 00:30:15,600
会随着第一壳层配位数改变而再次被改动

420
00:30:17,450 --> 00:30:21,900
点击 Fit

421
00:30:25,607 --> 00:30:27,607
看起来很好

422
00:30:27,975 --> 00:30:32,825
配位数的不确定值下来了 

423
00:30:32,825 --> 00:30:35,575
现在是 0.3

424
00:30:35,575 --> 00:30:40,125
这里配位数是 4 ± 0.3

425
00:30:40,125 --> 00:30:44,850
可以观察下各自路径的贡献

426
00:30:44,850 --> 00:30:49,750
看到第二条路径如何跟第一条错开

427
00:30:49,750 --> 00:30:53,500
造成了这个比较宽的信号

428
00:30:53,500 --> 00:30:56,975
还可以去 k 空间里面比较

429
00:30:57,450 --> 00:31:01,675
k 空间中拟合和实验数据接近

430
00:31:01,675 --> 00:31:07,350
主要由第一条铁硫路径影响

431
00:31:07,350 --> 00:31:09,350
这里有两个不同的信号

432
00:31:09,350 --> 00:31:14,425
我本想强调从 k = 0 到 14 以上一路的信号

433
00:31:14,857 --> 00:31:20,225
但包络显示这里信号衰减很快

434
00:31:20,225 --> 00:31:23,175
在 k 值高的区域，鉴于是硫原子

435
00:31:24,498 --> 00:31:29,125
还可以去反傅里叶变换的 q 空间

436
00:31:29,125 --> 00:31:34,000
踢掉噪音

437
00:31:34,000 --> 00:31:40,900
可以更好地观察路径各自贡献

438
00:31:42,738 --> 00:31:45,525
那么

439
00:31:46,100 --> 00:31:52,200
还有个我没太讲的部分是 log 文件

440
00:31:52,200 --> 00:31:54,200
来看下

441
00:31:54,529 --> 00:31:58,000
这里很方便检查错误

442
00:31:58,000 --> 00:31:59,850
可以看到 reduced chi-square

443
00:31:59,850 --> 00:32:03,875
还有之前没介绍的历史窗口

444
00:32:03,875 --> 00:32:06,350
往下翻 log 文件

445
00:32:06,600 --> 00:32:11,925
可以看到我们使用的两条路径

446
00:32:11,925 --> 00:32:17,400
Reff 是从 Feff 计算得到的距离

447
00:32:17,400 --> 00:32:19,750
R 是优化后的距离

448
00:32:19,750 --> 00:32:26,150
优化后的值是 2.4 和 2.59 Å

449
00:32:26,425 --> 00:32:28,425
这里是 e0

450
00:32:28,425 --> 00:32:30,250
σ2

451
00:32:30,250 --> 00:32:36,350
S02 是 amp 乘以配位数后的结果

452
00:32:36,675 --> 00:32:38,675
如果出现问题

453
00:32:38,675 --> 00:32:42,850
不知道为什么变量完全乱套

454
00:32:42,850 --> 00:32:45,450
这里很适合来寻找问题根源

455
00:32:45,450 --> 00:32:47,450
尤其如果会用到数学表达式

456
00:32:47,450 --> 00:32:49,800
可以看到它们各自的数值

457
00:32:52,325 --> 00:32:57,000
最后 reduced chi-square 是 219

458
00:32:57,000 --> 00:32:59,850
是个不错的值

459
00:32:59,850 --> 00:33:03,675
独立数据点 8

460
00:33:03,925 --> 00:33:06,750
变量数量是 5

461
00:33:06,750 --> 00:33:09,500
也不错

462
00:33:09,500 --> 00:33:16,100
独立数据点计算基于

463
00:33:16,667 --> 00:33:18,825
数据范围和拟合范围

464
00:33:18,825 --> 00:33:21,125
数据范围是 3 - 12

465
00:33:21,125 --> 00:33:23,750
拟合范围 1 - 2.5

466
00:33:23,750 --> 00:33:27,225
就能给出这套数据的独立数据点

467
00:33:28,575 --> 00:33:30,925
总的来说

468
00:33:30,925 --> 00:33:38,900
这节展示了如何拟合硫样品的第一壳层

469
00:33:38,900 --> 00:33:41,539
并且证明这是硫化铁

470
00:33:42,225 --> 00:33:44,100
谢谢

471
00:33:44,100 --> 00:33:46,100
Have a good day;)